svn commit: r282948 - head/lib/libthr/thread

[Bryan Drewery]

On 5/15/2015 3:40 AM, Konstantin Belousov wrote:
> Author: kib
> Date: Fri May 15 08:40:17 2015
> New Revision: 282948
> URL: https://svnweb.freebsd.org/changeset/base/282948
> 
> Log:
>   Some third-party malloc(3) implementations use pthread_setspecific(3)
>   to handle per-thread information.  Since our pthread_setspecific()
>   implementation calls calloc(3) to allocate per-thread specific data
>   storage, things get complicated.
>   
>   Switch the allocator to use bare mmap(2).  There is some loss of the
>   allocated page, since e.g. on amd64, PTHREAD_KEYS_MAX * sizeof(struct
>   pthread_specific_elem) is 3K (it actually spans whole page due to
>   padding), but I believe it is more acceptable than additional code for
>   specialized allocator().
>   
>   The alternatives would either to make the specific data array be part of
>   the struct thread, or use internal bindings to call the libc malloc,
>   avoiding interposing.
>   
>   Also do the style pass over the thr_spec.c, esp. simplify the
>   conditionals nesting by returning early when an error detected.
>   Remove trivial comments.
>   
>   Found by:	yuri at rawbw.com
>   PR:	200138
>   Sponsored by:	The FreeBSD Foundation
>   MFC after:	2 weeks
> 
> Modified:
>   head/lib/libthr/thread/thr_spec.c
> 

Thank you for this!

I have a lot of patches at Isilon for adding a libc cleanup function to
free all caching, for use in memory leak detectors such as valgrind. I
even fixed valgrind to utilize it. This particular code was problematic.

I wrote up an arch mail about my patch a few months ago and it ended up
being several pages about libthr. I never sent it because it became too
long. I need to revisit that work and get an arch@ discussion going on
my proposal and the various issues I ran into with libthr.

While developing this cleanup feature, and working on a malloc wrapper
(dmalloc), I had a lot of problems with pthread_mutext_t (and similar
primitives) using calloc(3). I found that David Xu actually did do most
of the work to change this but abandoned it (it's all in SVN still).
Using pthreads from a malloc wrapper is really painful to get right.

-- 
Regards,
Bryan Drewery

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/svn-src-all/attachments/20150516/a2ee0039/attachment.sig>