Messages in this thread | | | Date | Tue, 9 Apr 2013 14:08:06 -0400 (EDT) | From | Mikulas Patocka <> | Subject | Re: dm-crypt parallelization patches |
| |
On Tue, 9 Apr 2013, Tejun Heo wrote:
> On Tue, Apr 09, 2013 at 01:51:43PM -0400, Mikulas Patocka wrote: > > The patch dm-crypt-sort-requests.patch sorts write requests submitted by a > > single thread. The requests are sorted according to the sector number, > > rb-tree is used for efficient sorting. > > Hmmm? Why not just keep the issuing order along with plugging > boundaries?
What do you mean?
I used to have a patch that keeps order of requests as they were introduced, but sorting the requests according to sector number is a bit simpler.
> > So it seems that CFQ has some deficiency that it cannot merge adjacent > > requests done by different processes. > > As I wrote before, please use bio_associate_current(). Currently, > dm-crypt is completely messing up all the context information that cfq > depends on to schedule IOs. Of course, it doesn't perform well.
bio_associate_current() is only valid on a system with cgroups and there are no cgroups on the kernel where I tested it. It is an empty function:
static inline int bio_associate_current(struct bio *bio) { return -ENOENT; }
Mikulas
> Thanks. > > -- > tejun >
| |