* network tx_timeout guidance
@ 2006-07-19 20:53 Keith Chew
2006-07-19 21:12 ` David Miller
0 siblings, 1 reply; 2+ messages in thread
From: Keith Chew @ 2006-07-19 20:53 UTC (permalink / raw)
To: linux-kernel
Hi
We are stress testing a wireless network driver from a vendor (ZD1211
chip, we realise that there is a rewrite project happenning but that
it still not ready for production yet).
So far, everything is going very well except after several days when a
Tx Timeout occurs. In this scenario, the driver goes into a tight
loop, and we keep getting the Tx Timeout. I have already contacted the
vendor, but just wanted additional opinion/feedback from the experts
at LKML.
In the code, this is the callback from the assignment:
============================
#define ZD1211_TX_TIMEOUT (HZ*10)
...
dev->watchdog_timeo = ZD1211_TX_TIMEOUT;
dev->tx_timeout = &zd1211_tx_timeout;
============================
A couple of questions:
(1) How should the zd1211_tx_timeout handle a Tx Timeout? At present
it is calling the usb_kill_urb via the kevent. Is this what it is
suppose to do?
============================
static void zd1211_tx_timeout(struct net_device *dev)
struct zd1205_private *macp = dev->priv;
// Which will call zd1211_kill_tx_urb() below via kevent;
defer_kevent(macp, KEVENT_KILL_TX_URB);
}
static void zd1211_kill_tx_urb(void *_macp) {
struct zd1205_private *macp = (struct zd1205_private *)_macp;
// conditional defines are ommited to keep things clear
usb_kill_urb(macp->tx_urb);
}
============================
(2) What is the best way to simulate a Tx Timeout? Currently we have
to wait for several days (for 1 of the 10 PCs under stress testing)
for it to occur.
Regards
Keith
^ permalink raw reply [flat|nested] 2+ messages in thread* Re: network tx_timeout guidance
2006-07-19 20:53 network tx_timeout guidance Keith Chew
@ 2006-07-19 21:12 ` David Miller
0 siblings, 0 replies; 2+ messages in thread
From: David Miller @ 2006-07-19 21:12 UTC (permalink / raw)
To: keith.chew; +Cc: linux-kernel
From: "Keith Chew" <keith.chew@gmail.com>
Date: Thu, 20 Jul 2006 08:53:28 +1200
> (2) What is the best way to simulate a Tx Timeout? Currently we have
> to wait for several days (for 1 of the 10 PCs under stress testing)
> for it to occur.
Make your netdev->hard_start_xmit() method simply record the
SKB, but don't actually give it to the hardware.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2006-07-19 21:12 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2006-07-19 20:53 network tx_timeout guidance Keith Chew
2006-07-19 21:12 ` David Miller
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®