Re: [RFC] aio_down() for i386 and x86_64

[Suparna Bhattacharya <suparna@in.ibm.com>]

On Wed, Jun 15, 2005 at 03:18:30PM -0400, Benjamin LaHaise wrote:
> On Wed, Jun 15, 2005 at 10:23:49PM +0530, Suparna Bhattacharya wrote:
> > Interesting approach - using ki_wait.private for this.
> > Could we make aio_down take a wait queue parameter as well instead of
> > the iocb ?
> 
> Hmmm, I guess there might be instances where someone has to wait on 
> multiple wait queues.  Will add that to the next version of the patch.
> 
> > Need to think a little about impact on io cancellation.
> 
> It should be possible to cancel semaphore operations fairly easily -- 
> the aio_down function can set ->ki_cancel to point to a semaphore cancel 
> routine.  I'll give coding that a try.
> 
> > BTW, is the duplication of functions across architectures still needed ? I
> > thought that one of advantages of implementing a separate aio_down
> > routine vs modifiying down to become retryable was to get away from
> > that ... or wasn't it ?
> 
> Good point.  The fast path for down() will probably need to remain a 
> separate function, but we could well unify the code with the 
> down_interruptible() codepath.
> 
> > Meanwhile, I probably need to repost my aio_wait_bit patches - there
> > may be some impact here.
> 
> Sure -- any version of those would be useful to build on.  Cheers!

http://www.kernel.org/pub/linux/kernel/people/suparna/aio/2610-rc2/ has
the patchset. 

I just updated the AIO wait bit ones to 2.6.12-rc6, will post them
in a separate thread.

Regards
Suparna

> 
> 		-ben
> --
> To unsubscribe, send a message with 'unsubscribe linux-aio' in
> the body to majordomo@kvack.org.  For more info on Linux AIO,
> see: http://www.kvack.org/aio/
> Don't email: <a href=mailto:"aart@kvack.org">aart@kvack.org</a>

-- 
Suparna Bhattacharya (suparna@in.ibm.com)
Linux Technology Center
IBM Software Lab, India