On Fri, 8 Mar 2019, David Howells wrote: > > $ mount | grep nfs4 > > nfs:/usr/local/src on /usr/local/src type nfs4 > > (ro,relatime,vers=4.2,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=192.168.56.139,local_lock=none,addr=192.168.0.115) > > > > ...so FS-Cache ("fsc") isn't even used here. > > Interesting. Can you do: > > cat /proc/fs/nfsfs/volumes
That seems to confirm the mount options, fsc is disabled: # cat /proc/fs/nfsfs/volumes NV SERVER PORT DEV FSID FSC v4 c0a80073 801 0:46 1cfd45bf1921474d:a795870ea80f5ff7 no > See the attached for a patch that helps with certain kinds of collision, > though I can't see that it should help with what you're seeing since the > RELINQUISHED flag isn't set on the old cookie (fl=222, but 0x10 isn't in > there). You can monitor the number of waits by looking in > /proc/fs/fscache/stats for the: > > Acquire: n=289166 nul=0 noc=0 ok=286331 nbf=2 oom=0 wrq=23748 Ah, the wrq= field gets only introduced by this patch. OK, I'll see if I can build a test kernel with that and will report back. Thanks for looking in to this, Christian. -- BOFH excuse #290: The CPU has shifted, and become decentralized.