Hi Andrew,
On 2/10/21 3:52 AM, Andrew Lunn wrote:
> On Tue, Feb 09, 2021 at 10:53:20AM +0100, Michal Simek wrote:
>> +static int usb5744_i2c_probe(struct i2c_client *client,
>> + const struct i2c_device_id *id)
>> +{
>> + struct device *dev = &client->dev;
>> + int ret;
>> +
>> + /* Trigger gpio reset to the hub. */
>> + ret = usb5744_init_hw(dev);
>> + if (ret)
>> + return ret;
>> +
>> + /* Send SMBus command to boot hub. */
>> + ret = i2c_smbus_write_word_data(client, 0xAA, swab16(0x5600));
>
> Hi Michal
>
> This is not my area of the kernel. But that swab16() stood out, and
> made me wonder about endianness. Will this work correctly on big and
> little endian hosts?
thanks for bringing this up. I didn't test it on BE system.
I have grepped the kernel
[linux](xnext/usb5744)$ git grep i2c_smbus_write_word_data | grep swab
drivers/media/i2c/uda1342.c:17: i2c_smbus_write_word_data(client, reg,
swab16(value));
drivers/media/i2c/vpx3220.c:97: if (i2c_smbus_write_word_data(client,
0x27, swab16(fpaddr)) == -1) {
drivers/media/i2c/vpx3220.c:106: if (i2c_smbus_write_word_data(client,
0x28, swab16(data)) == -1) {
drivers/media/i2c/vpx3220.c:120: if (i2c_smbus_write_word_data(client,
0x26, swab16(fpaddr)) == -1) {
drivers/usb/misc/usb5744.c:50: ret = i2c_smbus_write_word_data(client,
0xAA, swab16(0x5600));
include/linux/i2c.h:168: return i2c_smbus_write_word_data(client,
command, swab16(value));
And last one is interesting
164 static inline s32
165 i2c_smbus_write_word_swapped(const struct i2c_client *client,
166 u8 command, u16 value)
167 {
168 return i2c_smbus_write_word_data(client, command,
swab16(value));
169 }
And this function is also used
[linux](xnext/usb5744)$ git grep i2c_smbus_write_word_swapped | wc -l
76
I think it would be the best to test it and see if this code works on BE
but I need to prepare it first.
And current code is aligned with others but it doesn't mean that it is
correct.
Thanks,
Michal
