> 01.11.2013, 23:30, "Andrew Morton" <a...@linux-foundation.org>:
>> set_bit() operates on longs. So if we do
>>
>> struct foo { u32 a; u32 b; } f;
>> set_bit(0, (long *)&f.a);
>>
>> then we'll scribble on f.b on a big-endian 64-bit machine.On Sat, Nov 02 2013, Рустафа Джамурахметов <z...@ioremap.net> wrote: > Argh, why would we just don't do that? Its in-memory field, it can be > anything, I wouldn't be surprised if it even can be non-atomic because > of proper locks already being held If the driver does not require an atomic set_bit operation for setting and testing the flag, feel free to prepare a patch replacing the whole thing with a plain bool.
signature.asc
Description: PGP signature
