No DMA RX on some BCM4321, on BCM43224 and BCM43225

[Jonas Gorski]

2011/7/19 Larry Finger <Larry.Finger at lwfinger.net>:
> On 07/19/2011 01:56 AM, Rafał Miłecki wrote:
>>
>> W dniu 19 lipca 2011 01:32 użytkownik Rafał Miłecki<zajec5 at gmail.com>
>>  napisał:
>>>
>>> Larry, for now, can you check:
>>>
>>> 1) When wl does set parity (0x800) bit
>>> 2) Does newer wl always use 38 offset for RX header?
>>
>> After adding routing (translation) address, using parity bit (0x800)
>> and cold booting, DMA seems to work =)
>
> Good to hear. I have not found anything about the offset of 38, but I have
> some confusing stuff on the parity enable bit.
>
> One routine in the reference driver does the following:
>
> 1. If bit 1 is set in the DMA control flags (this bit was set to 1 in the
> init)
>  a. Read DMA control
>  a. OR result with 0x800 and write to control
>  a. Read DMA control
>  a. If bit 0x800 is set
>  1. Write original contents back to control
>  a. Otherwise
>  1. Clear bit 1 in DMA control flags
>
> The above routine makes it look as if they have tested whether the device
> implements the parity bit. The funny part is that I cannot find any other
> place where they test the DMA control flags for bit 1 set. There is a spot
> where they unconditionally set bit 0x800 in the DMA control, which seems to
> match what you see.

The brcm80211 sources agree:
<http://git.kernel.org/?p=linux/kernel/git/gregkh/staging-2.6.git;a=blob;f=drivers/staging/brcm80211/brcmsmac/dma.c;h=ea17671efb63aa0c9556c8e46c842734ba398f86;hb=refs/heads/staging-next#l1147>


Regards,
Jonas