- 27 Sep, 2016 19 commits
-
-
Trond Myklebust authored
If the server tells us our stateid has expired, then handle that as if it was revoked. Signed-off-by:
Trond Myklebust <trond.myklebust@primarydata.com> Tested-by:
Oleg Drokin <green@linuxhacker.ru> Signed-off-by:
Anna Schumaker <Anna.Schumaker@Netapp.com>
-
Trond Myklebust authored
If the server tells us our stateid has expired, then handle that as if it was revoked. Signed-off-by:
-
Trond Myklebust authored
Modify the helper nfs_inode_find_state_and_recover() so that it can check all open/lock/delegation state trackers on that inode for whether or not they need are affected by a revoked stateid error. Signed-off-by:
-
Trond Myklebust authored
This fixes a potential infinite loop in nfs_reap_expired_delegations. Signed-off-by:
-
Trond Myklebust authored
If a server returns NFS4ERR_ADMIN_REVOKED, NFS4ERR_DELEG_REVOKED or NFS4ERR_EXPIRED on a call to close, open_downgrade, delegreturn, or locku, we should call FREE_STATEID before attempting to recover. Signed-off-by:
-
Trond Myklebust authored
Nothing should need to be serialised with FREE_STATEID on the client, so let's make the RPC call always asynchronous. Also constify the stateid argument. Signed-off-by:
-
Trond Myklebust authored
Right now, we're only running TEST/FREE_STATEID on the locks if the open stateid recovery succeeds. The protocol requires us to always do so. The fix would be to move the call to TEST/FREE_STATEID and do it before we attempt open recovery. Signed-off-by:
-
Trond Myklebust authored
In some cases (e.g. when the SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED sequence flag is set) we may already know that the stateid was revoked and that the only valid operation we can call is FREE_STATEID. In those cases, allow the stateid to carry the information in the type field, so that we skip the redundant call to TEST_STATEID. Signed-off-by:
-
Trond Myklebust authored
Ensure we don't spam the server with test_stateid() calls for delegations that have already been checked. Signed-off-by:
-
Trond Myklebust authored
Ensure that if the server reboots while we're testing and recovering from revoked delegations, we exit to allow the state manager to handle matters. Signed-off-by:
-
Trond Myklebust authored
According to RFC5661, if any of the SEQUENCE status bits SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED, SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED, SEQ4_STATUS_ADMIN_STATE_REVOKED, or SEQ4_STATUS_RECALLABLE_STATE_REVOKED are set, then we need to use TEST_STATEID to figure out which stateids have been revoked, so we can acknowledge the loss of state using FREE_STATEID. While we already do this for open and lock state, we have not been doing so for all the delegations. Signed-off-by:
-
Trond Myklebust authored
Allow the callers of nfs_remove_bad_delegation() to specify the stateid that needs to be marked as bad. Signed-off-by:
-
Trond Myklebust authored
In NFSv4.1 and newer, if the server decides to revoke some or all of the protocol state, the client is required to iterate through all the stateids that it holds and call TEST_STATEID to determine which stateids still correspond to valid state, and then call FREE_STATEID on the others. Signed-off-by:
-
Trond Myklebust authored
If the server crashes while we're testing stateids for validity, then we want to initiate session recovery. Usually, we will be calling from a state manager thread, though, so we don't really want to wait. Signed-off-by:
-
Trond Myklebust authored
If the delegation has been marked as revoked, we don't have to test it, because we should already have called FREE_STATEID on it. Signed-off-by:
-
Trond Myklebust authored
We must not allow the use of delegations that have been revoked or are being returned. Signed-off-by:
-
Trond Myklebust authored
If the delegation is revoked, then it can't be used for caching. Fixes: 869f9dfa ("NFSv4: Fix races between nfs_remove_bad_delegation()...") Signed-off-by:
-
Trond Myklebust authored
Due to inode number reuse in filesystems, we can end up corrupting the inode on our client if we apply the file attributes without ensuring that the filehandle matches. Typical symptoms include spurious "mode changed" reports in the syslog. We still do want to ensure that we don't invalidate the dentry if the inode number matches, but we don't have a filehandle. Fixes: fa923369 ("NFS: Don't require a filehandle to refresh...") Signed-off-by:
-
Trond Myklebust authored
As described in RFC5661, section 18.46, some of the status flags exist in order to tell the client when it needs to acknowledge the existence of revoked state on the server and/or to recover state. Those flags will then remain set until the recovery procedure is done. In order to avoid looping, the client therefore needs to ignore those particular flags while recovering. Signed-off-by:
-
- 23 Sep, 2016 3 commits
-
-
Daniel Wagner authored
There is only one waiter for the completion, therefore there is no need to use complete_all(). Let's make that clear by using complete() instead of complete_all(). The usage pattern of the completion is: waiter context waker context frwr_op_unmap_sync() reinit_completion() ib_post_send() wait_for_completion() frwr_wc_localinv_wake() complete() Signed-off-by:
Daniel Wagner <daniel.wagner@bmw-carit.de> Cc: Anna Schumaker <Anna.Schumaker@Netapp.com> Cc: Trond Myklebust <trond.myklebust@primarydata.com> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: linux-nfs@vger.kernel.org Cc: netdev@vger.kernel.org Signed-off-by:
-
Daniel Wagner authored
There is only one waiter for the completion, therefore there is no need to use complete_all(). Let's make that clear by using complete() instead of complete_all(). The generic caching code from sunrpc is calling revisit() only once. The usage pattern of the completion is: waiter context waker context do_cache_lookup_wait() nfs_cache_defer_req_alloc() init_completion() do_cache_lookup() nfs_cache_wait_for_upcall() wait_for_completion_timeout() nfs_dns_cache_revisit() complete() nfs_cache_defer_req_put() Signed-off-by: -
Daniel Wagner authored
There is only one waiter for the completion, therefore there is no need to use complete_all(). Let's make that clear by using complete() instead of complete_all(). nfs_file_direct_write() or nfs_file_direct_read() allocated a request object via nfs_direct_req_alloc(), which initializes the completion. The request object then is freed later in the exit path. Between the initialization and the release either nfs_direct_write_schedule_iovec() resp nfs_direct_read_schedule_iovec() are called which will asynchronously process the request. The calling function waits via nfs_direct_wait() till the async work has been done. Thus there is only one waiter on the completion. nfs_direct_pgio_init() and nfs_direct_read_completion() are passed via function pointers to nfs pageio. The first function does a ref counting (get_dreq() and put_dreq()) which ensures that nfs_direct_read_completion() and nfs_direct_read_schedule_iovec() only call the completion path once. The usage pattern of the completion is: waiter context waker context nfs_file_direct_write() dreq = nfs_direct_req_alloc() init_completion() nfs_direct_write_schedule_iovec() nfs_direct_wait() wait_for_completion_killable() nfs_direct_write_schedule_work() nfs_direct_complete() complete() nfs_file_direct_read() dreq = nfs_direct_req_all() init_completion() nfs_direct_read_schedule_iovec() nfs_direct_wait() wait_for_completion_killable() nfs_direct_read_schedule_iovec() nfs_direct_complete() complete() nfs_direct_read_completion() nfs_direct_complete() complete() Signed-off-by:
-
- 22 Sep, 2016 12 commits
-
-
Trond Myklebust authored
Use xdr->nwords to tell us how much buffer remains. Signed-off-by:
-
Trond Myklebust authored
When we copy the first part of the data, we need to ensure that value of xdr->nwords is updated as well. Do so by calling __xdr_inline_decode() Signed-off-by:
-
Trond Myklebust authored
Before we try to stash it in the dcache, we need to at least check that the filename passed to us by the server is non-empty and doesn't contain any illegal '\0' or '/' characters. Signed-off-by:
-
Jeff Layton authored
Add a waitqueue head to the client structure. Have clients set a wait on that queue prior to requesting a lock from the server. If the lock is blocked, then we can use that to wait for wakeups. Note that we do need to do this "manually" since we need to set the wait on the waitqueue prior to requesting the lock, but requesting a lock can involve activities that can block. However, only do that for NFSv4.1 locks, either by compiling out all of the waitqueue handling when CONFIG_NFS_V4_1 is disabled, or skipping all of it at runtime if we're dealing with v4.0, or v4.1 servers that don't send lock callbacks. Note too that even when we expect to get a lock callback, RFC5661 section 20.11.4 is pretty clear that we still need to poll for them, so we do still sleep on a timeout. We do however always poll at the longest interval in that case. Signed-off-by:
Jeff Layton <jlayton@redhat.com> [Anna: nfs4_retry_setlk() "status" should default to -ERESTARTSYS] Signed-off-by:
-
Jeff Layton authored
This also consolidates the waiting logic into a single function, instead of having it spread across two like it is now. Signed-off-by:
-
Jeff Layton authored
We need to have this info set up before adding the waiter to the waitqueue, so move this out of the _nfs4_proc_setlk and into the caller. That's more efficient anyway since we don't need to do this more than once if we end up waiting on the lock. Signed-off-by:
-
Jeff Layton authored
For now, the callback doesn't do anything. Support for that will be added in later patches. Signed-off-by:
-
Jeff Layton authored
We want to handle the two cases differently, such that we poll more aggressively when we don't expect a callback. Signed-off-by:
-
Jeff Layton authored
As defined in RFC 5661, section 18.16. Signed-off-by:
-
Jeff Layton authored
We actually want to use TASK_INTERRUPTIBLE sleeps when we're in the process of polling for a NFSv4 lock. If there is a signal pending when the task wakes up, then we'll be returning an error anyway. So, we might as well wake up immediately for non-fatal signals as well. That allows us to return to userland more quickly in that case, but won't change the error that userland sees. Also, there is no need to use the *_unsafe sleep variants here, as no vfs-layer locks should be held at this point. Signed-off-by:
-
Jeff Layton authored
Signed-off-by:
-
Jeff Layton authored
Since it gets passed through to xdr_inline_decode, we might as well have read_buf expect what it expects -- a size_t. Signed-off-by:
-
- 20 Sep, 2016 1 commit
-
-
Chao Yu authored
It will be more clean to use CONFIG_MIGRATION to cover nfs' private .migratepage in nfs_file_aops like we do in other part of nfs operations. Signed-off-by:
Chao Yu <yuchao0@huawei.com> Signed-off-by:
-
- 19 Sep, 2016 5 commits
-
-
David Vrabel authored
Write space becoming available may race with putting the task to sleep in xprt_wait_for_buffer_space(). The existing mechanism to avoid the race does not work. This (edited) partial trace illustrates the problem: [1] rpc_task_run_action: task:43546@5 ... action=call_transmit [2] xs_write_space <-xs_tcp_write_space [3] xprt_write_space <-xs_write_space [4] rpc_task_sleep: task:43546@5 ... [5] xs_write_space <-xs_tcp_write_space [1] Task 43546 runs but is out of write space. [2] Space becomes available, xs_write_space() clears the SOCKWQ_ASYNC_NOSPACE bit. [3] xprt_write_space() attemts to wake xprt->snd_task (== 43546), but this has not yet been queued and the wake up is lost. [4] xs_nospace() is called which calls xprt_wait_for_buffer_space() which queues task 43546. [5] The call to sk->sk_write_space() at the end of xs_nospace() (which is supposed to handle the above race) does not call xprt_write_space() as the SOCKWQ_ASYNC_NOSPACE bit is clear and thus the task is not woken. Fix the race by resetting the SOCKWQ_ASYNC_NOSPACE bit in xs_nospace() so the second call to sk->sk_write_space() calls xprt_write_space(). Suggested-by:Trond Myklebust <trondmy@primarydata.com> Signed-off-by:
David Vrabel <david.vrabel@citrix.com> cc: stable@vger.kernel.org # 4.4 Signed-off-by:
-
Jeff Layton authored
Currently, the layout driver selection code always chooses the first one from the list. That's not really ideal however, as the server can send the list of layout types in any order that it likes. It's up to the client to select the best one for its needs. This patch adds an ordered list of preferred driver types and has the selection code sort the list of available layout drivers according to it. Any unrecognized layout type is sorted to the end of the list. For now, the order of preference is hardcoded, but it should be possible to make this configurable in the future. Signed-off-by:
J. Bruce Fields <bfields@fieldses.org> Signed-off-by:
-
Chuck Lever authored
Clean up: the extra layer of indirection doesn't add value. Signed-off-by:
Chuck Lever <chuck.lever@oracle.com> Signed-off-by:
-
Chuck Lever authored
Clean up: When converting xprtrdma to use the new CQ API, I missed a spot. The naming convention elsewhere is: {svc_rdma,rpcrdma}_wc_{operation} Signed-off-by: -
Chuck Lever authored
Tie frwr debugging messages together by always reporting the address of the frwr. Signed-off-by:
-
