when starting io, both pnfs_read() and pnfs_write() need a guarantee that their range is covered by layout segments. because we have to drop the lock for LAYOUTGET and GETDEVICEINFO, earlier layout segments may be recalled during this process. to avoid this, new function pnfs_layout_state_prepare() gets called repeatedly until it can verify under a single lock that 1) the entire desired range is covered with layouts and 2) each of these layouts has an associated device. whenever pnfs_layout_state_prepare() has to drop its lock for LAYOUTGET or GETDEVICEINFO, it returns PNFS_PENDING
on PNFS_SUCCESS, the caller knows that all segments in the range are valid and can dispatch io to those segments without worrying about recalls, because it still holds the pnfs_layout_state lock
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
on creation of a new superblock, construct a bitmap for the default attribute mask to be used for GETATTR and READDIR requests on that filesystem. mask out any unsupported attributes, and store the bitmap in the field nfs41_superblock.default_getattr
replaced function init_getattr_request() with nfs41_superblock_getattr_mask(), which returns a copy of superblock->default_getattr
removed the locking in nfs41_superblock_supported_attrs() and nfs41_superblock_supported_attrs_exclcreat(), as the supported_attrs and suppattr_exclcreat fields are read-only after the superblock is first initialized. also factored out their common code into a bitmap_intersect() function in util.h
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
when we are doing non-pnfs and writing to a netapp filer, even though we
write UNSTABLE it returns us FILE_SYNC4. since we were doing unstable
writes we were not sending getattrs, and since the data servers we returning
stable commits we didn't send commit. in doing so, we never update our
attribute cache and had a wrong file size.
fixes an issue introduced by commit 1f7e560a9a, "pnfs: fix for short write on ds error"
when a write is satisfied completely by pnfs, the upcall was returning out_len=0. the only time i noticed this was against python server, where cp in cygwin would succeed but print 'No space left on device'
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
small mds writes were using DATA_SYNC4 and were not followed by COMMITs, so there was no guarantee of an updated file size before returning success
use FILE_SYNC4 for these small writes, and send COMMITs for both UNSTABLE4 and DATA_SYNC4
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
we were previously only verifying that the server didn't reboot between WRITEs. COMMIT returns a verifier that needs to be checked as well
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
added 2011 year to the copyright line
added authors info to the license
added UofM license to libtirpc files that we modified
(but i probably missed some)
combined nfs41_lock_stateid_arg() into nfs41_open_stateid_arg(). if a delegation is present, use the delegation stateid before looking at lock/open stateids. if a delegation recall is in progress, wait on its condition variable before falling back to the open stateid
made nfs41_lock_stateid_arg() static to lock.c because of its special semantics; open_to_lock_owner4 for LOCK won't accept a delegation stateid, so nfs41_delegation_to_open() is called to convert it
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
every upcall (except few) pass session and open_state pointer, so
add that to marshal_nfs41_header() in the driver. remove passing
of session and open_state elsewhere in marshal functions.
in the deamon, upcall.c now reads and stores pointers to session
and open_state in nfs41_upcall datastructure instead of having
each individual upcall store their own pointers. setattrl
and readdir args keeping pointer because the rest of the code
uses them a lot.
in upcall_parse() up refcounts on session and open_state if
valid handles were passed in. down refcounts upcall_cleanup() as
before. but need to be careful with count value for mount and open
upcalls. we need to take an extra ref because upcall_cleanup() now
will always decrement it.
nfs41_lock_stateid_arg() is now called only once in handle_read()/handle_write(), and pnfs_read()/pnfs_write() no longer depend on nfs41_open_state
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
operations that require a stateid now take stateid_arg for recovery information. these operations include close, setattr, lock/unlock, layoutget, and read/write (including pnfs)
nfs41_open_stateid_arg() locks nfs41_open_state and copies its stateid into a stateid_arg
nfs41_lock_stateid_arg() locks nfs41_open_state.last_lock and copies its stateid into a stateid_arg; if there is no lock state, it falls back to nfs41_open_stateid_arg()
pnfs_read/write() now take nfs41_open_state so they can generate stateid_args
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
very similar to the issue with nfs41_open_state, an abandoned upcall could outlive its mount. to prevent their nfs41_root from being freed, upcalls need to hold a reference until they're finished. this also keeps all of its clients/sessions/rpc connections alive
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
added call to upcall_cleanup() after both upcall_marshall() and upcall_cancel()
individual upcall operations define their nfs41_upcall_op structs locally, instead of putting tons of function prototypes in upcall.c
made the upcall_marshall() function optional; most marshall functions are noops
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
when open parsing fails, we were still returning upcall.status==NO_ERROR, so the driver assumed the open succeeded. other operations then sent up an open_state==NULL, and crashed the daemon. when upcall_parse() returns an error, set upcall.status to notify the driver
upcall_parse() prints a 'parsing of upcall <name> failed with <error>.' message on failure, so i removed redundant messages from the individual upcall parsing functions
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
