tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: NetBSD-5 ichlpcib watchdog timer seems to be broken for some revisions of the ichlpcib chip set.

        Hello.  I'd like to correct condition number 2 listed in my previous
I think the problem is not that the tickling function isn't working, but
that the timer is triggering early.  So, it's possible we're not correctly
setting the timeout value in the hardware when we set the mode.  That makes
sense to me, if I think about it, since the code path for those chips that
work versus those that don't is the same from what I can tell.  Also, if I
manually count to 30, while waiting for the timer to expire, the timer
kicks off between 10 and 15 seconds after I start counting.
So, I'll  probably look at the setmode function and see if I can notice any
obvious changes between NetBSD-4 and NetBSD-5.
On Jun 29, 11:27pm, Brian Buhrow wrote:
} Subject: Re: NetBSD-5 ichlpcib watchdog timer seems to be broken for some 
}       Hello.  Of the 4 examples I have working at the moment, I have two
} conditions:
} 1.  wdogctl works fine.  It sets the timer, and the tickling function
} works, either from user or kernel mode and, the machine only reboots if the
} tickling function fails to be called.
} 2.  The tickling function appears not to work. The timer is set and reboots
} after the appropriate amount of time of not being tickled. 
}       In all the cases I've reported where I see behavior number 2, I see
} behavior number 1, i.e. everything works fine, under NetBSD-4.  This seems
} to be a NetBSD-5 phenomenon.
}       The chip set which does work properly under NetBSD-5 reports as:
} ichlpcib0 at pci0 dev 31 function 0
} ichlpcib0: Intel 82801DB LPC Interface Bridge (rev. 0x02)
} Hope this helps.
} -Brian
} On Jun 29,  2:14pm, Joerg Sonnenberger wrote:
} } Subject: Re: NetBSD-5 ichlpcib watchdog timer seems to be broken for some 
} } On Mon, Jun 29, 2009 at 12:57:53AM -0700, Brian Buhrow wrote:
} } >   Hello.  I think so.  But if you mean that the timer is triggering
} } > early, before it's tickled, that's possible.  I guess I could try testing
} } > that theory by having the wdogctl program tickle  quicker than half the
} } > time left on the timer.
} } 
} } I more suspect that the tickling itself doesn't work, so it is just
} } always timing out. I'd like to know what of the different failure cases
} } we have first :)
} } 
} } Joerg
} >-- End of excerpt from Joerg Sonnenberger
>-- End of excerpt from Brian Buhrow

Home | Main Index | Thread Index | Old Index