2017-08-16 10:03:09 |
Alberto Milone |
bug |
|
|
added bug |
2017-08-16 10:03:19 |
Alberto Milone |
linux (Ubuntu): status |
New |
In Progress |
|
2017-08-16 10:03:23 |
Alberto Milone |
linux (Ubuntu): importance |
Undecided |
Medium |
|
2017-08-16 10:03:27 |
Alberto Milone |
linux (Ubuntu): assignee |
|
Alberto Milone (albertomilone) |
|
2017-08-16 10:03:36 |
Alberto Milone |
nominated for series |
|
Ubuntu Zesty |
|
2017-08-16 10:03:36 |
Alberto Milone |
bug task added |
|
linux (Ubuntu Zesty) |
|
2017-08-16 10:03:43 |
Alberto Milone |
linux (Ubuntu Zesty): status |
New |
In Progress |
|
2017-08-16 10:03:46 |
Alberto Milone |
linux (Ubuntu Zesty): importance |
Undecided |
Medium |
|
2017-08-16 10:03:48 |
Alberto Milone |
linux (Ubuntu Zesty): assignee |
|
Alberto Milone (albertomilone) |
|
2017-08-16 10:03:55 |
Alberto Milone |
linux (Ubuntu): assignee |
Alberto Milone (albertomilone) |
|
|
2017-08-16 10:03:59 |
Alberto Milone |
linux (Ubuntu): status |
In Progress |
Triaged |
|
2017-08-16 10:09:53 |
Alberto Milone |
description |
When a timeout of zero is specified, the caller is only interested in
the fence status.
In the current implementation, dma_fence_default_wait will always call
schedule_timeout() at least once for an unsignaled fence. This adds a
significant overhead to a fence status query.
Avoid this overhead by returning early if a zero timeout is specified. |
When a timeout of zero is specified, the caller is only interested in
the fence status.
In the current implementation, dma_fence_default_wait will always call
schedule_timeout() at least once for an unsignaled fence. This adds a
significant overhead to a fence status query.
Avoid this overhead by returning early if a zero timeout is specified.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=03c0c5f6641533f5fc14bf4e76d2304197402552 |
|
2017-08-16 11:18:17 |
Alberto Milone |
description |
When a timeout of zero is specified, the caller is only interested in
the fence status.
In the current implementation, dma_fence_default_wait will always call
schedule_timeout() at least once for an unsignaled fence. This adds a
significant overhead to a fence status query.
Avoid this overhead by returning early if a zero timeout is specified.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=03c0c5f6641533f5fc14bf4e76d2304197402552 |
SRU Request:
[Impact]
When a timeout of zero is specified, the caller is only interested in
the fence status.
In the current implementation, dma_fence_default_wait will always call
schedule_timeout() at least once for an unsignaled fence. This adds a
significant overhead to a fence status query.
Avoid this overhead by returning early if a zero timeout is specified.
[Test Case]
1) Install the new kernel
2)
[Regression Potential]
Low, the patch is very small, and it is already available upstream.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=03c0c5f6641533f5fc14bf4e76d2304197402552 |
|
2017-08-17 10:34:36 |
Alvin Huan |
description |
SRU Request:
[Impact]
When a timeout of zero is specified, the caller is only interested in
the fence status.
In the current implementation, dma_fence_default_wait will always call
schedule_timeout() at least once for an unsignaled fence. This adds a
significant overhead to a fence status query.
Avoid this overhead by returning early if a zero timeout is specified.
[Test Case]
1) Install the new kernel
2)
[Regression Potential]
Low, the patch is very small, and it is already available upstream.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=03c0c5f6641533f5fc14bf4e76d2304197402552 |
SRU Request:
[Impact]
When a timeout of zero is specified, the caller is only interested in
the fence status.
In the current implementation, dma_fence_default_wait will always call
schedule_timeout() at least once for an unsignaled fence. This adds a
significant overhead to a fence status query.
Avoid this overhead by returning early if a zero timeout is specified.
[Test Case]
1) Install the new kernel
2) Download amdgpu-pro-17.30-465504.tar.xz from http://support.amd.com/en-us/kb-articles/Pages/AMDGPU-PRO-Driver-for-Linux-Release-Notes.aspx, and follow the install guide to install the driver on Ubuntu 16.04.3
3) Install Unigine Heaven to get performance score
4) Rmeove DKMS, remove this patch http://git.amd.com:8080/gitweb?p=brahma/ec/linux.git;a=commit;h=8586edda6677b90f3971919f3c6babbbec4b5b8c from DKMS source code, and install DKMS
5) re-run the benchmark, and check if there's any obvious performance drop.
[Regression Potential]
Low, the patch is very small, and it is already available upstream.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=03c0c5f6641533f5fc14bf4e76d2304197402552 |
|
2017-08-17 12:33:52 |
Stéphane Verdy |
bug |
|
|
added subscriber Stéphane Verdy |
2017-08-24 09:46:24 |
Kleber Sacilotto de Souza |
linux (Ubuntu Zesty): status |
In Progress |
Fix Committed |
|
2017-09-01 08:25:57 |
Kleber Sacilotto de Souza |
tags |
|
verification-needed-zesty |
|
2017-09-07 07:00:45 |
Alberto Milone |
tags |
verification-needed-zesty |
verification-done-zesty |
|
2017-09-11 15:21:59 |
Launchpad Janitor |
linux (Ubuntu): status |
Triaged |
Fix Released |
|
2017-09-18 09:21:44 |
Launchpad Janitor |
linux (Ubuntu Zesty): status |
Fix Committed |
Fix Released |
|
2017-09-18 09:21:44 |
Launchpad Janitor |
cve linked |
|
2017-1000251 |
|
2017-09-18 09:21:44 |
Launchpad Janitor |
cve linked |
|
2017-7541 |
|