Bug 34982 - iwl4965: very bad performance after update to 2.6.35
Summary: iwl4965: very bad performance after update to 2.6.35
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: Drivers
Classification: Unclassified
Component: network-wireless (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Stanislaw Gruszka
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-05-12 11:11 UTC by Stanislaw Gruszka
Modified: 2012-08-10 10:03 UTC (History)
4 users (show)

See Also:
Kernel Version: 2.6.35
Subsystem:
Regression: No
Bisected commit-id:


Attachments
iwlagn_force_cts_to_self_protection.patch (1.97 KB, text/plain)
2011-05-12 12:21 UTC, Stanislaw Gruszka
Details
Patch on 2.6.37 (1.63 KB, patch)
2011-05-25 15:45 UTC, Arnaud Launay
Details | Diff
p1 (138.01 KB, text/plain)
2011-05-26 14:29 UTC, Stanislaw Gruszka
Details
p2 (67.20 KB, text/plain)
2011-05-26 14:29 UTC, Stanislaw Gruszka
Details
p3 (172.41 KB, text/plain)
2011-05-26 14:30 UTC, Stanislaw Gruszka
Details
p4 (282.58 KB, text/plain)
2011-05-26 14:30 UTC, Stanislaw Gruszka
Details
p5 (147.26 KB, application/octet-stream)
2011-05-26 14:31 UTC, Stanislaw Gruszka
Details

Comment 1 Stanislaw Gruszka 2011-05-12 11:12:32 UTC
There are still performance problems on 4965, what need to be solved.

Arnaud, I will provide some test patches soon.
Comment 2 Stanislaw Gruszka 2011-05-12 12:21:41 UTC
Created attachment 57572 [details]
iwlagn_force_cts_to_self_protection.patch

This use cts-to-self protection of aggregated frames, like it was in 2.6.34. Patch is on top of 2.6.35.13. It is small, so could be easily applied by hand on any other kernel version (up to 2.6.38). It helps on my setup with 5300 performance, not sure it would help in case of 4965.

Please test (with previous fix of course). If it does not help, I will provide a few patches that apply on top of 2.6.35-rc1 and convert driver to 2.6.34 level. Then we could see which patch fix things and process problem further.
Comment 3 Stanislaw Gruszka 2011-05-25 13:51:00 UTC
Can we get test results of the patch? I can not reproduce problem, I have 7MB/s on 2.4GHz and 9MB/s on 5GHz network here with 4965 device.
Comment 4 Arnaud Launay 2011-05-25 15:45:00 UTC
Created attachment 59422 [details]
Patch on 2.6.37

Didn't like my 2.6.37 :)

pomme iwlwifi # patch -p5 <~/patch.cts.patch 
patching file iwl-agn.c
Hunk #1 FAILED at 200.
Hunk #2 FAILED at 3335.
Hunk #3 FAILED at 3348.
Hunk #4 succeeded at 3790 with fuzz 2 (offset 420 lines).
3 out of 4 hunks FAILED -- saving rejects to file iwl-agn.c.rej
patching file iwl-core.c
Hunk #1 FAILED at 1908.
1 out of 1 hunk FAILED -- saving rejects to file iwl-core.c.rej

Patch on it enclosed, hope it's good, I had to modify the priv-> to a ctx -> (first hunk)

Anyway it's still (very) bad. It's a bit better when mixed with the previous patch on iwl-agn-tx.c ( "If this frame is broadcast or management, use broadcast station id" ), but it's around 1.22MB/s without it, 1.6/1.8MB/s with it.

ssh still feels like it's loosing packets.

Linux pomme 2.6.37-gentoo-r4 #3 SMP Wed May 25 17:20:14 CEST 2011 x86_64 Intel(R) Core(TM)2 Duo CPU T7100 @ 1.80GHz GenuineIntel GNU/Linux

03:00.0 Network controller: Intel Corporation PRO/Wireless 4965 AG or AGN [Kedron] Network Connection (rev 61)

hardware: "IBM Lenovo" Thinkpad R61, 8918-5WG:
http://www-307.ibm.com/pc/support/site.wss/quickPath.do?quickPathEntry=89185WG
Comment 5 Stanislaw Gruszka 2011-05-26 14:27:38 UTC
Your patch is fine, so this is different problem.

Below I'll attach 5 patches on top of 2.6.35.12 (need to build and install that version to test), that revert driver to 2.6.34 level + some minor mac80211 patches. I'm asking you to test them one by one and tell after which one the bug fix. If after last patch bug will still persist, that mean the problem is somewhere else i.e in mac80211 or net stack. First patch contains fix we have already - that brings speed from 0.1 to 1.2 MB/s.
Comment 6 Stanislaw Gruszka 2011-05-26 14:29:38 UTC
Created attachment 59522 [details]
p1
Comment 7 Stanislaw Gruszka 2011-05-26 14:29:59 UTC
Created attachment 59532 [details]
p2
Comment 8 Stanislaw Gruszka 2011-05-26 14:30:19 UTC
Created attachment 59542 [details]
p3
Comment 9 Stanislaw Gruszka 2011-05-26 14:30:44 UTC
Created attachment 59552 [details]
p4
Comment 10 Stanislaw Gruszka 2011-05-26 14:31:06 UTC
Created attachment 59562 [details]
p5
Comment 11 Stanislaw Gruszka 2012-01-15 10:20:40 UTC
Hi, bugzilla.kernel.org is back :-)

Arnaud, please install 2.6.35.12 and apply each of pX patches one by one and tell us after which one problem gone (if it gone, if not gone after p5, this will be a mac80211 or other subsystem problem). Otherwise this bug will not be fixed.

Oh, you may want to check the latest kernel (3.3-rc1) before applying the patches, it contains various iwlegacy patches, which by a chance could fix this issue.
Comment 12 Stanislaw Gruszka 2012-05-12 22:45:41 UTC
Without patches testing this bug will not be fixed.
Comment 13 Arnaud Launay 2012-08-10 10:03:56 UTC
Hello,
Unfortunately I can no longer test it, the R61 died some months ago. I now have a T530 which just work...
Sorry :(

Note You need to log in before you can comment on or make changes to this bug.