> Acked-by: David Brownell <[email protected]>
>
> Though the PIO bits still bother me ... I'd use *only* the
> io{read,write}8_rep() calls to load/unload the FIFO. While
> those calls haven't broken yet, it's rude to (a) open-code
> io{read,write}32_rep() like that, and (b) assume all buffers
> coming in to that function are 32-bit aligned. PIO is hardly
> ever used, given EDMA. Maybe just one or two calls when
> enumerating. (SDIO might use it more often.) That can be
> fixed a bit later.
>
> Also, it'd be good to submit a patch to the DaVinci tree
> to make it match whatever Pierre eventually merges. (Less
> the clock name change, of course, which is only needed to
> cope with the elderly clock logic now in mainline.)
In my testing, I have observed that io{read,write}8_rep() does not work
properly in pure PIO mode i.e. when use_dma is set 0. So, I have introduced
both io{read,write}8_rep() and io{read,write}32_rep() and we could handle
different size of buffers irrespective of whether it is multiple of 4 bytes or
not.
>
> Build bug in probe():
>
> > + host->clk = clk_get(&pdev->dev, MMCSDCLK);
>
> Put quotes around "MMCSDCLK".
I will add quotes and re-submit patch today or tomorrow.
-Purushotam
_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
