[pulseaudio-discuss] [PATCH] iochannel: Strictly specify PF_UNIX ancillary data boundaries

Tanu Kaskinen tanuk at iki.fi
Thu Nov 17 17:42:44 UTC 2016

On Tue, 2016-11-15 at 19:11 +0100, David Henningsson wrote:
> Interesting bug :-)
> Just so I get this right - this could just as well be fixed with instead 
> changing msg_controllen to this:
> mh.msg_controllen = CMSG_SPACE(sizeof(int) * nfd);
> That said, your version works equally well and is slightly cleaner. I 
> don't recall we have any problems with VLAs.

Have we had any VLAs before? If not, it's no wonder that we haven't had
problems with them...

I'm not too worried about VLAs, though. I don't expect people to build
pulseaudio with very exotic compilers. If someone has problems,
hopefully a bug report will be filed, and we can come up with another

> Perhaps adding a comment to your finding and a registered kernel 
> bug/patch/etc would be helpful.

I applied the patch now without any additional comments (I forgot to
add the Acked-by tag before pushing, sorry). I would have added the
comment myself, but I don't know if Ahmed filed any kernel bug. I agree
that a comment would be useful, if only to warn people to not
reintroduce this bug if VLAs turn out to be a problem at some point.
The comment can be submitted in a separate patch once we have the
kernel bug/patch reference, or a decision that no kernel bug or patch
will be submitted.



More information about the pulseaudio-discuss mailing list