Hi Bart,
On 9/4/24 2:15 PM, Bart Van Assche wrote:
On 9/4/24 2:00 PM, Badhri Jagan Sridharan wrote:
https://lore.kernel.org/all/ZsiYRAJST%2F2hAju1@xxxxxxxxxxxxxxxxx/ was
already accepted
Thanks, I hadn't noticed this yet.
and is perhaps better than what you are suggesting as
it does not use the internal methods of mutex_init().
Although I do not have a strong opinion about which patch is sent to
Linus, I think my patch has multiple advantages compared to the patch
mentioned above:
- Cleaner. lockdep_set_class() is not used. Hence, it is not possible
that the wrong lockdep key is used (the one assigned by
mutex_init()).
- The lock_class_key declaration occurs close to the sw->lock
declaration.
- The lockdep_register_key() call occurs close to __mutex_init() call
that uses the registered key.
- Needs less memory in debug kernels. The advantage of __mutex_init()
compared to mutex_init() is that it does not allocate (static) memory
for a lockdep key.
Thanks for the patch.
While I agree on (1) & (4), *may* be a good reason to reconsider.
However, I have seen almost 30+ instances of the prior
method
(https://lore.kernel.org/all/20240822223717.253433-1-amitsd@xxxxxxxxxx/)
of registering lockdep key, which is what I followed.
However, if that's is not the right way, it brings into question the purpose
of lockdep_set_class() considering I would always and unconditionally use
__mutex_init() if I want to manage the lockdep class keys myself or
mutex_init() if I didn't.
Thanks,
Amit
Thanks,
Bart.
