DLPX-82827 Fix for Solaris NFSv4 client mounts

[don-brady]

Problem:

We are unable to mount timeflows on Solaris targets when using NFSv4.

Solaris client mounts use a compound operation which contain 5 operations per path element in the mount path. So with a path like "/domain0/group-2/oracle_db_container-6/oracle_timeflow-7/datafile", the op count reaches 20 and exceeds the restriction of NFSD_MAX_OPS_PER_COMPOUND = 16.

Solution:

We bump NFSD_MAX_OPS_PER_COMPOUND to 40 to allow us to use both NFS v4.0 and v4.1 from a Solaris client.

Testing

1. Manually tested with a five element path for both NFSv4.0 and NFSv4.1
2. Provisioned an Oracle VDB on a Solaris target and confirmed that it was using a NFSv4 mount:
   NFS Protocol
   NFS Version 4
   NFS Reason
   Default
   Environment
   djb-sol11u4-ora19900-tgt

And as seen on the client:

oracle@djb-sol11u4-ora19900-tgt:~$ nfsstat -m
/mnt/provision/VDBOMSRB27FBB_NM8 from 10.110.216.69:/domain0/group-2/oracle_db_container-6/oracle_timeflow-7
 Flags:         vers=4,mvers=1,proto=tcp,sec=sys,hard,nointr,link,symlink,forcedirectio,rsize=1048576,wsize=1048576,retrans=5,timeo=600
 Attr cache:    acregmin=3,acregmax=60,acdirmin=30,acdirmax=60

/mnt/provision/VDBOMSRB27FBB_NM8/datafile from 10.110.216.69:/domain0/group-2/oracle_db_container-6/oracle_timeflow-7/datafile
 Flags:         vers=4,mvers=1,proto=tcp,sec=sys,hard,nointr,link,symlink,forcedirectio,rsize=1048576,wsize=1048576,retrans=5,timeo=600
 Attr cache:    acregmin=3,acregmax=60,acdirmin=30,acdirmax=60

/mnt/provision/VDBOMSRB27FBB_NM8/archive from 10.110.216.69:/domain0/group-2/oracle_db_container-6/oracle_timeflow-7/archive
 Flags:         vers=4,mvers=1,proto=tcp,sec=sys,hard,nointr,link,symlink,forcedirectio,rsize=1048576,wsize=1048576,retrans=5,timeo=600
 Attr cache:    acregmin=3,acregmax=60,acdirmin=30,acdirmax=60

/mnt/provision/VDBOMSRB27FBB_NM8/external from 10.110.216.69:/domain0/group-2/oracle_db_container-6/oracle_timeflow-7/external
 Flags:         vers=4,mvers=1,proto=tcp,sec=sys,hard,nointr,link,symlink,forcedirectio,rsize=1048576,wsize=1048576,retrans=5,timeo=600
 Attr cache:    acregmin=3,acregmax=60,acdirmin=30,acdirmax=60

/mnt/provision/VDBOMSRB27FBB_NM8/temp from 10.110.216.69:/domain0/group-2/oracle_db_container-6/oracle_timeflow-7/temp
 Flags:         vers=4,mvers=1,proto=tcp,sec=sys,hard,nointr,link,symlink,forcedirectio,rsize=1048576,wsize=1048576,retrans=5,timeo=600
 Attr cache:    acregmin=3,acregmax=60,acdirmin=30,acdirmax=60

/mnt/provision/VDBOMSRB27FBB_NM8/source-archive from 10.110.216.69:/domain0/group-2/oracle_db_container-6/oracle_timeflow-7/source-archive
 Flags:         vers=4,mvers=1,proto=tcp,sec=sys,hard,nointr,link,symlink,forcedirectio,rsize=1048576,wsize=1048576,retrans=5,timeo=600
 Attr cache:    acregmin=3,acregmax=60,acdirmin=30,acdirmax=60

Implementation:

We bumped NFSD_MAX_OPS_PER_COMPOUND to 40 operations, which allows for a 10 element path (5 levels below the datafile mount).

Deployment Plan:

This kernel change should land first followed by an app-gate change to remove the Solaris guardrail that was forcing NFSv3 mounts for Solaris.

Future work:

An upstream bug was filed and we will continue to monitor the progress of a potentially different solution to this problem
https://bugzilla.kernel.org/show_bug.cgi?id=216383