Skip to contentSkip to conversation

Adding mount option to kernel for NFS filesystem

DISCUSSION Gertjan Oude Lohuis 2 participants 2 messages Apr 29, 2009 VFS

lore
Share & tools
lore
#0 lore raw

Hi guys,

For fun and learning purposes I'm trying to add an mountflag for NFS-mounts to the kernel and mount. The flag's semantics are identical to ac/noac: 'flag' is default but has no special meaning, 'noflag' triggers a certain action.

I have a small question about adding the flag to include/linux/nfs_mount.h. Since NFS_MOUNT_FLAGMASK is defined as 0xFFFF, and all bits up until 0xFFFF are occupied by existing NFS_MOUNT-flags, how could I add my own flag? Would it be ok to define NFS_MOUNT_FLAGMASK as 0xFFFFF and my own flag as 0x10000? Or will this have other side effects? It seems to be working though ;-).
How would this be done if a real flag should be added by kernel maintainers?

I guess another option would be to create a new entry in the nfs_mount_data struct and treat it as an option (rsize-, wsize-, timeo-alike), but I'd like to do keep it the same as other flags.

Please advice! ;-)
Thanks in advance for your time.

Sincerely,
Gertjan

#1 lore raw
Quoted 7 lines · Gertjan Oude Lohuis

Hi guys,

For fun and learning purposes I'm trying to add an mountflag for NFS-mounts to the kernel and mount. The flag's semantics are identical to ac/noac: 'flag' is default but has no special meaning, 'noflag' triggers a certain action.

Quoted 8 lines

I have a small question about adding the flag to include/linux/nfs_mount.h. Since NFS_MOUNT_FLAGMASK is defined as 0xFFFF, and all bits up until 0xFFFF are occupied by existing NFS_MOUNT-flags, how could I add my own flag?
Would it be ok to define NFS_MOUNT_FLAGMASK as 0xFFFFF and my own flag as 0x10000? Or will this have other side effects? It seems to be working though ;-).
How would this be done if a real flag should be added by kernel maintainers?

The policy is that we don't add any more flags or features to the legacy binary mount interface. NFS_MOUNT_FLAGMASK is there to enforce that rule, so changing it is not acceptable.

All new mount features should be added using the text-based mount interface (see nfs_parse_mount_options()), which is supported by recent versions of the mount.nfs program in the nfs-utils package.

Cheers
Trond