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, 278 insertions, 0 deletions

diff --git a/drivers/firmware/efi/efi-init.c b/drivers/firmware/efi/efi-init.c
new file mode 100644
index 000000000..6103b1a08
--- /dev/null
+++ b/drivers/firmware/efi/efi-init.c
@@ -0,0 +1,278 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Extensible Firmware Interface
+ *
+ * Based on Extensible Firmware Interface Specification version 2.4
+ *
+ * Copyright (C) 2013 - 2015 Linaro Ltd.
+ */
+
+#define pr_fmt(fmt)	"efi: " fmt
+
+#include <linux/efi.h>
+#include <linux/fwnode.h>
+#include <linux/init.h>
+#include <linux/kexec_handover.h>
+#include <linux/memblock.h>
+#include <linux/mm_types.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_fdt.h>
+#include <linux/platform_device.h>
+#include <linux/sysfb.h>
+
+#include <asm/efi.h>
+
+unsigned long __initdata primary_display_table = EFI_INVALID_TABLE_ADDR;
+
+static int __init is_memory(efi_memory_desc_t *md)
+{
+	if (md->attribute & (EFI_MEMORY_WB|EFI_MEMORY_WT|EFI_MEMORY_WC))
+		return 1;
+	return 0;
+}
+
+/*
+ * Translate a EFI virtual address into a physical address: this is necessary,
+ * as some data members of the EFI system table are virtually remapped after
+ * SetVirtualAddressMap() has been called.
+ */
+static phys_addr_t __init efi_to_phys(unsigned long addr)
+{
+	efi_memory_desc_t *md;
+
+	for_each_efi_memory_desc(md) {
+		if (!(md->attribute & EFI_MEMORY_RUNTIME))
+			continue;
+		if (md->virt_addr == 0)
+			/* no virtual mapping has been installed by the stub */
+			break;
+		if (md->virt_addr <= addr &&
+		    (addr - md->virt_addr) < (md->num_pages << EFI_PAGE_SHIFT))
+			return md->phys_addr + addr - md->virt_addr;
+	}
+	return addr;
+}
+
+extern __weak const efi_config_table_type_t efi_arch_tables[];
+
+/*
+ * x86 defines its own instance of sysfb_primary_display and uses
+ * it even without EFI, everything else can get them from here.
+ */
+#if !defined(CONFIG_X86) && (defined(CONFIG_SYSFB) || defined(CONFIG_EFI_EARLYCON) || defined(CONFIG_FIRMWARE_EDID))
+struct sysfb_display_info sysfb_primary_display __section(".data");
+EXPORT_SYMBOL_GPL(sysfb_primary_display);
+#endif
+
+static void __init init_primary_display(void)
+{
+	struct sysfb_display_info *dpy;
+
+	if (primary_display_table != EFI_INVALID_TABLE_ADDR) {
+		dpy = early_memremap(primary_display_table, sizeof(*dpy));
+		if (!dpy) {
+			pr_err("Could not map primary_display config table\n");
+			return;
+		}
+		sysfb_primary_display = *dpy;
+		memset(dpy, 0, sizeof(*dpy));
+		early_memunmap(dpy, sizeof(*dpy));
+
+		if (memblock_is_map_memory(sysfb_primary_display.screen.lfb_base))
+			memblock_mark_nomap(sysfb_primary_display.screen.lfb_base,
+					    sysfb_primary_display.screen.lfb_size);
+
+		if (IS_ENABLED(CONFIG_EFI_EARLYCON))
+			efi_earlycon_reprobe();
+	}
+}
+
+static int __init uefi_init(u64 efi_system_table)
+{
+	efi_config_table_t *config_tables;
+	efi_system_table_t *systab;
+	size_t table_size;
+	int retval;
+
+	systab = early_memremap_ro(efi_system_table, sizeof(efi_system_table_t));
+	if (systab == NULL) {
+		pr_warn("Unable to map EFI system table.\n");
+		return -ENOMEM;
+	}
+
+	set_bit(EFI_BOOT, &efi.flags);
+	if (IS_ENABLED(CONFIG_64BIT))
+		set_bit(EFI_64BIT, &efi.flags);
+
+	retval = efi_systab_check_header(&systab->hdr);
+	if (retval)
+		goto out;
+
+	efi.runtime = systab->runtime;
+	efi.runtime_version = systab->hdr.revision;
+
+	efi_systab_report_header(&systab->hdr, efi_to_phys(systab->fw_vendor));
+
+	table_size = sizeof(efi_config_table_t) * systab->nr_tables;
+	config_tables = early_memremap_ro(efi_to_phys(systab->tables),
+					  table_size);
+	if (config_tables == NULL) {
+		pr_warn("Unable to map EFI config table array.\n");
+		retval = -ENOMEM;
+		goto out;
+	}
+	retval = efi_config_parse_tables(config_tables, systab->nr_tables,
+					 efi_arch_tables);
+
+	early_memunmap(config_tables, table_size);
+out:
+	early_memunmap(systab, sizeof(efi_system_table_t));
+	return retval;
+}
+
+/*
+ * Return true for regions that can be used as System RAM.
+ */
+static __init int is_usable_memory(efi_memory_desc_t *md)
+{
+	switch (md->type) {
+	case EFI_LOADER_CODE:
+	case EFI_LOADER_DATA:
+	case EFI_ACPI_RECLAIM_MEMORY:
+	case EFI_BOOT_SERVICES_CODE:
+	case EFI_BOOT_SERVICES_DATA:
+	case EFI_CONVENTIONAL_MEMORY:
+	case EFI_PERSISTENT_MEMORY:
+		/*
+		 * According to the spec, these regions are no longer reserved
+		 * after calling ExitBootServices(). However, we can only use
+		 * them as System RAM if they can be mapped writeback cacheable.
+		 */
+		return (md->attribute & EFI_MEMORY_WB);
+	default:
+		break;
+	}
+	return false;
+}
+
+static __init void reserve_regions(void)
+{
+	efi_memory_desc_t *md;
+	u64 paddr, npages, size;
+
+	if (efi_enabled(EFI_DBG))
+		pr_info("Processing EFI memory map:\n");
+
+	/*
+	 * Discard memblocks discovered so far except for KHO scratch
+	 * regions. Most memblocks at this point originate from memory nodes
+	 * in the DT and UEFI uses its own memory map instead. However, if
+	 * KHO is enabled, scratch regions, which are good known memory
+	 * must be preserved.
+	 */
+	memblock_dump_all();
+
+	if (is_kho_boot()) {
+		struct memblock_region *r;
+
+		/* Remove all non-KHO regions */
+		for_each_mem_region(r) {
+			if (!memblock_is_kho_scratch(r)) {
+				memblock_remove(r->base, r->size);
+				r--;
+			}
+		}
+	} else {
+		/*
+		 * KHO is disabled. Discard memblocks discovered so far:
+		 * if there are any at this point, they originate from memory
+		 * nodes in the DT, and UEFI uses its own memory map instead.
+		 */
+		memblock_remove(0, PHYS_ADDR_MAX);
+	}
+
+	for_each_efi_memory_desc(md) {
+		paddr = md->phys_addr;
+		npages = md->num_pages;
+
+		if (efi_enabled(EFI_DBG)) {
+			char buf[64];
+
+			pr_info("  0x%012llx-0x%012llx %s\n",
+				paddr, paddr + (npages << EFI_PAGE_SHIFT) - 1,
+				efi_md_typeattr_format(buf, sizeof(buf), md));
+		}
+
+		memrange_efi_to_native(&paddr, &npages);
+		size = npages << PAGE_SHIFT;
+
+		if (is_memory(md)) {
+			/*
+			 * Special purpose memory is 'soft reserved', which
+			 * means it is set aside initially. Don't add a memblock
+			 * for it now so that it can be hotplugged back in or
+			 * be assigned to the dax driver after boot.
+			 */
+			if (efi_soft_reserve_enabled() &&
+			    (md->attribute & EFI_MEMORY_SP))
+				continue;
+
+			early_init_dt_add_memory_arch(paddr, size);
+
+			if (!is_usable_memory(md))
+				memblock_mark_nomap(paddr, size);
+
+			/* keep ACPI reclaim memory intact for kexec etc. */
+			if (md->type == EFI_ACPI_RECLAIM_MEMORY)
+				memblock_reserve(paddr, size);
+		}
+	}
+}
+
+void __init efi_init(void)
+{
+	struct efi_memory_map_data data;
+	u64 efi_system_table;
+
+	/* Grab UEFI information placed in FDT by stub */
+	efi_system_table = efi_get_fdt_params(&data);
+	if (!efi_system_table)
+		return;
+
+	if (efi_memmap_init_early(&data) < 0) {
+		/*
+		* If we are booting via UEFI, the UEFI memory map is the only
+		* description of memory we have, so there is little point in
+		* proceeding if we cannot access it.
+		*/
+		panic("Unable to map EFI memory map.\n");
+	}
+
+	WARN(efi.memmap.desc_version != 1,
+	     "Unexpected EFI_MEMORY_DESCRIPTOR version %ld",
+	      efi.memmap.desc_version);
+
+	if (uefi_init(efi_system_table) < 0) {
+		efi_memmap_unmap();
+		return;
+	}
+
+	reserve_regions();
+	/*
+	 * For memblock manipulation, the cap should come after the memblock_add().
+	 * And now, memblock is fully populated, it is time to do capping.
+	 */
+	early_init_dt_check_for_usable_mem_range();
+	efi_find_mirror();
+	efi_esrt_init();
+	efi_mokvar_table_init();
+
+	memblock_reserve(data.phys_map & PAGE_MASK,
+			 PAGE_ALIGN(data.size + (data.phys_map & ~PAGE_MASK)));
+
+	if (IS_ENABLED(CONFIG_X86) ||
+	    IS_ENABLED(CONFIG_SYSFB) ||
+	    IS_ENABLED(CONFIG_EFI_EARLYCON))
+		init_primary_display();
+}