aboutsummaryrefslogtreecommitdiff
path: root/fs/nfs/nfs4proc.c
diff options
context:
space:
mode:
authorJeff Layton <jlayton@poochiereds.net>2016-05-17 12:28:47 -0400
committerAnna Schumaker <Anna.Schumaker@Netapp.com>2016-05-17 15:48:12 -0400
commit183d9e7b112aaed0d19c16ffcf0f8c3a86dc71e0 (patch)
treee78c3b12f7a309b7a364dff4ebcab69fab66cfae /fs/nfs/nfs4proc.c
parent83026d80a16ea6a4e4f06a994fc7831b1d8d6375 (diff)
downloadlinux-stericsson-183d9e7b112aaed0d19c16ffcf0f8c3a86dc71e0.tar.gz
pnfs: rework LAYOUTGET retry handling
There are several problems in the way a stateid is selected for a LAYOUTGET operation: We pick a stateid to use in the RPC prepare op, but that makes it difficult to serialize LAYOUTGETs that use the open stateid. That serialization is done in pnfs_update_layout, which occurs well before the rpc_prepare operation. Between those two events, the i_lock is dropped and reacquired. pnfs_update_layout can find that the list has lsegs in it and not do any serialization, but then later pnfs_choose_layoutget_stateid ends up choosing the open stateid. This patch changes the client to select the stateid to use in the LAYOUTGET earlier, when we're searching for a usable layout segment. This way we can do it all while holding the i_lock the first time, and ensure that we serialize any LAYOUTGET call that uses a non-layout stateid. This also means a rework of how LAYOUTGET replies are handled, as we must now get the latest stateid if we want to retransmit in response to a retryable error. Most of those errors boil down to the fact that the layout state has changed in some fashion. Thus, what we really want to do is to re-search for a layout when it fails with a retryable error, so that we can avoid reissuing the RPC at all if possible. While the LAYOUTGET RPC is async, the initiating thread always waits for it to complete, so it's effectively synchronous anyway. Currently, when we need to retry a LAYOUTGET because of an error, we drive that retry via the rpc state machine. This means that once the call has been submitted, it runs until it completes. So, we must move the error handling for this RPC out of the rpc_call_done operation and into the caller. In order to handle errors like NFS4ERR_DELAY properly, we must also pass a pointer to the sliding timeout, which is now moved to the stack in pnfs_update_layout. The complicating errors are -NFS4ERR_RECALLCONFLICT and -NFS4ERR_LAYOUTTRYLATER, as those involve a timeout after which we give up and return NULL back to the caller. So, there is some special handling for those errors to ensure that the layers driving the retries can handle that appropriately. Signed-off-by: Jeff Layton <jeff.layton@primarydata.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Diffstat (limited to 'fs/nfs/nfs4proc.c')
-rw-r--r--fs/nfs/nfs4proc.c115
1 files changed, 48 insertions, 67 deletions
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 2a29f5d12aeb..62d67f040906 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -427,6 +427,7 @@ static int nfs4_do_handle_exception(struct nfs_server *server,
case -NFS4ERR_DELAY:
nfs_inc_server_stats(server, NFSIOS_DELAY);
case -NFS4ERR_GRACE:
+ case -NFS4ERR_RECALLCONFLICT:
exception->delay = 1;
return 0;
@@ -7847,40 +7848,34 @@ nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
struct nfs4_layoutget *lgp = calldata;
struct nfs_server *server = NFS_SERVER(lgp->args.inode);
struct nfs4_session *session = nfs4_get_session(server);
- int ret;
dprintk("--> %s\n", __func__);
- /* Note the is a race here, where a CB_LAYOUTRECALL can come in
- * right now covering the LAYOUTGET we are about to send.
- * However, that is not so catastrophic, and there seems
- * to be no way to prevent it completely.
- */
- if (nfs41_setup_sequence(session, &lgp->args.seq_args,
- &lgp->res.seq_res, task))
- return;
- ret = pnfs_choose_layoutget_stateid(&lgp->args.stateid,
- NFS_I(lgp->args.inode)->layout,
- &lgp->args.range,
- lgp->args.ctx->state);
- if (ret < 0)
- rpc_exit(task, ret);
+ nfs41_setup_sequence(session, &lgp->args.seq_args,
+ &lgp->res.seq_res, task);
+ dprintk("<-- %s\n", __func__);
}
static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
{
struct nfs4_layoutget *lgp = calldata;
+
+ dprintk("--> %s\n", __func__);
+ nfs41_sequence_done(task, &lgp->res.seq_res);
+ dprintk("<-- %s\n", __func__);
+}
+
+static int
+nfs4_layoutget_handle_exception(struct rpc_task *task,
+ struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
+{
struct inode *inode = lgp->args.inode;
struct nfs_server *server = NFS_SERVER(inode);
struct pnfs_layout_hdr *lo;
- struct nfs4_state *state = NULL;
- unsigned long timeo, now, giveup;
+ int status = task->tk_status;
dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
- if (!nfs41_sequence_done(task, &lgp->res.seq_res))
- goto out;
-
- switch (task->tk_status) {
+ switch (status) {
case 0:
goto out;
@@ -7890,57 +7885,43 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
* retry go inband.
*/
case -NFS4ERR_LAYOUTUNAVAILABLE:
- task->tk_status = -ENODATA;
+ status = -ENODATA;
goto out;
/*
* NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
* length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
*/
case -NFS4ERR_BADLAYOUT:
- goto out_overflow;
+ status = -EOVERFLOW;
+ goto out;
/*
* NFS4ERR_LAYOUTTRYLATER is a conflict with another client
* (or clients) writing to the same RAID stripe except when
* the minlength argument is 0 (see RFC5661 section 18.43.3).
+ *
+ * Treat it like we would RECALLCONFLICT -- we retry for a little
+ * while, and then eventually give up.
*/
case -NFS4ERR_LAYOUTTRYLATER:
- if (lgp->args.minlength == 0)
- goto out_overflow;
- /*
- * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
- * existing layout before getting a new one).
- */
- case -NFS4ERR_RECALLCONFLICT:
- timeo = rpc_get_timeout(task->tk_client);
- giveup = lgp->args.timestamp + timeo;
- now = jiffies;
- if (time_after(giveup, now)) {
- unsigned long delay;
-
- /* Delay for:
- * - Not less then NFS4_POLL_RETRY_MIN.
- * - One last time a jiffie before we give up
- * - exponential backoff (time_now minus start_attempt)
- */
- delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
- min((giveup - now - 1),
- now - lgp->args.timestamp));
-
- dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
- __func__, delay);
- rpc_delay(task, delay);
- /* Do not call nfs4_async_handle_error() */
- goto out_restart;
+ if (lgp->args.minlength == 0) {
+ status = -EOVERFLOW;
+ goto out;
}
- break;
+ /* Fallthrough */
+ case -NFS4ERR_RECALLCONFLICT:
+ nfs4_handle_exception(server, -NFS4ERR_RECALLCONFLICT,
+ exception);
+ status = -ERECALLCONFLICT;
+ goto out;
case -NFS4ERR_EXPIRED:
case -NFS4ERR_BAD_STATEID:
+ exception->timeout = 0;
spin_lock(&inode->i_lock);
if (nfs4_stateid_match(&lgp->args.stateid,
&lgp->args.ctx->state->stateid)) {
spin_unlock(&inode->i_lock);
/* If the open stateid was bad, then recover it. */
- state = lgp->args.ctx->state;
+ exception->state = lgp->args.ctx->state;
break;
}
lo = NFS_I(inode)->layout;
@@ -7958,20 +7939,16 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
pnfs_free_lseg_list(&head);
} else
spin_unlock(&inode->i_lock);
- goto out_restart;
+ status = -EAGAIN;
+ goto out;
}
- if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN)
- goto out_restart;
+
+ status = nfs4_handle_exception(server, status, exception);
+ if (exception->retry)
+ status = -EAGAIN;
out:
dprintk("<-- %s\n", __func__);
- return;
-out_restart:
- task->tk_status = 0;
- rpc_restart_call_prepare(task);
- return;
-out_overflow:
- task->tk_status = -EOVERFLOW;
- goto out;
+ return status;
}
static size_t max_response_pages(struct nfs_server *server)
@@ -8040,7 +8017,7 @@ static const struct rpc_call_ops nfs4_layoutget_call_ops = {
};
struct pnfs_layout_segment *
-nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
+nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
{
struct inode *inode = lgp->args.inode;
struct nfs_server *server = NFS_SERVER(inode);
@@ -8060,6 +8037,7 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
.flags = RPC_TASK_ASYNC,
};
struct pnfs_layout_segment *lseg = NULL;
+ struct nfs4_exception exception = { .timeout = *timeout };
int status = 0;
dprintk("--> %s\n", __func__);
@@ -8073,7 +8051,6 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
return ERR_PTR(-ENOMEM);
}
lgp->args.layout.pglen = max_pages * PAGE_SIZE;
- lgp->args.timestamp = jiffies;
lgp->res.layoutp = &lgp->args.layout;
lgp->res.seq_res.sr_slot = NULL;
@@ -8083,13 +8060,17 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
if (IS_ERR(task))
return ERR_CAST(task);
status = nfs4_wait_for_completion_rpc_task(task);
- if (status == 0)
- status = task->tk_status;
+ if (status == 0) {
+ status = nfs4_layoutget_handle_exception(task, lgp, &exception);
+ *timeout = exception.timeout;
+ }
+
trace_nfs4_layoutget(lgp->args.ctx,
&lgp->args.range,
&lgp->res.range,
&lgp->res.stateid,
status);
+
/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
if (status == 0 && lgp->res.layoutp->len)
lseg = pnfs_layout_process(lgp);