On 09/27/2016 09:56 PM, Frank Filz wrote: >> We observed a ref leak of cache-inode entry (in V2.3-stable codepath) in >> case if there are multiple locks issued on that entry which may get merged. >> Attached the fix for it. >> >> Kindly review and merge it into V2.3-stable branch. > > Ok, I had to walk the code carefully... > > Looks good. > > Ah, and just looked at the 2.4 code... In the case of state_lock's call to > merge_lock_entry, it checks if the list is empty BEFORE calling > merge_lock_entry, if so, it takes the needed object reference to "pin" the > object.
The commit which Malahal pointed out in ganltc repo takes similar approach. We could probably take that patch instead for 2.3. Thanks, Soumya > > Another way to fix this would have been to add the new lock entry to the list > BEFORE calling merge_lock_entry... Then merge_lock_entry would NEVER empty > the list since the entry to be merged is already in the list (and > merge_lock_entry already skips it). > > Frank > > > --- > This email has been checked for viruses by Avast antivirus software. > https://www.avast.com/antivirus > ------------------------------------------------------------------------------ _______________________________________________ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
