Description of problem: Unmounting the filesystem fails if the NFS client has sent the blocking lock request to a file in the NFS exported filesystem. Stopping the nfs server does not also help. The server node needs to be rebooted inorder to unmount the filesystem Version-Release number of selected component (if applicable): nfs-utils-1.0.6-70.EL4 How reproducible: Always Steps to Reproduce: On Server 1. Mount the device(/dev/vgnfs/lvxfs) on the directory /NFS1. 2. Start nfs and nfslock. 3. Export the directory /NFS1. Need two client machines. 1. In one client, send lock request to a file in the NFS exported FS (which is a mounted filesystem) and hold lock. 2. In another client, send blocking lock request to the same file as above. Then, this request will be waiting for the lock to be released. On server 3. Unexport the NFS filesystem, Send SIGKILL to lockd thread and stop the NFS server (by shutting down nfs and nfslock) 4. Unmount the device Actual results: On server, unmounting the filesystem will fail umount: /NFS1: device is busy umount: /NFS1: device is busy To unmount the filesystem, the node needs to be rebooted. Expected results: After unexporting filessytems and releasing all the locks , umounting should happen without any problem. But this is not happening even after killing of nfsd Additional info: NFS server is on Red Hat Enterprise Linux AS release 4 (Nahant Update 4) and NFS clients are on RH4 U3 or RH4 U4. The code used for acquiring the blocking lock is attached. The program accepts the file to be locked as an argument.
(In reply to comment #0) Its a duplicate of bug with id "214172". When I clicked "Back" on my browser, it got registered once again in the database. Please ignore this and remove from the database if possible. > Description of problem: > Unmounting the filesystem fails if the NFS client has sent the blocking lock > request to a file in the NFS exported filesystem. Stopping the nfs server does > not also help. The server node needs to be rebooted inorder to unmount the > filesystem > Version-Release number of selected component (if applicable): > nfs-utils-1.0.6-70.EL4 > How reproducible: > Always > Steps to Reproduce: > On Server > 1. Mount the device(/dev/vgnfs/lvxfs) on the directory /NFS1. > 2. Start nfs and nfslock. > 3. Export the directory /NFS1. > Need two client machines. > 1. In one client, send lock request to a file in the NFS exported FS (which is > a mounted filesystem) and hold lock. > 2. In another client, send blocking lock request to the same file as above. > Then, this request will be waiting for the lock to be released. > On server > 3. Unexport the NFS filesystem, Send SIGKILL to lockd thread and stop the NFS > server (by shutting down nfs and nfslock) > 4. Unmount the device > > Actual results: > On server, unmounting the filesystem will fail > umount: /NFS1: device is busy > umount: /NFS1: device is busy > To unmount the filesystem, the node needs to be rebooted. > Expected results: > After unexporting filessytems and releasing all the locks , umounting should > happen without any problem. But this is not happening even after killing of nfsd > Additional info: > NFS server is on Red Hat Enterprise Linux AS release 4 (Nahant Update 4) and NFS > clients are on RH4 U3 or RH4 U4. > The code used for acquiring the blocking lock is attached. The program accepts > the file to be locked as an argument.
*** This bug has been marked as a duplicate of 214172 ***