Bug 200635 - javaldx[32038]: segfault at 000000000040300c rip 0000003d34503909 rsp 00007fff3bb74a60 error
Summary: javaldx[32038]: segfault at 000000000040300c rip 0000003d34503909 rsp 00007ff...
Keywords:
Status: CLOSED RAWHIDE
Alias: None
Product: Fedora
Classification: Fedora
Component: glibc
Version: rawhide
Hardware: x86_64
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Jakub Jelinek
QA Contact:
URL:
Whiteboard:
: 200313 200366 (view as bug list)
Depends On:
Blocks: FC6Blocker FC6Test3Blocker FC6Desktop
TreeView+ depends on / blocked
 
Reported: 2006-07-29 12:24 UTC by Nicolas Mailhot
Modified: 2009-02-02 15:49 UTC (History)
7 users (show)

Fixed In Version: 2.4.90-17
Clone Of:
Environment:
Last Closed: 2006-08-03 06:40:02 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)

Description Nicolas Mailhot 2006-07-29 12:24:05 UTC
Description of problem:

$ oocalc 
/usr/lib64/openoffice.org2.0/program/soffice: line 147: 32040 Segmentation fault
     "$sd_prog/$sd_binary" "$@"

in /var/log/messages
Jul 29 14:30:59 rousalka kernel: javaldx[32038]: segfault at 000000000040300c
rip 0000003d34503909 rsp 00007fff3bb74a60 error 4
Jul 29 14:30:59 rousalka kernel: scalc.bin[32040]: segfault at 000000000040100c
rip 0000003d34503909 rsp 00007fff4e89c950 error 4

Version-Release number of selected component (if applicable):
openoffice.org-core-2.0.3.x86_64

How reproducible:
Always

Comment 1 Caolan McNamara 2006-07-29 13:04:41 UTC
Does this happen with openoffice.org-core-2.0.3-7.6 ?, and if it does can you
give me the output of /usr/sbin/sestatus. 

Was there a dialog when office crashed which had some "paste me in" info ?

Comment 2 Nicolas Mailhot 2006-07-29 13:51:31 UTC
It happens with openoffice.org-core-2.0.3-7.6. These is no crash dialog
I forced reinstall of :

# rpm -Uvh --force openoffice.org-c* java-1.4.2-gcj-compat-*
libgcj-4.1.1-13.x86_64.rpm
   1:libgcj                 ########################################### [ 20%]
   2:openoffice.org-core    ########################################### [ 40%]
   3:java-1.4.2-gcj-compat  ########################################### [ 60%]
   4:openoffice.org-calc    ########################################### [ 80%]
   5:java-1.4.2-gcj-compat-p########################################### [100%]

and now it works.

So I suspect some form of prelink f-up

Do you still need /usr/sbin/sestatus ? I don't know if I can reproduce the bug
now I've workarounded it

Comment 3 Nicolas Mailhot 2006-07-29 15:35:01 UTC
Ok, I've manually forced preliking
# /etc/cron.daily/prelink

and calc is hosed again

Comment 4 Nicolas Mailhot 2006-07-29 15:35:46 UTC
$ /usr/sbin/sestatus
SELinux status:                 disabled


Comment 5 Caolan McNamara 2006-07-31 08:26:02 UTC
So, installed freshly, all ok, run prelink, not ok.

Comment 6 Jeremy Katz 2006-08-02 12:58:08 UTC
Did this just break after updating to prelink-0.3.9-2?

Comment 7 Jakub Jelinek 2006-08-02 16:05:50 UTC
This is actually a glibc bug.
71            const ElfW(Sym) *symtab
72              = (const ElfW(Sym) *) D_PTR (match, l_info[DT_SYMTAB]);
73            const char *strtab = (const char *) D_PTR (match,
l_info[DT_STRTAB]);
74
75            ElfW(Word) strtabsize = match->l_info[DT_STRSZ]->d_un.d_val;
76
77            const ElfW(Sym) *symtabend;
78            if (match->l_info[DT_HASH] != NULL)
79              symtabend = (symtab
80                           + ((Elf_Symndx *) D_PTR (match,
l_info[DT_HASH]))[1]);
81            else
82              /* There is no direct way to determine the number of symbols in
the
83                 dynamic symbol table and no hash table is present.  The ELF
84                 binary is ill-formed but what shall we do?  Use the beginning
of
85                 the string table which generally follows the symbol table.  */
86              symtabend = (const ElfW(Sym) *) strtab;

With prelinking (but even otherwise, there are no guarantees), the assunmption
that .dynstr immediately follows .dynsym is wrong.  And, for binaries and/or
libraries built with --hash-style=gnu there is no DT_HASH, which has been
superceeded by DT_GNU_HASH.  So, glibc needs to handle that.

Comment 8 Jakub Jelinek 2006-08-03 06:40:02 UTC
Should be fixed in glibc-2.4.90-17 in rawhide.

Comment 9 Jakub Jelinek 2006-08-03 06:44:02 UTC
*** Bug 200313 has been marked as a duplicate of this bug. ***

Comment 10 Caolan McNamara 2006-08-03 09:11:34 UTC
*** Bug 200366 has been marked as a duplicate of this bug. ***


Note You need to log in before you can comment on or make changes to this bug.