Toshiba ToPIC developer info

[Daniel Ritz]

On Sunday 26 June 2005 21.42, Ryan Underwood wrote:
> 
> On Sun, Jun 26, 2005 at 05:43:31PM +0200, Daniel Ritz wrote:
> > 
> > > and the system seems to be swamped with interrupts (or something
> > > polling/busy-waiting?) so it is almost unresponsive, until cardmgr is
> > > started again and the 16-bit card's driver(s) are loaded once cardmgr
> > > sees it.
> > > 
> > 
> > may be a problem with the assigned resources in config.opts. mine attached
> 
> I've ensured that there are no port/irq conflicts.  I have no idea what
> I would have to change in the memory areas.  Also, it's not an interrupt
> storm because I while true; do cat /proc/interrupts; done when removing
> and reinserting a card, and the only interrupts showing up besides the
> timer and ide0 are about 20-30 interrupts on yenta_socket's IRQ.  So I
> guess something is busy waiting somewhere.  I attached a call trace at
> the moment where the system appears hung in case this helps.
> 
> > also i'm not using 2.6.12 but the git tree from yesterday. this has
> > http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.12/2.6.12-mm1/broken-out/yenta-ti-turn-off-interrupts-during-card-power-on-more-2.patch
> > already applied. one of the effects is that it helps if the socket is
> > in a bad state after boot. please test with that one and see if those
> > 	[busy], [bat dead], [bat low], [req attn]
> > go away.
> 
> Nope, still here.  Also, the nobody cared IRQ is still here.  But the
> patch didn't break anything either.

now that i read the whole thread...you say in 2.4+pcmcia-cs you can
get the same flags if you comment out the topic_set_opts, topic_set_state,
topic_get_state in i82365.c. now can you figure out what exactly makes the
difference there?

> 
> > for the interrupt storm your seeing: i cannot reproduce them. i tried
> > both cards using ISA and using the PCI interrupt. stable here.
> 
> No interrupt storm, I checked - just a very slow system until it "clears up".
> 

yes interrupt storm. otherwise the irq wouldn't get disabled.

from you dmesg:
> rt2500: no version for "struct_module" found: kernel tainted.
> rt2500 1.1.0 BETA2 2005/02/21 http://rt2x00.serialmonkey.com
nice. an out of tree driver...which is even loaded before pcmcia.
not nice.

now before we look into the irq disabling: never plug your CB
wireless card and make very sure that crappy rt2500 driver never
loads! then plug your pcmcia network card and then we'll see.

rgds
-daniel