@@ -279,16 +279,6 @@ static inline int mm_has_notifiers(struct mm_struct *mm)structmmu_notifier*mmu_notifier_get_locked(conststructmmu_notifier_ops*ops,structmm_struct*mm);-staticinlinestructmmu_notifier*-mmu_notifier_get(conststructmmu_notifier_ops*ops,structmm_struct*mm)-{-structmmu_notifier*ret;--mmap_write_lock(mm);-ret=mmu_notifier_get_locked(ops,mm);-mmap_write_unlock(mm);-returnret;-}voidmmu_notifier_put(structmmu_notifier*subscription);voidmmu_notifier_synchronize(void);
The doc of mmu_notifier_get_locked() has a reference to this function.
"Each call to mmu_notifier_get() must be paired with a call to
mmu_notifier_put()." - that should have been mmu_notifier_get_locked(),
maybe fix that up right away.
--
Thanks,
David / dhildenb
@@ -279,16 +279,6 @@ static inline int mm_has_notifiers(struct mm_struct *mm)structmmu_notifier*mmu_notifier_get_locked(conststructmmu_notifier_ops*ops,structmm_struct*mm);-staticinlinestructmmu_notifier*-mmu_notifier_get(conststructmmu_notifier_ops*ops,structmm_struct*mm)-{-structmmu_notifier*ret;--mmap_write_lock(mm);-ret=mmu_notifier_get_locked(ops,mm);-mmap_write_unlock(mm);-returnret;-}voidmmu_notifier_put(structmmu_notifier*subscription);voidmmu_notifier_synchronize(void);
Looks like I missed this from
Fixes: f25a546e6529 ("RDMA/odp: Use mmu_interval_notifier_insert()")
Though, I have a vauge recollection that some SVA IOMMU patches posted
wanted to use this too? They can bring it back if it makes sense
Looks like David is right about the comment, but aside from that:
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
v2 should go through Andrew's tree
Thanks,
Jason