[ath9k-devel] big while doing DMA?

[Kai Scharwies <ath9k-devel@scharwies.de>]

2012/5/16 Holger Schurig <holgerschurig@googlemail.com>:
> Power save was off, from "iw wlan0 get power_save":
> Power save: off
>
> from "dmesg | grep Atheros":
> ? ?ieee80211 phy0: Atheros AR9285 Rev:2 mem=0xf9000000, irq=16
>
> A built-in pci-express card. From "lspci":
> ? ?01:00.0 Network controller [0280]: Atheros Communications Inc.
> AR9285 Wireless Network Adapter (PCI-Express) [168c:002b] (rev 01)
>
> I think ASPM is off, (some),from "lspci -vv":
> ? ? ? ? ? ? ? ?LnkCap: Port #0, Speed 2.5GT/s, Width x1, ASPM L0s L1,
> Latency L0 <512ns, L1 <64us
> ? ? ? ? ? ? ? ? ? ? ? ?ClockPM- Surprise- LLActRep- BwNot-
> ? ? ? ? ? ? ? ?LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- Retrain- CommClk-
> ? ? ? ? ? ? ? ? ? ? ? ?ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel

[ 3371.703890] ath: phy0: DMA failed to stop in 10 ms AR_CR=0xffffffff
AR_DIAG_SW=0xffffffff DMADBG_7=0xffffffff
[ 3371.713975] ath: phy0: Could not stop RX, we could be confusing the
DMA engine when we start RX up
[ 3371.723036] ------------[ cut here ]------------
[ 3371.727780] WARNING: at
/root/compat-wireless-3.5-rc3-1-sn/drivers/net/wireless/ath/ath9k/recv.c:494
ath_stoprecv+0xcb/0xe0 [ath9k]()
[ 3371.739892] Modules linked in: ipt_MASQUERADE iptable_nat nf_nat
nf_conntrack_ipv4 nf_conntrack nf_defrag_ipv4 ip_tables x_tables ipv6
arc4 ath9k mac80211 cfg80211 ath9k_common ftdi_sio ath9k_hw rtc_cmos
evdev usbserial ath rtc_core rtc_lib compat ext4 jbd2 crc16 ohci_hcd
ehci_hcd r6040 usbcore libphy [last unloaded: scsi_wait_scan]
[ 3371.770601] Pid: 840, comm: kworker/u:2 Not tainted 2.6.37-vortex86-sg #1
[ 3371.777465] Call Trace:
[ 3371.780000]  [<c102376f>] ? warn_slowpath_common+0x5f/0x80
[ 3371.785583]  [<d0e67d0b>] ? ath_stoprecv+0xcb/0xe0 [ath9k]
[ 3371.791156]  [<c102379d>] ? warn_slowpath_null+0xd/0x10
[ 3371.796475]  [<d0e67d0b>] ? ath_stoprecv+0xcb/0xe0 [ath9k]
[ 3371.802057]  [<d0e63e27>] ? ath_prepare_reset+0x47/0xc0 [ath9k]
[ 3371.808074]  [<d0e640a6>] ? ath_reset_internal+0x56/0x170 [ath9k]
[ 3371.814265]  [<d0e64aee>] ? ath_reset+0x1e/0x80 [ath9k]
[ 3371.819582]  [<c10388b4>] ? process_one_work+0x104/0x390
[ 3371.824989]  [<d0e64b50>] ? ath_reset_work+0x0/0x10 [ath9k]
[ 3371.830650]  [<c1038e07>] ? worker_thread+0xe7/0x3a0
[ 3371.835699]  [<c1038d20>] ? worker_thread+0x0/0x3a0
[ 3371.840674]  [<c103bdb4>] ? kthread+0x74/0x80
[ 3371.845113]  [<c103bd40>] ? kthread+0x0/0x80
[ 3371.849464]  [<c1002f56>] ? kernel_thread_helper+0x6/0x10
[ 3371.854941] ---[ end trace bf8d0dd91c11de04 ]---
[ 3371.950422] ath: phy0: Failed to stop TX DMA, queues=0x10f!
[ 3372.095127] ath: phy0: Chip reset failed
[ 3372.099122] ath: phy0: Unable to reset channel, reset status -22
[ 3372.219094] Clocksource tsc unstable (delta = 63274046 ns)
[ 3372.225271] Switching to clocksource pit

I am seeing above error log, which is similar to yours, when
establishing a mesh network using ath9k cards and running iperf.
A Mini PCI Express card on one side is not showing problems, but
multiple WNC DNMA-92 (ath9220 chipset) used on some small sized x86
SBCs do.
The crashes usually happen while benchmarking the mesh connection
(using HT40 bandwidth) with poor antenna conditions. This includes
great distances between the nodes, disconnecting the antennas or
plugging attenuators between the cards and the antennas, all resulting
in an unstable network, so the nodes timeout and rediscover eachothers
regularly. This however is part of a use case I am trying to
accomplish. In my opinion poor antenna conditions should not result in
driver crashes...

ASPM should not be an issue, as mentioned card is a mini PCI card.
I am not sure about ANI and powersaving features, which I read about
in other threads.

root at meshnode13:~# iw dev
phy#0
       Interface mesh0
               ifindex 4
               type mesh point
               channel 13 (2472 MHz) HT40-

root at meshnode13:~# iw mesh0 get power_save
Power save: on
root at meshnode13:~# iw mesh0 set power_save off
command failed: Operation not supported (-95)

It seems power_save is not available for mesh mode, but is reported
false as active?

root at meshnode13:~# lspci -vv

00:03.0 Network controller: Atheros Communications Inc. AR922X
Wireless Network Adapter (rev 01)
       Subsystem: Atheros Communications Inc. Device 2096
       Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop-
ParErr- Stepping- SERR+ FastB2B- DisINTx-
       Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium
>TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
       Latency: 168, Cache Line Size: 64 bytes
       Interrupt: pin A routed to IRQ 5
       Region 0: Memory at fefd0000 (32-bit, non-prefetchable) [size=64K]
       Capabilities: [44] Power Management version 2
               Flags: PMEClk- DSI- D1- D2- AuxCurrent=100mA
PME(D0+,D1-,D2-,D3hot+,D3cold-)
               Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
       Kernel driver in use: ath9k

As I am using a pretty recent compat-wireless Version all patches
mentioned in the mailinglists should already be applied.
Having used a couple of different cw builds the error seems to exist
for some time, or could have been even worse, as I remember kernel
freezes instead of module crashes some time ago.
When starting with the mesh experiments I did not have these troubles,
but then again,@this time there wasn't even HT support...

I am happy for receiving suggestions on how to make ath9k more stable,
or fix the issue some other way, if possible.
If there is more output, details or answers I can provide regarding my
setup, please ask.