Alex Williamson wrote:
Yeah, it kinda seems like there is.  With this change, the timer expires
and we go through this path:

pm_tmr_timer()
        -> pm_update_sci()
                -> get_pmsts()
                -> qemu_set_irq() [but not for a TMFOF_EN]
                -> qemu_mod_timer()
                bump tmr_overlfow_time

We bumped tmr_overflow_time in pm_update_sci after setting the timer to
expire on the old value.  Unless something goes horribly wrong with
timers, we'll always get the timer event before overflow time and
get_pmsts never adds in the TMROF_EN bit to the status flag.  We
therefore never toggle the SCI interrupt because of a timer overflow,
and we never report a timer overflow status to the guest.

The author of this patch is correct that the timer in the original code
only goes off a couple times before we del_timer().  However, I think
the way it's supposed to work is that we set the timer overflow status,
toggle the SCI, then wait for the OSPM to come in through
pm_ioport_writew() to clear the timer overflow status, at which point we
call pm_update_sci() mod_timer and start it all over again.  At least
that's the way I see it working after removing this change.

It doesn't make much sense to bump tmr_overflow_time so that we never
hit it, unless I'm completely misunderstanding the code.  Thanks,

You're right; a quick run with that patch reverted (and not) showed it clearly.

I guess it was merged before we had sci working properly; otherwise I can't explain how it worked then.

I reverted that patch.

--
Do not meddle in the internals of kernels, for they are subtle and quick to 
panic.

--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to