openvortex-dev
[Top][All Lists]
Advanced

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

Re: [Openvortex-dev] Findings...


From: Jeff Muizelaar
Subject: Re: [Openvortex-dev] Findings...
Date: Tue, 22 Apr 2003 22:24:55 -0400 (EDT)


On Tue, 22 Apr 2003, Shamus wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi all!
>
> I've finally been able to track down the problem on my machine with
> the ALSA driver. In vortex_interrupt(...) there is a call to
> snd_mpu401_uart_interrupt(...) and something in that call is causing
> the kernel panic (unable to dereference virtual address or some
> such--see previous email for exact wording).
>
> I also found out the reason I was having trouble with the source code
> (so it seemed). It seems that (at least as of the 2.4.19 kernel) as
> soon as you call request_irq(...) that it starts calling the
> vortex_interrupt(...) function *immediately* and the source mask has
> the IRQ_TIMER bit set--not good! So anyway, if you don't clear this
> IRQ it keeps calling the IRQ handler until you acknowledge it. Since I
> had a return; statement above where it does
>
>       if (source & IRQ_TIMER)
>               hwread(vortex->mmio, VORTEX_IRQ_STAT);
>
> it would simply keep calling the IRQ handler--effectively locking up
> the system. All *before* the driver had a chance to call
> vortex_core_init(chip)! I'm not sure why the call to
> snd_mpu401_uart_interrupt(...) is causing the crash, but at least I
> know where to look... BTW, the crash occurred *after* it returned from
> vortex_core_init(chip), when the card is supposed to be fully
> initialized.  :-/
>
> Other than that, it would probably be a better idea to initialize the
> Vortex hardware *before* attempting to grab the IRQ using
> request_irq(...), since then the card would be properly initialized
> and ready to handle IRQs properly at that point. It would probably
> stop the spurious IRQ_TIMER from happening as well when grabbing the
> IRQ...

Yes, this appears to be a good idea and is what the binary driver appears
to do. I have made a fix and committed it to cvs.

Manuel,

from vortex_core_init:

  // vortex_enable_int() must be first !!

        what does this apply to?

  hwwrite(vortex->mmio, VORTEX_IRQ_CTRL, 0);

        where is this from?

  vortex_enable_int(vortex);


-Jeff





reply via email to

[Prev in Thread] Current Thread [Next in Thread]