Re: [PATCH] rt2x00: rt2800lib: Fix rx path on SoC devices

[Gertjan van Wingerde <gwingerde@gmail.com>]

On 04/27/10 08:28, Helmut Schaa wrote:
> Am Montag 26 April 2010 schrieb Gertjan van Wingerde:
>> On 04/26/10 13:48, Helmut Schaa wrote:
>>> Restore the rfcsr initialization for RT305x SoC devices which was removed
>>> by "rt2x00: Finish rt3070 support in rt2800 register initialization.".
>>>
>>> This fixes the rx path on SoC devices.
>>>
>>> Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
>>> ---
>>>  drivers/net/wireless/rt2x00/rt2800lib.c |   34 ++++++++++++++++++++++++++++++-
>>>  1 files changed, 33 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/drivers/net/wireless/rt2x00/rt2800lib.c b/drivers/net/wireless/rt2x00/rt2800lib.c
>>> index 2648f31..1358d9a 100644
>>> --- a/drivers/net/wireless/rt2x00/rt2800lib.c
>>> +++ b/drivers/net/wireless/rt2x00/rt2800lib.c
>>> @@ -1703,7 +1703,8 @@ int rt2800_init_rfcsr(struct rt2x00_dev *rt2x00dev)
>>>  	if (!rt2x00_rt(rt2x00dev, RT3070) &&
>>>  	    !rt2x00_rt(rt2x00dev, RT3071) &&
>>>  	    !rt2x00_rt(rt2x00dev, RT3090) &&
>>> -	    !rt2x00_rt(rt2x00dev, RT3390))
>>> +	    !rt2x00_rt(rt2x00dev, RT3390) &&
>>> +	    !(rt2x00_is_soc(rt2x00dev) && rt2x00_rt(rt2x00dev, RT2872)))
>>>  		return 0;
>>
>> As indicated in the separate email, we should also check for the presence of an RF3020, RF3021
>> or RF3022 RF chipset for the RT2872 case (maybe this should be more generic, because I'm sure
>> this will also hold for the other RT chipsets).
> 
> Ok, do you mean we should only check for rfXXX instead of checking for SoC
> and rt2872? Or should we check for SoC + rfXXX?

I think that at the moment we should be looking at checking for SoC, RT2872 and RFxxxx.
To be honest I don't completely know what these RF CSR values represent, we can only find them
in a SoC Ralink driver, so FWIW it may simply be good settings for an AP (the main use case for
a SoC device) but not so good for an STA.

So properly applying this only to SoC devices with the right RT chipset and right RF chipset
seems to be the proper thing here.

> 
>>>  	/*
>>> @@ -1771,6 +1772,37 @@ int rt2800_init_rfcsr(struct rt2x00_dev *rt2x00dev)
>>>  		rt2800_rfcsr_write(rt2x00dev, 29, 0x8f);
>>>  		rt2800_rfcsr_write(rt2x00dev, 30, 0x20);
>>>  		rt2800_rfcsr_write(rt2x00dev, 31, 0x0f);
>>> +	} else if (rt2x00_rt(rt2x00dev, RT2872)) {
>>> +		rt2800_rfcsr_write(rt2x00dev, 0, 0x50);
>>> +		rt2800_rfcsr_write(rt2x00dev, 1, 0x01);
>>> +		rt2800_rfcsr_write(rt2x00dev, 2, 0xf7);
>>> +		rt2800_rfcsr_write(rt2x00dev, 3, 0x75);
>>> +		rt2800_rfcsr_write(rt2x00dev, 4, 0x40);
>>> +		rt2800_rfcsr_write(rt2x00dev, 5, 0x03);
>>> +		rt2800_rfcsr_write(rt2x00dev, 6, 0x02);
>>> +		rt2800_rfcsr_write(rt2x00dev, 7, 0x50);
>>> +		rt2800_rfcsr_write(rt2x00dev, 8, 0x39);
>>> +		rt2800_rfcsr_write(rt2x00dev, 9, 0x0f);
>>> +		rt2800_rfcsr_write(rt2x00dev, 10, 0x60);
>>> +		rt2800_rfcsr_write(rt2x00dev, 11, 0x21);
>>> +		rt2800_rfcsr_write(rt2x00dev, 12, 0x75);
>>> +		rt2800_rfcsr_write(rt2x00dev, 13, 0x75);
>>> +		rt2800_rfcsr_write(rt2x00dev, 14, 0x90);
>>> +		rt2800_rfcsr_write(rt2x00dev, 15, 0x58);
>>> +		rt2800_rfcsr_write(rt2x00dev, 16, 0xb3);
>>> +		rt2800_rfcsr_write(rt2x00dev, 17, 0x92);
>>> +		rt2800_rfcsr_write(rt2x00dev, 18, 0x2c);
>>> +		rt2800_rfcsr_write(rt2x00dev, 19, 0x02);
>>> +		rt2800_rfcsr_write(rt2x00dev, 20, 0xba);
>>> +		rt2800_rfcsr_write(rt2x00dev, 21, 0xdb);
>>> +		rt2800_rfcsr_write(rt2x00dev, 22, 0x00);
>>> +		rt2800_rfcsr_write(rt2x00dev, 23, 0x31);
>>> +		rt2800_rfcsr_write(rt2x00dev, 24, 0x08);
>>> +		rt2800_rfcsr_write(rt2x00dev, 25, 0x01);
>>> +		rt2800_rfcsr_write(rt2x00dev, 26, 0x25);
>>> +		rt2800_rfcsr_write(rt2x00dev, 27, 0x23);
>>> +		rt2800_rfcsr_write(rt2x00dev, 28, 0x13);
>>> +		rt2800_rfcsr_write(rt2x00dev, 29, 0x83);
>>>  	}
>>>  
>>>  	if (rt2x00_rt_rev_lt(rt2x00dev, RT3070, REV_RT3070F)) {
>>
>> Here we need 2 additional rfcsr_writes, for RF CSR 30 and RF CSR 31, both set to value 0x00.
>> Also, we can add an "return 0" at the end of the sequence, as nothing more needs to be done for RT2872.
> 
> Ok, thanks for looking that up.

And please add the rt2x00_is_soc(rt2x00dev) test to that hunk as well. It makes it easier to see that
that piece is for SoC RT2872 devices only.

P.S. It seems that John already applied this patch without even Ivo or me Acking it, so I guess an
incremental patch would be best now.

---
Gertjan