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, 119 insertions, 0 deletions

diff --git a/lib/zlib_inflate/inflate.h b/lib/zlib_inflate/inflate.h
new file mode 100644
index 000000000..f79337ddf
--- /dev/null
+++ b/lib/zlib_inflate/inflate.h
@@ -0,0 +1,119 @@
+#ifndef INFLATE_H
+#define INFLATE_H
+
+/* inflate.h -- internal inflate state definition
+ * Copyright (C) 1995-2004 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+#include "inftrees.h"
+
+/* Possible inflate modes between inflate() calls */
+typedef enum {
+    HEAD,       /* i: waiting for magic header */
+    FLAGS,      /* i: waiting for method and flags (gzip) */
+    TIME,       /* i: waiting for modification time (gzip) */
+    OS,         /* i: waiting for extra flags and operating system (gzip) */
+    EXLEN,      /* i: waiting for extra length (gzip) */
+    EXTRA,      /* i: waiting for extra bytes (gzip) */
+    NAME,       /* i: waiting for end of file name (gzip) */
+    COMMENT,    /* i: waiting for end of comment (gzip) */
+    HCRC,       /* i: waiting for header crc (gzip) */
+    DICTID,     /* i: waiting for dictionary check value */
+    DICT,       /* waiting for inflateSetDictionary() call */
+        TYPE,       /* i: waiting for type bits, including last-flag bit */
+        TYPEDO,     /* i: same, but skip check to exit inflate on new block */
+        STORED,     /* i: waiting for stored size (length and complement) */
+        COPY,       /* i/o: waiting for input or output to copy stored block */
+        TABLE,      /* i: waiting for dynamic block table lengths */
+        LENLENS,    /* i: waiting for code length code lengths */
+        CODELENS,   /* i: waiting for length/lit and distance code lengths */
+            LEN,        /* i: waiting for length/lit code */
+            LENEXT,     /* i: waiting for length extra bits */
+            DIST,       /* i: waiting for distance code */
+            DISTEXT,    /* i: waiting for distance extra bits */
+            MATCH,      /* o: waiting for output space to copy string */
+            LIT,        /* o: waiting for output space to write literal */
+    CHECK,      /* i: waiting for 32-bit check value */
+    LENGTH,     /* i: waiting for 32-bit length (gzip) */
+    DONE,       /* finished check, done -- remain here until reset */
+    BAD,        /* got a data error -- remain here until reset */
+    MEM,        /* got an inflate() memory error -- remain here until reset */
+    SYNC        /* looking for synchronization bytes to restart inflate() */
+} inflate_mode;
+
+/*
+    State transitions between above modes -
+
+    (most modes can go to the BAD or MEM mode -- not shown for clarity)
+
+    Process header:
+        HEAD -> (gzip) or (zlib)
+        (gzip) -> FLAGS -> TIME -> OS -> EXLEN -> EXTRA -> NAME
+        NAME -> COMMENT -> HCRC -> TYPE
+        (zlib) -> DICTID or TYPE
+        DICTID -> DICT -> TYPE
+    Read deflate blocks:
+            TYPE -> STORED or TABLE or LEN or CHECK
+            STORED -> COPY -> TYPE
+            TABLE -> LENLENS -> CODELENS -> LEN
+    Read deflate codes:
+                LEN -> LENEXT or LIT or TYPE
+                LENEXT -> DIST -> DISTEXT -> MATCH -> LEN
+                LIT -> LEN
+    Process trailer:
+        CHECK -> LENGTH -> DONE
+ */
+
+/* state maintained between inflate() calls.  Approximately 7K bytes. */
+struct inflate_state {
+    inflate_mode mode;          /* current inflate mode */
+    int last;                   /* true if processing last block */
+    int wrap;                   /* bit 0 true for zlib, bit 1 true for gzip */
+    int havedict;               /* true if dictionary provided */
+    int flags;                  /* gzip header method and flags (0 if zlib) */
+    unsigned dmax;              /* zlib header max distance (INFLATE_STRICT) */
+    unsigned long check;        /* protected copy of check value */
+    unsigned long total;        /* protected copy of output count */
+ /*   gz_headerp head; */           /* where to save gzip header information */
+        /* sliding window */
+    unsigned wbits;             /* log base 2 of requested window size */
+    unsigned wsize;             /* window size or zero if not using window */
+    unsigned whave;             /* valid bytes in the window */
+    unsigned write;             /* window write index */
+    unsigned char *window;  /* allocated sliding window, if needed */
+        /* bit accumulator */
+    unsigned long hold;         /* input bit accumulator */
+    unsigned bits;              /* number of bits in "in" */
+        /* for string and stored block copying */
+    unsigned length;            /* literal or length of data to copy */
+    unsigned offset;            /* distance back to copy string from */
+        /* for table and code decoding */
+    unsigned extra;             /* extra bits needed */
+        /* fixed and dynamic code tables */
+    code const *lencode;    /* starting table for length/literal codes */
+    code const *distcode;   /* starting table for distance codes */
+    unsigned lenbits;           /* index bits for lencode */
+    unsigned distbits;          /* index bits for distcode */
+        /* dynamic table building */
+    unsigned ncode;             /* number of code length code lengths */
+    unsigned nlen;              /* number of length code lengths */
+    unsigned ndist;             /* number of distance code lengths */
+    unsigned have;              /* number of code lengths in lens[] */
+    code *next;             /* next available space in codes[] */
+    unsigned short lens[320];   /* temporary storage for code lengths */
+    unsigned short work[288];   /* work area for code table building */
+    code codes[ENOUGH];         /* space for code tables */
+};
+
+/* Reverse the bytes in a 32-bit value */
+#define REVERSE(q) \
+    ((((q) >> 24) & 0xff) + (((q) >> 8) & 0xff00) + \
+     (((q) & 0xff00) << 8) + (((q) & 0xff) << 24))
+
+#endif