[PATCH 3/8] regmap: Add support for using regmap over ssbi
Stephen Boyd
sboyd at codeaurora.org
Tue Dec 10 20:32:51 EST 2013
On 12/10/13 16:51, Mark Brown wrote:
> On Tue, Dec 10, 2013 at 04:13:15PM -0800, Stephen Boyd wrote:
>> increment reg by 1 every time through this loop. Or should we just have
>> use_single_rw == true?
> No, it doesn't - it increments the address of reg by the size of a
> register value each time. Using use_single_rw might make sense, or if
> you can't do bulk I/O at all then hooking in via reg_read() and
> reg_write() in the config rather than trying to parse out the buffers
> might be even better (you can still make helpers to set that up).
Are you suggesting we implement the reg_read/reg_write as global helpers
that the config points to and then call regmap_init()? At a quick glance
it looks like we lose out on regmap_bulk_read() if we do that. There is
one driver that will use regmap_bulk_read(), but I suppose we can just
loop on regmap_read() and do our own increment? If we use use_single_rw
everything works and we can simplify this code to just pass the reg and
val buffers directly to ssbi_read/write.
This is what I have now.
static int regmap_ssbi_read(void *context,
const void *regp, size_t reg_size,
void *val, size_t val_size)
{
int ret;
u16 reg;
BUG_ON(reg_size != 2);
reg = *(u16 *)regp;
while (val_size) {
ret = ssbi_read(context, reg, val, 1);
if (ret)
return ret;
reg++;
val += sizeof(u8);
val_size -= sizeof(u8);
}
return 0;
}
With use_single_rw I think it can be this.
static int regmap_ssbi_read(void *context,
const void *reg, size_t reg_size,
void *val, size_t val_size)
{
BUG_ON(reg_size != 2);
return ssbi_read(context, *(u16 *)reg, val, 1);
}
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
More information about the linux-arm-kernel
mailing list