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, 186 insertions, 0 deletions

diff --git a/arch/alpha/lib/checksum.c b/arch/alpha/lib/checksum.c
new file mode 100644
index 000000000..27b2a9edf
--- /dev/null
+++ b/arch/alpha/lib/checksum.c
@@ -0,0 +1,186 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * arch/alpha/lib/checksum.c
+ *
+ * This file contains network checksum routines that are better done
+ * in an architecture-specific manner due to speed..
+ * Comments in other versions indicate that the algorithms are from RFC1071
+ *
+ * accelerated versions (and 21264 assembly versions ) contributed by
+ *	Rick Gorton	<rick.gorton@alpha-processor.com>
+ */
+ 
+#include <linux/module.h>
+#include <linux/string.h>
+#include <net/checksum.h>
+
+#include <asm/byteorder.h>
+#include <asm/checksum.h>
+
+static inline unsigned short from64to16(unsigned long x)
+{
+	/* Using extract instructions is a bit more efficient
+	   than the original shift/bitmask version.  */
+
+	union {
+		unsigned long	ul;
+		unsigned int	ui[2];
+		unsigned short	us[4];
+	} in_v, tmp_v, out_v;
+
+	in_v.ul = x;
+	tmp_v.ul = (unsigned long) in_v.ui[0] + (unsigned long) in_v.ui[1];
+
+	/* Since the bits of tmp_v.sh[3] are going to always be zero,
+	   we don't have to bother to add that in.  */
+	out_v.ul = (unsigned long) tmp_v.us[0] + (unsigned long) tmp_v.us[1]
+			+ (unsigned long) tmp_v.us[2];
+
+	/* Similarly, out_v.us[2] is always zero for the final add.  */
+	return out_v.us[0] + out_v.us[1];
+}
+
+/*
+ * computes the checksum of the TCP/UDP pseudo-header
+ * returns a 16-bit checksum, already complemented.
+ */
+__sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr,
+			  __u32 len, __u8 proto, __wsum sum)
+{
+	return (__force __sum16)~from64to16(
+		(__force u64)saddr + (__force u64)daddr +
+		(__force u64)sum + ((len + proto) << 8));
+}
+EXPORT_SYMBOL(csum_tcpudp_magic);
+
+__wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr,
+			  __u32 len, __u8 proto, __wsum sum)
+{
+	unsigned long result;
+
+	result = (__force u64)saddr + (__force u64)daddr +
+		 (__force u64)sum + ((len + proto) << 8);
+
+	/* Fold down to 32-bits so we don't lose in the typedef-less 
+	   network stack.  */
+	/* 64 to 33 */
+	result = (result & 0xffffffff) + (result >> 32);
+	/* 33 to 32 */
+	result = (result & 0xffffffff) + (result >> 32);
+	return (__force __wsum)result;
+}
+EXPORT_SYMBOL(csum_tcpudp_nofold);
+
+/*
+ * Do a 64-bit checksum on an arbitrary memory area..
+ *
+ * This isn't a great routine, but it's not _horrible_ either. The
+ * inner loop could be unrolled a bit further, and there are better
+ * ways to do the carry, but this is reasonable.
+ */
+static inline unsigned long do_csum(const unsigned char * buff, int len)
+{
+	int odd, count;
+	unsigned long result = 0;
+
+	if (len <= 0)
+		goto out;
+	odd = 1 & (unsigned long) buff;
+	if (odd) {
+		result = *buff << 8;
+		len--;
+		buff++;
+	}
+	count = len >> 1;		/* nr of 16-bit words.. */
+	if (count) {
+		if (2 & (unsigned long) buff) {
+			result += *(unsigned short *) buff;
+			count--;
+			len -= 2;
+			buff += 2;
+		}
+		count >>= 1;		/* nr of 32-bit words.. */
+		if (count) {
+			if (4 & (unsigned long) buff) {
+				result += *(unsigned int *) buff;
+				count--;
+				len -= 4;
+				buff += 4;
+			}
+			count >>= 1;	/* nr of 64-bit words.. */
+			if (count) {
+				unsigned long carry = 0;
+				do {
+					unsigned long w = *(unsigned long *) buff;
+					count--;
+					buff += 8;
+					result += carry;
+					result += w;
+					carry = (w > result);
+				} while (count);
+				result += carry;
+				result = (result & 0xffffffff) + (result >> 32);
+			}
+			if (len & 4) {
+				result += *(unsigned int *) buff;
+				buff += 4;
+			}
+		}
+		if (len & 2) {
+			result += *(unsigned short *) buff;
+			buff += 2;
+		}
+	}
+	if (len & 1)
+		result += *buff;
+	result = from64to16(result);
+	if (odd)
+		result = ((result >> 8) & 0xff) | ((result & 0xff) << 8);
+out:
+	return result;
+}
+
+/*
+ *	This is a version of ip_compute_csum() optimized for IP headers,
+ *	which always checksum on 4 octet boundaries.
+ */
+__sum16 ip_fast_csum(const void *iph, unsigned int ihl)
+{
+	return (__force __sum16)~do_csum(iph,ihl*4);
+}
+EXPORT_SYMBOL(ip_fast_csum);
+
+/*
+ * computes the checksum of a memory block at buff, length len,
+ * and adds in "sum" (32-bit)
+ *
+ * returns a 32-bit number suitable for feeding into itself
+ * or csum_tcpudp_magic
+ *
+ * this function must be called with even lengths, except
+ * for the last fragment, which may be odd
+ *
+ * it's best to have buff aligned on a 32-bit boundary
+ */
+__wsum csum_partial(const void *buff, int len, __wsum sum)
+{
+	unsigned long result = do_csum(buff, len);
+
+	/* add in old sum, and carry.. */
+	result += (__force u32)sum;
+	/* 32+c bits -> 32 bits */
+	result = (result & 0xffffffff) + (result >> 32);
+	return (__force __wsum)result;
+}
+
+EXPORT_SYMBOL(csum_partial);
+
+/*
+ * this routine is used for miscellaneous IP-like checksums, mainly
+ * in icmp.c
+ */
+__sum16 ip_compute_csum(const void *buff, int len)
+{
+	return (__force __sum16)~from64to16(do_csum(buff,len));
+}
+EXPORT_SYMBOL(ip_compute_csum);