Date: Fri, 19 Nov 1999 10:40:19 -0700 From: Warner Losh <imp@village.org> To: Josef Karthauser <joe@pavilion.net> Cc: Dan Moschuk <dan@FreeBSD.ORG>, hackers@FreeBSD.ORG Subject: Re: VAIO F270, ep0 and -current Message-ID: <199911191740.KAA33661@harmony.village.org> In-Reply-To: Your message of "Fri, 19 Nov 1999 17:20:25 GMT." <19991119172025.A97321@florence.pavilion.net> References: <19991119172025.A97321@florence.pavilion.net> <19991118192338.D55314@florence.pavilion.net> <19991117143759.A3011@spirit.jaded.net> <19991118192338.D55314@florence.pavilion.net> <199911182135.OAA28043@harmony.village.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In message <19991119172025.A97321@florence.pavilion.net> Josef Karthauser writes:
: If you can give me some clues I'll work on it. I've got zero familiarity
: with the code :(.
Sure. I posted this once to -mobile, so you might want to look in the
archives there as well as here.
The basic problem is that it appears that detach may be called AFTER
the power is removed from the card generating an interrupt causing the
interrupt routine to be called when the card is no longer there, which
leads to an instant hang. Either that, or the act of removing the
card causes an interupt to the card, then on the pcic which pulls the
rug out from under the card and then it just hangs. I'm not
completely sure and don't have a machine to test this on
extenstively right now, so here goes at an off the cuff analysis.
Looking at pcicintr in sys/pccard/pcic.c we see that the card_inserted
event gets called (pccard_event in sys/pccard/pccard.c). This calls
disable_slot and sets the state of the slot to empty and wakes up the
pccardd proces. Looking at code it appears that we go ahead and
delete the child in disable_slot (which should really be delayed until
spl has been lowered via a timeout) which should be right. We then
wait 1/2 second and power off the slot. Looks like we call the
controllers disable routine for the slot (which is pcic_disable in
pcic.c) where we turn off the card at the bridge chip.
It is when we power the card off that we hang in the cards interrupt
routine. Given this sequence of events, we should be safe. For the
ep card we if_down, teardown the interrupt and free the resources and
print unload.
Somewhere in the above, something goes wrong :-(. Maybe it is the
first parenthetical statement about waiting for the spl to be
lowered, which would be something like:
/*
* disable_slot - Disables the slot by removing
* the power and unmapping the I/O
*/
static void real_disable_slot(struct slot *slt);
static void
disable_slot(struct slot *slt)
{
slt->poff_ch = timeout(real_disable_slot, (caddr_t) slot, 0);
}
static void
real_disable_slot(struct slot *slt)
{
// guts of the old disable_slot goes here
}
but there may be some more complicated book keeping needed...
Lemme know if this helps...
Warner
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199911191740.KAA33661>