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, 166 insertions, 0 deletions

diff --git a/fs/btrfs/lru_cache.c b/fs/btrfs/lru_cache.c
new file mode 100644
index 000000000..f5a053b90
--- /dev/null
+++ b/fs/btrfs/lru_cache.c
@@ -0,0 +1,166 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/mm.h>
+#include "lru_cache.h"
+#include "messages.h"
+
+/*
+ * Initialize a cache object.
+ *
+ * @cache:      The cache.
+ * @max_size:   Maximum size (number of entries) for the cache.
+ *              Use 0 for unlimited size, it's the user's responsibility to
+ *              trim the cache in that case.
+ */
+void btrfs_lru_cache_init(struct btrfs_lru_cache *cache, unsigned int max_size)
+{
+	INIT_LIST_HEAD(&cache->lru_list);
+	mt_init(&cache->entries);
+	cache->size = 0;
+	cache->max_size = max_size;
+}
+
+static struct btrfs_lru_cache_entry *match_entry(struct list_head *head, u64 key,
+						 u64 gen)
+{
+	struct btrfs_lru_cache_entry *entry;
+
+	list_for_each_entry(entry, head, list) {
+		if (entry->key == key && entry->gen == gen)
+			return entry;
+	}
+
+	return NULL;
+}
+
+/*
+ * Lookup for an entry in the cache.
+ *
+ * @cache:      The cache.
+ * @key:        The key of the entry we are looking for.
+ * @gen:        Generation associated to the key.
+ *
+ * Returns the entry associated with the key or NULL if none found.
+ */
+struct btrfs_lru_cache_entry *btrfs_lru_cache_lookup(struct btrfs_lru_cache *cache,
+						     u64 key, u64 gen)
+{
+	struct list_head *head;
+	struct btrfs_lru_cache_entry *entry;
+
+	head = mtree_load(&cache->entries, key);
+	if (!head)
+		return NULL;
+
+	entry = match_entry(head, key, gen);
+	if (entry)
+		list_move_tail(&entry->lru_list, &cache->lru_list);
+
+	return entry;
+}
+
+/*
+ * Remove an entry from the cache.
+ *
+ * @cache:     The cache to remove from.
+ * @entry:     The entry to remove from the cache.
+ *
+ * Note: this also frees the memory used by the entry.
+ */
+void btrfs_lru_cache_remove(struct btrfs_lru_cache *cache,
+			    struct btrfs_lru_cache_entry *entry)
+{
+	struct list_head *prev = entry->list.prev;
+
+	ASSERT(cache->size > 0);
+	ASSERT(!mtree_empty(&cache->entries));
+
+	list_del(&entry->list);
+	list_del(&entry->lru_list);
+
+	if (list_empty(prev)) {
+		struct list_head *head;
+
+		/*
+		 * If previous element in the list entry->list is now empty, it
+		 * means it's a head entry not pointing to any cached entries,
+		 * so remove it from the maple tree and free it.
+		 */
+		head = mtree_erase(&cache->entries, entry->key);
+		ASSERT(head == prev);
+		kfree(head);
+	}
+
+	kfree(entry);
+	cache->size--;
+}
+
+/*
+ * Store an entry in the cache.
+ *
+ * @cache:      The cache.
+ * @entry:      The entry to store.
+ *
+ * Returns 0 on success and < 0 on error.
+ */
+int btrfs_lru_cache_store(struct btrfs_lru_cache *cache,
+			  struct btrfs_lru_cache_entry *new_entry,
+			  gfp_t gfp)
+{
+	const u64 key = new_entry->key;
+	struct list_head *head;
+	int ret;
+
+	head = kmalloc_obj(*head, gfp);
+	if (!head)
+		return -ENOMEM;
+
+	ret = mtree_insert(&cache->entries, key, head, gfp);
+	if (ret == 0) {
+		INIT_LIST_HEAD(head);
+		list_add_tail(&new_entry->list, head);
+	} else if (ret == -EEXIST) {
+		kfree(head);
+		head = mtree_load(&cache->entries, key);
+		ASSERT(head != NULL);
+		if (match_entry(head, key, new_entry->gen) != NULL)
+			return -EEXIST;
+		list_add_tail(&new_entry->list, head);
+	} else if (ret < 0) {
+		kfree(head);
+		return ret;
+	}
+
+	if (cache->max_size > 0 && cache->size == cache->max_size) {
+		struct btrfs_lru_cache_entry *lru_entry;
+
+		lru_entry = list_first_entry(&cache->lru_list,
+					     struct btrfs_lru_cache_entry,
+					     lru_list);
+		btrfs_lru_cache_remove(cache, lru_entry);
+	}
+
+	list_add_tail(&new_entry->lru_list, &cache->lru_list);
+	cache->size++;
+
+	return 0;
+}
+
+/*
+ * Empty a cache.
+ *
+ * @cache:     The cache to empty.
+ *
+ * Removes all entries from the cache.
+ */
+void btrfs_lru_cache_clear(struct btrfs_lru_cache *cache)
+{
+	struct btrfs_lru_cache_entry *entry;
+	struct btrfs_lru_cache_entry *tmp;
+
+	list_for_each_entry_safe(entry, tmp, &cache->lru_list, lru_list)
+		btrfs_lru_cache_remove(cache, entry);
+
+	ASSERT(cache->size == 0);
+	ASSERT(mtree_empty(&cache->entries));
+}