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, 169 insertions, 0 deletions

diff --git a/drivers/android/dbitmap.h b/drivers/android/dbitmap.h
new file mode 100644
index 000000000..c7299ce8b
--- /dev/null
+++ b/drivers/android/dbitmap.h
@@ -0,0 +1,169 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright 2024 Google LLC
+ *
+ * dbitmap - dynamically sized bitmap library.
+ *
+ * Used by the binder driver to optimize the allocation of the smallest
+ * available descriptor ID. Each bit in the bitmap represents the state
+ * of an ID.
+ *
+ * A dbitmap can grow or shrink as needed. This part has been designed
+ * considering that users might need to briefly release their locks in
+ * order to allocate memory for the new bitmap. These operations then,
+ * are verified to determine if the grow or shrink is sill valid.
+ *
+ * This library does not provide protection against concurrent access
+ * by itself. Binder uses the proc->outer_lock for this purpose.
+ */
+
+#ifndef _LINUX_DBITMAP_H
+#define _LINUX_DBITMAP_H
+#include <linux/bitmap.h>
+
+#define NBITS_MIN	BITS_PER_TYPE(unsigned long)
+
+struct dbitmap {
+	unsigned int nbits;
+	unsigned long *map;
+};
+
+static inline int dbitmap_enabled(struct dbitmap *dmap)
+{
+	return !!dmap->nbits;
+}
+
+static inline void dbitmap_free(struct dbitmap *dmap)
+{
+	dmap->nbits = 0;
+	kfree(dmap->map);
+	dmap->map = NULL;
+}
+
+/* Returns the nbits that a dbitmap can shrink to, 0 if not possible. */
+static inline unsigned int dbitmap_shrink_nbits(struct dbitmap *dmap)
+{
+	unsigned int bit;
+
+	if (dmap->nbits <= NBITS_MIN)
+		return 0;
+
+	/*
+	 * Determine if the bitmap can shrink based on the position of
+	 * its last set bit. If the bit is within the first quarter of
+	 * the bitmap then shrinking is possible. In this case, the
+	 * bitmap should shrink to half its current size.
+	 */
+	bit = find_last_bit(dmap->map, dmap->nbits);
+	if (bit < (dmap->nbits >> 2))
+		return dmap->nbits >> 1;
+
+	/* find_last_bit() returns dmap->nbits when no bits are set. */
+	if (bit == dmap->nbits)
+		return NBITS_MIN;
+
+	return 0;
+}
+
+/* Replace the internal bitmap with a new one of different size */
+static inline void
+dbitmap_replace(struct dbitmap *dmap, unsigned long *new, unsigned int nbits)
+{
+	bitmap_copy(new, dmap->map, min(dmap->nbits, nbits));
+	kfree(dmap->map);
+	dmap->map = new;
+	dmap->nbits = nbits;
+}
+
+static inline void
+dbitmap_shrink(struct dbitmap *dmap, unsigned long *new, unsigned int nbits)
+{
+	if (!new)
+		return;
+
+	/*
+	 * Verify that shrinking to @nbits is still possible. The @new
+	 * bitmap might have been allocated without locks, so this call
+	 * could now be outdated. In this case, free @new and move on.
+	 */
+	if (!dbitmap_enabled(dmap) || dbitmap_shrink_nbits(dmap) != nbits) {
+		kfree(new);
+		return;
+	}
+
+	dbitmap_replace(dmap, new, nbits);
+}
+
+/* Returns the nbits that a dbitmap can grow to. */
+static inline unsigned int dbitmap_grow_nbits(struct dbitmap *dmap)
+{
+	return dmap->nbits << 1;
+}
+
+static inline void
+dbitmap_grow(struct dbitmap *dmap, unsigned long *new, unsigned int nbits)
+{
+	/*
+	 * Verify that growing to @nbits is still possible. The @new
+	 * bitmap might have been allocated without locks, so this call
+	 * could now be outdated. In this case, free @new and move on.
+	 */
+	if (!dbitmap_enabled(dmap) || nbits <= dmap->nbits) {
+		kfree(new);
+		return;
+	}
+
+	/*
+	 * Check for ENOMEM after confirming the grow operation is still
+	 * required. This ensures we only disable the dbitmap when it's
+	 * necessary. Once the dbitmap is disabled, binder will fallback
+	 * to slow_desc_lookup_olocked().
+	 */
+	if (!new) {
+		dbitmap_free(dmap);
+		return;
+	}
+
+	dbitmap_replace(dmap, new, nbits);
+}
+
+/*
+ * Finds and sets the next zero bit in the bitmap. Upon success @bit
+ * is populated with the index and 0 is returned. Otherwise, -ENOSPC
+ * is returned to indicate that a dbitmap_grow() is needed.
+ */
+static inline int
+dbitmap_acquire_next_zero_bit(struct dbitmap *dmap, unsigned long offset,
+			      unsigned long *bit)
+{
+	unsigned long n;
+
+	n = find_next_zero_bit(dmap->map, dmap->nbits, offset);
+	if (n == dmap->nbits)
+		return -ENOSPC;
+
+	*bit = n;
+	set_bit(n, dmap->map);
+
+	return 0;
+}
+
+static inline void
+dbitmap_clear_bit(struct dbitmap *dmap, unsigned long bit)
+{
+	clear_bit(bit, dmap->map);
+}
+
+static inline int dbitmap_init(struct dbitmap *dmap)
+{
+	dmap->map = bitmap_zalloc(NBITS_MIN, GFP_KERNEL);
+	if (!dmap->map) {
+		dmap->nbits = 0;
+		return -ENOMEM;
+	}
+
+	dmap->nbits = NBITS_MIN;
+
+	return 0;
+}
+#endif