Bug 153299

Summary:	%ebp register is 0x0, %esp register has a correct value
Product:	Red Hat Enterprise Linux 3	Reporter:	Dragos MOINESCU <dmoinescu>
Component:	gcc	Assignee:	Jakub Jelinek <jakub>
Status:	CLOSED NOTABUG	QA Contact:
Severity:	medium	Docs Contact:
Priority:	medium
Version:	3.0	Keywords:	FutureFeature
Target Milestone:	---
Target Release:	---
Hardware:	i686
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Enhancement
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2005-04-04 18:50:44 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Dragos MOINESCU 2005-04-04 16:35:23 UTC

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.6) Gecko/20050317 Firefox/1.0.2

Description of problem:
THIS IS NOT A RHEL3 BUG, but I have no idea how to solve this problem.

I have developed a multi-threaded server that crashes if there are more than 500 threads. This server implements a simple HTTP Protocol. Each thread processes one request at a time.
When the number of threads are increased, the server receive a SIGSEGV un dumps a core file.
I have run "gdb -c core.pid server" and receive a strange $ebp register value. That value is 0x0. The server crashes in a function that tests if there is something to be read from the socket. The only thing there is done is a simple select((int)fd+1, &read_fs, NULL, NULL, &tv);
I have incresed maximum number of fds from 1024 to 65535 with ulimit -n. Also I have modified /usr/include/bits/typesize.h to increment __FD_SETSIZE from 1024 to 32678.
I have recompiled the server using -g and still receive that SIGSEGV.
Please tell me what can I do.

Thanx

Version-Release number of selected component (if applicable):


How reproducible:
Always

Steps to Reproduce:
you cannot reproduce this.

Additional info:

Comment 1 Jakub Jelinek 2005-04-04 18:50:44 UTC

If you need more than 1024 fds, then select is certainly a bad interface, use
poll instead.
Are you using NPTL or LinuxThreads (if the latter, non-FLOATING_STACKS or
FLOATING_STACKS)?  What is the thread stack size you are using?
With the default thread stack size ~ 500 threads are filling the whole address
space.  You can tweak that with ulimit -s or pthread_attr_setstacksize.

%ebp 0 is nothing strange in -fomit-frame-pointer code, in
-fno-omit-frame-pointer code it shouldn't be 0 with the exception of function
started by clone(2).

BTW, you should have asked about this through your support contact, bugzilla
is really only for reporting bugs.