Messages in this thread | | | Date | Wed, 24 Feb 2010 10:12:48 +0900 | From | Kiyoshi Ueda <> | Subject | Re: [Stable-review] [93/93] dm mpath: fix stall when requeueing io |
| |
Hi Alasdair, Linus,
On 02/24/2010 03:12 AM +0900, Alasdair G Kergon wrote: > On Mon, Feb 22, 2010 at 07:16:34PM +0900, Kiyoshi Ueda wrote: >> On 02/22/2010 01:07 AM +0900, Stefan Bader wrote: >>>> @@ -1568,12 +1575,16 @@ static void dm_request_fn(struct request >>>> >>>> blk_start_request(rq); >>>> spin_unlock(q->queue_lock); >>>> - map_request(ti, rq, md); >>>> + if (map_request(ti, rq, md)) >>>> + goto requeued; >>>> spin_lock_irq(q->queue_lock); >>>> } >> In the current device-mapper code, I would like to go with >> spin_unlock/lock here. >> However, there was a case to enable irq in map_requst() for request >> allocation, and this spin_lock_irq was a work-around for the case. >> Now, there is no such case in the device-mapper code, so spin_lock should >> be enough here. But I'm still using spin_lock_irq for safeness, since >> there might be some more cases to enable irq during request submission >> to underlying devices. >> I'll remove the _irq in the future after lots of testings. > > So, have I understood your reasoning? > > - This function (dm_request_fn) is always called with local interrupts disabled. > E.g. from generic_unplug_device() or blk_run_queue(). > > - The 'map_request()' function was found to re-enable interrupts in one case, but > that case got fixed. > > - The code still uses spin_lock_irq to ensure they remain disabled as protection > against there being other cases. This should be changed to spin_lock as a clean-up > but you are not aware of any current breakage.
That's correct. I think the spin_lock_irq can be changed to spin_lock as a clean-up. But I don't want to break things in this late stage of 2.6.33-rc and/or the stable tree. So I'll send the clean-up patch for 2.6.34 once I make sure it's ok.
Thanks, Kiyoshi Ueda
| |