[Bug 297638] ldd reports signal 13 when in pipeline

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=297638

--- Comment #1 from Markus Wild <[email protected]> ---
There seems to be a race whether the bug triggers or not. The problem itself
does seem to happen with previous systems as well, but the probability goes
down when I go back:

This is on a 13.2 system:

$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
pam_ssh.so.5: signal 13
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
pam_ssh.so.5: signal 13
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
pam_ssh.so.5: signal 13
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
pam_ssh.so.5: signal 13
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
pam_ssh.so.5: signal 13
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a pam_ssh.so.5 | fgrep -q 'not found'
pam_ssh.so.5: signal 13

This is on 11.2:

$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
./pam_ssh.so.5: signal 13
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
./pam_ssh.so.5: signal 13
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
./pam_ssh.so.5: signal 13
$ ldd -a ./pam_ssh.so.5 | fgrep -q 'not found'
./pam_ssh.so.5: signal 13


On 10.3, the bug didn't trigger at all.


I've run ktrace -i on a /bin/sh, where I did the above, and here's the end of 
kdump -E | egrep 'write|close|pipe|open|shut|read|"'

 [...]
 91610 ld-elf.so.1 4.465551813 RET   write 1
 91609 fgrep    4.465441180 CALL  read(0,0x53c5822404c0,0x8000)
 91610 ld-elf.so.1 4.465554672 CALL  write(0x1,0x23797c1fb6f,0xe)
       "libcrypto.so.7"
 91610 ld-elf.so.1 4.465566683 RET   write 14/0xe
 91609 fgrep    4.465559580 GIO   fd 0 read 30 bytes
       "libssh.so.5 => not found (0)
                "
 91609 fgrep    4.465572768 RET   read 30/0x1e
 91610 ld-elf.so.1 4.465569507 CALL  write(0x1,0x23796652eaf,0x1)
 91609 fgrep    4.465583554 CALL  close(0)
       " "
 91610 ld-elf.so.1 4.465592323 RET   write 1
 91610 ld-elf.so.1 4.465595462 CALL  write(0x1,0x23796652eaf,0x1)
 91609 fgrep    4.465606660 RET   close 0
       "="
 91610 ld-elf.so.1 4.465618571 RET   write -1 errno 32 Broken pipe
 91608 ldd      4.465734687 CALL  write(0x2,0x33e48d760510,0x18)
       "pam_ssh.so.5: signal 13
       "
 91608 ldd      4.465751997 RET   write 24/0x18


so, once fgrep -q sees the pattern it's looking for, it's essentially done and
closes fd 0. Since ld-elf.so.1 is not done yet, and continues writing, it will
wind up with SIGPIPE. This should not result in an error message in my
opinion...

-- 
You are receiving this mail because:
You are the assignee for the bug.