x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

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) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 118 insertions, 0 deletions

diff --git a/Documentation/scsi/ChangeLog.arcmsr b/Documentation/scsi/ChangeLog.arcmsr
new file mode 100644
index 000000000..038a3e6ec
--- /dev/null
+++ b/Documentation/scsi/ChangeLog.arcmsr
@@ -0,0 +1,118 @@
+**************************************************************************
+** History
+**
+**   REV#         DATE             NAME         DESCRIPTION
+** 1.00.00.00    3/31/2004       Erich Chen     First release
+** 1.10.00.04    7/28/2004       Erich Chen     modify for ioctl
+** 1.10.00.06    8/28/2004       Erich Chen     modify for 2.6.x
+** 1.10.00.08    9/28/2004       Erich Chen     modify for x86_64
+** 1.10.00.10   10/10/2004       Erich Chen     bug fix for SMP & ioctl
+** 1.20.00.00   11/29/2004       Erich Chen     bug fix with arcmsr_bus_reset when PHY error
+** 1.20.00.02   12/09/2004       Erich Chen     bug fix with over 2T bytes RAID Volume
+** 1.20.00.04    1/09/2005       Erich Chen     fits for Debian linux kernel version 2.2.xx
+** 1.20.00.05    2/20/2005       Erich Chen     cleanly as look like a Linux driver at 2.6.x
+**                                              thanks for peoples kindness comment
+**						Kornel Wieliczek
+**						Christoph Hellwig
+**						Adrian Bunk
+**						Andrew Morton
+**						Christoph Hellwig
+**						James Bottomley
+**						Arjan van de Ven
+** 1.20.00.06    3/12/2005       Erich Chen     fix with arcmsr_pci_unmap_dma "unsigned long" cast,
+**						modify PCCB POOL allocated by "dma_alloc_coherent"
+**						(Kornel Wieliczek's comment)
+** 1.20.00.07    3/23/2005       Erich Chen     bug fix with arcmsr_scsi_host_template_init
+**						occur segmentation fault,
+**						if RAID adapter does not on PCI slot
+**						and modprobe/rmmod this driver twice.
+**						bug fix enormous stack usage (Adrian Bunk's comment)
+** 1.20.00.08    6/23/2005       Erich Chen     bug fix with abort command,
+**						in case of heavy loading when sata cable
+**						working on low quality connection
+** 1.20.00.09    9/12/2005       Erich Chen     bug fix with abort command handling, firmware version check
+**						and firmware update notify for hardware bug fix
+** 1.20.00.10    9/23/2005       Erich Chen     enhance sysfs function for change driver's max tag Q number.
+**						add DMA_64BIT_MASK for backward compatible with all 2.6.x
+**						add some useful message for abort command
+**						add ioctl code 'ARCMSR_IOCTL_FLUSH_ADAPTER_CACHE'
+**						customer can send this command for sync raid volume data
+** 1.20.00.11    9/29/2005       Erich Chen     by comment of Arjan van de Ven fix incorrect msleep redefine
+**						cast off sizeof(dma_addr_t) condition for 64bit pci_set_dma_mask
+** 1.20.00.12    9/30/2005       Erich Chen     bug fix with 64bit platform's ccbs using if over 4G system memory
+**						change 64bit pci_set_consistent_dma_mask into 32bit
+**						increcct adapter count if adapter initialize fail.
+**						miss edit at arcmsr_build_ccb....
+**						psge += sizeof(struct _SG64ENTRY *) =>
+**						psge += sizeof(struct _SG64ENTRY)
+**						64 bits sg entry would be incorrectly calculated
+**						thanks Kornel Wieliczek give me kindly notify
+**						and detail description
+** 1.20.00.13   11/15/2005       Erich Chen     scheduling pending ccb with FIFO
+**						change the architecture of arcmsr command queue list
+**						for linux standard list
+**						enable usage of pci message signal interrupt
+**						follow Randy.Danlup kindness suggestion cleanup this code
+** 1.20.00.14   05/02/2007	 Erich Chen & Nick Cheng
+**						1.implement PCI-Express error recovery function and AER capability
+**						2.implement the selection of ARCMSR_MAX_XFER_SECTORS_B=4096
+**				 		if firmware version is newer than 1.42
+**						3.modify arcmsr_iop_reset to improve the ability
+**						4.modify the ISR, arcmsr_interrupt routine,to prevent the
+**						inconsistency with sg_mod driver if application	directly calls
+**						the arcmsr driver w/o passing through scsi mid layer
+**						specially thanks to Yanmin Zhang's openhanded help about AER
+** 1.20.00.15   08/30/2007	 Erich Chen & Nick Cheng
+**						1. support ARC1200/1201/1202 SATA RAID adapter, which is named
+**						ACB_ADAPTER_TYPE_B
+**						2. modify the arcmsr_pci_slot_reset function
+**						3. modify the arcmsr_pci_ers_disconnect_forepart function
+**						4. modify the arcmsr_pci_ers_need_reset_forepart function
+** 1.20.00.15   09/27/2007	 Erich Chen & Nick Cheng
+**						1. add arcmsr_enable_eoi_mode() on adapter Type B
+** 						2. add readl(reg->iop2drv_doorbell_reg) in arcmsr_handle_hbb_isr()
+**						in case of the doorbell interrupt clearance is cached
+** 1.20.00.15   10/01/2007	 Erich Chen & Nick Cheng
+**						1. modify acb->devstate[i][j]
+**						as ARECA_RAID_GOOD instead of
+**						ARECA_RAID_GONE in arcmsr_alloc_ccb_pool
+** 1.20.00.15   11/06/2007       Erich Chen & Nick Cheng
+**						1. add conditional declaration for
+** 						arcmsr_pci_error_detected() and
+**						arcmsr_pci_slot_reset
+** 1.20.00.15	11/23/2007       Erich Chen & Nick Cheng
+**						1.check if the sg list member number
+**						exceeds arcmsr default limit in arcmsr_build_ccb()
+**						2.change the returned value type of arcmsr_build_ccb()
+**						from "void" to "int"
+**						3.add the conditional check if arcmsr_build_ccb()
+**						returns FAILED
+** 1.20.00.15	12/04/2007	 Erich Chen & Nick Cheng
+**						1. modify arcmsr_drain_donequeue() to ignore unknown
+**						command and let kernel process command timeout.
+**						This could handle IO request violating max. segments
+**						while Linux XFS over DM-CRYPT.
+**						Thanks to Milan Broz's comments <mbroz@redhat.com>
+** 1.20.00.15	12/24/2007	 Erich Chen & Nick Cheng
+**						1.fix the portability problems
+**						2.fix type B where we should _not_ iounmap() acb->pmu;
+**						it's not ioremapped.
+**						3.add return -ENOMEM if ioremap() fails
+**						4.transfer IS_SG64_ADDR w/ cpu_to_le32()
+**						in arcmsr_build_ccb
+**						5. modify acb->devstate[i][j] as ARECA_RAID_GONE instead of
+**						ARECA_RAID_GOOD in arcmsr_alloc_ccb_pool()
+**						6.fix arcmsr_cdb->Context as (unsigned long)arcmsr_cdb
+**						7.add the checking state of
+**						(outbound_intstatus & ARCMSR_MU_OUTBOUND_HANDLE_INT) == 0
+**						in arcmsr_handle_hba_isr
+**						8.replace pci_alloc_consistent()/pci_free_consistent() with kmalloc()/kfree() in arcmsr_iop_message_xfer()
+**						9. fix the release of dma memory for type B in arcmsr_free_ccb_pool()
+**						10.fix the arcmsr_polling_hbb_ccbdone()
+** 1.20.00.15	02/27/2008	Erich Chen & Nick Cheng
+**						1.arcmsr_iop_message_xfer() is called from atomic context under the
+**						queuecommand scsi_host_template handler. James Bottomley pointed out
+**						that the current GFP_KERNEL|GFP_DMA flags are wrong: firstly we are in
+**						atomic context, secondly this memory is not used for DMA.
+**						Also removed some unneeded casts. Thanks to Daniel Drake <dsd@gentoo.org>
+**************************************************************************