iwlwifi: Intel 3945ABG speed issues finally identified.

iwlwifi: Intel 3945ABG speed issues finally identified.

[Eric Appleman]

I spent most of the day bisecting and testing numerous kernels and my work has
finally paid off.

I've found 2 candidate commits.

a6866ac93e6cb68091326e80b4fa4619a5957644
1402364162afbaac1b8a74ee21aeb013e817ac7d

iwl3945 definitely broke between 2.6.35-rc2 and -rc3.

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Eric Appleman]

2 more commits that I missed

7d47618a2ade0cb6d8a0b2597029c383c1662fa0
6db6340c42d027b6364d49fa99d69019aca24de4

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[John W. Linville]

On Thu, Dec 23, 2010 at 06:15:18AM +0000, Eric Appleman wrote:
> I spent most of the day bisecting and testing numerous kernels and my work has
> finally paid off.
> 
> I've found 2 candidate commits.
> 
> a6866ac93e6cb68091326e80b4fa4619a5957644
> 1402364162afbaac1b8a74ee21aeb013e817ac7d
> 
> iwl3945 definitely broke between 2.6.35-rc2 and -rc3.

On Thu, Dec 23, 2010 at 08:39:32AM +0000, Eric Appleman wrote:
> 2 more commits that I missed
> 
> 7d47618a2ade0cb6d8a0b2597029c383c1662fa0
> 6db6340c42d027b6364d49fa99d69019aca24de4

Did you revert any/all of these patches?  Can you quantify what effect
that had?

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Guy, Wey-Yi]

On Wed, 2010-12-22 at 22:15 -0800, Eric Appleman wrote:
> I spent most of the day bisecting and testing numerous kernels and my work has
> finally paid off.
> 
> I've found 2 candidate commits.
> 
> a6866ac93e6cb68091326e80b4fa4619a5957644
> 1402364162afbaac1b8a74ee21aeb013e817ac7d
> 
> iwl3945 definitely broke between 2.6.35-rc2 and -rc3.
> 
> --
For a6866ac93e6cb68091326e80b4fa4619a5957644, did you see any indication
in dmesg for firmware reload?

Wey

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Eric Appleman]

I have not tried testing the commits individually since kernel compilation is a
pain for me.

If I can trick a compat-wireless tree into a code splice, maybe I can test each
commit.

Do you guys know any tricks to compile iwlwifi individually?

However, I can say with utmost certainty that the regression is in the window
between rc2 and rc3 unless the Ubuntu daily kernels were mislabeled.

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Eric Appleman]

Hmmm.

Both "fix internal scan" and "stuck queue detection" made the cut for the rc2
kernel. Said kernel is unaffected by the bug.

For rc3, "rcu lock" and "move sysfs" both make the cut. "serialize station
management" barely misses the cut.

Does the sysfs commit apply to the 3945 or is it only for newer cards?

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Eric Appleman]

Btw, I've never seen a firmware reload message in association with this bug,
regardless of commit, whether it be from May or December.

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Guy, Wey-Yi]

On Thu, 2010-12-23 at 22:16 -0800, Eric Appleman wrote:
> Btw, I've never seen a firmware reload message in association with this bug,
> regardless of commit, whether it be from May or December.

> then most likely "stuck queue detection" is not the problem.

Wey
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Guy, Wey-Yi]

On Thu, 2010-12-23 at 18:08 -0800, Eric Appleman wrote:
> Hmmm.
> 
> Both "fix internal scan" and "stuck queue detection" made the cut for the rc2
> kernel. Said kernel is unaffected by the bug.
> 
> For rc3, "rcu lock" and "move sysfs" both make the cut. "serialize station
> management" barely misses the cut.
> 
> Does the sysfs commit apply to the 3945 or is it only for newer cards?
> 
> -
"move sysfs_create_group to post request firmware" patch only apply to
_agn, not for 3945. so it is not the problem :-(

Wey

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Eric Appleman]

So, unless I find something between 2.6.34 release and 2.6.35-rc2 (no changes
between rc1 and rc2), that leaves "fix internal scan", doesn't it?

Btw, this bug is ridiculously hard to test. The early rc kernels for 2.6.35 are
a pain in the ass to test because the iwl3945 driver was in flux at the time. If
I push too hard with speedtests, the driver will cause a kernel panic.

Re: iwlwifi: Intel 3945ABG speed issues finally identified.

[Stanislaw Gruszka]

On Thu, Dec 23, 2010 at 06:15:18AM +0000, Eric Appleman wrote:
> I spent most of the day bisecting and testing numerous kernels and my work has
> finally paid off.
> 
> I've found 2 candidate commits.
> 
> a6866ac93e6cb68091326e80b4fa4619a5957644
> 1402364162afbaac1b8a74ee21aeb013e817ac7d
> 
> iwl3945 definitely broke between 2.6.35-rc2 and -rc3.

For sure breakage is between -rc2 and -rc3 ?

We have these commits:
$ git log --pretty=oneline v2.6.35-rc2..v2.6.35-rc3 -- net/mac80211/ drivers/net/wireless/iwlwifi/
b054b747a694927879c94dd11af54d04346aed7d mac80211: fix deauth before assoc
6db6340c42d027b6364d49fa99d69019aca24de4 iwlwifi: add missing rcu_read_lock
35dd0509b21e4b5bab36b9eb80c8dab0322f5007 mac80211: fix function pointer check
7d47618a2ade0cb6d8a0b2597029c383c1662fa0 iwlwifi: move sysfs_create_group to post request firmware
1402364162afbaac1b8a74ee21aeb013e817ac7d iwl3945: fix internal scan
a6866ac93e6cb68091326e80b4fa4619a5957644 iwl3945: enable stuck queue detection on 3945
8b9a4e6e442756f670ef507f09bbc6c11dc0fca6 mac80211: process station blockack action frames from work

Also there are some general network commits.

What I can tell, none of them could cause performance problems in
obvious way. If breakage is really between 2.6.35 -rc2 and -rc3, could
you install 2.6.35 kernel from source, and revert these commits 
one by one to see revert of which commit fix the problem?

Stanislaw