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, 205 insertions, 0 deletions

diff --git a/drivers/md/dm-vdo/wait-queue.c b/drivers/md/dm-vdo/wait-queue.c
new file mode 100644
index 000000000..f81ed0cee
--- /dev/null
+++ b/drivers/md/dm-vdo/wait-queue.c
@@ -0,0 +1,205 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2023 Red Hat
+ */
+
+#include "wait-queue.h"
+
+#include <linux/device-mapper.h>
+
+#include "permassert.h"
+
+#include "status-codes.h"
+
+/**
+ * vdo_waitq_enqueue_waiter() - Add a waiter to the tail end of a waitq.
+ * @waitq: The vdo_wait_queue to which to add the waiter.
+ * @waiter: The waiter to add to the waitq.
+ *
+ * The waiter must not already be waiting in a waitq.
+ */
+void vdo_waitq_enqueue_waiter(struct vdo_wait_queue *waitq, struct vdo_waiter *waiter)
+{
+	BUG_ON(waiter->next_waiter != NULL);
+
+	if (waitq->last_waiter == NULL) {
+		/*
+		 * The waitq is empty, so form the initial circular list by self-linking the
+		 * initial waiter.
+		 */
+		waiter->next_waiter = waiter;
+	} else {
+		/* Splice the new waiter in at the end of the waitq. */
+		waiter->next_waiter = waitq->last_waiter->next_waiter;
+		waitq->last_waiter->next_waiter = waiter;
+	}
+
+	/* In both cases, the waiter we added to the list becomes the last waiter. */
+	waitq->last_waiter = waiter;
+	waitq->length += 1;
+}
+
+/**
+ * vdo_waitq_transfer_all_waiters() - Transfer all waiters from one waitq to
+ *                                    a second waitq, emptying the first waitq.
+ * @from_waitq: The waitq containing the waiters to move.
+ * @to_waitq: The waitq that will receive the waiters from the first waitq.
+ */
+void vdo_waitq_transfer_all_waiters(struct vdo_wait_queue *from_waitq,
+				    struct vdo_wait_queue *to_waitq)
+{
+	/* If the source waitq is empty, there's nothing to do. */
+	if (!vdo_waitq_has_waiters(from_waitq))
+		return;
+
+	if (vdo_waitq_has_waiters(to_waitq)) {
+		/*
+		 * Both are non-empty. Splice the two circular lists together
+		 * by swapping the next (head) pointers in the list tails.
+		 */
+		struct vdo_waiter *from_head = from_waitq->last_waiter->next_waiter;
+		struct vdo_waiter *to_head = to_waitq->last_waiter->next_waiter;
+
+		to_waitq->last_waiter->next_waiter = from_head;
+		from_waitq->last_waiter->next_waiter = to_head;
+	}
+
+	to_waitq->last_waiter = from_waitq->last_waiter;
+	to_waitq->length += from_waitq->length;
+	vdo_waitq_init(from_waitq);
+}
+
+/**
+ * vdo_waitq_notify_all_waiters() - Notify all the entries waiting in a waitq.
+ * @waitq: The vdo_wait_queue containing the waiters to notify.
+ * @callback: The function to call to notify each waiter, or NULL to invoke the callback field
+ *            registered in each waiter.
+ * @context: The context to pass to the callback function.
+ *
+ * Notifies all the entries waiting in a waitq to continue execution by invoking a callback
+ * function on each of them in turn. The waitq is copied and emptied before invoking any callbacks,
+ * and only the waiters that were in the waitq at the start of the call will be notified.
+ */
+void vdo_waitq_notify_all_waiters(struct vdo_wait_queue *waitq,
+				  vdo_waiter_callback_fn callback, void *context)
+{
+	/*
+	 * Copy and empty the waitq first, avoiding the possibility of an infinite
+	 * loop if entries are returned to the waitq by the callback function.
+	 */
+	struct vdo_wait_queue waiters;
+
+	vdo_waitq_init(&waiters);
+	vdo_waitq_transfer_all_waiters(waitq, &waiters);
+
+	/* Drain the copied waitq, invoking the callback on every entry. */
+	while (vdo_waitq_has_waiters(&waiters))
+		vdo_waitq_notify_next_waiter(&waiters, callback, context);
+}
+
+/**
+ * vdo_waitq_get_first_waiter() - Return the waiter that is at the head end of a waitq.
+ * @waitq: The vdo_wait_queue from which to get the first waiter.
+ *
+ * Return: The first (oldest) waiter in the waitq, or NULL if the waitq is empty.
+ */
+struct vdo_waiter *vdo_waitq_get_first_waiter(const struct vdo_wait_queue *waitq)
+{
+	struct vdo_waiter *last_waiter = waitq->last_waiter;
+
+	if (last_waiter == NULL) {
+		/* There are no waiters, so we're done. */
+		return NULL;
+	}
+
+	/* The waitq is circular, so the last entry links to the head of the waitq. */
+	return last_waiter->next_waiter;
+}
+
+/**
+ * vdo_waitq_dequeue_matching_waiters() - Remove all waiters that match based on the specified
+ *                                        matching method and append them to a vdo_wait_queue.
+ * @waitq: The vdo_wait_queue to process.
+ * @waiter_match: The method to determine matching.
+ * @match_context: Contextual info for the match method.
+ * @matched_waitq: A wait_waitq to store matches.
+ */
+void vdo_waitq_dequeue_matching_waiters(struct vdo_wait_queue *waitq,
+					vdo_waiter_match_fn waiter_match,
+					void *match_context,
+					struct vdo_wait_queue *matched_waitq)
+{
+	struct vdo_wait_queue iteration_waitq;
+
+	vdo_waitq_init(&iteration_waitq);
+	vdo_waitq_transfer_all_waiters(waitq, &iteration_waitq);
+
+	while (vdo_waitq_has_waiters(&iteration_waitq)) {
+		struct vdo_waiter *waiter = vdo_waitq_dequeue_waiter(&iteration_waitq);
+
+		vdo_waitq_enqueue_waiter((waiter_match(waiter, match_context) ?
+					  matched_waitq : waitq), waiter);
+	}
+}
+
+/**
+ * vdo_waitq_dequeue_waiter() - Remove the first (oldest) waiter from a waitq.
+ * @waitq: The vdo_wait_queue from which to remove the first entry.
+ *
+ * The caller will be responsible for waking the waiter by continuing its
+ * execution appropriately.
+ *
+ * Return: The first (oldest) waiter in the waitq, or NULL if the waitq is empty.
+ */
+struct vdo_waiter *vdo_waitq_dequeue_waiter(struct vdo_wait_queue *waitq)
+{
+	struct vdo_waiter *first_waiter = vdo_waitq_get_first_waiter(waitq);
+	struct vdo_waiter *last_waiter = waitq->last_waiter;
+
+	if (first_waiter == NULL)
+		return NULL;
+
+	if (first_waiter == last_waiter) {
+		/* The waitq has a single entry, so empty it by nulling the tail. */
+		waitq->last_waiter = NULL;
+	} else {
+		/*
+		 * The waitq has multiple waiters, so splice the first waiter out
+		 * of the circular waitq.
+		 */
+		last_waiter->next_waiter = first_waiter->next_waiter;
+	}
+
+	/* The waiter is no longer in a waitq. */
+	first_waiter->next_waiter = NULL;
+	waitq->length -= 1;
+
+	return first_waiter;
+}
+
+/**
+ * vdo_waitq_notify_next_waiter() - Notify the next entry waiting in a waitq.
+ * @waitq: The vdo_wait_queue containing the waiter to notify.
+ * @callback: The function to call to notify the waiter, or NULL to invoke the callback field
+ *            registered in the waiter.
+ * @context: The context to pass to the callback function.
+ *
+ * Notifies the next entry waiting in a waitq to continue execution by invoking a callback function
+ * on it after removing it from the waitq.
+ *
+ * Return: true if there was a waiter in the waitq.
+ */
+bool vdo_waitq_notify_next_waiter(struct vdo_wait_queue *waitq,
+				  vdo_waiter_callback_fn callback, void *context)
+{
+	struct vdo_waiter *waiter = vdo_waitq_dequeue_waiter(waitq);
+
+	if (waiter == NULL)
+		return false;
+
+	if (callback == NULL)
+		callback = waiter->callback;
+	callback(waiter, context);
+
+	return true;
+}