detected fid change

[James Harper]

More information...

/proc/interrupts looks like this:
           CPU0       CPU1
  0:    3312255         18    IO-APIC-edge  timer
  1:          9          1    IO-APIC-edge  i8042
  2:          0          0          XT-PIC  cascade
 14:       8668          1    IO-APIC-edge  ide0
 15:          7          3    IO-APIC-edge  ide1
 16:     227717          0   IO-APIC-level  eth0, nvidia
 17:       7384          0   IO-APIC-level  bttv0
 18:       7743          1   IO-APIC-level  ide2
 19:     325787          0   IO-APIC-level  uhci_hcd, wifi0, Ensoniq
AudioPCI
NMI:          0          0
LOC:    3312501    3312864
ERR:          0
MIS:          0

My isdn modem is on usb (uhci_hcd). If I'm only using wifi0 it's fine.
If I'm using the isdn modem too (eg flood pings on isdn interface and
wifi0) then more fid change errors occur. Using the sound card at the
same time means bad bad things happen.

Here's a dmesg output of a more severe crash:
br0: topology change detected, propgating
br0: port 2(wlan0) entering forwarding state
br0: topology change detected, propgating
br0: port 1(eth0) entering forwarding state
int
int
wifi0: detected fid change (try=0, reg=0044): 0000 017b 017b
int
int
int
int
int
int
int
int
int
int
NETDEV WATCHDOG: wifi0: transmit timed out
wifi0 Tx timed out! Resetting card
wifi0: CMD=010b EVSTAT=0000 OFFSET0=0000 OFFSET1=0000 SWSUPPORT0=8a32
wifi0: scheduled card reset
hostap_pci: wifi0: resetting card
wifi0: Original COR value: 0x32
wifi0: removed pending cmd_queue entry (type=1, cmd=0x010b,
param0=0x0127)
prism2_hw_init: initialized in 202 ms
wifi0: trying to read PDA from 0x007f0000: OK
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 55
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 20
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 10
wifi0: 00:50:8b:46:57:d5 auth_cb - alg=0 trans#=2 status=0 - STA
authenticated
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 55
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 20
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 10
wifi0: 00:50:8b:46:57:d5 auth_cb - alg=0 trans#=2 status=0 - frame was
not ACKed
wifi0: 00:50:8b:46:57:d5 assoc_cb - frame was not ACKed
wifi0: 00:50:8b:46:57:d5 auth_cb - alg=0 trans#=2 status=0 - STA
authenticated
wifi0: 00:50:8b:46:57:d5 assoc_cb - STA associated
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 55
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 20
wifi0: STA 00:50:8b:46:57:d5 TX rate lowered to 10
int
int
int
int
int
int
int
int
int
int
wifi0: 00:50:8b:46:57:d5 auth_cb - alg=0 trans#=2 status=0 - STA
authenticated
wifi0: 00:50:8b:46:57:d5 assoc_cb - STA associated
wifi0: Could not find STA 00:50:00:50:8b:46 for this TX error (@1300009)

the 'int' messages are a debug I put in... they reflect when
prism2_interrupt has been called whilst prism2_read_fid_reg is in
progress. Is this a bad thing? prism2_interrupt will do at least
HFA384X_INW(HFA384X_EVSTAT_OFF);
to find out if there is nothing to do (eg because it's actually a usb or
audio interrupt).

James

> -----Original Message-----
> From: hostap-bounces+james.harper=bigpond.com at shmoo.com
[mailto:hostap-
> bounces+james.harper=bigpond.com at shmoo.com] On Behalf Of James Harper
> Sent: Wednesday, 31 December 2003 21:01
> To: hostap at shmoo.com
> Subject: detected fid change
> 
> I'm still getting failures on my smp system under any sort of load. Eg
> flood pings with large packet sizes.
> 
> The sort of message I'm getting is:
> Dec 31 20:49:16 portal kernel: wifi0: detected fid change (try=0,
> reg=0040): 0000 0396 0396
> Dec 31 20:49:40 portal kernel: wifi0: detected fid change (try=0,
> reg=0040): 02b4 5d5c 02b4
> Dec 31 20:49:40 portal kernel: wifi0: detected fid change (try=0,
> reg=0040): 00e4 00e4 5400
> Dec 31 20:50:09 portal kernel: wifi0: detected fid change (try=0,
> reg=0040): 0000 0152 0152
> 
> Running 2.6.0 and CVS that was released the same day as 2.6.0.
> 
> I never got the chance to try Denis's patch to check race conditions,
is
> it worth trying?
> 
> James
> 
> _______________________________________________
> HostAP mailing list
> HostAP at shmoo.com
> http://lists.shmoo.com/mailman/listinfo/hostap