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, 225 insertions, 0 deletions

diff --git a/drivers/crypto/omap-crypto.c b/drivers/crypto/omap-crypto.c
new file mode 100644
index 000000000..7bd6869c2
--- /dev/null
+++ b/drivers/crypto/omap-crypto.c
@@ -0,0 +1,225 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * OMAP Crypto driver common support routines.
+ *
+ * Copyright (c) 2017 Texas Instruments Incorporated
+ *   Tero Kristo <t-kristo@ti.com>
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/scatterlist.h>
+#include <crypto/scatterwalk.h>
+
+#include "omap-crypto.h"
+
+static int omap_crypto_copy_sg_lists(int total, int bs,
+				     struct scatterlist **sg,
+				     struct scatterlist *new_sg, u16 flags)
+{
+	int n = sg_nents(*sg);
+	struct scatterlist *tmp;
+
+	if (!(flags & OMAP_CRYPTO_FORCE_SINGLE_ENTRY)) {
+		new_sg = kmalloc_objs(*new_sg, n);
+		if (!new_sg)
+			return -ENOMEM;
+
+		sg_init_table(new_sg, n);
+	}
+
+	tmp = new_sg;
+
+	while (*sg && total) {
+		int len = (*sg)->length;
+
+		if (total < len)
+			len = total;
+
+		if (len > 0) {
+			total -= len;
+			sg_set_page(tmp, sg_page(*sg), len, (*sg)->offset);
+			if (total <= 0)
+				sg_mark_end(tmp);
+			tmp = sg_next(tmp);
+		}
+
+		*sg = sg_next(*sg);
+	}
+
+	*sg = new_sg;
+
+	return 0;
+}
+
+static int omap_crypto_copy_sgs(int total, int bs, struct scatterlist **sg,
+				struct scatterlist *new_sg, u16 flags)
+{
+	void *buf;
+	int pages;
+	int new_len;
+
+	new_len = ALIGN(total, bs);
+	pages = get_order(new_len);
+
+	buf = (void *)__get_free_pages(GFP_ATOMIC, pages);
+	if (!buf) {
+		pr_err("%s: Couldn't allocate pages for unaligned cases.\n",
+		       __func__);
+		return -ENOMEM;
+	}
+
+	if (flags & OMAP_CRYPTO_COPY_DATA) {
+		scatterwalk_map_and_copy(buf, *sg, 0, total, 0);
+		if (flags & OMAP_CRYPTO_ZERO_BUF)
+			memset(buf + total, 0, new_len - total);
+	}
+
+	if (!(flags & OMAP_CRYPTO_FORCE_SINGLE_ENTRY))
+		sg_init_table(new_sg, 1);
+
+	sg_set_buf(new_sg, buf, new_len);
+
+	*sg = new_sg;
+
+	return 0;
+}
+
+static int omap_crypto_check_sg(struct scatterlist *sg, int total, int bs,
+				u16 flags)
+{
+	int len = 0;
+	int num_sg = 0;
+
+	if (!IS_ALIGNED(total, bs))
+		return OMAP_CRYPTO_NOT_ALIGNED;
+
+	while (sg) {
+		num_sg++;
+
+		if (!IS_ALIGNED(sg->offset, 4))
+			return OMAP_CRYPTO_NOT_ALIGNED;
+		if (!IS_ALIGNED(sg->length, bs))
+			return OMAP_CRYPTO_NOT_ALIGNED;
+#ifdef CONFIG_ZONE_DMA
+		if (page_zonenum(sg_page(sg)) != ZONE_DMA)
+			return OMAP_CRYPTO_NOT_ALIGNED;
+#endif
+
+		len += sg->length;
+		sg = sg_next(sg);
+
+		if (len >= total)
+			break;
+	}
+
+	if ((flags & OMAP_CRYPTO_FORCE_SINGLE_ENTRY) && num_sg > 1)
+		return OMAP_CRYPTO_NOT_ALIGNED;
+
+	if (len != total)
+		return OMAP_CRYPTO_BAD_DATA_LENGTH;
+
+	return 0;
+}
+
+int omap_crypto_align_sg(struct scatterlist **sg, int total, int bs,
+			 struct scatterlist *new_sg, u16 flags,
+			 u8 flags_shift, unsigned long *dd_flags)
+{
+	int ret;
+
+	*dd_flags &= ~(OMAP_CRYPTO_COPY_MASK << flags_shift);
+
+	if (flags & OMAP_CRYPTO_FORCE_COPY)
+		ret = OMAP_CRYPTO_NOT_ALIGNED;
+	else
+		ret = omap_crypto_check_sg(*sg, total, bs, flags);
+
+	if (ret == OMAP_CRYPTO_NOT_ALIGNED) {
+		ret = omap_crypto_copy_sgs(total, bs, sg, new_sg, flags);
+		if (ret)
+			return ret;
+		*dd_flags |= OMAP_CRYPTO_DATA_COPIED << flags_shift;
+	} else if (ret == OMAP_CRYPTO_BAD_DATA_LENGTH) {
+		ret = omap_crypto_copy_sg_lists(total, bs, sg, new_sg, flags);
+		if (ret)
+			return ret;
+		if (!(flags & OMAP_CRYPTO_FORCE_SINGLE_ENTRY))
+			*dd_flags |= OMAP_CRYPTO_SG_COPIED << flags_shift;
+	} else if (flags & OMAP_CRYPTO_FORCE_SINGLE_ENTRY) {
+		sg_set_buf(new_sg, sg_virt(*sg), (*sg)->length);
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(omap_crypto_align_sg);
+
+static void omap_crypto_copy_data(struct scatterlist *src,
+				  struct scatterlist *dst,
+				  int offset, int len)
+{
+	int amt;
+	void *srcb, *dstb;
+	int srco = 0, dsto = offset;
+
+	while (src && dst && len) {
+		if (srco >= src->length) {
+			srco -= src->length;
+			src = sg_next(src);
+			continue;
+		}
+
+		if (dsto >= dst->length) {
+			dsto -= dst->length;
+			dst = sg_next(dst);
+			continue;
+		}
+
+		amt = min(src->length - srco, dst->length - dsto);
+		amt = min(len, amt);
+
+		srcb = kmap_atomic(sg_page(src)) + srco + src->offset;
+		dstb = kmap_atomic(sg_page(dst)) + dsto + dst->offset;
+
+		memcpy(dstb, srcb, amt);
+
+		flush_dcache_page(sg_page(dst));
+
+		kunmap_atomic(srcb);
+		kunmap_atomic(dstb);
+
+		srco += amt;
+		dsto += amt;
+		len -= amt;
+	}
+}
+
+void omap_crypto_cleanup(struct scatterlist *sg, struct scatterlist *orig,
+			 int offset, int len, u8 flags_shift,
+			 unsigned long flags)
+{
+	void *buf;
+	int pages;
+
+	flags >>= flags_shift;
+	flags &= OMAP_CRYPTO_COPY_MASK;
+
+	if (!flags)
+		return;
+
+	buf = sg_virt(sg);
+	pages = get_order(len);
+
+	if (orig && (flags & OMAP_CRYPTO_DATA_COPIED))
+		omap_crypto_copy_data(sg, orig, offset, len);
+
+	if (flags & OMAP_CRYPTO_DATA_COPIED)
+		free_pages((unsigned long)buf, pages);
+	else if (flags & OMAP_CRYPTO_SG_COPIED)
+		kfree(sg);
+}
+EXPORT_SYMBOL_GPL(omap_crypto_cleanup);
+
+MODULE_DESCRIPTION("OMAP crypto support library.");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Tero Kristo <t-kristo@ti.com>");