[PATCH 5/5] Use size_t for memory offsets

[Sascha Hauer]

On Fri, Oct 14, 2011 at 10:59:03AM +0200, Loïc Minier wrote:
> On Fri, Oct 14, 2011, Sascha Hauer wrote:
> > You should change the prototypes in include/driver.h aswell.
> 
>  Ah thanks, now it strikes me that the very same constructs are present
>  in many file_operations implementations; e.g. imx_iim_cdev_read and
>  imx_iim_cdev_write also use an ulong offset, as well as
>  ubi_volume_cdev_read/ubi_volume_cdev_write (unsigned long), lp_read,
>  miidev_read/_write etc.
> 
>  I had a look at file_operations in linux now, and it uses
>  size_t/ssize_t and a loff_t type for regular read/write:
> struct file_operations {
>         loff_t (*llseek) (struct file *, loff_t, int);
>         ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
>         ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
> 
>  however for aio:
>         ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
>         ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
> 
>  loff_t is defined as long long on 32-bits and 64-bits arches, which I
>  believe are both 8 bytes.
> 
>  So perhaps it's better to switch from ulong to unsigned long long for
>  offsets?  This isn't important for mem_read/mem_write, but it would be
>  for e.g. MMC accesses as it's of course valid to seek after the first
>  4 G of a MMC on a 32-bits system.

Yes, an 8 byte type would be definitely better for file offsets. It
would be a first step to support SD cards > 4G. I have looked into
this several times before and ended up with huge patches everytime
I tried.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |