Re: [PATCH 1/6] mm/mmu_notifier: Allow two-pass struct mmu_interval_notifiers

["Thomas Hellström" <thomas.hellstrom@linux.intel.com>]

Hi, Jason

On 9/9/25 19:18, Jason Gunthorpe wrote:
> On Thu, Aug 21, 2025 at 01:46:21PM +0200, Thomas Hellström wrote:
>> +struct mmu_interval_notifier_finish {
>> +	struct list_head link;
>> +	/**
>> +	 * @finish: Driver callback for the finish pass.
>> +	 * @final: Pointer to the mmu_interval_notifier_finish structure.
>> +	 * @range: The mmu_notifier_range.
>> +	 * @cur_seq: The current sequence set by the first pass.
>> +	 *
>> +	 * Note that there is no error reporting for additional passes.
>> +	 */
>> +	void (*finish)(struct mmu_interval_notifier_finish *final,
>> +		       const struct mmu_notifier_range *range,
>> +		       unsigned long cur_seq);
> I would rather this be in mmu_interval_notifier_ops

Thanks for reviewing.

We could have the struct mmu_interval_notifier_finish have a pointer to 
mmu_interval_notifier_ops or even to
mmu_interval_notifier. Now that you mention it IIRC Linus has pointed 
out that he strongly prefer function pointers in const ops wherever 
possible. Would like to keep the linked list, though, as we've discussed 
before, for passing of state and to avoid a second interval tree traversal.

Thanks,

Thomas



>   though I guess I
> see why it was done like this, I don't think it is a great idea for
> DRM to wrapper the notifier library with yet another library :\
>
> Regardless
>
> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
>
> Jason