A separate tracepoint can be left enabled all the time to capture
rare but important retransmission events. So for example:

kworker/u26:3-568   [009]   156.967933: xprt_retransmit:      task:44093@5 xid=0xa25dbc79 nfsv3 WRITE ntrans=2

Or, for example, enable all nfs and nfs4 tracepoints, and set up a
trigger to disable tracing when xprt_retransmit fires to capture
everything that leads up to it.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

I've hit some crashes that occur in the xprt_rdma_inject_disconnect
path. It appears that, for some provides, rdma_disconnect() can
take so long that the transport can disconnect and release its
hardware resources while rdma_disconnect() is still running,
resulting in a UAF in the provider.

The transport's fault injection method may depend on the stability
of transport data structures. That means it needs to be invoked
only from contexts that hold the transport write lock.

Fixes: 4a068258 ("SUNRPC: Transport fault injection")
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Currently the client ignores the value of the sr_eof of the SEEK
operation. According to the spec, if the server didn't find the
requested extent and reached the end of the file, the server
would return sr_eof=true. In case the request for DATA and no
data was found (ie in the middle of the hole), then the lseek
expects that ENXIO would be returned.

Fixes: 1c6dcbe5 ("NFS: Implement SEEK")
Signed-off-by: NOlga Kornievskaia <kolga@netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We (adam zabrocki, alexander matrosov, alexander tereshkin, maksym
bazalii) observed the check:

	if (fh->size > sizeof(struct nfs_fh))

should not use the size of the nfs_fh struct which includes an extra two
bytes from the size field.

struct nfs_fh {
	unsigned short         size;
	unsigned char          data[NFS_MAXFHSIZE];
}

but should determine the size from data[NFS_MAXFHSIZE] so the memcpy
will not write 2 bytes beyond destination.  The proposed fix is to
compare against the NFS_MAXFHSIZE directly, as is done elsewhere in fs
code base.

Fixes: d67ae825 ("pnfs/flexfiles: Add the FlexFile Layout Driver")
Signed-off-by: NNikola Livic <nlivic@gmail.com>
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the server returns a filehandle with an invalid length, then trace
that, and return an EREMOTEIO error.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We would like the ability to record other XDR errors, particularly
those that are due to server bugs.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

When the server returns a different operation than we expected, then
trace that.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

There are lots of attributes, and they are crowding out the bit space.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

NFS_INO_REVAL_FORCED is intended to tell us that the cache needs
revalidation despite the fact that we hold a delegation. We shouldn't
need to store it anymore, though.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the nlink changes, then so will the ctime.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If we're trying to update the inode because a previous update left the
cache in a partially unrevalidated state, then allow the update if the
change attrs match.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the NFSv4.2 server supports the 'change_attr_type' attribute, then
allow the client to optimise its attribute cache update strategy.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

The change_attr_type allows the server to provide a description of how
the change attribute will behave. This again will allow the client to
optimise its behaviour w.r.t. attribute revalidation.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

When the client is caching data and a write delegation is held, then the
server may send a CB_GETATTR to query the attributes. When this happens,
the client is supposed to bump the change attribute value that it
returns if it holds cached data.
However that process uses a value that is stored in the delegation. We
do not want to bump the change attribute held in the inode.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

At least two extra fields have been added to fsinfo since this was last
updated.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We should not be invalidating the access or acl caches in
nfs_check_inode_attributes(), since the point is we're unsure about
whether the contents of the struct nfs_fattr are fully up to date.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Commit 0b467264 ("NFS: Fix attribute revalidation") changed the way
we populate the 'invalid' attribute, and made the line that strips away
the NFS_INO_INVALID_ATTR bits redundant.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If there is an outstanding layoutcommit, then the list of attributes
whose values are expected to change is not the full set. So let's
be explicit about the full list.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

chown()/chgrp() and chmod() are separate operations, and in addition,
there are mode operations that are performed automatically by the
server. So let's track mode validity separately from the file ownership
validity.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Rename can cause us to revalidate the access cache, so lets track the
nlinks separately from the mode/uid/gid.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Don't remove flags from the set retrieved from the cache_validity.
We do want to retrieve all attributes that are listed as being
invalid, whether or not there is a delegation set.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

It is no longer necessary to preserve the NFS_INO_REVAL_PAGECACHE flag.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

When checking cache validity, be more specific than just 'we want to
check the page cache validity'. In almost all cases, we want to check
that change attribute, and possibly also the size.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Add an argument to nfs_revalidate_inode() to allow callers to specify
which attributes they need to check for validity.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

When we do a 'chown' or 'chgrp', the server will clear the suid/sgid
bits. Ensure that we mirror that in nfs_setattr_update_inode().
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If statx has valid attributes available that weren't asked for, then
return them and set the result mask appropriately.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the user doesn't set STATX_UID/GID/MODE, then don't care if they are
known to be stale. Ditto if we're not being asked for the file size.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Ensure that when the change attribute or the size change, we also
remember to revalidate the space used.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

When we're looking to revalidate the page cache, we should just ensure
that we mark the change attribute invalid.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We don't currently support STATX_BTIME, so don't advertise it in the
return values for nfs_getattr().
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Add missing enum definitions and missing entries for
nfs_show_cache_validity().
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We need to use unsigned long subtraction and then convert to signed in
order to deal correcly with C overflow rules.

Fixes: f5062003 ("NFS: Set an attribute barrier on all updates")
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Whether we're allocating or delallocating space, we should flush out the
pending writes in order to avoid races with attribute updates.

Fixes: 1e564d3d ("NFSv4.2: Fix a race in nfs42_proc_deallocate()")
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We can't use nfs4_fattr_bitmap as a bitmask, because it hasn't been
filtered to represent the attributes supported by the server. Instead,
let's revert to using server->cache_consistency_bitmask after adding in
the missing SPACE_USED attribute.

Fixes: 913eca1a ("NFS: Fallocate should use the nfs4_fattr_bitmap")
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

As currently set, the calls to nfs4_bitmask_adjust() will end up
overwriting the contents of the nfs_server cache_consistency_bitmask
field.
The intention here should be to modify a private copy of that mask in
the close/delegreturn/write arguments.

Fixes: 76bd5c01 ("NFSv4: make cache consistency bitmask dynamic")
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Ensure that we invalidate the fscache before we strip the
NFS_INO_INVALID_DATA flag.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

The change attribute is always set by all NFS client versions so get rid
of the open-coded version.

Fixes: 3cc55f44 ("nfs: use change attribute for NFS re-exports")
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the server sends CB_ calls on a connection that is not associated
with the backchannel, refuse to process the call and shut down the
connection.  This avoids a NULL dereference crash in
xprt_complete_bc_request().  There's not much more we can do in this
situation unless we want to look into allowing all connections to be
associated with the fore and back channel.
Signed-off-by: NBenjamin Coddington <bcodding@redhat.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Mounting NFSv3 uses default timeout parameters specified by underlying
sunrpc transport, and mount options like 'timeo' and 'retrans', unlike
NFSv4, are not honored.

But sometimes we want to set non-default timeout value when mounting
NFSv3, so pass 'timeo' and 'retrans' to nfs_mount() and fill the
'timeout' field of struct rpc_create_args before creating RPC
connection. This is also consistent with NFSv4 behavior.

Note that this only sets the timeout value of rpc connection to mountd,
but the timeout of rpcbind connection should be set as well. A later
patch will fix the rpcbind part.
Signed-off-by: NEryu Guan <eguan@linux.alibaba.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>