Hi Amir! On Tue 03-08-21 21:03:40, Amir Goldstein wrote: > This idea was suggested on year ago [1], but I never got to test > its performance benefits. > > Following the performance improvement report from kernel robot [2], > please consider these changes. > > I have other optimization patches for no ignored mask etc, but the > "no marks" case is the most low hanging improvement. Thanks for the improvement! The series looks good except for that one comment I had. If you respin the series with that addressed, I can take it to my tree. Honza > > Thanks, > Amir. > > [1] https://lore.kernel.org/linux-fsdevel/CAOQ4uxgYpufPyhivOQyEhUQ0g+atKLwAAuefkSwaWXYAyMgw5Q@xxxxxxxxxxxxxx/ > [2] https://lore.kernel.org/linux-fsdevel/CAOQ4uxisyDjVpWX1M6O4ugxBbcX+LWWf4NQJ+LQY1-3-9tN+BA@xxxxxxxxxxxxxx/ > > Amir Goldstein (4): > fsnotify: replace igrab() with ihold() on attach connector > fsnotify: count s_fsnotify_inode_refs for attached connectors > fsnotify: count all objects with attached connectors > fsnotify: optimize the case of no marks of any type > > fs/notify/fsnotify.c | 6 ++-- > fs/notify/mark.c | 73 +++++++++++++++++++++++++++++++++------- > include/linux/fs.h | 4 +-- > include/linux/fsnotify.h | 9 +++++ > 4 files changed, 75 insertions(+), 17 deletions(-) > > -- > 2.25.1 > -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
