-
Bug
-
Resolution: Done
-
Undefined
-
None
-
rhel-8.8.0.z
-
None
-
None
-
sst_filesystems
-
ssg_filesystems_storage_and_HA
-
None
-
False
-
-
None
-
Red Hat Enterprise Linux
-
None
-
None
-
None
-
None
What were you trying to do that didn't work?
After a disconnection, under some circumstances the client may not always reuse the source tcp port.
Please provide the package NVR for which bug is seen:
kernel-4.18.0-477.26.1.el8_8.x86_64
How reproducible:
easy
Steps to reproduce
on the nfs server:
# echo 4 > /proc/fs/nfsd/threads # echo 4 > /proc/fs/nfsd/max_connections add 5 IPs to the network interface # ip addr add dev bond0 192.168.122.101 # ip addr add dev bond0 192.168.122.102 # ip addr add dev bond0 192.168.122.103 # ip addr add dev bond0 192.168.122.104 # ip addr add dev bond0 192.168.122.105
nfs client 1
# mount server:/exports /mnt/tmp -overs=3 # echo foo > /mnt/tmp/testfile ; while [[ 42 ]] ; do mv -f /mnt/tmp/testfile /mnt/tmp/testfile2 || break ; mv -f /mnt/tmp/testfile2 /mnt/tmp/testfile || break ; echo -n . ; done
nfs client 2
# mkdir -p /mnt/tmp/mount{1,2,3,4,5}
# mount 192.168.122.101:/exports /mnt/tmp/mount1 -overs=3,noac
# mount 192.168.122.102:/exports /mnt/tmp/mount2 -overs=3,noac
# mount 192.168.122.103:/exports /mnt/tmp/mount3 -overs=3,noac
# mount 192.168.122.104:/exports /mnt/tmp/mount4 -overs=3,noac
# mount 192.168.122.105:/exports /mnt/tmp/mount5 -overs=3,noac
# while [[ 42 ]] ; do stat /mnt/tmp/mount* >/dev/null ; sleep 1 ; done
Expected results
The clients reuse the source port, so the DRC can reply correctly to the 'mv' (RENAME)
Actual results
The client doing the 'mv' changes source port when reconnecting, so a replayed RENAME is not matched in the DRC:
# echo foo > /mnt/vm23/testfile ; while [[ 42 ]] ; do mv -f /mnt/vm23/testfile /mnt/vm23/testfile2 || break ; mv -f /mnt/vm23/testfile2 /mnt/vm23/testfile || break ; echo -n . ; done ..........................................................................................mv: cannot stat '/mnt/vm23/testfile2': No such file or directory
