For bulk I/O it is both convenient and more sensible to pre-swap the data
rather than doing the swap as part of the I/O operation so move the byte
swaps we're currently doing into the core write function into the register
based functions, giving the bulk write function a straight pass through
to the chip.
This leaves reads inconsistent, this will be addressed as a followup patch.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
for (i = 0; i < bytes / 2; i++) {
dev_vdbg(wm8994->dev, "Write %04x to R%d(0x%x)\n",
for (i = 0; i < bytes / 2; i++) {
dev_vdbg(wm8994->dev, "Write %04x to R%d(0x%x)\n",
- buf[i], reg + i, reg + i);
-
- buf[i] = cpu_to_be16(buf[i]);
+ be16_to_cpu(buf[i]), reg + i, reg + i);
}
return wm8994->write_dev(wm8994, reg, bytes, src);
}
return wm8994->write_dev(wm8994, reg, bytes, src);
+ val = cpu_to_be16(val);
+
mutex_lock(&wm8994->io_lock);
ret = wm8994_write(wm8994, reg, 2, &val);
mutex_lock(&wm8994->io_lock);
ret = wm8994_write(wm8994, reg, 2, &val);
* @wm8994: Device to write to
* @reg: First register
* @count: Number of registers
* @wm8994: Device to write to
* @reg: First register
* @count: Number of registers
- * @buf: Buffer to write from.
+ * @buf: Buffer to write from. Data must be big-endian formatted.
*/
int wm8994_bulk_write(struct wm8994 *wm8994, unsigned short reg,
int count, u16 *buf)
*/
int wm8994_bulk_write(struct wm8994 *wm8994, unsigned short reg,
int count, u16 *buf)
ret = wm8994_write(wm8994, reg, 2, &r);
out:
ret = wm8994_write(wm8994, reg, 2, &r);
out: