Re: Mis-backport in af_unix patch for Linux 3.10.95

Hello,

Thank you very much for the warm welcome :-)

You're right, the noblock variable is used elsewhere in the stream
receive function, so nothing is needed there after the interruptible
logic is restored for the dgram function.

Your patch looks good to me. I had picked the Linux 3.12.52 version of
the patch (where the interruptible locking was removed from the right
place in the stream receive function) onto my 3.10 kernel branch a
couple weeks ago and it has been working fine for me.

Thanks,
Sultan

On Sun, Jan 24, 2016 at 3:31 AM, Willy Tarreau [off-list ref] wrote:

Hello,

On Sun, Jan 24, 2016 at 12:10:35AM -0500, Sultan Qasim wrote:

quoted

Hello all,

I'm an outsider to the Linux kernel community, so I apologize if this
is not the right channel to mention this.

The simple fact that you participate, inspect the code and report bugs
makes you part of this community :-)  It's indeed the right place.
Usually when reporting an issue with a commit, we also CC the whole
signed-off-by / CC chain of that commit (which I'm doing now). For
bugs related to networking, we usually CC the netdev list as well.

quoted

I noticed that the
backported version of the patch "af_unix: Revert 'lock_interruptible'
in stream receive code" in Linux 3.10.95 seems to have removed the
mutex_lock_interruptible from the wrong function.

Here is the backported patch:
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/?id=3a57e783016bf43ab9326172217f564941b85b17

Here is the original:
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/net/unix/af_unix.c?id=3822b5c2fc62e3de8a0f33806ff279fb7df92432

Was it not meant to be removed from unix_stream_recvmsg instead of
unix_dgram_recvmsg?

You're absolutely right, good catch! Similar controls were added to
both functions resulting in the same code appearing there, which
confused the patch process, causing the change to be applied to the
wrong location. This happens from time to time in such circumstances
when backporting to older kernels.

quoted

Also, the variable called "noblock" needs to be
removed from the function being changed to prevent unused variable
warnings.

If you mean this variable in function unix_dgram_recvmsg(), it would
indeed report a warning but only due to the patch being mis-applied.
In unix_stream_recvmsg(), it's still used as well.

Does the attached patch seem better to you (not compile-tested) ?

Greg/Ben, both 3.2.76 and 3.14.59 are OK regarding this, it seems
like only 3.10.95 was affected.

Thanks,
Willy