MIME-Version: 1.0
In-Reply-To: <CAMgrJ3b7yZru5P-bYpmmfjYCoSJqUhAMrTzfKnKLPN3CgdeJ4w@mail.gmail.com>
References: <CAMgrJ3Y9=K4umoLD=htrrW-Wn4Hgi7qfaKDpY55_GKd=KgD==A@mail.gmail.com>
 <CAJ-VmonoNE1azzwZrjr+BX2un=oz9D_Tf3AAWdw9twJjyzFMOA@mail.gmail.com>
 <CAJ-Vmo=XMnm9b3t102q305Kj2LiVqEhTJ+X1+mrW=1X8cvArSg@mail.gmail.com>
 <201303251712.19808.chunkeey@googlemail.com> <CAJ-VmonVKm+SQrbB_D+0R-V5airK8fUZpE2Y_TK22TuMczMu6A@mail.gmail.com>
 <CAMgrJ3afo=woPwXo6HO16b1CPuo4aRYM+w0__gQQYRvvJT3D-g@mail.gmail.com>
 <CAJ-Vmok8p9T0anY_1zGvArA5tZ4Fx4gdHHTVmSOGgprfNDqS+w@mail.gmail.com>
 <CAMgrJ3Zy1EKBcdcVouK9=r6=ErU8R72-=aC149VPT1Cg9=wiTg@mail.gmail.com>
 <5151D752.3080303@candelatech.com> <CAMgrJ3b7yZru5P-bYpmmfjYCoSJqUhAMrTzfKnKLPN3CgdeJ4w@mail.gmail.com>
From: Robert Shade <robert.shade@gmail.com>
Date: Tue, 26 Mar 2013 23:11:53 -0400
Message-ID: <CAMgrJ3bENvfezMSge9_oOJiSsoMpMVFGk+QaSgsHSDASczgn0A@mail.gmail.com> (sfid-20130327_041226_241421_C6B8095B)
Subject: Re: [ath9k-devel] Auth Packet TX Delay
To: Ben Greear <greearb@candelatech.com>,
	Adrian Chadd <adrian@freebsd.org>
Cc: Christian Lamparter <chunkeey@googlemail.com>,
	ath9k-devel@venema.h4ckr.net, linux-wireless@vger.kernel.org,
	Marco Fonseca <marco@tampabay.rr.com>
Content-Type: text/plain; charset=ISO-8859-1
Sender: linux-wireless-owner@vger.kernel.org

Think I found it:  In ath_reset_internal, before we reset, we call
ath_prepare_reset, which calls ath9k_hw_disable_interrupts.  If the
channel change fails, we never call ath_complete_reset, which calls
ath9k_hw_enable_interrupts.

ath9k_hw_{enable|disable}_interrupts calls definitely needs to be
balanced.  Looking at an old log, IER never gets re-enabled after the
channel set failure.

Any suggestions on the proper cleanup?  Could we just call
ath_complete_reset anyway?  Some sort of special handling?

On Tue, Mar 26, 2013 at 8:55 PM, Robert Shade <robert.shade@gmail.com> wrote:
> On Tue, Mar 26, 2013 at 1:13 PM, Ben Greear <greearb@candelatech.com> wrote:
>> Can you cat out the /debug/ieee80211/wiphy0/ath9k/xmit
>> file when it gets in this hung state?  I saw issues where
>> the xmit queues got hung in our AR9380 NIC systems and
>> ended up adding an ugly bit of cleanup/reset logic to
>> get them working again...
>
> Dumps are attached on to the bug.  I remember being hopeful when
> reading your thread that we stumbled on the same issue, but the
> symptoms don't look the same to me because pending-frames goes back to
> 0 in my case.