[Bug symtab/10457] New: gdb Segfaults due stack overflow (excessive usage of alloca)

[Bug symtab/10457] New: gdb Segfaults due stack overflow (excessive usage of alloca)

[zecke at selfish dot org]

Debugging a process that loaded a debug enabled libWebKitQt will lead to a segfault in gdb. This 
is due a stack overflow. valgrind reports the event the following way and a backtrace from a core 
file is also in this description.

valgrind:
==11373== Process terminating with default action of signal 11 (SIGSEGV)
==11373==  Access not within mapped region at address 0xBE26EFDC
==11373==    at 0x827D2C2: cplus_demangle_type (in /usr/bin/gdb)
==11373==  If you believe this happened as a result of a stack overflow in your
==11373==  program's main thread (unlikely but possible), you can try to increase
==11373==  the size of the main thread stack using the --main-stacksize= flag.
==11373==  The main thread stack size used in this run was 8388608.
==11373== Stack overflow in thread 1: can't grow stack to 0xbe26efd8


gdb on the core:
Core was generated by `gdb --args ./bin/QtLauncher http://index.hu'.
Program terminated with signal 11, Segmentation fault.
#0  0x0827e538 in ?? ()
(gdb) bt
#0  0x0827e538 in ?? ()
#1  0x0827d2c7 in cplus_demangle_type ()
#2  0x0827d39e in cplus_demangle_type ()
#3  0x0827df55 in ?? ()
#4  0x0827e6b5 in ?? ()
#5  0x0827e76c in ?? ()
#6  0x08280b2b in ?? ()
#7  0x08280c95 in ?? ()
#8  0x08280d04 in cplus_demangle_v3 ()
#9  0x08279e03 in cplus_demangle ()
#10 0x08139fc4 in symbol_set_names ()
#11 0x0808dc80 in prim_record_minimal_symbol_and_info ()
#12 0x080fce83 in ?? ()
#13 0x080fd0ce in ?? ()
#14 0x080fda39 in ?? ()
#15 0x0813dcf0 in syms_from_objfile ()
#16 0x0813f383 in ?? ()
#17 0x080a879e in solib_read_symbols ()
#18 0x080a8dcc in solib_add ()
#19 0x08152341 in handle_inferior_event ()
#20 0x08154760 in wait_for_inferior ()
#21 0x08154b2c in proceed ()
#22 0x0814c157 in ?? ()
#23 0x08090ea2 in execute_command ()
#24 0x08161a9b in ?? ()
#25 0x081627ea in ?? ()
#26 0xb7ec1277 in rl_callback_read_char () from /lib/libreadline.so.5
#27 0x08161bdb in ?? ()
#28 0x081610b9 in ?? ()
#29 0x081608db in ?? ()
#30 0x08161796 in gdb_do_one_event ()
#31 0x0815baa3 in catch_errors ()
#32 0x080ee2d4 in ?? ()
#33 0x0815c173 in current_interp_command_loop ()
#34 0x0808890b in ?? ()
#35 0x0815baa3 in catch_errors ()
#36 0x080894a4 in ?? ()
#37 0x0815baa3 in catch_errors ()
#38 0x08088832 in gdb_main ()
#39 0x080887f3 in main ()
(gdb)

-- 
           Summary: gdb Segfaults due stack overflow (excessive usage of
                    alloca)
           Product: gdb
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: symtab
        AssignedTo: unassigned at sourceware dot org
        ReportedBy: zecke at selfish dot org
                CC: gdb-prs at sourceware dot org


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[zecke at selfish dot org]

------- Additional Comments From zecke at selfish dot org  2009-07-29 05:41 -------
Created an attachment (id=4095)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=4095&action=view)
Hack to change alloca to malloc

The following workaround is working for me. The idea is to exchange alloca with
malloc.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[ppluzhnikov at google dot com]

------- Additional Comments From ppluzhnikov at google dot com  2009-07-29 23:01 -------
A related patch for the same problem (but in different area of GDB):
http://sourceware.org/ml/gdb-patches/2009-07/msg00598.html

Holger, could you send me (or otherwise make available) the bin/QtLauncher
binary which makes GDB crash?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[zecke at selfish dot org]

------- Additional Comments From zecke at selfish dot org  2009-08-11 14:44 -------
(In reply to comment #2)
> A related patch for the same problem (but in different area of GDB):
> http://sourceware.org/ml/gdb-patches/2009-07/msg00598.html
> 
> Holger, could you send me (or otherwise make available) the bin/QtLauncher
> binary which makes GDB crash?

The binary + WebKit library is +100mb of size. Where should I send it or upload it?



-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[ppluzhnikov at google dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ppluzhnikov at google dot
                   |                            |com


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[zecke at selfish dot org]

------- Additional Comments From zecke at selfish dot org  2009-10-01 06:31 -------
Sorry for delaying it, I will try to compile and provide the necessary libraries
by next week.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[ppluzhnikov at google dot com]

------- Additional Comments From ppluzhnikov at google dot com  2009-10-08 05:30 -------
Following Holger's directions and additional directions here:
http://trac.webkit.org/wiki/BuildingQtOnLinux#DependenciesforspecificLinuxdistributions

I built debug version of Webkit with:
  WebKitTools/Scripts/build-webkit --qt --debug
on Fedora 11/i686 and tested it with GDB from CVS Head, which did not crash with
(default) 'ulimit -s' of 10240

The resulting libQtWebKit.so.4.5.2 is 472223763 bytes.

It does crash with 6M stack though.
The problem is alloca in elf_symtab_read in a loop.

Patch sent:
http://sourceware.org/ml/gdb-patches/2009-10/msg00163.html

With the patch, GDB runs in 16K stack (ulimit -s 16) on this test case without
any problems.



-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at sourceware dot|ppluzhnikov at google dot
                   |org                         |com
             Status|WAITING                     |ASSIGNED
     Ever Confirmed|                            |1


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2009-10-08 16:38 -------
Subject: Bug 10457

CVSROOT:	/cvs/src
Module name:	src
Changes by:	ppluzhnikov@sourceware.org	2009-10-08 16:38:42

Modified files:
	gdb            : ChangeLog elfread.c 

Log message:
	2009-10-08  Paul Pluzhnikov  <ppluzhnikov@google.com>
	
	PR gdb/10457
	* elfread.c (elf_symtab_read): Don't use alloca in a loop.

Patches:
http://sources.redhat.com/cgi-bin/cvsweb.cgi/src/gdb/ChangeLog.diff?cvsroot=src&r1=1.10941&r2=1.10942
http://sources.redhat.com/cgi-bin/cvsweb.cgi/src/gdb/elfread.c.diff?cvsroot=src&r1=1.77&r2=1.78



-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[ppluzhnikov at google dot com]

------- Additional Comments From ppluzhnikov at google dot com  2009-10-08 16:41 -------
Fixed on trunk.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug symtab/10457] gdb Segfaults due stack overflow (excessive usage of alloca)

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2009-10-08 17:42 -------
Subject: Bug 10457

CVSROOT:	/cvs/src
Module name:	src
Branch: 	gdb_7_0-branch
Changes by:	ppluzhnikov@sourceware.org	2009-10-08 17:42:10

Modified files:
	gdb            : ChangeLog elfread.c 

Log message:
	2009-10-08  Paul Pluzhnikov  <ppluzhnikov@google.com>
	
	PR gdb/10457
	* elfread.c (elf_symtab_read): Don't use alloca in a loop.

Patches:
http://sources.redhat.com/cgi-bin/cvsweb.cgi/src/gdb/ChangeLog.diff?cvsroot=src&only_with_tag=gdb_7_0-branch&r1=1.10874.2.48&r2=1.10874.2.49
http://sources.redhat.com/cgi-bin/cvsweb.cgi/src/gdb/elfread.c.diff?cvsroot=src&only_with_tag=gdb_7_0-branch&r1=1.77&r2=1.77.4.1



-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10457

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.