Re: [PATCH] [Rev2] MPC5121 FEC support

[Scott Wood <scottwood@freescale.com>]

Sam Ravnborg wrote:
>> +	data = of_get_property(ofdev->node, "fsl,align-tx-packets", &len);
>> +	if (data && len == 4)
>> +		fpi->align_tx_packets = *data;
>> +
> Where did "4" come from. USe a define with a desriptive name.

It's sizeof(u32), i.e. one device tree cell.  This is fairly normal.

>>  	fpi->rx_ring = 32;
>>  	fpi->tx_ring = 32;
> Same for "32"
>>  	fpi->rx_copybreak = 240;
> Same for "240".

They're arbitrary tuning parameters.  How is a #define any more 
descriptive than the field name?

Besides, that's pre-existing, and has nothing to do with this patch.

>> --- a/drivers/net/fs_enet/fs_enet.h
>> +++ b/drivers/net/fs_enet/fs_enet.h
>> @@ -10,12 +10,17 @@
>>  
>>  #include <linux/fs_enet_pd.h>
>>  #include <asm/fs_pd.h>
>> +#ifdef CONFIG_FS_ENET_MPC5121_FEC
>> +#include "fec_mpc5121.h"
>> +#endif
> 
> Which is this include ifdeffed - looks like some wrong concept.

This has already been discussed.  There are two similar but different 
ethernet controllers that are being targeted, and the chips they are a 
part of (8xx and 512x) are already mutually exclusive with respect to 
multiplatform kernels due to core differences.

> The amount of ifdef introduced looks bad..

Yes, it's bad -- but it's a matter of which is the lesser evil, a few 
ifdefs or large amounts of mostly duplicated code.

> And try to run the patch through scripts/checkpatch.pl
> And try to split it up a bit.

Other than the fec_t thing, I don't see any needed splitting...

-Scott