Re: ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x02000020 DMADBG_7=0x00008040
From: Justin P. Mattock <hidden>
Date: 2012-03-20 05:54:13
Also in:
lkml, netdev
On 03/19/2012 08:33 AM, Felix Fietkau wrote:
On 2012-03-19 4:25 PM, Justin P. Mattock wrote:quoted
On 03/19/2012 08:20 AM, Mohammed Shafi wrote:quoted
Hi Justin,quoted
seems I got this when waking up from suspend, was trying to connect with NM then all of a sudden Xorg shut off and this appeared on the screen(image links below I could not save dmesg only pic) http://flic.kr/p/bEhf2K http://flic.kr/p/bEhf2x http://flic.kr/p/bEhf2n http://flic.kr/p/bEhf26 I see ath9k in the image, but am unsure if its related. kernel is: 3.3.0-rc7-next-20120316this looks like some new issue in ath9k, lets try to recreate it herequoted
Justin P. Mattockwould be nice to recreate, then bisecting will be more accurate to do.. I am trying userspace tweaks here and there but no luck(hostapd etc..) keep in mind this is for the RX DMA, the above just occured.. as for writing a small app I still need to try, but my C skills only go so far(anything to trigger!).Was your kernel compiled with debug info? If so, please run gdb on the ath9k.ko module and enter this: l *ath9k_tx+start+0x284 Then send me the output of that. - Felix
I changed the Cc''s so hopefully its correct..
anyway I tried the above but might have gotten it wrong..:
Reading symbols from
/home/kernel/linux-next/drivers/net/wireless/ath/ath9k/ath9k.o...done.
(gdb) list *ath9k_tx+0xa3
0x4673 is in ath9k_tx (drivers/net/wireless/ath/ath9k/main.c:1136).
1131 memset(&txctl, 0, sizeof(struct ath_tx_control));
1132 txctl.txq = sc->tx.txq_map[skb_get_queue_mapping(skb)];
1133
1134 ath_dbg(common, XMIT, "transmitting packet, skb: %p\n", skb);
1135
1136 if (ath_tx_start(hw, skb, &txctl) != 0) {
1137 ath_dbg(common, XMIT, "TX failed\n");
1138 goto exit;
1139 }
1140
(gdb) list *ath9k_tx+start+0x284
No symbol "start" in current context.
before receiving the email to do the above code with gdb I had cleaned
the tree out with git clean -fx so rebuilding might have misalighned
things(but could be wrong)..