From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- arch/um/.gitignore | 5 + arch/um/Kbuild | 3 + arch/um/Kconfig | 306 ++++++ arch/um/Kconfig.debug | 38 + arch/um/Makefile | 156 +++ arch/um/Makefile-os-Linux | 9 + arch/um/Makefile-skas | 17 + arch/um/configs/i386_defconfig | 64 ++ arch/um/configs/x86_64_defconfig | 63 ++ arch/um/drivers/Kconfig | 190 ++++ arch/um/drivers/Makefile | 55 + arch/um/drivers/chan.h | 49 + arch/um/drivers/chan_kern.c | 610 +++++++++++ arch/um/drivers/chan_user.c | 323 ++++++ arch/um/drivers/chan_user.h | 54 + arch/um/drivers/cow.h | 26 + arch/um/drivers/cow_sys.h | 41 + arch/um/drivers/cow_user.c | 450 ++++++++ arch/um/drivers/fd.c | 95 ++ arch/um/drivers/harddog.h | 9 + arch/um/drivers/harddog_kern.c | 174 +++ arch/um/drivers/harddog_user.c | 128 +++ arch/um/drivers/harddog_user_exp.c | 9 + arch/um/drivers/hostaudio_kern.c | 353 ++++++ arch/um/drivers/line.c | 771 ++++++++++++++ arch/um/drivers/line.h | 96 ++ arch/um/drivers/mconsole.h | 98 ++ arch/um/drivers/mconsole_kern.c | 871 +++++++++++++++ arch/um/drivers/mconsole_kern.h | 52 + arch/um/drivers/mconsole_user.c | 221 ++++ arch/um/drivers/null.c | 51 + arch/um/drivers/port.h | 20 + arch/um/drivers/port_kern.c | 305 ++++++ arch/um/drivers/port_user.c | 217 ++++ arch/um/drivers/pty.c | 165 +++ arch/um/drivers/random.c | 121 +++ arch/um/drivers/rtc.h | 15 + arch/um/drivers/rtc_kern.c | 213 ++++ arch/um/drivers/rtc_user.c | 81 ++ arch/um/drivers/ssl.c | 205 ++++ arch/um/drivers/stderr_console.c | 63 ++ arch/um/drivers/stdio_console.c | 198 ++++ arch/um/drivers/stdio_console.h | 11 + arch/um/drivers/tty.c | 81 ++ arch/um/drivers/ubd.h | 22 + arch/um/drivers/ubd_kern.c | 1540 +++++++++++++++++++++++++++ arch/um/drivers/ubd_user.c | 78 ++ arch/um/drivers/vector_kern.c | 1769 +++++++++++++++++++++++++++++++ arch/um/drivers/vector_kern.h | 139 +++ arch/um/drivers/vector_transports.c | 494 +++++++++ arch/um/drivers/vector_user.c | 941 ++++++++++++++++ arch/um/drivers/vector_user.h | 107 ++ arch/um/drivers/vfio_kern.c | 707 ++++++++++++ arch/um/drivers/vfio_user.c | 327 ++++++ arch/um/drivers/vfio_user.h | 44 + arch/um/drivers/vhost_user.h | 123 +++ arch/um/drivers/virt-pci.c | 618 +++++++++++ arch/um/drivers/virt-pci.h | 41 + arch/um/drivers/virtio_pcidev.c | 634 +++++++++++ arch/um/drivers/virtio_uml.c | 1544 +++++++++++++++++++++++++++ arch/um/drivers/xterm.c | 226 ++++ arch/um/drivers/xterm.h | 12 + arch/um/drivers/xterm_kern.c | 83 ++ arch/um/include/asm/Kbuild | 30 + arch/um/include/asm/archrandom.h | 25 + arch/um/include/asm/asm-prototypes.h | 6 + arch/um/include/asm/bpf_perf_event.h | 9 + arch/um/include/asm/cache.h | 18 + arch/um/include/asm/cacheflush.h | 9 + arch/um/include/asm/common.lds.S | 101 ++ arch/um/include/asm/cpufeature.h | 133 +++ arch/um/include/asm/current.h | 24 + arch/um/include/asm/delay.h | 30 + arch/um/include/asm/dma.h | 11 + arch/um/include/asm/fpu/api.h | 22 + arch/um/include/asm/futex.h | 14 + arch/um/include/asm/hardirq.h | 31 + arch/um/include/asm/io.h | 26 + arch/um/include/asm/irq.h | 40 + arch/um/include/asm/irqflags.h | 38 + arch/um/include/asm/kasan.h | 31 + arch/um/include/asm/kvm_para.h | 1 + arch/um/include/asm/mmu.h | 32 + arch/um/include/asm/mmu_context.h | 29 + arch/um/include/asm/msi.h | 1 + arch/um/include/asm/page.h | 98 ++ arch/um/include/asm/pci.h | 19 + arch/um/include/asm/pgalloc.h | 45 + arch/um/include/asm/pgtable-2level.h | 42 + arch/um/include/asm/pgtable-4level.h | 110 ++ arch/um/include/asm/pgtable.h | 328 ++++++ arch/um/include/asm/processor-generic.h | 88 ++ arch/um/include/asm/ptrace-generic.h | 49 + arch/um/include/asm/sections.h | 10 + arch/um/include/asm/setup.h | 11 + arch/um/include/asm/smp.h | 20 + arch/um/include/asm/stacktrace.h | 43 + arch/um/include/asm/syscall-generic.h | 86 ++ arch/um/include/asm/thread_info.h | 62 ++ arch/um/include/asm/timex.h | 13 + arch/um/include/asm/tlb.h | 11 + arch/um/include/asm/tlbflush.h | 59 ++ arch/um/include/asm/uaccess.h | 66 ++ arch/um/include/asm/unwind.h | 8 + arch/um/include/asm/vmalloc.h | 4 + arch/um/include/asm/vmlinux.lds.h | 2 + arch/um/include/linux/smp-internal.h | 17 + arch/um/include/linux/time-internal.h | 96 ++ arch/um/include/linux/virtio-uml.h | 13 + arch/um/include/shared/arch.h | 17 + arch/um/include/shared/as-layout.h | 57 + arch/um/include/shared/elf_user.h | 19 + arch/um/include/shared/frame_kern.h | 15 + arch/um/include/shared/init.h | 127 +++ arch/um/include/shared/irq_kern.h | 80 ++ arch/um/include/shared/irq_user.h | 27 + arch/um/include/shared/kern.h | 22 + arch/um/include/shared/kern_util.h | 70 ++ arch/um/include/shared/longjmp.h | 23 + arch/um/include/shared/mem.h | 22 + arch/um/include/shared/mem_user.h | 42 + arch/um/include/shared/os.h | 359 +++++++ arch/um/include/shared/ptrace_user.h | 15 + arch/um/include/shared/registers.h | 16 + arch/um/include/shared/sigio.h | 12 + arch/um/include/shared/skas/mm_id.h | 30 + arch/um/include/shared/skas/skas.h | 21 + arch/um/include/shared/skas/stub-data.h | 76 ++ arch/um/include/shared/smp.h | 20 + arch/um/include/shared/timetravel.h | 30 + arch/um/include/shared/um_malloc.h | 20 + arch/um/include/shared/user.h | 68 ++ arch/um/include/uapi/asm/Kbuild | 1 + arch/um/kernel/Makefile | 65 ++ arch/um/kernel/asm-offsets.c | 49 + arch/um/kernel/config.c.in | 26 + arch/um/kernel/dtb.c | 42 + arch/um/kernel/dyn.lds.S | 182 ++++ arch/um/kernel/early_printk.c | 32 + arch/um/kernel/exec.c | 37 + arch/um/kernel/exitcode.c | 79 ++ arch/um/kernel/gprof_syms.c | 9 + arch/um/kernel/initrd.c | 46 + arch/um/kernel/irq.c | 728 +++++++++++++ arch/um/kernel/kmsg_dump.c | 65 ++ arch/um/kernel/ksyms.c | 48 + arch/um/kernel/load_file.c | 59 ++ arch/um/kernel/mem.c | 139 +++ arch/um/kernel/physmem.c | 129 +++ arch/um/kernel/process.c | 309 ++++++ arch/um/kernel/ptrace.c | 161 +++ arch/um/kernel/reboot.c | 76 ++ arch/um/kernel/sigio.c | 22 + arch/um/kernel/signal.c | 129 +++ arch/um/kernel/skas/.gitignore | 2 + arch/um/kernel/skas/Makefile | 50 + arch/um/kernel/skas/mmu.c | 165 +++ arch/um/kernel/skas/process.c | 79 ++ arch/um/kernel/skas/stub.c | 181 ++++ arch/um/kernel/skas/stub_exe.c | 230 ++++ arch/um/kernel/skas/stub_exe_embed.S | 11 + arch/um/kernel/skas/syscall.c | 74 ++ arch/um/kernel/skas/uaccess.c | 368 +++++++ arch/um/kernel/smp.c | 242 +++++ arch/um/kernel/stacktrace.c | 75 ++ arch/um/kernel/sysrq.c | 50 + arch/um/kernel/time.c | 1092 +++++++++++++++++++ arch/um/kernel/tlb.c | 225 ++++ arch/um/kernel/trap.c | 430 ++++++++ arch/um/kernel/um_arch.c | 539 ++++++++++ arch/um/kernel/um_arch.h | 16 + arch/um/kernel/umid.c | 38 + arch/um/kernel/uml.lds.S | 120 +++ arch/um/kernel/vmlinux.lds.S | 8 + arch/um/os-Linux/Makefile | 23 + arch/um/os-Linux/elf_aux.c | 50 + arch/um/os-Linux/execvp.c | 150 +++ arch/um/os-Linux/file.c | 726 +++++++++++++ arch/um/os-Linux/helper.c | 236 +++++ arch/um/os-Linux/internal.h | 36 + arch/um/os-Linux/irq.c | 145 +++ arch/um/os-Linux/main.c | 265 +++++ arch/um/os-Linux/mem.c | 246 +++++ arch/um/os-Linux/process.c | 211 ++++ arch/um/os-Linux/registers.c | 43 + arch/um/os-Linux/sigio.c | 304 ++++++ arch/um/os-Linux/signal.c | 506 +++++++++ arch/um/os-Linux/skas/Makefile | 10 + arch/um/os-Linux/skas/mem.c | 284 +++++ arch/um/os-Linux/skas/process.c | 908 ++++++++++++++++ arch/um/os-Linux/smp.c | 148 +++ arch/um/os-Linux/start_up.c | 495 +++++++++ arch/um/os-Linux/time.c | 157 +++ arch/um/os-Linux/tty.c | 60 ++ arch/um/os-Linux/umid.c | 409 +++++++ arch/um/os-Linux/user_syms.c | 37 + arch/um/os-Linux/util.c | 209 ++++ arch/um/scripts/Makefile.rules | 27 + 198 files changed, 31793 insertions(+) create mode 100644 arch/um/.gitignore create mode 100644 arch/um/Kbuild create mode 100644 arch/um/Kconfig create mode 100644 arch/um/Kconfig.debug create mode 100644 arch/um/Makefile create mode 100644 arch/um/Makefile-os-Linux create mode 100644 arch/um/Makefile-skas create mode 100644 arch/um/configs/i386_defconfig create mode 100644 arch/um/configs/x86_64_defconfig create mode 100644 arch/um/drivers/Kconfig create mode 100644 arch/um/drivers/Makefile create mode 100644 arch/um/drivers/chan.h create mode 100644 arch/um/drivers/chan_kern.c create mode 100644 arch/um/drivers/chan_user.c create mode 100644 arch/um/drivers/chan_user.h create mode 100644 arch/um/drivers/cow.h create mode 100644 arch/um/drivers/cow_sys.h create mode 100644 arch/um/drivers/cow_user.c create mode 100644 arch/um/drivers/fd.c create mode 100644 arch/um/drivers/harddog.h create mode 100644 arch/um/drivers/harddog_kern.c create mode 100644 arch/um/drivers/harddog_user.c create mode 100644 arch/um/drivers/harddog_user_exp.c create mode 100644 arch/um/drivers/hostaudio_kern.c create mode 100644 arch/um/drivers/line.c create mode 100644 arch/um/drivers/line.h create mode 100644 arch/um/drivers/mconsole.h create mode 100644 arch/um/drivers/mconsole_kern.c create mode 100644 arch/um/drivers/mconsole_kern.h create mode 100644 arch/um/drivers/mconsole_user.c create mode 100644 arch/um/drivers/null.c create mode 100644 arch/um/drivers/port.h create mode 100644 arch/um/drivers/port_kern.c create mode 100644 arch/um/drivers/port_user.c create mode 100644 arch/um/drivers/pty.c create mode 100644 arch/um/drivers/random.c create mode 100644 arch/um/drivers/rtc.h create mode 100644 arch/um/drivers/rtc_kern.c create mode 100644 arch/um/drivers/rtc_user.c create mode 100644 arch/um/drivers/ssl.c create mode 100644 arch/um/drivers/stderr_console.c create mode 100644 arch/um/drivers/stdio_console.c create mode 100644 arch/um/drivers/stdio_console.h create mode 100644 arch/um/drivers/tty.c create mode 100644 arch/um/drivers/ubd.h create mode 100644 arch/um/drivers/ubd_kern.c create mode 100644 arch/um/drivers/ubd_user.c create mode 100644 arch/um/drivers/vector_kern.c create mode 100644 arch/um/drivers/vector_kern.h create mode 100644 arch/um/drivers/vector_transports.c create mode 100644 arch/um/drivers/vector_user.c create mode 100644 arch/um/drivers/vector_user.h create mode 100644 arch/um/drivers/vfio_kern.c create mode 100644 arch/um/drivers/vfio_user.c create mode 100644 arch/um/drivers/vfio_user.h create mode 100644 arch/um/drivers/vhost_user.h create mode 100644 arch/um/drivers/virt-pci.c create mode 100644 arch/um/drivers/virt-pci.h create mode 100644 arch/um/drivers/virtio_pcidev.c create mode 100644 arch/um/drivers/virtio_uml.c create mode 100644 arch/um/drivers/xterm.c create mode 100644 arch/um/drivers/xterm.h create mode 100644 arch/um/drivers/xterm_kern.c create mode 100644 arch/um/include/asm/Kbuild create mode 100644 arch/um/include/asm/archrandom.h create mode 100644 arch/um/include/asm/asm-prototypes.h create mode 100644 arch/um/include/asm/bpf_perf_event.h create mode 100644 arch/um/include/asm/cache.h create mode 100644 arch/um/include/asm/cacheflush.h create mode 100644 arch/um/include/asm/common.lds.S create mode 100644 arch/um/include/asm/cpufeature.h create mode 100644 arch/um/include/asm/current.h create mode 100644 arch/um/include/asm/delay.h create mode 100644 arch/um/include/asm/dma.h create mode 100644 arch/um/include/asm/fpu/api.h create mode 100644 arch/um/include/asm/futex.h create mode 100644 arch/um/include/asm/hardirq.h create mode 100644 arch/um/include/asm/io.h create mode 100644 arch/um/include/asm/irq.h create mode 100644 arch/um/include/asm/irqflags.h create mode 100644 arch/um/include/asm/kasan.h create mode 100644 arch/um/include/asm/kvm_para.h create mode 100644 arch/um/include/asm/mmu.h create mode 100644 arch/um/include/asm/mmu_context.h create mode 100644 arch/um/include/asm/msi.h create mode 100644 arch/um/include/asm/page.h create mode 100644 arch/um/include/asm/pci.h create mode 100644 arch/um/include/asm/pgalloc.h create mode 100644 arch/um/include/asm/pgtable-2level.h create mode 100644 arch/um/include/asm/pgtable-4level.h create mode 100644 arch/um/include/asm/pgtable.h create mode 100644 arch/um/include/asm/processor-generic.h create mode 100644 arch/um/include/asm/ptrace-generic.h create mode 100644 arch/um/include/asm/sections.h create mode 100644 arch/um/include/asm/setup.h create mode 100644 arch/um/include/asm/smp.h create mode 100644 arch/um/include/asm/stacktrace.h create mode 100644 arch/um/include/asm/syscall-generic.h create mode 100644 arch/um/include/asm/thread_info.h create mode 100644 arch/um/include/asm/timex.h create mode 100644 arch/um/include/asm/tlb.h create mode 100644 arch/um/include/asm/tlbflush.h create mode 100644 arch/um/include/asm/uaccess.h create mode 100644 arch/um/include/asm/unwind.h create mode 100644 arch/um/include/asm/vmalloc.h create mode 100644 arch/um/include/asm/vmlinux.lds.h create mode 100644 arch/um/include/linux/smp-internal.h create mode 100644 arch/um/include/linux/time-internal.h create mode 100644 arch/um/include/linux/virtio-uml.h create mode 100644 arch/um/include/shared/arch.h create mode 100644 arch/um/include/shared/as-layout.h create mode 100644 arch/um/include/shared/elf_user.h create mode 100644 arch/um/include/shared/frame_kern.h create mode 100644 arch/um/include/shared/init.h create mode 100644 arch/um/include/shared/irq_kern.h create mode 100644 arch/um/include/shared/irq_user.h create mode 100644 arch/um/include/shared/kern.h create mode 100644 arch/um/include/shared/kern_util.h create mode 100644 arch/um/include/shared/longjmp.h create mode 100644 arch/um/include/shared/mem.h create mode 100644 arch/um/include/shared/mem_user.h create mode 100644 arch/um/include/shared/os.h create mode 100644 arch/um/include/shared/ptrace_user.h create mode 100644 arch/um/include/shared/registers.h create mode 100644 arch/um/include/shared/sigio.h create mode 100644 arch/um/include/shared/skas/mm_id.h create mode 100644 arch/um/include/shared/skas/skas.h create mode 100644 arch/um/include/shared/skas/stub-data.h create mode 100644 arch/um/include/shared/smp.h create mode 100644 arch/um/include/shared/timetravel.h create mode 100644 arch/um/include/shared/um_malloc.h create mode 100644 arch/um/include/shared/user.h create mode 100644 arch/um/include/uapi/asm/Kbuild create mode 100644 arch/um/kernel/Makefile create mode 100644 arch/um/kernel/asm-offsets.c create mode 100644 arch/um/kernel/config.c.in create mode 100644 arch/um/kernel/dtb.c create mode 100644 arch/um/kernel/dyn.lds.S create mode 100644 arch/um/kernel/early_printk.c create mode 100644 arch/um/kernel/exec.c create mode 100644 arch/um/kernel/exitcode.c create mode 100644 arch/um/kernel/gprof_syms.c create mode 100644 arch/um/kernel/initrd.c create mode 100644 arch/um/kernel/irq.c create mode 100644 arch/um/kernel/kmsg_dump.c create mode 100644 arch/um/kernel/ksyms.c create mode 100644 arch/um/kernel/load_file.c create mode 100644 arch/um/kernel/mem.c create mode 100644 arch/um/kernel/physmem.c create mode 100644 arch/um/kernel/process.c create mode 100644 arch/um/kernel/ptrace.c create mode 100644 arch/um/kernel/reboot.c create mode 100644 arch/um/kernel/sigio.c create mode 100644 arch/um/kernel/signal.c create mode 100644 arch/um/kernel/skas/.gitignore create mode 100644 arch/um/kernel/skas/Makefile create mode 100644 arch/um/kernel/skas/mmu.c create mode 100644 arch/um/kernel/skas/process.c create mode 100644 arch/um/kernel/skas/stub.c create mode 100644 arch/um/kernel/skas/stub_exe.c create mode 100644 arch/um/kernel/skas/stub_exe_embed.S create mode 100644 arch/um/kernel/skas/syscall.c create mode 100644 arch/um/kernel/skas/uaccess.c create mode 100644 arch/um/kernel/smp.c create mode 100644 arch/um/kernel/stacktrace.c create mode 100644 arch/um/kernel/sysrq.c create mode 100644 arch/um/kernel/time.c create mode 100644 arch/um/kernel/tlb.c create mode 100644 arch/um/kernel/trap.c create mode 100644 arch/um/kernel/um_arch.c create mode 100644 arch/um/kernel/um_arch.h create mode 100644 arch/um/kernel/umid.c create mode 100644 arch/um/kernel/uml.lds.S create mode 100644 arch/um/kernel/vmlinux.lds.S create mode 100644 arch/um/os-Linux/Makefile create mode 100644 arch/um/os-Linux/elf_aux.c create mode 100644 arch/um/os-Linux/execvp.c create mode 100644 arch/um/os-Linux/file.c create mode 100644 arch/um/os-Linux/helper.c create mode 100644 arch/um/os-Linux/internal.h create mode 100644 arch/um/os-Linux/irq.c create mode 100644 arch/um/os-Linux/main.c create mode 100644 arch/um/os-Linux/mem.c create mode 100644 arch/um/os-Linux/process.c create mode 100644 arch/um/os-Linux/registers.c create mode 100644 arch/um/os-Linux/sigio.c create mode 100644 arch/um/os-Linux/signal.c create mode 100644 arch/um/os-Linux/skas/Makefile create mode 100644 arch/um/os-Linux/skas/mem.c create mode 100644 arch/um/os-Linux/skas/process.c create mode 100644 arch/um/os-Linux/smp.c create mode 100644 arch/um/os-Linux/start_up.c create mode 100644 arch/um/os-Linux/time.c create mode 100644 arch/um/os-Linux/tty.c create mode 100644 arch/um/os-Linux/umid.c create mode 100644 arch/um/os-Linux/user_syms.c create mode 100644 arch/um/os-Linux/util.c create mode 100644 arch/um/scripts/Makefile.rules (limited to 'arch/um') diff --git a/arch/um/.gitignore b/arch/um/.gitignore new file mode 100644 index 000000000..d69ea5b56 --- /dev/null +++ b/arch/um/.gitignore @@ -0,0 +1,5 @@ +# SPDX-License-Identifier: GPL-2.0-only +kernel/config.c +kernel/config.tmp +kernel/vmlinux.lds +kernel/capflags.c diff --git a/arch/um/Kbuild b/arch/um/Kbuild new file mode 100644 index 000000000..6cf0c1e59 --- /dev/null +++ b/arch/um/Kbuild @@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0-only + +obj-y += kernel/ drivers/ os-Linux/ diff --git a/arch/um/Kconfig b/arch/um/Kconfig new file mode 100644 index 000000000..d9541d13d --- /dev/null +++ b/arch/um/Kconfig @@ -0,0 +1,306 @@ +# SPDX-License-Identifier: GPL-2.0 + +menu "UML-specific options" + +config UML + bool + default y + select ARCH_DISABLE_KASAN_INLINE if STATIC_LINK + select ARCH_NEEDS_DEFER_KASAN if STATIC_LINK + select ARCH_WANTS_DYNAMIC_TASK_STRUCT + select ARCH_HAS_CACHE_LINE_SIZE + select ARCH_HAS_CPU_FINALIZE_INIT + select ARCH_HAS_FORTIFY_SOURCE + # Clang 20 & 21 miscompute __builtin_object_size() under -fprofile-arcs + # on 32-bit, causing spurious compile-time errors in check_copy_size(). + select ARCH_HAS_GCOV_PROFILE_ALL if !(!64BIT && CLANG_VERSION >= 200000 && CLANG_VERSION < 220100) + select ARCH_HAS_KCOV + select ARCH_HAS_STRNCPY_FROM_USER + select ARCH_HAS_STRNLEN_USER + select ARCH_HAS_STRICT_KERNEL_RWX + select HAVE_ARCH_AUDITSYSCALL + select HAVE_ARCH_KASAN if X86_64 + select HAVE_ARCH_KASAN_VMALLOC if HAVE_ARCH_KASAN + select HAVE_ARCH_SECCOMP_FILTER + select HAVE_ASM_MODVERSIONS + select HAVE_UID16 + select HAVE_DEBUG_KMEMLEAK + select HAVE_DEBUG_BUGVERBOSE + select HAVE_PAGE_SIZE_4KB + select NO_DMA if !UML_DMA_EMULATION + select OF_EARLY_FLATTREE if OF + select GENERIC_IRQ_SHOW + select GENERIC_CPU_DEVICES + select GENERIC_SMP_IDLE_THREAD + select HAVE_GCC_PLUGINS + select ARCH_SUPPORTS_LTO_CLANG + select ARCH_SUPPORTS_LTO_CLANG_THIN + select TRACE_IRQFLAGS_SUPPORT + select TTY # Needed for line.c + select HAVE_ARCH_VMAP_STACK + select HAVE_RUST + select ARCH_HAS_UBSAN + select HAVE_ARCH_TRACEHOOK + select HAVE_SYSCALL_TRACEPOINTS + select THREAD_INFO_IN_TASK + select SPARSE_IRQ + select MMU_GATHER_RCU_TABLE_FREE + +config MMU + bool + default y + +config UML_DMA_EMULATION + bool + +config NO_IOMEM + bool "disable IOMEM" if EXPERT + depends on !INDIRECT_IOMEM + default y + +config UML_IOMEM_EMULATION + bool + select INDIRECT_IOMEM + select GENERIC_PCI_IOMAP + +config ISA + bool + +config SBUS + bool + +config LOCKDEP_SUPPORT + bool + default y + +config STACKTRACE_SUPPORT + bool + default y + select STACKTRACE + +config GENERIC_CALIBRATE_DELAY + bool + default y + +config HZ + int + default 100 + +config UML_SUBARCH_SUPPORTS_SMP + bool + +config SMP + bool "Symmetric multi-processing support" + default n + depends on UML_SUBARCH_SUPPORTS_SMP + help + This option enables UML SMP support. + + With this enabled, users can tell UML to start multiple virtual + processors. Each virtual processor is represented as a separate + host thread. + + In UML, kthreads and normal threads (when running in kernel mode) + can be scheduled and executed simultaneously on different virtual + processors. However, the userspace code of normal threads still + runs within their respective single-threaded stubs. + + That is, SMP support is available both within the kernel and + across different processes, but remains limited within threads + of the same process in userspace. + +config NR_CPUS_RANGE_BEGIN + int + default 1 if !SMP + default 2 + +config NR_CPUS_RANGE_END + int + default 1 if !SMP + default 64 + +config NR_CPUS_DEFAULT + int + default 1 if !SMP + default 2 + +config NR_CPUS + int "Maximum number of CPUs" if SMP + range NR_CPUS_RANGE_BEGIN NR_CPUS_RANGE_END + default NR_CPUS_DEFAULT + +source "arch/$(HEADER_ARCH)/um/Kconfig" + +config MAY_HAVE_RUNTIME_DEPS + bool + +config STATIC_LINK + bool "Force a static link" + depends on !MAY_HAVE_RUNTIME_DEPS + help + This option gives you the ability to force a static link of UML. + Normally, UML is linked as a shared binary. This is inconvenient for + use in a chroot jail. So, if you intend to run UML inside a chroot, + you probably want to say Y here. + Additionally, this option enables using higher memory spaces (up to + 2.75G) for UML. + + NOTE: This option is incompatible with some networking features which + depend on features that require being dynamically loaded (like NSS). + +config LD_SCRIPT_STATIC + bool + default y + depends on STATIC_LINK + +config LD_SCRIPT_DYN + bool + default y + depends on !LD_SCRIPT_STATIC + +config LD_SCRIPT_DYN_RPATH + bool "set rpath in the binary" if EXPERT + default y + depends on LD_SCRIPT_DYN + help + Add /lib (and /lib64 for 64-bit) to the linux binary's rpath + explicitly. + + You may need to turn this off if compiling for nix systems + that have their libraries in random /nix directories and + might otherwise unexpected use libraries from /lib or /lib64 + instead of the desired ones. + +config HOSTFS + tristate "Host filesystem" + help + While the User-Mode Linux port uses its own root file system for + booting and normal file access, this module lets the UML user + access files stored on the host. It does not require any + network connection between the Host and UML. An example use of + this might be: + + mount none /tmp/fromhost -t hostfs -o /tmp/umlshare + + where /tmp/fromhost is an empty directory inside UML and + /tmp/umlshare is a directory on the host with files the UML user + wishes to access. + + For more information, see + . + + If you'd like to be able to work with files stored on the host, + say Y or M here; otherwise say N. + +config MCONSOLE + bool "Management console" + depends on PROC_FS + default y + help + The user mode linux management console is a low-level interface to + the kernel, somewhat like the i386 SysRq interface. Since there is + a full-blown operating system running under every user mode linux + instance, there is much greater flexibility possible than with the + SysRq mechanism. + + If you answer 'Y' to this option, to use this feature, you need the + mconsole client (called uml_mconsole) which is present in CVS in + 2.4.5-9um and later (path /tools/mconsole), and is also in the + distribution RPM package in 2.4.6 and later. + + It is safe to say 'Y' here. + +config MAGIC_SYSRQ + bool "Magic SysRq key" + depends on MCONSOLE + help + If you say Y here, you will have some control over the system even + if the system crashes for example during kernel debugging (e.g., you + will be able to flush the buffer cache to disk, reboot the system + immediately or dump some status information). A key for each of the + possible requests is provided. + + This is the feature normally accomplished by pressing a key + while holding SysRq (Alt+PrintScreen). + + On UML, this is accomplished by sending a "sysrq" command with + mconsole, followed by the letter for the requested command. + + The keys are documented in . Don't say Y + unless you really know what this hack does. + +config KERNEL_STACK_ORDER + int "Kernel stack size order" + default 2 if 64BIT + range 2 10 if 64BIT + default 1 if !64BIT + help + This option determines the size of UML kernel stacks. They will + be 1 << order pages. The default is OK unless you're running Valgrind + on UML, in which case, set this to 3. + It is possible to reduce the stack to 1 for 64BIT and 0 for 32BIT on + older (pre-2017) CPUs. It is not recommended on newer CPUs due to the + increase in the size of the state which needs to be saved when handling + signals. + +config PGTABLE_LEVELS + int + default 4 if 64BIT + default 2 if !64BIT + +config UML_TIME_TRAVEL_SUPPORT + bool + prompt "Support time-travel mode (e.g. for test execution)" + # inf-cpu mode is incompatible with the benchmarking + depends on !RAID6_PQ_BENCHMARK + depends on !SMP + help + Enable this option to support time travel inside the UML instance. + + After enabling this option, two modes are accessible at runtime + (selected by the kernel command line), see the kernel's command- + line help for more details. + + It is safe to say Y, but you probably don't need this. + +config UML_MAX_USERSPACE_ITERATIONS + int + prompt "Maximum number of unscheduled userspace iterations" + default 10000 + depends on UML_TIME_TRAVEL_SUPPORT + help + In UML inf-cpu and ext time-travel mode userspace can run without being + interrupted. This will eventually overwhelm the kernel and create OOM + situations (mainly RCU not running). This setting specifies the number + of kernel/userspace switches (minor/major page fault, signal or syscall) + for the same userspace thread before the sched_clock is advanced by a + jiffie to trigger scheduling. + + Setting it to zero disables the feature. + +config KASAN_SHADOW_OFFSET + hex + depends on KASAN + default 0x100000000000 + help + This is the offset at which the ~16TB of shadow memory is + mapped and used by KASAN for memory debugging. This can be any + address that has at least KASAN_SHADOW_SIZE (total address space divided + by 8) amount of space so that the KASAN shadow memory does not conflict + with anything. The default is 0x100000000000, which works even if mem is + set to a large value. On low-memory systems, try 0x7fff8000, as it fits + into the immediate of most instructions, improving performance. + +endmenu + +source "arch/um/drivers/Kconfig" + +config ARCH_SUSPEND_POSSIBLE + def_bool y + depends on !SMP + +menu "Power management options" + +source "kernel/power/Kconfig" + +endmenu diff --git a/arch/um/Kconfig.debug b/arch/um/Kconfig.debug new file mode 100644 index 000000000..1dfb2959c --- /dev/null +++ b/arch/um/Kconfig.debug @@ -0,0 +1,38 @@ +# SPDX-License-Identifier: GPL-2.0 + +config GPROF + bool "Enable gprof support" + depends on DEBUG_INFO && FRAME_POINTER + help + This allows profiling of a User-Mode Linux kernel with the gprof + utility. + + See for more + details. + + If you're involved in UML kernel development and want to use gprof, + say Y. If you're unsure, say N. + +config GCOV + bool "Enable gcov support" + depends on DEBUG_INFO + depends on !KCOV + depends on !MODULES + help + This option allows developers to retrieve coverage data from a UML + session. + + See for more + details. + + If you're involved in UML kernel development and want to use gcov, + say Y. If you're unsure, say N. + +config EARLY_PRINTK + bool "Early printk" + default y + help + Write kernel log output directly to stdout. + + This is useful for kernel debugging when your machine crashes very + early before the console code is initialized. diff --git a/arch/um/Makefile b/arch/um/Makefile new file mode 100644 index 000000000..937639edc --- /dev/null +++ b/arch/um/Makefile @@ -0,0 +1,156 @@ +# +# This file is included by the global makefile so that you can add your own +# architecture-specific flags and dependencies. +# +# Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) +# Licensed under the GPL +# + +# select defconfig based on actual architecture +ifeq ($(SUBARCH),x86) + ifeq ($(shell uname -m),x86_64) + KBUILD_DEFCONFIG := x86_64_defconfig + else + KBUILD_DEFCONFIG := i386_defconfig + endif +else + KBUILD_DEFCONFIG := $(SUBARCH)_defconfig +endif + +ARCH_DIR := arch/um +# We require bash because the vmlinux link and loader script cpp use bash +# features. +SHELL := bash + +MODE_INCLUDE += -I$(srctree)/$(ARCH_DIR)/include/shared/skas + +HEADER_ARCH := $(SUBARCH) + +ifneq ($(filter $(SUBARCH),x86 x86_64 i386),) + HEADER_ARCH := x86 +endif + +ifdef CONFIG_64BIT + KBUILD_CFLAGS += -mcmodel=large +endif + +HOST_DIR := arch/$(HEADER_ARCH) + +include $(srctree)/$(ARCH_DIR)/Makefile-skas +include $(srctree)/$(HOST_DIR)/Makefile.um + +core-y += $(HOST_DIR)/um/ + +SHARED_HEADERS := $(ARCH_DIR)/include/shared +ARCH_INCLUDE := -I$(srctree)/$(SHARED_HEADERS) +ARCH_INCLUDE += -I$(srctree)/$(HOST_DIR)/um/shared +KBUILD_CPPFLAGS += -I$(srctree)/$(HOST_DIR)/um + +# -Dstrrchr=kernel_strrchr (as well as the various in6addr symbols) prevents +# anything from referencing +# libc symbols with the same name, which can cause a linker error. +# +# -Dlongjmp=kernel_longjmp prevents anything from referencing the libpthread.a +# embedded copy of longjmp, same thing for setjmp. +# +# These apply to USER_CFLAGS too. + +KBUILD_CFLAGS += $(CFLAGS) $(CFLAGS-y) -D__arch_um__ \ + $(ARCH_INCLUDE) $(MODE_INCLUDE) \ + -Dlongjmp=kernel_longjmp -Dsetjmp=kernel_setjmp \ + -Din6addr_loopback=kernel_in6addr_loopback \ + -Din6addr_any=kernel_in6addr_any -Dstrrchr=kernel_strrchr \ + -D__close_range=kernel__close_range + +KBUILD_RUSTFLAGS += -Crelocation-model=pie + +KBUILD_AFLAGS += $(ARCH_INCLUDE) + +USER_CFLAGS = $(patsubst $(KERNEL_DEFINES),,$(patsubst -I%,,$(KBUILD_CFLAGS))) \ + $(ARCH_INCLUDE) $(MODE_INCLUDE) $(filter -I%,$(CFLAGS)) \ + -D_FILE_OFFSET_BITS=64 -idirafter $(srctree)/include \ + -idirafter $(objtree)/include -D__KERNEL__ -D__UM_HOST__ \ + -include $(srctree)/include/linux/compiler-version.h \ + -include $(srctree)/include/linux/kconfig.h \ + -idirafter $(ARCH_DIR)/include/generated + +#This will adjust *FLAGS accordingly to the platform. +include $(srctree)/$(ARCH_DIR)/Makefile-os-Linux + +KBUILD_CPPFLAGS += -I$(srctree)/$(HOST_DIR)/include \ + -I$(srctree)/$(HOST_DIR)/include/uapi \ + -I$(objtree)/$(HOST_DIR)/include/generated \ + -I$(objtree)/$(HOST_DIR)/include/generated/uapi + +# -Derrno=kernel_errno - This turns all kernel references to errno into +# kernel_errno to separate them from the libc errno. This allows -fno-common +# in KBUILD_CFLAGS. Otherwise, it would cause ld to complain about the two different +# errnos. +# These apply to kernelspace only. +# +# strip leading and trailing whitespace to make the USER_CFLAGS removal of these +# defines more robust + +KERNEL_DEFINES = $(strip -Derrno=kernel_errno -Dsigprocmask=kernel_sigprocmask \ + -Dmktime=kernel_mktime $(ARCH_KERNEL_DEFINES)) +KBUILD_CFLAGS += $(KERNEL_DEFINES) + +PHONY += linux + +all: linux + +linux: vmlinux + @echo ' LINK $@' + $(Q)ln -f $< $@ + +define archhelp + echo '* linux - Binary kernel image (./linux) - for backward' + echo ' compatibility only, this creates a hard link to the' + echo ' real kernel binary, the "vmlinux" binary you' + echo ' find in the kernel root.' +endef + +archheaders: + $(Q)$(MAKE) -f $(srctree)/Makefile ARCH=$(HEADER_ARCH) asm-generic archheaders + +archprepare: + $(Q)$(MAKE) $(build)=$(HOST_DIR)/um include/generated/user_constants.h + +LINK-$(CONFIG_LD_SCRIPT_STATIC) += -static +LINK-$(CONFIG_LD_SCRIPT_DYN) += -no-pie +LINK-$(CONFIG_LD_SCRIPT_DYN_RPATH) += -Wl,-rpath,/lib + +CFLAGS_NO_HARDENING := $(call cc-option, -fno-PIC,) $(call cc-option, -fno-pic,) \ + -fno-stack-protector $(call cc-option, -fno-stack-protector-all) + +# Options used by linker script +export LDS_START := $(START) +export LDS_ELF_ARCH := $(ELF_ARCH) +export LDS_ELF_FORMAT := $(ELF_FORMAT) + +# The wrappers will select whether using "malloc" or the kernel allocator. +LINK_WRAPS = -Wl,--wrap,malloc -Wl,--wrap,free -Wl,--wrap,calloc + +# Avoid binutils 2.39+ warnings by marking the stack non-executable and +# ignorning warnings for the kallsyms sections. +LDFLAGS_EXECSTACK = -z noexecstack +ifeq ($(CONFIG_LD_IS_BFD),y) +LDFLAGS_EXECSTACK += $(call ld-option,--no-warn-rwx-segments) +endif + +LD_FLAGS_CMDLINE = $(foreach opt,$(KBUILD_LDFLAGS) $(LDFLAGS_EXECSTACK),-Wl,$(opt)) + +# Used by link-vmlinux.sh which has special support for um link +export CFLAGS_vmlinux := $(LINK-y) $(LINK_WRAPS) $(LD_FLAGS_CMDLINE) $(CC_FLAGS_LTO) + +# When cleaning we don't include .config, so we don't include +# TT or skas makefiles and don't clean skas_ptregs.h. +CLEAN_FILES += linux x.i gmon.out +MRPROPER_FILES += $(HOST_DIR)/include/generated + +archclean: + @find . \( -name '*.bb' -o -name '*.bbg' -o -name '*.da' \ + -o -name '*.gcov' \) -type f -print | xargs rm -f + $(Q)$(MAKE) -f $(srctree)/Makefile ARCH=$(HEADER_ARCH) clean + +export HEADER_ARCH SUBARCH USER_CFLAGS CFLAGS_NO_HARDENING DEV_NULL_PATH diff --git a/arch/um/Makefile-os-Linux b/arch/um/Makefile-os-Linux new file mode 100644 index 000000000..2c8a598ec --- /dev/null +++ b/arch/um/Makefile-os-Linux @@ -0,0 +1,9 @@ +# +# Copyright (C) 2000 Jeff Dike (jdike@karaya.com) +# Licensed under the GPL +# + +# To get a definition of F_SETSIG +USER_CFLAGS += -D_GNU_SOURCE -D_LARGEFILE64_SOURCE +KBUILD_CFLAGS += -D_LARGEFILE64_SOURCE +DEV_NULL_PATH = \"/dev/null\" diff --git a/arch/um/Makefile-skas b/arch/um/Makefile-skas new file mode 100644 index 000000000..1a27e65bc --- /dev/null +++ b/arch/um/Makefile-skas @@ -0,0 +1,17 @@ +# +# Copyright (C) 2002 Jeff Dike (jdike@karaya.com) +# Licensed under the GPL +# + +export UM_GPROF_OPT += -pg + +ifdef CONFIG_CC_IS_CLANG +export UM_GCOV_OPT += -fprofile-instr-generate -fcoverage-mapping +else +export UM_GCOV_OPT += -fprofile-arcs -ftest-coverage +endif + +CFLAGS-$(CONFIG_GCOV) += $(UM_GCOV_OPT) +CFLAGS-$(CONFIG_GPROF) += $(UM_GPROF_OPT) +LINK-$(CONFIG_GCOV) += $(UM_GCOV_OPT) +LINK-$(CONFIG_GPROF) += $(UM_GPROF_OPT) diff --git a/arch/um/configs/i386_defconfig b/arch/um/configs/i386_defconfig new file mode 100644 index 000000000..29d9666ec --- /dev/null +++ b/arch/um/configs/i386_defconfig @@ -0,0 +1,64 @@ +# CONFIG_COMPACTION is not set +CONFIG_BINFMT_MISC=m +CONFIG_HOSTFS=y +CONFIG_MAGIC_SYSRQ=y +CONFIG_KERNEL_STACK_ORDER=1 +CONFIG_SYSVIPC=y +CONFIG_POSIX_MQUEUE=y +CONFIG_NO_HZ=y +CONFIG_HIGH_RES_TIMERS=y +CONFIG_BSD_PROCESS_ACCT=y +CONFIG_IKCONFIG=y +CONFIG_IKCONFIG_PROC=y +CONFIG_LOG_BUF_SHIFT=14 +CONFIG_CGROUPS=y +CONFIG_CGROUP_FREEZER=y +CONFIG_CGROUP_DEVICE=y +CONFIG_CPUSETS=y +CONFIG_CGROUP_CPUACCT=y +CONFIG_CGROUP_SCHED=y +CONFIG_BLK_CGROUP=y +# CONFIG_PID_NS is not set +CONFIG_SYSFS_DEPRECATED=y +CONFIG_CC_OPTIMIZE_FOR_SIZE=y +CONFIG_MODULES=y +CONFIG_MODULE_UNLOAD=y +# CONFIG_BLK_DEV_BSG is not set +CONFIG_IOSCHED_BFQ=m +CONFIG_SSL=y +CONFIG_NULL_CHAN=y +CONFIG_PORT_CHAN=y +CONFIG_PTY_CHAN=y +CONFIG_TTY_CHAN=y +CONFIG_XTERM_CHAN=y +CONFIG_CON_CHAN="pts" +CONFIG_SSL_CHAN="pts" +CONFIG_SOUND=m +CONFIG_UML_SOUND=m +CONFIG_DEVTMPFS=y +CONFIG_DEVTMPFS_MOUNT=y +CONFIG_BLK_DEV_UBD=y +CONFIG_BLK_DEV_LOOP=m +CONFIG_BLK_DEV_NBD=m +CONFIG_DUMMY=m +CONFIG_TUN=m +CONFIG_PPP=m +CONFIG_SLIP=m +CONFIG_LEGACY_PTY_COUNT=32 +# CONFIG_HW_RANDOM is not set +CONFIG_UML_RANDOM=y +CONFIG_NET=y +CONFIG_PACKET=y +CONFIG_UNIX=y +CONFIG_INET=y +# CONFIG_IPV6 is not set +CONFIG_EXT4_FS=y +CONFIG_QUOTA=y +CONFIG_AUTOFS_FS=m +CONFIG_ISO9660_FS=m +CONFIG_JOLIET=y +CONFIG_PROC_KCORE=y +CONFIG_TMPFS=y +CONFIG_NLS=y +CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y +CONFIG_DEBUG_KERNEL=y diff --git a/arch/um/configs/x86_64_defconfig b/arch/um/configs/x86_64_defconfig new file mode 100644 index 000000000..af6ff784e --- /dev/null +++ b/arch/um/configs/x86_64_defconfig @@ -0,0 +1,63 @@ +# CONFIG_COMPACTION is not set +CONFIG_BINFMT_MISC=m +CONFIG_HOSTFS=y +CONFIG_MAGIC_SYSRQ=y +CONFIG_SYSVIPC=y +CONFIG_POSIX_MQUEUE=y +CONFIG_NO_HZ=y +CONFIG_HIGH_RES_TIMERS=y +CONFIG_BSD_PROCESS_ACCT=y +CONFIG_IKCONFIG=y +CONFIG_IKCONFIG_PROC=y +CONFIG_LOG_BUF_SHIFT=14 +CONFIG_CGROUPS=y +CONFIG_CGROUP_FREEZER=y +CONFIG_CGROUP_DEVICE=y +CONFIG_CPUSETS=y +CONFIG_CGROUP_CPUACCT=y +CONFIG_CGROUP_SCHED=y +CONFIG_BLK_CGROUP=y +# CONFIG_PID_NS is not set +CONFIG_SYSFS_DEPRECATED=y +CONFIG_CC_OPTIMIZE_FOR_SIZE=y +CONFIG_MODULES=y +CONFIG_MODULE_UNLOAD=y +# CONFIG_BLK_DEV_BSG is not set +CONFIG_IOSCHED_BFQ=m +CONFIG_SSL=y +CONFIG_NULL_CHAN=y +CONFIG_PORT_CHAN=y +CONFIG_PTY_CHAN=y +CONFIG_TTY_CHAN=y +CONFIG_XTERM_CHAN=y +CONFIG_CON_CHAN="pts" +CONFIG_SSL_CHAN="pts" +CONFIG_SOUND=m +CONFIG_UML_SOUND=m +CONFIG_DEVTMPFS=y +CONFIG_DEVTMPFS_MOUNT=y +CONFIG_BLK_DEV_UBD=y +CONFIG_BLK_DEV_LOOP=m +CONFIG_BLK_DEV_NBD=m +CONFIG_DUMMY=m +CONFIG_TUN=m +CONFIG_PPP=m +CONFIG_SLIP=m +CONFIG_LEGACY_PTY_COUNT=32 +# CONFIG_HW_RANDOM is not set +CONFIG_UML_RANDOM=y +CONFIG_NET=y +CONFIG_PACKET=y +CONFIG_UNIX=y +CONFIG_INET=y +# CONFIG_IPV6 is not set +CONFIG_EXT4_FS=y +CONFIG_QUOTA=y +CONFIG_AUTOFS_FS=m +CONFIG_ISO9660_FS=m +CONFIG_JOLIET=y +CONFIG_PROC_KCORE=y +CONFIG_TMPFS=y +CONFIG_NLS=y +CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y +CONFIG_DEBUG_KERNEL=y diff --git a/arch/um/drivers/Kconfig b/arch/um/drivers/Kconfig new file mode 100644 index 000000000..6a0354ca0 --- /dev/null +++ b/arch/um/drivers/Kconfig @@ -0,0 +1,190 @@ +# SPDX-License-Identifier: GPL-2.0 + +menu "UML Character Devices" + +config STDERR_CONSOLE + bool "stderr console" + default y + help + console driver which dumps all printk messages to stderr. + +config SSL + bool "Virtual serial line" + help + The User-Mode Linux environment allows you to create virtual serial + lines on the UML that are usually made to show up on the host as + ttys or ptys. + + See for more + information and command line examples of how to use this facility. + + Unless you have a specific reason for disabling this, say Y. + +config NULL_CHAN + bool "null channel support" + help + This option enables support for attaching UML consoles and serial + lines to a device similar to /dev/null. Data written to it disappears + and there is never any data to be read. + +config PORT_CHAN + bool "port channel support" + help + This option enables support for attaching UML consoles and serial + lines to host portals. They may be accessed with 'telnet + '. Any number of consoles and serial lines may be + attached to a single portal, although what UML device you get when + you telnet to that portal will be unpredictable. + It is safe to say 'Y' here. + +config PTY_CHAN + bool "pty channel support" + help + This option enables support for attaching UML consoles and serial + lines to host pseudo-terminals. Access to both traditional + pseudo-terminals (/dev/pty*) and pts pseudo-terminals are controlled + with this option. The assignment of UML devices to host devices + will be announced in the kernel message log. + It is safe to say 'Y' here. + +config TTY_CHAN + bool "tty channel support" + help + This option enables support for attaching UML consoles and serial + lines to host terminals. Access to both virtual consoles + (/dev/tty*) and the slave side of pseudo-terminals (/dev/ttyp* and + /dev/pts/*) are controlled by this option. + It is safe to say 'Y' here. + +config XTERM_CHAN + bool "xterm channel support" + help + This option enables support for attaching UML consoles and serial + lines to xterms. Each UML device so assigned will be brought up in + its own xterm. + It is safe to say 'Y' here. + +config XTERM_CHAN_DEFAULT_EMULATOR + string "xterm channel default terminal emulator" + depends on XTERM_CHAN + default "xterm" + help + This option allows changing the default terminal emulator. + +config NOCONFIG_CHAN + bool + default !(XTERM_CHAN && TTY_CHAN && PTY_CHAN && PORT_CHAN && NULL_CHAN) + +config CON_ZERO_CHAN + string "Default main console channel initialization" + default "fd:0,fd:1" + help + This is the string describing the channel to which the main console + will be attached by default. This value can be overridden from the + command line. The default value is "fd:0,fd:1", which attaches the + main console to stdin and stdout. + It is safe to leave this unchanged. + +config CON_CHAN + string "Default console channel initialization" + default "xterm" + help + This is the string describing the channel to which all consoles + except the main console will be attached by default. This value can + be overridden from the command line. The default value is "xterm", + which brings them up in xterms. + It is safe to leave this unchanged, although you may wish to change + this if you expect the UML that you build to be run in environments + which don't have X or xterm available. + +config SSL_CHAN + string "Default serial line channel initialization" + default "pty" + help + This is the string describing the channel to which the serial lines + will be attached by default. This value can be overridden from the + command line. The default value is "pty", which attaches them to + traditional pseudo-terminals. + It is safe to leave this unchanged, although you may wish to change + this if you expect the UML that you build to be run in environments + which don't have a set of /dev/pty* devices. + +config UML_SOUND + tristate "Sound support" + depends on SOUND + select SOUND_OSS_CORE + help + This option enables UML sound support. If enabled, it will pull in + the UML hostaudio relay, which acts as a intermediary + between the host's dsp and mixer devices and the UML sound system. + It is safe to say 'Y' here. + +endmenu + +menu "UML Network Devices" + depends on NET + +config UML_NET_VECTOR + bool "Vector I/O high performance network devices" + select MAY_HAVE_RUNTIME_DEPS + help + This User-Mode Linux network driver uses multi-message send + and receive functions. The host running the UML guest must have + a linux kernel version above 3.0 and a libc version > 2.13. + This driver provides tap, raw, gre and l2tpv3 network transports. + + For more information, including explanations of the networking + and sample configurations, see + . + +endmenu + +config VIRTIO_UML + bool "UML driver for virtio devices" + select VIRTIO + help + This driver provides support for virtio based paravirtual device + drivers over vhost-user sockets. + +config UML_RTC + bool "UML RTC driver" + depends on RTC_CLASS + # there's no use in this if PM_SLEEP isn't enabled ... + depends on PM_SLEEP + help + When PM_SLEEP is configured, it may be desirable to wake up using + rtcwake, especially in time-travel mode. This driver enables that + by providing a fake RTC clock that causes a wakeup at the right + time. + +config UML_PCI + bool + select FORCE_PCI + select IRQ_MSI_LIB + select UML_IOMEM_EMULATION + select UML_DMA_EMULATION + select PCI_MSI + select PCI_LOCKLESS_CONFIG + +config UML_PCI_OVER_VIRTIO + bool "Enable PCI over VIRTIO device simulation" + # in theory, just VIRTIO is enough, but that causes recursion + depends on VIRTIO_UML + select UML_PCI + +config UML_PCI_OVER_VIRTIO_DEVICE_ID + int "set the virtio device ID for PCI emulation" + default -1 + depends on UML_PCI_OVER_VIRTIO + help + There's no official device ID assigned (yet), set the one you + wish to use for experimentation here. The default of -1 is + not valid and will cause the driver to fail at probe. + +config UML_PCI_OVER_VFIO + bool "Enable VFIO-based PCI passthrough" + select UML_PCI + help + This driver provides support for VFIO-based PCI passthrough. + Currently, only MSI-X capable devices are supported, and it + is assumed that drivers will use MSI-X. diff --git a/arch/um/drivers/Makefile b/arch/um/drivers/Makefile new file mode 100644 index 000000000..36dc57840 --- /dev/null +++ b/arch/um/drivers/Makefile @@ -0,0 +1,55 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Copyright (C) 2000, 2002, 2003 Jeff Dike (jdike@karaya.com) +# + +# pcap is broken in 2.5 because kbuild doesn't allow pcap.a to be linked +# in to pcap.o + +vector-objs := vector_kern.o vector_user.o vector_transports.o +mconsole-objs := mconsole_kern.o mconsole_user.o +hostaudio-objs := hostaudio_kern.o +ubd-objs := ubd_kern.o ubd_user.o +port-objs := port_kern.o port_user.o +harddog-objs := harddog_kern.o +harddog-builtin-$(CONFIG_UML_WATCHDOG) := harddog_user.o harddog_user_exp.o +rtc-objs := rtc_kern.o rtc_user.o +vfio_uml-objs := vfio_kern.o vfio_user.o + +#XXX: The call below does not work because the flags are added before the +# object name, so nothing from the library gets linked. +#$(call if_changed,ld) + +# When the above is fixed, don't forget to add this too! +#targets += $(obj)/pcap.o + +obj-y := stdio_console.o fd.o chan_kern.o chan_user.o line.o +obj-$(CONFIG_SSL) += ssl.o +obj-$(CONFIG_STDERR_CONSOLE) += stderr_console.o + +obj-$(CONFIG_UML_NET_VECTOR) += vector.o +obj-$(CONFIG_MCONSOLE) += mconsole.o +obj-$(CONFIG_BLK_DEV_UBD) += ubd.o +obj-$(CONFIG_UML_SOUND) += hostaudio.o +obj-$(CONFIG_NULL_CHAN) += null.o +obj-$(CONFIG_PORT_CHAN) += port.o +obj-$(CONFIG_PTY_CHAN) += pty.o +obj-$(CONFIG_TTY_CHAN) += tty.o +obj-$(CONFIG_XTERM_CHAN) += xterm.o xterm_kern.o +obj-$(CONFIG_UML_WATCHDOG) += harddog.o +obj-y += $(harddog-builtin-y) $(harddog-builtin-m) +obj-$(CONFIG_BLK_DEV_COW_COMMON) += cow_user.o +obj-$(CONFIG_UML_RANDOM) += random.o +obj-$(CONFIG_VIRTIO_UML) += virtio_uml.o +obj-$(CONFIG_UML_RTC) += rtc.o +obj-$(CONFIG_UML_PCI) += virt-pci.o +obj-$(CONFIG_UML_PCI_OVER_VIRTIO) += virtio_pcidev.o +obj-$(CONFIG_UML_PCI_OVER_VFIO) += vfio_uml.o + +# pcap_user.o must be added explicitly. +USER_OBJS := fd.o null.o pty.o tty.o xterm.o vector_user.o +CFLAGS_null.o = -DDEV_NULL=$(DEV_NULL_PATH) + +CFLAGS_xterm.o += '-DCONFIG_XTERM_CHAN_DEFAULT_EMULATOR="$(CONFIG_XTERM_CHAN_DEFAULT_EMULATOR)"' + +include $(srctree)/arch/um/scripts/Makefile.rules diff --git a/arch/um/drivers/chan.h b/arch/um/drivers/chan.h new file mode 100644 index 000000000..5a61db512 --- /dev/null +++ b/arch/um/drivers/chan.h @@ -0,0 +1,49 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2000, 2001 Jeff Dike (jdike@karaya.com) + */ + +#ifndef __CHAN_KERN_H__ +#define __CHAN_KERN_H__ + +#include +#include +#include +#include "chan_user.h" +#include "line.h" + +struct chan { + struct list_head list; + struct list_head free_list; + struct line *line; + char *dev; + unsigned int primary:1; + unsigned int input:1; + unsigned int output:1; + unsigned int opened:1; + unsigned int enabled:1; + int fd_in; + int fd_out; /* only different to fd_in if blocking output is needed */ + const struct chan_ops *ops; + void *data; +}; + +extern void chan_interrupt(struct line *line, int irq); +extern int parse_chan_pair(char *str, struct line *line, int device, + const struct chan_opts *opts, char **error_out); +extern int write_chan(struct chan *chan, const u8 *buf, size_t len, + int write_irq); +extern int console_write_chan(struct chan *chan, const char *buf, + int len); +extern int console_open_chan(struct line *line, struct console *co); +extern void deactivate_chan(struct chan *chan, int irq); +extern void chan_enable_winch(struct chan *chan, struct tty_port *port); +extern int enable_chan(struct line *line); +extern void close_chan(struct line *line); +extern int chan_window_size(struct line *line, + unsigned short *rows_out, + unsigned short *cols_out); +extern int chan_config_string(struct line *line, char *str, int size, + char **error_out); + +#endif diff --git a/arch/um/drivers/chan_kern.c b/arch/um/drivers/chan_kern.c new file mode 100644 index 000000000..db2bb032b --- /dev/null +++ b/arch/um/drivers/chan_kern.c @@ -0,0 +1,610 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include "chan.h" +#include +#include + +#ifdef CONFIG_NOCONFIG_CHAN +static void *not_configged_init(char *str, int device, + const struct chan_opts *opts) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); + return NULL; +} + +static int not_configged_open(int input, int output, int primary, void *data, + char **dev_out) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); + return -ENODEV; +} + +static void not_configged_close(int fd, void *data) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); +} + +static int not_configged_read(int fd, u8 *c_out, void *data) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); + return -EIO; +} + +static int not_configged_write(int fd, const u8 *buf, size_t len, void *data) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); + return -EIO; +} + +static int not_configged_console_write(int fd, const char *buf, int len) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); + return -EIO; +} + +static int not_configged_window_size(int fd, void *data, unsigned short *rows, + unsigned short *cols) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); + return -ENODEV; +} + +static void not_configged_free(void *data) +{ + printk(KERN_ERR "Using a channel type which is configured out of " + "UML\n"); +} + +static const struct chan_ops not_configged_ops = { + .init = not_configged_init, + .open = not_configged_open, + .close = not_configged_close, + .read = not_configged_read, + .write = not_configged_write, + .console_write = not_configged_console_write, + .window_size = not_configged_window_size, + .free = not_configged_free, + .winch = 0, +}; +#endif /* CONFIG_NOCONFIG_CHAN */ + +static inline bool need_output_blocking(void) +{ + return time_travel_mode == TT_MODE_INFCPU || + time_travel_mode == TT_MODE_EXTERNAL; +} + +static int open_one_chan(struct chan *chan) +{ + int fd, err; + + if (chan->opened) + return 0; + + if (chan->ops->open == NULL) + fd = 0; + else fd = (*chan->ops->open)(chan->input, chan->output, chan->primary, + chan->data, &chan->dev); + if (fd < 0) + return fd; + + err = os_set_fd_block(fd, 0); + if (err) + goto out_close; + + chan->fd_in = fd; + chan->fd_out = fd; + + /* + * In time-travel modes infinite-CPU and external we need to guarantee + * that any writes to the output succeed immdiately from the point of + * the VM. The best way to do this is to put the FD in blocking mode + * and simply wait/retry until everything is written. + * As every write is guaranteed to complete, we also do not need to + * request an IRQ for the output. + * + * Note that input cannot happen in a time synchronized way. We permit + * it, but time passes very quickly if anything waits for a read. + */ + if (chan->output && need_output_blocking()) { + err = os_dup_file(chan->fd_out); + if (err < 0) + goto out_close; + + chan->fd_out = err; + + err = os_set_fd_block(chan->fd_out, 1); + if (err) { + os_close_file(chan->fd_out); + goto out_close; + } + } + + chan->opened = 1; + return 0; + +out_close: + (*chan->ops->close)(fd, chan->data); + return err; +} + +static int open_chan(struct list_head *chans) +{ + struct list_head *ele; + struct chan *chan; + int ret, err = 0; + + list_for_each(ele, chans) { + chan = list_entry(ele, struct chan, list); + ret = open_one_chan(chan); + if (chan->primary) + err = ret; + } + return err; +} + +void chan_enable_winch(struct chan *chan, struct tty_port *port) +{ + if (chan && chan->primary && chan->ops->winch) + register_winch(chan->fd_in, port); +} + +static void line_timer_cb(struct work_struct *work) +{ + struct line *line = container_of(work, struct line, task.work); + + if (!line->throttled) + chan_interrupt(line, line->read_irq); +} + +int enable_chan(struct line *line) +{ + struct list_head *ele; + struct chan *chan; + int err; + + INIT_DELAYED_WORK(&line->task, line_timer_cb); + + list_for_each(ele, &line->chan_list) { + chan = list_entry(ele, struct chan, list); + err = open_one_chan(chan); + if (err) { + if (chan->primary) + goto out_close; + + continue; + } + + if (chan->enabled) + continue; + err = line_setup_irq(chan->fd_in, chan->input, + chan->output && !need_output_blocking(), + line, chan); + if (err) + goto out_close; + + chan->enabled = 1; + } + + return 0; + + out_close: + close_chan(line); + return err; +} + +/* Items are added in IRQ context, when free_irq can't be called, and + * removed in process context, when it can. + * This handles interrupt sources which disappear, and which need to + * be permanently disabled. This is discovered in IRQ context, but + * the freeing of the IRQ must be done later. + */ +static DEFINE_RAW_SPINLOCK(irqs_to_free_lock); +static LIST_HEAD(irqs_to_free); + +void free_irqs(void) +{ + struct chan *chan; + LIST_HEAD(list); + struct list_head *ele; + unsigned long flags; + + raw_spin_lock_irqsave(&irqs_to_free_lock, flags); + list_splice_init(&irqs_to_free, &list); + raw_spin_unlock_irqrestore(&irqs_to_free_lock, flags); + + list_for_each(ele, &list) { + chan = list_entry(ele, struct chan, free_list); + + if (chan->input && chan->enabled) + um_free_irq(chan->line->read_irq, chan); + if (chan->output && chan->enabled && + !need_output_blocking()) + um_free_irq(chan->line->write_irq, chan); + chan->enabled = 0; + } +} + +static void close_one_chan(struct chan *chan, int delay_free_irq) +{ + unsigned long flags; + + if (!chan->opened) + return; + + if (delay_free_irq) { + raw_spin_lock_irqsave(&irqs_to_free_lock, flags); + list_add(&chan->free_list, &irqs_to_free); + raw_spin_unlock_irqrestore(&irqs_to_free_lock, flags); + } else { + if (chan->input && chan->enabled) + um_free_irq(chan->line->read_irq, chan); + if (chan->output && chan->enabled && + !need_output_blocking()) + um_free_irq(chan->line->write_irq, chan); + chan->enabled = 0; + } + if (chan->fd_out != chan->fd_in) + os_close_file(chan->fd_out); + if (chan->ops->close != NULL) + (*chan->ops->close)(chan->fd_in, chan->data); + + chan->opened = 0; + chan->fd_in = -1; + chan->fd_out = -1; +} + +void close_chan(struct line *line) +{ + struct chan *chan; + + /* Close in reverse order as open in case more than one of them + * refers to the same device and they save and restore that device's + * state. Then, the first one opened will have the original state, + * so it must be the last closed. + */ + list_for_each_entry_reverse(chan, &line->chan_list, list) { + close_one_chan(chan, 0); + } +} + +void deactivate_chan(struct chan *chan, int irq) +{ + if (chan && chan->enabled) + deactivate_fd(chan->fd_in, irq); +} + +int write_chan(struct chan *chan, const u8 *buf, size_t len, int write_irq) +{ + int n, ret = 0; + + if (len == 0 || !chan || !chan->ops->write) + return 0; + + n = chan->ops->write(chan->fd_out, buf, len, chan->data); + if (chan->primary) { + ret = n; + } + return ret; +} + +int console_write_chan(struct chan *chan, const char *buf, int len) +{ + int n, ret = 0; + + if (!chan || !chan->ops->console_write) + return 0; + + n = chan->ops->console_write(chan->fd_out, buf, len); + if (chan->primary) + ret = n; + return ret; +} + +int console_open_chan(struct line *line, struct console *co) +{ + int err; + + err = open_chan(&line->chan_list); + if (err) + return err; + + printk(KERN_INFO "Console initialized on /dev/%s%d\n", co->name, + co->index); + return 0; +} + +int chan_window_size(struct line *line, unsigned short *rows_out, + unsigned short *cols_out) +{ + struct chan *chan; + + chan = line->chan_in; + if (chan && chan->primary) { + if (chan->ops->window_size == NULL) + return 0; + return chan->ops->window_size(chan->fd_in, chan->data, + rows_out, cols_out); + } + chan = line->chan_out; + if (chan && chan->primary) { + if (chan->ops->window_size == NULL) + return 0; + return chan->ops->window_size(chan->fd_in, chan->data, + rows_out, cols_out); + } + return 0; +} + +static void free_one_chan(struct chan *chan) +{ + list_del(&chan->list); + + close_one_chan(chan, 0); + + if (chan->ops->free != NULL) + (*chan->ops->free)(chan->data); + + if (chan->primary && chan->output) + ignore_sigio_fd(chan->fd_in); + kfree(chan); +} + +static void free_chan(struct list_head *chans) +{ + struct list_head *ele, *next; + struct chan *chan; + + list_for_each_safe(ele, next, chans) { + chan = list_entry(ele, struct chan, list); + free_one_chan(chan); + } +} + +static int one_chan_config_string(struct chan *chan, char *str, int size, + char **error_out) +{ + int n = 0; + + if (chan == NULL) { + CONFIG_CHUNK(str, size, n, "none", 1); + return n; + } + + CONFIG_CHUNK(str, size, n, chan->ops->type, 0); + + if (chan->dev == NULL) { + CONFIG_CHUNK(str, size, n, "", 1); + return n; + } + + CONFIG_CHUNK(str, size, n, ":", 0); + CONFIG_CHUNK(str, size, n, chan->dev, 0); + + return n; +} + +static int chan_pair_config_string(struct chan *in, struct chan *out, + char *str, int size, char **error_out) +{ + int n; + + n = one_chan_config_string(in, str, size, error_out); + str += n; + size -= n; + + if (in == out) { + CONFIG_CHUNK(str, size, n, "", 1); + return n; + } + + CONFIG_CHUNK(str, size, n, ",", 1); + n = one_chan_config_string(out, str, size, error_out); + str += n; + size -= n; + CONFIG_CHUNK(str, size, n, "", 1); + + return n; +} + +int chan_config_string(struct line *line, char *str, int size, + char **error_out) +{ + struct chan *in = line->chan_in, *out = line->chan_out; + + if (in && !in->primary) + in = NULL; + if (out && !out->primary) + out = NULL; + + return chan_pair_config_string(in, out, str, size, error_out); +} + +struct chan_type { + char *key; + const struct chan_ops *ops; +}; + +static const struct chan_type chan_table[] = { + { "fd", &fd_ops }, + +#ifdef CONFIG_NULL_CHAN + { "null", &null_ops }, +#else + { "null", ¬_configged_ops }, +#endif + +#ifdef CONFIG_PORT_CHAN + { "port", &port_ops }, +#else + { "port", ¬_configged_ops }, +#endif + +#ifdef CONFIG_PTY_CHAN + { "pty", &pty_ops }, + { "pts", &pts_ops }, +#else + { "pty", ¬_configged_ops }, + { "pts", ¬_configged_ops }, +#endif + +#ifdef CONFIG_TTY_CHAN + { "tty", &tty_ops }, +#else + { "tty", ¬_configged_ops }, +#endif + +#ifdef CONFIG_XTERM_CHAN + { "xterm", &xterm_ops }, +#else + { "xterm", ¬_configged_ops }, +#endif +}; + +static struct chan *parse_chan(struct line *line, char *str, int device, + const struct chan_opts *opts, char **error_out) +{ + const struct chan_type *entry; + const struct chan_ops *ops; + struct chan *chan; + void *data; + int i; + + ops = NULL; + data = NULL; + for(i = 0; i < ARRAY_SIZE(chan_table); i++) { + entry = &chan_table[i]; + if (!strncmp(str, entry->key, strlen(entry->key))) { + ops = entry->ops; + str += strlen(entry->key); + break; + } + } + if (ops == NULL) { + *error_out = "No match for configured backends"; + return NULL; + } + + data = (*ops->init)(str, device, opts); + if (data == NULL) { + *error_out = "Configuration failed"; + return NULL; + } + + chan = kmalloc_obj(*chan, GFP_ATOMIC); + if (chan == NULL) { + *error_out = "Memory allocation failed"; + return NULL; + } + *chan = ((struct chan) { .list = LIST_HEAD_INIT(chan->list), + .free_list = + LIST_HEAD_INIT(chan->free_list), + .line = line, + .primary = 1, + .input = 0, + .output = 0, + .opened = 0, + .enabled = 0, + .fd_in = -1, + .fd_out = -1, + .ops = ops, + .data = data }); + return chan; +} + +int parse_chan_pair(char *str, struct line *line, int device, + const struct chan_opts *opts, char **error_out) +{ + struct list_head *chans = &line->chan_list; + struct chan *new; + char *in, *out; + + if (!list_empty(chans)) { + line->chan_in = line->chan_out = NULL; + free_chan(chans); + INIT_LIST_HEAD(chans); + } + + if (!str) + return 0; + + out = strchr(str, ','); + if (out != NULL) { + in = str; + *out = '\0'; + out++; + new = parse_chan(line, in, device, opts, error_out); + if (new == NULL) + return -1; + + new->input = 1; + list_add(&new->list, chans); + line->chan_in = new; + + new = parse_chan(line, out, device, opts, error_out); + if (new == NULL) + return -1; + + list_add(&new->list, chans); + new->output = 1; + line->chan_out = new; + } + else { + new = parse_chan(line, str, device, opts, error_out); + if (new == NULL) + return -1; + + list_add(&new->list, chans); + new->input = 1; + new->output = 1; + line->chan_in = line->chan_out = new; + } + return 0; +} + +void chan_interrupt(struct line *line, int irq) +{ + struct tty_port *port = &line->port; + struct chan *chan = line->chan_in; + int err; + u8 c; + + if (!chan || !chan->ops->read) + goto out; + + do { + if (!tty_buffer_request_room(port, 1)) { + schedule_delayed_work(&line->task, 1); + goto out; + } + err = chan->ops->read(chan->fd_in, &c, chan->data); + if (err > 0) + tty_insert_flip_char(port, c, TTY_NORMAL); + } while (err > 0); + + if (err == -EIO) { + if (chan->primary) { + tty_port_tty_hangup(&line->port, false); + if (line->chan_out != chan) + close_one_chan(line->chan_out, 1); + } + close_one_chan(chan, 1); + if (chan->primary) + return; + } + out: + tty_flip_buffer_push(port); +} diff --git a/arch/um/drivers/chan_user.c b/arch/um/drivers/chan_user.c new file mode 100644 index 000000000..35f9beeb1 --- /dev/null +++ b/arch/um/drivers/chan_user.c @@ -0,0 +1,323 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include "chan_user.h" +#include +#include + +void generic_close(int fd, void *unused) +{ + close(fd); +} + +int generic_read(int fd, __u8 *c_out, void *unused) +{ + int n; + + CATCH_EINTR(n = read(fd, c_out, sizeof(*c_out))); + if (n > 0) + return n; + else if (n == 0) + return -EIO; + else if (errno == EAGAIN) + return 0; + return -errno; +} + +/* XXX Trivial wrapper around write */ + +int generic_write(int fd, const __u8 *buf, size_t n, void *unused) +{ + int written = 0; + int err; + + /* The FD may be in blocking mode, as such, need to retry short writes, + * they may have been interrupted by a signal. + */ + do { + errno = 0; + err = write(fd, buf + written, n - written); + if (err > 0) { + written += err; + continue; + } + } while (err < 0 && errno == EINTR); + + if (written > 0) + return written; + else if (errno == EAGAIN) + return 0; + else if (err == 0) + return -EIO; + return -errno; +} + +int generic_window_size(int fd, void *unused, unsigned short *rows_out, + unsigned short *cols_out) +{ + struct winsize size; + int ret; + + if (ioctl(fd, TIOCGWINSZ, &size) < 0) + return -errno; + + ret = ((*rows_out != size.ws_row) || (*cols_out != size.ws_col)); + + *rows_out = size.ws_row; + *cols_out = size.ws_col; + + return ret; +} + +void generic_free(void *data) +{ + kfree(data); +} + +int generic_console_write(int fd, const char *buf, int n) +{ + sigset_t old, no_sigio; + struct termios save, new; + int err; + + if (isatty(fd)) { + sigemptyset(&no_sigio); + sigaddset(&no_sigio, SIGIO); + if (sigprocmask(SIG_BLOCK, &no_sigio, &old)) + goto error; + + CATCH_EINTR(err = tcgetattr(fd, &save)); + if (err) + goto error; + new = save; + /* + * The terminal becomes a bit less raw, to handle \n also as + * "Carriage Return", not only as "New Line". Otherwise, the new + * line won't start at the first column. + */ + new.c_oflag |= OPOST; + CATCH_EINTR(err = tcsetattr(fd, TCSAFLUSH, &new)); + if (err) + goto error; + } + err = generic_write(fd, buf, n, NULL); + /* + * Restore raw mode, in any case; we *must* ignore any error apart + * EINTR, except for debug. + */ + if (isatty(fd)) { + CATCH_EINTR(tcsetattr(fd, TCSAFLUSH, &save)); + sigprocmask(SIG_SETMASK, &old, NULL); + } + + return err; +error: + return -errno; +} + +/* + * UML SIGWINCH handling + * + * The point of this is to handle SIGWINCH on consoles which have host + * ttys and relay them inside UML to whatever might be running on the + * console and cares about the window size (since SIGWINCH notifies + * about terminal size changes). + * + * So, we have a separate thread for each host tty attached to a UML + * device (side-issue - I'm annoyed that one thread can't have + * multiple controlling ttys for the purpose of handling SIGWINCH, but + * I imagine there are other reasons that doesn't make any sense). + * + * SIGWINCH can't be received synchronously, so you have to set up to + * receive it as a signal. That being the case, if you are going to + * wait for it, it is convenient to sit in sigsuspend() and wait for + * the signal to bounce you out of it (see below for how we make sure + * to exit only on SIGWINCH). + */ + +static void winch_handler(int sig) +{ +} + +struct winch_data { + int pty_fd; + int pipe_fd; +}; + +static __noreturn int winch_thread(void *arg) +{ + struct winch_data *data = arg; + sigset_t sigs; + int pty_fd, pipe_fd; + int count; + char c = 1; + + os_set_pdeathsig(); + + pty_fd = data->pty_fd; + pipe_fd = data->pipe_fd; + count = write(pipe_fd, &c, sizeof(c)); + if (count != sizeof(c)) + os_info("winch_thread : failed to write synchronization byte, err = %d\n", + -count); + + /* + * We are not using SIG_IGN on purpose, so don't fix it as I thought to + * do! If using SIG_IGN, the sigsuspend() call below would not stop on + * SIGWINCH. + */ + + signal(SIGWINCH, winch_handler); + sigfillset(&sigs); + /* Block all signals possible. */ + if (sigprocmask(SIG_SETMASK, &sigs, NULL) < 0) { + os_info("winch_thread : sigprocmask failed, errno = %d\n", + errno); + goto wait_kill; + } + /* In sigsuspend(), block anything else than SIGWINCH. */ + sigdelset(&sigs, SIGWINCH); + + if (setsid() < 0) { + os_info("winch_thread : setsid failed, errno = %d\n", + errno); + goto wait_kill; + } + + if (ioctl(pty_fd, TIOCSCTTY, 0) < 0) { + os_info("winch_thread : TIOCSCTTY failed on " + "fd %d err = %d\n", pty_fd, errno); + goto wait_kill; + } + + if (tcsetpgrp(pty_fd, os_getpid()) < 0) { + os_info("winch_thread : tcsetpgrp failed on fd %d err = %d\n", + pty_fd, errno); + goto wait_kill; + } + + /* + * These are synchronization calls between various UML threads on the + * host - since they are not different kernel threads, we cannot use + * kernel semaphores. We don't use SysV semaphores because they are + * persistent. + */ + count = read(pipe_fd, &c, sizeof(c)); + if (count != sizeof(c)) + os_info("winch_thread : failed to read synchronization byte, err = %d\n", + errno); + + while(1) { + /* + * This will be interrupted by SIGWINCH only, since + * other signals are blocked. + */ + sigsuspend(&sigs); + + count = write(pipe_fd, &c, sizeof(c)); + if (count != sizeof(c)) + os_info("winch_thread : write failed, err = %d\n", + errno); + } + +wait_kill: + c = 2; + count = write(pipe_fd, &c, sizeof(c)); + while (1) + pause(); +} + +static int winch_tramp(int fd, struct tty_port *port, int *fd_out, + unsigned long *stack_out) +{ + struct winch_data data; + int fds[2], n, err, pid; + char c; + + err = os_pipe(fds, 1, 1); + if (err < 0) { + printk(UM_KERN_ERR "winch_tramp : os_pipe failed, err = %d\n", + -err); + goto out; + } + + data = ((struct winch_data) { .pty_fd = fd, + .pipe_fd = fds[1] } ); + /* + * CLONE_FILES so this thread doesn't hold open files which are open + * now, but later closed in a different thread. This is a + * problem with /dev/net/tun, which if held open by this + * thread, prevents the TUN/TAP device from being reused. + */ + pid = run_helper_thread(winch_thread, &data, CLONE_FILES, stack_out); + if (pid < 0) { + err = pid; + printk(UM_KERN_ERR "fork of winch_thread failed - errno = %d\n", + -err); + goto out_close; + } + + *fd_out = fds[0]; + n = read(fds[0], &c, sizeof(c)); + if (n != sizeof(c)) { + printk(UM_KERN_ERR "winch_tramp : failed to read " + "synchronization byte\n"); + printk(UM_KERN_ERR "read failed, err = %d\n", errno); + printk(UM_KERN_ERR "fd %d will not support SIGWINCH\n", fd); + err = -EINVAL; + goto out_close; + } + + err = os_set_fd_block(*fd_out, 0); + if (err) { + printk(UM_KERN_ERR "winch_tramp: failed to set thread_fd " + "non-blocking.\n"); + goto out_close; + } + + return pid; + + out_close: + close(fds[1]); + close(fds[0]); + out: + return err; +} + +void register_winch(int fd, struct tty_port *port) +{ + unsigned long stack; + int pid, thread, count, thread_fd = -1; + char c = 1; + + if (!isatty(fd)) + return; + + pid = tcgetpgrp(fd); + if (is_skas_winch(pid, fd, port)) { + register_winch_irq(-1, fd, -1, port, 0); + return; + } + + if (pid == -1) { + thread = winch_tramp(fd, port, &thread_fd, &stack); + if (thread < 0) + return; + + register_winch_irq(thread_fd, fd, thread, port, stack); + + count = write(thread_fd, &c, sizeof(c)); + if (count != sizeof(c)) + printk(UM_KERN_ERR "register_winch : failed to write " + "synchronization byte, err = %d\n", errno); + } +} diff --git a/arch/um/drivers/chan_user.h b/arch/um/drivers/chan_user.h new file mode 100644 index 000000000..e158e16fb --- /dev/null +++ b/arch/um/drivers/chan_user.h @@ -0,0 +1,54 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2000, 2001 Jeff Dike (jdike@karaya.com) + */ + +#ifndef __CHAN_USER_H__ +#define __CHAN_USER_H__ + +#include +#include + +struct chan_opts { + void (*const announce)(char *dev_name, int dev); + char *xterm_title; + int raw; +}; + +struct chan_ops { + char *type; + void *(*init)(char *, int, const struct chan_opts *); + int (*open)(int, int, int, void *, char **); + void (*close)(int, void *); + int (*read)(int, __u8 *, void *); + int (*write)(int, const __u8 *, size_t, void *); + int (*console_write)(int, const char *, int); + int (*window_size)(int, void *, unsigned short *, unsigned short *); + void (*free)(void *); + int winch; +}; + +extern const struct chan_ops fd_ops, null_ops, port_ops, pts_ops, pty_ops, + tty_ops, xterm_ops; + +extern void generic_close(int fd, void *unused); +extern int generic_read(int fd, __u8 *c_out, void *unused); +extern int generic_write(int fd, const __u8 *buf, size_t n, void *unused); +extern int generic_console_write(int fd, const char *buf, int n); +extern int generic_window_size(int fd, void *unused, unsigned short *rows_out, + unsigned short *cols_out); +extern void generic_free(void *data); + +struct tty_port; +extern void register_winch(int fd, struct tty_port *port); +extern void register_winch_irq(int fd, int tty_fd, int pid, + struct tty_port *port, unsigned long stack); + +#define __channel_help(fn, prefix) \ +__uml_help(fn, prefix "[0-9]*=\n" \ +" Attach a console or serial line to a host channel. See\n" \ +" http://user-mode-linux.sourceforge.net/old/input.html for a complete\n" \ +" description of this switch.\n\n" \ +); + +#endif diff --git a/arch/um/drivers/cow.h b/arch/um/drivers/cow.h new file mode 100644 index 000000000..9a67c0170 --- /dev/null +++ b/arch/um/drivers/cow.h @@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __COW_H__ +#define __COW_H__ + +#include + +extern int init_cow_file(int fd, char *cow_file, char *backing_file, + int sectorsize, int alignment, int *bitmap_offset_out, + unsigned long *bitmap_len_out, int *data_offset_out); + +extern int file_reader(__u64 offset, char *buf, int len, void *arg); +extern int read_cow_header(int (*reader)(__u64, char *, int, void *), + void *arg, __u32 *version_out, + char **backing_file_out, long long *mtime_out, + unsigned long long *size_out, int *sectorsize_out, + __u32 *align_out, int *bitmap_offset_out); + +extern int write_cow_header(char *cow_file, int fd, char *backing_file, + int sectorsize, int alignment, + unsigned long long *size); + +extern void cow_sizes(int version, __u64 size, int sectorsize, int align, + int bitmap_offset, unsigned long *bitmap_len_out, + int *data_offset_out); + +#endif diff --git a/arch/um/drivers/cow_sys.h b/arch/um/drivers/cow_sys.h new file mode 100644 index 000000000..916811ef5 --- /dev/null +++ b/arch/um/drivers/cow_sys.h @@ -0,0 +1,41 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __COW_SYS_H__ +#define __COW_SYS_H__ + +#include +#include +#include + +static inline void *cow_malloc(int size) +{ + return uml_kmalloc(size, UM_GFP_KERNEL); +} + +static inline void cow_free(void *ptr) +{ + kfree(ptr); +} + +#define cow_printf printk + +static inline char *cow_strdup(char *str) +{ + return uml_strdup(str); +} + +static inline int cow_seek_file(int fd, __u64 offset) +{ + return os_seek_file(fd, offset); +} + +static inline int cow_file_size(char *file, unsigned long long *size_out) +{ + return os_file_size(file, size_out); +} + +static inline int cow_write_file(int fd, void *buf, int size) +{ + return os_write_file(fd, buf, size); +} + +#endif diff --git a/arch/um/drivers/cow_user.c b/arch/um/drivers/cow_user.c new file mode 100644 index 000000000..dc1d1bcd8 --- /dev/null +++ b/arch/um/drivers/cow_user.c @@ -0,0 +1,450 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +/* + * _XOPEN_SOURCE is needed for pread, but we define _GNU_SOURCE, which defines + * that. + */ +#include +#include +#include +#include +#include +#include "cow.h" +#include "cow_sys.h" + +/* + * arch/um/Makefile remaps strrchr to kernel_strrchr; call the kernel + * name directly to avoid glibc >= 2.43's C23 strrchr macro. + */ +char *kernel_strrchr(const char *, int); + +#define PATH_LEN_V1 256 + +/* unsigned time_t works until year 2106 */ +typedef __u32 time32_t; + +struct cow_header_v1 { + __s32 magic; + __s32 version; + char backing_file[PATH_LEN_V1]; + time32_t mtime; + __u64 size; + __s32 sectorsize; +} __attribute__((packed)); + +/* + * Define PATH_LEN_V3 as the usual value of MAXPATHLEN, just hard-code it in + * case other systems have different values for MAXPATHLEN. + * + * The same must hold for V2 - we want file format compatibility, not anything + * else. + */ +#define PATH_LEN_V3 4096 +#define PATH_LEN_V2 PATH_LEN_V3 + +struct cow_header_v2 { + __u32 magic; + __u32 version; + char backing_file[PATH_LEN_V2]; + time32_t mtime; + __u64 size; + __s32 sectorsize; +} __attribute__((packed)); + +/* + * Changes from V2 - + * PATH_LEN_V3 as described above + * Explicitly specify field bit lengths for systems with different + * lengths for the usual C types. Not sure whether char or + * time_t should be changed, this can be changed later without + * breaking compatibility + * Add alignment field so that different alignments can be used for the + * bitmap and data + * Add cow_format field to allow for the possibility of different ways + * of specifying the COW blocks. For now, the only value is 0, + * for the traditional COW bitmap. + * Move the backing_file field to the end of the header. This allows + * for the possibility of expanding it into the padding required + * by the bitmap alignment. + * The bitmap and data portions of the file will be aligned as specified + * by the alignment field. This is to allow COW files to be + * put on devices with restrictions on access alignments, such as + * /dev/raw, with a 512 byte alignment restriction. This also + * allows the data to be more aligned more strictly than on + * sector boundaries. This is needed for ubd-mmap, which needs + * the data to be page aligned. + * Fixed (finally!) the rounding bug + */ + +/* + * Until Dec2005, __attribute__((packed)) was left out from the below + * definition, leading on 64-bit systems to 4 bytes of padding after mtime, to + * align size to 8-byte alignment. This shifted all fields above (no padding + * was present on 32-bit, no other padding was added). + * + * However, this _can be detected_: it means that cow_format (always 0 until + * now) is shifted onto the first 4 bytes of backing_file, where it is otherwise + * impossible to find 4 zeros. -bb */ + +struct cow_header_v3 { + __u32 magic; + __u32 version; + __u32 mtime; + __u64 size; + __u32 sectorsize; + __u32 alignment; + __u32 cow_format; + char backing_file[PATH_LEN_V3]; +} __attribute__((packed)); + +/* This is the broken layout used by some 64-bit binaries. */ +struct cow_header_v3_broken { + __u32 magic; + __u32 version; + __s64 mtime; + __u64 size; + __u32 sectorsize; + __u32 alignment; + __u32 cow_format; + char backing_file[PATH_LEN_V3]; +}; + +/* COW format definitions - for now, we have only the usual COW bitmap */ +#define COW_BITMAP 0 + +union cow_header { + struct cow_header_v1 v1; + struct cow_header_v2 v2; + struct cow_header_v3 v3; + struct cow_header_v3_broken v3_b; +}; + +#define COW_MAGIC 0x4f4f4f4d /* MOOO */ +#define COW_VERSION 3 + +#define DIV_ROUND(x, len) (((x) + (len) - 1) / (len)) +#define ROUND_UP(x, align) DIV_ROUND(x, align) * (align) + +void cow_sizes(int version, __u64 size, int sectorsize, int align, + int bitmap_offset, unsigned long *bitmap_len_out, + int *data_offset_out) +{ + if (version < 3) { + *bitmap_len_out = (size + sectorsize - 1) / (8 * sectorsize); + + *data_offset_out = bitmap_offset + *bitmap_len_out; + *data_offset_out = (*data_offset_out + sectorsize - 1) / + sectorsize; + *data_offset_out *= sectorsize; + } + else { + *bitmap_len_out = DIV_ROUND(size, sectorsize); + *bitmap_len_out = DIV_ROUND(*bitmap_len_out, 8); + + *data_offset_out = bitmap_offset + *bitmap_len_out; + *data_offset_out = ROUND_UP(*data_offset_out, align); + } +} + +static int absolutize(char *to, int size, char *from) +{ + char save_cwd[256], *slash; + int remaining; + + if (getcwd(save_cwd, sizeof(save_cwd)) == NULL) { + cow_printf("absolutize : unable to get cwd - errno = %d\n", + errno); + return -1; + } + slash = kernel_strrchr(from, '/'); + if (slash != NULL) { + *slash = '\0'; + if (chdir(from)) { + *slash = '/'; + cow_printf("absolutize : Can't cd to '%s' - " + "errno = %d\n", from, errno); + return -1; + } + *slash = '/'; + if (getcwd(to, size) == NULL) { + cow_printf("absolutize : unable to get cwd of '%s' - " + "errno = %d\n", from, errno); + return -1; + } + remaining = size - strlen(to); + if (strlen(slash) + 1 > remaining) { + cow_printf("absolutize : unable to fit '%s' into %d " + "chars\n", from, size); + return -1; + } + strcat(to, slash); + } + else { + if (strlen(save_cwd) + 1 + strlen(from) + 1 > size) { + cow_printf("absolutize : unable to fit '%s' into %d " + "chars\n", from, size); + return -1; + } + strcpy(to, save_cwd); + strcat(to, "/"); + strcat(to, from); + } + if (chdir(save_cwd)) { + cow_printf("absolutize : Can't cd to '%s' - " + "errno = %d\n", save_cwd, errno); + return -1; + } + return 0; +} + +int write_cow_header(char *cow_file, int fd, char *backing_file, + int sectorsize, int alignment, unsigned long long *size) +{ + struct cow_header_v3 *header; + long long modtime; + int err; + + err = cow_seek_file(fd, 0); + if (err < 0) { + cow_printf("write_cow_header - lseek failed, err = %d\n", -err); + goto out; + } + + err = -ENOMEM; + header = cow_malloc(sizeof(*header)); + if (header == NULL) { + cow_printf("write_cow_header - failed to allocate COW V3 " + "header\n"); + goto out; + } + header->magic = htobe32(COW_MAGIC); + header->version = htobe32(COW_VERSION); + + err = -EINVAL; + if (strlen(backing_file) > sizeof(header->backing_file) - 1) { + /* Below, %zd is for a size_t value */ + cow_printf("Backing file name \"%s\" is too long - names are " + "limited to %zd characters\n", backing_file, + sizeof(header->backing_file) - 1); + goto out_free; + } + + if (absolutize(header->backing_file, sizeof(header->backing_file), + backing_file)) + goto out_free; + + err = os_file_modtime(header->backing_file, &modtime); + if (err < 0) { + cow_printf("write_cow_header - backing file '%s' mtime " + "request failed, err = %d\n", header->backing_file, + -err); + goto out_free; + } + + err = cow_file_size(header->backing_file, size); + if (err < 0) { + cow_printf("write_cow_header - couldn't get size of " + "backing file '%s', err = %d\n", + header->backing_file, -err); + goto out_free; + } + + header->mtime = htobe32(modtime); + header->size = htobe64(*size); + header->sectorsize = htobe32(sectorsize); + header->alignment = htobe32(alignment); + header->cow_format = COW_BITMAP; + + err = cow_write_file(fd, header, sizeof(*header)); + if (err != sizeof(*header)) { + cow_printf("write_cow_header - write of header to " + "new COW file '%s' failed, err = %d\n", cow_file, + -err); + goto out_free; + } + err = 0; + out_free: + cow_free(header); + out: + return err; +} + +int file_reader(__u64 offset, char *buf, int len, void *arg) +{ + int fd = *((int *) arg); + + return pread(fd, buf, len, offset); +} + +/* XXX Need to sanity-check the values read from the header */ + +int read_cow_header(int (*reader)(__u64, char *, int, void *), void *arg, + __u32 *version_out, char **backing_file_out, + long long *mtime_out, unsigned long long *size_out, + int *sectorsize_out, __u32 *align_out, + int *bitmap_offset_out) +{ + union cow_header *header; + char *file; + int err, n; + unsigned long version, magic; + + header = cow_malloc(sizeof(*header)); + if (header == NULL) { + cow_printf("read_cow_header - Failed to allocate header\n"); + return -ENOMEM; + } + err = -EINVAL; + n = (*reader)(0, (char *) header, sizeof(*header), arg); + if (n < offsetof(typeof(header->v1), backing_file)) { + cow_printf("read_cow_header - short header\n"); + goto out; + } + + magic = header->v1.magic; + if (magic == COW_MAGIC) + version = header->v1.version; + else if (magic == be32toh(COW_MAGIC)) + version = be32toh(header->v1.version); + /* No error printed because the non-COW case comes through here */ + else goto out; + + *version_out = version; + + if (version == 1) { + if (n < sizeof(header->v1)) { + cow_printf("read_cow_header - failed to read V1 " + "header\n"); + goto out; + } + *mtime_out = header->v1.mtime; + *size_out = header->v1.size; + *sectorsize_out = header->v1.sectorsize; + *bitmap_offset_out = sizeof(header->v1); + *align_out = *sectorsize_out; + file = header->v1.backing_file; + } + else if (version == 2) { + if (n < sizeof(header->v2)) { + cow_printf("read_cow_header - failed to read V2 " + "header\n"); + goto out; + } + *mtime_out = be32toh(header->v2.mtime); + *size_out = be64toh(header->v2.size); + *sectorsize_out = be32toh(header->v2.sectorsize); + *bitmap_offset_out = sizeof(header->v2); + *align_out = *sectorsize_out; + file = header->v2.backing_file; + } + /* This is very subtle - see above at union cow_header definition */ + else if (version == 3 && (*((int*)header->v3.backing_file) != 0)) { + if (n < sizeof(header->v3)) { + cow_printf("read_cow_header - failed to read V3 " + "header\n"); + goto out; + } + *mtime_out = be32toh(header->v3.mtime); + *size_out = be64toh(header->v3.size); + *sectorsize_out = be32toh(header->v3.sectorsize); + *align_out = be32toh(header->v3.alignment); + if (*align_out == 0) { + cow_printf("read_cow_header - invalid COW header, " + "align == 0\n"); + } + *bitmap_offset_out = ROUND_UP(sizeof(header->v3), *align_out); + file = header->v3.backing_file; + } + else if (version == 3) { + cow_printf("read_cow_header - broken V3 file with" + " 64-bit layout - recovering content.\n"); + + if (n < sizeof(header->v3_b)) { + cow_printf("read_cow_header - failed to read V3 " + "header\n"); + goto out; + } + + /* + * this was used until Dec2005 - 64bits are needed to represent + * 2106+. I.e. we can safely do this truncating cast. + * + * Additionally, we must use be32toh() instead of be64toh(), since + * the program used to use the former (tested - I got mtime + * mismatch "0 vs whatever"). + * + * Ever heard about bug-to-bug-compatibility ? ;-) */ + *mtime_out = (time32_t) be32toh(header->v3_b.mtime); + + *size_out = be64toh(header->v3_b.size); + *sectorsize_out = be32toh(header->v3_b.sectorsize); + *align_out = be32toh(header->v3_b.alignment); + if (*align_out == 0) { + cow_printf("read_cow_header - invalid COW header, " + "align == 0\n"); + } + *bitmap_offset_out = ROUND_UP(sizeof(header->v3_b), *align_out); + file = header->v3_b.backing_file; + } + else { + cow_printf("read_cow_header - invalid COW version\n"); + goto out; + } + err = -ENOMEM; + *backing_file_out = cow_strdup(file); + if (*backing_file_out == NULL) { + cow_printf("read_cow_header - failed to allocate backing " + "file\n"); + goto out; + } + err = 0; + out: + cow_free(header); + return err; +} + +int init_cow_file(int fd, char *cow_file, char *backing_file, int sectorsize, + int alignment, int *bitmap_offset_out, + unsigned long *bitmap_len_out, int *data_offset_out) +{ + unsigned long long size, offset; + char zero = 0; + int err; + + err = write_cow_header(cow_file, fd, backing_file, sectorsize, + alignment, &size); + if (err) + goto out; + + *bitmap_offset_out = ROUND_UP(sizeof(struct cow_header_v3), alignment); + cow_sizes(COW_VERSION, size, sectorsize, alignment, *bitmap_offset_out, + bitmap_len_out, data_offset_out); + + offset = *data_offset_out + size - sizeof(zero); + err = cow_seek_file(fd, offset); + if (err < 0) { + cow_printf("cow bitmap lseek failed : err = %d\n", -err); + goto out; + } + + /* + * does not really matter how much we write it is just to set EOF + * this also sets the entire COW bitmap + * to zero without having to allocate it + */ + err = cow_write_file(fd, &zero, sizeof(zero)); + if (err != sizeof(zero)) { + cow_printf("Write of bitmap to new COW file '%s' failed, " + "err = %d\n", cow_file, -err); + if (err >= 0) + err = -EINVAL; + goto out; + } + + return 0; + out: + return err; +} diff --git a/arch/um/drivers/fd.c b/arch/um/drivers/fd.c new file mode 100644 index 000000000..082d739dc --- /dev/null +++ b/arch/um/drivers/fd.c @@ -0,0 +1,95 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include +#include +#include "chan_user.h" +#include +#include + +struct fd_chan { + int fd; + int raw; + struct termios tt; + char str[sizeof("1234567890\0")]; +}; + +static void *fd_init(char *str, int device, const struct chan_opts *opts) +{ + struct fd_chan *data; + char *end; + int n; + + if (*str != ':') { + printk(UM_KERN_ERR "fd_init : channel type 'fd' must specify a " + "file descriptor\n"); + return NULL; + } + str++; + n = strtoul(str, &end, 0); + if ((*end != '\0') || (end == str)) { + printk(UM_KERN_ERR "fd_init : couldn't parse file descriptor " + "'%s'\n", str); + return NULL; + } + + data = uml_kmalloc(sizeof(*data), UM_GFP_KERNEL); + if (data == NULL) + return NULL; + + *data = ((struct fd_chan) { .fd = n, + .raw = opts->raw }); + return data; +} + +static int fd_open(int input, int output, int primary, void *d, char **dev_out) +{ + struct fd_chan *data = d; + int err; + + if (data->raw && isatty(data->fd)) { + CATCH_EINTR(err = tcgetattr(data->fd, &data->tt)); + if (err) + return err; + + err = raw(data->fd); + if (err) + return err; + } + sprintf(data->str, "%d", data->fd); + *dev_out = data->str; + return data->fd; +} + +static void fd_close(int fd, void *d) +{ + struct fd_chan *data = d; + int err; + + if (!data->raw || !isatty(fd)) + return; + + CATCH_EINTR(err = tcsetattr(fd, TCSAFLUSH, &data->tt)); + if (err) + printk(UM_KERN_ERR "Failed to restore terminal state - " + "errno = %d\n", -err); + data->raw = 0; +} + +const struct chan_ops fd_ops = { + .type = "fd", + .init = fd_init, + .open = fd_open, + .close = fd_close, + .read = generic_read, + .write = generic_write, + .console_write = generic_console_write, + .window_size = generic_window_size, + .free = generic_free, + .winch = 1, +}; diff --git a/arch/um/drivers/harddog.h b/arch/um/drivers/harddog.h new file mode 100644 index 000000000..6d9ea60e7 --- /dev/null +++ b/arch/um/drivers/harddog.h @@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef UM_WATCHDOG_H +#define UM_WATCHDOG_H + +int start_watchdog(int *in_fd_ret, int *out_fd_ret, char *sock); +void stop_watchdog(int in_fd, int out_fd); +int ping_watchdog(int fd); + +#endif /* UM_WATCHDOG_H */ diff --git a/arch/um/drivers/harddog_kern.c b/arch/um/drivers/harddog_kern.c new file mode 100644 index 000000000..819aabb4e --- /dev/null +++ b/arch/um/drivers/harddog_kern.c @@ -0,0 +1,174 @@ +/* UML hardware watchdog, shamelessly stolen from: + * + * SoftDog 0.05: A Software Watchdog Device + * + * (c) Copyright 1996 Alan Cox , All Rights Reserved. + * http://www.redhat.com + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + * + * Neither Alan Cox nor CymruNet Ltd. admit liability nor provide + * warranty for any of this software. This material is provided + * "AS-IS" and at no charge. + * + * (c) Copyright 1995 Alan Cox + * + * Software only watchdog driver. Unlike its big brother the WDT501P + * driver this won't always recover a failed machine. + * + * 03/96: Angelo Haritsis : + * Modularised. + * Added soft_margin; use upon insmod to change the timer delay. + * NB: uses same minor as wdt (WATCHDOG_MINOR); we could use separate + * minors. + * + * 19980911 Alan Cox + * Made SMP safe for 2.3.x + * + * 20011127 Joel Becker (jlbec@evilplan.org> + * Added soft_noboot; Allows testing the softdog trigger without + * requiring a recompile. + * Added WDIOC_GETTIMEOUT and WDIOC_SETTIMOUT. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "mconsole.h" +#include "harddog.h" + +MODULE_DESCRIPTION("UML hardware watchdog"); +MODULE_LICENSE("GPL"); + +static DEFINE_MUTEX(harddog_mutex); +static DEFINE_SPINLOCK(lock); +static int timer_alive; +static int harddog_in_fd = -1; +static int harddog_out_fd = -1; + +/* + * Allow only one person to hold it open + */ + +static int harddog_open(struct inode *inode, struct file *file) +{ + int err = -EBUSY; + char *sock = NULL; + + mutex_lock(&harddog_mutex); + spin_lock(&lock); + if(timer_alive) + goto err; +#ifdef CONFIG_WATCHDOG_NOWAYOUT + __module_get(THIS_MODULE); +#endif + +#ifdef CONFIG_MCONSOLE + sock = mconsole_notify_socket(); +#endif + err = start_watchdog(&harddog_in_fd, &harddog_out_fd, sock); + if(err) + goto err; + + timer_alive = 1; + spin_unlock(&lock); + mutex_unlock(&harddog_mutex); + return stream_open(inode, file); +err: + spin_unlock(&lock); + mutex_unlock(&harddog_mutex); + return err; +} + +static int harddog_release(struct inode *inode, struct file *file) +{ + /* + * Shut off the timer. + */ + + spin_lock(&lock); + + stop_watchdog(harddog_in_fd, harddog_out_fd); + harddog_in_fd = -1; + harddog_out_fd = -1; + + timer_alive=0; + spin_unlock(&lock); + + return 0; +} + +static ssize_t harddog_write(struct file *file, const char __user *data, size_t len, + loff_t *ppos) +{ + /* + * Refresh the timer. + */ + if(len) + return ping_watchdog(harddog_out_fd); + return 0; +} + +static int harddog_ioctl_unlocked(struct file *file, + unsigned int cmd, unsigned long arg) +{ + void __user *argp= (void __user *)arg; + static struct watchdog_info ident = { + WDIOC_SETTIMEOUT, + 0, + "UML Hardware Watchdog" + }; + switch (cmd) { + default: + return -ENOTTY; + case WDIOC_GETSUPPORT: + if(copy_to_user(argp, &ident, sizeof(ident))) + return -EFAULT; + return 0; + case WDIOC_GETSTATUS: + case WDIOC_GETBOOTSTATUS: + return put_user(0,(int __user *)argp); + case WDIOC_KEEPALIVE: + return ping_watchdog(harddog_out_fd); + } +} + +static long harddog_ioctl(struct file *file, + unsigned int cmd, unsigned long arg) +{ + long ret; + + mutex_lock(&harddog_mutex); + ret = harddog_ioctl_unlocked(file, cmd, arg); + mutex_unlock(&harddog_mutex); + + return ret; +} + +static const struct file_operations harddog_fops = { + .owner = THIS_MODULE, + .write = harddog_write, + .unlocked_ioctl = harddog_ioctl, + .compat_ioctl = compat_ptr_ioctl, + .open = harddog_open, + .release = harddog_release, +}; + +static struct miscdevice harddog_miscdev = { + .minor = WATCHDOG_MINOR, + .name = "watchdog", + .fops = &harddog_fops, +}; +module_misc_device(harddog_miscdev); diff --git a/arch/um/drivers/harddog_user.c b/arch/um/drivers/harddog_user.c new file mode 100644 index 000000000..9ed893049 --- /dev/null +++ b/arch/um/drivers/harddog_user.c @@ -0,0 +1,128 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + */ + +#include +#include +#include +#include +#include "harddog.h" + +struct dog_data { + int stdin_fd; + int stdout_fd; + int close_me[2]; +}; + +static void pre_exec(void *d) +{ + struct dog_data *data = d; + + dup2(data->stdin_fd, 0); + dup2(data->stdout_fd, 1); + dup2(data->stdout_fd, 2); + close(data->stdin_fd); + close(data->stdout_fd); + close(data->close_me[0]); + close(data->close_me[1]); +} + +int start_watchdog(int *in_fd_ret, int *out_fd_ret, char *sock) +{ + struct dog_data data; + int in_fds[2], out_fds[2], pid, n, err; + char pid_buf[sizeof("nnnnnnn\0")], c; + char *pid_args[] = { "/usr/bin/uml_watchdog", "-pid", pid_buf, NULL }; + char *mconsole_args[] = { "/usr/bin/uml_watchdog", "-mconsole", NULL, + NULL }; + char **args = NULL; + + err = os_pipe(in_fds, 1, 0); + if (err < 0) { + printk("harddog_open - os_pipe failed, err = %d\n", -err); + goto out; + } + + err = os_pipe(out_fds, 1, 0); + if (err < 0) { + printk("harddog_open - os_pipe failed, err = %d\n", -err); + goto out_close_in; + } + + data.stdin_fd = out_fds[0]; + data.stdout_fd = in_fds[1]; + data.close_me[0] = out_fds[1]; + data.close_me[1] = in_fds[0]; + + if (sock != NULL) { + mconsole_args[2] = sock; + args = mconsole_args; + } + else { + /* XXX The os_getpid() is not SMP correct */ + sprintf(pid_buf, "%d", os_getpid()); + args = pid_args; + } + + pid = run_helper(pre_exec, &data, args); + + close(out_fds[0]); + close(in_fds[1]); + + if (pid < 0) { + err = -pid; + printk("harddog_open - run_helper failed, errno = %d\n", -err); + goto out_close_out; + } + + n = read(in_fds[0], &c, sizeof(c)); + if (n == 0) { + printk("harddog_open - EOF on watchdog pipe\n"); + helper_wait(pid); + err = -EIO; + goto out_close_out; + } + else if (n < 0) { + printk("harddog_open - read of watchdog pipe failed, " + "err = %d\n", errno); + helper_wait(pid); + err = n; + goto out_close_out; + } + *in_fd_ret = in_fds[0]; + *out_fd_ret = out_fds[1]; + return 0; + + out_close_in: + close(in_fds[0]); + close(in_fds[1]); + out_close_out: + close(out_fds[0]); + close(out_fds[1]); + out: + return err; +} + +void stop_watchdog(int in_fd, int out_fd) +{ + close(in_fd); + close(out_fd); +} + +int ping_watchdog(int fd) +{ + int n; + char c = '\n'; + + n = write(fd, &c, sizeof(c)); + if (n != sizeof(c)) { + printk("ping_watchdog - write failed, ret = %d, err = %d\n", + n, errno); + if (n < 0) + return n; + return -EIO; + } + return 1; + +} diff --git a/arch/um/drivers/harddog_user_exp.c b/arch/um/drivers/harddog_user_exp.c new file mode 100644 index 000000000..c74d4b815 --- /dev/null +++ b/arch/um/drivers/harddog_user_exp.c @@ -0,0 +1,9 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include "harddog.h" + +#if IS_MODULE(CONFIG_UML_WATCHDOG) +EXPORT_SYMBOL(start_watchdog); +EXPORT_SYMBOL(stop_watchdog); +EXPORT_SYMBOL(ping_watchdog); +#endif diff --git a/arch/um/drivers/hostaudio_kern.c b/arch/um/drivers/hostaudio_kern.c new file mode 100644 index 000000000..7ce98f148 --- /dev/null +++ b/arch/um/drivers/hostaudio_kern.c @@ -0,0 +1,353 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2002 Steve Schmidtke + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +struct hostaudio_state { + int fd; +}; + +struct hostmixer_state { + int fd; +}; + +#define HOSTAUDIO_DEV_DSP "/dev/sound/dsp" +#define HOSTAUDIO_DEV_MIXER "/dev/sound/mixer" + +/* + * Changed either at boot time or module load time. At boot, this is + * single-threaded; at module load, multiple modules would each have + * their own copy of these variables. + */ +static char *dsp = HOSTAUDIO_DEV_DSP; +static char *mixer = HOSTAUDIO_DEV_MIXER; + +#define DSP_HELP \ +" This is used to specify the host dsp device to the hostaudio driver.\n" \ +" The default is \"" HOSTAUDIO_DEV_DSP "\".\n\n" + +#define MIXER_HELP \ +" This is used to specify the host mixer device to the hostaudio driver.\n"\ +" The default is \"" HOSTAUDIO_DEV_MIXER "\".\n\n" + +module_param(dsp, charp, 0644); +MODULE_PARM_DESC(dsp, DSP_HELP); +module_param(mixer, charp, 0644); +MODULE_PARM_DESC(mixer, MIXER_HELP); + +#ifndef MODULE +static int set_dsp(char *name, int *add) +{ + *add = 0; + dsp = name; + return 0; +} + +__uml_setup("dsp=", set_dsp, "dsp=\n" DSP_HELP); + +static int set_mixer(char *name, int *add) +{ + *add = 0; + mixer = name; + return 0; +} + +__uml_setup("mixer=", set_mixer, "mixer=\n" MIXER_HELP); +#endif + +static DEFINE_MUTEX(hostaudio_mutex); + +/* /dev/dsp file operations */ + +static ssize_t hostaudio_read(struct file *file, char __user *buffer, + size_t count, loff_t *ppos) +{ + struct hostaudio_state *state = file->private_data; + void *kbuf; + int err; + +#ifdef DEBUG + printk(KERN_DEBUG "hostaudio: read called, count = %d\n", count); +#endif + + kbuf = kmalloc(count, GFP_KERNEL); + if (kbuf == NULL) + return -ENOMEM; + + err = os_read_file(state->fd, kbuf, count); + if (err < 0) + goto out; + + if (copy_to_user(buffer, kbuf, err)) + err = -EFAULT; + +out: + kfree(kbuf); + return err; +} + +static ssize_t hostaudio_write(struct file *file, const char __user *buffer, + size_t count, loff_t *ppos) +{ + struct hostaudio_state *state = file->private_data; + void *kbuf; + int err; + +#ifdef DEBUG + printk(KERN_DEBUG "hostaudio: write called, count = %d\n", count); +#endif + + kbuf = memdup_user(buffer, count); + if (IS_ERR(kbuf)) + return PTR_ERR(kbuf); + + err = os_write_file(state->fd, kbuf, count); + if (err < 0) + goto out; + *ppos += err; + + out: + kfree(kbuf); + return err; +} + +static __poll_t hostaudio_poll(struct file *file, + struct poll_table_struct *wait) +{ +#ifdef DEBUG + printk(KERN_DEBUG "hostaudio: poll called (unimplemented)\n"); +#endif + + return 0; +} + +static long hostaudio_ioctl(struct file *file, + unsigned int cmd, unsigned long arg) +{ + struct hostaudio_state *state = file->private_data; + unsigned long data = 0; + int err; + +#ifdef DEBUG + printk(KERN_DEBUG "hostaudio: ioctl called, cmd = %u\n", cmd); +#endif + switch(cmd){ + case SNDCTL_DSP_SPEED: + case SNDCTL_DSP_STEREO: + case SNDCTL_DSP_GETBLKSIZE: + case SNDCTL_DSP_CHANNELS: + case SNDCTL_DSP_SUBDIVIDE: + case SNDCTL_DSP_SETFRAGMENT: + if (get_user(data, (int __user *) arg)) + return -EFAULT; + break; + default: + break; + } + + err = os_ioctl_generic(state->fd, cmd, (unsigned long) &data); + + switch(cmd){ + case SNDCTL_DSP_SPEED: + case SNDCTL_DSP_STEREO: + case SNDCTL_DSP_GETBLKSIZE: + case SNDCTL_DSP_CHANNELS: + case SNDCTL_DSP_SUBDIVIDE: + case SNDCTL_DSP_SETFRAGMENT: + if (put_user(data, (int __user *) arg)) + return -EFAULT; + break; + default: + break; + } + + return err; +} + +static int hostaudio_open(struct inode *inode, struct file *file) +{ + struct hostaudio_state *state; + int r = 0, w = 0; + int ret; + +#ifdef DEBUG + kernel_param_lock(THIS_MODULE); + printk(KERN_DEBUG "hostaudio: open called (host: %s)\n", dsp); + kernel_param_unlock(THIS_MODULE); +#endif + + state = kmalloc_obj(struct hostaudio_state); + if (state == NULL) + return -ENOMEM; + + if (file->f_mode & FMODE_READ) + r = 1; + if (file->f_mode & FMODE_WRITE) + w = 1; + + kernel_param_lock(THIS_MODULE); + mutex_lock(&hostaudio_mutex); + ret = os_open_file(dsp, of_set_rw(OPENFLAGS(), r, w), 0); + mutex_unlock(&hostaudio_mutex); + kernel_param_unlock(THIS_MODULE); + + if (ret < 0) { + kfree(state); + return ret; + } + state->fd = ret; + file->private_data = state; + return 0; +} + +static int hostaudio_release(struct inode *inode, struct file *file) +{ + struct hostaudio_state *state = file->private_data; + +#ifdef DEBUG + printk(KERN_DEBUG "hostaudio: release called\n"); +#endif + os_close_file(state->fd); + kfree(state); + + return 0; +} + +/* /dev/mixer file operations */ + +static long hostmixer_ioctl_mixdev(struct file *file, + unsigned int cmd, unsigned long arg) +{ + struct hostmixer_state *state = file->private_data; + +#ifdef DEBUG + printk(KERN_DEBUG "hostmixer: ioctl called\n"); +#endif + + return os_ioctl_generic(state->fd, cmd, arg); +} + +static int hostmixer_open_mixdev(struct inode *inode, struct file *file) +{ + struct hostmixer_state *state; + int r = 0, w = 0; + int ret; + +#ifdef DEBUG + printk(KERN_DEBUG "hostmixer: open called (host: %s)\n", mixer); +#endif + + state = kmalloc_obj(struct hostmixer_state); + if (state == NULL) + return -ENOMEM; + + if (file->f_mode & FMODE_READ) + r = 1; + if (file->f_mode & FMODE_WRITE) + w = 1; + + kernel_param_lock(THIS_MODULE); + mutex_lock(&hostaudio_mutex); + ret = os_open_file(mixer, of_set_rw(OPENFLAGS(), r, w), 0); + mutex_unlock(&hostaudio_mutex); + kernel_param_unlock(THIS_MODULE); + + if (ret < 0) { + kernel_param_lock(THIS_MODULE); + printk(KERN_ERR "hostaudio_open_mixdev failed to open '%s', " + "err = %d\n", dsp, -ret); + kernel_param_unlock(THIS_MODULE); + kfree(state); + return ret; + } + + file->private_data = state; + return 0; +} + +static int hostmixer_release(struct inode *inode, struct file *file) +{ + struct hostmixer_state *state = file->private_data; + +#ifdef DEBUG + printk(KERN_DEBUG "hostmixer: release called\n"); +#endif + + os_close_file(state->fd); + kfree(state); + + return 0; +} + +/* kernel module operations */ + +static const struct file_operations hostaudio_fops = { + .owner = THIS_MODULE, + .read = hostaudio_read, + .write = hostaudio_write, + .poll = hostaudio_poll, + .unlocked_ioctl = hostaudio_ioctl, + .compat_ioctl = compat_ptr_ioctl, + .mmap = NULL, + .open = hostaudio_open, + .release = hostaudio_release, +}; + +static const struct file_operations hostmixer_fops = { + .owner = THIS_MODULE, + .unlocked_ioctl = hostmixer_ioctl_mixdev, + .open = hostmixer_open_mixdev, + .release = hostmixer_release, +}; + +static struct { + int dev_audio; + int dev_mixer; +} module_data; + +MODULE_AUTHOR("Steve Schmidtke"); +MODULE_DESCRIPTION("UML Audio Relay"); +MODULE_LICENSE("GPL"); + +static int __init hostaudio_init_module(void) +{ + kernel_param_lock(THIS_MODULE); + printk(KERN_INFO "UML Audio Relay (host dsp = %s, host mixer = %s)\n", + dsp, mixer); + kernel_param_unlock(THIS_MODULE); + + module_data.dev_audio = register_sound_dsp(&hostaudio_fops, -1); + if (module_data.dev_audio < 0) { + printk(KERN_ERR "hostaudio: couldn't register DSP device!\n"); + return -ENODEV; + } + + module_data.dev_mixer = register_sound_mixer(&hostmixer_fops, -1); + if (module_data.dev_mixer < 0) { + printk(KERN_ERR "hostmixer: couldn't register mixer " + "device!\n"); + unregister_sound_dsp(module_data.dev_audio); + return -ENODEV; + } + + return 0; +} + +static void __exit hostaudio_cleanup_module (void) +{ + unregister_sound_mixer(module_data.dev_mixer); + unregister_sound_dsp(module_data.dev_audio); +} + +module_init(hostaudio_init_module); +module_exit(hostaudio_cleanup_module); diff --git a/arch/um/drivers/line.c b/arch/um/drivers/line.c new file mode 100644 index 000000000..c851a5928 --- /dev/null +++ b/arch/um/drivers/line.c @@ -0,0 +1,771 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + */ + +#include +#include +#include +#include + +#include "chan.h" +#include +#include +#include +#include + +#define LINE_BUFSIZE 4096 + +static irqreturn_t line_interrupt(int irq, void *data) +{ + struct chan *chan = data; + struct line *line = chan->line; + + if (line) + chan_interrupt(line, irq); + + return IRQ_HANDLED; +} + +/* + * Returns the free space inside the ring buffer of this line. + * + * Should be called while holding line->lock (this does not modify data). + */ +static unsigned int write_room(struct line *line) +{ + int n; + + if (line->buffer == NULL) + return LINE_BUFSIZE - 1; + + /* This is for the case where the buffer is wrapped! */ + n = line->head - line->tail; + + if (n <= 0) + n += LINE_BUFSIZE; /* The other case */ + return n - 1; +} + +unsigned int line_write_room(struct tty_struct *tty) +{ + struct line *line = tty->driver_data; + unsigned long flags; + unsigned int room; + + spin_lock_irqsave(&line->lock, flags); + room = write_room(line); + spin_unlock_irqrestore(&line->lock, flags); + + return room; +} + +unsigned int line_chars_in_buffer(struct tty_struct *tty) +{ + struct line *line = tty->driver_data; + unsigned long flags; + unsigned int ret; + + spin_lock_irqsave(&line->lock, flags); + /* write_room subtracts 1 for the needed NULL, so we readd it.*/ + ret = LINE_BUFSIZE - (write_room(line) + 1); + spin_unlock_irqrestore(&line->lock, flags); + + return ret; +} + +/* + * This copies the content of buf into the circular buffer associated with + * this line. + * The return value is the number of characters actually copied, i.e. the ones + * for which there was space: this function is not supposed to ever flush out + * the circular buffer. + * + * Must be called while holding line->lock! + */ +static int buffer_data(struct line *line, const u8 *buf, size_t len) +{ + int end, room; + + if (line->buffer == NULL) { + line->buffer = kmalloc(LINE_BUFSIZE, GFP_ATOMIC); + if (line->buffer == NULL) { + printk(KERN_ERR "buffer_data - atomic allocation " + "failed\n"); + return 0; + } + line->head = line->buffer; + line->tail = line->buffer; + } + + room = write_room(line); + len = (len > room) ? room : len; + + end = line->buffer + LINE_BUFSIZE - line->tail; + + if (len < end) { + memcpy(line->tail, buf, len); + line->tail += len; + } + else { + /* The circular buffer is wrapping */ + memcpy(line->tail, buf, end); + buf += end; + memcpy(line->buffer, buf, len - end); + line->tail = line->buffer + len - end; + } + + return len; +} + +/* + * Flushes the ring buffer to the output channels. That is, write_chan is + * called, passing it line->head as buffer, and an appropriate count. + * + * On exit, returns 1 when the buffer is empty, + * 0 when the buffer is not empty on exit, + * and -errno when an error occurred. + * + * Must be called while holding line->lock!*/ +static int flush_buffer(struct line *line) +{ + int n, count; + + if ((line->buffer == NULL) || (line->head == line->tail)) + return 1; + + if (line->tail < line->head) { + /* line->buffer + LINE_BUFSIZE is the end of the buffer! */ + count = line->buffer + LINE_BUFSIZE - line->head; + + n = write_chan(line->chan_out, line->head, count, + line->write_irq); + if (n < 0) + return n; + if (n == count) { + /* + * We have flushed from ->head to buffer end, now we + * must flush only from the beginning to ->tail. + */ + line->head = line->buffer; + } else { + line->head += n; + return 0; + } + } + + count = line->tail - line->head; + n = write_chan(line->chan_out, line->head, count, + line->write_irq); + + if (n < 0) + return n; + + line->head += n; + return line->head == line->tail; +} + +void line_flush_buffer(struct tty_struct *tty) +{ + struct line *line = tty->driver_data; + unsigned long flags; + + spin_lock_irqsave(&line->lock, flags); + flush_buffer(line); + spin_unlock_irqrestore(&line->lock, flags); +} + +/* + * We map both ->flush_chars and ->put_char (which go in pair) onto + * ->flush_buffer and ->write. Hope it's not that bad. + */ +void line_flush_chars(struct tty_struct *tty) +{ + line_flush_buffer(tty); +} + +ssize_t line_write(struct tty_struct *tty, const u8 *buf, size_t len) +{ + struct line *line = tty->driver_data; + unsigned long flags; + int n, ret = 0; + + spin_lock_irqsave(&line->lock, flags); + if (line->head != line->tail) + ret = buffer_data(line, buf, len); + else { + n = write_chan(line->chan_out, buf, len, + line->write_irq); + if (n < 0) { + ret = n; + goto out_up; + } + + len -= n; + ret += n; + if (len > 0) + ret += buffer_data(line, buf + n, len); + } +out_up: + spin_unlock_irqrestore(&line->lock, flags); + return ret; +} + +void line_throttle(struct tty_struct *tty) +{ + struct line *line = tty->driver_data; + + deactivate_chan(line->chan_in, line->read_irq); + line->throttled = 1; +} + +void line_unthrottle(struct tty_struct *tty) +{ + struct line *line = tty->driver_data; + + line->throttled = 0; + chan_interrupt(line, line->read_irq); +} + +static irqreturn_t line_write_interrupt(int irq, void *data) +{ + struct chan *chan = data; + struct line *line = chan->line; + int err; + + /* + * Interrupts are disabled here because genirq keep irqs disabled when + * calling the action handler. + */ + + spin_lock(&line->lock); + err = flush_buffer(line); + if (err == 0) { + spin_unlock(&line->lock); + return IRQ_NONE; + } else if ((err < 0) && (err != -EAGAIN)) { + line->head = line->buffer; + line->tail = line->buffer; + } + spin_unlock(&line->lock); + + tty_port_tty_wakeup(&line->port); + + return IRQ_HANDLED; +} + +int line_setup_irq(int fd, int input, int output, struct line *line, void *data) +{ + const struct line_driver *driver = line->driver; + int err; + + if (input) { + err = um_request_irq(UM_IRQ_ALLOC, fd, IRQ_READ, + line_interrupt, 0, + driver->read_irq_name, data); + if (err < 0) + return err; + + line->read_irq = err; + } + + if (output) { + err = um_request_irq(UM_IRQ_ALLOC, fd, IRQ_WRITE, + line_write_interrupt, 0, + driver->write_irq_name, data); + if (err < 0) + return err; + + line->write_irq = err; + } + + return 0; +} + +static int line_activate(struct tty_port *port, struct tty_struct *tty) +{ + int ret; + struct line *line = tty->driver_data; + + ret = enable_chan(line); + if (ret) + return ret; + + if (!line->sigio) { + chan_enable_winch(line->chan_out, port); + line->sigio = 1; + } + + chan_window_size(line, &tty->winsize.ws_row, + &tty->winsize.ws_col); + + return 0; +} + +static void unregister_winch(struct tty_struct *tty); + +static void line_destruct(struct tty_port *port) +{ + struct tty_struct *tty = tty_port_tty_get(port); + struct line *line = tty->driver_data; + + if (line->sigio) { + unregister_winch(tty); + line->sigio = 0; + } +} + +static const struct tty_port_operations line_port_ops = { + .activate = line_activate, + .destruct = line_destruct, +}; + +int line_open(struct tty_struct *tty, struct file *filp) +{ + struct line *line = tty->driver_data; + + return tty_port_open(&line->port, tty, filp); +} + +int line_install(struct tty_driver *driver, struct tty_struct *tty, + struct line *line) +{ + int ret; + + ret = tty_standard_install(driver, tty); + if (ret) + return ret; + + tty->driver_data = line; + + return 0; +} + +void line_close(struct tty_struct *tty, struct file * filp) +{ + struct line *line = tty->driver_data; + + tty_port_close(&line->port, tty, filp); +} + +void line_hangup(struct tty_struct *tty) +{ + struct line *line = tty->driver_data; + + tty_port_hangup(&line->port); +} + +void close_lines(struct line *lines, int nlines) +{ + int i; + + for(i = 0; i < nlines; i++) + close_chan(&lines[i]); +} + +int setup_one_line(struct line *lines, int n, char *init, + const struct chan_opts *opts, char **error_out) +{ + struct line *line = &lines[n]; + struct tty_driver *driver = line->driver->driver; + int err = -EINVAL; + + if (line->port.count) { + *error_out = "Device is already open"; + goto out; + } + + if (!strcmp(init, "none")) { + if (line->valid) { + line->valid = 0; + kfree(line->init_str); + tty_unregister_device(driver, n); + parse_chan_pair(NULL, line, n, opts, error_out); + err = 0; + } + *error_out = "configured as 'none'"; + } else { + char *new = kstrdup(init, GFP_KERNEL); + if (!new) { + *error_out = "Failed to allocate memory"; + return -ENOMEM; + } + if (line->valid) { + tty_unregister_device(driver, n); + kfree(line->init_str); + } + line->init_str = new; + line->valid = 1; + err = parse_chan_pair(new, line, n, opts, error_out); + if (!err) { + struct device *d = tty_port_register_device(&line->port, + driver, n, NULL); + if (IS_ERR(d)) { + *error_out = "Failed to register device"; + err = PTR_ERR(d); + parse_chan_pair(NULL, line, n, opts, error_out); + } + } + if (err) { + *error_out = "failed to parse channel pair"; + line->init_str = NULL; + line->valid = 0; + kfree(new); + } + } +out: + return err; +} + +/* + * Common setup code for both startup command line and mconsole initialization. + * @lines contains the array (of size @num) to modify; + * @init is the setup string; + * @error_out is an error string in the case of failure; + */ + +int line_setup(char **conf, unsigned int num, char **def, + char *init, char *name) +{ + char *error; + + if (*init == '=') { + /* + * We said con=/ssl= instead of con#=, so we are configuring all + * consoles at once. + */ + *def = init + 1; + } else { + char *end; + unsigned n = simple_strtoul(init, &end, 0); + + if (*end != '=') { + error = "Couldn't parse device number"; + goto out; + } + if (n >= num) { + error = "Device number out of range"; + goto out; + } + conf[n] = end + 1; + } + return 0; + +out: + printk(KERN_ERR "Failed to set up %s with " + "configuration string \"%s\" : %s\n", name, init, error); + return -EINVAL; +} + +int line_config(struct line *lines, unsigned int num, char *str, + const struct chan_opts *opts, char **error_out) +{ + char *end; + int n; + + if (*str == '=') { + *error_out = "Can't configure all devices from mconsole"; + return -EINVAL; + } + + n = simple_strtoul(str, &end, 0); + if (*end++ != '=') { + *error_out = "Couldn't parse device number"; + return -EINVAL; + } + if (n >= num) { + *error_out = "Device number out of range"; + return -EINVAL; + } + + return setup_one_line(lines, n, end, opts, error_out); +} + +int line_get_config(char *name, struct line *lines, unsigned int num, char *str, + int size, char **error_out) +{ + struct line *line; + char *end; + int dev, n = 0; + + dev = simple_strtoul(name, &end, 0); + if ((*end != '\0') || (end == name)) { + *error_out = "line_get_config failed to parse device number"; + return 0; + } + + if ((dev < 0) || (dev >= num)) { + *error_out = "device number out of range"; + return 0; + } + + line = &lines[dev]; + + if (!line->valid) + CONFIG_CHUNK(str, size, n, "none", 1); + else { + struct tty_struct *tty = tty_port_tty_get(&line->port); + if (tty == NULL) { + CONFIG_CHUNK(str, size, n, line->init_str, 1); + } else { + n = chan_config_string(line, str, size, error_out); + tty_kref_put(tty); + } + } + + return n; +} + +int line_id(char **str, int *start_out, int *end_out) +{ + char *end; + int n; + + n = simple_strtoul(*str, &end, 0); + if ((*end != '\0') || (end == *str)) + return -1; + + *str = end; + *start_out = n; + *end_out = n; + return n; +} + +int line_remove(struct line *lines, unsigned int num, int n, char **error_out) +{ + if (n >= num) { + *error_out = "Device number out of range"; + return -EINVAL; + } + return setup_one_line(lines, n, "none", NULL, error_out); +} + +int register_lines(struct line_driver *line_driver, + const struct tty_operations *ops, + struct line *lines, int nlines) +{ + struct tty_driver *driver; + int err; + int i; + + driver = tty_alloc_driver(nlines, TTY_DRIVER_REAL_RAW | + TTY_DRIVER_DYNAMIC_DEV); + if (IS_ERR(driver)) + return PTR_ERR(driver); + + driver->driver_name = line_driver->name; + driver->name = line_driver->device_name; + driver->major = line_driver->major; + driver->minor_start = line_driver->minor_start; + driver->type = line_driver->type; + driver->subtype = line_driver->subtype; + driver->init_termios = tty_std_termios; + + for (i = 0; i < nlines; i++) { + tty_port_init(&lines[i].port); + lines[i].port.ops = &line_port_ops; + spin_lock_init(&lines[i].lock); + lines[i].driver = line_driver; + INIT_LIST_HEAD(&lines[i].chan_list); + } + tty_set_operations(driver, ops); + + err = tty_register_driver(driver); + if (err) { + printk(KERN_ERR "register_lines : can't register %s driver\n", + line_driver->name); + tty_driver_kref_put(driver); + for (i = 0; i < nlines; i++) + tty_port_destroy(&lines[i].port); + return err; + } + + line_driver->driver = driver; + mconsole_register_dev(&line_driver->mc); + return 0; +} + +static DEFINE_SPINLOCK(winch_handler_lock); +static LIST_HEAD(winch_handlers); + +struct winch { + struct list_head list; + int fd; + int tty_fd; + int pid; + struct tty_port *port; + unsigned long stack; + struct work_struct work; +}; + +static void __free_winch(struct work_struct *work) +{ + struct winch *winch = container_of(work, struct winch, work); + um_free_irq(WINCH_IRQ, winch); + + if (winch->pid != -1) + os_kill_process(winch->pid, 1); + if (winch->stack != 0) + free_stack(winch->stack, 0); + kfree(winch); +} + +static void free_winch(struct winch *winch) +{ + int fd = winch->fd; + winch->fd = -1; + if (fd != -1) + os_close_file(fd); + __free_winch(&winch->work); +} + +static irqreturn_t winch_interrupt(int irq, void *data) +{ + struct winch *winch = data; + struct tty_struct *tty; + struct line *line; + int fd = winch->fd; + int err; + char c; + struct pid *pgrp; + + if (fd != -1) { + err = generic_read(fd, &c, NULL); + /* A read of 2 means the winch thread failed and has warned */ + if (err < 0 || (err == 1 && c == 2)) { + if (err != -EAGAIN) { + winch->fd = -1; + list_del(&winch->list); + os_close_file(fd); + if (err < 0) { + printk(KERN_ERR "winch_interrupt : read failed, errno = %d\n", + -err); + printk(KERN_ERR "fd %d is losing SIGWINCH support\n", + winch->tty_fd); + } + INIT_WORK(&winch->work, __free_winch); + schedule_work(&winch->work); + return IRQ_HANDLED; + } + goto out; + } + } + tty = tty_port_tty_get(winch->port); + if (tty != NULL) { + line = tty->driver_data; + if (line != NULL) { + chan_window_size(line, &tty->winsize.ws_row, + &tty->winsize.ws_col); + pgrp = tty_get_pgrp(tty); + if (pgrp) + kill_pgrp(pgrp, SIGWINCH, 1); + put_pid(pgrp); + } + tty_kref_put(tty); + } + out: + return IRQ_HANDLED; +} + +void register_winch_irq(int fd, int tty_fd, int pid, struct tty_port *port, + unsigned long stack) +{ + struct winch *winch; + + winch = kmalloc_obj(*winch); + if (winch == NULL) { + printk(KERN_ERR "register_winch_irq - kmalloc failed\n"); + goto cleanup; + } + + *winch = ((struct winch) { .fd = fd, + .tty_fd = tty_fd, + .pid = pid, + .port = port, + .stack = stack }); + + spin_lock(&winch_handler_lock); + list_add(&winch->list, &winch_handlers); + spin_unlock(&winch_handler_lock); + + if (um_request_irq(WINCH_IRQ, fd, IRQ_READ, winch_interrupt, + IRQF_SHARED, "winch", winch) < 0) { + printk(KERN_ERR "register_winch_irq - failed to register " + "IRQ\n"); + spin_lock(&winch_handler_lock); + list_del(&winch->list); + spin_unlock(&winch_handler_lock); + goto out_free; + } + + return; + + out_free: + kfree(winch); + cleanup: + os_kill_process(pid, 1); + os_close_file(fd); + if (stack != 0) + free_stack(stack, 0); +} + +static void unregister_winch(struct tty_struct *tty) +{ + struct list_head *ele, *next; + struct winch *winch; + struct tty_struct *wtty; + + spin_lock(&winch_handler_lock); + + list_for_each_safe(ele, next, &winch_handlers) { + winch = list_entry(ele, struct winch, list); + wtty = tty_port_tty_get(winch->port); + if (wtty == tty) { + list_del(&winch->list); + spin_unlock(&winch_handler_lock); + free_winch(winch); + break; + } + tty_kref_put(wtty); + } + spin_unlock(&winch_handler_lock); +} + +static void winch_cleanup(void) +{ + struct winch *winch; + + spin_lock(&winch_handler_lock); + while ((winch = list_first_entry_or_null(&winch_handlers, + struct winch, list))) { + list_del(&winch->list); + spin_unlock(&winch_handler_lock); + + free_winch(winch); + + spin_lock(&winch_handler_lock); + } + + spin_unlock(&winch_handler_lock); +} +__uml_exitcall(winch_cleanup); + +char *add_xterm_umid(char *base) +{ + char *umid, *title; + int len; + + umid = get_umid(); + if (*umid == '\0') + return base; + + len = strlen(base) + strlen(" ()") + strlen(umid) + 1; + title = kmalloc(len, GFP_KERNEL); + if (title == NULL) { + printk(KERN_ERR "Failed to allocate buffer for xterm title\n"); + return base; + } + + snprintf(title, len, "%s (%s)", base, umid); + return title; +} diff --git a/arch/um/drivers/line.h b/arch/um/drivers/line.h new file mode 100644 index 000000000..e8bd6f3df --- /dev/null +++ b/arch/um/drivers/line.h @@ -0,0 +1,96 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2001, 2002 Jeff Dike (jdike@karaya.com) + */ + +#ifndef __LINE_H__ +#define __LINE_H__ + +#include +#include +#include +#include +#include +#include +#include "chan_user.h" +#include "mconsole_kern.h" + +/* There's only two modifiable fields in this - .mc.list and .driver */ +struct line_driver { + const char *name; + const char *device_name; + const short major; + const short minor_start; + const short type; + const short subtype; + const char *read_irq_name; + const char *write_irq_name; + struct mc_device mc; + struct tty_driver *driver; +}; + +struct line { + struct tty_port port; + int valid; + + int read_irq, write_irq; + + char *init_str; + struct list_head chan_list; + struct chan *chan_in, *chan_out; + + /*This lock is actually, mostly, local to*/ + spinlock_t lock; + int throttled; + /* Yes, this is a real circular buffer. + * XXX: And this should become a struct kfifo! + * + * buffer points to a buffer allocated on demand, of length + * LINE_BUFSIZE, head to the start of the ring, tail to the end.*/ + u8 *buffer; + u8 *head; + u8 *tail; + + int sigio; + struct delayed_work task; + const struct line_driver *driver; +}; + +extern void line_close(struct tty_struct *tty, struct file * filp); +extern int line_open(struct tty_struct *tty, struct file *filp); +extern int line_install(struct tty_driver *driver, struct tty_struct *tty, + struct line *line); +extern void line_cleanup(struct tty_struct *tty); +extern void line_hangup(struct tty_struct *tty); +extern int line_setup(char **conf, unsigned nlines, char **def, + char *init, char *name); +extern ssize_t line_write(struct tty_struct *tty, const u8 *buf, size_t len); +extern unsigned int line_chars_in_buffer(struct tty_struct *tty); +extern void line_flush_buffer(struct tty_struct *tty); +extern void line_flush_chars(struct tty_struct *tty); +extern unsigned int line_write_room(struct tty_struct *tty); +extern void line_throttle(struct tty_struct *tty); +extern void line_unthrottle(struct tty_struct *tty); + +extern char *add_xterm_umid(char *base); +extern int line_setup_irq(int fd, int input, int output, struct line *line, + void *data); +extern void line_close_chan(struct line *line); +extern int register_lines(struct line_driver *line_driver, + const struct tty_operations *driver, + struct line *lines, int nlines); +extern int setup_one_line(struct line *lines, int n, char *init, + const struct chan_opts *opts, char **error_out); +extern void close_lines(struct line *lines, int nlines); + +extern int line_config(struct line *lines, unsigned int sizeof_lines, + char *str, const struct chan_opts *opts, + char **error_out); +extern int line_id(char **str, int *start_out, int *end_out); +extern int line_remove(struct line *lines, unsigned int sizeof_lines, int n, + char **error_out); +extern int line_get_config(char *dev, struct line *lines, + unsigned int sizeof_lines, char *str, + int size, char **error_out); + +#endif diff --git a/arch/um/drivers/mconsole.h b/arch/um/drivers/mconsole.h new file mode 100644 index 000000000..635637830 --- /dev/null +++ b/arch/um/drivers/mconsole.h @@ -0,0 +1,98 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2001 Lennert Buytenhek (buytenh@gnu.org) + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + */ + +#ifndef __MCONSOLE_H__ +#define __MCONSOLE_H__ + +#ifdef __UM_HOST__ +#include +#define u32 uint32_t +#endif + +#include + +#define MCONSOLE_MAGIC (0xcafebabe) +#define MCONSOLE_MAX_DATA (512) +#define MCONSOLE_VERSION 2 + +struct mconsole_request { + u32 magic; + u32 version; + u32 len; + char data[MCONSOLE_MAX_DATA]; +}; + +struct mconsole_reply { + u32 err; + u32 more; + u32 len; + char data[MCONSOLE_MAX_DATA]; +}; + +struct mconsole_notify { + u32 magic; + u32 version; + enum { MCONSOLE_SOCKET, MCONSOLE_PANIC, MCONSOLE_HANG, + MCONSOLE_USER_NOTIFY } type; + u32 len; + char data[MCONSOLE_MAX_DATA]; +}; + +struct mc_request; + +enum mc_context { MCONSOLE_INTR, MCONSOLE_PROC }; + +struct mconsole_command +{ + char *command; + void (*handler)(struct mc_request *req); + enum mc_context context; +}; + +struct mc_request +{ + int len; + int as_interrupt; + + int originating_fd; + unsigned int originlen; + unsigned char origin[128]; /* sockaddr_un */ + + struct mconsole_request request; + struct mconsole_command *cmd; + struct uml_pt_regs regs; +}; + +extern char mconsole_socket_name[]; + +extern int mconsole_unlink_socket(void); +extern int mconsole_reply_len(struct mc_request *req, const char *reply, + int len, int err, int more); +extern int mconsole_reply(struct mc_request *req, const char *str, int err, + int more); + +extern void mconsole_version(struct mc_request *req); +extern void mconsole_help(struct mc_request *req); +extern void mconsole_halt(struct mc_request *req); +extern void mconsole_reboot(struct mc_request *req); +extern void mconsole_config(struct mc_request *req); +extern void mconsole_remove(struct mc_request *req); +extern void mconsole_sysrq(struct mc_request *req); +extern void mconsole_cad(struct mc_request *req); +extern void mconsole_stop(struct mc_request *req); +extern void mconsole_go(struct mc_request *req); +extern void mconsole_log(struct mc_request *req); +extern void mconsole_proc(struct mc_request *req); +extern void mconsole_stack(struct mc_request *req); + +extern int mconsole_get_request(int fd, struct mc_request *req); +extern int mconsole_notify(char *sock_name, int type, const void *data, + int len); +extern char *mconsole_notify_socket(void); +extern void lock_notify(void); +extern void unlock_notify(void); + +#endif diff --git a/arch/um/drivers/mconsole_kern.c b/arch/um/drivers/mconsole_kern.c new file mode 100644 index 000000000..e2a9e8879 --- /dev/null +++ b/arch/um/drivers/mconsole_kern.c @@ -0,0 +1,871 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 Lennert Buytenhek (buytenh@gnu.org) + * Copyright (C) 2001 - 2008 Jeff Dike (jdike@{addtoit,linux.intel}.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include "mconsole.h" +#include "mconsole_kern.h" +#include + +static struct vfsmount *proc_mnt = NULL; + +static int do_unlink_socket(struct notifier_block *notifier, + unsigned long what, void *data) +{ + return mconsole_unlink_socket(); +} + + +static struct notifier_block reboot_notifier = { + .notifier_call = do_unlink_socket, + .priority = 0, +}; + +/* Safe without explicit locking for now. Tasklets provide their own + * locking, and the interrupt handler is safe because it can't interrupt + * itself and it can only happen on CPU 0. + */ + +static LIST_HEAD(mc_requests); + +static void mc_work_proc(struct work_struct *unused) +{ + struct mconsole_entry *req; + unsigned long flags; + + while (!list_empty(&mc_requests)) { + local_irq_save(flags); + req = list_entry(mc_requests.next, struct mconsole_entry, list); + list_del(&req->list); + local_irq_restore(flags); + req->request.cmd->handler(&req->request); + kfree(req); + } +} + +static DECLARE_WORK(mconsole_work, mc_work_proc); + +static irqreturn_t mconsole_interrupt(int irq, void *dev_id) +{ + /* long to avoid size mismatch warnings from gcc */ + long fd; + struct mconsole_entry *new; + static struct mc_request req; /* that's OK */ + + fd = (long) dev_id; + while (mconsole_get_request(fd, &req)) { + if (req.cmd->context == MCONSOLE_INTR) + (*req.cmd->handler)(&req); + else { + new = kmalloc_obj(*new, GFP_NOWAIT); + if (new == NULL) + mconsole_reply(&req, "Out of memory", 1, 0); + else { + new->request = req; + new->request.regs = get_irq_regs()->regs; + list_add(&new->list, &mc_requests); + } + } + } + if (!list_empty(&mc_requests)) + schedule_work(&mconsole_work); + return IRQ_HANDLED; +} + +void mconsole_version(struct mc_request *req) +{ + char version[256]; + + sprintf(version, "%s %s %s %s %s", utsname()->sysname, + utsname()->nodename, utsname()->release, utsname()->version, + utsname()->machine); + mconsole_reply(req, version, 0, 0); +} + +void mconsole_log(struct mc_request *req) +{ + int len; + char *ptr = req->request.data; + + ptr += strlen("log "); + + len = req->len - (ptr - req->request.data); + printk(KERN_WARNING "%.*s", len, ptr); + mconsole_reply(req, "", 0, 0); +} + +void mconsole_proc(struct mc_request *req) +{ + struct vfsmount *mnt = proc_mnt; + char *buf; + int len; + struct file *file; + int first_chunk = 1; + char *ptr = req->request.data; + loff_t pos = 0; + + ptr += strlen("proc"); + ptr = skip_spaces(ptr); + + if (!mnt) { + mconsole_reply(req, "Proc not available", 1, 0); + goto out; + } + file = file_open_root_mnt(mnt, ptr, O_RDONLY, 0); + if (IS_ERR(file)) { + mconsole_reply(req, "Failed to open file", 1, 0); + printk(KERN_ERR "open /proc/%s: %ld\n", ptr, PTR_ERR(file)); + goto out; + } + + buf = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (buf == NULL) { + mconsole_reply(req, "Failed to allocate buffer", 1, 0); + goto out_fput; + } + + do { + len = kernel_read(file, buf, PAGE_SIZE - 1, &pos); + if (len < 0) { + mconsole_reply(req, "Read of file failed", 1, 0); + goto out_free; + } + /* Begin the file content on his own line. */ + if (first_chunk) { + mconsole_reply(req, "\n", 0, 1); + first_chunk = 0; + } + buf[len] = '\0'; + mconsole_reply(req, buf, 0, (len > 0)); + } while (len > 0); + out_free: + kfree(buf); + out_fput: + fput(file); + out: ; +} + +#define UML_MCONSOLE_HELPTEXT \ +"Commands: \n\ + version - Get kernel version \n\ + help - Print this message \n\ + halt - Halt UML \n\ + reboot - Reboot UML \n\ + config = - Add a new device to UML; \n\ + same syntax as command line \n\ + config - Query the configuration of a device \n\ + remove - Remove a device from UML \n\ + sysrq - Performs the SysRq action controlled by the letter \n\ + cad - invoke the Ctrl-Alt-Del handler \n\ + stop - pause the UML; it will do nothing until it receives a 'go' \n\ + go - continue the UML after a 'stop' \n\ + log - make UML enter into the kernel log\n\ + proc - returns the contents of the UML's /proc/\n\ + stack - returns the stack of the specified pid\n\ +" + +void mconsole_help(struct mc_request *req) +{ + mconsole_reply(req, UML_MCONSOLE_HELPTEXT, 0, 0); +} + +void mconsole_halt(struct mc_request *req) +{ + mconsole_reply(req, "", 0, 0); + machine_halt(); +} + +void mconsole_reboot(struct mc_request *req) +{ + mconsole_reply(req, "", 0, 0); + machine_restart(NULL); +} + +void mconsole_cad(struct mc_request *req) +{ + mconsole_reply(req, "", 0, 0); + ctrl_alt_del(); +} + +void mconsole_go(struct mc_request *req) +{ + mconsole_reply(req, "Not stopped", 1, 0); +} + +void mconsole_stop(struct mc_request *req) +{ + block_signals(); + os_set_fd_block(req->originating_fd, 1); + mconsole_reply(req, "stopped", 0, 0); + for (;;) { + if (!mconsole_get_request(req->originating_fd, req)) + continue; + if (req->cmd->handler == mconsole_go) + break; + if (req->cmd->handler == mconsole_stop) { + mconsole_reply(req, "Already stopped", 1, 0); + continue; + } + if (req->cmd->handler == mconsole_sysrq) { + struct pt_regs *old_regs; + old_regs = set_irq_regs((struct pt_regs *)&req->regs); + mconsole_sysrq(req); + set_irq_regs(old_regs); + continue; + } + (*req->cmd->handler)(req); + } + os_set_fd_block(req->originating_fd, 0); + mconsole_reply(req, "", 0, 0); + unblock_signals(); +} + +static DEFINE_SPINLOCK(mc_devices_lock); +static LIST_HEAD(mconsole_devices); + +void mconsole_register_dev(struct mc_device *new) +{ + spin_lock(&mc_devices_lock); + BUG_ON(!list_empty(&new->list)); + list_add(&new->list, &mconsole_devices); + spin_unlock(&mc_devices_lock); +} + +static struct mc_device *mconsole_find_dev(char *name) +{ + struct list_head *ele; + struct mc_device *dev; + + list_for_each(ele, &mconsole_devices) { + dev = list_entry(ele, struct mc_device, list); + if (!strncmp(name, dev->name, strlen(dev->name))) + return dev; + } + return NULL; +} + +#define UNPLUGGED_PER_PAGE \ + ((PAGE_SIZE - sizeof(struct list_head)) / sizeof(unsigned long)) + +struct unplugged_pages { + struct list_head list; + void *pages[UNPLUGGED_PER_PAGE]; +}; + +static DEFINE_MUTEX(plug_mem_mutex); +static unsigned long long unplugged_pages_count; +static LIST_HEAD(unplugged_pages); +static int unplug_index = UNPLUGGED_PER_PAGE; + +static int mem_config(char *str, char **error_out) +{ + unsigned long long diff; + int err = -EINVAL, i, add; + char *ret; + + if (str[0] != '=') { + *error_out = "Expected '=' after 'mem'"; + goto out; + } + + str++; + if (str[0] == '-') + add = 0; + else if (str[0] == '+') { + add = 1; + } + else { + *error_out = "Expected increment to start with '-' or '+'"; + goto out; + } + + str++; + diff = memparse(str, &ret); + if (*ret != '\0') { + *error_out = "Failed to parse memory increment"; + goto out; + } + + diff /= PAGE_SIZE; + + mutex_lock(&plug_mem_mutex); + for (i = 0; i < diff; i++) { + struct unplugged_pages *unplugged; + void *addr; + + if (add) { + if (list_empty(&unplugged_pages)) + break; + + unplugged = list_entry(unplugged_pages.next, + struct unplugged_pages, list); + if (unplug_index > 0) + addr = unplugged->pages[--unplug_index]; + else { + list_del(&unplugged->list); + addr = unplugged; + unplug_index = UNPLUGGED_PER_PAGE; + } + + free_page((unsigned long) addr); + unplugged_pages_count--; + } + else { + struct page *page; + + page = alloc_page(GFP_ATOMIC); + if (page == NULL) + break; + + unplugged = page_address(page); + if (unplug_index == UNPLUGGED_PER_PAGE) { + list_add(&unplugged->list, &unplugged_pages); + unplug_index = 0; + } + else { + struct list_head *entry = unplugged_pages.next; + addr = unplugged; + + unplugged = list_entry(entry, + struct unplugged_pages, + list); + err = os_drop_memory(addr, PAGE_SIZE); + if (err) { + printk(KERN_ERR "Failed to release " + "memory - errno = %d\n", err); + *error_out = "Failed to release memory"; + goto out_unlock; + } + unplugged->pages[unplug_index++] = addr; + } + + unplugged_pages_count++; + } + } + + err = 0; +out_unlock: + mutex_unlock(&plug_mem_mutex); +out: + return err; +} + +static int mem_get_config(char *name, char *str, int size, char **error_out) +{ + char buf[sizeof("18446744073709551615")]; + int len = 0; + + sprintf(buf, "%ld", uml_physmem); + CONFIG_CHUNK(str, size, len, buf, 1); + + return len; +} + +static int mem_id(char **str, int *start_out, int *end_out) +{ + *start_out = 0; + *end_out = 0; + + return 0; +} + +static int mem_remove(int n, char **error_out) +{ + *error_out = "Memory doesn't support the remove operation"; + return -EBUSY; +} + +static struct mc_device mem_mc = { + .list = LIST_HEAD_INIT(mem_mc.list), + .name = "mem", + .config = mem_config, + .get_config = mem_get_config, + .id = mem_id, + .remove = mem_remove, +}; + +static int __init mem_mc_init(void) +{ + if (can_drop_memory()) + mconsole_register_dev(&mem_mc); + else printk(KERN_ERR "Can't release memory to the host - memory " + "hotplug won't be supported\n"); + return 0; +} + +__initcall(mem_mc_init); + +#define CONFIG_BUF_SIZE 64 + +static void mconsole_get_config(int (*get_config)(char *, char *, int, + char **), + struct mc_request *req, char *name) +{ + char default_buf[CONFIG_BUF_SIZE], *error, *buf; + int n, size; + + if (get_config == NULL) { + mconsole_reply(req, "No get_config routine defined", 1, 0); + return; + } + + error = NULL; + size = ARRAY_SIZE(default_buf); + buf = default_buf; + + while (1) { + n = (*get_config)(name, buf, size, &error); + if (error != NULL) { + mconsole_reply(req, error, 1, 0); + goto out; + } + + if (n <= size) { + mconsole_reply(req, buf, 0, 0); + goto out; + } + + if (buf != default_buf) + kfree(buf); + + size = n; + buf = kmalloc(size, GFP_KERNEL); + if (buf == NULL) { + mconsole_reply(req, "Failed to allocate buffer", 1, 0); + return; + } + } + out: + if (buf != default_buf) + kfree(buf); +} + +void mconsole_config(struct mc_request *req) +{ + struct mc_device *dev; + char *ptr = req->request.data, *name, *error_string = ""; + int err; + + ptr += strlen("config"); + ptr = skip_spaces(ptr); + dev = mconsole_find_dev(ptr); + if (dev == NULL) { + mconsole_reply(req, "Bad configuration option", 1, 0); + return; + } + + name = &ptr[strlen(dev->name)]; + ptr = name; + while ((*ptr != '=') && (*ptr != '\0')) + ptr++; + + if (*ptr == '=') { + err = (*dev->config)(name, &error_string); + mconsole_reply(req, error_string, err, 0); + } + else mconsole_get_config(dev->get_config, req, name); +} + +void mconsole_remove(struct mc_request *req) +{ + struct mc_device *dev; + char *ptr = req->request.data, *err_msg = ""; + char error[256]; + int err, start, end, n; + + ptr += strlen("remove"); + ptr = skip_spaces(ptr); + dev = mconsole_find_dev(ptr); + if (dev == NULL) { + mconsole_reply(req, "Bad remove option", 1, 0); + return; + } + + ptr = &ptr[strlen(dev->name)]; + + err = 1; + n = (*dev->id)(&ptr, &start, &end); + if (n < 0) { + err_msg = "Couldn't parse device number"; + goto out; + } + else if ((n < start) || (n > end)) { + sprintf(error, "Invalid device number - must be between " + "%d and %d", start, end); + err_msg = error; + goto out; + } + + err_msg = NULL; + err = (*dev->remove)(n, &err_msg); + switch(err) { + case 0: + err_msg = ""; + break; + case -ENODEV: + if (err_msg == NULL) + err_msg = "Device doesn't exist"; + break; + case -EBUSY: + if (err_msg == NULL) + err_msg = "Device is currently open"; + break; + default: + break; + } +out: + mconsole_reply(req, err_msg, err, 0); +} + +struct mconsole_output { + struct list_head list; + struct mc_request *req; +}; + +static DEFINE_SPINLOCK(client_lock); +static LIST_HEAD(clients); +static char console_buf[MCONSOLE_MAX_DATA] __nonstring; + +static void console_write(struct console *console, const char *string, + unsigned int len) +{ + struct list_head *ele; + int n; + + if (list_empty(&clients)) + return; + + while (len > 0) { + n = min((size_t) len, ARRAY_SIZE(console_buf)); + memcpy(console_buf, string, n); + string += n; + len -= n; + + list_for_each(ele, &clients) { + struct mconsole_output *entry; + + entry = list_entry(ele, struct mconsole_output, list); + mconsole_reply_len(entry->req, console_buf, n, 0, 1); + } + } +} + +static struct console mc_console = { .name = "mc", + .write = console_write, + .flags = CON_ENABLED, + .index = -1 }; + +static int mc_add_console(void) +{ + register_console(&mc_console); + return 0; +} + +late_initcall(mc_add_console); + +static void with_console(struct mc_request *req, void (*proc)(void *), + void *arg) +{ + struct mconsole_output entry; + unsigned long flags; + + entry.req = req; + spin_lock_irqsave(&client_lock, flags); + list_add(&entry.list, &clients); + spin_unlock_irqrestore(&client_lock, flags); + + (*proc)(arg); + + mconsole_reply_len(req, "", 0, 0, 0); + + spin_lock_irqsave(&client_lock, flags); + list_del(&entry.list); + spin_unlock_irqrestore(&client_lock, flags); +} + +#ifdef CONFIG_MAGIC_SYSRQ + +#include + +static void sysrq_proc(void *arg) +{ + char *op = arg; + handle_sysrq(*op); +} + +void mconsole_sysrq(struct mc_request *req) +{ + char *ptr = req->request.data; + + ptr += strlen("sysrq"); + ptr = skip_spaces(ptr); + + /* + * With 'b', the system will shut down without a chance to reply, + * so in this case, we reply first. + */ + if (*ptr == 'b') + mconsole_reply(req, "", 0, 0); + + with_console(req, sysrq_proc, ptr); +} +#else +void mconsole_sysrq(struct mc_request *req) +{ + mconsole_reply(req, "Sysrq not compiled in", 1, 0); +} +#endif + +static void stack_proc(void *arg) +{ + struct task_struct *task = arg; + + show_stack(task, NULL, KERN_INFO); +} + +/* + * Mconsole stack trace + * Added by Allan Graves, Jeff Dike + * Dumps a stacks registers to the linux console. + * Usage stack . + */ +void mconsole_stack(struct mc_request *req) +{ + char *ptr = req->request.data; + int pid_requested= -1; + struct task_struct *to = NULL; + + /* + * Would be nice: + * 1) Send showregs output to mconsole. + * 2) Add a way to stack dump all pids. + */ + + ptr += strlen("stack"); + ptr = skip_spaces(ptr); + + /* + * Should really check for multiple pids or reject bad args here + */ + /* What do the arguments in mconsole_reply mean? */ + if (sscanf(ptr, "%d", &pid_requested) == 0) { + mconsole_reply(req, "Please specify a pid", 1, 0); + return; + } + + to = find_task_by_pid_ns(pid_requested, &init_pid_ns); + if ((to == NULL) || (pid_requested == 0)) { + mconsole_reply(req, "Couldn't find that pid", 1, 0); + return; + } + with_console(req, stack_proc, to); +} + +static int __init mount_proc(void) +{ + struct file_system_type *proc_fs_type; + struct vfsmount *mnt; + + proc_fs_type = get_fs_type("proc"); + if (!proc_fs_type) + return -ENODEV; + + mnt = kern_mount(proc_fs_type); + put_filesystem(proc_fs_type); + if (IS_ERR(mnt)) + return PTR_ERR(mnt); + + proc_mnt = mnt; + return 0; +} + +/* + * Changed by mconsole_setup, which is __setup, and called before SMP is + * active. + */ +static char *notify_socket = NULL; + +static int __init mconsole_init(void) +{ + /* long to avoid size mismatch warnings from gcc */ + long sock; + int err; + char file[UNIX_PATH_MAX]; + + mount_proc(); + + if (umid_file_name("mconsole", file, sizeof(file))) + return -1; + snprintf(mconsole_socket_name, sizeof(file), "%s", file); + + sock = os_create_unix_socket(file, sizeof(file), 1); + if (sock < 0) { + printk(KERN_ERR "Failed to initialize management console\n"); + return 1; + } + if (os_set_fd_block(sock, 0)) + goto out; + + register_reboot_notifier(&reboot_notifier); + + err = um_request_irq(MCONSOLE_IRQ, sock, IRQ_READ, mconsole_interrupt, + IRQF_SHARED, "mconsole", (void *)sock); + if (err < 0) { + printk(KERN_ERR "Failed to get IRQ for management console\n"); + goto out; + } + + if (notify_socket != NULL) { + notify_socket = kstrdup(notify_socket, GFP_KERNEL); + if (notify_socket != NULL) + mconsole_notify(notify_socket, MCONSOLE_SOCKET, + mconsole_socket_name, + strlen(mconsole_socket_name) + 1); + else printk(KERN_ERR "mconsole_setup failed to strdup " + "string\n"); + } + + printk(KERN_INFO "mconsole (version %d) initialized on %s\n", + MCONSOLE_VERSION, mconsole_socket_name); + return 0; + + out: + os_close_file(sock); + return 1; +} + +__initcall(mconsole_init); + +static ssize_t mconsole_proc_write(struct file *file, + const char __user *buffer, size_t count, loff_t *pos) +{ + char *buf; + + buf = memdup_user_nul(buffer, count); + if (IS_ERR(buf)) + return PTR_ERR(buf); + + mconsole_notify(notify_socket, MCONSOLE_USER_NOTIFY, buf, count); + kfree(buf); + return count; +} + +static const struct proc_ops mconsole_proc_ops = { + .proc_write = mconsole_proc_write, + .proc_lseek = noop_llseek, +}; + +static int create_proc_mconsole(void) +{ + struct proc_dir_entry *ent; + + if (notify_socket == NULL) + return 0; + + ent = proc_create("mconsole", 0200, NULL, &mconsole_proc_ops); + if (ent == NULL) { + printk(KERN_INFO "create_proc_mconsole : proc_create failed\n"); + return 0; + } + return 0; +} + +static DEFINE_SPINLOCK(notify_spinlock); + +void lock_notify(void) +{ + spin_lock(¬ify_spinlock); +} + +void unlock_notify(void) +{ + spin_unlock(¬ify_spinlock); +} + +__initcall(create_proc_mconsole); + +#define NOTIFY "notify:" + +static int mconsole_setup(char *str) +{ + if (!strncmp(str, NOTIFY, strlen(NOTIFY))) { + str += strlen(NOTIFY); + notify_socket = str; + } + else printk(KERN_ERR "mconsole_setup : Unknown option - '%s'\n", str); + return 1; +} + +__setup("mconsole=", mconsole_setup); + +__uml_help(mconsole_setup, +"mconsole=notify:\n" +" Requests that the mconsole driver send a message to the named Unix\n" +" socket containing the name of the mconsole socket. This also serves\n" +" to notify outside processes when UML has booted far enough to respond\n" +" to mconsole requests.\n\n" +); + +static int notify_panic(struct notifier_block *self, unsigned long unused1, + void *ptr) +{ + char *message = ptr; + + if (notify_socket == NULL) + return 0; + + mconsole_notify(notify_socket, MCONSOLE_PANIC, message, + strlen(message) + 1); + return NOTIFY_DONE; +} + +static struct notifier_block panic_exit_notifier = { + .notifier_call = notify_panic, + .priority = INT_MAX, /* run as soon as possible */ +}; + +static int add_notifier(void) +{ + atomic_notifier_chain_register(&panic_notifier_list, + &panic_exit_notifier); + return 0; +} + +__initcall(add_notifier); + +char *mconsole_notify_socket(void) +{ + return notify_socket; +} + +EXPORT_SYMBOL(mconsole_notify_socket); diff --git a/arch/um/drivers/mconsole_kern.h b/arch/um/drivers/mconsole_kern.h new file mode 100644 index 000000000..56d8d6a3f --- /dev/null +++ b/arch/um/drivers/mconsole_kern.h @@ -0,0 +1,52 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2001, 2002 Jeff Dike (jdike@karaya.com) + */ + +#ifndef __MCONSOLE_KERN_H__ +#define __MCONSOLE_KERN_H__ + +#include +#include "mconsole.h" + +struct mconsole_entry { + struct list_head list; + struct mc_request request; +}; + +/* All these methods are called in process context. */ +struct mc_device { + struct list_head list; + char *name; + int (*config)(char *, char **); + int (*get_config)(char *, char *, int, char **); + int (*id)(char **, int *, int *); + int (*remove)(int, char **); +}; + +#define CONFIG_CHUNK(str, size, current, chunk, end) \ +do { \ + current += strlen(chunk); \ + if(current >= size) \ + str = NULL; \ + if(str != NULL){ \ + strcpy(str, chunk); \ + str += strlen(chunk); \ + } \ + if(end) \ + current++; \ +} while(0) + +#ifdef CONFIG_MCONSOLE + +extern void mconsole_register_dev(struct mc_device *new); + +#else + +static inline void mconsole_register_dev(struct mc_device *new) +{ +} + +#endif + +#endif diff --git a/arch/um/drivers/mconsole_user.c b/arch/um/drivers/mconsole_user.c new file mode 100644 index 000000000..a04cd13c6 --- /dev/null +++ b/arch/um/drivers/mconsole_user.c @@ -0,0 +1,221 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 Lennert Buytenhek (buytenh@gnu.org) + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + */ + +#include +#include +#include +#include +#include +#include +#include "mconsole.h" + +static struct mconsole_command commands[] = { + /* + * With uts namespaces, uts information becomes process-specific, so + * we need a process context. If we try handling this in interrupt + * context, we may hit an exiting process without a valid uts + * namespace. + */ + { "version", mconsole_version, MCONSOLE_PROC }, + { "halt", mconsole_halt, MCONSOLE_PROC }, + { "reboot", mconsole_reboot, MCONSOLE_PROC }, + { "config", mconsole_config, MCONSOLE_PROC }, + { "remove", mconsole_remove, MCONSOLE_PROC }, + { "sysrq", mconsole_sysrq, MCONSOLE_INTR }, + { "help", mconsole_help, MCONSOLE_INTR }, + { "cad", mconsole_cad, MCONSOLE_INTR }, + { "stop", mconsole_stop, MCONSOLE_PROC }, + { "go", mconsole_go, MCONSOLE_INTR }, + { "log", mconsole_log, MCONSOLE_INTR }, + { "proc", mconsole_proc, MCONSOLE_PROC }, + { "stack", mconsole_stack, MCONSOLE_INTR }, +}; + +/* Initialized in mconsole_init, which is an initcall */ +char mconsole_socket_name[256]; + +static int mconsole_reply_v0(struct mc_request *req, char *reply) +{ + struct iovec iov; + struct msghdr msg; + + iov.iov_base = reply; + iov.iov_len = strlen(reply); + + msg.msg_name = &(req->origin); + msg.msg_namelen = req->originlen; + msg.msg_iov = &iov; + msg.msg_iovlen = 1; + msg.msg_control = NULL; + msg.msg_controllen = 0; + msg.msg_flags = 0; + + return sendmsg(req->originating_fd, &msg, 0); +} + +static struct mconsole_command *mconsole_parse(struct mc_request *req) +{ + struct mconsole_command *cmd; + int i; + + for (i = 0; i < ARRAY_SIZE(commands); i++) { + cmd = &commands[i]; + if (!strncmp(req->request.data, cmd->command, + strlen(cmd->command))) { + return cmd; + } + } + return NULL; +} + +#ifndef MIN +#define MIN(a,b) ((a)<(b) ? (a):(b)) +#endif + +#define STRINGX(x) #x +#define STRING(x) STRINGX(x) + +int mconsole_get_request(int fd, struct mc_request *req) +{ + int len; + + req->originlen = sizeof(req->origin); + req->len = recvfrom(fd, &req->request, sizeof(req->request), 0, + (struct sockaddr *) req->origin, &req->originlen); + if (req->len < 0) + return 0; + + req->originating_fd = fd; + + if (req->request.magic != MCONSOLE_MAGIC) { + /* Unversioned request */ + len = MIN(sizeof(req->request.data) - 1, + strlen((char *) &req->request)); + memmove(req->request.data, &req->request, len); + req->request.data[len] = '\0'; + + req->request.magic = MCONSOLE_MAGIC; + req->request.version = 0; + req->request.len = len; + + mconsole_reply_v0(req, "ERR Version 0 mconsole clients are " + "not supported by this driver"); + return 0; + } + + if (req->request.len >= MCONSOLE_MAX_DATA) { + mconsole_reply(req, "Request too large", 1, 0); + return 0; + } + if (req->request.version != MCONSOLE_VERSION) { + mconsole_reply(req, "This driver only supports version " + STRING(MCONSOLE_VERSION) " clients", 1, 0); + } + + req->request.data[req->request.len] = '\0'; + req->cmd = mconsole_parse(req); + if (req->cmd == NULL) { + mconsole_reply(req, "Unknown command", 1, 0); + return 0; + } + + return 1; +} + +int mconsole_reply_len(struct mc_request *req, const char *str, int total, + int err, int more) +{ + /* + * XXX This is a stack consumption problem. It'd be nice to + * make it global and serialize access to it, but there are a + * ton of callers to this function. + */ + struct mconsole_reply reply; + int len, n; + + do { + reply.err = err; + + /* err can only be true on the first packet */ + err = 0; + + len = MIN(total, MCONSOLE_MAX_DATA - 1); + + if (len == total) reply.more = more; + else reply.more = 1; + + memcpy(reply.data, str, len); + reply.data[len] = '\0'; + total -= len; + str += len; + reply.len = len + 1; + + len = sizeof(reply) + reply.len - sizeof(reply.data); + + n = sendto(req->originating_fd, &reply, len, 0, + (struct sockaddr *) req->origin, req->originlen); + + if (n < 0) + return -errno; + } while (total > 0); + return 0; +} + +int mconsole_reply(struct mc_request *req, const char *str, int err, int more) +{ + return mconsole_reply_len(req, str, strlen(str), err, more); +} + + +int mconsole_unlink_socket(void) +{ + unlink(mconsole_socket_name); + return 0; +} + +static int notify_sock = -1; + +int mconsole_notify(char *sock_name, int type, const void *data, int len) +{ + struct sockaddr_un target; + struct mconsole_notify packet; + int n, err = 0; + + lock_notify(); + if (notify_sock < 0) { + notify_sock = socket(PF_UNIX, SOCK_DGRAM, 0); + if (notify_sock < 0) { + err = -errno; + printk(UM_KERN_ERR "mconsole_notify - socket failed, " + "errno = %d\n", errno); + } + } + unlock_notify(); + + if (err) + return err; + + target.sun_family = AF_UNIX; + strcpy(target.sun_path, sock_name); + + packet.magic = MCONSOLE_MAGIC; + packet.version = MCONSOLE_VERSION; + packet.type = type; + len = (len > sizeof(packet.data)) ? sizeof(packet.data) : len; + packet.len = len; + memcpy(packet.data, data, len); + + err = 0; + len = sizeof(packet) + packet.len - sizeof(packet.data); + n = sendto(notify_sock, &packet, len, 0, (struct sockaddr *) &target, + sizeof(target)); + if (n < 0) { + err = -errno; + printk(UM_KERN_ERR "mconsole_notify - sendto failed, " + "errno = %d\n", errno); + } + return err; +} diff --git a/arch/um/drivers/null.c b/arch/um/drivers/null.c new file mode 100644 index 000000000..30d59b848 --- /dev/null +++ b/arch/um/drivers/null.c @@ -0,0 +1,51 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2002 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include "chan_user.h" +#include + +/* This address is used only as a unique identifier */ +static int null_chan; + +static void *null_init(char *str, int device, const struct chan_opts *opts) +{ + return &null_chan; +} + +static int null_open(int input, int output, int primary, void *d, + char **dev_out) +{ + int fd; + + *dev_out = NULL; + + fd = open(DEV_NULL, O_RDWR); + return (fd < 0) ? -errno : fd; +} + +static int null_read(int fd, __u8 *c_out, void *unused) +{ + return -ENODEV; +} + +static void null_free(void *data) +{ +} + +const struct chan_ops null_ops = { + .type = "null", + .init = null_init, + .open = null_open, + .close = generic_close, + .read = null_read, + .write = generic_write, + .console_write = generic_console_write, + .window_size = generic_window_size, + .free = null_free, + .winch = 0, +}; diff --git a/arch/um/drivers/port.h b/arch/um/drivers/port.h new file mode 100644 index 000000000..9085b336e --- /dev/null +++ b/arch/um/drivers/port.h @@ -0,0 +1,20 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2001 Jeff Dike (jdike@karaya.com) + */ + +#ifndef __PORT_H__ +#define __PORT_H__ + +extern void *port_data(int port); +extern int port_wait(void *data); +extern void port_kern_close(void *d); +extern int port_connection(int fd, int *socket_out, int *pid_out); +extern int port_listen_fd(int port); +extern void port_read(int fd, void *data); +extern void port_kern_free(void *d); +extern int port_rcv_fd(int fd); +extern void port_remove_dev(void *d); + +#endif + diff --git a/arch/um/drivers/port_kern.c b/arch/um/drivers/port_kern.c new file mode 100644 index 000000000..7b3ff4af1 --- /dev/null +++ b/arch/um/drivers/port_kern.c @@ -0,0 +1,305 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "port.h" + +struct port_list { + struct list_head list; + atomic_t wait_count; + int has_connection; + struct completion done; + int port; + int fd; + spinlock_t lock; + struct list_head pending; + struct list_head connections; +}; + +struct port_dev { + struct port_list *port; + int helper_pid; + int telnetd_pid; +}; + +struct connection { + struct list_head list; + int fd; + int helper_pid; + int socket[2]; + int telnetd_pid; + struct port_list *port; +}; + +static irqreturn_t pipe_interrupt(int irq, void *data) +{ + struct connection *conn = data; + int n_fds = 1, fd = -1; + ssize_t ret; + + ret = os_rcv_fd_msg(conn->socket[0], &fd, n_fds, &conn->helper_pid, + sizeof(conn->helper_pid)); + if (ret != sizeof(conn->helper_pid)) { + if (ret == -EAGAIN) + return IRQ_NONE; + + printk(KERN_ERR "pipe_interrupt : os_rcv_fd_msg returned %zd\n", + ret); + os_close_file(conn->fd); + } + + list_del(&conn->list); + + conn->fd = fd; + list_add(&conn->list, &conn->port->connections); + + complete(&conn->port->done); + return IRQ_HANDLED; +} + +#define NO_WAITER_MSG \ + "****\n" \ + "There are currently no UML consoles waiting for port connections.\n" \ + "Either disconnect from one to make it available or activate some more\n" \ + "by enabling more consoles in the UML /etc/inittab.\n" \ + "****\n" + +static int port_accept(struct port_list *port) +{ + struct connection *conn; + int fd, socket[2], pid; + + fd = port_connection(port->fd, socket, &pid); + if (fd < 0) { + if (fd != -EAGAIN) + printk(KERN_ERR "port_accept : port_connection " + "returned %d\n", -fd); + goto out; + } + + conn = kmalloc_obj(*conn, GFP_ATOMIC); + if (conn == NULL) { + printk(KERN_ERR "port_accept : failed to allocate " + "connection\n"); + goto out_close; + } + *conn = ((struct connection) + { .list = LIST_HEAD_INIT(conn->list), + .fd = fd, + .socket = { socket[0], socket[1] }, + .telnetd_pid = pid, + .port = port }); + + if (um_request_irq(TELNETD_IRQ, socket[0], IRQ_READ, pipe_interrupt, + IRQF_SHARED, "telnetd", conn) < 0) { + printk(KERN_ERR "port_accept : failed to get IRQ for " + "telnetd\n"); + goto out_free; + } + + if (atomic_read(&port->wait_count) == 0) { + os_write_file(fd, NO_WAITER_MSG, sizeof(NO_WAITER_MSG)); + printk(KERN_ERR "No one waiting for port\n"); + } + list_add(&conn->list, &port->pending); + return 1; + + out_free: + kfree(conn); + out_close: + os_close_file(fd); + os_kill_process(pid, 1); + out: + return 0; +} + +static DEFINE_MUTEX(ports_mutex); +static LIST_HEAD(ports); + +static void port_work_proc(struct work_struct *unused) +{ + struct port_list *port; + struct list_head *ele; + unsigned long flags; + + local_irq_save(flags); + list_for_each(ele, &ports) { + port = list_entry(ele, struct port_list, list); + if (!port->has_connection) + continue; + + while (port_accept(port)) + ; + port->has_connection = 0; + } + local_irq_restore(flags); +} + +static DECLARE_WORK(port_work, port_work_proc); + +static irqreturn_t port_interrupt(int irq, void *data) +{ + struct port_list *port = data; + + port->has_connection = 1; + schedule_work(&port_work); + return IRQ_HANDLED; +} + +void *port_data(int port_num) +{ + struct list_head *ele; + struct port_list *port; + struct port_dev *dev = NULL; + int fd; + + mutex_lock(&ports_mutex); + list_for_each(ele, &ports) { + port = list_entry(ele, struct port_list, list); + if (port->port == port_num) + goto found; + } + port = kmalloc_obj(struct port_list); + if (port == NULL) { + printk(KERN_ERR "Allocation of port list failed\n"); + goto out; + } + + fd = port_listen_fd(port_num); + if (fd < 0) { + printk(KERN_ERR "binding to port %d failed, errno = %d\n", + port_num, -fd); + goto out_free; + } + + if (um_request_irq(ACCEPT_IRQ, fd, IRQ_READ, port_interrupt, + IRQF_SHARED, "port", port) < 0) { + printk(KERN_ERR "Failed to get IRQ for port %d\n", port_num); + goto out_close; + } + + *port = ((struct port_list) + { .list = LIST_HEAD_INIT(port->list), + .wait_count = ATOMIC_INIT(0), + .has_connection = 0, + .port = port_num, + .fd = fd, + .pending = LIST_HEAD_INIT(port->pending), + .connections = LIST_HEAD_INIT(port->connections) }); + spin_lock_init(&port->lock); + init_completion(&port->done); + list_add(&port->list, &ports); + + found: + dev = kmalloc_obj(struct port_dev); + if (dev == NULL) { + printk(KERN_ERR "Allocation of port device entry failed\n"); + goto out; + } + + *dev = ((struct port_dev) { .port = port, + .helper_pid = -1, + .telnetd_pid = -1 }); + goto out; + + out_close: + os_close_file(fd); + out_free: + kfree(port); + out: + mutex_unlock(&ports_mutex); + return dev; +} + +int port_wait(void *data) +{ + struct port_dev *dev = data; + struct connection *conn; + struct port_list *port = dev->port; + int fd; + + atomic_inc(&port->wait_count); + while (1) { + fd = -ERESTARTSYS; + if (wait_for_completion_interruptible(&port->done)) + goto out; + + spin_lock(&port->lock); + + conn = list_entry(port->connections.next, struct connection, + list); + list_del(&conn->list); + spin_unlock(&port->lock); + + os_shutdown_socket(conn->socket[0], 1, 1); + os_close_file(conn->socket[0]); + os_shutdown_socket(conn->socket[1], 1, 1); + os_close_file(conn->socket[1]); + + /* This is done here because freeing an IRQ can't be done + * within the IRQ handler. So, pipe_interrupt always ups + * the semaphore regardless of whether it got a successful + * connection. Then we loop here throwing out failed + * connections until a good one is found. + */ + um_free_irq(TELNETD_IRQ, conn); + + if (conn->fd >= 0) + break; + os_close_file(conn->fd); + kfree(conn); + } + + fd = conn->fd; + dev->helper_pid = conn->helper_pid; + dev->telnetd_pid = conn->telnetd_pid; + kfree(conn); + out: + atomic_dec(&port->wait_count); + return fd; +} + +void port_remove_dev(void *d) +{ + struct port_dev *dev = d; + + if (dev->helper_pid != -1) + os_kill_process(dev->helper_pid, 0); + if (dev->telnetd_pid != -1) + os_kill_process(dev->telnetd_pid, 1); + dev->helper_pid = -1; + dev->telnetd_pid = -1; +} + +void port_kern_free(void *d) +{ + struct port_dev *dev = d; + + port_remove_dev(dev); + kfree(dev); +} + +static void free_port(void) +{ + struct list_head *ele; + struct port_list *port; + + list_for_each(ele, &ports) { + port = list_entry(ele, struct port_list, list); + free_irq_by_fd(port->fd); + os_close_file(port->fd); + } +} + +__uml_exitcall(free_port); diff --git a/arch/um/drivers/port_user.c b/arch/um/drivers/port_user.c new file mode 100644 index 000000000..3c62ae81d --- /dev/null +++ b/arch/um/drivers/port_user.c @@ -0,0 +1,217 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include "chan_user.h" +#include +#include "port.h" +#include + +struct port_chan { + int raw; + struct termios tt; + void *kernel_data; + char dev[sizeof("32768\0")]; +}; + +static void *port_init(char *str, int device, const struct chan_opts *opts) +{ + struct port_chan *data; + void *kern_data; + char *end; + int port; + + if (*str != ':') { + printk(UM_KERN_ERR "port_init : channel type 'port' must " + "specify a port number\n"); + return NULL; + } + str++; + port = strtoul(str, &end, 0); + if ((*end != '\0') || (end == str)) { + printk(UM_KERN_ERR "port_init : couldn't parse port '%s'\n", + str); + return NULL; + } + + kern_data = port_data(port); + if (kern_data == NULL) + return NULL; + + data = uml_kmalloc(sizeof(*data), UM_GFP_KERNEL); + if (data == NULL) + goto err; + + *data = ((struct port_chan) { .raw = opts->raw, + .kernel_data = kern_data }); + sprintf(data->dev, "%d", port); + + return data; + err: + port_kern_free(kern_data); + return NULL; +} + +static void port_free(void *d) +{ + struct port_chan *data = d; + + port_kern_free(data->kernel_data); + kfree(data); +} + +static int port_open(int input, int output, int primary, void *d, + char **dev_out) +{ + struct port_chan *data = d; + int fd, err; + + fd = port_wait(data->kernel_data); + if ((fd >= 0) && data->raw) { + CATCH_EINTR(err = tcgetattr(fd, &data->tt)); + if (err) + return err; + + err = raw(fd); + if (err) + return err; + } + *dev_out = data->dev; + return fd; +} + +static void port_close(int fd, void *d) +{ + struct port_chan *data = d; + + port_remove_dev(data->kernel_data); + os_close_file(fd); +} + +const struct chan_ops port_ops = { + .type = "port", + .init = port_init, + .open = port_open, + .close = port_close, + .read = generic_read, + .write = generic_write, + .console_write = generic_console_write, + .window_size = generic_window_size, + .free = port_free, + .winch = 1, +}; + +int port_listen_fd(int port) +{ + struct sockaddr_in addr; + int fd, err, arg; + + fd = socket(PF_INET, SOCK_STREAM, 0); + if (fd == -1) + return -errno; + + arg = 1; + if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &arg, sizeof(arg)) < 0) { + err = -errno; + goto out; + } + + addr.sin_family = AF_INET; + addr.sin_port = htons(port); + addr.sin_addr.s_addr = htonl(INADDR_ANY); + if (bind(fd, (struct sockaddr *) &addr, sizeof(addr)) < 0) { + err = -errno; + goto out; + } + + if (listen(fd, 1) < 0) { + err = -errno; + goto out; + } + + err = os_set_fd_block(fd, 0); + if (err < 0) + goto out; + + return fd; + out: + close(fd); + return err; +} + +struct port_pre_exec_data { + int sock_fd; + int pipe_fd; +}; + +static void port_pre_exec(void *arg) +{ + struct port_pre_exec_data *data = arg; + + dup2(data->sock_fd, 0); + dup2(data->sock_fd, 1); + dup2(data->sock_fd, 2); + close(data->sock_fd); + dup2(data->pipe_fd, 3); + shutdown(3, SHUT_RD); + close(data->pipe_fd); +} + +int port_connection(int fd, int *socket, int *pid_out) +{ + int new, err; + char *env; + char *argv[] = { "in.telnetd", "-L", + OS_LIB_PATH "/uml/port-helper", NULL }; + struct port_pre_exec_data data; + + if ((env = getenv("UML_PORT_HELPER"))) + argv[2] = env; + + new = accept(fd, NULL, 0); + if (new < 0) + return -errno; + + err = os_access(argv[2], X_OK); + if (err < 0) { + printk(UM_KERN_ERR "port_connection : error accessing port-helper " + "executable at %s: %s\n", argv[2], strerror(-err)); + if (env == NULL) + printk(UM_KERN_ERR "Set UML_PORT_HELPER environment " + "variable to path to uml-utilities port-helper " + "binary\n"); + goto out_close; + } + + err = os_pipe(socket, 0, 0); + if (err < 0) + goto out_close; + + data = ((struct port_pre_exec_data) + { .sock_fd = new, + .pipe_fd = socket[1] }); + + err = run_helper(port_pre_exec, &data, argv); + if (err < 0) + goto out_shutdown; + + *pid_out = err; + return new; + + out_shutdown: + shutdown(socket[0], SHUT_RDWR); + close(socket[0]); + shutdown(socket[1], SHUT_RDWR); + close(socket[1]); + out_close: + close(new); + return err; +} diff --git a/arch/um/drivers/pty.c b/arch/um/drivers/pty.c new file mode 100644 index 000000000..39c60068c --- /dev/null +++ b/arch/um/drivers/pty.c @@ -0,0 +1,165 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include "chan_user.h" +#include +#include + +struct pty_chan { + void (*announce)(char *dev_name, int dev); + int dev; + int raw; + struct termios tt; + char dev_name[sizeof("/dev/pts/0123456\0")]; +}; + +static void *pty_chan_init(char *str, int device, const struct chan_opts *opts) +{ + struct pty_chan *data; + + data = uml_kmalloc(sizeof(*data), UM_GFP_KERNEL); + if (data == NULL) + return NULL; + + *data = ((struct pty_chan) { .announce = opts->announce, + .dev = device, + .raw = opts->raw }); + return data; +} + +static int pts_open(int input, int output, int primary, void *d, + char **dev_out) +{ + struct pty_chan *data = d; + char *dev; + int fd, err; + + fd = get_pty(); + if (fd < 0) { + err = -errno; + printk(UM_KERN_ERR "open_pts : Failed to open pts\n"); + return err; + } + + if (data->raw) { + CATCH_EINTR(err = tcgetattr(fd, &data->tt)); + if (err) + goto out_close; + + err = raw(fd); + if (err) + goto out_close; + } + + dev = ptsname(fd); + sprintf(data->dev_name, "%s", dev); + *dev_out = data->dev_name; + + if (data->announce) + (*data->announce)(dev, data->dev); + + return fd; + +out_close: + close(fd); + return err; +} + +static int getmaster(char *line) +{ + struct stat buf; + char *pty, *bank, *cp; + int master, err; + + pty = &line[strlen("/dev/ptyp")]; + for (bank = "pqrs"; *bank; bank++) { + line[strlen("/dev/pty")] = *bank; + *pty = '0'; + /* Did we hit the end ? */ + if ((stat(line, &buf) < 0) && (errno == ENOENT)) + break; + + for (cp = "0123456789abcdef"; *cp; cp++) { + *pty = *cp; + master = open(line, O_RDWR); + if (master >= 0) { + char *tp = &line[strlen("/dev/")]; + + /* verify slave side is usable */ + *tp = 't'; + err = access(line, R_OK | W_OK); + *tp = 'p'; + if (!err) + return master; + close(master); + } + } + } + + printk(UM_KERN_ERR "getmaster - no usable host pty devices\n"); + return -ENOENT; +} + +static int pty_open(int input, int output, int primary, void *d, + char **dev_out) +{ + struct pty_chan *data = d; + int fd, err; + char dev[sizeof("/dev/ptyxx\0")] = "/dev/ptyxx"; + + fd = getmaster(dev); + if (fd < 0) + return fd; + + if (data->raw) { + err = raw(fd); + if (err) { + close(fd); + return err; + } + } + + if (data->announce) + (*data->announce)(dev, data->dev); + + sprintf(data->dev_name, "%s", dev); + *dev_out = data->dev_name; + + return fd; +} + +const struct chan_ops pty_ops = { + .type = "pty", + .init = pty_chan_init, + .open = pty_open, + .close = generic_close, + .read = generic_read, + .write = generic_write, + .console_write = generic_console_write, + .window_size = generic_window_size, + .free = generic_free, + .winch = 0, +}; + +const struct chan_ops pts_ops = { + .type = "pts", + .init = pty_chan_init, + .open = pts_open, + .close = generic_close, + .read = generic_read, + .write = generic_write, + .console_write = generic_console_write, + .window_size = generic_window_size, + .free = generic_free, + .winch = 0, +}; diff --git a/arch/um/drivers/random.c b/arch/um/drivers/random.c new file mode 100644 index 000000000..ca08c91f4 --- /dev/null +++ b/arch/um/drivers/random.c @@ -0,0 +1,121 @@ +/* Copyright (C) 2005 - 2008 Jeff Dike */ + +/* Much of this ripped from drivers/char/hw_random.c, see there for other + * copyright. + * + * This software may be used and distributed according to the terms + * of the GNU General Public License, incorporated herein by reference. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* + * core module information + */ +#define RNG_MODULE_NAME "hw_random" + +/* Changed at init time, in the non-modular case, and at module load + * time, in the module case. Presumably, the module subsystem + * protects against a module being loaded twice at the same time. + */ +static int random_fd = -1; +static struct hwrng hwrng; +static DECLARE_COMPLETION(have_data); + +static int rng_dev_read(struct hwrng *rng, void *buf, size_t max, bool block) +{ + int ret; + + for (;;) { + ret = os_read_file(random_fd, buf, max); + if (block && ret == -EAGAIN) { + add_sigio_fd(random_fd); + + ret = wait_for_completion_killable(&have_data); + + ignore_sigio_fd(random_fd); + deactivate_fd(random_fd, RANDOM_IRQ); + + if (ret < 0) + break; + } else { + break; + } + } + + return ret != -EAGAIN ? ret : 0; +} + +static irqreturn_t random_interrupt(int irq, void *data) +{ + complete(&have_data); + + return IRQ_HANDLED; +} + +/* + * rng_init - initialize RNG module + */ +static int __init rng_init (void) +{ + int err; + + err = os_open_file("/dev/random", of_read(OPENFLAGS()), 0); + if (err < 0) + goto out; + + random_fd = err; + err = um_request_irq(RANDOM_IRQ, random_fd, IRQ_READ, random_interrupt, + 0, "random", NULL); + if (err < 0) + goto err_out_cleanup_hw; + + sigio_broken(); + hwrng.name = RNG_MODULE_NAME; + hwrng.read = rng_dev_read; + + err = hwrng_register(&hwrng); + if (err) { + pr_err(RNG_MODULE_NAME " registering failed (%d)\n", err); + goto err_out_cleanup_hw; + } +out: + return err; + +err_out_cleanup_hw: + os_close_file(random_fd); + random_fd = -1; + goto out; +} + +/* + * rng_cleanup - shutdown RNG module + */ + +static void cleanup(void) +{ + free_irq_by_fd(random_fd); + os_close_file(random_fd); +} + +static void __exit rng_cleanup(void) +{ + hwrng_unregister(&hwrng); + os_close_file(random_fd); +} + +module_init (rng_init); +module_exit (rng_cleanup); +__uml_exitcall(cleanup); + +MODULE_DESCRIPTION("UML Host Random Number Generator (RNG) driver"); +MODULE_LICENSE("GPL"); diff --git a/arch/um/drivers/rtc.h b/arch/um/drivers/rtc.h new file mode 100644 index 000000000..95e41c7d3 --- /dev/null +++ b/arch/um/drivers/rtc.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2020 Intel Corporation + * Author: Johannes Berg + */ +#ifndef __UM_RTC_H__ +#define __UM_RTC_H__ + +int uml_rtc_start(bool timetravel); +int uml_rtc_enable_alarm(unsigned long long delta_seconds); +void uml_rtc_disable_alarm(void); +void uml_rtc_stop(bool timetravel); +void uml_rtc_send_timetravel_alarm(void); + +#endif /* __UM_RTC_H__ */ diff --git a/arch/um/drivers/rtc_kern.c b/arch/um/drivers/rtc_kern.c new file mode 100644 index 000000000..9158c936c --- /dev/null +++ b/arch/um/drivers/rtc_kern.c @@ -0,0 +1,213 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2020 Intel Corporation + * Author: Johannes Berg + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include "rtc.h" + +static time64_t uml_rtc_alarm_time; +static bool uml_rtc_alarm_enabled; +static struct rtc_device *uml_rtc; +static int uml_rtc_irq_fd, uml_rtc_irq; + +#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT + +static void uml_rtc_time_travel_alarm(struct time_travel_event *ev) +{ + uml_rtc_send_timetravel_alarm(); +} + +static struct time_travel_event uml_rtc_alarm_event = { + .fn = uml_rtc_time_travel_alarm, +}; +#endif + +static int uml_rtc_read_time(struct device *dev, struct rtc_time *tm) +{ + struct timespec64 ts; + + /* Use this to get correct time in time-travel mode */ + read_persistent_clock64(&ts); + rtc_time64_to_tm(timespec64_to_ktime(ts) / NSEC_PER_SEC, tm); + + return 0; +} + +static int uml_rtc_read_alarm(struct device *dev, struct rtc_wkalrm *alrm) +{ + rtc_time64_to_tm(uml_rtc_alarm_time, &alrm->time); + alrm->enabled = uml_rtc_alarm_enabled; + + return 0; +} + +static int uml_rtc_alarm_irq_enable(struct device *dev, unsigned int enable) +{ + struct timespec64 ts; + unsigned long long secs; + + if (!enable && !uml_rtc_alarm_enabled) + return 0; + + uml_rtc_alarm_enabled = enable; + + read_persistent_clock64(&ts); + secs = uml_rtc_alarm_time - ts.tv_sec; + + if (time_travel_mode == TT_MODE_OFF) { + if (!enable) { + uml_rtc_disable_alarm(); + return 0; + } + + /* enable or update */ + return uml_rtc_enable_alarm(secs); + } else { + time_travel_del_event(¨_rtc_alarm_event); + + if (enable) + time_travel_add_event_rel(¨_rtc_alarm_event, + secs * NSEC_PER_SEC - + ts.tv_nsec); + } + + return 0; +} + +static int uml_rtc_set_alarm(struct device *dev, struct rtc_wkalrm *alrm) +{ + uml_rtc_alarm_irq_enable(dev, 0); + uml_rtc_alarm_time = rtc_tm_to_time64(&alrm->time); + uml_rtc_alarm_irq_enable(dev, alrm->enabled); + + return 0; +} + +static const struct rtc_class_ops uml_rtc_ops = { + .read_time = uml_rtc_read_time, + .read_alarm = uml_rtc_read_alarm, + .alarm_irq_enable = uml_rtc_alarm_irq_enable, + .set_alarm = uml_rtc_set_alarm, +}; + +static irqreturn_t uml_rtc_interrupt(int irq, void *data) +{ + unsigned long long c = 0; + + /* alarm triggered, it's now off */ + uml_rtc_alarm_enabled = false; + + os_read_file(uml_rtc_irq_fd, &c, sizeof(c)); + WARN_ON(c == 0); + + pm_system_wakeup(); + rtc_update_irq(uml_rtc, 1, RTC_IRQF | RTC_AF); + + return IRQ_HANDLED; +} + +static int uml_rtc_setup(void) +{ + int err; + + err = uml_rtc_start(time_travel_mode != TT_MODE_OFF); + if (WARN(err < 0, "err = %d\n", err)) + return err; + + uml_rtc_irq_fd = err; + + err = um_request_irq(UM_IRQ_ALLOC, uml_rtc_irq_fd, IRQ_READ, + uml_rtc_interrupt, 0, "rtc", NULL); + if (err < 0) { + uml_rtc_stop(time_travel_mode != TT_MODE_OFF); + return err; + } + + irq_set_irq_wake(err, 1); + + uml_rtc_irq = err; + return 0; +} + +static void uml_rtc_cleanup(void) +{ + um_free_irq(uml_rtc_irq, NULL); + uml_rtc_stop(time_travel_mode != TT_MODE_OFF); +} + +static int uml_rtc_probe(struct platform_device *pdev) +{ + int err; + + err = uml_rtc_setup(); + if (err) + return err; + + uml_rtc = devm_rtc_allocate_device(&pdev->dev); + if (IS_ERR(uml_rtc)) { + err = PTR_ERR(uml_rtc); + goto cleanup; + } + + uml_rtc->ops = ¨_rtc_ops; + + device_init_wakeup(&pdev->dev, 1); + + err = devm_rtc_register_device(uml_rtc); + if (err) + goto cleanup; + + return 0; +cleanup: + uml_rtc_cleanup(); + return err; +} + +static void uml_rtc_remove(struct platform_device *pdev) +{ + device_init_wakeup(&pdev->dev, 0); + uml_rtc_cleanup(); +} + +static struct platform_driver uml_rtc_driver = { + .probe = uml_rtc_probe, + .remove = uml_rtc_remove, + .driver = { + .name = "uml-rtc", + }, +}; + +static int __init uml_rtc_init(void) +{ + struct platform_device *pdev; + int err; + + err = platform_driver_register(¨_rtc_driver); + if (err) + return err; + + pdev = platform_device_alloc("uml-rtc", 0); + if (!pdev) { + err = -ENOMEM; + goto unregister; + } + + err = platform_device_add(pdev); + if (err) + goto unregister; + return 0; + +unregister: + platform_device_put(pdev); + platform_driver_unregister(¨_rtc_driver); + return err; +} +device_initcall(uml_rtc_init); diff --git a/arch/um/drivers/rtc_user.c b/arch/um/drivers/rtc_user.c new file mode 100644 index 000000000..67912fcf7 --- /dev/null +++ b/arch/um/drivers/rtc_user.c @@ -0,0 +1,81 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2020 Intel Corporation + * Author: Johannes Berg + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "rtc.h" + +static int uml_rtc_irq_fds[2]; + +void uml_rtc_send_timetravel_alarm(void) +{ + unsigned long long c = 1; + + CATCH_EINTR(write(uml_rtc_irq_fds[1], &c, sizeof(c))); +} + +int uml_rtc_start(bool timetravel) +{ + int err; + + if (timetravel) { + err = os_pipe(uml_rtc_irq_fds, 1, 1); + if (err) + goto fail; + } else { + uml_rtc_irq_fds[0] = timerfd_create(CLOCK_REALTIME, TFD_CLOEXEC); + if (uml_rtc_irq_fds[0] < 0) { + err = -errno; + goto fail; + } + + /* apparently timerfd won't send SIGIO, use workaround */ + sigio_broken(); + err = add_sigio_fd(uml_rtc_irq_fds[0]); + if (err < 0) { + close(uml_rtc_irq_fds[0]); + goto fail; + } + } + + return uml_rtc_irq_fds[0]; +fail: + uml_rtc_stop(timetravel); + return err; +} + +int uml_rtc_enable_alarm(unsigned long long delta_seconds) +{ + struct itimerspec it = { + .it_value = { + .tv_sec = delta_seconds, + }, + }; + + if (timerfd_settime(uml_rtc_irq_fds[0], 0, &it, NULL)) + return -errno; + return 0; +} + +void uml_rtc_disable_alarm(void) +{ + uml_rtc_enable_alarm(0); +} + +void uml_rtc_stop(bool timetravel) +{ + if (timetravel) + os_close_file(uml_rtc_irq_fds[1]); + else + ignore_sigio_fd(uml_rtc_irq_fds[0]); + os_close_file(uml_rtc_irq_fds[0]); +} diff --git a/arch/um/drivers/ssl.c b/arch/um/drivers/ssl.c new file mode 100644 index 000000000..8006a5bd5 --- /dev/null +++ b/arch/um/drivers/ssl.c @@ -0,0 +1,205 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2000, 2002 Jeff Dike (jdike@karaya.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "chan.h" +#include +#include +#include "mconsole_kern.h" + +static const int ssl_version = 1; + +#define NR_PORTS 64 + +static void ssl_announce(char *dev_name, int dev) +{ + printk(KERN_INFO "Serial line %d assigned device '%s'\n", dev, + dev_name); +} + +/* Almost const, except that xterm_title may be changed in an initcall */ +static struct chan_opts opts = { + .announce = ssl_announce, + .xterm_title = "Serial Line #%d", + .raw = 1, +}; + +static int ssl_config(char *str, char **error_out); +static int ssl_get_config(char *dev, char *str, int size, char **error_out); +static int ssl_remove(int n, char **error_out); + + +/* Const, except for .mc.list */ +static struct line_driver driver = { + .name = "UML serial line", + .device_name = "ttyS", + .major = TTY_MAJOR, + .minor_start = 64, + .type = TTY_DRIVER_TYPE_SERIAL, + .subtype = 0, + .read_irq_name = "ssl", + .write_irq_name = "ssl-write", + .mc = { + .list = LIST_HEAD_INIT(driver.mc.list), + .name = "ssl", + .config = ssl_config, + .get_config = ssl_get_config, + .id = line_id, + .remove = ssl_remove, + }, +}; + +/* The array is initialized by line_init, at initcall time. The + * elements are locked individually as needed. + */ +static char *conf[NR_PORTS]; +static char *def_conf = CONFIG_SSL_CHAN; +static struct line serial_lines[NR_PORTS]; + +static int ssl_config(char *str, char **error_out) +{ + return line_config(serial_lines, ARRAY_SIZE(serial_lines), str, &opts, + error_out); +} + +static int ssl_get_config(char *dev, char *str, int size, char **error_out) +{ + return line_get_config(dev, serial_lines, ARRAY_SIZE(serial_lines), str, + size, error_out); +} + +static int ssl_remove(int n, char **error_out) +{ + return line_remove(serial_lines, ARRAY_SIZE(serial_lines), n, + error_out); +} + +static int ssl_install(struct tty_driver *driver, struct tty_struct *tty) +{ + return line_install(driver, tty, &serial_lines[tty->index]); +} + +static const struct tty_operations ssl_ops = { + .open = line_open, + .close = line_close, + .write = line_write, + .write_room = line_write_room, + .chars_in_buffer = line_chars_in_buffer, + .flush_buffer = line_flush_buffer, + .flush_chars = line_flush_chars, + .throttle = line_throttle, + .unthrottle = line_unthrottle, + .install = ssl_install, + .hangup = line_hangup, +}; + +/* Changed by ssl_init and referenced by ssl_exit, which are both serialized + * by being an initcall and exitcall, respectively. + */ +static int ssl_init_done; + +static void ssl_console_write(struct console *c, const char *string, + unsigned len) +{ + struct line *line = &serial_lines[c->index]; + unsigned long flags; + + spin_lock_irqsave(&line->lock, flags); + console_write_chan(line->chan_out, string, len); + spin_unlock_irqrestore(&line->lock, flags); +} + +static struct tty_driver *ssl_console_device(struct console *c, int *index) +{ + *index = c->index; + return driver.driver; +} + +static int ssl_console_setup(struct console *co, char *options) +{ + struct line *line = &serial_lines[co->index]; + + return console_open_chan(line, co); +} + +/* No locking for register_console call - relies on single-threaded initcalls */ +static struct console ssl_cons = { + .name = "ttyS", + .write = ssl_console_write, + .device = ssl_console_device, + .setup = ssl_console_setup, + .flags = CON_PRINTBUFFER|CON_ANYTIME, + .index = -1, +}; + +static int ssl_init(void) +{ + char *new_title; + int err; + int i; + + printk(KERN_INFO "Initializing software serial port version %d\n", + ssl_version); + + err = register_lines(&driver, &ssl_ops, serial_lines, + ARRAY_SIZE(serial_lines)); + if (err) + return err; + + new_title = add_xterm_umid(opts.xterm_title); + if (new_title != NULL) + opts.xterm_title = new_title; + + for (i = 0; i < NR_PORTS; i++) { + char *error; + char *s = conf[i]; + if (!s) + s = def_conf; + if (setup_one_line(serial_lines, i, s, &opts, &error)) + printk(KERN_ERR "setup_one_line failed for " + "device %d : %s\n", i, error); + } + + ssl_init_done = 1; + register_console(&ssl_cons); + return 0; +} +late_initcall(ssl_init); + +static void ssl_exit(void) +{ + if (!ssl_init_done) + return; + close_lines(serial_lines, ARRAY_SIZE(serial_lines)); +} +__uml_exitcall(ssl_exit); + +static int ssl_chan_setup(char *str) +{ + line_setup(conf, NR_PORTS, &def_conf, str, "serial line"); + return 1; +} + +__setup("ssl", ssl_chan_setup); +__channel_help(ssl_chan_setup, "ssl"); + +static int ssl_non_raw_setup(char *str) +{ + opts.raw = 0; + return 1; +} +__setup("ssl-non-raw", ssl_non_raw_setup); +__uml_help(ssl_non_raw_setup, +"ssl-non-raw\n" +" Set serial lines to non-raw mode.\n\n" +); diff --git a/arch/um/drivers/stderr_console.c b/arch/um/drivers/stderr_console.c new file mode 100644 index 000000000..ecc3a5814 --- /dev/null +++ b/arch/um/drivers/stderr_console.c @@ -0,0 +1,63 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include +#include + +#include "chan_user.h" + +/* ----------------------------------------------------------------------------- */ +/* trivial console driver -- simply dump everything to stderr */ + +/* + * Don't register by default -- as this registers very early in the + * boot process it becomes the default console. + * + * Initialized at init time. + */ +static int use_stderr_console = 0; + +static void stderr_console_write(struct console *console, const char *string, + unsigned len) +{ + generic_write(2 /* stderr */, string, len, NULL); +} + +static struct console stderr_console = { + .name = "stderr", + .write = stderr_console_write, + .flags = CON_PRINTBUFFER, +}; + +static int __init stderr_console_init(void) +{ + if (use_stderr_console) + register_console(&stderr_console); + return 0; +} +console_initcall(stderr_console_init); + +static int stderr_setup(char *str) +{ + if (!str) + return 0; + use_stderr_console = simple_strtoul(str,&str,0); + return 1; +} +__setup("stderr=", stderr_setup); + +/* The previous behavior of not unregistering led to /dev/console being + * impossible to open. My FC5 filesystem started having init die, and the + * system panicing because of this. Unregistering causes the real + * console to become the default console, and /dev/console can then be + * opened. Making this an initcall makes this happen late enough that + * there is no added value in dumping everything to stderr, and the + * normal console is good enough to show you all available output. + */ +static int __init unregister_stderr(void) +{ + unregister_console(&stderr_console); + + return 0; +} + +__initcall(unregister_stderr); diff --git a/arch/um/drivers/stdio_console.c b/arch/um/drivers/stdio_console.c new file mode 100644 index 000000000..1c239737d --- /dev/null +++ b/arch/um/drivers/stdio_console.c @@ -0,0 +1,198 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2000, 2001 Jeff Dike (jdike@karaya.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "stdio_console.h" +#include "chan.h" +#include +#include "mconsole_kern.h" +#include + +#define MAX_TTYS (16) + +static void stdio_announce(char *dev_name, int dev) +{ + printk(KERN_INFO "Virtual console %d assigned device '%s'\n", dev, + dev_name); +} + +/* Almost const, except that xterm_title may be changed in an initcall */ +static struct chan_opts opts = { + .announce = stdio_announce, + .xterm_title = "Virtual Console #%d", + .raw = 1, +}; + +static int con_config(char *str, char **error_out); +static int con_get_config(char *dev, char *str, int size, char **error_out); +static int con_remove(int n, char **con_remove); + + +/* Const, except for .mc.list */ +static struct line_driver driver = { + .name = "UML console", + .device_name = "tty", + .major = TTY_MAJOR, + .minor_start = 0, + .type = TTY_DRIVER_TYPE_CONSOLE, + .subtype = SYSTEM_TYPE_CONSOLE, + .read_irq_name = "console", + .write_irq_name = "console-write", + .mc = { + .list = LIST_HEAD_INIT(driver.mc.list), + .name = "con", + .config = con_config, + .get_config = con_get_config, + .id = line_id, + .remove = con_remove, + }, +}; + +/* The array is initialized by line_init, at initcall time. The + * elements are locked individually as needed. + */ +static char *vt_conf[MAX_TTYS]; +static char *def_conf; +static struct line vts[MAX_TTYS]; + +static int con_config(char *str, char **error_out) +{ + return line_config(vts, ARRAY_SIZE(vts), str, &opts, error_out); +} + +static int con_get_config(char *dev, char *str, int size, char **error_out) +{ + return line_get_config(dev, vts, ARRAY_SIZE(vts), str, size, error_out); +} + +static int con_remove(int n, char **error_out) +{ + return line_remove(vts, ARRAY_SIZE(vts), n, error_out); +} + +/* Set in an initcall, checked in an exitcall */ +static int con_init_done; + +static int con_install(struct tty_driver *driver, struct tty_struct *tty) +{ + return line_install(driver, tty, &vts[tty->index]); +} + +static const struct tty_operations console_ops = { + .open = line_open, + .install = con_install, + .close = line_close, + .write = line_write, + .write_room = line_write_room, + .chars_in_buffer = line_chars_in_buffer, + .flush_buffer = line_flush_buffer, + .flush_chars = line_flush_chars, + .throttle = line_throttle, + .unthrottle = line_unthrottle, + .hangup = line_hangup, +}; + +static void uml_console_write(struct console *console, const char *string, + unsigned len) +{ + struct line *line = &vts[console->index]; + unsigned long flags; + + spin_lock_irqsave(&line->lock, flags); + console_write_chan(line->chan_out, string, len); + spin_unlock_irqrestore(&line->lock, flags); +} + +static struct tty_driver *uml_console_device(struct console *c, int *index) +{ + *index = c->index; + return driver.driver; +} + +static int uml_console_setup(struct console *co, char *options) +{ + struct line *line = &vts[co->index]; + + return console_open_chan(line, co); +} + +/* No locking for register_console call - relies on single-threaded initcalls */ +static struct console stdiocons = { + .name = "tty", + .write = uml_console_write, + .device = uml_console_device, + .setup = uml_console_setup, + .flags = CON_PRINTBUFFER|CON_ANYTIME, + .index = -1, +}; + +static int stdio_init(void) +{ + char *new_title; + int err; + int i; + + err = register_lines(&driver, &console_ops, vts, + ARRAY_SIZE(vts)); + if (err) + return err; + + printk(KERN_INFO "Initialized stdio console driver\n"); + + new_title = add_xterm_umid(opts.xterm_title); + if(new_title != NULL) + opts.xterm_title = new_title; + + for (i = 0; i < MAX_TTYS; i++) { + char *error; + char *s = vt_conf[i]; + if (!s) + s = def_conf; + if (!s) + s = i ? CONFIG_CON_CHAN : CONFIG_CON_ZERO_CHAN; + if (setup_one_line(vts, i, s, &opts, &error)) + printk(KERN_ERR "setup_one_line failed for " + "device %d : %s\n", i, error); + } + + con_init_done = 1; + register_console(&stdiocons); + return 0; +} +late_initcall(stdio_init); + +static void console_exit(void) +{ + if (!con_init_done) + return; + close_lines(vts, ARRAY_SIZE(vts)); +} +__uml_exitcall(console_exit); + +static int console_chan_setup(char *str) +{ + if (!strncmp(str, "sole=", 5)) /* console= option specifies tty */ + return 0; + + line_setup(vt_conf, MAX_TTYS, &def_conf, str, "console"); + return 1; +} +__setup("con", console_chan_setup); +__channel_help(console_chan_setup, "con"); diff --git a/arch/um/drivers/stdio_console.h b/arch/um/drivers/stdio_console.h new file mode 100644 index 000000000..3a409ec23 --- /dev/null +++ b/arch/um/drivers/stdio_console.h @@ -0,0 +1,11 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2000 Jeff Dike (jdike@karaya.com) + */ + +#ifndef __STDIO_CONSOLE_H +#define __STDIO_CONSOLE_H + +extern void save_console_flags(void); +#endif + diff --git a/arch/um/drivers/tty.c b/arch/um/drivers/tty.c new file mode 100644 index 000000000..884a762d2 --- /dev/null +++ b/arch/um/drivers/tty.c @@ -0,0 +1,81 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{linux.intel,addtoit}.com) + */ + +#include +#include +#include +#include "chan_user.h" +#include +#include + +struct tty_chan { + char *dev; + int raw; + struct termios tt; +}; + +static void *tty_chan_init(char *str, int device, const struct chan_opts *opts) +{ + struct tty_chan *data; + + if (*str != ':') { + printk(UM_KERN_ERR "tty_init : channel type 'tty' must specify " + "a device\n"); + return NULL; + } + str++; + + data = uml_kmalloc(sizeof(*data), UM_GFP_KERNEL); + if (data == NULL) + return NULL; + *data = ((struct tty_chan) { .dev = str, + .raw = opts->raw }); + + return data; +} + +static int tty_open(int input, int output, int primary, void *d, + char **dev_out) +{ + struct tty_chan *data = d; + int fd, err, mode = 0; + + if (input && output) + mode = O_RDWR; + else if (input) + mode = O_RDONLY; + else if (output) + mode = O_WRONLY; + + fd = open(data->dev, mode); + if (fd < 0) + return -errno; + + if (data->raw) { + CATCH_EINTR(err = tcgetattr(fd, &data->tt)); + if (err) + return err; + + err = raw(fd); + if (err) + return err; + } + + *dev_out = data->dev; + return fd; +} + +const struct chan_ops tty_ops = { + .type = "tty", + .init = tty_chan_init, + .open = tty_open, + .close = generic_close, + .read = generic_read, + .write = generic_write, + .console_write = generic_console_write, + .window_size = generic_window_size, + .free = generic_free, + .winch = 0, +}; diff --git a/arch/um/drivers/ubd.h b/arch/um/drivers/ubd.h new file mode 100644 index 000000000..2985c1466 --- /dev/null +++ b/arch/um/drivers/ubd.h @@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2000 Jeff Dike (jdike@karaya.com) + * Copyright (C) 2001 RidgeRun, Inc (glonnon@ridgerun.com) + */ + +#ifndef __UM_UBD_USER_H +#define __UM_UBD_USER_H + +#include + +int start_io_thread(struct os_helper_thread **td_out, int *fd_out); +void *io_thread(void *arg); +extern int kernel_fd; + +extern int ubd_read_poll(int timeout); +extern int ubd_write_poll(int timeout); + +#define UBD_REQ_BUFFER_SIZE 64 + +#endif + diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c new file mode 100644 index 000000000..20fc33300 --- /dev/null +++ b/arch/um/drivers/ubd_kern.c @@ -0,0 +1,1540 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2018 Cambridge Greys Ltd + * Copyright (C) 2015-2016 Anton Ivanov (aivanov@brocade.com) + * Copyright (C) 2000 Jeff Dike (jdike@karaya.com) + */ + +/* 2001-09-28...2002-04-17 + * Partition stuff by James_McMechan@hotmail.com + * old style ubd by setting UBD_SHIFT to 0 + * 2002-09-27...2002-10-18 massive tinkering for 2.5 + * partitions have changed in 2.5 + * 2003-01-29 more tinkering for 2.5.59-1 + * This should now address the sysfs problems and has + * the symlink for devfs to allow for booting with + * the common /dev/ubd/discX/... names rather than + * only /dev/ubdN/discN this version also has lots of + * clean ups preparing for ubd-many. + * James McMechan + */ + +#define UBD_SHIFT 4 + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "mconsole_kern.h" +#include +#include +#include "ubd.h" +#include +#include "cow.h" + +/* Max request size is determined by sector mask - 32K */ +#define UBD_MAX_REQUEST (8 * sizeof(long)) + +struct io_desc { + char *buffer; + unsigned long length; + unsigned long sector_mask; + unsigned long long cow_offset; + unsigned long bitmap_words[2]; +}; + +struct io_thread_req { + struct request *req; + int fds[2]; + unsigned long offsets[2]; + unsigned long long offset; + int sectorsize; + int error; + + int desc_cnt; + /* io_desc has to be the last element of the struct */ + struct io_desc io_desc[]; +}; + + +static struct io_thread_req **irq_req_buffer; +static struct io_thread_req *irq_remainder; +static int irq_remainder_size; + +static struct io_thread_req **io_req_buffer; +static struct io_thread_req *io_remainder; +static int io_remainder_size; + + + +static inline int ubd_test_bit(__u64 bit, unsigned char *data) +{ + __u64 n; + int bits, off; + + bits = sizeof(data[0]) * 8; + n = bit / bits; + off = bit % bits; + return (data[n] & (1 << off)) != 0; +} + +static inline void ubd_set_bit(__u64 bit, unsigned char *data) +{ + __u64 n; + int bits, off; + + bits = sizeof(data[0]) * 8; + n = bit / bits; + off = bit % bits; + data[n] |= (1 << off); +} +/*End stuff from ubd_user.h*/ + +#define DRIVER_NAME "uml-blkdev" + +static DEFINE_MUTEX(ubd_lock); + +static int ubd_ioctl(struct block_device *bdev, blk_mode_t mode, + unsigned int cmd, unsigned long arg); +static int ubd_getgeo(struct gendisk *disk, struct hd_geometry *geo); + +#define MAX_DEV (16) + +static const struct block_device_operations ubd_blops = { + .owner = THIS_MODULE, + .ioctl = ubd_ioctl, + .compat_ioctl = blkdev_compat_ptr_ioctl, + .getgeo = ubd_getgeo, +}; + +#ifdef CONFIG_BLK_DEV_UBD_SYNC +#define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 1, .c = 0, \ + .cl = 1 }) +#else +#define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 0, .c = 0, \ + .cl = 1 }) +#endif +static struct openflags global_openflags = OPEN_FLAGS; + +struct cow { + /* backing file name */ + char *file; + /* backing file fd */ + int fd; + unsigned long *bitmap; + unsigned long bitmap_len; + int bitmap_offset; + int data_offset; +}; + +#define MAX_SG 64 + +struct ubd { + /* name (and fd, below) of the file opened for writing, either the + * backing or the cow file. */ + char *file; + char *serial; + int fd; + __u64 size; + struct openflags boot_openflags; + struct openflags openflags; + unsigned shared:1; + unsigned no_cow:1; + unsigned no_trim:1; + struct cow cow; + struct platform_device pdev; + struct gendisk *disk; + struct blk_mq_tag_set tag_set; + spinlock_t lock; +}; + +#define DEFAULT_COW { \ + .file = NULL, \ + .fd = -1, \ + .bitmap = NULL, \ + .bitmap_offset = 0, \ + .data_offset = 0, \ +} + +#define DEFAULT_UBD { \ + .file = NULL, \ + .serial = NULL, \ + .fd = -1, \ + .size = -1, \ + .boot_openflags = OPEN_FLAGS, \ + .openflags = OPEN_FLAGS, \ + .no_cow = 0, \ + .no_trim = 0, \ + .shared = 0, \ + .cow = DEFAULT_COW, \ + .lock = __SPIN_LOCK_UNLOCKED(ubd_devs.lock), \ +} + +/* Protected by ubd_lock */ +static struct ubd ubd_devs[MAX_DEV] = { [0 ... MAX_DEV - 1] = DEFAULT_UBD }; + +static blk_status_t ubd_queue_rq(struct blk_mq_hw_ctx *hctx, + const struct blk_mq_queue_data *bd); + +static int fake_ide_setup(char *str) +{ + pr_warn("The fake_ide option has been removed\n"); + return 1; +} +__setup("fake_ide", fake_ide_setup); + +__uml_help(fake_ide_setup, +"fake_ide\n" +" Obsolete stub.\n\n" +); + +static int parse_unit(char **ptr) +{ + char *str = *ptr, *end; + int n = -1; + + if(isdigit(*str)) { + n = simple_strtoul(str, &end, 0); + if(end == str) + return -1; + *ptr = end; + } + else if (('a' <= *str) && (*str <= 'z')) { + n = *str - 'a'; + str++; + *ptr = str; + } + return n; +} + +/* If *index_out == -1 at exit, the passed option was a general one; + * otherwise, the str pointer is used (and owned) inside ubd_devs array, so it + * should not be freed on exit. + */ +static int ubd_setup_common(char *str, int *index_out, char **error_out) +{ + struct ubd *ubd_dev; + struct openflags flags = global_openflags; + char *file, *backing_file, *serial; + int n, err = 0, i; + + if(index_out) *index_out = -1; + n = *str; + if(n == '='){ + str++; + if(!strcmp(str, "sync")){ + global_openflags = of_sync(global_openflags); + return err; + } + + pr_warn("fake major not supported any more\n"); + return 0; + } + + n = parse_unit(&str); + if(n < 0){ + *error_out = "Couldn't parse device number"; + return -EINVAL; + } + if(n >= MAX_DEV){ + *error_out = "Device number out of range"; + return 1; + } + + err = -EBUSY; + mutex_lock(&ubd_lock); + + ubd_dev = &ubd_devs[n]; + if(ubd_dev->file != NULL){ + *error_out = "Device is already configured"; + goto out; + } + + if (index_out) + *index_out = n; + + err = -EINVAL; + for (i = 0; i < sizeof("rscdt="); i++) { + switch (*str) { + case 'r': + flags.w = 0; + break; + case 's': + flags.s = 1; + break; + case 'd': + ubd_dev->no_cow = 1; + break; + case 'c': + ubd_dev->shared = 1; + break; + case 't': + ubd_dev->no_trim = 1; + break; + case '=': + str++; + goto break_loop; + default: + *error_out = "Expected '=' or flag letter " + "(r, s, c, t or d)"; + goto out; + } + str++; + } + + if (*str == '=') + *error_out = "Too many flags specified"; + else + *error_out = "Missing '='"; + goto out; + +break_loop: + file = strsep(&str, ",:"); + if (*file == '\0') + file = NULL; + + backing_file = strsep(&str, ",:"); + if (backing_file && *backing_file == '\0') + backing_file = NULL; + + serial = strsep(&str, ",:"); + if (serial && *serial == '\0') + serial = NULL; + + if (backing_file && ubd_dev->no_cow) { + *error_out = "Can't specify both 'd' and a cow file"; + goto out; + } + + err = 0; + ubd_dev->file = file; + ubd_dev->cow.file = backing_file; + ubd_dev->serial = serial; + ubd_dev->boot_openflags = flags; +out: + mutex_unlock(&ubd_lock); + return err; +} + +static int ubd_setup(char *str) +{ + char *error; + int err; + + err = ubd_setup_common(str, NULL, &error); + if(err) + printk(KERN_ERR "Failed to initialize device with \"%s\" : " + "%s\n", str, error); + return 1; +} + +__setup("ubd", ubd_setup); +__uml_help(ubd_setup, +"ubd=[(:|,)][(:|,)]\n" +" This is used to associate a device with a file in the underlying\n" +" filesystem. When specifying two filenames, the first one is the\n" +" COW name and the second is the backing file name. As separator you can\n" +" use either a ':' or a ',': the first one allows writing things like;\n" +" ubd0=~/Uml/root_cow:~/Uml/root_backing_file\n" +" while with a ',' the shell would not expand the 2nd '~'.\n" +" When using only one filename, UML will detect whether to treat it like\n" +" a COW file or a backing file. To override this detection, add the 'd'\n" +" flag:\n" +" ubd0d=BackingFile\n" +" Usually, there is a filesystem in the file, but \n" +" that's not required. Swap devices containing swap files can be\n" +" specified like this. Also, a file which doesn't contain a\n" +" filesystem can have its contents read in the virtual \n" +" machine by running 'dd' on the device. must be in the range\n" +" 0 to 7. Appending an 'r' to the number will cause that device\n" +" to be mounted read-only. For example ubd1r=./ext_fs. Appending\n" +" an 's' will cause data to be written to disk on the host immediately.\n" +" 'c' will cause the device to be treated as being shared between multiple\n" +" UMLs and file locking will be turned off - this is appropriate for a\n" +" cluster filesystem and inappropriate at almost all other times.\n\n" +" 't' will disable trim/discard support on the device (enabled by default).\n\n" +" An optional device serial number can be exposed using the serial parameter\n" +" on the cmdline which is exposed as a sysfs entry. This is particularly\n" +" useful when a unique number should be given to the device. Note when\n" +" specifying a label, the filename2 must be also presented. It can be\n" +" an empty string, in which case the backing file is not used:\n" +" ubd0=File,,Serial\n\n" +); + +static int udb_setup(char *str) +{ + printk("udb%s specified on command line is almost certainly a ubd -> " + "udb TYPO\n", str); + return 1; +} + +__setup("udb", udb_setup); +__uml_help(udb_setup, +"udb\n" +" This option is here solely to catch ubd -> udb typos, which can be\n" +" to impossible to catch visually unless you specifically look for\n" +" them. The only result of any option starting with 'udb' is an error\n" +" in the boot output.\n\n" +); + +/* Only changed by ubd_init, which is an initcall. */ +static int thread_fd = -1; + +/* Function to read several request pointers at a time +* handling fractional reads if (and as) needed +*/ + +static int bulk_req_safe_read( + int fd, + struct io_thread_req **request_buffer, + struct io_thread_req **remainder, + int *remainder_size, + int max_recs + ) +{ + int n = 0; + int res = 0; + + if (*remainder_size > 0) { + memmove( + (char *) request_buffer, + (char *) remainder, *remainder_size + ); + n = *remainder_size; + } + + res = os_read_file( + fd, + ((char *) request_buffer) + *remainder_size, + sizeof(struct io_thread_req *)*max_recs + - *remainder_size + ); + if (res > 0) { + n += res; + if ((n % sizeof(struct io_thread_req *)) > 0) { + /* + * Read somehow returned not a multiple of dword + * theoretically possible, but never observed in the + * wild, so read routine must be able to handle it + */ + *remainder_size = n % sizeof(struct io_thread_req *); + WARN(*remainder_size > 0, "UBD IPC read returned a partial result"); + memmove( + remainder, + ((char *) request_buffer) + + (n/sizeof(struct io_thread_req *))*sizeof(struct io_thread_req *), + *remainder_size + ); + n = n - *remainder_size; + } + } else { + n = res; + } + return n; +} + +static void ubd_end_request(struct io_thread_req *io_req) +{ + if (io_req->error == BLK_STS_NOTSUPP) { + if (req_op(io_req->req) == REQ_OP_DISCARD) + blk_queue_disable_discard(io_req->req->q); + else if (req_op(io_req->req) == REQ_OP_WRITE_ZEROES) + blk_queue_disable_write_zeroes(io_req->req->q); + } + blk_mq_end_request(io_req->req, io_req->error); + kfree(io_req); +} + +static irqreturn_t ubd_intr(int irq, void *dev) +{ + int len, i; + + while ((len = bulk_req_safe_read(thread_fd, irq_req_buffer, + &irq_remainder, &irq_remainder_size, + UBD_REQ_BUFFER_SIZE)) >= 0) { + for (i = 0; i < len / sizeof(struct io_thread_req *); i++) + ubd_end_request(irq_req_buffer[i]); + } + + if (len < 0 && len != -EAGAIN) + pr_err("spurious interrupt in %s, err = %d\n", __func__, len); + return IRQ_HANDLED; +} + +/* Only changed by ubd_init, which is an initcall. */ +static struct os_helper_thread *io_td; + +static void kill_io_thread(void) +{ + if (io_td) + os_kill_helper_thread(io_td); +} + +__uml_exitcall(kill_io_thread); + +static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out) +{ + char *file; + int fd; + int err; + + __u32 version; + __u32 align; + char *backing_file; + time64_t mtime; + unsigned long long size; + int sector_size; + int bitmap_offset; + + if (ubd_dev->file && ubd_dev->cow.file) { + file = ubd_dev->cow.file; + + goto out; + } + + fd = os_open_file(ubd_dev->file, of_read(OPENFLAGS()), 0); + if (fd < 0) + return fd; + + err = read_cow_header(file_reader, &fd, &version, &backing_file, \ + &mtime, &size, §or_size, &align, &bitmap_offset); + os_close_file(fd); + + if(err == -EINVAL) + file = ubd_dev->file; + else + file = backing_file; + +out: + return os_file_size(file, size_out); +} + +static int read_cow_bitmap(int fd, void *buf, int offset, int len) +{ + int err; + + err = os_pread_file(fd, buf, len, offset); + if (err < 0) + return err; + + return 0; +} + +static int backing_file_mismatch(char *file, __u64 size, time64_t mtime) +{ + time64_t modtime; + unsigned long long actual; + int err; + + err = os_file_modtime(file, &modtime); + if (err < 0) { + printk(KERN_ERR "Failed to get modification time of backing " + "file \"%s\", err = %d\n", file, -err); + return err; + } + + err = os_file_size(file, &actual); + if (err < 0) { + printk(KERN_ERR "Failed to get size of backing file \"%s\", " + "err = %d\n", file, -err); + return err; + } + + if (actual != size) { + /*__u64 can be a long on AMD64 and with %lu GCC complains; so + * the typecast.*/ + printk(KERN_ERR "Size mismatch (%llu vs %llu) of COW header " + "vs backing file\n", (unsigned long long) size, actual); + return -EINVAL; + } + if (modtime != mtime) { + printk(KERN_ERR "mtime mismatch (%lld vs %lld) of COW header vs " + "backing file\n", mtime, modtime); + return -EINVAL; + } + return 0; +} + +static int path_requires_switch(char *from_cmdline, char *from_cow, char *cow) +{ + struct uml_stat buf1, buf2; + int err; + + if (from_cmdline == NULL) + return 0; + if (!strcmp(from_cmdline, from_cow)) + return 0; + + err = os_stat_file(from_cmdline, &buf1); + if (err < 0) { + printk(KERN_ERR "Couldn't stat '%s', err = %d\n", from_cmdline, + -err); + return 0; + } + err = os_stat_file(from_cow, &buf2); + if (err < 0) { + printk(KERN_ERR "Couldn't stat '%s', err = %d\n", from_cow, + -err); + return 1; + } + if ((buf1.ust_dev == buf2.ust_dev) && (buf1.ust_ino == buf2.ust_ino)) + return 0; + + printk(KERN_ERR "Backing file mismatch - \"%s\" requested, " + "\"%s\" specified in COW header of \"%s\"\n", + from_cmdline, from_cow, cow); + return 1; +} + +static int open_ubd_file(char *file, struct openflags *openflags, int shared, + char **backing_file_out, int *bitmap_offset_out, + unsigned long *bitmap_len_out, int *data_offset_out, + int *create_cow_out) +{ + time64_t mtime; + unsigned long long size; + __u32 version, align; + char *backing_file; + int fd, err, sectorsize, asked_switch, mode = 0644; + + fd = os_open_file(file, *openflags, mode); + if (fd < 0) { + if ((fd == -ENOENT) && (create_cow_out != NULL)) + *create_cow_out = 1; + if (!openflags->w || + ((fd != -EROFS) && (fd != -EACCES))) + return fd; + openflags->w = 0; + fd = os_open_file(file, *openflags, mode); + if (fd < 0) + return fd; + } + + if (shared) + printk(KERN_INFO "Not locking \"%s\" on the host\n", file); + else { + err = os_lock_file(fd, openflags->w); + if (err < 0) { + printk(KERN_ERR "Failed to lock '%s', err = %d\n", + file, -err); + goto out_close; + } + } + + /* Successful return case! */ + if (backing_file_out == NULL) + return fd; + + err = read_cow_header(file_reader, &fd, &version, &backing_file, &mtime, + &size, §orsize, &align, bitmap_offset_out); + if (err && (*backing_file_out != NULL)) { + printk(KERN_ERR "Failed to read COW header from COW file " + "\"%s\", errno = %d\n", file, -err); + goto out_close; + } + if (err) + return fd; + + asked_switch = path_requires_switch(*backing_file_out, backing_file, + file); + + /* Allow switching only if no mismatch. */ + if (asked_switch && !backing_file_mismatch(*backing_file_out, size, + mtime)) { + printk(KERN_ERR "Switching backing file to '%s'\n", + *backing_file_out); + err = write_cow_header(file, fd, *backing_file_out, + sectorsize, align, &size); + if (err) { + printk(KERN_ERR "Switch failed, errno = %d\n", -err); + goto out_close; + } + } else { + *backing_file_out = backing_file; + err = backing_file_mismatch(*backing_file_out, size, mtime); + if (err) + goto out_close; + } + + cow_sizes(version, size, sectorsize, align, *bitmap_offset_out, + bitmap_len_out, data_offset_out); + + return fd; + out_close: + os_close_file(fd); + return err; +} + +static int create_cow_file(char *cow_file, char *backing_file, + struct openflags flags, + int sectorsize, int alignment, int *bitmap_offset_out, + unsigned long *bitmap_len_out, int *data_offset_out) +{ + int err, fd; + + flags.c = 1; + fd = open_ubd_file(cow_file, &flags, 0, NULL, NULL, NULL, NULL, NULL); + if (fd < 0) { + err = fd; + printk(KERN_ERR "Open of COW file '%s' failed, errno = %d\n", + cow_file, -err); + goto out; + } + + err = init_cow_file(fd, cow_file, backing_file, sectorsize, alignment, + bitmap_offset_out, bitmap_len_out, + data_offset_out); + if (!err) + return fd; + os_close_file(fd); + out: + return err; +} + +static void ubd_close_dev(struct ubd *ubd_dev) +{ + os_close_file(ubd_dev->fd); + if(ubd_dev->cow.file == NULL) + return; + + os_close_file(ubd_dev->cow.fd); + vfree(ubd_dev->cow.bitmap); + ubd_dev->cow.bitmap = NULL; +} + +static int ubd_open_dev(struct ubd *ubd_dev) +{ + struct openflags flags; + char **back_ptr; + int err, create_cow, *create_ptr; + int fd; + + ubd_dev->openflags = ubd_dev->boot_openflags; + create_cow = 0; + create_ptr = (ubd_dev->cow.file != NULL) ? &create_cow : NULL; + back_ptr = ubd_dev->no_cow ? NULL : &ubd_dev->cow.file; + + fd = open_ubd_file(ubd_dev->file, &ubd_dev->openflags, ubd_dev->shared, + back_ptr, &ubd_dev->cow.bitmap_offset, + &ubd_dev->cow.bitmap_len, &ubd_dev->cow.data_offset, + create_ptr); + + if((fd == -ENOENT) && create_cow){ + fd = create_cow_file(ubd_dev->file, ubd_dev->cow.file, + ubd_dev->openflags, SECTOR_SIZE, PAGE_SIZE, + &ubd_dev->cow.bitmap_offset, + &ubd_dev->cow.bitmap_len, + &ubd_dev->cow.data_offset); + if(fd >= 0){ + printk(KERN_INFO "Creating \"%s\" as COW file for " + "\"%s\"\n", ubd_dev->file, ubd_dev->cow.file); + } + } + + if(fd < 0){ + printk("Failed to open '%s', errno = %d\n", ubd_dev->file, + -fd); + return fd; + } + ubd_dev->fd = fd; + + if(ubd_dev->cow.file != NULL){ + err = -ENOMEM; + ubd_dev->cow.bitmap = vmalloc(ubd_dev->cow.bitmap_len); + if(ubd_dev->cow.bitmap == NULL){ + printk(KERN_ERR "Failed to vmalloc COW bitmap\n"); + goto error; + } + + err = read_cow_bitmap(ubd_dev->fd, ubd_dev->cow.bitmap, + ubd_dev->cow.bitmap_offset, + ubd_dev->cow.bitmap_len); + if(err < 0) + goto error; + + flags = ubd_dev->openflags; + flags.w = 0; + err = open_ubd_file(ubd_dev->cow.file, &flags, ubd_dev->shared, NULL, + NULL, NULL, NULL, NULL); + if(err < 0) goto error; + ubd_dev->cow.fd = err; + } + return 0; + error: + os_close_file(ubd_dev->fd); + return err; +} + +static void ubd_device_release(struct device *dev) +{ + struct ubd *ubd_dev = container_of(dev, struct ubd, pdev.dev); + + blk_mq_free_tag_set(&ubd_dev->tag_set); + *ubd_dev = ((struct ubd) DEFAULT_UBD); +} + +static ssize_t serial_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct gendisk *disk = dev_to_disk(dev); + struct ubd *ubd_dev = disk->private_data; + + if (!ubd_dev) + return 0; + + return sprintf(buf, "%s", ubd_dev->serial); +} + +static DEVICE_ATTR_RO(serial); + +static struct attribute *ubd_attrs[] = { + &dev_attr_serial.attr, + NULL, +}; + +static umode_t ubd_attrs_are_visible(struct kobject *kobj, + struct attribute *a, int n) +{ + return a->mode; +} + +static const struct attribute_group ubd_attr_group = { + .attrs = ubd_attrs, + .is_visible = ubd_attrs_are_visible, +}; + +static const struct attribute_group *ubd_attr_groups[] = { + &ubd_attr_group, + NULL, +}; + +#define ROUND_BLOCK(n) ((n + (SECTOR_SIZE - 1)) & (-SECTOR_SIZE)) + +static const struct blk_mq_ops ubd_mq_ops = { + .queue_rq = ubd_queue_rq, +}; + +static int ubd_add(int n, char **error_out) +{ + struct ubd *ubd_dev = &ubd_devs[n]; + struct queue_limits lim = { + .max_segments = MAX_SG, + .seg_boundary_mask = PAGE_SIZE - 1, + .features = BLK_FEAT_WRITE_CACHE, + }; + struct gendisk *disk; + int err = 0; + + if(ubd_dev->file == NULL) + goto out; + + if (ubd_dev->cow.file) + lim.max_hw_sectors = 8 * sizeof(long); + if (!ubd_dev->no_trim) { + lim.max_hw_discard_sectors = UBD_MAX_REQUEST; + lim.max_write_zeroes_sectors = UBD_MAX_REQUEST; + } + + err = ubd_file_size(ubd_dev, &ubd_dev->size); + if(err < 0){ + *error_out = "Couldn't determine size of device's file"; + goto out; + } + + err = ubd_open_dev(ubd_dev); + if (err) { + pr_err("ubd%c: Can't open \"%s\": errno = %d\n", + 'a' + n, ubd_dev->file, -err); + goto out; + } + + ubd_dev->size = ROUND_BLOCK(ubd_dev->size); + + ubd_dev->tag_set.ops = &ubd_mq_ops; + ubd_dev->tag_set.queue_depth = 64; + ubd_dev->tag_set.numa_node = NUMA_NO_NODE; + ubd_dev->tag_set.driver_data = ubd_dev; + ubd_dev->tag_set.nr_hw_queues = 1; + + err = blk_mq_alloc_tag_set(&ubd_dev->tag_set); + if (err) + goto out_close; + + disk = blk_mq_alloc_disk(&ubd_dev->tag_set, &lim, ubd_dev); + if (IS_ERR(disk)) { + err = PTR_ERR(disk); + goto out_cleanup_tags; + } + + disk->major = UBD_MAJOR; + disk->first_minor = n << UBD_SHIFT; + disk->minors = 1 << UBD_SHIFT; + disk->fops = &ubd_blops; + set_capacity(disk, ubd_dev->size / 512); + sprintf(disk->disk_name, "ubd%c", 'a' + n); + disk->private_data = ubd_dev; + set_disk_ro(disk, !ubd_dev->openflags.w); + + ubd_dev->pdev.id = n; + ubd_dev->pdev.name = DRIVER_NAME; + ubd_dev->pdev.dev.release = ubd_device_release; + dev_set_drvdata(&ubd_dev->pdev.dev, ubd_dev); + platform_device_register(&ubd_dev->pdev); + + err = device_add_disk(&ubd_dev->pdev.dev, disk, ubd_attr_groups); + if (err) + goto out_cleanup_disk; + + ubd_dev->disk = disk; + + return 0; + +out_cleanup_disk: + put_disk(disk); +out_cleanup_tags: + blk_mq_free_tag_set(&ubd_dev->tag_set); +out_close: + ubd_close_dev(ubd_dev); +out: + return err; +} + +static int ubd_config(char *str, char **error_out) +{ + int n, ret; + + /* This string is possibly broken up and stored, so it's only + * freed if ubd_setup_common fails, or if only general options + * were set. + */ + str = kstrdup(str, GFP_KERNEL); + if (str == NULL) { + *error_out = "Failed to allocate memory"; + return -ENOMEM; + } + + ret = ubd_setup_common(str, &n, error_out); + if (ret) + goto err_free; + + if (n == -1) { + ret = 0; + goto err_free; + } + + mutex_lock(&ubd_lock); + ret = ubd_add(n, error_out); + if (ret) + ubd_devs[n].file = NULL; + mutex_unlock(&ubd_lock); + +out: + return ret; + +err_free: + kfree(str); + goto out; +} + +static int ubd_get_config(char *name, char *str, int size, char **error_out) +{ + struct ubd *ubd_dev; + int n, len = 0; + + n = parse_unit(&name); + if((n >= MAX_DEV) || (n < 0)){ + *error_out = "ubd_get_config : device number out of range"; + return -1; + } + + ubd_dev = &ubd_devs[n]; + mutex_lock(&ubd_lock); + + if(ubd_dev->file == NULL){ + CONFIG_CHUNK(str, size, len, "", 1); + goto out; + } + + CONFIG_CHUNK(str, size, len, ubd_dev->file, 0); + + if(ubd_dev->cow.file != NULL){ + CONFIG_CHUNK(str, size, len, ",", 0); + CONFIG_CHUNK(str, size, len, ubd_dev->cow.file, 1); + } + else CONFIG_CHUNK(str, size, len, "", 1); + + out: + mutex_unlock(&ubd_lock); + return len; +} + +static int ubd_id(char **str, int *start_out, int *end_out) +{ + int n; + + n = parse_unit(str); + *start_out = 0; + *end_out = MAX_DEV - 1; + return n; +} + +static int ubd_remove(int n, char **error_out) +{ + struct ubd *ubd_dev; + int err = -ENODEV; + + mutex_lock(&ubd_lock); + + ubd_dev = &ubd_devs[n]; + + if(ubd_dev->file == NULL) + goto out; + + if (ubd_dev->disk) { + /* you cannot remove a open disk */ + err = -EBUSY; + if (disk_openers(ubd_dev->disk)) + goto out; + + del_gendisk(ubd_dev->disk); + ubd_close_dev(ubd_dev); + put_disk(ubd_dev->disk); + } + + err = 0; + platform_device_unregister(&ubd_dev->pdev); +out: + mutex_unlock(&ubd_lock); + return err; +} + +/* All these are called by mconsole in process context and without + * ubd-specific locks. The structure itself is const except for .list. + */ +static struct mc_device ubd_mc = { + .list = LIST_HEAD_INIT(ubd_mc.list), + .name = "ubd", + .config = ubd_config, + .get_config = ubd_get_config, + .id = ubd_id, + .remove = ubd_remove, +}; + +static int __init ubd_mc_init(void) +{ + mconsole_register_dev(&ubd_mc); + return 0; +} + +__initcall(ubd_mc_init); + +static int __init ubd0_init(void) +{ + struct ubd *ubd_dev = &ubd_devs[0]; + + mutex_lock(&ubd_lock); + if(ubd_dev->file == NULL) + ubd_dev->file = "root_fs"; + mutex_unlock(&ubd_lock); + + return 0; +} + +__initcall(ubd0_init); + +/* Used in ubd_init, which is an initcall */ +static struct platform_driver ubd_driver = { + .driver = { + .name = DRIVER_NAME, + }, +}; + +static int __init ubd_init(void) +{ + char *error; + int i, err; + + if (register_blkdev(UBD_MAJOR, "ubd")) + return -1; + + irq_req_buffer = kmalloc_objs(struct io_thread_req *, + UBD_REQ_BUFFER_SIZE); + irq_remainder = 0; + + if (irq_req_buffer == NULL) { + printk(KERN_ERR "Failed to initialize ubd buffering\n"); + return -ENOMEM; + } + io_req_buffer = kmalloc_objs(struct io_thread_req *, + UBD_REQ_BUFFER_SIZE); + + io_remainder = 0; + + if (io_req_buffer == NULL) { + printk(KERN_ERR "Failed to initialize ubd buffering\n"); + return -ENOMEM; + } + platform_driver_register(&ubd_driver); + mutex_lock(&ubd_lock); + for (i = 0; i < MAX_DEV; i++){ + err = ubd_add(i, &error); + if(err) + printk(KERN_ERR "Failed to initialize ubd device %d :" + "%s\n", i, error); + } + mutex_unlock(&ubd_lock); + return 0; +} + +late_initcall(ubd_init); + +static int __init ubd_driver_init(void) +{ + int err; + + /* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/ + if(global_openflags.s){ + printk(KERN_INFO "ubd: Synchronous mode\n"); + /* Letting ubd=sync be like using ubd#s= instead of ubd#= is + * enough. So use anyway the io thread. */ + } + err = start_io_thread(&io_td, &thread_fd); + if (err < 0) { + printk(KERN_ERR + "ubd : Failed to start I/O thread (errno = %d) - " + "falling back to synchronous I/O\n", -err); + return 0; + } + err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr, + 0, "ubd", ubd_devs); + if(err < 0) + printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err); + return 0; +} + +device_initcall(ubd_driver_init); + +static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask, + __u64 *cow_offset, unsigned long *bitmap, + __u64 bitmap_offset, unsigned long *bitmap_words, + __u64 bitmap_len) +{ + __u64 sector = io_offset >> SECTOR_SHIFT; + int i, update_bitmap = 0; + + for (i = 0; i < length >> SECTOR_SHIFT; i++) { + if(cow_mask != NULL) + ubd_set_bit(i, (unsigned char *) cow_mask); + if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) + continue; + + update_bitmap = 1; + ubd_set_bit(sector + i, (unsigned char *) bitmap); + } + + if(!update_bitmap) + return; + + *cow_offset = sector / (sizeof(unsigned long) * 8); + + /* This takes care of the case where we're exactly at the end of the + * device, and *cow_offset + 1 is off the end. So, just back it up + * by one word. Thanks to Lynn Kerby for the fix and James McMechan + * for the original diagnosis. + */ + if (*cow_offset == (DIV_ROUND_UP(bitmap_len, + sizeof(unsigned long)) - 1)) + (*cow_offset)--; + + bitmap_words[0] = bitmap[*cow_offset]; + bitmap_words[1] = bitmap[*cow_offset + 1]; + + *cow_offset *= sizeof(unsigned long); + *cow_offset += bitmap_offset; +} + +static void cowify_req(struct io_thread_req *req, struct io_desc *segment, + unsigned long offset, unsigned long *bitmap, + __u64 bitmap_offset, __u64 bitmap_len) +{ + __u64 sector = offset >> SECTOR_SHIFT; + int i; + + if (segment->length > (sizeof(segment->sector_mask) * 8) << SECTOR_SHIFT) + panic("Operation too long"); + + if (req_op(req->req) == REQ_OP_READ) { + for (i = 0; i < segment->length >> SECTOR_SHIFT; i++) { + if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) + ubd_set_bit(i, (unsigned char *) + &segment->sector_mask); + } + } else { + cowify_bitmap(offset, segment->length, &segment->sector_mask, + &segment->cow_offset, bitmap, bitmap_offset, + segment->bitmap_words, bitmap_len); + } +} + +static void ubd_map_req(struct ubd *dev, struct io_thread_req *io_req, + struct request *req) +{ + struct bio_vec bvec; + struct req_iterator iter; + int i = 0; + unsigned long byte_offset = io_req->offset; + enum req_op op = req_op(req); + + if (op == REQ_OP_WRITE_ZEROES || op == REQ_OP_DISCARD) { + io_req->io_desc[0].buffer = NULL; + io_req->io_desc[0].length = blk_rq_bytes(req); + } else { + rq_for_each_segment(bvec, req, iter) { + BUG_ON(i >= io_req->desc_cnt); + + io_req->io_desc[i].buffer = bvec_virt(&bvec); + io_req->io_desc[i].length = bvec.bv_len; + i++; + } + } + + if (dev->cow.file) { + for (i = 0; i < io_req->desc_cnt; i++) { + cowify_req(io_req, &io_req->io_desc[i], byte_offset, + dev->cow.bitmap, dev->cow.bitmap_offset, + dev->cow.bitmap_len); + byte_offset += io_req->io_desc[i].length; + } + + } +} + +static struct io_thread_req *ubd_alloc_req(struct ubd *dev, struct request *req, + int desc_cnt) +{ + struct io_thread_req *io_req; + int i; + + io_req = kmalloc(sizeof(*io_req) + + (desc_cnt * sizeof(struct io_desc)), + GFP_ATOMIC); + if (!io_req) + return NULL; + + io_req->req = req; + if (dev->cow.file) + io_req->fds[0] = dev->cow.fd; + else + io_req->fds[0] = dev->fd; + io_req->error = 0; + io_req->sectorsize = SECTOR_SIZE; + io_req->fds[1] = dev->fd; + io_req->offset = (u64) blk_rq_pos(req) << SECTOR_SHIFT; + io_req->offsets[0] = 0; + io_req->offsets[1] = dev->cow.data_offset; + + for (i = 0 ; i < desc_cnt; i++) { + io_req->io_desc[i].sector_mask = 0; + io_req->io_desc[i].cow_offset = -1; + } + + return io_req; +} + +static int ubd_submit_request(struct ubd *dev, struct request *req) +{ + int segs = 0; + struct io_thread_req *io_req; + int ret; + enum req_op op = req_op(req); + + if (op == REQ_OP_FLUSH) + segs = 0; + else if (op == REQ_OP_WRITE_ZEROES || op == REQ_OP_DISCARD) + segs = 1; + else + segs = blk_rq_nr_phys_segments(req); + + io_req = ubd_alloc_req(dev, req, segs); + if (!io_req) + return -ENOMEM; + + io_req->desc_cnt = segs; + if (segs) + ubd_map_req(dev, io_req, req); + + ret = os_write_file(thread_fd, &io_req, sizeof(io_req)); + if (ret != sizeof(io_req)) { + if (ret != -EAGAIN) + pr_err("write to io thread failed: %d\n", -ret); + kfree(io_req); + } + return ret; +} + +static blk_status_t ubd_queue_rq(struct blk_mq_hw_ctx *hctx, + const struct blk_mq_queue_data *bd) +{ + struct ubd *ubd_dev = hctx->queue->queuedata; + struct request *req = bd->rq; + int ret = 0, res = BLK_STS_OK; + + blk_mq_start_request(req); + + spin_lock_irq(&ubd_dev->lock); + + switch (req_op(req)) { + case REQ_OP_FLUSH: + case REQ_OP_READ: + case REQ_OP_WRITE: + case REQ_OP_DISCARD: + case REQ_OP_WRITE_ZEROES: + ret = ubd_submit_request(ubd_dev, req); + break; + default: + WARN_ON_ONCE(1); + res = BLK_STS_NOTSUPP; + } + + spin_unlock_irq(&ubd_dev->lock); + + if (ret < 0) { + if (ret == -ENOMEM) + res = BLK_STS_RESOURCE; + else + res = BLK_STS_DEV_RESOURCE; + } + + return res; +} + +static int ubd_getgeo(struct gendisk *disk, struct hd_geometry *geo) +{ + struct ubd *ubd_dev = disk->private_data; + + geo->heads = 128; + geo->sectors = 32; + geo->cylinders = ubd_dev->size / (128 * 32 * 512); + return 0; +} + +static int ubd_ioctl(struct block_device *bdev, blk_mode_t mode, + unsigned int cmd, unsigned long arg) +{ + struct ubd *ubd_dev = bdev->bd_disk->private_data; + u16 ubd_id[ATA_ID_WORDS]; + + switch (cmd) { + struct cdrom_volctrl volume; + case HDIO_GET_IDENTITY: + memset(&ubd_id, 0, ATA_ID_WORDS * 2); + ubd_id[ATA_ID_CYLS] = ubd_dev->size / (128 * 32 * 512); + ubd_id[ATA_ID_HEADS] = 128; + ubd_id[ATA_ID_SECTORS] = 32; + if(copy_to_user((char __user *) arg, (char *) &ubd_id, + sizeof(ubd_id))) + return -EFAULT; + return 0; + + case CDROMVOLREAD: + if(copy_from_user(&volume, (char __user *) arg, sizeof(volume))) + return -EFAULT; + volume.channel0 = 255; + volume.channel1 = 255; + volume.channel2 = 255; + volume.channel3 = 255; + if(copy_to_user((char __user *) arg, &volume, sizeof(volume))) + return -EFAULT; + return 0; + } + return -EINVAL; +} + +static int map_error(int error_code) +{ + switch (error_code) { + case 0: + return BLK_STS_OK; + case ENOSYS: + case EOPNOTSUPP: + return BLK_STS_NOTSUPP; + case ENOSPC: + return BLK_STS_NOSPC; + } + return BLK_STS_IOERR; +} + +/* + * Everything from here onwards *IS NOT PART OF THE KERNEL* + * + * The following functions are part of UML hypervisor code. + * All functions from here onwards are executed as a helper + * thread and are not allowed to execute any kernel functions. + * + * Any communication must occur strictly via shared memory and IPC. + * + * Do not add printks, locks, kernel memory operations, etc - it + * will result in unpredictable behaviour and/or crashes. + */ + +static int update_bitmap(struct io_thread_req *req, struct io_desc *segment) +{ + int n; + + if (segment->cow_offset == -1) + return map_error(0); + + n = os_pwrite_file(req->fds[1], &segment->bitmap_words, + sizeof(segment->bitmap_words), segment->cow_offset); + if (n != sizeof(segment->bitmap_words)) + return map_error(-n); + + return map_error(0); +} + +static void do_io(struct io_thread_req *req, struct io_desc *desc) +{ + char *buf = NULL; + unsigned long len; + int n, nsectors, start, end, bit; + __u64 off; + + /* FLUSH is really a special case, we cannot "case" it with others */ + + if (req_op(req->req) == REQ_OP_FLUSH) { + /* fds[0] is always either the rw image or our cow file */ + req->error = map_error(-os_sync_file(req->fds[0])); + return; + } + + nsectors = desc->length / req->sectorsize; + start = 0; + do { + bit = ubd_test_bit(start, (unsigned char *) &desc->sector_mask); + end = start; + while((end < nsectors) && + (ubd_test_bit(end, (unsigned char *) &desc->sector_mask) == bit)) + end++; + + off = req->offset + req->offsets[bit] + + start * req->sectorsize; + len = (end - start) * req->sectorsize; + if (desc->buffer != NULL) + buf = &desc->buffer[start * req->sectorsize]; + + switch (req_op(req->req)) { + case REQ_OP_READ: + n = 0; + do { + buf = &buf[n]; + len -= n; + n = os_pread_file(req->fds[bit], buf, len, off); + if (n < 0) { + req->error = map_error(-n); + return; + } + } while((n < len) && (n != 0)); + if (n < len) memset(&buf[n], 0, len - n); + break; + case REQ_OP_WRITE: + n = os_pwrite_file(req->fds[bit], buf, len, off); + if(n != len){ + req->error = map_error(-n); + return; + } + break; + case REQ_OP_DISCARD: + n = os_falloc_punch(req->fds[bit], off, len); + if (n) { + req->error = map_error(-n); + return; + } + break; + case REQ_OP_WRITE_ZEROES: + n = os_falloc_zeroes(req->fds[bit], off, len); + if (n) { + req->error = map_error(-n); + return; + } + break; + default: + WARN_ON_ONCE(1); + req->error = BLK_STS_NOTSUPP; + return; + } + + start = end; + } while(start < nsectors); + + req->offset += len; + req->error = update_bitmap(req, desc); +} + +/* Changed in start_io_thread, which is serialized by being called only + * from ubd_init, which is an initcall. + */ +int kernel_fd = -1; + +/* Only changed by the io thread. XXX: currently unused. */ +static int io_count; + +void *io_thread(void *arg) +{ + int n, count, written, res; + + os_fix_helper_thread_signals(); + + while(1){ + n = bulk_req_safe_read( + kernel_fd, + io_req_buffer, + &io_remainder, + &io_remainder_size, + UBD_REQ_BUFFER_SIZE + ); + if (n <= 0) { + if (n == -EAGAIN) + ubd_read_poll(-1); + + continue; + } + + for (count = 0; count < n/sizeof(struct io_thread_req *); count++) { + struct io_thread_req *req = io_req_buffer[count]; + int i; + + io_count++; + for (i = 0; !req->error && i < req->desc_cnt; i++) + do_io(req, &(req->io_desc[i])); + + } + + written = 0; + + do { + res = os_write_file(kernel_fd, + ((char *) io_req_buffer) + written, + n - written); + if (res >= 0) { + written += res; + } + if (written < n) { + ubd_write_poll(-1); + } + } while (written < n); + } + + return NULL; +} diff --git a/arch/um/drivers/ubd_user.c b/arch/um/drivers/ubd_user.c new file mode 100644 index 000000000..8e8a8bf51 --- /dev/null +++ b/arch/um/drivers/ubd_user.c @@ -0,0 +1,78 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2016 Anton Ivanov (aivanov@brocade.com) + * Copyright (C) 2000, 2001, 2002 Jeff Dike (jdike@karaya.com) + * Copyright (C) 2001 Ridgerun,Inc (glonnon@ridgerun.com) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "ubd.h" +#include +#include + +static struct pollfd kernel_pollfd; + +int start_io_thread(struct os_helper_thread **td_out, int *fd_out) +{ + int fds[2], err; + + err = os_pipe(fds, 1, 1); + if(err < 0){ + printk("start_io_thread - os_pipe failed, err = %d\n", -err); + goto out; + } + + kernel_fd = fds[0]; + kernel_pollfd.fd = kernel_fd; + kernel_pollfd.events = POLLIN; + *fd_out = fds[1]; + + err = os_set_fd_block(*fd_out, 0); + err |= os_set_fd_block(kernel_fd, 0); + if (err) { + printk("start_io_thread - failed to set nonblocking I/O.\n"); + goto out_close; + } + + err = os_run_helper_thread(td_out, io_thread, NULL); + if (err < 0) { + printk("%s - failed to run helper thread, err = %d\n", + __func__, -err); + goto out_close; + } + + return 0; + + out_close: + os_close_file(fds[0]); + os_close_file(fds[1]); + kernel_fd = -1; + *fd_out = -1; + out: + return err; +} + +int ubd_read_poll(int timeout) +{ + kernel_pollfd.events = POLLIN; + return poll(&kernel_pollfd, 1, timeout); +} +int ubd_write_poll(int timeout) +{ + kernel_pollfd.events = POLLOUT; + return poll(&kernel_pollfd, 1, timeout); +} + diff --git a/arch/um/drivers/vector_kern.c b/arch/um/drivers/vector_kern.c new file mode 100644 index 000000000..8a70b3a62 --- /dev/null +++ b/arch/um/drivers/vector_kern.c @@ -0,0 +1,1769 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2017 - 2019 Cambridge Greys Limited + * Copyright (C) 2011 - 2014 Cisco Systems Inc + * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) + * Copyright (C) 2001 Lennert Buytenhek (buytenh@gnu.org) and + * James Leu (jleu@mindspring.net). + * Copyright (C) 2001 by various other people who didn't put their name here. + */ + +#define pr_fmt(fmt) "uml-vector: " fmt + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "mconsole_kern.h" +#include "vector_user.h" +#include "vector_kern.h" + +/* + * Adapted from network devices with the following major changes: + * All transports are static - simplifies the code significantly + * Multiple FDs/IRQs per device + * Vector IO optionally used for read/write, falling back to legacy + * based on configuration and/or availability + * Configuration is no longer positional - L2TPv3 and GRE require up to + * 10 parameters, passing this as positional is not fit for purpose. + * Only socket transports are supported + */ + + +#define DRIVER_NAME "uml-vector" +struct vector_cmd_line_arg { + struct list_head list; + int unit; + char *arguments; +}; + +struct vector_device { + struct list_head list; + struct net_device *dev; + struct platform_device pdev; + int unit; + int opened; +}; + +static LIST_HEAD(vec_cmd_line); + +static DEFINE_SPINLOCK(vector_devices_lock); +static LIST_HEAD(vector_devices); + +static int driver_registered; + +static void vector_eth_configure(int n, struct arglist *def); +static int vector_mmsg_rx(struct vector_private *vp, int budget); + +/* Argument accessors to set variables (and/or set default values) + * mtu, buffer sizing, default headroom, etc + */ + +#define DEFAULT_HEADROOM 2 +#define SAFETY_MARGIN 32 +#define DEFAULT_VECTOR_SIZE 64 +#define TX_SMALL_PACKET 128 +#define MAX_IOV_SIZE (MAX_SKB_FRAGS + 1) + +static const struct { + const char string[ETH_GSTRING_LEN]; +} ethtool_stats_keys[] = { + { "rx_queue_max" }, + { "rx_queue_running_average" }, + { "tx_queue_max" }, + { "tx_queue_running_average" }, + { "rx_encaps_errors" }, + { "tx_timeout_count" }, + { "tx_restart_queue" }, + { "tx_kicks" }, + { "tx_flow_control_xon" }, + { "tx_flow_control_xoff" }, + { "rx_csum_offload_good" }, + { "rx_csum_offload_errors"}, + { "sg_ok"}, + { "sg_linearized"}, +}; + +#define VECTOR_NUM_STATS ARRAY_SIZE(ethtool_stats_keys) + +static void vector_reset_stats(struct vector_private *vp) +{ + /* We reuse the existing queue locks for stats */ + + /* RX stats are modified with RX head_lock held + * in vector_poll. + */ + + spin_lock(&vp->rx_queue->head_lock); + vp->estats.rx_queue_max = 0; + vp->estats.rx_queue_running_average = 0; + vp->estats.rx_encaps_errors = 0; + vp->estats.sg_ok = 0; + vp->estats.sg_linearized = 0; + spin_unlock(&vp->rx_queue->head_lock); + + /* TX stats are modified with TX head_lock held + * in vector_send. + */ + + spin_lock(&vp->tx_queue->head_lock); + vp->estats.tx_timeout_count = 0; + vp->estats.tx_restart_queue = 0; + vp->estats.tx_kicks = 0; + vp->estats.tx_flow_control_xon = 0; + vp->estats.tx_flow_control_xoff = 0; + vp->estats.tx_queue_max = 0; + vp->estats.tx_queue_running_average = 0; + spin_unlock(&vp->tx_queue->head_lock); +} + +static int get_mtu(struct arglist *def) +{ + char *mtu = uml_vector_fetch_arg(def, "mtu"); + long result; + + if (mtu != NULL) { + if (kstrtoul(mtu, 10, &result) == 0) + if ((result < (1 << 16) - 1) && (result >= 576)) + return result; + } + return ETH_MAX_PACKET; +} + +static char *get_bpf_file(struct arglist *def) +{ + return uml_vector_fetch_arg(def, "bpffile"); +} + +static bool get_bpf_flash(struct arglist *def) +{ + char *allow = uml_vector_fetch_arg(def, "bpfflash"); + long result; + + if (allow != NULL) { + if (kstrtoul(allow, 10, &result) == 0) + return result > 0; + } + return false; +} + +static int get_depth(struct arglist *def) +{ + char *mtu = uml_vector_fetch_arg(def, "depth"); + long result; + + if (mtu != NULL) { + if (kstrtoul(mtu, 10, &result) == 0) + return result; + } + return DEFAULT_VECTOR_SIZE; +} + +static int get_headroom(struct arglist *def) +{ + char *mtu = uml_vector_fetch_arg(def, "headroom"); + long result; + + if (mtu != NULL) { + if (kstrtoul(mtu, 10, &result) == 0) + return result; + } + return DEFAULT_HEADROOM; +} + +static int get_req_size(struct arglist *def) +{ + char *gro = uml_vector_fetch_arg(def, "gro"); + long result; + + if (gro != NULL) { + if (kstrtoul(gro, 10, &result) == 0) { + if (result > 0) + return 65536; + } + } + return get_mtu(def) + ETH_HEADER_OTHER + + get_headroom(def) + SAFETY_MARGIN; +} + + +static int get_transport_options(struct arglist *def) +{ + char *transport = uml_vector_fetch_arg(def, "transport"); + char *vector = uml_vector_fetch_arg(def, "vec"); + + int vec_rx = VECTOR_RX; + int vec_tx = VECTOR_TX; + long parsed; + int result = 0; + + if (transport == NULL) + return -EINVAL; + + if (vector != NULL) { + if (kstrtoul(vector, 10, &parsed) == 0) { + if (parsed == 0) { + vec_rx = 0; + vec_tx = 0; + } + } + } + + if (get_bpf_flash(def)) + result = VECTOR_BPF_FLASH; + + if (strncmp(transport, TRANS_TAP, TRANS_TAP_LEN) == 0) + return result; + if (strncmp(transport, TRANS_HYBRID, TRANS_HYBRID_LEN) == 0) + return (result | vec_rx | VECTOR_BPF); + if (strncmp(transport, TRANS_RAW, TRANS_RAW_LEN) == 0) + return (result | vec_rx | vec_tx | VECTOR_QDISC_BYPASS); + return (result | vec_rx | vec_tx); +} + + +/* A mini-buffer for packet drop read + * All of our supported transports are datagram oriented and we always + * read using recvmsg or recvmmsg. If we pass a buffer which is smaller + * than the packet size it still counts as full packet read and will + * clean the incoming stream to keep sigio/epoll happy + */ + +#define DROP_BUFFER_SIZE 32 + +static char *drop_buffer; + + +/* + * Advance the mmsg queue head by n = advance. Resets the queue to + * maximum enqueue/dequeue-at-once capacity if possible. Called by + * dequeuers. Caller must hold the head_lock! + */ + +static int vector_advancehead(struct vector_queue *qi, int advance) +{ + qi->head = + (qi->head + advance) + % qi->max_depth; + + + atomic_sub(advance, &qi->queue_depth); + return atomic_read(&qi->queue_depth); +} + +/* Advance the queue tail by n = advance. + * This is called by enqueuers which should hold the + * head lock already + */ + +static int vector_advancetail(struct vector_queue *qi, int advance) +{ + qi->tail = + (qi->tail + advance) + % qi->max_depth; + atomic_add(advance, &qi->queue_depth); + return atomic_read(&qi->queue_depth); +} + +static int prep_msg(struct vector_private *vp, + struct sk_buff *skb, + struct iovec *iov) +{ + int iov_index = 0; + int nr_frags, frag; + skb_frag_t *skb_frag; + + nr_frags = skb_shinfo(skb)->nr_frags; + if (nr_frags > MAX_IOV_SIZE) { + if (skb_linearize(skb) != 0) + goto drop; + } + if (vp->header_size > 0) { + iov[iov_index].iov_len = vp->header_size; + vp->form_header(iov[iov_index].iov_base, skb, vp); + iov_index++; + } + iov[iov_index].iov_base = skb->data; + if (nr_frags > 0) { + iov[iov_index].iov_len = skb->len - skb->data_len; + vp->estats.sg_ok++; + } else + iov[iov_index].iov_len = skb->len; + iov_index++; + for (frag = 0; frag < nr_frags; frag++) { + skb_frag = &skb_shinfo(skb)->frags[frag]; + iov[iov_index].iov_base = skb_frag_address_safe(skb_frag); + iov[iov_index].iov_len = skb_frag_size(skb_frag); + iov_index++; + } + return iov_index; +drop: + return -1; +} +/* + * Generic vector enqueue with support for forming headers using transport + * specific callback. Allows GRE, L2TPv3, RAW and other transports + * to use a common enqueue procedure in vector mode + */ + +static int vector_enqueue(struct vector_queue *qi, struct sk_buff *skb) +{ + struct vector_private *vp = netdev_priv(qi->dev); + int queue_depth; + int packet_len; + struct mmsghdr *mmsg_vector = qi->mmsg_vector; + int iov_count; + + spin_lock(&qi->tail_lock); + queue_depth = atomic_read(&qi->queue_depth); + + if (skb) + packet_len = skb->len; + + if (queue_depth < qi->max_depth) { + + *(qi->skbuff_vector + qi->tail) = skb; + mmsg_vector += qi->tail; + iov_count = prep_msg( + vp, + skb, + mmsg_vector->msg_hdr.msg_iov + ); + if (iov_count < 1) + goto drop; + mmsg_vector->msg_hdr.msg_iovlen = iov_count; + mmsg_vector->msg_hdr.msg_name = vp->fds->remote_addr; + mmsg_vector->msg_hdr.msg_namelen = vp->fds->remote_addr_size; + wmb(); /* Make the packet visible to the NAPI poll thread */ + queue_depth = vector_advancetail(qi, 1); + } else + goto drop; + spin_unlock(&qi->tail_lock); + return queue_depth; +drop: + qi->dev->stats.tx_dropped++; + if (skb != NULL) { + packet_len = skb->len; + dev_consume_skb_any(skb); + netdev_completed_queue(qi->dev, 1, packet_len); + } + spin_unlock(&qi->tail_lock); + return queue_depth; +} + +static int consume_vector_skbs(struct vector_queue *qi, int count) +{ + struct sk_buff *skb; + int skb_index; + int bytes_compl = 0; + + for (skb_index = qi->head; skb_index < qi->head + count; skb_index++) { + skb = *(qi->skbuff_vector + skb_index); + /* mark as empty to ensure correct destruction if + * needed + */ + bytes_compl += skb->len; + *(qi->skbuff_vector + skb_index) = NULL; + dev_consume_skb_any(skb); + } + qi->dev->stats.tx_bytes += bytes_compl; + qi->dev->stats.tx_packets += count; + netdev_completed_queue(qi->dev, count, bytes_compl); + return vector_advancehead(qi, count); +} + +/* + * Generic vector dequeue via sendmmsg with support for forming headers + * using transport specific callback. Allows GRE, L2TPv3, RAW and + * other transports to use a common dequeue procedure in vector mode + */ + + +static int vector_send(struct vector_queue *qi) +{ + struct vector_private *vp = netdev_priv(qi->dev); + struct mmsghdr *send_from; + int result = 0, send_len; + + if (spin_trylock(&qi->head_lock)) { + /* update queue_depth to current value */ + while (atomic_read(&qi->queue_depth) > 0) { + /* Calculate the start of the vector */ + send_len = atomic_read(&qi->queue_depth); + send_from = qi->mmsg_vector; + send_from += qi->head; + /* Adjust vector size if wraparound */ + if (send_len + qi->head > qi->max_depth) + send_len = qi->max_depth - qi->head; + /* Try to TX as many packets as possible */ + if (send_len > 0) { + result = uml_vector_sendmmsg( + vp->fds->tx_fd, + send_from, + send_len, + 0 + ); + vp->in_write_poll = + (result != send_len); + } + /* For some of the sendmmsg error scenarios + * we may end being unsure in the TX success + * for all packets. It is safer to declare + * them all TX-ed and blame the network. + */ + if (result < 0) { + if (net_ratelimit()) + netdev_err(vp->dev, "sendmmsg err=%i\n", + result); + vp->in_error = true; + result = send_len; + } + if (result > 0) { + consume_vector_skbs(qi, result); + /* This is equivalent to an TX IRQ. + * Restart the upper layers to feed us + * more packets. + */ + if (result > vp->estats.tx_queue_max) + vp->estats.tx_queue_max = result; + vp->estats.tx_queue_running_average = + (vp->estats.tx_queue_running_average + result) >> 1; + } + netif_wake_queue(qi->dev); + /* if TX is busy, break out of the send loop, + * poll write IRQ will reschedule xmit for us. + */ + if (result != send_len) { + vp->estats.tx_restart_queue++; + break; + } + } + spin_unlock(&qi->head_lock); + } + return atomic_read(&qi->queue_depth); +} + +/* Queue destructor. Deliberately stateless so we can use + * it in queue cleanup if initialization fails. + */ + +static void destroy_queue(struct vector_queue *qi) +{ + int i; + struct iovec *iov; + struct vector_private *vp = netdev_priv(qi->dev); + struct mmsghdr *mmsg_vector; + + if (qi == NULL) + return; + /* deallocate any skbuffs - we rely on any unused to be + * set to NULL. + */ + if (qi->skbuff_vector != NULL) { + for (i = 0; i < qi->max_depth; i++) { + if (*(qi->skbuff_vector + i) != NULL) + dev_kfree_skb_any(*(qi->skbuff_vector + i)); + } + kfree(qi->skbuff_vector); + } + /* deallocate matching IOV structures including header buffs */ + if (qi->mmsg_vector != NULL) { + mmsg_vector = qi->mmsg_vector; + for (i = 0; i < qi->max_depth; i++) { + iov = mmsg_vector->msg_hdr.msg_iov; + if (iov != NULL) { + if ((vp->header_size > 0) && + (iov->iov_base != NULL)) + kfree(iov->iov_base); + kfree(iov); + } + mmsg_vector++; + } + kfree(qi->mmsg_vector); + } + kfree(qi); +} + +/* + * Queue constructor. Create a queue with a given side. + */ +static struct vector_queue *create_queue( + struct vector_private *vp, + int max_size, + int header_size, + int num_extra_frags) +{ + struct vector_queue *result; + int i; + struct iovec *iov; + struct mmsghdr *mmsg_vector; + + result = kmalloc_obj(struct vector_queue); + if (result == NULL) + return NULL; + result->max_depth = max_size; + result->dev = vp->dev; + result->mmsg_vector = kmalloc( + (sizeof(struct mmsghdr) * max_size), GFP_KERNEL); + if (result->mmsg_vector == NULL) + goto out_mmsg_fail; + result->skbuff_vector = kmalloc( + (sizeof(void *) * max_size), GFP_KERNEL); + if (result->skbuff_vector == NULL) + goto out_skb_fail; + + /* further failures can be handled safely by destroy_queue*/ + + mmsg_vector = result->mmsg_vector; + for (i = 0; i < max_size; i++) { + /* Clear all pointers - we use non-NULL as marking on + * what to free on destruction + */ + *(result->skbuff_vector + i) = NULL; + mmsg_vector->msg_hdr.msg_iov = NULL; + mmsg_vector++; + } + mmsg_vector = result->mmsg_vector; + result->max_iov_frags = num_extra_frags; + for (i = 0; i < max_size; i++) { + if (vp->header_size > 0) + iov = kmalloc_objs(struct iovec, 3 + num_extra_frags); + else + iov = kmalloc_objs(struct iovec, 2 + num_extra_frags); + if (iov == NULL) + goto out_fail; + mmsg_vector->msg_hdr.msg_iov = iov; + mmsg_vector->msg_hdr.msg_iovlen = 1; + mmsg_vector->msg_hdr.msg_control = NULL; + mmsg_vector->msg_hdr.msg_controllen = 0; + mmsg_vector->msg_hdr.msg_flags = MSG_DONTWAIT; + mmsg_vector->msg_hdr.msg_name = NULL; + mmsg_vector->msg_hdr.msg_namelen = 0; + if (vp->header_size > 0) { + iov->iov_base = kmalloc(header_size, GFP_KERNEL); + if (iov->iov_base == NULL) + goto out_fail; + iov->iov_len = header_size; + mmsg_vector->msg_hdr.msg_iovlen = 2; + iov++; + } + iov->iov_base = NULL; + iov->iov_len = 0; + mmsg_vector++; + } + spin_lock_init(&result->head_lock); + spin_lock_init(&result->tail_lock); + atomic_set(&result->queue_depth, 0); + result->head = 0; + result->tail = 0; + return result; +out_skb_fail: + kfree(result->mmsg_vector); +out_mmsg_fail: + kfree(result); + return NULL; +out_fail: + destroy_queue(result); + return NULL; +} + +/* + * We do not use the RX queue as a proper wraparound queue for now + * This is not necessary because the consumption via napi_gro_receive() + * happens in-line. While we can try using the return code of + * netif_rx() for flow control there are no drivers doing this today. + * For this RX specific use we ignore the tail/head locks and + * just read into a prepared queue filled with skbuffs. + */ + +static struct sk_buff *prep_skb( + struct vector_private *vp, + struct user_msghdr *msg) +{ + int linear = vp->max_packet + vp->headroom + SAFETY_MARGIN; + struct sk_buff *result; + int iov_index = 0, len; + struct iovec *iov = msg->msg_iov; + int err, nr_frags, frag; + skb_frag_t *skb_frag; + + if (vp->req_size <= linear) + len = linear; + else + len = vp->req_size; + result = alloc_skb_with_frags( + linear, + len - vp->max_packet, + 3, + &err, + GFP_ATOMIC + ); + if (vp->header_size > 0) + iov_index++; + if (result == NULL) { + iov[iov_index].iov_base = NULL; + iov[iov_index].iov_len = 0; + goto done; + } + skb_reserve(result, vp->headroom); + result->dev = vp->dev; + skb_put(result, vp->max_packet); + result->data_len = len - vp->max_packet; + result->len += len - vp->max_packet; + skb_reset_mac_header(result); + result->ip_summed = CHECKSUM_NONE; + iov[iov_index].iov_base = result->data; + iov[iov_index].iov_len = vp->max_packet; + iov_index++; + + nr_frags = skb_shinfo(result)->nr_frags; + for (frag = 0; frag < nr_frags; frag++) { + skb_frag = &skb_shinfo(result)->frags[frag]; + iov[iov_index].iov_base = skb_frag_address_safe(skb_frag); + if (iov[iov_index].iov_base != NULL) + iov[iov_index].iov_len = skb_frag_size(skb_frag); + else + iov[iov_index].iov_len = 0; + iov_index++; + } +done: + msg->msg_iovlen = iov_index; + return result; +} + + +/* Prepare queue for recvmmsg one-shot rx - fill with fresh sk_buffs */ + +static void prep_queue_for_rx(struct vector_queue *qi) +{ + struct vector_private *vp = netdev_priv(qi->dev); + struct mmsghdr *mmsg_vector = qi->mmsg_vector; + void **skbuff_vector = qi->skbuff_vector; + int i, queue_depth; + + queue_depth = atomic_read(&qi->queue_depth); + + if (queue_depth == 0) + return; + + /* RX is always emptied 100% during each cycle, so we do not + * have to do the tail wraparound math for it. + */ + + qi->head = qi->tail = 0; + + for (i = 0; i < queue_depth; i++) { + /* it is OK if allocation fails - recvmmsg with NULL data in + * iov argument still performs an RX, just drops the packet + * This allows us stop faffing around with a "drop buffer" + */ + + *skbuff_vector = prep_skb(vp, &mmsg_vector->msg_hdr); + skbuff_vector++; + mmsg_vector++; + } + atomic_set(&qi->queue_depth, 0); +} + +static struct vector_device *find_device(int n) +{ + struct vector_device *device; + struct list_head *ele; + + spin_lock(&vector_devices_lock); + list_for_each(ele, &vector_devices) { + device = list_entry(ele, struct vector_device, list); + if (device->unit == n) + goto out; + } + device = NULL; + out: + spin_unlock(&vector_devices_lock); + return device; +} + +static int vector_parse(char *str, int *index_out, char **str_out, + char **error_out) +{ + int n, err; + char *start = str; + + while ((*str != ':') && (strlen(str) > 1)) + str++; + if (*str != ':') { + *error_out = "Expected ':' after device number"; + return -EINVAL; + } + *str = '\0'; + + err = kstrtouint(start, 0, &n); + if (err < 0) { + *error_out = "Bad device number"; + return err; + } + + str++; + if (find_device(n)) { + *error_out = "Device already configured"; + return -EINVAL; + } + + *index_out = n; + *str_out = str; + return 0; +} + +static int vector_config(char *str, char **error_out) +{ + int err, n; + char *params; + struct arglist *parsed; + + err = vector_parse(str, &n, ¶ms, error_out); + if (err != 0) + return err; + + /* This string is broken up and the pieces used by the underlying + * driver. We should copy it to make sure things do not go wrong + * later. + */ + + params = kstrdup(params, GFP_KERNEL); + if (params == NULL) { + *error_out = "vector_config failed to strdup string"; + return -ENOMEM; + } + + parsed = uml_parse_vector_ifspec(params); + + if (parsed == NULL) { + *error_out = "vector_config failed to parse parameters"; + kfree(params); + return -EINVAL; + } + + vector_eth_configure(n, parsed); + return 0; +} + +static int vector_id(char **str, int *start_out, int *end_out) +{ + char *end; + int n; + + n = simple_strtoul(*str, &end, 0); + if ((*end != '\0') || (end == *str)) + return -1; + + *start_out = n; + *end_out = n; + *str = end; + return n; +} + +static int vector_remove(int n, char **error_out) +{ + struct vector_device *vec_d; + struct net_device *dev; + struct vector_private *vp; + + vec_d = find_device(n); + if (vec_d == NULL) + return -ENODEV; + dev = vec_d->dev; + vp = netdev_priv(dev); + if (vp->fds != NULL) + return -EBUSY; + unregister_netdev(dev); + platform_device_unregister(&vec_d->pdev); + return 0; +} + +/* + * There is no shared per-transport initialization code, so + * we will just initialize each interface one by one and + * add them to a list + */ + +static struct platform_driver uml_net_driver = { + .driver = { + .name = DRIVER_NAME, + }, +}; + + +static void vector_device_release(struct device *dev) +{ + struct vector_device *device = + container_of(dev, struct vector_device, pdev.dev); + struct net_device *netdev = device->dev; + + list_del(&device->list); + kfree(device); + free_netdev(netdev); +} + +/* Bog standard recv using recvmsg - not used normally unless the user + * explicitly specifies not to use recvmmsg vector RX. + */ + +static int vector_legacy_rx(struct vector_private *vp) +{ + int pkt_len; + struct user_msghdr hdr; + struct iovec iov[2 + MAX_IOV_SIZE]; /* header + data use case only */ + int iovpos = 0; + struct sk_buff *skb; + int header_check; + + hdr.msg_name = NULL; + hdr.msg_namelen = 0; + hdr.msg_iov = (struct iovec *) &iov; + hdr.msg_control = NULL; + hdr.msg_controllen = 0; + hdr.msg_flags = 0; + + if (vp->header_size > 0) { + iov[0].iov_base = vp->header_rxbuffer; + iov[0].iov_len = vp->header_size; + } + + skb = prep_skb(vp, &hdr); + + if (skb == NULL) { + /* Read a packet into drop_buffer and don't do + * anything with it. + */ + iov[iovpos].iov_base = drop_buffer; + iov[iovpos].iov_len = DROP_BUFFER_SIZE; + hdr.msg_iovlen = 1; + vp->dev->stats.rx_dropped++; + } + + pkt_len = uml_vector_recvmsg(vp->fds->rx_fd, &hdr, 0); + if (pkt_len < 0) { + vp->in_error = true; + return pkt_len; + } + + if (skb != NULL) { + if (pkt_len > vp->header_size) { + if (vp->header_size > 0) { + header_check = vp->verify_header( + vp->header_rxbuffer, skb, vp); + if (header_check < 0) { + dev_kfree_skb_irq(skb); + vp->dev->stats.rx_dropped++; + vp->estats.rx_encaps_errors++; + return 0; + } + if (header_check > 0) { + vp->estats.rx_csum_offload_good++; + skb->ip_summed = CHECKSUM_UNNECESSARY; + } + } + pskb_trim(skb, pkt_len - vp->rx_header_size); + skb->protocol = eth_type_trans(skb, skb->dev); + vp->dev->stats.rx_bytes += skb->len; + vp->dev->stats.rx_packets++; + napi_gro_receive(&vp->napi, skb); + } else { + dev_kfree_skb_irq(skb); + } + } + return pkt_len; +} + +/* + * Packet at a time TX which falls back to vector TX if the + * underlying transport is busy. + */ + + + +static int writev_tx(struct vector_private *vp, struct sk_buff *skb) +{ + struct iovec iov[3 + MAX_IOV_SIZE]; + int iov_count, pkt_len = 0; + + iov[0].iov_base = vp->header_txbuffer; + iov_count = prep_msg(vp, skb, (struct iovec *) &iov); + + if (iov_count < 1) + goto drop; + + pkt_len = uml_vector_writev( + vp->fds->tx_fd, + (struct iovec *) &iov, + iov_count + ); + + if (pkt_len < 0) + goto drop; + + netif_trans_update(vp->dev); + netif_wake_queue(vp->dev); + + if (pkt_len > 0) { + vp->dev->stats.tx_bytes += skb->len; + vp->dev->stats.tx_packets++; + } else { + vp->dev->stats.tx_dropped++; + } + consume_skb(skb); + return pkt_len; +drop: + vp->dev->stats.tx_dropped++; + consume_skb(skb); + if (pkt_len < 0) + vp->in_error = true; + return pkt_len; +} + +/* + * Receive as many messages as we can in one call using the special + * mmsg vector matched to an skb vector which we prepared earlier. + */ + +static int vector_mmsg_rx(struct vector_private *vp, int budget) +{ + int packet_count, i; + struct vector_queue *qi = vp->rx_queue; + struct sk_buff *skb; + struct mmsghdr *mmsg_vector = qi->mmsg_vector; + void **skbuff_vector = qi->skbuff_vector; + int header_check; + + /* Refresh the vector and make sure it is with new skbs and the + * iovs are updated to point to them. + */ + + prep_queue_for_rx(qi); + + /* Fire the Lazy Gun - get as many packets as we can in one go. */ + + if (budget > qi->max_depth) + budget = qi->max_depth; + + packet_count = uml_vector_recvmmsg( + vp->fds->rx_fd, qi->mmsg_vector, budget, 0); + + if (packet_count < 0) + vp->in_error = true; + + if (packet_count <= 0) + return packet_count; + + /* We treat packet processing as enqueue, buffer refresh as dequeue + * The queue_depth tells us how many buffers have been used and how + * many do we need to prep the next time prep_queue_for_rx() is called. + */ + + atomic_add(packet_count, &qi->queue_depth); + + for (i = 0; i < packet_count; i++) { + skb = (*skbuff_vector); + if (mmsg_vector->msg_len > vp->header_size) { + if (vp->header_size > 0) { + header_check = vp->verify_header( + mmsg_vector->msg_hdr.msg_iov->iov_base, + skb, + vp + ); + if (header_check < 0) { + /* Overlay header failed to verify - discard. + * We can actually keep this skb and reuse it, + * but that will make the prep logic too + * complex. + */ + dev_kfree_skb_irq(skb); + vp->estats.rx_encaps_errors++; + (*skbuff_vector) = NULL; + mmsg_vector++; + skbuff_vector++; + continue; + } + if (header_check > 0) { + vp->estats.rx_csum_offload_good++; + skb->ip_summed = CHECKSUM_UNNECESSARY; + } + } + pskb_trim(skb, + mmsg_vector->msg_len - vp->rx_header_size); + skb->protocol = eth_type_trans(skb, skb->dev); + /* + * We do not need to lock on updating stats here + * The interrupt loop is non-reentrant. + */ + vp->dev->stats.rx_bytes += skb->len; + vp->dev->stats.rx_packets++; + napi_gro_receive(&vp->napi, skb); + } else { + /* Overlay header too short to do anything - discard. + * We can actually keep this skb and reuse it, + * but that will make the prep logic too complex. + */ + if (skb != NULL) + dev_kfree_skb_irq(skb); + } + (*skbuff_vector) = NULL; + /* Move to the next buffer element */ + mmsg_vector++; + skbuff_vector++; + } + if (packet_count > 0) { + if (vp->estats.rx_queue_max < packet_count) + vp->estats.rx_queue_max = packet_count; + vp->estats.rx_queue_running_average = + (vp->estats.rx_queue_running_average + packet_count) >> 1; + } + return packet_count; +} + +static int vector_net_start_xmit(struct sk_buff *skb, struct net_device *dev) +{ + struct vector_private *vp = netdev_priv(dev); + int queue_depth = 0; + + if (vp->in_error) { + deactivate_fd(vp->fds->rx_fd, vp->rx_irq); + if ((vp->fds->rx_fd != vp->fds->tx_fd) && (vp->tx_irq != 0)) + deactivate_fd(vp->fds->tx_fd, vp->tx_irq); + return NETDEV_TX_BUSY; + } + + if ((vp->options & VECTOR_TX) == 0) { + writev_tx(vp, skb); + return NETDEV_TX_OK; + } + + /* We do BQL only in the vector path, no point doing it in + * packet at a time mode as there is no device queue + */ + + netdev_sent_queue(vp->dev, skb->len); + queue_depth = vector_enqueue(vp->tx_queue, skb); + + if (queue_depth < vp->tx_queue->max_depth && netdev_xmit_more()) { + mod_timer(&vp->tl, vp->coalesce); + return NETDEV_TX_OK; + } else { + queue_depth = vector_send(vp->tx_queue); + if (queue_depth > 0) + napi_schedule(&vp->napi); + } + + return NETDEV_TX_OK; +} + +static irqreturn_t vector_rx_interrupt(int irq, void *dev_id) +{ + struct net_device *dev = dev_id; + struct vector_private *vp = netdev_priv(dev); + + if (!netif_running(dev)) + return IRQ_NONE; + napi_schedule(&vp->napi); + return IRQ_HANDLED; + +} + +static irqreturn_t vector_tx_interrupt(int irq, void *dev_id) +{ + struct net_device *dev = dev_id; + struct vector_private *vp = netdev_priv(dev); + + if (!netif_running(dev)) + return IRQ_NONE; + /* We need to pay attention to it only if we got + * -EAGAIN or -ENOBUFFS from sendmmsg. Otherwise + * we ignore it. In the future, it may be worth + * it to improve the IRQ controller a bit to make + * tweaking the IRQ mask less costly + */ + + napi_schedule(&vp->napi); + return IRQ_HANDLED; + +} + +static int irq_rr; + +static int vector_net_close(struct net_device *dev) +{ + struct vector_private *vp = netdev_priv(dev); + + netif_stop_queue(dev); + timer_delete(&vp->tl); + + vp->opened = false; + + if (vp->fds == NULL) + return 0; + + /* Disable and free all IRQS */ + if (vp->rx_irq > 0) { + um_free_irq(vp->rx_irq, dev); + vp->rx_irq = 0; + } + if (vp->tx_irq > 0) { + um_free_irq(vp->tx_irq, dev); + vp->tx_irq = 0; + } + napi_disable(&vp->napi); + netif_napi_del(&vp->napi); + if (vp->fds->rx_fd > 0) { + if (vp->bpf) + uml_vector_detach_bpf(vp->fds->rx_fd, vp->bpf); + os_close_file(vp->fds->rx_fd); + vp->fds->rx_fd = -1; + } + if (vp->fds->tx_fd > 0) { + os_close_file(vp->fds->tx_fd); + vp->fds->tx_fd = -1; + } + if (vp->bpf != NULL) + kfree(vp->bpf->filter); + kfree(vp->bpf); + vp->bpf = NULL; + kfree(vp->fds->remote_addr); + kfree(vp->transport_data); + kfree(vp->header_rxbuffer); + kfree(vp->header_txbuffer); + if (vp->rx_queue != NULL) + destroy_queue(vp->rx_queue); + if (vp->tx_queue != NULL) + destroy_queue(vp->tx_queue); + kfree(vp->fds); + vp->fds = NULL; + vp->in_error = false; + return 0; +} + +static int vector_poll(struct napi_struct *napi, int budget) +{ + struct vector_private *vp = container_of(napi, struct vector_private, napi); + int work_done = 0; + int err; + bool tx_enqueued = false; + + if ((vp->options & VECTOR_TX) != 0) + tx_enqueued = (vector_send(vp->tx_queue) > 0); + spin_lock(&vp->rx_queue->head_lock); + if ((vp->options & VECTOR_RX) > 0) + err = vector_mmsg_rx(vp, budget); + else { + err = vector_legacy_rx(vp); + if (err > 0) + err = 1; + } + spin_unlock(&vp->rx_queue->head_lock); + if (err > 0) + work_done += err; + + if (tx_enqueued || err > 0) + napi_schedule(napi); + if (work_done <= budget) + napi_complete_done(napi, work_done); + return work_done; +} + +static void vector_reset_tx(struct work_struct *work) +{ + struct vector_private *vp = + container_of(work, struct vector_private, reset_tx); + netdev_reset_queue(vp->dev); + netif_start_queue(vp->dev); + netif_wake_queue(vp->dev); +} + +static int vector_net_open(struct net_device *dev) +{ + struct vector_private *vp = netdev_priv(dev); + int err = -EINVAL; + struct vector_device *vdevice; + + if (vp->opened) + return -ENXIO; + vp->opened = true; + + vp->bpf = uml_vector_user_bpf(get_bpf_file(vp->parsed)); + + vp->fds = uml_vector_user_open(vp->unit, vp->parsed); + + if (vp->fds == NULL) + goto out_close; + + if (build_transport_data(vp) < 0) + goto out_close; + + if ((vp->options & VECTOR_RX) > 0) { + vp->rx_queue = create_queue( + vp, + get_depth(vp->parsed), + vp->rx_header_size, + MAX_IOV_SIZE + ); + atomic_set(&vp->rx_queue->queue_depth, get_depth(vp->parsed)); + } else { + vp->header_rxbuffer = kmalloc( + vp->rx_header_size, + GFP_KERNEL + ); + if (vp->header_rxbuffer == NULL) + goto out_close; + } + if ((vp->options & VECTOR_TX) > 0) { + vp->tx_queue = create_queue( + vp, + get_depth(vp->parsed), + vp->header_size, + MAX_IOV_SIZE + ); + } else { + vp->header_txbuffer = kmalloc(vp->header_size, GFP_KERNEL); + if (vp->header_txbuffer == NULL) + goto out_close; + } + + netif_napi_add_weight(vp->dev, &vp->napi, vector_poll, + get_depth(vp->parsed)); + napi_enable(&vp->napi); + + /* READ IRQ */ + err = um_request_irq( + irq_rr + VECTOR_BASE_IRQ, vp->fds->rx_fd, + IRQ_READ, vector_rx_interrupt, + IRQF_SHARED, dev->name, dev); + if (err < 0) { + netdev_err(dev, "vector_open: failed to get rx irq(%d)\n", err); + err = -ENETUNREACH; + goto out_close; + } + vp->rx_irq = irq_rr + VECTOR_BASE_IRQ; + dev->irq = irq_rr + VECTOR_BASE_IRQ; + irq_rr = (irq_rr + 1) % VECTOR_IRQ_SPACE; + + /* WRITE IRQ - we need it only if we have vector TX */ + if ((vp->options & VECTOR_TX) > 0) { + err = um_request_irq( + irq_rr + VECTOR_BASE_IRQ, vp->fds->tx_fd, + IRQ_WRITE, vector_tx_interrupt, + IRQF_SHARED, dev->name, dev); + if (err < 0) { + netdev_err(dev, + "vector_open: failed to get tx irq(%d)\n", err); + err = -ENETUNREACH; + goto out_close; + } + vp->tx_irq = irq_rr + VECTOR_BASE_IRQ; + irq_rr = (irq_rr + 1) % VECTOR_IRQ_SPACE; + } + + if ((vp->options & VECTOR_QDISC_BYPASS) != 0) { + if (!uml_raw_enable_qdisc_bypass(vp->fds->rx_fd)) + vp->options |= VECTOR_BPF; + } + if (((vp->options & VECTOR_BPF) != 0) && (vp->bpf == NULL)) + vp->bpf = uml_vector_default_bpf(dev->dev_addr); + + if (vp->bpf != NULL) + uml_vector_attach_bpf(vp->fds->rx_fd, vp->bpf); + + netif_start_queue(dev); + vector_reset_stats(vp); + + /* clear buffer - it can happen that the host side of the interface + * is full when we get here. In this case, new data is never queued, + * SIGIOs never arrive, and the net never works. + */ + + napi_schedule(&vp->napi); + + vdevice = find_device(vp->unit); + vdevice->opened = 1; + + if ((vp->options & VECTOR_TX) != 0) + add_timer(&vp->tl); + return 0; +out_close: + vector_net_close(dev); + return err; +} + + +static void vector_net_set_multicast_list(struct net_device *dev) +{ + /* TODO: - we can do some BPF games here */ + return; +} + +static void vector_net_tx_timeout(struct net_device *dev, unsigned int txqueue) +{ + struct vector_private *vp = netdev_priv(dev); + + vp->estats.tx_timeout_count++; + netif_trans_update(dev); + schedule_work(&vp->reset_tx); +} + +static netdev_features_t vector_fix_features(struct net_device *dev, + netdev_features_t features) +{ + features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); + return features; +} + +static int vector_set_features(struct net_device *dev, + netdev_features_t features) +{ + struct vector_private *vp = netdev_priv(dev); + /* Adjust buffer sizes for GSO/GRO. Unfortunately, there is + * no way to negotiate it on raw sockets, so we can change + * only our side. + */ + if (features & NETIF_F_GRO) + /* All new frame buffers will be GRO-sized */ + vp->req_size = 65536; + else + /* All new frame buffers will be normal sized */ + vp->req_size = vp->max_packet + vp->headroom + SAFETY_MARGIN; + return 0; +} + +#ifdef CONFIG_NET_POLL_CONTROLLER +static void vector_net_poll_controller(struct net_device *dev) +{ + disable_irq(dev->irq); + vector_rx_interrupt(dev->irq, dev); + enable_irq(dev->irq); +} +#endif + +static void vector_net_get_drvinfo(struct net_device *dev, + struct ethtool_drvinfo *info) +{ + strscpy(info->driver, DRIVER_NAME); +} + +static int vector_net_load_bpf_flash(struct net_device *dev, + struct ethtool_flash *efl) +{ + struct vector_private *vp = netdev_priv(dev); + struct vector_device *vdevice; + const struct firmware *fw; + int result = 0; + + if (!(vp->options & VECTOR_BPF_FLASH)) { + netdev_err(dev, "loading firmware not permitted: %s\n", efl->data); + return -1; + } + + if (vp->bpf != NULL) { + if (vp->opened) + uml_vector_detach_bpf(vp->fds->rx_fd, vp->bpf); + kfree(vp->bpf->filter); + vp->bpf->filter = NULL; + } else { + vp->bpf = kmalloc_obj(struct sock_fprog, GFP_ATOMIC); + if (vp->bpf == NULL) { + netdev_err(dev, "failed to allocate memory for firmware\n"); + goto flash_fail; + } + } + + vdevice = find_device(vp->unit); + + if (request_firmware(&fw, efl->data, &vdevice->pdev.dev)) + goto flash_fail; + + vp->bpf->filter = kmemdup(fw->data, fw->size, GFP_ATOMIC); + if (!vp->bpf->filter) + goto free_buffer; + + vp->bpf->len = fw->size / sizeof(struct sock_filter); + release_firmware(fw); + + if (vp->opened) + result = uml_vector_attach_bpf(vp->fds->rx_fd, vp->bpf); + + return result; + +free_buffer: + release_firmware(fw); + +flash_fail: + if (vp->bpf != NULL) + kfree(vp->bpf->filter); + kfree(vp->bpf); + vp->bpf = NULL; + return -1; +} + +static void vector_get_ringparam(struct net_device *netdev, + struct ethtool_ringparam *ring, + struct kernel_ethtool_ringparam *kernel_ring, + struct netlink_ext_ack *extack) +{ + struct vector_private *vp = netdev_priv(netdev); + + ring->rx_max_pending = vp->rx_queue->max_depth; + ring->tx_max_pending = vp->tx_queue->max_depth; + ring->rx_pending = vp->rx_queue->max_depth; + ring->tx_pending = vp->tx_queue->max_depth; +} + +static void vector_get_strings(struct net_device *dev, u32 stringset, u8 *buf) +{ + switch (stringset) { + case ETH_SS_TEST: + *buf = '\0'; + break; + case ETH_SS_STATS: + memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys)); + break; + default: + WARN_ON(1); + break; + } +} + +static int vector_get_sset_count(struct net_device *dev, int sset) +{ + switch (sset) { + case ETH_SS_TEST: + return 0; + case ETH_SS_STATS: + return VECTOR_NUM_STATS; + default: + return -EOPNOTSUPP; + } +} + +static void vector_get_ethtool_stats(struct net_device *dev, + struct ethtool_stats *estats, + u64 *tmp_stats) +{ + struct vector_private *vp = netdev_priv(dev); + + /* Stats are modified in the dequeue portions of + * rx/tx which are protected by the head locks + * grabbing these locks here ensures they are up + * to date. + */ + + spin_lock(&vp->tx_queue->head_lock); + spin_lock(&vp->rx_queue->head_lock); + memcpy(tmp_stats, &vp->estats, sizeof(struct vector_estats)); + spin_unlock(&vp->rx_queue->head_lock); + spin_unlock(&vp->tx_queue->head_lock); +} + +static int vector_get_coalesce(struct net_device *netdev, + struct ethtool_coalesce *ec, + struct kernel_ethtool_coalesce *kernel_coal, + struct netlink_ext_ack *extack) +{ + struct vector_private *vp = netdev_priv(netdev); + + ec->tx_coalesce_usecs = (vp->coalesce * 1000000) / HZ; + return 0; +} + +static int vector_set_coalesce(struct net_device *netdev, + struct ethtool_coalesce *ec, + struct kernel_ethtool_coalesce *kernel_coal, + struct netlink_ext_ack *extack) +{ + struct vector_private *vp = netdev_priv(netdev); + + vp->coalesce = (ec->tx_coalesce_usecs * HZ) / 1000000; + if (vp->coalesce == 0) + vp->coalesce = 1; + return 0; +} + +static const struct ethtool_ops vector_net_ethtool_ops = { + .supported_coalesce_params = ETHTOOL_COALESCE_TX_USECS, + .get_drvinfo = vector_net_get_drvinfo, + .get_link = ethtool_op_get_link, + .get_ts_info = ethtool_op_get_ts_info, + .get_ringparam = vector_get_ringparam, + .get_strings = vector_get_strings, + .get_sset_count = vector_get_sset_count, + .get_ethtool_stats = vector_get_ethtool_stats, + .get_coalesce = vector_get_coalesce, + .set_coalesce = vector_set_coalesce, + .flash_device = vector_net_load_bpf_flash, +}; + + +static const struct net_device_ops vector_netdev_ops = { + .ndo_open = vector_net_open, + .ndo_stop = vector_net_close, + .ndo_start_xmit = vector_net_start_xmit, + .ndo_set_rx_mode = vector_net_set_multicast_list, + .ndo_tx_timeout = vector_net_tx_timeout, + .ndo_set_mac_address = eth_mac_addr, + .ndo_validate_addr = eth_validate_addr, + .ndo_fix_features = vector_fix_features, + .ndo_set_features = vector_set_features, +#ifdef CONFIG_NET_POLL_CONTROLLER + .ndo_poll_controller = vector_net_poll_controller, +#endif +}; + +static void vector_timer_expire(struct timer_list *t) +{ + struct vector_private *vp = timer_container_of(vp, t, tl); + + vp->estats.tx_kicks++; + napi_schedule(&vp->napi); +} + +static void vector_setup_etheraddr(struct net_device *dev, char *str) +{ + u8 addr[ETH_ALEN]; + + if (str == NULL) + goto random; + + if (!mac_pton(str, addr)) { + netdev_err(dev, + "Failed to parse '%s' as an ethernet address\n", str); + goto random; + } + if (is_multicast_ether_addr(addr)) { + netdev_err(dev, + "Attempt to assign a multicast ethernet address to a device disallowed\n"); + goto random; + } + if (!is_valid_ether_addr(addr)) { + netdev_err(dev, + "Attempt to assign an invalid ethernet address to a device disallowed\n"); + goto random; + } + if (!is_local_ether_addr(addr)) { + netdev_warn(dev, "Warning: Assigning a globally valid ethernet address to a device\n"); + netdev_warn(dev, "You should set the 2nd rightmost bit in the first byte of the MAC,\n"); + netdev_warn(dev, "i.e. %02x:%02x:%02x:%02x:%02x:%02x\n", + addr[0] | 0x02, addr[1], addr[2], addr[3], addr[4], addr[5]); + } + eth_hw_addr_set(dev, addr); + return; + +random: + netdev_info(dev, "Choosing a random ethernet address\n"); + eth_hw_addr_random(dev); +} + +static void vector_eth_configure( + int n, + struct arglist *def + ) +{ + struct vector_device *device; + struct net_device *dev; + struct vector_private *vp; + int err; + + device = kzalloc_obj(*device); + if (device == NULL) { + pr_err("Failed to allocate struct vector_device for vec%d\n", n); + return; + } + dev = alloc_etherdev(sizeof(struct vector_private)); + if (dev == NULL) { + pr_err("Failed to allocate struct net_device for vec%d\n", n); + goto out_free_device; + } + + dev->mtu = get_mtu(def); + + INIT_LIST_HEAD(&device->list); + device->unit = n; + + /* If this name ends up conflicting with an existing registered + * netdevice, that is OK, register_netdev{,ice}() will notice this + * and fail. + */ + snprintf(dev->name, sizeof(dev->name), "vec%d", n); + vector_setup_etheraddr(dev, uml_vector_fetch_arg(def, "mac")); + vp = netdev_priv(dev); + + /* sysfs register */ + if (!driver_registered) { + platform_driver_register(¨_net_driver); + driver_registered = 1; + } + device->pdev.id = n; + device->pdev.name = DRIVER_NAME; + device->pdev.dev.release = vector_device_release; + dev_set_drvdata(&device->pdev.dev, device); + if (platform_device_register(&device->pdev)) + goto out_free_netdev; + SET_NETDEV_DEV(dev, &device->pdev.dev); + + device->dev = dev; + + INIT_LIST_HEAD(&vp->list); + vp->dev = dev; + vp->unit = n; + vp->options = get_transport_options(def); + vp->parsed = def; + vp->max_packet = get_mtu(def) + ETH_HEADER_OTHER; + /* + * TODO - we need to calculate headroom so that ip header + * is 16 byte aligned all the time + */ + vp->headroom = get_headroom(def); + vp->coalesce = 2; + vp->req_size = get_req_size(def); + + dev->features = dev->hw_features = (NETIF_F_SG | NETIF_F_FRAGLIST); + INIT_WORK(&vp->reset_tx, vector_reset_tx); + + timer_setup(&vp->tl, vector_timer_expire, 0); + + /* FIXME */ + dev->netdev_ops = &vector_netdev_ops; + dev->ethtool_ops = &vector_net_ethtool_ops; + dev->watchdog_timeo = (HZ >> 1); + /* primary IRQ - fixme */ + dev->irq = 0; /* we will adjust this once opened */ + + rtnl_lock(); + err = register_netdevice(dev); + rtnl_unlock(); + if (err) + goto out_undo_user_init; + + spin_lock(&vector_devices_lock); + list_add(&device->list, &vector_devices); + spin_unlock(&vector_devices_lock); + + return; + +out_undo_user_init: + return; +out_free_netdev: + free_netdev(dev); +out_free_device: + kfree(device); +} + + + + +/* + * Invoked late in the init + */ + +static int __init vector_init(void) +{ + struct list_head *ele; + struct vector_cmd_line_arg *def; + struct arglist *parsed; + + list_for_each(ele, &vec_cmd_line) { + def = list_entry(ele, struct vector_cmd_line_arg, list); + parsed = uml_parse_vector_ifspec(def->arguments); + if (parsed != NULL) + vector_eth_configure(def->unit, parsed); + } + return 0; +} + + +/* Invoked at initial argument parsing, only stores + * arguments until a proper vector_init is called + * later + */ + +static int __init vector_setup(char *str) +{ + char *error; + int n, err; + struct vector_cmd_line_arg *new; + + err = vector_parse(str, &n, &str, &error); + if (err) { + pr_err("Couldn't parse '%s': %s\n", str, error); + return 1; + } + new = memblock_alloc_or_panic(sizeof(*new), SMP_CACHE_BYTES); + INIT_LIST_HEAD(&new->list); + new->unit = n; + new->arguments = str; + list_add_tail(&new->list, &vec_cmd_line); + return 1; +} + +__setup("vec", vector_setup); +__uml_help(vector_setup, +"vec[0-9]+: