[LTP] [PATCH] mremap06: fallocate is not supported on nfsv3

[samasth.norway.ananda@oracle.com]

On 4/10/24 6:33 AM, Petr Vorel wrote:
>> Hi!
>>> The solution would be to at least check for tst_fs_type(".") == TST_NFS_MAGIC.
> 
>> I would go for this at the moment, that is the easiest fix at the
>> moment.
> 
> So you mean we risk we lost fallocate support in NFSv4.2 for the simplicity. OK
> for me.
> 
>>> Perfect solution would IMHO be to check for NFS version (e.g. in /proc/mounts)
>>> and TCONF only on NFSv3.
> 
>> And is nfsv4 enough, wasn't it added in nfsv4.2? Do we need to parse
>> /proc/net/rpc/nfs to figure out if nfs fallocate() is supported?
> 
> Yes. First, I got confused 95d871f03cae ("nfsd: Add ALLOCATE support") [1] which is
> not mentioning NFS version, because it's for nfsd. But looking other commits,
> e.g. 95d871f03cae ("nfsd: Add ALLOCATE support") [2], 624bd5b7b683 ("nfs: Add
> DEALLOCATE support") [3] and the docs from merge commit in 3.19 [4] and
> kernelnewbies [5], it's clear: it's for NFSv4.2.
> 
> BTW is that whole fallocate() support? [4] mentions "NFSv4.2 client support for
> hole punching and preallocation.", commit [3]: "This patch adds support for
> using the NFS v4.2 operation DEALLOCATE to punch holes in a file.", commit
> b0cb9085239a ("nfsd: Add DEALLOCATE support") [6] uses FALLOC_FL_PUNCH_HOLE
> flag. All that IMHO means it's adding just FALLOC_FL_PUNCH_HOLE support into
> NFSv4.2. Thus other (basic) fallocate() might work in all NFS versions, right?
> 
> That would suggest when addin safe_fallocate() we should check for NFS
> (tst_fs_type(".") == TST_NFS_MAGIC) only when FALLOC_FL_PUNCH_HOLE is used.
> (I got from your answer we ignore missing support for older filesystems,
> we can always add the check once somebody reports invalid error).

Thanks for the detailed explanation of the issue. I will send out a v2 
patch which would check the file system using tst_fs_type(".") before 
printing TCONF.

I will also send a patch to update the man page to include the 
information about nfs.

Thanks,
Samasth.
> 
> Kind regards,
> Petr
> 
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=95d871f03cae6b49de040265cf88cbe2a16b9f05
> [2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=f4ac1674f5da420ef17896f0f222c5215ebcde80
> [3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=624bd5b7b683c978c6d5f4e9f6142cfb3470983d
> [4] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e20db597b6264de55ea6636fc79b1e4aaa89d129
> [5] https://kernelnewbies.org/Linux_3.19#NFSv4.2_support_for_hole_punching_and_preallocation
> [6] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b0cb9085239a20b7482ddd4839dd1d5476801dfa