dcbw at redhat.com
Thu Sep 4 09:27:17 EDT 2008
On Thu, 2008-09-04 at 10:00 +0000, Cliff Cai wrote:
> >> Yes,at last I found an AP without encryption,so I successfully associated with it.
> >> Everything is good except the bad performance,any suggestion about how to improve it?
> >>>> One more question,I want to know if there is available firmware support SDIO SPI mode now?
> >>>> I saw a post said it was not support that time ,how about now?
> >>>> http://lists.infradead.org/pipermail/libertas-dev/2008-March/001382.html
> >>> Various parts have had G-SPI support for a while but of course that
> >>> requires the firmware which people need to get form Marvell. If you've
> >>> got firmware and the necessary host interface, we could try to give it a
> >>> shot. But I have to admin I'm not familiar with how the host controller
> >>> needs to be set up to handle SPI mode. Does it require a specific SDIO
> >>> controller type, or can most SDIO controllers handle SPI? Does it
> >>> require changes to the kernel SDIO stack to accommodate SPI? Basically,
> >>> changes to the actual libertas driver aren't the only thing required
> >>> here because I'm pretty sure we'd need changes to the host controller
> >>> drivers and the stack too.
> >> I have another board which has no SD host,but has a SPI bus.
> >> AFAIK,Currently the G-SPI driver for Marvell 8686 is non-opensource,
> >> I won't try it.Now there is a SPI simulated SD host driver in the kernel tree,see /drivers/mmc/host/mmc_spi.c
> >> we've successfully tried it with a SD card,So I think SDIO SPI mode is a way to go,but I can't make it work.
> >> that's why I ask this question.
> > Single or dual-stage firmware?
> What does this mean?I just use the same firmwares for SDIO to try SDIO SPI mode
> >We can probably make SDIO SPI mode work
> > if we try hard enough, but there are some quirks with it.
> Maybe,I haven't heard anyone who have successfully made it work yet,actually seldom people have tried it:)
> Do you have suggestions to improve the performance,My SDIO WIFI seems has a bad performance:(
> maybe I should open another thread to discuss it,shouldn't I?
While the transfer rate varies, I've been getting up to 500K/s (yes K)
from an offsite HTTP archive with the sd8686 b2 reference board, using
v9 firmware and 2.6.27-based wireless-testing git with driver logging
(0x4043a7) enabled. The SD host controller might have a lot to do with
it; mine is a Fujitsu Lifebook P1510D with a Ricoh R5C822 (rev 13).
One thing I did notice was that the card's reported rate never got above
6 Mbps even though I'm about 30 feet away from the AP in light RF
environment (~5 APs visible). It might have to do with rate adaptation
which is certainly something to look at.
But the point being that at least with the SD cards, we don't yet know
whether the problem is the controller, the driver/firmware, or both.
More information about the libertas-dev