On Sat, 7 Feb 2009, Stuart Henderson wrote:
In gmane.os.openbsd.misc, you wrote:
In a net5501 I have a rt2860 ral card, running the Feb 04 snapshot:
ral0 at pci0 dev 14 function 0 "Ralink RT2860" rev 0x00: irq 10
ral0: MAC/BBP RT2860 (rev 0x0101), RF RT2820 (MIMO 2T3R)

I've been trying snapshots off and on since damien@ started tinkering with
the rt2860 code two months ago. With any snapshot from the last 2 months, I
can't get the box to stay up for more than 2 hours (or less) without it
rebooting. If I turn off the watchdog timer, it will just hang without
printing any messages. If I "ifconfig ral0 down", the box is rock stable.

enable ddb.console=1 and send it a BREAK, see if you can get some
trace out of ddb.

Thanks for the suggestion. I tried it, but the kernel's not responding to the break :(

leave some sessions open, run things like "top -s.1", "systat vmstat .1"
and see what the system's doing when it freezes.

Right, the top is not showing anything out of the ordinary, the vmstat shows 7.1% interrupt load and nothing else on the processor at that time:

   7.1%Int   0.0%Sys   0.0%Usr   0.0%Nic  92.9%Idle

Interrupts
732 total
229 vr0
192 vr3
 82 ral0
    pciide0
    ohci0
    com0
101 clock
128 rtc

Proc:r  d  s  w    Csw   Trp   Sys   Int   Sof  Flt
           7        55     9   841   731   110  357

All seems fairly standard to me, some light load on lan/wlan.

send dmesg :-)

I'd rather not spam the list, it's just an ordinary net5501, dmesg is easily googled.

Is anyone else seeing this with -current or a snapshot, with this ral or a
different one? I'd file a problem report but there's nothing to go on,
other than my suspicions that the changes to rt2860 in the last 2 months
are the cause.
I can try to narrow it down to a specific commit if that will help?

I've been running recent snaps on an ALIX board with RT2860 with
no trouble.

That's.. unfortunate. I keep thinking that since some people don't even see the problems with traffic stalling in PR 5958, there might be something specific to the location of the AP, like load or some specific client that makes it go boom. Grasping at straws, here.

Thanks for the suggestion,

bbee

Reply via email to