I think the root cause in hda_intel driver's self.
It gets io-apic irq initialized at first, and it will use
azx_acquire_irq to install handler after check if MSI can be enabled.
And when it try to enable the MSI, that will start the int in the chip.
Even handler for MSI is not installed.
YH
On Wednesday 15 November 2006 23:25, Lu, Yinghai wrote:
> I think the root cause in hda_intel driver's self.
>
> It gets io-apic irq initialized at first, and it will use
> azx_acquire_irq to install handler after check if MSI can be enabled.
> And when it try to enable the MSI, that will start the int in the chip.
> Even handler for MSI is not installed.
>
> YH
That seems rather stupid. Perhaps installing a conditional handler before
doing anything with MSI will solve that problem... The problem then just
revolves around replacing the conditional handler with the real one when MSI
is fully enabled. If I understood the code for this driver better I'd try to
make this work. As it stands I'm still trying to recover from the loss of the
work I did on another project. (DRM based uniform kernel graphics interface -
had a blackout here that caused my devel machine to crash and left me with a
trashed drive)
DRH