7 Aug
2007
7 Aug
'07
11:23 a.m.
At Tue, 07 Aug 2007 14:45:12 +0200,
I wrote:
At Mon, 6 Aug 2007 21:58:29 +0200,
Fons Adriaensen wrote:
... well, to be a bit more specific, suppose the case of a wrong IRQ
assignment. ACPI has many built-in quirks as workaround of broken
BIOS. It might be your device. It might be fixed in the later
release.
The PCI core code change may influence. Possibly the default IRQ
routing is changed for your device, which you needed a special boot
option.
Or, the driver using MSI may behave utterly differently depending on
the kernel version because of the core MSI implementation.
It's more or less a same problem from the user-space, "wrong IRQ".
But, the cause can vary indeed.
Takashi
On Mon, Aug 06, 2007 at 07:13:12PM +0200, Takashi Iwai wrote:
A "kernel bug" is too generic wording. In most cases, it's a bug in
other components that a driver depends on. For example, the kernel
memory management, scheduler, ACPI, IRQ handler, etc. are involved
with the sound driver. If one of them is broken, the driver doesn't
work. Then, well, you made it really too dramatic.
Requesting the update to
the latest kernel is a very standard procedure. Otherwise, the
developer cannot start real debugging at all.
Not being a kernel/driver specialist at all, I do have some
difficulty in believing this.
The kernel has to call some routines in the driver, and provide
some services to it. I find it difficult to imagine
- how a kernel bug could affect just one driver and have no
impact on all the others.