[PATCH] fix-gta02-irq-arbiter-priority.patch

[Werner Almesberger]

Andy Green wrote:
> Since SDI stuff is fully sycnchronous, allowing preemption inside the
> interrupt service for it doesn't obviously make trouble.

It appears that is encourages some not-so-obvious gremlins to show
more presence, though. "jc" has reported seeing the following AR6k
initialization failure at boot:

[21474538.910000] mmc1: queuing CIS tuple 0x80 length 1
[21474538.910000] mmc1: queuing CIS tuple 0x81 length 1
[21474538.910000] mmc1: queuing CIS tuple 0x82 length 1
[21474538.910000] mmc1: new SDIO card at address 0001
[21474538.915000] sdio_ar6000 mmc1:0001:1: deviceInsertedHandler: -1
[21474538.915000] sdio_ar6000 mmc1:0001:1: kthread_stop (ar6000_io): -4
[21474538.920000] sdio_ar6000: probe of mmc1:0001:1 failed with error -4

Using commit de8460932576502fb6bf61968a37c4d2aed2d6e2 with a modified
gta02_moredrivers: http://prostor.hopto.org/~jc/config

Apparently, this is reproducible. As a work-around, the usual
unbind/bind brings up WLAN properly:

echo s3c2440-sdi >/sys/bus/platform/drivers/s3c2440-sdi/unbind
echo s3c2440-sdi >/sys/bus/platform/drivers/s3c2440-sdi/bind

I built and tried "the same" kernel but if course WLAN came up
without a hitch.

I also saw one stray failure myself:

[274661.070000] mmc1: new SDIO card at address 0001
[274661.105000] ar6000_available
[274661.445000] Unable to decrement the command credit count register
[274661.445000] Unable to write to the device

I've never seen this one before or since (in maybe a dozen boots
since it happened.)

Just a "heads up." I still need to hit something I can reproduce.

- Werner