Re: more problems with em(4)

  Hi, Michal!

On Tue, Mar 07, 2006 at 12:29:34PM +0100, Michal Mertl wrote:
M> Today I've updated one machine acting as a router to RELENG_6 and I've
M> also got several watchdog timeouts on the machine. One em card (em1)
M> shares the IRQ with USB (as on the notebook) whereas the other doesn't.

All new interrupt handling done by Scott is solely in HEAD, nothing
is merged to RELENG_6.

M> The machine didn't stop forwarding packets so it is probably a different
M> problem that the one with wedged card, possibly completely harmless. It
M> will be pretty bad if we shipped 6.1 with broken em(4) driver though.
M> The first watchdog message appeared after almost 2 hours of operation
M> (so it wasn't some stabilizing on boot or something). It has never
M> appeared before in more than a year with 5.x (the router is only lighly
M> loader - it pushes 1Mbps on average).
M> 
M> The em card in the notebook identifies itself in 'pciconf -lv' output as
M> "82540EP Gigabit Ethernet Controller (Mobile)" whereas on the router the
M> card is "82541EI Gigabit Ethernet Controller (Copper)".
M> 
M> Mar  6 21:56:26 sx-a kernel: em1: watchdog timeout -- resetting
M> 
M> I've changed the media settings of em0 from 100-full to auto here (em1
M> has been set to auto and runs 1000-full all the time).
M> Mar  7 01:35:45 sx-a kernel: em0: Link is up 100 Mbps Full Duplex
M> 
M> Mar  7 01:43:16 sx-a kernel: em0: watchdog timeout -- resetting
M> Mar  7 05:33:57 sx-a kernel: em1: watchdog timeout -- resetting
M> 
M> Verbose dmesg available on request.

Before the latest change these line was just suppressed due to typo. So
the card actually got timeout and was reseted, but nothing was printed.
Please see the latest change in RELENG_6 and comments to it.

-- 
Totus tuus, Glebius.
GLEBIUS-RIPN GLEB-RIPE