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, 232 insertions, 0 deletions

diff --git a/drivers/mtd/tests/stresstest.c b/drivers/mtd/tests/stresstest.c
new file mode 100644
index 000000000..944f960a4
--- /dev/null
+++ b/drivers/mtd/tests/stresstest.c
@@ -0,0 +1,232 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2006-2008 Nokia Corporation
+ *
+ * Test random reads, writes and erases on MTD device.
+ *
+ * Author: Adrian Hunter <ext-adrian.hunter@nokia.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/err.h>
+#include <linux/mtd/mtd.h>
+#include <linux/slab.h>
+#include <linux/sched.h>
+#include <linux/vmalloc.h>
+#include <linux/random.h>
+
+#include "mtd_test.h"
+
+static int dev = -EINVAL;
+module_param(dev, int, S_IRUGO);
+MODULE_PARM_DESC(dev, "MTD device number to use");
+
+static int count = 10000;
+module_param(count, int, S_IRUGO);
+MODULE_PARM_DESC(count, "Number of operations to do (default is 10000)");
+
+static struct mtd_info *mtd;
+static unsigned char *writebuf;
+static unsigned char *readbuf;
+static unsigned char *bbt;
+static int *offsets;
+
+static int pgsize;
+static int bufsize;
+static int ebcnt;
+static int pgcnt;
+
+static int rand_eb(void)
+{
+	unsigned int eb;
+
+again:
+	/* Read or write up 2 eraseblocks at a time - hence 'ebcnt - 1' */
+	eb = get_random_u32_below(ebcnt - 1);
+	if (bbt[eb])
+		goto again;
+	return eb;
+}
+
+static int rand_offs(void)
+{
+	return get_random_u32_below(bufsize);
+}
+
+static int rand_len(int offs)
+{
+	return get_random_u32_below(bufsize - offs);
+}
+
+static int do_read(void)
+{
+	int eb = rand_eb();
+	int offs = rand_offs();
+	int len = rand_len(offs);
+	loff_t addr;
+
+	if (bbt[eb + 1]) {
+		if (offs >= mtd->erasesize)
+			offs -= mtd->erasesize;
+		if (offs + len > mtd->erasesize)
+			len = mtd->erasesize - offs;
+	}
+	addr = (loff_t)eb * mtd->erasesize + offs;
+	return mtdtest_read(mtd, addr, len, readbuf);
+}
+
+static int do_write(void)
+{
+	int eb = rand_eb(), offs, err, len;
+	loff_t addr;
+
+	offs = offsets[eb];
+	if (offs >= mtd->erasesize) {
+		err = mtdtest_erase_eraseblock(mtd, eb);
+		if (err)
+			return err;
+		offs = offsets[eb] = 0;
+	}
+	len = rand_len(offs);
+	len = ((len + pgsize - 1) / pgsize) * pgsize;
+	if (offs + len > mtd->erasesize) {
+		if (bbt[eb + 1])
+			len = mtd->erasesize - offs;
+		else {
+			err = mtdtest_erase_eraseblock(mtd, eb + 1);
+			if (err)
+				return err;
+			offsets[eb + 1] = 0;
+		}
+	}
+	addr = (loff_t)eb * mtd->erasesize + offs;
+	err = mtdtest_write(mtd, addr, len, writebuf);
+	if (unlikely(err))
+		return err;
+	offs += len;
+	while (offs > mtd->erasesize) {
+		offsets[eb++] = mtd->erasesize;
+		offs -= mtd->erasesize;
+	}
+	offsets[eb] = offs;
+	return 0;
+}
+
+static int do_operation(void)
+{
+	if (get_random_u32_below(2))
+		return do_read();
+	else
+		return do_write();
+}
+
+static int __init mtd_stresstest_init(void)
+{
+	int err;
+	int i, op;
+	uint64_t tmp;
+
+	printk(KERN_INFO "\n");
+	printk(KERN_INFO "=================================================\n");
+
+	if (dev < 0) {
+		pr_info("Please specify a valid mtd-device via module parameter\n");
+		pr_crit("CAREFUL: This test wipes all data on the specified MTD device!\n");
+		return -EINVAL;
+	}
+
+	pr_info("MTD device: %d\n", dev);
+
+	mtd = get_mtd_device(NULL, dev);
+	if (IS_ERR(mtd)) {
+		err = PTR_ERR(mtd);
+		pr_err("error: cannot get MTD device\n");
+		return err;
+	}
+
+	if (mtd->writesize == 1) {
+		pr_info("not NAND flash, assume page size is 512 "
+		       "bytes.\n");
+		pgsize = 512;
+	} else
+		pgsize = mtd->writesize;
+
+	tmp = mtd->size;
+	do_div(tmp, mtd->erasesize);
+	ebcnt = tmp;
+	pgcnt = mtd->erasesize / pgsize;
+
+	pr_info("MTD device size %llu, eraseblock size %u, "
+	       "page size %u, count of eraseblocks %u, pages per "
+	       "eraseblock %u, OOB size %u\n",
+	       (unsigned long long)mtd->size, mtd->erasesize,
+	       pgsize, ebcnt, pgcnt, mtd->oobsize);
+
+	if (ebcnt < 2) {
+		pr_err("error: need at least 2 eraseblocks\n");
+		err = -ENOSPC;
+		goto out_put_mtd;
+	}
+
+	/* Read or write up 2 eraseblocks at a time */
+	bufsize = mtd->erasesize * 2;
+
+	err = -ENOMEM;
+	readbuf = vmalloc(bufsize);
+	writebuf = vmalloc(bufsize);
+	offsets = kmalloc_objs(int, ebcnt);
+	if (!readbuf || !writebuf || !offsets)
+		goto out;
+	for (i = 0; i < ebcnt; i++)
+		offsets[i] = mtd->erasesize;
+	get_random_bytes(writebuf, bufsize);
+
+	bbt = kzalloc(ebcnt, GFP_KERNEL);
+	if (!bbt)
+		goto out;
+	err = mtdtest_scan_for_bad_eraseblocks(mtd, bbt, 0, ebcnt);
+	if (err)
+		goto out;
+
+	/* Do operations */
+	pr_info("doing operations\n");
+	for (op = 0; op < count; op++) {
+		if ((op & 1023) == 0)
+			pr_info("%d operations done\n", op);
+		err = do_operation();
+		if (err)
+			goto out;
+
+		err = mtdtest_relax();
+		if (err)
+			goto out;
+	}
+	pr_info("finished, %d operations done\n", op);
+
+out:
+	kfree(offsets);
+	kfree(bbt);
+	vfree(writebuf);
+	vfree(readbuf);
+out_put_mtd:
+	put_mtd_device(mtd);
+	if (err)
+		pr_info("error %d occurred\n", err);
+	printk(KERN_INFO "=================================================\n");
+	return err;
+}
+module_init(mtd_stresstest_init);
+
+static void __exit mtd_stresstest_exit(void)
+{
+	return;
+}
+module_exit(mtd_stresstest_exit);
+
+MODULE_DESCRIPTION("Stress test module");
+MODULE_AUTHOR("Adrian Hunter");
+MODULE_LICENSE("GPL");