if any of (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED | SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED | SEQ4_STATUS_ADMIN_STATE_REVOKED) are set in the session flags returned by SEQUENCE:
* enter client recovery mode
* determine which state was lost with TEST_STATEID (consider all delegations, opens, locks, and layouts)
* send FREE_STATEID for each stateid revoked
* recall all layouts and forget devices (required by 12.7.2: Dealing with Lease Expiration on the Client)
* call recover_delegation(), recover_open() or recover_locks() to reclaim each lock
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
in the case of ds writes returning FILE_SYNC, we don't need to send a COMMIT or LAYOUTCOMMIT to the mds. COMMIT and LAYOUTCOMMIT, however, are the places where we do GETATTR(size) to update the attribute cache. so we must add a separate call to GETATTR to accomplish this after ds writes return FILE_SYNC
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>
avoid unnecessary calls to SystemTimeToFileTime() in get_file_epoch() by hardcoding the FILETIME value of jan_1_1970
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
on CB_GETATTR, search for a delegation with the given filehandle. if found, use its fileid to get its cached attributes. when encoding the response, only include change and size attributes
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
the out_err_deleg: case of nfs41_name_cache_insert() was forgetting to increment cache->delegations, leading to a decrement past 0. once that happens, the cache stops accepting delegations because it's comparing cache->max_delegations against UINT_MAX
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
iozone tests were failing intermittently against emc-2 due to out-of-order LAYOUTCOMMITs that both specified new_last_offset
nfs41_open_state now maintains a cached value of the last_offset, and avoids sending it with LAYOUTCOMMIT unless the new last_offset is greater than the cached offset. this cached value is initialized on open/delegation, and updated on setattr for size
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
sorry, earlier Casey, but the patch 'threading by io unit instead of stripe' from 6/20/2010 was nuts! with PNFS_THREAD_BY_SERVER disabled, we definitely -don't- want to create a separate thread for each io unit (each READ/WRITE request to a ds). we just want the one per stripe, as the intended alternative to PNFS_THREAD_BY_SERVER
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
cancel_lock() can't rely on checking upcall.status, because a handle_lock() success could be overwritten by upcall_marshall() or failure to allocate the upcall reply buffer. added new flag lock_upcall_args.acquired for this purpose
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
now that cancel_lock() and handle_unlock() depend on finding lock state in the open, the comment /* ignore errors from open_lock_add(); they just mean we won't be able to recover the lock after reboot */ no longer applies. allocate the lock state at the top of handle_lock(), so we can bail immediately on failure
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
removed unused variable 'grandparent'
factored out the call to name_cache_entry_update() from root/non-root paths
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
renamed name_cache_entry_refresh() to name_cache_entry_updated(). this function resets the entry's expiration timer. new function name_cache_entry_accessed() deals with moving the entry to the end of exp_entries to prevent eviction
name_cache_entry_accessed() now does the same for its parents as well. cache consistency of the parents is maintained because their expiration timers are unchanged. by 'refreshing' the parents of each name cache entry inserted, the eviction policy will always prefer child nodes (which are far less costly to evict)
Signed-off-by: Casey Bodley <cbodley@citi.umich.edu>
