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, 287 insertions, 0 deletions

diff --git a/drivers/tty/vt/conmakehash.c b/drivers/tty/vt/conmakehash.c
new file mode 100644
index 000000000..a931fcde7
--- /dev/null
+++ b/drivers/tty/vt/conmakehash.c
@@ -0,0 +1,287 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * conmakehash.c
+ *
+ * Create arrays for initializing the kernel folded tables (using a hash
+ * table turned out to be to limiting...)  Unfortunately we can't simply
+ * preinitialize the tables at compile time since kfree() cannot accept
+ * memory not allocated by kmalloc(), and doing our own memory management
+ * just for this seems like massive overkill.
+ *
+ * Copyright (C) 1995-1997 H. Peter Anvin
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <sysexits.h>
+#include <string.h>
+#include <ctype.h>
+
+#define MAX_FONTLEN 256
+
+typedef unsigned short unicode;
+
+static void usage(char *argv0)
+{
+  fprintf(stderr, "Usage: \n"
+         "        %s chartable [hashsize] [hashstep] [maxhashlevel]\n", argv0);
+  exit(EX_USAGE);
+}
+
+static int getunicode(char **p0)
+{
+  char *p = *p0;
+
+  while (*p == ' ' || *p == '\t')
+    p++;
+  if (*p != 'U' || p[1] != '+' ||
+      !isxdigit(p[2]) || !isxdigit(p[3]) || !isxdigit(p[4]) ||
+      !isxdigit(p[5]) || isxdigit(p[6]))
+    return -1;
+  *p0 = p+6;
+  return strtol(p+2,0,16);
+}
+
+unicode unitable[MAX_FONTLEN][255];
+				/* Massive overkill, but who cares? */
+int unicount[MAX_FONTLEN];
+
+static void addpair(int fp, int un)
+{
+  int i;
+
+  if ( un <= 0xfffe )
+    {
+      /* Check it isn't a duplicate */
+
+      for ( i = 0 ; i < unicount[fp] ; i++ )
+	if ( unitable[fp][i] == un )
+	  return;
+
+      /* Add to list */
+
+      if ( unicount[fp] > 254 )
+	{
+	  fprintf(stderr, "ERROR: Only 255 unicodes/glyph permitted!\n");
+	  exit(EX_DATAERR);
+	}
+
+      unitable[fp][unicount[fp]] = un;
+      unicount[fp]++;
+    }
+
+  /* otherwise: ignore */
+}
+
+int main(int argc, char *argv[])
+{
+  FILE *ctbl;
+  const char *tblname;
+  char buffer[65536];
+  int fontlen;
+  int i, nuni, nent;
+  int fp0, fp1, un0, un1;
+  char *p, *p1;
+
+  if ( argc < 2 || argc > 5 )
+    usage(argv[0]);
+
+  if ( !strcmp(argv[1],"-") )
+    {
+      ctbl = stdin;
+      tblname = "stdin";
+    }
+  else
+    {
+      ctbl = fopen(tblname = argv[1], "r");
+      if ( !ctbl )
+	{
+	  perror(tblname);
+	  exit(EX_NOINPUT);
+	}
+    }
+
+  /* For now we assume the default font is always 256 characters. */
+  fontlen = 256;
+
+  /* Initialize table */
+
+  for ( i = 0 ; i < fontlen ; i++ )
+    unicount[i] = 0;
+
+  /* Now we come to the tricky part.  Parse the input table. */
+
+  while ( fgets(buffer, sizeof(buffer), ctbl) != NULL )
+    {
+      if ( (p = strchr(buffer, '\n')) != NULL )
+	*p = '\0';
+      else
+	fprintf(stderr, "%s: Warning: line too long\n", tblname);
+
+      p = buffer;
+
+/*
+ * Syntax accepted:
+ *	<fontpos>	<unicode> <unicode> ...
+ *	<range>		idem
+ *	<range>		<unicode range>
+ *
+ * where <range> ::= <fontpos>-<fontpos>
+ * and <unicode> ::= U+<h><h><h><h>
+ * and <h> ::= <hexadecimal digit>
+ */
+
+      while (*p == ' ' || *p == '\t')
+	p++;
+      if (!*p || *p == '#')
+	continue;	/* skip comment or blank line */
+
+      fp0 = strtol(p, &p1, 0);
+      if (p1 == p)
+	{
+	  fprintf(stderr, "Bad input line: %s\n", buffer);
+	  exit(EX_DATAERR);
+        }
+      p = p1;
+
+      while (*p == ' ' || *p == '\t')
+	p++;
+      if (*p == '-')
+	{
+	  p++;
+	  fp1 = strtol(p, &p1, 0);
+	  if (p1 == p)
+	    {
+	      fprintf(stderr, "Bad input line: %s\n", buffer);
+	      exit(EX_DATAERR);
+	    }
+	  p = p1;
+        }
+      else
+	fp1 = 0;
+
+      if ( fp0 < 0 || fp0 >= fontlen )
+	{
+	    fprintf(stderr,
+		    "%s: Glyph number (0x%x) larger than font length\n",
+		    tblname, fp0);
+	    exit(EX_DATAERR);
+	}
+      if ( fp1 && (fp1 < fp0 || fp1 >= fontlen) )
+	{
+	    fprintf(stderr,
+		    "%s: Bad end of range (0x%x)\n",
+		    tblname, fp1);
+	    exit(EX_DATAERR);
+	}
+
+      if (fp1)
+	{
+	  /* we have a range; expect the word "idem" or a Unicode range of the
+	     same length */
+	  while (*p == ' ' || *p == '\t')
+	    p++;
+	  if (!strncmp(p, "idem", 4))
+	    {
+	      for (i=fp0; i<=fp1; i++)
+		addpair(i,i);
+	      p += 4;
+	    }
+	  else
+	    {
+	      un0 = getunicode(&p);
+	      while (*p == ' ' || *p == '\t')
+		p++;
+	      if (*p != '-')
+		{
+		  fprintf(stderr,
+"%s: Corresponding to a range of font positions, there should be a Unicode range\n",
+			  tblname);
+		  exit(EX_DATAERR);
+	        }
+	      p++;
+	      un1 = getunicode(&p);
+	      if (un0 < 0 || un1 < 0)
+		{
+		  fprintf(stderr,
+"%s: Bad Unicode range corresponding to font position range 0x%x-0x%x\n",
+			  tblname, fp0, fp1);
+		  exit(EX_DATAERR);
+	        }
+	      if (un1 - un0 != fp1 - fp0)
+		{
+		  fprintf(stderr,
+"%s: Unicode range U+%x-U+%x not of the same length as font position range 0x%x-0x%x\n",
+			  tblname, un0, un1, fp0, fp1);
+		  exit(EX_DATAERR);
+	        }
+	      for(i=fp0; i<=fp1; i++)
+		addpair(i,un0-fp0+i);
+	    }
+        }
+      else
+	{
+	    /* no range; expect a list of unicode values for a single font position */
+
+	    while ( (un0 = getunicode(&p)) >= 0 )
+	      addpair(fp0, un0);
+	}
+      while (*p == ' ' || *p == '\t')
+	p++;
+      if (*p && *p != '#')
+	fprintf(stderr, "%s: trailing junk (%s) ignored\n", tblname, p);
+    }
+
+  /* Okay, we hit EOF, now output hash table */
+
+  fclose(ctbl);
+
+
+  /* Compute total size of Unicode list */
+  nuni = 0;
+  for ( i = 0 ; i < fontlen ; i++ )
+    nuni += unicount[i];
+
+  printf("\
+/*\n\
+ * Automatically generated file; Do not edit.\n\
+ */\n\
+\n\
+#include <linux/types.h>\n\
+\n\
+u8 dfont_unicount[%d] = \n\
+{\n\t", fontlen);
+
+  for ( i = 0 ; i < fontlen ; i++ )
+    {
+      printf("%3d", unicount[i]);
+      if ( i == fontlen-1 )
+        printf("\n};\n");
+      else if ( i % 8 == 7 )
+        printf(",\n\t");
+      else
+        printf(", ");
+    }
+
+  printf("\nu16 dfont_unitable[%d] = \n{\n\t", nuni);
+
+  fp0 = 0;
+  nent = 0;
+  for ( i = 0 ; i < nuni ; i++ )
+    {
+      while ( nent >= unicount[fp0] )
+	{
+	  fp0++;
+	  nent = 0;
+	}
+      printf("0x%04x", unitable[fp0][nent++]);
+      if ( i == nuni-1 )
+         printf("\n};\n");
+       else if ( i % 8 == 7 )
+         printf(",\n\t");
+       else
+         printf(", ");
+    }
+
+  exit(EX_OK);
+}