httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ruediger Pluem <rpl...@apache.org>
Subject Re: segfaults / core dumps caused by ap_internal_fast_redirect
Date Sun, 19 Apr 2009 11:58:24 GMT


On 04/08/2009 08:31 AM, Plüm, Rüdiger, VF-Group wrote:
>  
> 
>> -----Ursprüngliche Nachricht-----
>> Von: Paul Querna 
>> Gesendet: Dienstag, 7. April 2009 20:15
>> An: dev@httpd.apache.org
>> Betreff: Re: segfaults / core dumps caused by 
>> ap_internal_fast_redirect
>>
>> On Tue, Apr 7, 2009 at 10:01 AM, William A. Rowe, Jr.
>> <wrowe@rowe-clan.net> wrote:
>>> Plüm, Rüdiger, VF-Group wrote:
>>>
>>>> How to solve?
>>>>
>>>> I don't know. This is why I write this mail :-).
>>> We eliminate internal_fast_redirect ;-)
>>>
>> hell yeah, I would love to eliminate internal redirects completely.
>>
> 
> Not surprised by these answers :-).
> But what are the alternatives to internal redirects. AFAIKT 
> ap_internal_fast_redirect is currently used by mod_dir and
> mod_negotiation. So what should they do instead?
> Furthermore we also have this crashing risk in 2.2.x as well.
> It does not seem to happen there that often because we
> do not use eor buckets there and the used APR has the old
> pool implementation. We cannot axe it in 2.2.x, but of course
> approaches for 2.2.x and trunk can be different.

As I have seen no further proposals in the last 14 days for a replacement
of ap_internal_fast_redirect (as said I have no idea either) I would like
to fix the current code to avoid the segfaults that are currently occurring.
My outlined approach is the following:

1. Create _ex versions of ap_sub_req_lookup_uri / ap_sub_req_lookup_dirent
   and ap_sub_req_lookup_file that allow to supply a pool from which the
   subrequest allocates memory.

2. Let ap_internal_fast_redirect spit out a warning to the error log if
   the subrequest pool is different from the request pool.

3. Fix the current internal consumers of ap_internal_fast_redirect (mod_dir /
   mod_negotiation to use the _ex versions and feed them with the request pool.


Any opinions / objections?

Regards

Rüdiger

Mime
View raw message