Re: [patch] pipe: add support for shrinking and growing pipes

[Jens Axboe <jens.axboe@oracle.com>]

On Thu, May 20 2010, Miklos Szeredi wrote:
> On Wed, 19 May 2010, Jens Axboe wrote:
> > On Wed, May 19 2010, Jens Axboe wrote:
> > > On Wed, May 19 2010, Linus Torvalds wrote:
> > > > 
> > > > 
> > > > On Wed, 19 May 2010, Miklos Szeredi wrote:
> > > > > 
> > > > > One issue I see is that it's possible to grow pipes indefinitely.
> > > > > Should this be restricted to privileged users?
> > > > 
> > > > Yes. But perhaps only if it grows past the default (or perhaps "default*2" 
> > > > or similar). That way a normal user could shrink the pipe buffers, and 
> > > > then grow them again if he wants to.
> > > 
> > > That's still a bit arbitrary, I don't think allowing default*2 only for
> > > non-root is going to be hugely interesting. But limiting makes sense,
> > > but lets at least allow a larger max limit for the normal user. I'm
> > > suspecting that the media application that wants to use this will not be
> > > running as root, and we don't make the feature properly available to the
> > > ones that want to use it, then we may as well not do it.
> > > 
> > > Or we could expose a sysctl for instance that holds the max non-root
> > > size. And make that default to default*16 or something. How does that
> > > sound?
> > > 
> > > > Oh, and I think you need to also require that there be at least two 
> > > > buffers. Otherwise we can't guarantee POSIX behavior, I think.
> > > 
> > > Good point, and at least that part is easily doable :-)
> > 
> > So I updated the patch, that branch was pretty ancient... The fcntl pipe
> > numbers were also screwed up, so got that fixed. New patch is here:
> > 
> > http://git.kernel.dk/?p=linux-2.6-block.git;a=commit;h=23dcb845246946aeda5a5e398c6911381ad28365
> 
> Not sure why you didn't take my updated patch.  Yours misses
> conversion of kernel/trace/trace.c as well as some hunks below (that
> renaming the "pages" and "partial" arrays and a compile test would
> have revealed).

Sorry, I missed that you had done extra changes. I had to rebase the
whole thing anyway, and as I said it was totally untested. I'll get it
updated.

The main reason I didn't push it before is that I wasn't completely sure
that we wanted to use fcntl() for this. But Linus doesn't seem to object
on that side at least.

-- 
Jens Axboe