Bug 77968 - cannot parse /proc/self/maps on Linux 2.6.4 kernel with PaX
Summary: cannot parse /proc/self/maps on Linux 2.6.4 kernel with PaX
Status: RESOLVED INTENTIONAL
Alias: None
Product: valgrind
Classification: Developer tools
Component: general (show other bugs)
Version: unspecified
Platform: Compiled Sources Linux
: NOR normal
Target Milestone: ---
Assignee: Julian Seward
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-03-19 01:46 UTC by Alexander
Modified: 2004-07-16 20:43 UTC (History)
0 users

See Also:
Latest Commit:
Version Fixed In:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Alexander 2004-03-19 01:46:52 UTC
The format of /proc/(pid)/maps is changed (I guess, generally in 2.6+ series, 
however tested only on 2.6.4).

The differences are - flags (modes) are uppercase, and also plus sign may be 
instead of minus sign. Sample:

4020e000-40214000 R-+s 00000000 03:02 17056      /usr/lib/gconv/gconv-
modules.cache
40214000-40216000 R+Xp 00000000 03:02 17185      /usr/lib/gconv/ISO8859-1.so
40216000-40217000 RW+p 00001000 03:02 17185      /usr/lib/gconv/ISO8859-1.so
40217000-40257000 RW+p 00000000 00:00 0 
bfffa000-c0000000 RWXp ffffc000 00:00 0 
ffffe000-fffff000 ---p 00000000 00:00 0 

The parser expects only uppercase flags, and it won't accept '+' as well.

This problem affects all versions (up to 2.1.1).
Comment 1 Tom Hughes 2004-03-19 08:21:19 UTC
In message <20040319004653.13501.qmail@ktown.kde.org> you wrote:

> The format of /proc/(pid)/maps is changed (I guess, generally in 2.6+
> series,  however tested only on 2.6.4).

Actually that isn't a standard 2.6 feature, it's am additional kernel
patch that causes that change. There was another bug report about it a
few weeks ago.

Tom

Comment 2 Nicholas Nethercote 2004-03-19 10:44:41 UTC
Yes, it's the PaX extension for control over memory mappings.  There is some
discussion about it, and a patch, in bug #69616.
Comment 3 Alexander 2004-03-19 13:44:00 UTC
The bug #69616 is closed, but the problem is not solved.

One may argue that this is not valgrind problem - yes, partially true, but it won't help those who uses PAX...

I can fix it for myself, that's not a problem - but about others?
Comment 4 Nicholas Nethercote 2004-03-28 14:39:54 UTC
Bug #69616 is closed because the original problem has been fixed;  the PaX problem was identified while fixing that, but was unrelated to the original problem.
I suggested to the person having a problem with PaX that they open another bug;
they didn't, but now you have, so now there is a problem report for it.

As for fixing it:  it's hard to say what best to do... if someone writes a patch
that changes the kernel's interface, it's inevitable that this will break some programs.  I think we should leave it for the moment, and if more people complain, then we can consider handling it.  For the moment, the patch in #69616 is available for people affected.

Comment 5 Nicholas Nethercote 2004-07-16 20:43:31 UTC
I'm closing this because nobody has complained for several months now.  If that turns out to be a problem, it can be re-opened.