GPFS 184.108.40.206 is now available from IBM Fix Central:
Problems fixed in GPFS 220.127.116.11
September 20, 2016
* Fix code for handling mmfsctl suspend-write operation to prevent blocking of close operation after reading file/directory. This issue only affect nodes running AIX.
* Fix a GNR logging issue that may cause RG recovery failure and data lost.
* Fix a SIGV problem when running mmdefrag on FPO file system with trace enabled.
* Make GSKit the default encryption engine on s390.
* Fix a bug in which start protocol tracing (with mmprotocoltrace) would fail on nodes where daemon node name and admin node name were not the same.
* Fix a problem in which updating quota files will fail after running mmfsck.
* Fix a problem in which GPFS may have incorrect contact nodes information in memory even though the correct contact nodes information was provided by the mmremotecluster add command.
This could happen if the same cluster was previously added with incorrect contact nodes.
* Fix a race condition that could cause incorrect file size being reported to an application. This only affect Linux and only when the application checks the file size right after inc
reasing file size via writing past end of file.
* This fix will make sure the NFSv3 files are closed when using Ganesha.
* Resolve interdependencies of Mkdirs and Renames so as to avoid queue flush getting stuck.
* Fix a deadlock that can occur during look up of a .snapshots/ directory of a fileset.
* Fix a problem that causes mmsdrrestore on all CCR nodes to fail if the command is issued from a nonquorum node.
* Make GPFS work on Intel Broadwell CPU with kernel 3.7 or later.
* Workloads with heavy use of medium-sized writes (100K-500K) may see improved performance.
* Fix Assert exp(threadP != __null) in file rbmd11628b/src/avs/fs/mmfs/ts/classes/tscomm/llcomm.C which can occur under heavy stress.
* Fix a high contention in read code path which improves the performance when large number of threads/processes read the same block of a file concurrently.
* Fix a GPFS daemon failure during QosIoMon::startManager - logAssertFailed - acquireMgrMutex. Apply this fix if you use QOS features.
* Fix an issue in the AFM environment where fileset recovery causes gateway node to crash if recovery detects that the symlink have attribute(Example. chown -h user:name) changes.
* Fix snapshot restore when building restore operation lists. This issue only happens when there are files with inode numbers bigger than the maximum value of 32 bits integer in the file system.
* Fix a problem in which gpfs.snap prematurely exits if the GUI process is not running.
* Fix the "Assert on Structure Error on FileMetadata::getSnapDataBlockDiskAddr" when reading a clone child file from a snapshot file system. This issue would only occur on a clone chi
ld file and only when the data of the clone child file is modified in the root file system after creating the snapshot.
* Fix a signal 11 error that was caused by a race condition of mmadddisk and block allocation.
* Fix an assert "fsdaP->getNValidAddrs() == nAllocated" which happens on a FPO file system while doing fragment block allocation.
* Fix signal 11 caused by a very rare race condition between recovery group resign/relinquish and pdisk I/O activities.
* Fix a bug in which mmgetstate reports the wrong status during heavy stress.
* Fix a problem in which the mmsmb exportacl getid command failed to return the correct SID of a user or group.
* Fix a GNR server crash that can occur during pdisk discovery.
* Fix a deadlock that can occur during a heavy IO-load while using LROC devices while requesting pagepool buffer allocations.
* This fix allows GPFS on Windows to fetch IMU/RFC2307 mappings from a single alternate trusted domain.
* Fix a deadlock that can occur when accessing snapshot file data.
* Fix a 760(E_CDITTO_LOCK) error which can occur on a punch hole operation on a clone child file.
* Fix a not implemented error that can occur calling GPFS APIs on a system in which we have reached the maximum number of open files.
* Fix a potential deadlock during a PIT repair/restripe run if a file is truncated when the new end of file is being repaired/restriped by the PIT process.
* Fix an assert which happens when logtipback vdisk is created before loghome vdisk.
* Fix a NT_STATUS_NONE_MAPPED error which can occur using the 'mmsmb export change' command when attempting to set the 'admin users' SMBOption.
* The logic for preserving the state of world for expel data collection is adjusted to avoid causing mutex contention resulting in an unstable cluster.
* Fix an assertion in GPFS send message when shutting down an invalid socket number.
* Fix a deadlock that can occur when prefetch recovery is happening and applications are trying to access the files in the fileset.
* mmchattr --delete-attr can remove gpfs.BGF, gpfs.WAD or gpfs.WADFG now.
* Fix a problem when both maxblocsize and scatterBufferSize are equal. The daemon asserts when the buffer memory usage is high.
* Fix an assert that is caused by a race between SGExceptionLocalPanicThread and repairFile.
* Fix a problem in which SGDesc migrate code handles multiple disk failures incorrectly.
* Fix information messages from mmapplypolicy which contain "hit" counts for EXCLUDE rules, which were incorrectly reported as 0 when -N and -g options are used.
* Fix a segmentation fault that can occur running the mmsetquota command. This issue would only happen when GPFS overwrite tracing is enabled on Linux.
* Fix a clone parent file restore failure. This issue would only happen when the clone files need to be restored and the clone parent files have "immutable" attribute in restoring snapsh* Fix rename issues with Recovery for Directories and files when the special block files inside a directory gets renamed.
* Before linux-2.6.38, the implementation of fallocate(2) on gpfs needs an open instance which has delay close semantics. This may cause a timing window after the open instance is closed. read(2) operation cannot see up to date mmap modification until the timing window is gone (Assume no msync(2) is called). This fix closes the time window.
* Fix output for mmgetstate with -Y flag so that it matches GPFS documentation.
* GPFS commands may fail when public key expired. This fix re-generated the key.
* Fix a possible segmentation fault in offline fsck during mutlipass dir scan. Multipass directory scan happens when there is insufficient memory to hold the fsck data structures required for the directory scan phase.
* Fix resync thread not getting killed by deadlock detection code if it takes a long time to complete.
* Fix a rare assert "isSGPanicked" in the Asynchronous Direct I/O code path.
* Fixed an assert on when DBGASSERT is enabled and dm_get_dmattr is called on files in .snapshot directory.
* Fix a ENOMEM failure in tslsenclslot. The error condition occurs under heavy loads only. This fix is required for all platforms.
* Fix an assert that can occur on a GW node if the dr primary conversion is stopped, and the fileset is unlinked and AFM is disabled.
* Fix errors that occur after a file set manager is moved from nodeA to nodeB and new filesets are created. The errors occur because nodeA can not see the new filesets.
* Fix a problem that causes ACL inconsistency on the sgmgr node. This may cause an FSSTRUCT or an assert.
* Fix code to prevent a potential GPFS daemon assert during file system restripe.
* Fix an alloc segment steal problem that can lead to a "More than 22 minutes searching for a free buffer" assert.
* This fix reduces the ThreadStateMutex hold time in deadlock detection and waiter related functions to make GPFS run more smoothly.
* mmexportfs: Fix an issue where the file system exported but the output data file is missing.
* Fix a minor problem by not sending a delete snapshot request to secondary.
* Fix a problem with PCACHE_REMOVES on renamed NFS files that contain special characters in their name returning E_BUSY from home.
* Fix the long waiter "InodeDeleteThread, 'waiting for XW lock'" caused by a self-deadlock issue, by correcting the release orders of GPFS file lock and Linux inode. This issue would see under Linux system only.
* Fix an issue in AFM environment where prefetch recovery could deadlock with already running management programs (ex. create snapshot).
* Add an undocumented config parameter dataCollectionPendingDelay which controls how long we try to preserve the state of the world while collecting expel debug data. The code path is also optimized in general.
* Fix allocation code to prevent an infinite loop during restripe that could cause long waiters. This could occur when disks are full in multiple failure groups.
* Add a "flushOnClose" config parameter to force dirty data to be flushed to disk whenever a file is closed.
* Fix revalidation issue with resync over gpfs backend.
* Fix a problem when creating iSCSI volumes with more than 16 bytes name.
* Fix a problem in which Ganesha fails to start when netgroup name starts with numeric character.
* Fix a problem in which mmbackup --rebuild fails. This occurs on AIX only.
* Fix an assert that can occur after a nfs error getting sparse info of a file using control interface.
* Fix the "Node name is not valid" failure for mmrestorefs command.
* Fix the way errorcodes are being decoded and converted to AFM range from GPFS or system range.
* Fix an issue in AFM DR environment where DR failback fails to restore from the RPO snapshot. This happens when fileset restore tries to punch hole on a file and file was sparse in RPO snapshot.
* Fix a bug where mmchcluster may fail when run on a non-configure server to disable CCR.
* This fix will not show old open file or directory names that are renamed while they are open.
* Fix a problem in which AFM does not honor async delay if softQMem threshold is set as 0. It plays ops to home immediately.
* This update addresses the following APARs: IV85089 IV87625 IV87626 IV87627 IV87628 IV87629 IV87630 IV87631 IV87632 IV87633 IV87634 IV87635 IV87636 IV87637 IV87638 IV87639 IV87645 IV87809 IV87813 IV88298 IV88855.