Re: [RFC] Atmel-supplied hardware headers for AT91RM9200 SoC processor

[Andrew Victor <andrew@sanpeople.com>]

hi Alexey,

Thanks for the constructive comments.


> And check for non-NULL data in at91_add_device_usbh() is useless:

There are many more custom AT91RM9200-based boards.  That NULL check is
really just there to help the developers when they write their own
board-XX.c file.


> at91_wdt_ioctl() isn't __user annotated. Let alone it is ioctl.

All the other watchdog drivers use ioctl?  That is how it's described in
Documentation/watchdog/watchdog-api.txt.


> > +       char* command = kmalloc(2, GFP_KERNEL);
> 
> Anyone remembers 1 kmallocated byte?

That command buffer is passed down to the SPI driver, which then DMA's
directly from/to it.  We can't DMA to an address on the stack (atleast
not in 2.4 when that driver was written).


Regards,
  Andrew Victor