Details
-
Bug
-
Resolution: Fixed
-
Minor
-
Lustre 2.5.1
-
3
-
14818
Description
In some case the memory allocated by lustre could be zero-out twice (kmalloc followed by memset in LIBCFS_ALLOC_*, followed by another memset in the source code).
Additionally if GFP_ZERO is set, then the kernel will zero out the
block too. Surely that can't be too good for performance.
cfs_cpt_malloc() has a similar issue.