Re: [PATCH 1/1] ath5k: fix hw rate index condition

[Stefan Lippers-Hollmann <s.L-H@gmx.de>]

Hi

On Sonntag, 15. März 2009, Stefan Lippers-Hollmann wrote:
> Hi
> 
> On Mittwoch, 7. Januar 2009, Jiri Slaby wrote:
> > On 01/07/2009 02:51 PM, Jiri Slaby wrote:
> > > Dhaval Giani wrote:
> > >> I see this on current git. Not sure how to reproduce it, has happened on
> > >> two random occasions. At both times, I was not connected to a wireless
> > >> network, but to wired networks.
> > >>
> > >> ------------[ cut here ]------------
> > >> WARNING: at net/mac80211/rx.c:2234 __ieee80211_rx+0x7f/0x559
> > >> ...
> > >> Call Trace:
> > >>  [<f80d4192>] __ieee80211_rx+0x7f/0x559 [mac80211]
> > >>  [<f80a19f4>] ath5k_tasklet_rx+0x4f7/0x53b [ath5k]
> > >> ...
> > > 
> > > Hmm, maybe ath5k is culprit. Could you apply the attached patch and
> > > use the kernel till the problem appears again?
> 
> It seems as if this problem wouldn't be restricted to ath5k, I just 
> triggered something very similar on b43 and 2.6.29-rc8-git1 (i386, hard 
> preemption):
> 
> b43-phy0: Broadcom 4306 WLAN found (core revision 5)
[...]
> wlan1: no IPv6 routers present
> b43-phy0 ERROR: PHY transmission error
> b43-phy0 ERROR: PHY transmission error
> 
> [ lots of these, likely to be caused by minstrel being a little too 
>   optimistic about the possible wlan rates (it was more conservative in 
>   2.6.28 and didn't happen there); the distance between both stations is 
>   on the upper end ]
> 
> b43-phy0 ERROR: PHY transmission error
> __ratelimit: 9 callbacks suppressed
> b43-phy0 ERROR: PHY transmission error
> b43-phy0 ERROR: PHY transmission error
> ------------[ cut here ]------------
> WARNING: at net/mac80211/rx.c:2234 __ieee80211_rx+0xa2/0x6a0 [mac80211]()
> Hardware name: Amilo D-Series
> Modules linked in: ppdev lp aes_i586 aes_generic ipv6 af_packet rfkill_input arc4 ecb b43 rfkill rng_core mac80211 cfg80211 led_class input_polldev ssb joydev pcmcia snd_via82xx gameport snd_ac97_codec ac97_bus snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc snd_mpu401_uart snd_rawmidi snd_seq_device i2c_viapro serio_raw snd i2c_core pcspkr psmouse evdev soundcore via686a via_agp shpchp yenta_socket rsrc_nonstatic pcmcia_core pci_hotplug rtc_cmos battery rtc_core rtc_lib parport_pc parport ac button ext3 jbd mbcache sg sr_mod cdrom sd_mod ata_generic pata_acpi pata_via uhci_hcd ehci_hcd floppy firewire_ohci libata tulip firewire_core crc_itu_t usbcore scsi_mod thermal processor fan
> Pid: 0, comm: swapper Not tainted 2.6.29-rc8-sidux-686 #1
> Call Trace:
>  [<c01319d7>] warn_slowpath+0x87/0xe0
>  [<d00523b7>] op32_set_current_rxslot+0x27/0x40 [b43]
>  [<d0052d93>] b43_dma_rx+0x193/0x420 [b43]
>  [<c0124fc3>] __wake_up_common+0x43/0x70
>  [<cfffcc62>] __ieee80211_rx+0xa2/0x6a0 [mac80211]
>  [<c011e9a5>] default_spin_lock_flags+0x5/0x10
>  [<c03a3f2e>] _spin_lock_irqsave+0x3e/0x60
>  [<cffeb337>] ieee80211_tasklet_handler+0x107/0x130 [mac80211]
>  [<c013692c>] tasklet_action+0x6c/0xf0
>  [<c0137147>] __do_softirq+0x87/0x140
>  [<c011e9a5>] default_spin_lock_flags+0x5/0x10
>  [<c03a3f2e>] _spin_lock_irqsave+0x3e/0x60
>  [<c0137255>] do_softirq+0x55/0x60
>  [<c0137495>] irq_exit+0x75/0x90
>  [<c0106378>] do_IRQ+0x48/0x90
>  [<c0104527>] common_interrupt+0x27/0x2c
>  [<cf8372e4>] acpi_idle_enter_simple+0x17a/0x1f4 [processor]
>  [<c02fd3bf>] cpuidle_idle_call+0x6f/0xc0
>  [<c0102de6>] cpu_idle+0x66/0xa0
> ---[ end trace c754f566bbe5ac47 ]---
[...]
> Sometimes even the firmware crashes and gets reloaded continously.
[...]
> Setting a fixed wlan rate (like 11M) seems to avoid this problem.
> 
> > I don't think this will print anything, the rate won't be 32, it's rather
> > too high. Could you apply also the appended debug one?
> 
> I will apply this patch and give it some more testing tomorrow evening, 
> this problem is almost 100% reproducable for me at the end of my router's
> range and doesn't happen in closer proximity.
[...]

Sorry for the late response, but I've been unexpectedly away from my 
BCM4306 system until today.

Thanks to the following (not yet mainline) patches by Michael Buesch and 
Lorenzo Nava on top of 2.6.29-rc8-git5, these problems seem to be "fixed" 
(well, the PHY errors are basically just hidden, but as they don't 
trigger the firmware watchdog anymore, it's much less of a problem and 
isn't actually a user visible problem anymore).

[PATCH] b43: Mask PHY TX error interrupt, if not debugging
	http://marc.info/?l=linux-wireless&m=123748731831778&w=2

[PATCH] b43: fix b43_plcp_get_bitrate_idx_ofdm return type
	http://marc.info/?l=linux-wireless&m=123774585529189&w=2


Confirming the patch descriptions, Jiri Slaby's debugging patch did reveal 
a signedness problem of the return value of in 
b43_plcp_get_bitrate_idx_ofdm(), which has been fixed by the patch above:

[ this trace happened *without* "b43: fix b43_plcp_get_bitrate_idx_ofdm 
  return type", and only "b43: Mask PHY TX error interrupt, if not 
  debugging" applied on top of 2.6.29-rc8-git5 ]
------------[ cut here ]------------
WARNING: at net/mac80211/rx.c:2236 __ieee80211_rx+0xab/0x6b0 [mac80211]()
Hardware name: Amilo D-Series
RATE=255, BAND=c
Modules linked in: ppdev lp aes_i586 aes_generic ipv6 af_packet rfkill_input arc4 ecb b43 rfkill rng_core mac80211 cfg80211 led_class input_polldev ssb joydev snd_via82xx gameport snd_ac97_codec ac97_bus snd_pcm_oss snd_mixer_oss pcmcia snd_pcm snd_timer snd_page_alloc snd_mpu401_uart snd_rawmidi i2c_viapro serio_raw snd_seq_device pcspkr i2c_core psmouse snd evdev soundcore via686a shpchp yenta_socket rsrc_nonstatic pcmcia_core via_agp pci_hotplug rtc_cmos parport_pc battery rtc_core rtc_lib parport ac button ext3 jbd mbcache sg sr_mod cdrom sd_mod ata_generic pata_acpi pata_via uhci_hcd ehci_hcd floppy firewire_ohci libata tulip firewire_core crc_itu_t usbcore scsi_mod thermal processor fan
Pid: 0, comm: swapper Not tainted 2.6.29-rc8-sidux-686 #1
Call Trace:
 [<c0131a67>] warn_slowpath+0x87/0xe0
 [<d002d377>] op32_set_current_rxslot+0x27/0x40 [b43]
 [<d002dd53>] b43_dma_rx+0x193/0x420 [b43]
 [<c01ae229>] add_partial+0x19/0x70
 [<cfcd834f>] ieee80211_tasklet_handler+0x11f/0x130 [mac80211]
 [<c03a4195>] _spin_unlock+0x5/0x20
 [<cfce9c6b>] __ieee80211_rx+0xab/0x6b0 [mac80211]
 [<c011ea35>] default_spin_lock_flags+0x5/0x10
 [<c03a3d7e>] _spin_lock_irqsave+0x3e/0x60
 [<cfcd8337>] ieee80211_tasklet_handler+0x107/0x130 [mac80211]
 [<c01369bc>] tasklet_action+0x6c/0xf0
 [<c01371d7>] __do_softirq+0x87/0x140
 [<c011ea35>] default_spin_lock_flags+0x5/0x10
 [<c03a3d7e>] _spin_lock_irqsave+0x3e/0x60
 [<c01372e5>] do_softirq+0x55/0x60
 [<c0137525>] irq_exit+0x75/0x90
 [<c0106378>] do_IRQ+0x48/0x90
 [<c0104527>] common_interrupt+0x27/0x2c
 [<cf8372cb>] acpi_idle_enter_simple+0x17a/0x1f4 [processor]
 [<c02fcfcf>] cpuidle_idle_call+0x6f/0xc0
 [<c0102de6>] cpu_idle+0x66/0xa0
---[ end trace ba8601a4d52a20d2 ]---
------------[ cut here ]------------		

So far (after 2.9 GB continuous kernel tarball downloads from a local 
mirror) b43 seems to be fine again:

wlan1     IEEE 802.11bg  ESSID:"gemini"
          Mode:Managed  Frequency:2.412 GHz  Access Point: 00:21:27:FF:51:A8
          Bit Rate=54 Mb/s   Tx-Power=20 dBm
          Retry min limit:7   RTS thr:off   Fragment thr=2352 B
          Power Management:off
          Link Quality=54/100  Signal level:-82 dBm  Noise level=-69 dBm
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  Invalid misc:0   Missed beacon:0

wlan1     Link encap:Ethernet  HWaddr 00:0f:66:d8:67:ca
          inet addr:192.168.0.70  Bcast:192.168.0.255  Mask:255.255.255.0
          inet6 addr: fe80::20f:66ff:fed8:67ca/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:2090104 errors:0 dropped:0 overruns:0 frame:0
          TX packets:1082081 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:3146865411 (2.9 GiB)  TX bytes:93054386 (88.7 MiB)

Fetched 83.2MB in 1min18s (1058kB/s)
[...]
Fetched 83.2MB in 1min1s (1362kB/s)

Thank you and sorry about the late response.

Regards
	Stefan Lippers-Hollmann


Post scriptum: I'm not able to trigger this trace with ath5k/ AR2425.
-- 
> >  net/mac80211/rx.c |    6 ++++--
> >  1 files changed, 4 insertions(+), 2 deletions(-)
> > 
> > diff --git a/net/mac80211/rx.c b/net/mac80211/rx.c
> > index 7175ae8..5e17e57 100644
> > --- a/net/mac80211/rx.c
> > +++ b/net/mac80211/rx.c
> > @@ -2230,8 +2230,10 @@ void __ieee80211_rx(struct ieee80211_hw *hw, struct sk_buff *skb,
> >  		 * MCS aware. */
> >  		rate = &sband->bitrates[sband->n_bitrates - 1];
> >  	} else {
> > -		if (WARN_ON(status->rate_idx < 0 ||
> > -			    status->rate_idx >= sband->n_bitrates))
> > +		if (WARN(status->rate_idx < 0 ||
> > +			    status->rate_idx >= sband->n_bitrates,
> > +			    "RATE=%u, BAND=%x\n", status->rate_idx,
> > +			    sband->n_bitrates))
> >  			return;
> >  		rate = &sband->bitrates[status->rate_idx];
> >  	}