ath9k: Still seeing DMA issues in latest wireless-testing.

ath9k: Still seeing DMA issues in latest wireless-testing.

[Ben Greear]

Looks like some potentially nice ath9k patches just got merged into
wireless-testing.  It has been stable and fast so far, but I
do still see DMA issues in the logs.

This is on a system with 30 STAs associated with a flakey netgear running
HT40-, and 30 associated with a cisco running HT20.  (Through some luck,
the netgear managed to send proper ht-info channel this time,
so the NIC went HT40-).

ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
sta0: authenticate with 30:46:9a:10:0b:9a (try 1)
sta0: authenticated
sta0: associate with 30:46:9a:10:0b:9a (try 1)
sta0: RX ReassocResp from 30:46:9a:10:0b:9a (capab=0x421 status=0 aid=31)
sta0: associated
ieee80211 wiphy0: Allocated STA 30:46:9a:10:0b:9a
ieee80211 wiphy0: Inserted STA 30:46:9a:10:0b:9a
ieee80211 wiphy0: WMM queue=2 aci=0 acm=0 aifs=3 cWmin=15 cWmax=1023 txop=0 uapsd=0
ieee80211 wiphy0: WMM queue=3 aci=1 acm=0 aifs=7 cWmin=15 cWmax=1023 txop=0 uapsd=0
ieee80211 wiphy0: WMM queue=1 aci=2 acm=0 aifs=2 cWmin=7 cWmax=15 txop=94 uapsd=0
ieee80211 wiphy0: WMM queue=0 aci=3 acm=0 aifs=2 cWmin=3 cWmax=7 txop=47 uapsd=0
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ADDRCONF(NETDEV_CHANGE): sta0: link becomes ready
sta0: no IPv6 routers present
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020

Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com

Re: ath9k: Still seeing DMA issues in latest wireless-testing.

[Björn Smedman]

On Sat, Jan 29, 2011 at 5:43 AM, Ben Greear <greearb@candelatech.com> wrote:
> Looks like some potentially nice ath9k patches just got merged into
> wireless-testing.  It has been stable and fast so far, but I
> do still see DMA issues in the logs.
>
> This is on a system with 30 STAs associated with a flakey netgear running
> HT40-, and 30 associated with a cisco running HT20.  (Through some luck,
> the netgear managed to send proper ht-info channel this time,
> so the NIC went HT40-).

Ben, the fact that you see much more of these DMA issues when you're
running your virtual STAs suggests to me that root cause is some race
condition between dma and cpu. The main difference between the 60 STA
use-case and the one STA use case is that rx irq processing will take
longer with multiple STAs, right? Is there some other difference?

/Björn

Re: ath9k: Still seeing DMA issues in latest wireless-testing.

[Ben Greear]

On 01/30/2011 07:54 AM, Björn Smedman wrote:
> On Sat, Jan 29, 2011 at 5:43 AM, Ben Greear<greearb@candelatech.com>  wrote:
>> Looks like some potentially nice ath9k patches just got merged into
>> wireless-testing.  It has been stable and fast so far, but I
>> do still see DMA issues in the logs.
>>
>> This is on a system with 30 STAs associated with a flakey netgear running
>> HT40-, and 30 associated with a cisco running HT20.  (Through some luck,
>> the netgear managed to send proper ht-info channel this time,
>> so the NIC went HT40-).
>
> Ben, the fact that you see much more of these DMA issues when you're
> running your virtual STAs suggests to me that root cause is some race
> condition between dma and cpu. The main difference between the 60 STA
> use-case and the one STA use case is that rx irq processing will take
> longer with multiple STAs, right? Is there some other difference?

I don't know.  I see it even when not running much traffic.

I also see DMA on my AP system, which has a single AP and one idle
station.  Granted, I have 128 or so stations connected to the AP,
but again, the stations are mostly idle.

Other folks see the DMA issues too...and most probably just don't report
them anymore...

Thanks,
Ben

>
> /Björn


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com