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, 193 insertions, 0 deletions

diff --git a/net/ceph/decode.c b/net/ceph/decode.c
new file mode 100644
index 000000000..2f21af38c
--- /dev/null
+++ b/net/ceph/decode.c
@@ -0,0 +1,193 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/ceph/ceph_debug.h>
+
+#include <linux/inet.h>
+
+#include <linux/ceph/decode.h>
+#include <linux/ceph/messenger.h>  /* for ceph_pr_addr() */
+
+static int
+ceph_decode_entity_addr_versioned(void **p, void *end,
+				  struct ceph_entity_addr *addr)
+{
+	int ret;
+	u8 struct_v;
+	u32 struct_len, addr_len;
+	void *struct_end;
+
+	ret = ceph_start_decoding(p, end, 1, "entity_addr_t", &struct_v,
+				  &struct_len);
+	if (ret)
+		goto bad;
+
+	ret = -EINVAL;
+	struct_end = *p + struct_len;
+
+	ceph_decode_copy_safe(p, end, &addr->type, sizeof(addr->type), bad);
+
+	ceph_decode_copy_safe(p, end, &addr->nonce, sizeof(addr->nonce), bad);
+
+	ceph_decode_32_safe(p, end, addr_len, bad);
+	if (addr_len > sizeof(addr->in_addr))
+		goto bad;
+
+	memset(&addr->in_addr, 0, sizeof(addr->in_addr));
+	if (addr_len) {
+		ceph_decode_copy_safe(p, end, &addr->in_addr, addr_len, bad);
+
+		addr->in_addr.ss_family =
+			le16_to_cpu((__force __le16)addr->in_addr.ss_family);
+	}
+
+	/* Advance past anything the client doesn't yet understand */
+	*p = struct_end;
+	ret = 0;
+bad:
+	return ret;
+}
+
+static int
+ceph_decode_entity_addr_legacy(void **p, void *end,
+			       struct ceph_entity_addr *addr)
+{
+	int ret = -EINVAL;
+
+	/* Skip rest of type field */
+	ceph_decode_skip_n(p, end, 3, bad);
+
+	/*
+	 * Clients that don't support ADDR2 always send TYPE_NONE, change it
+	 * to TYPE_LEGACY for forward compatibility.
+	 */
+	addr->type = CEPH_ENTITY_ADDR_TYPE_LEGACY;
+	ceph_decode_copy_safe(p, end, &addr->nonce, sizeof(addr->nonce), bad);
+	memset(&addr->in_addr, 0, sizeof(addr->in_addr));
+	ceph_decode_copy_safe(p, end, &addr->in_addr,
+			      sizeof(addr->in_addr), bad);
+	addr->in_addr.ss_family =
+			be16_to_cpu((__force __be16)addr->in_addr.ss_family);
+	ret = 0;
+bad:
+	return ret;
+}
+
+int
+ceph_decode_entity_addr(void **p, void *end, struct ceph_entity_addr *addr)
+{
+	u8 marker;
+
+	ceph_decode_8_safe(p, end, marker, bad);
+	if (marker == 1)
+		return ceph_decode_entity_addr_versioned(p, end, addr);
+	else if (marker == 0)
+		return ceph_decode_entity_addr_legacy(p, end, addr);
+bad:
+	return -EINVAL;
+}
+EXPORT_SYMBOL(ceph_decode_entity_addr);
+
+/*
+ * Return addr of desired type (MSGR2 or LEGACY) or error.  In case of
+ * multiple matches, use the first one for compatibility with userspace
+ * messenger.
+ *
+ * Assume encoding with MSG_ADDR2.
+ */
+int ceph_decode_entity_addrvec(void **p, void *end, bool msgr2,
+			       struct ceph_entity_addr *addr)
+{
+	__le32 my_type = msgr2 ? CEPH_ENTITY_ADDR_TYPE_MSGR2 :
+				 CEPH_ENTITY_ADDR_TYPE_LEGACY;
+	struct ceph_entity_addr tmp_addr;
+	int addr_cnt;
+	bool found;
+	u8 marker;
+	int ret;
+	int i;
+
+	ceph_decode_8_safe(p, end, marker, e_inval);
+	if (marker != 2) {
+		pr_err("bad addrvec marker %d\n", marker);
+		return -EINVAL;
+	}
+
+	ceph_decode_32_safe(p, end, addr_cnt, e_inval);
+	dout("%s addr_cnt %d\n", __func__, addr_cnt);
+
+	found = false;
+	for (i = 0; i < addr_cnt; i++) {
+		ret = ceph_decode_entity_addr(p, end, &tmp_addr);
+		if (ret)
+			return ret;
+
+		dout("%s i %d addr %s\n", __func__, i, ceph_pr_addr(&tmp_addr));
+		if (tmp_addr.type == my_type) {
+			if (!found) {
+				memcpy(addr, &tmp_addr, sizeof(*addr));
+				found = true;
+			} else {
+				dout("%s skipping extra match of type %d in addrvec\n",
+				     __func__, le32_to_cpu(my_type));
+			}
+		}
+	}
+
+	if (found)
+		return 0;
+
+	if (!addr_cnt)
+		return 0;  /* normal -- e.g. unused OSD id/slot */
+
+	if (addr_cnt == 1 && !memchr_inv(&tmp_addr, 0, sizeof(tmp_addr)))
+		return 0;  /* weird but effectively the same as !addr_cnt */
+
+	pr_err("no match of type %d in addrvec\n", le32_to_cpu(my_type));
+	return -ENOENT;
+
+e_inval:
+	return -EINVAL;
+}
+EXPORT_SYMBOL(ceph_decode_entity_addrvec);
+
+static int get_sockaddr_encoding_len(sa_family_t family)
+{
+	union {
+		struct sockaddr sa;
+		struct sockaddr_in sin;
+		struct sockaddr_in6 sin6;
+	} u;
+
+	switch (family) {
+	case AF_INET:
+		return sizeof(u.sin);
+	case AF_INET6:
+		return sizeof(u.sin6);
+	default:
+		return sizeof(u);
+	}
+}
+
+int ceph_entity_addr_encoding_len(const struct ceph_entity_addr *addr)
+{
+	sa_family_t family = get_unaligned(&addr->in_addr.ss_family);
+	int addr_len = get_sockaddr_encoding_len(family);
+
+	return 1 + CEPH_ENCODING_START_BLK_LEN + 4 + 4 + 4 + addr_len;
+}
+
+void ceph_encode_entity_addr(void **p, const struct ceph_entity_addr *addr)
+{
+	sa_family_t family = get_unaligned(&addr->in_addr.ss_family);
+	int addr_len = get_sockaddr_encoding_len(family);
+
+	ceph_encode_8(p, 1);  /* marker */
+	ceph_start_encoding(p, 1, 1, sizeof(addr->type) +
+				     sizeof(addr->nonce) +
+				     sizeof(u32) + addr_len);
+	ceph_encode_copy(p, &addr->type, sizeof(addr->type));
+	ceph_encode_copy(p, &addr->nonce, sizeof(addr->nonce));
+
+	ceph_encode_32(p, addr_len);
+	ceph_encode_16(p, family);
+	ceph_encode_copy(p, addr->in_addr.__data, addr_len - sizeof(family));
+}