297193 - crash after flush

Status: RESOLVED INCOMPLETE

(Directory :: LDAP C SDK, defect)

Description

[jagadish]

User-Agent:       Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0)
Build Identifier: 

Memcache genrates core after flushing the items for immediate next request

Reproducible: Always

Steps to Reproduce:
1.Allocate the cache using ldap_memcache_init
2.Items will be cached for group search
3.try to flush all the items using ldap_memcache_flush.
4. flushes all the items.
5. send another request , generates core

Actual Results:  
Generates core dump, crashes the authentication module.

Expected Results:  
It ahould append new items, with out crashing

Comment 1

[Mark Smith [:mcs]]

Thanks for reporting this problem.  Can you provide a stack trace to show where
the code is crashing?  What product or software does "authentication module"
refer to?

Comment 2

[jagadish]

I can't give the core file as its confidential. I will tell where exactly its 
crashing. The ldap_memcache_flush call exits successfully without any probs, 
for the next immediate request from our module we call ldap_search_s(), after 
that thf low will be as follows from stack trace in seacrh.c at line 1000 
insldapi_search() -> ldap_memcache_result() -> memcache_seacrh() -> 
memcache_access () with mode 3(FIND) -> htable_get() -> attrkey_getdata () here 
while accessing the member pRes->ldmemcr_crc_key it crashes.

I think the above information will give and idea for the problem. It seems to 
be some problem while flushing all the items using htable_removaall, Is that we 
are removing all the items and all the pointer variables are linked properly.
 

(In reply to comment #1)
> Thanks for reporting this problem.  Can you provide a stack trace to show 
where
> the code is crashing?  What product or software does "authentication module"
> refer to?

Comment 3

[Rich Megginson]

I've looked at the code - it's quite hard to follow all of the pointer references and back references.  The code in htable_removeall() just does not look like it would work correctly, so perhaps fixing that will solve this problem.  I would say it's going to be about 3 solid days of work to try to untangle the pointers and test a fix for this problem.

Comment 4

[Wayne Mery (:wsmwk)]

Rich, can you still reproduce this, and if so get a stacktrace with symbols?
 https://developer.mozilla.org/En/How_to_get_a_stacktrace_for_a_bug_report#Linux

Comment 5

[Robert Kaiser]

No reply to the request in comment #4, no info on current versions, so let's resolve this. If new info comes around, please reopen based on that.