On 25/02/13 17:49, Konrad Rzeszutek Wilk wrote:On Fri, Feb 15, 2013 at 08:12:52PM +0100, Roger Pau Monné wrote:Yes, I was going to resubmit it as part of my indirect descriptorsOn 15/02/13 19:56, Konrad Rzeszutek Wilk wrote:Can you rebase this patch on top of 'stable/for-jens-3.9' when you get a chanceWill recheck it, but blkif_completition is called with the io_lock hold,Should be backported to 3.8 stable.Lets do one thing at a time.
The patch I have in the tree (and that I've asked Jens to pull for 3.9 - so
he might have already in his tree) is the old hybrid where we still use llist
but change the loop from 'for' to 'while'.
This is the stable/for-jens-3.8 tree in git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git
Could you rebase your patch on top of that tree and just simplify the loop?
Sorry about the mess about this - but I had already pulled the trigger
so to say - hoping that Jens would pull the tree and do a git pull to Linus.
And you are absolutly sure that we don't need any extra locking when switching
over to list_head? Say if blkif_completion is called while we are
processing in blkif_queue_request and doing ?
as does blkif_queue_request (I've added extra assert_spin_locked before
every list modification).
series (which I'm hoping to send before the end of the week), but if you
are in a rush I can send it earlier, I already have it rebased on top of