Re: [patch 1/2] epoll fix own poll()

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 29 Jan 2009 10:16:31 -0800 (PST) Davide Libenzi <davidel@xmailserver.org> wrote:

> On Thu, 29 Jan 2009, Andrew Morton wrote:
> 
> > >  fs/eventpoll.c |  510 +++++++++++++++++++++++++++++++++------------------------
> > >  1 file changed, 304 insertions(+), 206 deletions(-)
> > 
> > Holy cow man, this patch is HUGE!  I don't have a clue what it does nor
> > how it does it.  I'd be somewhat scared to merge it into 2.6.29.  How
> > serious is this bug?
> 
> It is a 3 in a scale of 5. The reason the patch is HUGE is because the 
> epoll ->poll() code now has to perform an operation similar to what was 
> performing in epoll_wait(), and under the same constraints (check out for 
> recursions and too long nesting chains) that were checked in the wakeups.
> So instead of duplicating the code, I made the two core operations such 
> that they get a function pointer for the core operation they have to 
> perform. That required some code movement.
> 

But which kernel version are you looking to get this merged into?

> 
> You always confuse me with your comments. Before you comment, then you 
> merge w/out giving me time to change.
>

yeah, I often do that.  If I see nothing fatal in the patch I'll merge
it so that it gets a bit of testing and so that it doesn't get
forgotten about.  Especially if it's a bugfix.

Sometimes I'll send a reply and the originator goes to sleep for weeks,
and stuff could even get lost.  I hate losing stuff.  I'll mark the patch
as "needs an update" in the series file so I don't accidentally merge it
upstream.

> Would you like the updated patches?

Sure, when convenient.