Subject: Re: or DELAY() ?
To: Reinoud Zandijk <email@example.com>
From: Frank Kardel <firstname.lastname@example.org>
Date: 09/09/2006 15:56:15
Reinoud Zandijk wrote:
>On Fri, Sep 08, 2006 at 10:23:39PM +0200, Reinoud Zandijk wrote:
>>i'm having a curious problem with my auixp driver (PCI). Using a clean
>>sourcetree dated 20060907 it all works fine but when updating the
>>sourcetree to 20060908 it gets into trouble communicating with its codecs.
>>The codecs detect fine but it times out all the times when reading in the
>>function auixp.c:auixp_read_codec(). I noticed christos (?) changed some
>>PCI reset functions that might be the cause though when i reverse that
>>patch it still doesn't work anymore so i dont think thats the cause.
>>Running a GENERIC or a GENERIC.MPACPI doesn't matter and the problem
>>resolves itself when i disable the auixp driver.
>>Did something change in the PCI framework that could cause it ? Could any
>>of you PCI guru's have a look at the driver to see what it does wrong?
>Greg Oster pointed to me that it could also be due to the
>timecounter/DELAY() changes... has that interface/resolution changed?
no, at least not intentionally. The x86 delay() loop should still be fine.
But maybe someone can cross check.
Perry/gdamore merged the isa/clock.c implementations for i386 and amd64
>code uses DELAY(2) a few times to wait for the obliged few microsecs to
>give the device time to respond.
Just the data points above.
>Thanks in advance,