Rik,
I have added calls to run_task_queue(&tq_disk) to the code, and it makes
the AIO performance go way up. I noticed that the current task is
always set to UNINTERRUPTIBLE right before this is called for some
reason. I think I have another problem, though, because I keep getting
an Ooops with the swapper process. I think it's related to the memory
pools I am allocating buffer heads from. I am still running this one
down. Thought we had it, but it is still showing up during regression
testing with lots of swap activity.
:-)
Jeff
Rik van Riel wrote:
>
> On Tue, 11 Apr 2000, Jeff V. Merkey wrote:
>
> > 1. ll_rw_blk (and add_request/make_request) (oink, oink.....
> > oink, oink ... rooting around down in the hardware -- I think
> > it's looking for truffles)
> >
> > and a close second:
> >
> > 2. ide_do_request
> > ide_delay_50ms (huge!!!)
> > ide_ioctl
> >
> > I am also seeing an occasional Ooops when using **FULL** linux AIO (I
> > call ll_rw_blk and don't call wait on buffer, instead signal the NWFS
> > LRU directly from the b_end_io callback routine). Also, do I have to
> > call run_task_queue(&tq_disk) call all the time to get the AIO subsystem
> > to post the buffer-head request. If I don't call it, then the AIO's
> > just sit there and get processed about one every 10 seconds (???).
>
> Ben LaHaise and me were talking about run_taskqueue(&tq_disk),
> which had "magically" disappeared from various places in the
> kernel source.
>
> Maybe this should be called every clock tick if:
> - outstanding swap requests
> - outstanding io requests
>
> Could you add something like this and see if it helps?
>
> (no, I have no idea if it would help or not, but maybe it
> does)
>
> Rik
> --
> The Internet is not a network of computers. It is a network
> of people. That is its real strength.
>
> Wanna talk about the kernel? irc.openprojects.net / #kernelnewbies
> http://www.conectiva.com/ http://www.surriel.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Apr 15 2000 - 21:00:19 EST