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

3 files changed, 552 insertions, 0 deletions

diff --git a/net/strparser/Kconfig b/net/strparser/Kconfig
new file mode 100644
index 000000000..e6146c21a
--- /dev/null
+++ b/net/strparser/Kconfig
@@ -0,0 +1,3 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config STREAM_PARSER
+	def_bool n
diff --git a/net/strparser/Makefile b/net/strparser/Makefile
new file mode 100644
index 000000000..931319153
--- /dev/null
+++ b/net/strparser/Makefile
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+obj-$(CONFIG_STREAM_PARSER) += strparser.o
diff --git a/net/strparser/strparser.c b/net/strparser/strparser.c
new file mode 100644
index 000000000..a23f4b4df
--- /dev/null
+++ b/net/strparser/strparser.c
@@ -0,0 +1,547 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Stream Parser
+ *
+ * Copyright (c) 2016 Tom Herbert <tom@herbertland.com>
+ */
+
+#include <linux/bpf.h>
+#include <linux/errno.h>
+#include <linux/errqueue.h>
+#include <linux/file.h>
+#include <linux/in.h>
+#include <linux/kernel.h>
+#include <linux/export.h>
+#include <linux/init.h>
+#include <linux/net.h>
+#include <linux/netdevice.h>
+#include <linux/poll.h>
+#include <linux/rculist.h>
+#include <linux/skbuff.h>
+#include <linux/socket.h>
+#include <linux/uaccess.h>
+#include <linux/workqueue.h>
+#include <net/strparser.h>
+#include <net/netns/generic.h>
+#include <net/sock.h>
+
+static struct workqueue_struct *strp_wq;
+
+static inline struct _strp_msg *_strp_msg(struct sk_buff *skb)
+{
+	return (struct _strp_msg *)((void *)skb->cb +
+		offsetof(struct sk_skb_cb, strp));
+}
+
+/* Lower lock held */
+static void strp_abort_strp(struct strparser *strp, int err)
+{
+	/* Unrecoverable error in receive */
+
+	cancel_delayed_work(&strp->msg_timer_work);
+
+	if (strp->stopped)
+		return;
+
+	strp->stopped = 1;
+
+	if (strp->skb_head) {
+		kfree_skb(strp->skb_head);
+		strp->skb_head = NULL;
+	}
+
+	strp->skb_nextp = NULL;
+	strp->need_bytes = 0;
+
+	if (strp->sk) {
+		struct sock *sk = strp->sk;
+
+		/* Report an error on the lower socket */
+		sk->sk_err = -err;
+		sk_error_report(sk);
+	}
+}
+
+static void strp_start_timer(struct strparser *strp, long timeo)
+{
+	if (timeo && timeo != LONG_MAX)
+		mod_delayed_work(strp_wq, &strp->msg_timer_work, timeo);
+}
+
+/* Lower lock held */
+static void strp_parser_err(struct strparser *strp, int err,
+			    read_descriptor_t *desc)
+{
+	desc->error = err;
+	kfree_skb(strp->skb_head);
+	strp->skb_head = NULL;
+	strp->cb.abort_parser(strp, err);
+}
+
+static inline int strp_peek_len(struct strparser *strp)
+{
+	if (strp->sk) {
+		struct socket *sock = strp->sk->sk_socket;
+
+		return sock->ops->peek_len(sock);
+	}
+
+	/* If we don't have an associated socket there's nothing to peek.
+	 * Return int max to avoid stopping the strparser.
+	 */
+
+	return INT_MAX;
+}
+
+/* Lower socket lock held */
+static int __strp_recv(read_descriptor_t *desc, struct sk_buff *orig_skb,
+		       unsigned int orig_offset, size_t orig_len,
+		       size_t max_msg_size, long timeo)
+{
+	struct strparser *strp = (struct strparser *)desc->arg.data;
+	struct _strp_msg *stm;
+	struct sk_buff *head, *skb;
+	size_t eaten = 0, cand_len;
+	ssize_t extra;
+	int err;
+	bool cloned_orig = false;
+
+	if (strp->paused)
+		return 0;
+
+	head = strp->skb_head;
+	if (head) {
+		/* Message already in progress */
+		if (unlikely(orig_offset)) {
+			/* Getting data with a non-zero offset when a message is
+			 * in progress is not expected. If it does happen, we
+			 * need to clone and pull since we can't deal with
+			 * offsets in the skbs for a message expect in the head.
+			 */
+			orig_skb = skb_clone(orig_skb, GFP_ATOMIC);
+			if (!orig_skb) {
+				STRP_STATS_INCR(strp->stats.mem_fail);
+				desc->error = -ENOMEM;
+				return 0;
+			}
+			if (!pskb_pull(orig_skb, orig_offset)) {
+				STRP_STATS_INCR(strp->stats.mem_fail);
+				kfree_skb(orig_skb);
+				desc->error = -ENOMEM;
+				return 0;
+			}
+			cloned_orig = true;
+			orig_offset = 0;
+		}
+
+		if (!strp->skb_nextp) {
+			/* We are going to append to the frag_list of head.
+			 * Need to unshare the frag_list.
+			 */
+			err = skb_unclone(head, GFP_ATOMIC);
+			if (err) {
+				STRP_STATS_INCR(strp->stats.mem_fail);
+				desc->error = err;
+				return 0;
+			}
+
+			if (unlikely(skb_shinfo(head)->frag_list)) {
+				/* We can't append to an sk_buff that already
+				 * has a frag_list. We create a new head, point
+				 * the frag_list of that to the old head, and
+				 * then are able to use the old head->next for
+				 * appending to the message.
+				 */
+				if (WARN_ON(head->next)) {
+					desc->error = -EINVAL;
+					return 0;
+				}
+
+				skb = alloc_skb_for_msg(head);
+				if (!skb) {
+					STRP_STATS_INCR(strp->stats.mem_fail);
+					desc->error = -ENOMEM;
+					return 0;
+				}
+
+				strp->skb_nextp = &head->next;
+				strp->skb_head = skb;
+				head = skb;
+			} else {
+				strp->skb_nextp =
+				    &skb_shinfo(head)->frag_list;
+			}
+		}
+	}
+
+	while (eaten < orig_len) {
+		/* Always clone since we will consume something */
+		skb = skb_clone(orig_skb, GFP_ATOMIC);
+		if (!skb) {
+			STRP_STATS_INCR(strp->stats.mem_fail);
+			desc->error = -ENOMEM;
+			break;
+		}
+
+		cand_len = orig_len - eaten;
+
+		head = strp->skb_head;
+		if (!head) {
+			head = skb;
+			strp->skb_head = head;
+			/* Will set skb_nextp on next packet if needed */
+			strp->skb_nextp = NULL;
+			stm = _strp_msg(head);
+			memset(stm, 0, sizeof(*stm));
+			stm->strp.offset = orig_offset + eaten;
+		} else {
+			/* Unclone if we are appending to an skb that we
+			 * already share a frag_list with.
+			 */
+			if (skb_has_frag_list(skb)) {
+				err = skb_unclone(skb, GFP_ATOMIC);
+				if (err) {
+					STRP_STATS_INCR(strp->stats.mem_fail);
+					desc->error = err;
+					break;
+				}
+			}
+
+			stm = _strp_msg(head);
+			*strp->skb_nextp = skb;
+			strp->skb_nextp = &skb->next;
+			head->data_len += skb->len;
+			head->len += skb->len;
+			head->truesize += skb->truesize;
+		}
+
+		if (!stm->strp.full_len) {
+			ssize_t len;
+
+			len = (*strp->cb.parse_msg)(strp, head);
+
+			if (!len) {
+				/* Need more header to determine length */
+				if (!stm->accum_len) {
+					/* Start RX timer for new message */
+					strp_start_timer(strp, timeo);
+				}
+				stm->accum_len += cand_len;
+				eaten += cand_len;
+				STRP_STATS_INCR(strp->stats.need_more_hdr);
+				WARN_ON(eaten != orig_len);
+				break;
+			} else if (len < 0) {
+				if (len == -ESTRPIPE && stm->accum_len) {
+					len = -ENODATA;
+					strp->unrecov_intr = 1;
+				} else {
+					strp->interrupted = 1;
+				}
+				strp_parser_err(strp, len, desc);
+				break;
+			} else if (len > max_msg_size) {
+				/* Message length exceeds maximum allowed */
+				STRP_STATS_INCR(strp->stats.msg_too_big);
+				strp_parser_err(strp, -EMSGSIZE, desc);
+				break;
+			} else if (len <= (ssize_t)head->len -
+					  (ssize_t)skb->len - stm->strp.offset) {
+				/* Length must be into new skb (and also
+				 * greater than zero)
+				 */
+				STRP_STATS_INCR(strp->stats.bad_hdr_len);
+				strp_parser_err(strp, -EPROTO, desc);
+				break;
+			}
+
+			stm->strp.full_len = len;
+		}
+
+		extra = (ssize_t)(stm->accum_len + cand_len) -
+			stm->strp.full_len;
+
+		if (extra < 0) {
+			/* Message not complete yet. */
+			if (stm->strp.full_len - stm->accum_len >
+			    strp_peek_len(strp)) {
+				/* Don't have the whole message in the socket
+				 * buffer. Set strp->need_bytes to wait for
+				 * the rest of the message. Also, set "early
+				 * eaten" since we've already buffered the skb
+				 * but don't consume yet per strp_read_sock.
+				 */
+
+				if (!stm->accum_len) {
+					/* Start RX timer for new message */
+					strp_start_timer(strp, timeo);
+				}
+
+				stm->accum_len += cand_len;
+				eaten += cand_len;
+				strp->need_bytes = stm->strp.full_len -
+						       stm->accum_len;
+				STRP_STATS_ADD(strp->stats.bytes, cand_len);
+				desc->count = 0; /* Stop reading socket */
+				break;
+			}
+			stm->accum_len += cand_len;
+			eaten += cand_len;
+			WARN_ON(eaten != orig_len);
+			break;
+		}
+
+		/* Positive extra indicates more bytes than needed for the
+		 * message
+		 */
+
+		WARN_ON(extra > cand_len);
+
+		eaten += (cand_len - extra);
+
+		/* Hurray, we have a new message! */
+		cancel_delayed_work(&strp->msg_timer_work);
+		strp->skb_head = NULL;
+		strp->need_bytes = 0;
+		STRP_STATS_INCR(strp->stats.msgs);
+
+		/* Give skb to upper layer */
+		strp->cb.rcv_msg(strp, head);
+
+		if (unlikely(strp->paused)) {
+			/* Upper layer paused strp */
+			break;
+		}
+	}
+
+	if (cloned_orig)
+		kfree_skb(orig_skb);
+
+	STRP_STATS_ADD(strp->stats.bytes, eaten);
+
+	return eaten;
+}
+
+int strp_process(struct strparser *strp, struct sk_buff *orig_skb,
+		 unsigned int orig_offset, size_t orig_len,
+		 size_t max_msg_size, long timeo)
+{
+	read_descriptor_t desc; /* Dummy arg to strp_recv */
+
+	desc.arg.data = strp;
+
+	return __strp_recv(&desc, orig_skb, orig_offset, orig_len,
+			   max_msg_size, timeo);
+}
+EXPORT_SYMBOL_GPL(strp_process);
+
+static int strp_recv(read_descriptor_t *desc, struct sk_buff *orig_skb,
+		     unsigned int orig_offset, size_t orig_len)
+{
+	struct strparser *strp = (struct strparser *)desc->arg.data;
+
+	return __strp_recv(desc, orig_skb, orig_offset, orig_len,
+			   strp->sk->sk_rcvbuf, READ_ONCE(strp->sk->sk_rcvtimeo));
+}
+
+static int default_read_sock_done(struct strparser *strp, int err)
+{
+	return err;
+}
+
+/* Called with lock held on lower socket */
+static int strp_read_sock(struct strparser *strp)
+{
+	struct socket *sock = strp->sk->sk_socket;
+	read_descriptor_t desc;
+
+	if (unlikely(!sock || !sock->ops))
+		return -EBUSY;
+
+	if (unlikely(!strp->cb.read_sock && !sock->ops->read_sock))
+		return -EBUSY;
+
+	desc.arg.data = strp;
+	desc.error = 0;
+	desc.count = 1; /* give more than one skb per call */
+
+	/* sk should be locked here, so okay to do read_sock */
+	if (strp->cb.read_sock)
+		strp->cb.read_sock(strp, &desc, strp_recv);
+	else
+		sock->ops->read_sock(strp->sk, &desc, strp_recv);
+
+	desc.error = strp->cb.read_sock_done(strp, desc.error);
+
+	return desc.error;
+}
+
+/* Lower sock lock held */
+void strp_data_ready(struct strparser *strp)
+{
+	if (unlikely(strp->stopped) || strp->paused)
+		return;
+
+	/* This check is needed to synchronize with do_strp_work.
+	 * do_strp_work acquires a process lock (lock_sock) whereas
+	 * the lock held here is bh_lock_sock. The two locks can be
+	 * held by different threads at the same time, but bh_lock_sock
+	 * allows a thread in BH context to safely check if the process
+	 * lock is held. In this case, if the lock is held, queue work.
+	 */
+	if (sock_owned_by_user_nocheck(strp->sk)) {
+		queue_work(strp_wq, &strp->work);
+		return;
+	}
+
+	if (strp->need_bytes) {
+		if (strp_peek_len(strp) < strp->need_bytes)
+			return;
+	}
+
+	if (strp_read_sock(strp) == -ENOMEM)
+		queue_work(strp_wq, &strp->work);
+}
+EXPORT_SYMBOL_GPL(strp_data_ready);
+
+static void do_strp_work(struct strparser *strp)
+{
+	/* We need the read lock to synchronize with strp_data_ready. We
+	 * need the socket lock for calling strp_read_sock.
+	 */
+	strp->cb.lock(strp);
+
+	if (unlikely(strp->stopped))
+		goto out;
+
+	if (strp->paused)
+		goto out;
+
+	if (strp_read_sock(strp) == -ENOMEM)
+		queue_work(strp_wq, &strp->work);
+
+out:
+	strp->cb.unlock(strp);
+}
+
+static void strp_work(struct work_struct *w)
+{
+	do_strp_work(container_of(w, struct strparser, work));
+}
+
+static void strp_msg_timeout(struct work_struct *w)
+{
+	struct strparser *strp = container_of(w, struct strparser,
+					      msg_timer_work.work);
+
+	/* Message assembly timed out */
+	STRP_STATS_INCR(strp->stats.msg_timeouts);
+	strp->cb.lock(strp);
+	strp->cb.abort_parser(strp, -ETIMEDOUT);
+	strp->cb.unlock(strp);
+}
+
+static void strp_sock_lock(struct strparser *strp)
+{
+	lock_sock(strp->sk);
+}
+
+static void strp_sock_unlock(struct strparser *strp)
+{
+	release_sock(strp->sk);
+}
+
+int strp_init(struct strparser *strp, struct sock *sk,
+	      const struct strp_callbacks *cb)
+{
+
+	if (!cb || !cb->rcv_msg || !cb->parse_msg)
+		return -EINVAL;
+
+	/* The sk (sock) arg determines the mode of the stream parser.
+	 *
+	 * If the sock is set then the strparser is in receive callback mode.
+	 * The upper layer calls strp_data_ready to kick receive processing
+	 * and strparser calls the read_sock function on the socket to
+	 * get packets.
+	 *
+	 * If the sock is not set then the strparser is in general mode.
+	 * The upper layer calls strp_process for each skb to be parsed.
+	 */
+
+	if (!sk) {
+		if (!cb->lock || !cb->unlock)
+			return -EINVAL;
+	}
+
+	memset(strp, 0, sizeof(*strp));
+
+	strp->sk = sk;
+
+	strp->cb.lock = cb->lock ? : strp_sock_lock;
+	strp->cb.unlock = cb->unlock ? : strp_sock_unlock;
+	strp->cb.rcv_msg = cb->rcv_msg;
+	strp->cb.parse_msg = cb->parse_msg;
+	strp->cb.read_sock = cb->read_sock;
+	strp->cb.read_sock_done = cb->read_sock_done ? : default_read_sock_done;
+	strp->cb.abort_parser = cb->abort_parser ? : strp_abort_strp;
+
+	INIT_DELAYED_WORK(&strp->msg_timer_work, strp_msg_timeout);
+	INIT_WORK(&strp->work, strp_work);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(strp_init);
+
+void strp_unpause(struct strparser *strp)
+{
+	strp->paused = 0;
+
+	/* Sync setting paused with RX work */
+	smp_mb();
+
+	queue_work(strp_wq, &strp->work);
+}
+EXPORT_SYMBOL_GPL(strp_unpause);
+
+/* strp must already be stopped so that strp_recv will no longer be called.
+ * Note that strp_done is not called with the lower socket held.
+ */
+void strp_done(struct strparser *strp)
+{
+	WARN_ON(!strp->stopped);
+
+	cancel_delayed_work_sync(&strp->msg_timer_work);
+	cancel_work_sync(&strp->work);
+
+	if (strp->skb_head) {
+		kfree_skb(strp->skb_head);
+		strp->skb_head = NULL;
+	}
+}
+EXPORT_SYMBOL_GPL(strp_done);
+
+void strp_stop(struct strparser *strp)
+{
+	strp->stopped = 1;
+}
+EXPORT_SYMBOL_GPL(strp_stop);
+
+void strp_check_rcv(struct strparser *strp)
+{
+	queue_work(strp_wq, &strp->work);
+}
+EXPORT_SYMBOL_GPL(strp_check_rcv);
+
+static int __init strp_dev_init(void)
+{
+	BUILD_BUG_ON(sizeof(struct sk_skb_cb) >
+		     sizeof_field(struct sk_buff, cb));
+
+	strp_wq = create_singlethread_workqueue("kstrp");
+	if (unlikely(!strp_wq))
+		return -ENOMEM;
+
+	return 0;
+}
+device_initcall(strp_dev_init);