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, 260 insertions, 0 deletions

diff --git a/fs/smb/server/compress.c b/fs/smb/server/compress.c
new file mode 100644
index 000000000..5162fb84c
--- /dev/null
+++ b/fs/smb/server/compress.c
@@ -0,0 +1,260 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * SMB2 compression support for ksmbd.
+ *
+ * Receive and send SMB 3.1.1 compression transforms using the common helpers.
+ *
+ * Copyright (C) 2026 Namjae Jeon <linkinjeon@kernel.org>
+ */
+#include <linux/slab.h>
+
+#include "compress.h"
+#include "smb_common.h"
+#include "../common/compress/lz77.h"
+
+#define SMB_COMPRESS_MIN_LEN	PAGE_SIZE
+
+static int __ksmbd_decompress_request(struct ksmbd_conn *conn,
+					     void *request_buf, void **out_buf)
+{
+	struct smb2_compression_hdr *hdr;
+	unsigned int pdu_size = get_rfc1002_len(request_buf);
+	u32 orig_size, offset, out_size;
+	u32 max_allowed_pdu_size;
+	char *out;
+	int rc;
+
+	if (pdu_size < sizeof(struct smb2_compression_hdr))
+		return -EINVAL;
+
+	if (conn->dialect != SMB311_PROT_ID ||
+	    conn->compress_algorithm == SMB3_COMPRESS_NONE)
+		return -EINVAL;
+
+	hdr = smb_get_msg(request_buf);
+	if (hdr->ProtocolId != SMB2_COMPRESSION_TRANSFORM_ID)
+		return -EINVAL;
+
+	orig_size = le32_to_cpu(hdr->OriginalCompressedSegmentSize);
+	/*
+	 * For chained transforms the top-level header is only eight bytes; the
+	 * Flags field overlays the first payload header. Reject unknown Flags
+	 * and unnegotiated chained mode before allocating the output buffer.
+	 */
+	if (hdr->Flags == cpu_to_le16(SMB2_COMPRESSION_FLAG_CHAINED)) {
+		if (!conn->compress_chained)
+			return -EINVAL;
+		out_size = orig_size;
+	} else if (hdr->Flags == cpu_to_le16(SMB2_COMPRESSION_FLAG_NONE)) {
+		offset = le32_to_cpu(hdr->Offset);
+		if (offset > pdu_size - sizeof(*hdr) ||
+		    check_add_overflow(orig_size, offset, &out_size))
+			return -EINVAL;
+	} else {
+		return -EINVAL;
+	}
+
+	max_allowed_pdu_size = ksmbd_max_allowed_pdu_size(conn);
+	if (out_size < sizeof(struct smb2_pdu) ||
+	    out_size > max_allowed_pdu_size ||
+	    out_size > MAX_STREAM_PROT_LEN)
+		return -EINVAL;
+
+	out = kvmalloc(out_size + 4 + 1, KSMBD_DEFAULT_GFP);
+	if (!out)
+		return -ENOMEM;
+
+	*(__be32 *)out = cpu_to_be32(out_size);
+	rc = smb_compression_decompress(conn->compress_algorithm,
+					conn->compress_chained,
+					conn->compress_pattern,
+					(char *)hdr, pdu_size, out + 4, out_size);
+	if (rc) {
+		kvfree(out);
+		return rc;
+	}
+
+	*out_buf = out;
+	return 0;
+}
+
+/**
+ * ksmbd_decompress_request() - replace a compressed request with its SMB2 PDU
+ * @conn: connection which owns the current RFC1002 request buffer
+ *
+ * Derive the uncompressed size from the transform variant, enforce ksmbd's
+ * normal message limits, and ask the common decoder to validate every payload.
+ * On success, replace conn->request_buf with a regular RFC1002-framed SMB2
+ * message so the rest of the request path needs no compression awareness.
+ *
+ * Return: 0 on success, otherwise a negative errno.
+ */
+int ksmbd_decompress_request(struct ksmbd_conn *conn)
+{
+	void *out_buf;
+	int rc;
+
+	rc = __ksmbd_decompress_request(conn, conn->request_buf, &out_buf);
+	if (rc)
+		return rc;
+
+	kvfree(conn->request_buf);
+	conn->request_buf = out_buf;
+	return 0;
+}
+
+/**
+ * ksmbd_decompress_work_request() - decompress an encrypted work request
+ * @work: work item whose request buffer contains a compression transform
+ *
+ * SMB3 encrypts a compressed message by applying compression first and
+ * encryption second.  The receive loop can therefore only decode the
+ * compression transform before work allocation for an unencrypted request;
+ * an encrypted request must be decompressed after its encryption layer has
+ * been removed.
+ *
+ * Return: 0 on success, otherwise a negative errno.
+ */
+int ksmbd_decompress_work_request(struct ksmbd_work *work)
+{
+	void *out_buf;
+	int rc;
+
+	rc = __ksmbd_decompress_request(work->conn, work->request_buf,
+					&out_buf);
+	if (rc)
+		return rc;
+
+	kvfree(work->request_buf);
+	work->request_buf = out_buf;
+	return 0;
+}
+
+/**
+ * ksmbd_compress_response() - compress an eligible ksmbd response
+ * @work: request work item containing the response iov
+ *
+ * Compression transforms describe one contiguous SMB2 message, while ksmbd
+ * builds responses from multiple iov entries. Flatten the response first,
+ * produce the negotiated transform, and replace the response iov only when the
+ * result is smaller than the original message.
+ *
+ * Encrypted and compound responses are intentionally left unchanged. The
+ * caller may still continue sending the original response when this function
+ * returns zero.
+ *
+ * Return: 1 if the response was replaced, 0 if compression was skipped, or a
+ * negative errno on failure.
+ */
+int ksmbd_compress_response(struct ksmbd_work *work)
+{
+	struct smb2_compression_hdr *chdr;
+	struct smb2_hdr *req_hdr;
+	u32 src_len, dst_len, compressed_pdu_len, max_dst_len;
+	u8 *src = NULL, *out = NULL, *p;
+	int i, rc;
+
+	if (!work->compress_response || work->encrypted ||
+	    work->conn->compress_algorithm != SMB3_COMPRESS_LZ77)
+		return 0;
+
+	req_hdr = smb_get_msg(work->request_buf);
+	if (req_hdr->NextCommand || work->next_smb2_rcv_hdr_off ||
+	    work->next_smb2_rsp_hdr_off)
+		return 0;
+
+	src_len = get_rfc1002_len(work->iov[0].iov_base);
+	if (src_len < SMB_COMPRESS_MIN_LEN)
+		return 0;
+
+	src = kvmalloc(src_len, KSMBD_DEFAULT_GFP);
+	if (!src)
+		return -ENOMEM;
+
+	p = src;
+	/* iov[0] contains only the RFC1002 length; the SMB2 PDU starts at iov[1]. */
+	for (i = 1; i < work->iov_cnt; i++) {
+		if (work->iov[i].iov_len > src + src_len - p) {
+			rc = -EINVAL;
+			goto out;
+		}
+		memcpy(p, work->iov[i].iov_base, work->iov[i].iov_len);
+		p += work->iov[i].iov_len;
+	}
+	if (p != src + src_len) {
+		rc = -EINVAL;
+		goto out;
+	}
+
+	max_dst_len = smb_lz77_compressed_alloc_size(src_len) +
+		sizeof(struct smb2_compression_hdr) +
+		3 * sizeof(struct smb2_compression_payload_hdr) +
+		2 * sizeof(struct smb2_compression_pattern_v1);
+	out = kvzalloc(sizeof(__be32) + max_dst_len,
+		       KSMBD_DEFAULT_GFP);
+	if (!out) {
+		rc = -ENOMEM;
+		goto out;
+	}
+
+	if (work->conn->compress_chained) {
+		dst_len = max_dst_len;
+		rc = smb_compression_compress_chained(SMB3_COMPRESS_LZ77,
+						      work->conn->compress_pattern,
+						      src, src_len,
+						      out + sizeof(__be32),
+						      &dst_len);
+		if (rc == -EMSGSIZE || dst_len >= src_len) {
+			rc = 0;
+			goto out;
+		}
+		if (rc)
+			goto out;
+		compressed_pdu_len = dst_len;
+	} else {
+		/*
+		 * Peers which did not negotiate chained compression still use
+		 * the original 16-byte unchained transform format.
+		 */
+		dst_len = smb_lz77_compressed_alloc_size(src_len);
+		rc = smb_lz77_compress(src, src_len,
+				       out + sizeof(__be32) + sizeof(*chdr),
+				       &dst_len);
+		if (rc == -EMSGSIZE ||
+		    dst_len + sizeof(*chdr) >= src_len) {
+			rc = 0;
+			goto out;
+		}
+		if (rc)
+			goto out;
+
+		compressed_pdu_len = sizeof(*chdr) + dst_len;
+		chdr = (struct smb2_compression_hdr *)(out + sizeof(__be32));
+		chdr->ProtocolId = SMB2_COMPRESSION_TRANSFORM_ID;
+		chdr->OriginalCompressedSegmentSize = cpu_to_le32(src_len);
+		chdr->CompressionAlgorithm = SMB3_COMPRESS_LZ77;
+		chdr->Flags = cpu_to_le16(SMB2_COMPRESSION_FLAG_NONE);
+		chdr->Offset = 0;
+	}
+
+	*(__be32 *)out = cpu_to_be32(compressed_pdu_len);
+
+	/*
+	 * Keep the transform in work->compress_buf until send completion.
+	 * Existing response iovs can then be replaced without changing their
+	 * individual ownership rules.
+	 */
+	work->compress_buf = out;
+	work->iov[0].iov_base = out;
+	work->iov[0].iov_len = sizeof(__be32);
+	work->iov[1].iov_base = out + sizeof(__be32);
+	work->iov[1].iov_len = compressed_pdu_len;
+	work->iov_cnt = 2;
+	work->iov_idx = 1;
+	out = NULL;
+	rc = 1;
+out:
+	kvfree(out);
+	kvfree(src);
+	return rc;
+}