[lkml]   [2004]   [Mar]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: smp dead lock of io_request_lock/queue_lock patch
On Mon, 2004-01-19 at 16:36, Martin Peschke3 wrote:
> Doug Ledford wrote:
> > On Tue, 2004-01-13 at 15:55, Marcelo Tosatti wrote:
> > > What is this mlqueue patchset fixing ? To tell you the truth, I'm not
> > > aware of the SCSI stack problems.
> >
> > OK. This is a different problem entirely. The current 2.4 scsi stack
> > has a problem handling QUEUE_FULL and BUSY status commands. Amongst
> > other things, it retries the command immediately and will basically hang
> > the SCSI bus if the low level driver doesn't intervene. Most well
> > written drivers have workaround code in them (including my aic7xxx_old
> > driver). The mlqueue patch changes it so that the mid layer implements
> > delayed retries of QUEUE_FULL and BUSY commands. It also makes the mid
> > layer recognize the situation where you get a QUEUE_FULL or BUSY command
> > status with 0 outstanding commands. This used to cause the mid layer to
> > hang on that device because there would be no command completion to kick
> > the queue back into gear.
> We also stumbled about this problem last year. Our patch wasn't
> as sophisticated as Doug's new patch. However, it worked,
> and some fix is certainly needed.

I got tied up on other stuff and Marcelo's tree went into -rc mode. So,
I waited out the next release. At this point, there is some stuff up
and available for testing. There are some very minimal patches in the
bk:// tree. The 2.4-drivers tree
is just a clone of the 2.4-for-marcelo tree because I haven't done the
drivers yet. The 2.4-everything tree includes the io_request_lock
patch, the few patches the io_request_lock patch depended on, the
softirq patch, the mlqueue fixes patch, and some more misc. stuff. I'm
not done with the 2.4-everything tree, but what I've got so far is now
working for me. I haven't done much testing yet so caveat emptor. I'll
make another announcement when I've both A) finished putting everything
in the 2.4-everything tree that I was planning on putting in and B)
completed the audit of the code to make sure the merges were all
successful. But I know Jens had asked to see the mlqueue patch in
particular, so I figured I would announce that the forward ported
version is in the 2.4-everything tree for people to look at.

The number of pure bugfix patches in the 2.4-for-marcelo tree is
actually pretty small. The biggest bugfix patch by far is the mlqueue
patch in the 2.4-everything tree, but it depends on a lot of other stuff
so it would be hard to stick it into the 2.4-for-marcelo tree. Quite a
few of the patches in the 2.4-everything tree are there for performance
reasons only, but they do a pretty good job of making the 2.4 scsi stack
not feel like it's moving at about the same speed as a heavily salted

Doug Ledford <> 919-754-3700 x44233
Red Hat, Inc.
1801 Varsity Dr.
Raleigh, NC 27606

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 14:01    [W:0.051 / U:0.036 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site