- 18 Oct, 2018 5 commits
-
-
git://git.linux-nfs.org/projects/anna/linux-nfsTrond Myklebust authored
NFS RDMA client updates for Linux 4.20 Stable bugfixes: - Reset credit grant properly after a disconnect Other bugfixes and cleanups: - xprt_release_rqst_cong is called outside of transport_lock - Create more MRs at a time and toss out old ones during recovery - Various improvements to the RDMA connection and disconnection code: - Improve naming of trace events, functions, and variables - Add documenting comments - Fix metrics and stats reporting - Fix a tracepoint sparse warning Signed-off-by:Trond Myklebust <trond.myklebust@hammerspace.com>
-
J. Bruce Fields authored
Commits ffb6ca33 and e08ea3a9 prevent setting xprt_min_resvport greater than xprt_max_resvport, but may also break simple code that sets one parameter then the other, if the new range does not overlap the old. Also it looks racy to me, unless there's some serialization I'm not seeing. Granted it would probably require malicious privileged processes (unless there's a chance these might eventually be settable in unprivileged containers), but still it seems better not to let userspace panic the kernel. Simpler seems to be to allow setting the parameters to whatever you want but interpret xprt_min_resvport > xprt_max_resvport as the empty range. Fixes: ffb6ca33 "sunrpc: Prevent resvport min/max inversion..." Fixes: e08ea3a9 "sunrpc: Prevent rexvport min/max inversion..." Signed-off-by:
J. Bruce Fields <bfields@redhat.com> Signed-off-by:
-
Benjamin Coddington authored
We don't need to call this in the direct, read, or pnfs resend paths and the only other caller is the write path in nfs_page_async_flush() which already checks and sets the pg_error on the context. Signed-off-by:
Benjamin Coddington <bcodding@redhat.com> Signed-off-by:
-
Benjamin Coddington authored
We must check pg_error and call error_cleanup after any call to pg_doio. Currently, we are skipping the unlock of a page if we encounter an error in nfs_pageio_complete() before handing off the work to the RPC layer. Signed-off-by:
-
Trond Myklebust authored
Signed-off-by:
-
- 05 Oct, 2018 2 commits
-
-
Olga Kornievskaia authored
Running "./nfstest_delegation --runtest recall26" uncovers that client doesn't recover the lock when we have an appending open, where the initial open got a write delegation. Instead of checking for the passed in open context against the file lock's open context. Check that the state is the same. Signed-off-by:
Olga Kornievskaia <kolga@netapp.com> Signed-off-by:
-
Arnd Bergmann authored
The newly introduced gss_seq_send64_fetch_and_inc() fails to build on 32-bit architectures: net/sunrpc/auth_gss/gss_krb5_seal.c:144:14: note: in expansion of macro 'cmpxchg' seq_send = cmpxchg(&ctx->seq_send64, old, old + 1); ^~~~~~~ arch/x86/include/asm/cmpxchg.h:128:3: error: call to '__cmpxchg_wrong_size' declared with attribute error: Bad argument size for cmpxchg __cmpxchg_wrong_size(); \ As the message tells us, cmpxchg() cannot be used on 64-bit arguments, that's what cmpxchg64() does. Fixes: 571ed1fd ("SUNRPC: Replace krb5_seq_lock with a lockless scheme") Signed-off-by:Arnd Bergmann <arnd@arndb.de> Signed-off-by:
-
- 03 Oct, 2018 10 commits
-
-
Chuck Lever authored
linux/include/trace/events/rpcrdma.h:501:1: warning: expression using sizeof bool linux/include/trace/events/rpcrdma.h:501:1: warning: odd constant _Bool cast (ffffffffffffffff becomes 1) Fixes: ab03eff5 ("xprtrdma: Add trace points in RPC Call ... ") Signed-off-by:
Chuck Lever <chuck.lever@oracle.com> Signed-off-by:
Anna Schumaker <Anna.Schumaker@Netapp.com>
-
Chuck Lever authored
Clean up: Use the appropriate C macro instead of open-coding container_of() . Signed-off-by:
-
Chuck Lever authored
Clean up: fill in or update documenting comments for transport switch entry points. For xprt_rdma_allocate: The first paragraph is no longer true since commit 5a6d1db4 ("SUNRPC: Add a transport-specific private field in rpc_rqst"). The second paragraph is no longer true since commit 54cbd6b0 ("xprtrdma: Delay DMA mapping Send and Receive buffers"). Signed-off-by:
-
Chuck Lever authored
To show that a caller did attempt to allocate and post more Receive buffers, the trace point in rpcrdma_post_recvs() should report when rpcrdma_post_recvs() was invoked but no new Receive buffers were posted. Signed-off-by:
-
Chuck Lever authored
Clean up: rb_flags might be used for other things besides RPCRDMA_BUF_F_EMPTY_SCQ, so initialize it in a generic spot instead of in a send-completion-queue-related helper. Signed-off-by:
-
Chuck Lever authored
Clean up: This code was copied from xprtsock.c and backchannel_rqst.c. For rpcrdma, the backchannel server runs exclusively in process context, thus disabling bottom-halves is unnecessary. Signed-off-by:
-
Chuck Lever authored
Clean up: Replace the hashed memory address of the target rpcrdma_ep with the server's IP address and port. The server address is more useful in an administrative error message. Signed-off-by:
-
Chuck Lever authored
Clean up: Use a function name that is consistent with the RDMA core API and with other consumers. Because this is a function that is invoked from outside the rpcrdma.ko module, add an appropriate documenting comment. Signed-off-by:
-
Chuck Lever authored
Currently, when a connection is established, rpcrdma_conn_upcall invokes rpcrdma_conn_func and then wake_up_all(&ep->rep_connect_wait). The former wakes waiting RPCs, but the connect worker is not done yet, and that leads to races, double wakes, and difficulty understanding how this logic is supposed to work. Instead, collect all the "connection established" logic in the connect worker (xprt_rdma_connect_worker). A disconnect worker is retained to handle provider upcalls safely. Fixes: 254f91e2 ("xprtrdma: RPC/RDMA must invoke ... ") Signed-off-by:
-
Chuck Lever authored
Clean up: Eliminate the FALLTHROUGH into the default arm to make the switch easier to understand. Also, as long as I'm here, do not display the memory address of the target rpcrdma_ep. A hashed memory address is of marginal use here. Signed-off-by:
-
- 02 Oct, 2018 10 commits
-
-
Chuck Lever authored
Clean up. Since commit 173b8f49 ("xprtrdma: Demote "connect" log messages") there has been no need to initialize connstat to zero. In fact, in this code path there's now no reason not to set rep_connected directly. Signed-off-by:
-
Chuck Lever authored
Clean up: The convention throughout other parts of xprtrdma is to name variables of type struct rpcrdma_xprt "r_xprt", not "xprt". This convention enables the use of the name "xprt" for a "struct rpc_xprt" type variable, as in other parts of the RPC client. Signed-off-by:
-
Chuck Lever authored
Clean up: Use a function name that is consistent with the RDMA core API and with other consumers. Because this is a function that is invoked from outside the rpcrdma.ko module, add an appropriate documenting comment. Signed-off-by:
-
Chuck Lever authored
The way connection-oriented transports report connect_time is wrong: it's supposed to be in seconds, not in jiffies. Signed-off-by:
-
Chuck Lever authored
For TCP, the logic in xprt_connect_status is currently never invoked to record a successful connection. Commit 2a491991 ("SUNRPC: Return EAGAIN instead of ENOTCONN when waking up xprt->pending") changed the way TCP xprt's are awoken after a connect succeeds. Instead, change connection-oriented transports to bump connect_count and compute connect_time the moment that XPRT_CONNECTED is set. Signed-off-by:
-
Chuck Lever authored
Clean up the names of trace events related to MRs so that it's easy to enable these with a glob. Signed-off-by:
-
Chuck Lever authored
When a memory operation fails, the MR's driver state might not match its hardware state. The only reliable recourse is to dereg the MR. This is done in ->ro_recover_mr, which then attempts to allocate a fresh MR to replace the released MR. Since commit e2ac236c ("xprtrdma: Allocate MRs on demand"), xprtrdma dynamically allocates MRs. It can add more MRs whenever they are needed. That makes it possible to simply release an MR when a memory operation fails, instead of "recovering" it. It will automatically be replaced by the on-demand MR allocator. This commit is a little larger than I wanted, but it replaces ->ro_recover_mr, rb_recovery_lock, rb_recovery_worker, and the rb_stale_mrs list with a generic work queue. Since MRs are no longer orphaned, the mrs_orphaned metric is no longer used. Signed-off-by:
-
Chuck Lever authored
Some devices require more than 3 MRs to build a single 1MB I/O. Ensure that rpcrdma_mrs_create() will add enough MRs to build that I/O. In a subsequent patch I'm changing the MR recovery logic to just toss out the MRs. In that case it's possible for ->send_request to loop acquiring some MRs, not getting enough, getting called again, recycling the previous MRs, then not getting enough, lather rinse repeat. Thus first we need to ensure enough MRs are created to prevent that loop. I'm "reusing" ia->ri_max_segs. All of its accessors seem to want the maximum number of data segments plus two, so I'm going to bake that into the initial calculation. Signed-off-by:
-
Chuck Lever authored
On a fresh connection, an RPC/RDMA client is supposed to send only one RPC Call until it gets a credit grant in the first RPC Reply from the server [RFC 8166, Section 3.3.3]. There is a bug in the Linux client's credit accounting mechanism introduced by commit e7ce710a ("xprtrdma: Avoid deadlock when credit window is reset"). On connect, it simply dumps all pending RPC Calls onto the new connection. Servers have been tolerant of this bad behavior. Currently no server implementation ever changes its credit grant over reconnects, and servers always repost enough Receives before connections are fully established. To correct this issue, ensure that the client resets both the credit grant _and_ the congestion window when handling a reconnect. Fixes: e7ce710a ("xprtrdma: Avoid deadlock when credit ... ") Signed-off-by:
-
Chuck Lever authored
Since commit ce7c252a ("SUNRPC: Add a separate spinlock to protect the RPC request receive list") the RPC/RDMA reply handler has been calling xprt_release_rqst_cong without holding xprt->transport_lock. I think the only way this call is ever made is if the credit grant increases and there are RPCs pending. Current server implementations do not change their credit grant during operation (except at connect time). Commit e7ce710a ("xprtrdma: Avoid deadlock when credit window is reset") added the ->release_rqst call because UDP invokes xprt_adjust_cwnd(), which calls __xprt_put_cong() after adjusting xprt->cwnd. Both xprt_release() and ->xprt_release_xprt already wake another task in this case, so it is safe to remove this call from the reply handler. Signed-off-by:
-
- 30 Sep, 2018 13 commits
-
-
Trond Myklebust authored
If we're revalidating an existing dentry in order to open a file, we need to ensure that we check the directory has not changed before we optimise away the lookup. Signed-off-by: -
Trond Myklebust authored
Refactor the code in nfs_lookup_revalidate() as a stepping stone towards optimising and fixing nfs4_lookup_revalidate(). Signed-off-by: -
Trond Myklebust authored
We need to ensure that inode and dentry revalidation occurs correctly on reopen of a file that is already open. Currently, we can end up not revalidating either in the case of NFSv4.0, due to the 'cached open' path. Let's fix that by ensuring that we only do cached open for the special cases of open recovery and delegation return. Reported-by:
Stan Hu <stanhu@gmail.com> Signed-off-by:
-
Trond Myklebust authored
Signed-off-by: -
Trond Myklebust authored
Use RCU protected lookups for discovering the supported mechanisms. Signed-off-by: -
Trond Myklebust authored
Module removal is RCU safe by design, so we really have no need to lock the auth_flavors[] array. Substitute a lockless scheme to add/remove entries in the array, and then use rcu. Signed-off-by: -
Trond Myklebust authored
Now that each struct nfs_pgio_header corresponds to one RPC call, we only have one writer to the struct nfs_pgio_header. Signed-off-by: -
Trond Myklebust authored
Save a few bytes by allowing the read/write specific fields of the structures to share storage. Signed-off-by: -
Trond Myklebust authored
When the server fails to return post-op attributes, the client's attempt to place read data directly in the page cache fails, and so we have to do an extra copy in order to realign the data with page borders. This patch attempts to detect servers that don't return post-op attributes on read (e.g. for pNFS) and adjusts the placement calculation accordingly. Signed-off-by: -
Anna Schumaker authored
The convention in the rest of the code is to have a separate function for anything that might be ifdef-ed out. Signed-off-by:
-
Anna Schumaker authored
This is to match kernel coding style for switch statements. Signed-off-by:
-
Anna Schumaker authored
Most places in the kernel tend to line up cases with the switch to reduce indentation, so move this over to match that style. Additionally, I handle the (status >= 0) case in the switch so that we only "goto restart" from a single place after error handling. Signed-off-by:
-
Anna Schumaker authored
Moving all of this into a new function removes the need for cramped indentation, making the code overall easier to look at. I also take this chance to switch copy recovery over to using nfs4_stateid_match_other() Signed-off-by:
-
