23

Given a kernel stack trace as below, how do you determine the specific line of code where the issue occurred?

kernel:  [<ffffffff80009a14>] __link_path_walk+0x173/0xfb9
kernel:  [<ffffffff8002cbec>] mntput_no_expire+0x19/0x89
kernel:  [<ffffffff8000eb94>] link_path_walk+0xa6/0xb2
kernel:  [<ffffffff80063c4f>] __mutex_lock_slowpath+0x60/0x9b
kernel:  [<ffffffff800238de>] __path_lookup_intent_open+0x56/0x97
kernel:  [<ffffffff80063c99>] .text.lock.mutex+0xf/0x14
kernel:  [<ffffffff8001b222>] open_namei+0xea/0x712
kernel:  [<ffffffff8006723e>] do_page_fault+0x4fe/0x874
kernel:  [<ffffffff80027660>] do_filp_open+0x1c/0x38
kernel:  [<ffffffff8001a061>] do_sys_open+0x44/0xbe
kernel:  [<ffffffff8005d28d>] tracesys+0xd5/0xe0

While I have no trouble finding the function call -- but translating __link_path_walk plus an offset to an actual line number is the difficult part.

Assuming this is for a standard distribution-provided kernel for which I know the exact version and build number, what's the process for fetching the necessary metadata and doing the corresponding lookup?

tylerl
  • 14,885
  • 7
  • 49
  • 71

5 Answers5

17

I don't have a ~= RHEL5 at hand, so the output shown is from a Fedora 20, though the process should be mostly the same (the name of the function has changed).

You'd need to install the appropriate kernel-debug-debuginfo package for your kernel (assuming RHEL or derivative distro). This package provides a vmlinux image (an uncompressed not stripped version of the kernel):

# rpm -ql kernel-debug-debuginfo | grep vmlinux
/usr/lib/debug/lib/modules/3.14.7-200.fc20.x86_64+debug/vmlinux

that image can be used directly with gdb

# gdb /usr/lib/debug/lib/modules/3.14.7-200.fc20.x86_64+debug/vmlinux
GNU gdb (GDB) Fedora 7.7.1-13.fc20
Copyright (C) 2014 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
...
Reading symbols from /usr/lib/debug/lib/modules/3.14.7-200.fc20.x86_64+debug/vmlinux...done.
(gdb) disassemble link_path_walk
Dump of assembler code for function link_path_walk:
   0xffffffff81243d50 <+0>:     callq  0xffffffff817ea840 <__fentry__>
   0xffffffff81243d55 <+5>:     push   %rbp
   0xffffffff81243d56 <+6>:     mov    %rsp,%rbp
   0xffffffff81243d59 <+9>:     push   %r15
   0xffffffff81243d5b <+11>:    mov    %rsi,%r15
   0xffffffff81243d5e <+14>:    push   %r14
   0xffffffff81243d60 <+16>:    push   %r13
   0xffffffff81243d62 <+18>:    push   %r12
   0xffffffff81243d64 <+20>:    push   %rbx
   0xffffffff81243d65 <+21>:    mov    %rdi,%rbx
   0xffffffff81243d68 <+24>:    sub    $0x78,%rsp
   0xffffffff81243d6c <+28>:    mov    %gs:0x28,%rax
   0xffffffff81243d75 <+37>:    mov    %rax,0x70(%rsp)
   0xffffffff81243d7a <+42>:    xor    %eax,%eax
   0xffffffff81243d7c <+44>:    movzbl (%rdi),%eax
   0xffffffff81243d7f <+47>:    cmp    $0x2f,%al
   ....

You can also use objdump(1) on the vmlinux image:

# objdump -rDlS /usr/lib/debug/lib/modules/3.14.7-200.fc20.x86_64+debug/vmlinux > vmlinux.out

The flags are:

   -D
   --disassemble-all
       Like -d, but disassemble the contents of all sections, not just those expected to contain instructions.
   -r
   --reloc
       Print the relocation entries of the file.  If used with -d or -D, the relocations are printed interspersed with the
       disassembly.
   -S
   --source
       Display source code intermixed with disassembly, if possible.  Implies -d.
   -l
   --line-numbers
       Label the display (using debugging information) with the filename and source line numbers corresponding to the object
       code or relocs shown.  Only useful with -d, -D, or -r.

You can lookup the function there:

ffffffff81243d50 <link_path_walk>:
link_path_walk():
/usr/src/debug/kernel-3.14.fc20/linux-3.14.7-200.fc20.x86_64/fs/namei.c:1729
 *
 * Returns 0 and nd will have valid dentry and mnt on success.
 * Returns error and drops reference to input namei data on failure.
 */
static int link_path_walk(const char *name, struct nameidata *nd)
{
ffffffff81243d50:       e8 eb 6a 5a 00          callq  ffffffff817ea840 <__entry_text_start>
ffffffff81243d55:       55                      push   %rbp
ffffffff81243d56:       48 89 e5                mov    %rsp,%rbp
ffffffff81243d59:       41 57                   push   %r15
ffffffff81243d5b:       49 89 f7                mov    %rsi,%r15
ffffffff81243d5e:       41 56                   push   %r14
ffffffff81243d60:       41 55                   push   %r13
ffffffff81243d62:       41 54                   push   %r12
ffffffff81243d64:       53                      push   %rbx
ffffffff81243d65:       48 89 fb                mov    %rdi,%rbx
ffffffff81243d68:       48 83 ec 78             sub    $0x78,%rsp
ffffffff81243d6c:       65 48 8b 04 25 28 00    mov    %gs:0x28,%rax
ffffffff81243d73:       00 00
ffffffff81243d75:       48 89 44 24 70          mov    %rax,0x70(%rsp)
ffffffff81243d7a:       31 c0                   xor    %eax,%eax
/usr/src/debug/kernel-3.14.fc20/linux-3.14.7-200.fc20.x86_64/fs/namei.c:1733
        struct path next;
        int err;

        while (*name=='/')
ffffffff81243d7c:       0f b6 07                movzbl (%rdi),%eax
ffffffff81243d7f:       3c 2f                   cmp    $0x2f,%al
ffffffff81243d81:       75 10                   jne    ffffffff81243d93 <link_path_walk+0x43>
ffffffff81243d83:       0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)
/usr/src/debug/kernel-3.14.fc20/linux-3.14.7-200.fc20.x86_64/fs/namei.c:1734
                name++;
ffffffff81243d88:       48 83 c3 01             add    $0x1,%rbx
/usr/src/debug/kernel-3.14.fc20/linux-3.14.7-200.fc20.x86_64/fs/namei.c:1733
static int link_path_walk(const char *name, struct nameidata *nd)
{
        struct path next;
        int err;

        while (*name=='/')
....

and match the offset to the actual line of code.

dawud
  • 14,918
  • 3
  • 41
  • 61
15

Given an unstripped vmlinux with debugging symbols (typically included with "linux-devel" or "linux-headers" packages matching your kernel version), you can use the addr2line program included with binutils to translate addresses to lines in source files.

Consider this call trace:

Call Trace:
 [<ffffffff8107bf5d>] ? finish_task_switch+0x3d/0x120
 [<ffffffff815f3130>] __schedule+0x3b0/0x9d0
 [<ffffffff815f3779>] schedule+0x29/0x70
 [<ffffffff815f2ccc>] schedule_hrtimeout_range_clock.part.24+0xdc/0xf0
 [<ffffffff81076440>] ? hrtimer_get_res+0x50/0x50
 [<ffffffff815f2c6f>] ? schedule_hrtimeout_range_clock.part.24+0x7f/0xf0
 [<ffffffff815f2cf9>] schedule_hrtimeout_range_clock+0x19/0x60
 [<ffffffff815f2d53>] schedule_hrtimeout_range+0x13/0x20
 [<ffffffff811a8aa9>] poll_schedule_timeout+0x49/0x70
 [<ffffffff811aa203>] do_sys_poll+0x423/0x550
 [<ffffffff814eaf8c>] ? sock_recvmsg+0x9c/0xd0
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811a8c50>] ? poll_select_copy_remaining+0x140/0x140
 [<ffffffff811aa3fe>] SyS_poll+0x5e/0x100
 [<ffffffff816015d2>] system_call_fastpath+0x16/0x1b

Then the address of the caller in poll_select_copy_remaining can be found with:

$ addr2line -e /tmp/vmlinux ffffffff811a8c50
/tmp/linux-3.15-rc8/fs/select.c:209
Lekensteyn
  • 6,111
  • 6
  • 37
  • 55
  • 2
    My panic stack traces don't show the full addresses, only func+line. Is there a configuration to enable it? – Ciro Santilli OurBigBook.com Apr 14 '18 at 22:12
  • 1
    @CiroSantilli包子露宪六四事件法轮功 Perhaps you could post a new question about that? You have to include more information: architecture (x86-64/ARM/...), Linux distribution, kernel version, and a sample of the output that does not match what you would expect. – Lekensteyn Apr 15 '18 at 08:46
  • 3
    OK, symbols only show when you have `CONFIG_KALLSYMS`: https://github.com/cirosantilli/linux-kernel-module-cheat/tree/b8f190cc24b4f7474894b68a5510a8f3d767843d#panic-trace-show-addresses-instead-of-symbols – Ciro Santilli OurBigBook.com May 02 '18 at 17:05
  • 1
    @CiroSantilli新疆棉花TRUMPBANBAD the other answer here that is about using `faddr2line` works for `func+offset`, so no need to enable or recompile anything. – Hi-Angel May 16 '21 at 11:43
8
  1. Install kernel-debuginfo

  2. Download decode_stacktrace.sh which is in the kernel source tree.

  3. Make stack dump output useful again.

     # ./decode_stacktrace.sh /usr/lib/debug/lib/modules/`uname -r`/vmlinux /usr/lib/debug/lib/modules/4.1.12-112.14.14.el7uek.x86_64/ < ./trace > out
     # cat out
     [102820.087367] Call Trace:
     [102820.087371] dump_stack (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/lib/dump_stack.c:53)
     [102820.087375] warn_slowpath_common (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/kernel/panic.c:499)
     [102820.087378] warn_slowpath_null (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/kernel/panic.c:533)
     [102820.087380] af_alg_accept (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/include/net/sock.h:1689 /usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/crypto/af_alg.c:287)
     [102820.087382] alg_accept (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/crypto/af_alg.c:326)
     [102820.087385] SYSC_accept4 (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/net/socket.c:1485)
     [102820.087388] ? release_sock (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/net/core/sock.c:2415)
     [102820.087390] ? alg_setsockopt (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/crypto/af_alg.c:264)
     [102820.087393] SyS_accept (/usr/src/debug/kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/net/socket.c:1515)
     [102820.087395] system_call_fastpath (/usr/src/debug////////kernel-4.1.12/linux-4.1.12-112.14.14.el7uek/arch/x86/kernel/entry_64.S:277)
     [102820.087397] ---[ end trace 1315ff0b8d6ff7d8 ]---
    
  4. For a handful of function offsets, try faddr2line which is also in the kernel source.

     $ wget https://raw.githubusercontent.com/torvalds/linux/master/scripts/faddr2line
     $ bash faddr2line /usr/lib/debug/lib/modules/`uname -r`/vmlinux __do_softirq+0x92/0x320
     __do_softirq+0x92/0x320:
     ffs at arch/x86/include/asm/bitops.h:410
     (inlined by) __do_softirq at kernel/softirq.c:261
    
wenjianhn
  • 211
  • 2
  • 3
  • ⁺¹ for `faddr2line` as it works with kernel modules too, like for example: `faddr2line /lib/modules/5.12.0-arch1-1/build/drivers/gpu/drm/i915/i915.ko i915_gem_prime_import+0x2c/0x12`. The `decode_stacktrace.sh` I did not manage to make work: whatever path I give it, it always refuses to decode addresses in modules with the exact same complaint `readelf: Error: Not an ELF file - it has the wrong magic bytes at the start WARNING! Modules path isn't set, but is needed to parse this symbol` – Hi-Angel May 16 '21 at 11:38
1

If addr2line should print a question mark for line number or objdump fails to inline source code and you have a custom kernel, be sure to recompile the kernel with CONFIG_DEBUG_INFO set. You might need to reproduce the error with the kernel just built.

0

With gdb, you can also use this command to find the line number quickly:

(gdb) list *(some_function+0x12c)

accessory
  • 101
  • 2