On Sat, Nov 25, 2000 at 05:48:40PM -0800, Mark Vojkovich wrote:
>
>
> On Sat, 25 Nov 2000 [EMAIL PROTECTED] wrote:
>
> >
> > Andrew Morton reports that the S3 and Matrox cards are known to
> > have problems giving up the PCI bus.
>
> On many cards, Matrox for sure, it's virtually impossible
> to prevent PCI disconnects from happening. Matrox discusses
> this in their documentation, for instance, the G400 docs,
> section 4.1.2
>
> "PCI Retry Handling
>
> In certian situations the chip may not be able to repond
> to a PCI access immediately, therefore, a number of retry
> cycles will be generated. A retry will be asserted when:" ...
>
> [ a whole mess of cases including direct framebuffer accesses ]
>
> For real-time platforms you folks probably need to be using
> old ISA cards or something. Some PCI chipsets make the retry
> problem worse. LX chipset is infamous.
Our measurements show that for most cards, PCI is far better.
The problem seems to arise on a combination of
certain video cards, certain chipsets. certain drivers
Andrew quotes Alan Cox :
"The S3 cards dont error the transaction with a busy when the fifo
is full. They lock the bus until it isn't".
This seems a violation of PCI spec.
--
---------------------------------------------------------
Victor Yodaiken
Finite State Machine Labs: The RTLinux Company.
www.fsmlabs.com www.rtlinux.com
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/
