Re: linux-next: manual merge of the kvm tree with Linus' tree

[Gregory Haskins <ghaskins@novell.com>]

[-- Attachment #1: Type: text/plain, Size: 2534 bytes --]

Avi Kivity wrote:
> On 07/01/2009 10:10 AM, Davide Libenzi wrote:
>> On Wed, 1 Jul 2009, Stephen Rothwell wrote:
>>
>>   
>>> Hi Avi,
>>>
>>> Today's linux-next merge of the kvm tree got a conflict in fs/eventfd.c
>>> between commit 133890103b9de08904f909995973e4b5c08a780e ("eventfd:
>>> revised interface and cleanups") from Linus' tree and commit
>>> 28ddf0aebbf546e56efd1951725d5457ce1ebf98 ("eventfd: Allow waiters to be
>>> notified about the eventfd file* going away") from the kvm tree.
>>>
>>> Overlapping changes.  I fixed it up (see below), but don't know if this
>>> is the correct fix.
>>>
>>> -- 
>>> Cheers,
>>> Stephen Rothwell                    sfr@canb.auug.org.au
>>>
>>> diff --cc fs/eventfd.c
>>> index 31d12de,72f5f8d..0000000
>>> --- a/fs/eventfd.c
>>> +++ b/fs/eventfd.c
>>> @@@ -105,8 -63,13 +105,13 @@@ static int eventfd_release(struct inod
>>>    {
>>>        struct eventfd_ctx *ctx = file->private_data;
>>>
>>> -     wake_up_poll(&ctx->wqh, POLLHUP);
>>> +     /*
>>> +      * No need to hold the lock here, since we are on the file
>>> cleanup
>>> +      * path and the ones still attached to the wait queue will be
>>> +      * serialized by wake_up_locked_poll().
>>> +      */
>>> +     wake_up_locked_poll(&ctx->wqh, POLLHUP);
>>>   -    kfree(ctx);
>>>   +    eventfd_ctx_put(ctx);
>>>        return 0;
>>>    }
>>>      
>>
>> That's fine.
>> There are a couple of extra spaces before the last two -+ in that patch
>> though ;)
>>    
>
> No, that's a git N-way diff format.  The first column shows the
> changes relative to mainline by kvm.git, and the second the changes to
> kvm.git made by mainline.
>
> I've merged and will push soon, which will resolve the conflict, but I
> think the patch wake_up_locked_poll() is better off in mainline rather
> than kvm.git.

Just to be clear: the final form will need to be wake_up_poll(), not
wake_up_locked_poll.  However, I think what Steven did was optimal
because converting to the locked form without coordinating with kvm.git
would break bisectability.

In the irqfd-fixes series, I had 3 patches related to this situation
(1/5 to prepare to change, 2/5 was Davide's patch, and 3/5 did the final
change-over).  Now that the vast majority of Davide's work is in
mainline+kvm.git, here is my proposal:

*) drop 2/5 (already upstream, sans the locked POLLHUP)
*) fold 1/5 + 3/5, and add new fs/eventfd.c hunk to convert to locked
variant
*) drop 5/5

Sound good?
-Greg




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 266 bytes --]