74 lines
2.9 KiB
Plaintext
74 lines
2.9 KiB
Plaintext
PURPOSE of /tools/glibc/Regression/bz455360-RHEL4-vfprintf-call-goes-into-recursive
|
|
Description: Test for bz455360 ([RHEL4] vfprintf() call goes into recursive)
|
|
Author: Petr Muller <pmuller@redhat.com>
|
|
Bug summary: [RHEL4] vfprintf() call goes into recursive overflow and crashes with a segfault
|
|
Bugzilla link: https://bugzilla.redhat.com/show_bug.cgi?id=455360
|
|
|
|
Description:
|
|
|
|
Description of problem in short (details sent next thru IT):
|
|
|
|
Sporadically, when under load, the client is seeing tibco EMS processes be
|
|
killed by segfaults. Getting a robust diagnosis has proven difficult, but from
|
|
what can be told, it appears that somehow a vfprintf() call is getting into a
|
|
state where it loops with:
|
|
|
|
<snip>
|
|
6 0x008602fa in *__GI___overflow (f=0xf7fe8688, ch=0) at genops.c:240
|
|
#7 0x0083afe6 in _IO_helper_overflow (s=0xf7fe8688, c=Variable "c" is not
|
|
available.
|
|
) at vfprintf.c:2058
|
|
</snip>
|
|
|
|
It recurses for quite some time in glibc and then crashes with a segfault. I've
|
|
just gotten the application binary (stripped unfortunately) and thus could
|
|
actually get something from the core.
|
|
|
|
Customer's analysis:
|
|
we have seen this issue before. What's going wrong is that we are calling
|
|
vfprintf to perform a message trace including the body and the functions in
|
|
glibc are recursing approximately 373909 frames and then it crashed.
|
|
|
|
|
|
<snip>
|
|
#373902 0x008602fa in *__GI___overflow (f=0xf7fe8688, ch=0) at genops.c:240
|
|
#373903 0x0083afe6 in _IO_helper_overflow (s=0xf7fe8688, c=-1)
|
|
at vfprintf.c:2058
|
|
#373904 0x0086096f in _IO_default_xsputn (f=0xf7fe8688, data=0x32046928,
|
|
n=10285) at genops.c:478
|
|
#373905 0x0083e88e in _IO_vfprintf (s=0xf7fe8688,
|
|
format=0x8321890 "%s %s from %s: connID=%lld prodID=%lld msgID='%s' %s
|
|
mode=%s %s='%s'%s%s", ap=0xf7feae90 "ïJ©\236À¦\004\204:Z\\Æ\\ò%$Ü\223K")
|
|
at vfprintf.c:1553
|
|
#373906 0x0083b0ac in buffered_vfprintf (s=0x929460,
|
|
format=0x8321890 "%s %s from %s: connID=%lld prodID=%lld msgID='%s' %s
|
|
mode=%s %s='%s'%s%s", args=) at vfprintf.c:2144
|
|
#373907 0x0083b2eb in _IO_vfprintf (s=0x929460,
|
|
format=0x8321890 "%s %s from %s: connID=%lld prodID=%lld msgID='%s' %s
|
|
mode=%s %s='%s'%s%s", ap=0xf7feae58 "Яþ÷lê2\bP°þ÷;\216K") at vfprintf.c:1264
|
|
#373908 0x0811ba34 in _authenticate ()
|
|
#373909 0x0811bc4c in _authenticate ()
|
|
#373910 0x080f7f3c in _authenticate ()
|
|
#373911 0x080e8e6a in _authenticate ()
|
|
#373912 0x080e6d88 in _authenticate ()
|
|
#373913 0x080e6ed0 in _authenticate ()
|
|
#373914 0x0807d4fd in ?? ()
|
|
#373915 0x2846afc8 in ?? ()
|
|
#373916 0x0830e880 in _IO_stdin_used ()
|
|
#373917 0x00000432 in ?? ()
|
|
#373918 0x0830e86d in _IO_stdin_used ()
|
|
#373919 0x0830e86d in _IO_stdin_used ()
|
|
#373920 0x00000000 in ?? ()
|
|
</snip>
|
|
|
|
|
|
Version-Release number of selected component (if applicable):
|
|
glibc-2.3.4-2.39
|
|
|
|
Additional info:
|
|
|
|
I have the core and binary set up on a lab host which has the correct version of
|
|
glibc installed. Feel free to work on that host:
|
|
|
|
dhcp139.gsslab.rdu.redhat.com (root:redhat)
|