From: Jouni Malinen (jkmaline_at_cc.hut.fi)
Date: 2002-09-03 03:26:58 UTC
On Tue, Sep 03, 2002 at 09:27:24AM +0800, Lei chuanhua wrote:
> 1. Wireless to wired PCs, using iperf or other tcp testing tools, under heavy traffic. After a few hours(sometimes 30min, sometimes 12 hours), HOSTAP stopped working while other parts worked properly in the system. I checked it carefully, and found that no interrupts coming in. However, after the card stops working, if I call hfa384x_reset_port(disable, and enable card), interrupt comes again. And it can work properly now.
I haven't seen exactly this behavior, but if reset_port is enough to fix it, there might be something odd happening with the card firmware state.
> At first , I suspect hostap driver has a problem. But I use the agere/orinoco card wiht firmware-based AP to test the same case in the same hardware platform, and it also failed. Once again, I suspect my hardware has a problem. The key problem is, after the card stopped working, how the card can work again if I disable first, then enable wireless card.
> Even if I can solve the problem by using hfa384x_reset_port, where should I call it? I've tried to spawn one task to call it every 30 mininutes, but I failed.
I wouldn't call it a solution to the problem since the problem would be the thing causing this.. This would be more a workaround for the symptom.. Anyway, I added detection for IRQ delivery problems that is run during card initialization. This same method could be left running (i.e., producing small amount of tick events, like 1 per second) and then have the driver check the tick event counter every 10 seconds or so. If no tick events have been received, the port would be resetted.
-- Jouni Malinen PGP id EFC895FA