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, 187 insertions, 0 deletions

diff --git a/drivers/mtd/devices/mtdram.c b/drivers/mtd/devices/mtdram.c
new file mode 100644
index 000000000..3fff3cdd9
--- /dev/null
+++ b/drivers/mtd/devices/mtdram.c
@@ -0,0 +1,187 @@
+/*
+ * mtdram - a test mtd device
+ * Author: Alexander Larsson <alex@cendio.se>
+ *
+ * Copyright (c) 1999 Alexander Larsson <alex@cendio.se>
+ * Copyright (c) 2005 Joern Engel <joern@wh.fh-wedel.de>
+ *
+ * This code is GPL
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/ioport.h>
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/mtdram.h>
+
+static unsigned long total_size = CONFIG_MTDRAM_TOTAL_SIZE;
+static unsigned long erase_size = CONFIG_MTDRAM_ERASE_SIZE;
+static unsigned long writebuf_size = 64;
+#define MTDRAM_TOTAL_SIZE (total_size * 1024)
+#define MTDRAM_ERASE_SIZE (erase_size * 1024)
+
+module_param(total_size, ulong, 0);
+MODULE_PARM_DESC(total_size, "Total device size in KiB");
+module_param(erase_size, ulong, 0);
+MODULE_PARM_DESC(erase_size, "Device erase block size in KiB");
+module_param(writebuf_size, ulong, 0);
+MODULE_PARM_DESC(writebuf_size, "Device write buf size in Bytes (Default: 64)");
+
+// We could store these in the mtd structure, but we only support 1 device..
+static struct mtd_info *mtd_info;
+
+static int check_offs_len(struct mtd_info *mtd, loff_t ofs, uint64_t len)
+{
+	int ret = 0;
+
+	/* Start address must align on block boundary */
+	if (mtd_mod_by_eb(ofs, mtd)) {
+		pr_debug("%s: unaligned address\n", __func__);
+		ret = -EINVAL;
+	}
+
+	/* Length must align on block boundary */
+	if (mtd_mod_by_eb(len, mtd)) {
+		pr_debug("%s: length not block aligned\n", __func__);
+		ret = -EINVAL;
+	}
+
+	return ret;
+}
+
+static int ram_erase(struct mtd_info *mtd, struct erase_info *instr)
+{
+	if (check_offs_len(mtd, instr->addr, instr->len))
+		return -EINVAL;
+	memset((char *)mtd->priv + instr->addr, 0xff, instr->len);
+
+	return 0;
+}
+
+static int ram_point(struct mtd_info *mtd, loff_t from, size_t len,
+		size_t *retlen, void **virt, resource_size_t *phys)
+{
+	*virt = mtd->priv + from;
+	*retlen = len;
+
+	if (phys) {
+		/* limit retlen to the number of contiguous physical pages */
+		unsigned long page_ofs = offset_in_page(*virt);
+		void *addr = *virt - page_ofs;
+		unsigned long pfn1, pfn0 = vmalloc_to_pfn(addr);
+
+		*phys = __pfn_to_phys(pfn0) + page_ofs;
+		len += page_ofs;
+		while (len > PAGE_SIZE) {
+			len -= PAGE_SIZE;
+			addr += PAGE_SIZE;
+			pfn0++;
+			pfn1 = vmalloc_to_pfn(addr);
+			if (pfn1 != pfn0) {
+				*retlen = addr - *virt;
+				break;
+			}
+		}
+	}
+
+	return 0;
+}
+
+static int ram_unpoint(struct mtd_info *mtd, loff_t from, size_t len)
+{
+	return 0;
+}
+
+static int ram_read(struct mtd_info *mtd, loff_t from, size_t len,
+		size_t *retlen, u_char *buf)
+{
+	memcpy(buf, mtd->priv + from, len);
+	*retlen = len;
+	return 0;
+}
+
+static int ram_write(struct mtd_info *mtd, loff_t to, size_t len,
+		size_t *retlen, const u_char *buf)
+{
+	memcpy((char *)mtd->priv + to, buf, len);
+	*retlen = len;
+	return 0;
+}
+
+static void __exit cleanup_mtdram(void)
+{
+	if (mtd_info) {
+		mtd_device_unregister(mtd_info);
+		vfree(mtd_info->priv);
+		kfree(mtd_info);
+	}
+}
+
+int mtdram_init_device(struct mtd_info *mtd, void *mapped_address,
+		unsigned long size, const char *name)
+{
+	memset(mtd, 0, sizeof(*mtd));
+
+	/* Setup the MTD structure */
+	mtd->name = name;
+	mtd->type = MTD_RAM;
+	mtd->flags = MTD_CAP_RAM;
+	mtd->size = size;
+	mtd->writesize = 1;
+	mtd->writebufsize = writebuf_size;
+	mtd->erasesize = MTDRAM_ERASE_SIZE;
+	mtd->priv = mapped_address;
+
+	mtd->owner = THIS_MODULE;
+	mtd->_erase = ram_erase;
+	mtd->_point = ram_point;
+	mtd->_unpoint = ram_unpoint;
+	mtd->_read = ram_read;
+	mtd->_write = ram_write;
+
+	if (mtd_device_register(mtd, NULL, 0))
+		return -EIO;
+
+	return 0;
+}
+
+static int __init init_mtdram(void)
+{
+	void *addr;
+	int err;
+
+	if (!total_size)
+		return -EINVAL;
+
+	/* Allocate some memory */
+	mtd_info = kmalloc_obj(struct mtd_info);
+	if (!mtd_info)
+		return -ENOMEM;
+
+	addr = vmalloc(MTDRAM_TOTAL_SIZE);
+	if (!addr) {
+		kfree(mtd_info);
+		mtd_info = NULL;
+		return -ENOMEM;
+	}
+	err = mtdram_init_device(mtd_info, addr, MTDRAM_TOTAL_SIZE, "mtdram test device");
+	if (err) {
+		vfree(addr);
+		kfree(mtd_info);
+		mtd_info = NULL;
+		return err;
+	}
+	memset(mtd_info->priv, 0xff, MTDRAM_TOTAL_SIZE);
+	return err;
+}
+
+module_init(init_mtdram);
+module_exit(cleanup_mtdram);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Alexander Larsson <alexl@redhat.com>");
+MODULE_DESCRIPTION("Simulated MTD driver for testing");