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, 241 insertions, 0 deletions

diff --git a/crypto/md4.c b/crypto/md4.c
new file mode 100644
index 000000000..55bf47e23
--- /dev/null
+++ b/crypto/md4.c
@@ -0,0 +1,241 @@
+/* 
+ * Cryptographic API.
+ *
+ * MD4 Message Digest Algorithm (RFC1320).
+ *
+ * Implementation derived from Andrew Tridgell and Steve French's
+ * CIFS MD4 implementation, and the cryptoapi implementation
+ * originally based on the public domain implementation written
+ * by Colin Plumb in 1993.
+ *
+ * Copyright (c) Andrew Tridgell 1997-1998.
+ * Modified by Steve French (sfrench@us.ibm.com) 2002
+ * Copyright (c) Cryptoapi developers.
+ * Copyright (c) 2002 David S. Miller (davem@redhat.com)
+ * Copyright (c) 2002 James Morris <jmorris@intercode.com.au>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ */
+#include <crypto/internal/hash.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <asm/byteorder.h>
+
+#define MD4_DIGEST_SIZE		16
+#define MD4_HMAC_BLOCK_SIZE	64
+#define MD4_BLOCK_WORDS		16
+#define MD4_HASH_WORDS		4
+
+struct md4_ctx {
+	u32 hash[MD4_HASH_WORDS];
+	u32 block[MD4_BLOCK_WORDS];
+	u64 byte_count;
+};
+
+static inline u32 lshift(u32 x, unsigned int s)
+{
+	x &= 0xFFFFFFFF;
+	return ((x << s) & 0xFFFFFFFF) | (x >> (32 - s));
+}
+
+static inline u32 F(u32 x, u32 y, u32 z)
+{
+	return (x & y) | ((~x) & z);
+}
+
+static inline u32 G(u32 x, u32 y, u32 z)
+{
+	return (x & y) | (x & z) | (y & z);
+}
+
+static inline u32 H(u32 x, u32 y, u32 z)
+{
+	return x ^ y ^ z;
+}
+
+#define ROUND1(a,b,c,d,k,s) (a = lshift(a + F(b,c,d) + k, s))
+#define ROUND2(a,b,c,d,k,s) (a = lshift(a + G(b,c,d) + k + (u32)0x5A827999,s))
+#define ROUND3(a,b,c,d,k,s) (a = lshift(a + H(b,c,d) + k + (u32)0x6ED9EBA1,s))
+
+static void md4_transform(u32 *hash, u32 const *in)
+{
+	u32 a, b, c, d;
+
+	a = hash[0];
+	b = hash[1];
+	c = hash[2];
+	d = hash[3];
+
+	ROUND1(a, b, c, d, in[0], 3);
+	ROUND1(d, a, b, c, in[1], 7);
+	ROUND1(c, d, a, b, in[2], 11);
+	ROUND1(b, c, d, a, in[3], 19);
+	ROUND1(a, b, c, d, in[4], 3);
+	ROUND1(d, a, b, c, in[5], 7);
+	ROUND1(c, d, a, b, in[6], 11);
+	ROUND1(b, c, d, a, in[7], 19);
+	ROUND1(a, b, c, d, in[8], 3);
+	ROUND1(d, a, b, c, in[9], 7);
+	ROUND1(c, d, a, b, in[10], 11);
+	ROUND1(b, c, d, a, in[11], 19);
+	ROUND1(a, b, c, d, in[12], 3);
+	ROUND1(d, a, b, c, in[13], 7);
+	ROUND1(c, d, a, b, in[14], 11);
+	ROUND1(b, c, d, a, in[15], 19);
+
+	ROUND2(a, b, c, d,in[ 0], 3);
+	ROUND2(d, a, b, c, in[4], 5);
+	ROUND2(c, d, a, b, in[8], 9);
+	ROUND2(b, c, d, a, in[12], 13);
+	ROUND2(a, b, c, d, in[1], 3);
+	ROUND2(d, a, b, c, in[5], 5);
+	ROUND2(c, d, a, b, in[9], 9);
+	ROUND2(b, c, d, a, in[13], 13);
+	ROUND2(a, b, c, d, in[2], 3);
+	ROUND2(d, a, b, c, in[6], 5);
+	ROUND2(c, d, a, b, in[10], 9);
+	ROUND2(b, c, d, a, in[14], 13);
+	ROUND2(a, b, c, d, in[3], 3);
+	ROUND2(d, a, b, c, in[7], 5);
+	ROUND2(c, d, a, b, in[11], 9);
+	ROUND2(b, c, d, a, in[15], 13);
+
+	ROUND3(a, b, c, d,in[ 0], 3);
+	ROUND3(d, a, b, c, in[8], 9);
+	ROUND3(c, d, a, b, in[4], 11);
+	ROUND3(b, c, d, a, in[12], 15);
+	ROUND3(a, b, c, d, in[2], 3);
+	ROUND3(d, a, b, c, in[10], 9);
+	ROUND3(c, d, a, b, in[6], 11);
+	ROUND3(b, c, d, a, in[14], 15);
+	ROUND3(a, b, c, d, in[1], 3);
+	ROUND3(d, a, b, c, in[9], 9);
+	ROUND3(c, d, a, b, in[5], 11);
+	ROUND3(b, c, d, a, in[13], 15);
+	ROUND3(a, b, c, d, in[3], 3);
+	ROUND3(d, a, b, c, in[11], 9);
+	ROUND3(c, d, a, b, in[7], 11);
+	ROUND3(b, c, d, a, in[15], 15);
+
+	hash[0] += a;
+	hash[1] += b;
+	hash[2] += c;
+	hash[3] += d;
+}
+
+static inline void md4_transform_helper(struct md4_ctx *ctx)
+{
+	le32_to_cpu_array(ctx->block, ARRAY_SIZE(ctx->block));
+	md4_transform(ctx->hash, ctx->block);
+}
+
+static int md4_init(struct shash_desc *desc)
+{
+	struct md4_ctx *mctx = shash_desc_ctx(desc);
+
+	mctx->hash[0] = 0x67452301;
+	mctx->hash[1] = 0xefcdab89;
+	mctx->hash[2] = 0x98badcfe;
+	mctx->hash[3] = 0x10325476;
+	mctx->byte_count = 0;
+
+	return 0;
+}
+
+static int md4_update(struct shash_desc *desc, const u8 *data, unsigned int len)
+{
+	struct md4_ctx *mctx = shash_desc_ctx(desc);
+	const u32 avail = sizeof(mctx->block) - (mctx->byte_count & 0x3f);
+
+	mctx->byte_count += len;
+
+	if (avail > len) {
+		memcpy((char *)mctx->block + (sizeof(mctx->block) - avail),
+		       data, len);
+		return 0;
+	}
+
+	memcpy((char *)mctx->block + (sizeof(mctx->block) - avail),
+	       data, avail);
+
+	md4_transform_helper(mctx);
+	data += avail;
+	len -= avail;
+
+	while (len >= sizeof(mctx->block)) {
+		memcpy(mctx->block, data, sizeof(mctx->block));
+		md4_transform_helper(mctx);
+		data += sizeof(mctx->block);
+		len -= sizeof(mctx->block);
+	}
+
+	memcpy(mctx->block, data, len);
+
+	return 0;
+}
+
+static int md4_final(struct shash_desc *desc, u8 *out)
+{
+	struct md4_ctx *mctx = shash_desc_ctx(desc);
+	const unsigned int offset = mctx->byte_count & 0x3f;
+	char *p = (char *)mctx->block + offset;
+	int padding = 56 - (offset + 1);
+
+	*p++ = 0x80;
+	if (padding < 0) {
+		memset(p, 0x00, padding + sizeof (u64));
+		md4_transform_helper(mctx);
+		p = (char *)mctx->block;
+		padding = 56;
+	}
+
+	memset(p, 0, padding);
+	mctx->block[14] = mctx->byte_count << 3;
+	mctx->block[15] = mctx->byte_count >> 29;
+	le32_to_cpu_array(mctx->block, (sizeof(mctx->block) -
+	                  sizeof(u64)) / sizeof(u32));
+	md4_transform(mctx->hash, mctx->block);
+	cpu_to_le32_array(mctx->hash, ARRAY_SIZE(mctx->hash));
+	memcpy(out, mctx->hash, sizeof(mctx->hash));
+	memset(mctx, 0, sizeof(*mctx));
+
+	return 0;
+}
+
+static struct shash_alg alg = {
+	.digestsize	=	MD4_DIGEST_SIZE,
+	.init		=	md4_init,
+	.update		=	md4_update,
+	.final		=	md4_final,
+	.descsize	=	sizeof(struct md4_ctx),
+	.base		=	{
+		.cra_name	 =	"md4",
+		.cra_driver_name =	"md4-generic",
+		.cra_blocksize	 =	MD4_HMAC_BLOCK_SIZE,
+		.cra_module	 =	THIS_MODULE,
+	}
+};
+
+static int __init md4_mod_init(void)
+{
+	return crypto_register_shash(&alg);
+}
+
+static void __exit md4_mod_fini(void)
+{
+	crypto_unregister_shash(&alg);
+}
+
+module_init(md4_mod_init);
+module_exit(md4_mod_fini);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("MD4 Message Digest Algorithm");
+MODULE_ALIAS_CRYPTO("md4");