io_request_lock patch?
I have heard that a patch to reduce io_request_lock contention by breaking it into per queue locks was released in the past. Does anyone know where I could find this patch if it exists?
Thanks,
Jonathan
Signature · 5 lines
Jonathan Lahr IBM Linux Technology Center Beaverton, Oregon [email protected] 503-578-3385
Quoted 4 lines · Jonathan Lahr
I have heard that a patch to reduce
io_request_lockcontention by breaking it into per queue locks was released in the past. Does anyone know where I could find this patch if it exists?
I had a patch about a year ago that did it safely for the block layer and IDE at least, and also for selected SCSI hba's. Some of the latter variety are pretty hard and/or tedious to fixup, Eric Y has done some work automating this process almost completely. Until that is done, the general patch has no chance of being integrated.
It's also interesting to take a look at _why_ there's contention on the io_request_lock. And fix those up first.
Signature · 1 line
Jens Axboe
Quoted 11 lines · Jens Axboe [[email protected]] wrote:
Quoted 4 lines · Jonathan Lahr
I have heard that a patch to reduce
io_request_lockcontention by breaking it into per queue locks was released in the past. Does anyone know where I could find this patch if it exists?I had a patch about a year ago that did it safely for the block layer and IDE at least, and also for selected SCSI hba's. Some of the latter variety are pretty hard and/or tedious to fixup, Eric Y has done some work automating this process almost completely. Until that is done, the general patch has no chance of being integrated.
I am investigating reducing io_request_lock contention in the shorter term if possible with smaller incremental modifications. So I'm first trying to discover any previous work that might have been done toward this purpose.
Signature · 5 lines
Jonathan Lahr IBM Linux Technology Center Beaverton, Oregon [email protected] 503-578-3385
The bit that I had automated was to essentially fix each and every
low-level SCSI driver such that each low-level driver would be responsible for it's own locking. At this point the patches and the tool are on hold - once the 2.5 kernel series gets underway, I can generate some fairly massive patchsets.
-Eric
----- Original Message -----
From: "Jonathan Lahr" <[email protected]>
To: "Jens Axboe" <[email protected]>
Cc: <[email protected]>; <[email protected]>
Sent: Tuesday, July 10, 2001 3:49 PM
Subject: Re: io_request_lock patch?
Quoted 7 lines
Quoted 6 lines · Jens Axboe [[email protected]] wrote:
Quoted 4 lines · Jonathan Lahr
I have heard that a patch to reduce
io_request_lockcontention by breaking it into per queue locks was released in the past. Does anyone know where I could find this patch if it exists?
Quoted 8 lines
Quoted 5 lines
I had a patch about a year ago that did it safely for the block layer and IDE at least, and also for selected SCSI hba's. Some of the latter variety are pretty hard and/or tedious to fixup, Eric Y has done some work automating this process almost completely. Until that is done, the general patch has no chance of being integrated.
I am investigating reducing
io_request_lockcontention in the shorter term if possible with smaller incremental modifications. So I'm first trying
to
Quoted 13 lines
discover any previous work that might have been done toward this purpose.
--
Jonathan Lahr
IBM Linux Technology Center
Beaverton, Oregon
[email protected]
503-578-3385-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected]
Quoted 6 lines · Eric Youngdale
The bit that I had automated was to essentially fix each and every
low-level SCSI driver such that each low-level driver would be responsible for it's own locking. At this point the patches and the tool are on hold - once the 2.5 kernel series gets underway, I can generate some fairly massive patchsets.
Cool, sounds good Eric. I'll probably go ahead and complete the initial ll_rw_blk and IDE work, along with a few selected SCSI drivers. Then leave the grunt of the work for your tools.
Signature · 1 line
Jens Axboe