Re: [Qemu-devel] [PATCH 2/2] qemu-ga: Add the guest-suspend command

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1472 bytes --]

On 01/17/2012 03:57 AM, Jamie Lokier wrote:
> You're right, but it looks like it may be removed in the next edition:
> 
>    https://www.opengroup.org/austin/docs/austin_446.txt
> 
>> it was only pthread_atfork() which was removed.
> 
> I didn't think pthread_atfork() ever was async-signal-safe.
> 
>> That is, fork() is _required_
>> to be async-signal-safe (and usable from signal handlers), provided that
>> the actions following the fork also follow safety rules.
> 
> Nonethless, Glibc fork() isn't async-signal-safe even if it should be:
> 
>     http://sourceware.org/bugzilla/show_bug.cgi?id=4737

Thanks for the (depressing) pointers.  You posted the link to the Austin
Group meeting where fork() was discusses; here's a further link to the
actual defect and resolution, which is that the next version of POSIX
_will_ be removing fork() from the list of async-signal-safe functions,
by replacing it with _Fork() which does _not_ call any of the
pthread_atfork() handlers:

http://austingroupbugs.net/view.php?id=62

You are right that the only reason that fork() is not signal-safe is
because of pthread_atfork(), so I was almost right in my above
characterization that pthread_atfork() was the culprit.  Maybe we should
start probing at configure time whether _Fork already exists, and if so,
use it instead of fork().

-- 
Eric Blake   eblake@redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 620 bytes --]