From a20c6bec0b8ae775e2e8f350819cef98eea9a832 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:36 -0400 Subject: NFS: grab open context in direct read Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 481be7f7bdd3..8a8942326758 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -260,7 +260,7 @@ static void nfs_direct_read_release(void *calldata) if (put_dreq(dreq)) nfs_direct_complete(dreq); - nfs_readdata_free(data); + nfs_readdata_release(data); } static const struct rpc_call_ops nfs_read_direct_ops = { @@ -337,7 +337,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, data->inode = inode; data->cred = msg.rpc_cred; data->args.fh = NFS_FH(inode); - data->args.context = ctx; + data->args.context = get_nfs_open_context(ctx); data->args.lock_context = dreq->l_ctx; data->args.offset = pos; data->args.pgbase = pgbase; -- cgit v1.2.3 From 0b7c01533aa9f4a228d07d2768d084acb3a387bc Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:39 -0400 Subject: NFS: add a struct nfs_commit_data to replace nfs_write_data in commits Commits don't need the vectors of pages, etc. that writes do. Split out a separate structure for the commit operation. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 17 ++++++------- fs/nfs/internal.h | 13 +++++----- fs/nfs/nfs3proc.c | 10 ++++++-- fs/nfs/nfs3xdr.c | 6 ++--- fs/nfs/nfs4filelayout.c | 65 ++++++++++++++++++++++++++++++++++--------------- fs/nfs/nfs4proc.c | 23 ++++++++++++----- fs/nfs/nfs4xdr.c | 8 +++--- fs/nfs/proc.c | 8 +++++- fs/nfs/write.c | 50 ++++++++++++++++++++++--------------- include/linux/nfs_fs.h | 4 +-- include/linux/nfs_xdr.h | 45 +++++++++++++++++++++++++++++++--- 11 files changed, 173 insertions(+), 76 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 8a8942326758..5897dfe48118 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -82,7 +82,7 @@ struct nfs_direct_req { /* commit state */ struct list_head rewrite_list; /* saved nfs_write_data structs */ - struct nfs_write_data * commit_data; /* special write_data for commits */ + struct nfs_commit_data *commit_data; /* special write_data for commits */ int flags; #define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */ #define NFS_ODIRECT_RESCHED_WRITES (2) /* write verification failed */ @@ -524,7 +524,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) static void nfs_direct_commit_result(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = calldata; + struct nfs_commit_data *data = calldata; /* Call the NFS version-specific code */ NFS_PROTO(data->inode)->commit_done(task, data); @@ -532,8 +532,8 @@ static void nfs_direct_commit_result(struct rpc_task *task, void *calldata) static void nfs_direct_commit_release(void *calldata) { - struct nfs_write_data *data = calldata; - struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req; + struct nfs_commit_data *data = calldata; + struct nfs_direct_req *dreq = data->dreq; int status = data->task.tk_status; if (status < 0) { @@ -551,14 +551,14 @@ static void nfs_direct_commit_release(void *calldata) } static const struct rpc_call_ops nfs_commit_direct_ops = { - .rpc_call_prepare = nfs_write_prepare, + .rpc_call_prepare = nfs_commit_prepare, .rpc_call_done = nfs_direct_commit_result, .rpc_release = nfs_direct_commit_release, }; static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq) { - struct nfs_write_data *data = dreq->commit_data; + struct nfs_commit_data *data = dreq->commit_data; struct rpc_task *task; struct rpc_message msg = { .rpc_argp = &data->args, @@ -581,9 +581,6 @@ static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq) data->args.fh = NFS_FH(data->inode); data->args.offset = 0; data->args.count = 0; - data->args.context = dreq->ctx; - data->args.lock_context = dreq->l_ctx; - data->res.count = 0; data->res.fattr = &data->fattr; data->res.verf = &data->verf; nfs_fattr_init(&data->fattr); @@ -625,7 +622,7 @@ static void nfs_alloc_commit_data(struct nfs_direct_req *dreq) { dreq->commit_data = nfs_commitdata_alloc(); if (dreq->commit_data != NULL) - dreq->commit_data->req = (struct nfs_page *) dreq; + dreq->commit_data->dreq = dreq; } #else static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq) diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index b777bdaba4c5..29ab441b22b1 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -314,24 +314,25 @@ extern void nfs_pageio_init_write_mds(struct nfs_pageio_descriptor *pgio, struct inode *inode, int ioflags); extern void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio); extern void nfs_writedata_release(struct nfs_write_data *wdata); -extern void nfs_commit_free(struct nfs_write_data *p); +extern void nfs_commit_free(struct nfs_commit_data *p); extern int nfs_initiate_write(struct nfs_write_data *data, struct rpc_clnt *clnt, const struct rpc_call_ops *call_ops, int how); extern void nfs_write_prepare(struct rpc_task *task, void *calldata); -extern int nfs_initiate_commit(struct nfs_write_data *data, - struct rpc_clnt *clnt, +extern void nfs_commit_prepare(struct rpc_task *task, void *calldata); +extern int nfs_initiate_commit(struct rpc_clnt *clnt, + struct nfs_commit_data *data, const struct rpc_call_ops *call_ops, int how); -extern void nfs_init_commit(struct nfs_write_data *data, +extern void nfs_init_commit(struct nfs_commit_data *data, struct list_head *head, struct pnfs_layout_segment *lseg); void nfs_retry_commit(struct list_head *page_list, struct pnfs_layout_segment *lseg); void nfs_commit_clear_lock(struct nfs_inode *nfsi); -void nfs_commitdata_release(void *data); -void nfs_commit_release_pages(struct nfs_write_data *data); +void nfs_commitdata_release(struct nfs_commit_data *data); +void nfs_commit_release_pages(struct nfs_commit_data *data); void nfs_request_add_commit_list(struct nfs_page *req, struct list_head *head); void nfs_request_remove_commit_list(struct nfs_page *req); diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index 5242eae6711a..b1daca7f0f7b 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -848,7 +848,12 @@ static void nfs3_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_ rpc_call_start(task); } -static int nfs3_commit_done(struct rpc_task *task, struct nfs_write_data *data) +static void nfs3_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data) +{ + rpc_call_start(task); +} + +static int nfs3_commit_done(struct rpc_task *task, struct nfs_commit_data *data) { if (nfs3_async_handle_jukebox(task, data->inode)) return -EAGAIN; @@ -856,7 +861,7 @@ static int nfs3_commit_done(struct rpc_task *task, struct nfs_write_data *data) return 0; } -static void nfs3_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg) +static void nfs3_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg) { msg->rpc_proc = &nfs3_procedures[NFS3PROC_COMMIT]; } @@ -907,6 +912,7 @@ const struct nfs_rpc_ops nfs_v3_clientops = { .write_rpc_prepare = nfs3_proc_write_rpc_prepare, .write_done = nfs3_write_done, .commit_setup = nfs3_proc_commit_setup, + .commit_rpc_prepare = nfs3_proc_commit_rpc_prepare, .commit_done = nfs3_commit_done, .lock = nfs3_proc_lock, .clear_acl_cache = nfs3_forget_cached_acls, diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index a77cc9a3ce55..01e53e94f53d 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -1287,7 +1287,7 @@ static void nfs3_xdr_enc_readdirplus3args(struct rpc_rqst *req, * }; */ static void encode_commit3args(struct xdr_stream *xdr, - const struct nfs_writeargs *args) + const struct nfs_commitargs *args) { __be32 *p; @@ -1300,7 +1300,7 @@ static void encode_commit3args(struct xdr_stream *xdr, static void nfs3_xdr_enc_commit3args(struct rpc_rqst *req, struct xdr_stream *xdr, - const struct nfs_writeargs *args) + const struct nfs_commitargs *args) { encode_commit3args(xdr, args); } @@ -2319,7 +2319,7 @@ out_status: */ static int nfs3_xdr_dec_commit3res(struct rpc_rqst *req, struct xdr_stream *xdr, - struct nfs_writeres *result) + struct nfs_commitres *result) { enum nfs_stat status; int error; diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 15aeba20d57d..675ce3b8663c 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -250,7 +250,7 @@ static int filelayout_write_done_cb(struct rpc_task *task, } /* Fake up some data that will cause nfs_commit_release to retry the writes. */ -static void prepare_to_resend_writes(struct nfs_write_data *data) +static void prepare_to_resend_writes(struct nfs_commit_data *data) { struct nfs_page *first = nfs_list_entry(data->pages.next); @@ -261,11 +261,11 @@ static void prepare_to_resend_writes(struct nfs_write_data *data) } static int filelayout_commit_done_cb(struct rpc_task *task, - struct nfs_write_data *data) + struct nfs_commit_data *data) { int reset = 0; - if (filelayout_async_handle_error(task, data->args.context->state, + if (filelayout_async_handle_error(task, data->context->state, data->ds_clp, &reset) == -EAGAIN) { dprintk("%s calling restart ds_clp %p ds_clp->cl_session %p\n", __func__, data->ds_clp, data->ds_clp->cl_session); @@ -315,15 +315,42 @@ static void filelayout_write_release(void *data) wdata->mds_ops->rpc_release(data); } -static void filelayout_commit_release(void *data) +static void filelayout_commit_prepare(struct rpc_task *task, void *data) { - struct nfs_write_data *wdata = (struct nfs_write_data *)data; + struct nfs_commit_data *wdata = data; - nfs_commit_release_pages(wdata); - if (atomic_dec_and_test(&NFS_I(wdata->inode)->commits_outstanding)) - nfs_commit_clear_lock(NFS_I(wdata->inode)); - put_lseg(wdata->lseg); - nfs_commitdata_release(wdata); + if (nfs41_setup_sequence(wdata->ds_clp->cl_session, + &wdata->args.seq_args, &wdata->res.seq_res, + task)) + return; + + rpc_call_start(task); +} + +static void filelayout_write_commit_done(struct rpc_task *task, void *data) +{ + struct nfs_commit_data *wdata = data; + + /* Note this may cause RPC to be resent */ + wdata->mds_ops->rpc_call_done(task, data); +} + +static void filelayout_commit_count_stats(struct rpc_task *task, void *data) +{ + struct nfs_commit_data *cdata = data; + + rpc_count_iostats(task, NFS_SERVER(cdata->inode)->client->cl_metrics); +} + +static void filelayout_commit_release(void *calldata) +{ + struct nfs_commit_data *data = calldata; + + nfs_commit_release_pages(data); + if (atomic_dec_and_test(&NFS_I(data->inode)->commits_outstanding)) + nfs_commit_clear_lock(NFS_I(data->inode)); + put_lseg(data->lseg); + nfs_commitdata_release(data); } static const struct rpc_call_ops filelayout_read_call_ops = { @@ -341,9 +368,9 @@ static const struct rpc_call_ops filelayout_write_call_ops = { }; static const struct rpc_call_ops filelayout_commit_call_ops = { - .rpc_call_prepare = filelayout_write_prepare, - .rpc_call_done = filelayout_write_call_done, - .rpc_count_stats = filelayout_write_count_stats, + .rpc_call_prepare = filelayout_commit_prepare, + .rpc_call_done = filelayout_write_commit_done, + .rpc_count_stats = filelayout_commit_count_stats, .rpc_release = filelayout_commit_release, }; @@ -922,7 +949,7 @@ select_ds_fh_from_commit(struct pnfs_layout_segment *lseg, u32 i) return flseg->fh_array[i]; } -static int filelayout_initiate_commit(struct nfs_write_data *data, int how) +static int filelayout_initiate_commit(struct nfs_commit_data *data, int how) { struct pnfs_layout_segment *lseg = data->lseg; struct nfs4_pnfs_ds *ds; @@ -941,12 +968,12 @@ static int filelayout_initiate_commit(struct nfs_write_data *data, int how) return -EAGAIN; } dprintk("%s ino %lu, how %d\n", __func__, data->inode->i_ino, how); - data->write_done_cb = filelayout_commit_done_cb; + data->commit_done_cb = filelayout_commit_done_cb; data->ds_clp = ds->ds_clp; fh = select_ds_fh_from_commit(lseg, data->ds_commit_index); if (fh) data->args.fh = fh; - return nfs_initiate_commit(data, ds->ds_clp->cl_rpcclient, + return nfs_initiate_commit(ds->ds_clp->cl_rpcclient, data, &filelayout_commit_call_ops, how); } @@ -1008,7 +1035,7 @@ alloc_ds_commits(struct inode *inode, struct list_head *list) { struct nfs4_fl_commit_info *fl_cinfo; struct nfs4_fl_commit_bucket *bucket; - struct nfs_write_data *data; + struct nfs_commit_data *data; int i, j; unsigned int nreq = 0; @@ -1044,7 +1071,7 @@ static int filelayout_commit_pagelist(struct inode *inode, struct list_head *mds_pages, int how) { - struct nfs_write_data *data, *tmp; + struct nfs_commit_data *data, *tmp; LIST_HEAD(list); unsigned int nreq = 0; @@ -1071,7 +1098,7 @@ filelayout_commit_pagelist(struct inode *inode, struct list_head *mds_pages, list_del_init(&data->pages); if (!data->lseg) { nfs_init_commit(data, mds_pages, NULL); - nfs_initiate_commit(data, NFS_CLIENT(inode), + nfs_initiate_commit(NFS_CLIENT(inode), data, data->mds_ops, how); } else { struct nfs4_fl_commit_info *fl_cinfo; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 75eb883ed4ce..cc04b6e409ed 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3468,7 +3468,17 @@ static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_ rpc_call_start(task); } -static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data) +static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data) +{ + if (nfs4_setup_sequence(NFS_SERVER(data->inode), + &data->args.seq_args, + &data->res.seq_res, + task)) + return; + rpc_call_start(task); +} + +static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data) { struct inode *inode = data->inode; @@ -3480,14 +3490,14 @@ static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *dat return 0; } -static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data) +static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data) { if (!nfs4_sequence_done(task, &data->res.seq_res)) return -EAGAIN; - return data->write_done_cb(task, data); + return data->commit_done_cb(task, data); } -static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg) +static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg) { struct nfs_server *server = NFS_SERVER(data->inode); @@ -3496,8 +3506,8 @@ static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_messa data->res.fattr = NULL; } else data->args.bitmask = server->cache_consistency_bitmask; - if (!data->write_done_cb) - data->write_done_cb = nfs4_commit_done_cb; + if (data->commit_done_cb == NULL) + data->commit_done_cb = nfs4_commit_done_cb; data->res.server = server; msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT]; nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1); @@ -6591,6 +6601,7 @@ const struct nfs_rpc_ops nfs_v4_clientops = { .write_rpc_prepare = nfs4_proc_write_rpc_prepare, .write_done = nfs4_write_done, .commit_setup = nfs4_proc_commit_setup, + .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare, .commit_done = nfs4_commit_done, .lock = nfs4_proc_lock, .clear_acl_cache = nfs4_zap_acl_attr, diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index c54aae364bee..4c3cc0ed9543 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1103,7 +1103,7 @@ static void encode_close(struct xdr_stream *xdr, const struct nfs_closeargs *arg encode_nfs4_stateid(xdr, arg->stateid); } -static void encode_commit(struct xdr_stream *xdr, const struct nfs_writeargs *args, struct compound_hdr *hdr) +static void encode_commit(struct xdr_stream *xdr, const struct nfs_commitargs *args, struct compound_hdr *hdr) { __be32 *p; @@ -2448,7 +2448,7 @@ static void nfs4_xdr_enc_write(struct rpc_rqst *req, struct xdr_stream *xdr, * a COMMIT request */ static void nfs4_xdr_enc_commit(struct rpc_rqst *req, struct xdr_stream *xdr, - struct nfs_writeargs *args) + struct nfs_commitargs *args) { struct compound_hdr hdr = { .minorversion = nfs4_xdr_minorversion(&args->seq_args), @@ -4102,7 +4102,7 @@ static int decode_verifier(struct xdr_stream *xdr, void *verifier) return decode_opaque_fixed(xdr, verifier, NFS4_VERIFIER_SIZE); } -static int decode_commit(struct xdr_stream *xdr, struct nfs_writeres *res) +static int decode_commit(struct xdr_stream *xdr, struct nfs_commitres *res) { int status; @@ -6353,7 +6353,7 @@ out: * Decode COMMIT response */ static int nfs4_xdr_dec_commit(struct rpc_rqst *rqstp, struct xdr_stream *xdr, - struct nfs_writeres *res) + struct nfs_commitres *res) { struct compound_hdr hdr; int status; diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index b63b6f4d14fb..bf80503200f5 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -688,8 +688,13 @@ static void nfs_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_d rpc_call_start(task); } +static void nfs_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data) +{ + BUG(); +} + static void -nfs_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg) +nfs_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg) { BUG(); } @@ -764,6 +769,7 @@ const struct nfs_rpc_ops nfs_v2_clientops = { .write_rpc_prepare = nfs_proc_write_rpc_prepare, .write_done = nfs_write_done, .commit_setup = nfs_proc_commit_setup, + .commit_rpc_prepare = nfs_proc_commit_rpc_prepare, .lock = nfs_proc_lock, .lock_check_bounds = nfs_lock_check_bounds, .close_context = nfs_close_context, diff --git a/fs/nfs/write.c b/fs/nfs/write.c index c07462320f6b..54f7c0ffe5c3 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -48,11 +48,12 @@ static const struct rpc_call_ops nfs_commit_ops; static struct kmem_cache *nfs_wdata_cachep; static mempool_t *nfs_wdata_mempool; +static struct kmem_cache *nfs_cdata_cachep; static mempool_t *nfs_commit_mempool; -struct nfs_write_data *nfs_commitdata_alloc(void) +struct nfs_commit_data *nfs_commitdata_alloc(void) { - struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOFS); + struct nfs_commit_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOFS); if (p) { memset(p, 0, sizeof(*p)); @@ -62,10 +63,8 @@ struct nfs_write_data *nfs_commitdata_alloc(void) } EXPORT_SYMBOL_GPL(nfs_commitdata_alloc); -void nfs_commit_free(struct nfs_write_data *p) +void nfs_commit_free(struct nfs_commit_data *p) { - if (p && (p->pagevec != &p->page_array[0])) - kfree(p->pagevec); mempool_free(p, nfs_commit_mempool); } EXPORT_SYMBOL_GPL(nfs_commit_free); @@ -1179,6 +1178,13 @@ void nfs_write_prepare(struct rpc_task *task, void *calldata) NFS_PROTO(data->inode)->write_rpc_prepare(task, data); } +void nfs_commit_prepare(struct rpc_task *task, void *calldata) +{ + struct nfs_commit_data *data = calldata; + + NFS_PROTO(data->inode)->commit_rpc_prepare(task, data); +} + static const struct rpc_call_ops nfs_write_partial_ops = { .rpc_call_prepare = nfs_write_prepare, .rpc_call_done = nfs_writeback_done_partial, @@ -1355,16 +1361,14 @@ void nfs_commit_clear_lock(struct nfs_inode *nfsi) } EXPORT_SYMBOL_GPL(nfs_commit_clear_lock); -void nfs_commitdata_release(void *data) +void nfs_commitdata_release(struct nfs_commit_data *data) { - struct nfs_write_data *wdata = data; - - put_nfs_open_context(wdata->args.context); - nfs_commit_free(wdata); + put_nfs_open_context(data->context); + nfs_commit_free(data); } EXPORT_SYMBOL_GPL(nfs_commitdata_release); -int nfs_initiate_commit(struct nfs_write_data *data, struct rpc_clnt *clnt, +int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data, const struct rpc_call_ops *call_ops, int how) { @@ -1403,7 +1407,7 @@ EXPORT_SYMBOL_GPL(nfs_initiate_commit); /* * Set up the argument/result storage required for the RPC call. */ -void nfs_init_commit(struct nfs_write_data *data, +void nfs_init_commit(struct nfs_commit_data *data, struct list_head *head, struct pnfs_layout_segment *lseg) { @@ -1424,8 +1428,7 @@ void nfs_init_commit(struct nfs_write_data *data, /* Note: we always request a commit of the entire inode */ data->args.offset = 0; data->args.count = 0; - data->args.context = get_nfs_open_context(first->wb_context); - data->res.count = 0; + data->context = get_nfs_open_context(first->wb_context); data->res.fattr = &data->fattr; data->res.verf = &data->verf; nfs_fattr_init(&data->fattr); @@ -1455,7 +1458,7 @@ EXPORT_SYMBOL_GPL(nfs_retry_commit); static int nfs_commit_list(struct inode *inode, struct list_head *head, int how) { - struct nfs_write_data *data; + struct nfs_commit_data *data; data = nfs_commitdata_alloc(); @@ -1464,7 +1467,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how) /* Set up the argument struct */ nfs_init_commit(data, head, NULL); - return nfs_initiate_commit(data, NFS_CLIENT(inode), data->mds_ops, how); + return nfs_initiate_commit(NFS_CLIENT(inode), data, data->mds_ops, how); out_bad: nfs_retry_commit(head, NULL); nfs_commit_clear_lock(NFS_I(inode)); @@ -1476,7 +1479,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how) */ static void nfs_commit_done(struct rpc_task *task, void *calldata) { - struct nfs_write_data *data = calldata; + struct nfs_commit_data *data = calldata; dprintk("NFS: %5u nfs_commit_done (status %d)\n", task->tk_pid, task->tk_status); @@ -1485,7 +1488,7 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata) NFS_PROTO(data->inode)->commit_done(task, data); } -void nfs_commit_release_pages(struct nfs_write_data *data) +void nfs_commit_release_pages(struct nfs_commit_data *data) { struct nfs_page *req; int status = data->task.tk_status; @@ -1526,7 +1529,7 @@ EXPORT_SYMBOL_GPL(nfs_commit_release_pages); static void nfs_commit_release(void *calldata) { - struct nfs_write_data *data = calldata; + struct nfs_commit_data *data = calldata; nfs_commit_release_pages(data); nfs_commit_clear_lock(NFS_I(data->inode)); @@ -1534,7 +1537,7 @@ static void nfs_commit_release(void *calldata) } static const struct rpc_call_ops nfs_commit_ops = { - .rpc_call_prepare = nfs_write_prepare, + .rpc_call_prepare = nfs_commit_prepare, .rpc_call_done = nfs_commit_done, .rpc_release = nfs_commit_release, }; @@ -1753,6 +1756,13 @@ int __init nfs_init_writepagecache(void) if (nfs_wdata_mempool == NULL) return -ENOMEM; + nfs_cdata_cachep = kmem_cache_create("nfs_commit_data", + sizeof(struct nfs_commit_data), + 0, SLAB_HWCACHE_ALIGN, + NULL); + if (nfs_cdata_cachep == NULL) + return -ENOMEM; + nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT, nfs_wdata_cachep); if (nfs_commit_mempool == NULL) diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 52a1bdb4ee2b..d5d68f322bf0 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -552,8 +552,8 @@ extern int nfs_wb_page(struct inode *inode, struct page* page); extern int nfs_wb_page_cancel(struct inode *inode, struct page* page); #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) extern int nfs_commit_inode(struct inode *, int); -extern struct nfs_write_data *nfs_commitdata_alloc(void); -extern void nfs_commit_free(struct nfs_write_data *wdata); +extern struct nfs_commit_data *nfs_commitdata_alloc(void); +extern void nfs_commit_free(struct nfs_commit_data *data); #else static inline int nfs_commit_inode(struct inode *inode, int how) diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 7ba3551a0414..8fb036a0d489 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -518,6 +518,24 @@ struct nfs_writeres { struct nfs4_sequence_res seq_res; }; +/* + * Arguments to the commit call. + */ +struct nfs_commitargs { + struct nfs_fh *fh; + __u64 offset; + __u32 count; + const u32 *bitmask; + struct nfs4_sequence_args seq_args; +}; + +struct nfs_commitres { + struct nfs_fattr *fattr; + struct nfs_writeverf *verf; + const struct nfs_server *server; + struct nfs4_sequence_res seq_res; +}; + /* * Common arguments to the unlink call */ @@ -1171,6 +1189,8 @@ struct nfs_read_data { struct page *page_array[NFS_PAGEVEC_SIZE]; }; +struct nfs_direct_req; + struct nfs_write_data { struct rpc_task task; struct inode *inode; @@ -1186,7 +1206,6 @@ struct nfs_write_data { struct nfs_writeres res; /* result struct */ struct pnfs_layout_segment *lseg; struct nfs_client *ds_clp; /* pNFS data server */ - int ds_commit_index; const struct rpc_call_ops *mds_ops; int (*write_done_cb) (struct rpc_task *task, struct nfs_write_data *data); #ifdef CONFIG_NFS_V4 @@ -1197,6 +1216,25 @@ struct nfs_write_data { struct page *page_array[NFS_PAGEVEC_SIZE]; }; +struct nfs_commit_data { + struct rpc_task task; + struct inode *inode; + struct rpc_cred *cred; + struct nfs_fattr fattr; + struct nfs_writeverf verf; + struct list_head pages; /* Coalesced requests we wish to flush */ + struct list_head list; /* lists of struct nfs_write_data */ + struct nfs_direct_req *dreq; /* O_DIRECT request */ + struct nfs_commitargs args; /* argument struct */ + struct nfs_commitres res; /* result struct */ + struct nfs_open_context *context; + struct pnfs_layout_segment *lseg; + struct nfs_client *ds_clp; /* pNFS data server */ + int ds_commit_index; + const struct rpc_call_ops *mds_ops; + int (*commit_done_cb) (struct rpc_task *task, struct nfs_commit_data *data); +}; + struct nfs_unlinkdata { struct hlist_node list; struct nfs_removeargs args; @@ -1277,8 +1315,9 @@ struct nfs_rpc_ops { void (*write_setup) (struct nfs_write_data *, struct rpc_message *); void (*write_rpc_prepare)(struct rpc_task *, struct nfs_write_data *); int (*write_done) (struct rpc_task *, struct nfs_write_data *); - void (*commit_setup) (struct nfs_write_data *, struct rpc_message *); - int (*commit_done) (struct rpc_task *, struct nfs_write_data *); + void (*commit_setup) (struct nfs_commit_data *, struct rpc_message *); + void (*commit_rpc_prepare)(struct rpc_task *, struct nfs_commit_data *); + int (*commit_done) (struct rpc_task *, struct nfs_commit_data *); int (*lock)(struct file *, int, struct file_lock *); int (*lock_check_bounds)(const struct file_lock *); void (*clear_acl_cache)(struct inode *); -- cgit v1.2.3 From 31f6852a4c187c031456581b35e146c0d5bbdecd Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:40 -0400 Subject: NFS: dprintks in directio code were referencing task after put Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 5897dfe48118..fb7fbaa79c20 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -357,15 +357,15 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, task = rpc_run_task(&task_setup_data); if (IS_ERR(task)) break; - rpc_put_task(task); dprintk("NFS: %5u initiated direct read call " "(req %s/%Ld, %zu bytes @ offset %Lu)\n", - data->task.tk_pid, + task->tk_pid, inode->i_sb->s_id, (long long)NFS_FILEID(inode), bytes, (unsigned long long)data->args.offset); + rpc_put_task(task); started += bytes; user_addr += bytes; @@ -784,15 +784,15 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, task = rpc_run_task(&task_setup_data); if (IS_ERR(task)) break; - rpc_put_task(task); dprintk("NFS: %5u initiated direct write call " "(req %s/%Ld, %zu bytes @ offset %Lu)\n", - data->task.tk_pid, + task->tk_pid, inode->i_sb->s_id, (long long)NFS_FILEID(inode), bytes, (unsigned long long)data->args.offset); + rpc_put_task(task); started += bytes; user_addr += bytes; -- cgit v1.2.3 From cd841605f7a721878d8a2d1362484723d8abf569 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:44 -0400 Subject: NFS: create common nfs_pgio_header for both read and write In order to avoid duplicating all the data in nfs_read_data whenever we split it up into multiple RPC calls (either due to a short read result or due to rsize < PAGE_SIZE), we split out the bits that are the same per RPC call into a separate "header" structure. The goal this patch moves towards is to have a single header refcounted by several rpc_data structures. Thus, want to always refer from rpc_data to the header, and not the other way. This patch comes close to that ideal, but the directio code currently needs some special casing, isolated in the nfs_direct_[read_write]hdr_release() functions. This will be dealt with in a future patch. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/blocklayout.c | 79 +++++++++++++++-------------- fs/nfs/direct.c | 73 ++++++++++++++++++--------- fs/nfs/internal.h | 4 ++ fs/nfs/nfs3proc.c | 14 ++++-- fs/nfs/nfs4filelayout.c | 40 ++++++++------- fs/nfs/nfs4proc.c | 44 ++++++++++------- fs/nfs/objlayout/objio_osd.c | 16 +++--- fs/nfs/objlayout/objlayout.c | 19 ++++--- fs/nfs/pnfs.c | 102 ++++++++++++++++++++++---------------- fs/nfs/proc.c | 10 ++-- fs/nfs/read.c | 89 +++++++++++++++++++-------------- fs/nfs/write.c | 104 ++++++++++++++++++++++----------------- include/linux/nfs_fs.h | 12 ----- include/linux/nfs_xdr.h | 48 ++++++++++-------- 14 files changed, 376 insertions(+), 278 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c index 7f6a23f0244e..7a482517f4c6 100644 --- a/fs/nfs/blocklayout/blocklayout.c +++ b/fs/nfs/blocklayout/blocklayout.c @@ -187,7 +187,6 @@ static void bl_end_io_read(struct bio *bio, int err) struct parallel_io *par = bio->bi_private; const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct nfs_read_data *rdata = (struct nfs_read_data *)par->data; do { struct page *page = bvec->bv_page; @@ -198,9 +197,12 @@ static void bl_end_io_read(struct bio *bio, int err) SetPageUptodate(page); } while (bvec >= bio->bi_io_vec); if (!uptodate) { - if (!rdata->pnfs_error) - rdata->pnfs_error = -EIO; - pnfs_set_lo_fail(rdata->lseg); + struct nfs_read_data *rdata = par->data; + struct nfs_pgio_header *header = rdata->header; + + if (!header->pnfs_error) + header->pnfs_error = -EIO; + pnfs_set_lo_fail(header->lseg); } bio_put(bio); put_parallel(par); @@ -221,7 +223,7 @@ bl_end_par_io_read(void *data, int unused) { struct nfs_read_data *rdata = data; - rdata->task.tk_status = rdata->pnfs_error; + rdata->task.tk_status = rdata->header->pnfs_error; INIT_WORK(&rdata->task.u.tk_work, bl_read_cleanup); schedule_work(&rdata->task.u.tk_work); } @@ -229,6 +231,7 @@ bl_end_par_io_read(void *data, int unused) static enum pnfs_try_status bl_read_pagelist(struct nfs_read_data *rdata) { + struct nfs_pgio_header *header = rdata->header; int i, hole; struct bio *bio = NULL; struct pnfs_block_extent *be = NULL, *cow_read = NULL; @@ -256,10 +259,10 @@ bl_read_pagelist(struct nfs_read_data *rdata) bl_put_extent(cow_read); bio = bl_submit_bio(READ, bio); /* Get the next one */ - be = bl_find_get_extent(BLK_LSEG2EXT(rdata->lseg), + be = bl_find_get_extent(BLK_LSEG2EXT(header->lseg), isect, &cow_read); if (!be) { - rdata->pnfs_error = -EIO; + header->pnfs_error = -EIO; goto out; } extent_length = be->be_length - @@ -286,7 +289,7 @@ bl_read_pagelist(struct nfs_read_data *rdata) isect, pages[i], be_read, bl_end_io_read, par); if (IS_ERR(bio)) { - rdata->pnfs_error = PTR_ERR(bio); + header->pnfs_error = PTR_ERR(bio); bio = NULL; goto out; } @@ -294,9 +297,9 @@ bl_read_pagelist(struct nfs_read_data *rdata) isect += PAGE_CACHE_SECTORS; extent_length -= PAGE_CACHE_SECTORS; } - if ((isect << SECTOR_SHIFT) >= rdata->inode->i_size) { + if ((isect << SECTOR_SHIFT) >= header->inode->i_size) { rdata->res.eof = 1; - rdata->res.count = rdata->inode->i_size - f_offset; + rdata->res.count = header->inode->i_size - f_offset; } else { rdata->res.count = (isect << SECTOR_SHIFT) - f_offset; } @@ -345,7 +348,6 @@ static void bl_end_io_write_zero(struct bio *bio, int err) struct parallel_io *par = bio->bi_private; const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct nfs_write_data *wdata = (struct nfs_write_data *)par->data; do { struct page *page = bvec->bv_page; @@ -358,9 +360,12 @@ static void bl_end_io_write_zero(struct bio *bio, int err) } while (bvec >= bio->bi_io_vec); if (unlikely(!uptodate)) { - if (!wdata->pnfs_error) - wdata->pnfs_error = -EIO; - pnfs_set_lo_fail(wdata->lseg); + struct nfs_write_data *data = par->data; + struct nfs_pgio_header *header = data->header; + + if (!header->pnfs_error) + header->pnfs_error = -EIO; + pnfs_set_lo_fail(header->lseg); } bio_put(bio); put_parallel(par); @@ -370,12 +375,13 @@ static void bl_end_io_write(struct bio *bio, int err) { struct parallel_io *par = bio->bi_private; const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); - struct nfs_write_data *wdata = (struct nfs_write_data *)par->data; + struct nfs_write_data *data = par->data; + struct nfs_pgio_header *header = data->header; if (!uptodate) { - if (!wdata->pnfs_error) - wdata->pnfs_error = -EIO; - pnfs_set_lo_fail(wdata->lseg); + if (!header->pnfs_error) + header->pnfs_error = -EIO; + pnfs_set_lo_fail(header->lseg); } bio_put(bio); put_parallel(par); @@ -391,9 +397,9 @@ static void bl_write_cleanup(struct work_struct *work) dprintk("%s enter\n", __func__); task = container_of(work, struct rpc_task, u.tk_work); wdata = container_of(task, struct nfs_write_data, task); - if (likely(!wdata->pnfs_error)) { + if (likely(!wdata->header->pnfs_error)) { /* Marks for LAYOUTCOMMIT */ - mark_extents_written(BLK_LSEG2EXT(wdata->lseg), + mark_extents_written(BLK_LSEG2EXT(wdata->header->lseg), wdata->args.offset, wdata->args.count); } pnfs_ld_write_done(wdata); @@ -404,12 +410,12 @@ static void bl_end_par_io_write(void *data, int num_se) { struct nfs_write_data *wdata = data; - if (unlikely(wdata->pnfs_error)) { - bl_free_short_extents(&BLK_LSEG2EXT(wdata->lseg)->bl_inval, + if (unlikely(wdata->header->pnfs_error)) { + bl_free_short_extents(&BLK_LSEG2EXT(wdata->header->lseg)->bl_inval, num_se); } - wdata->task.tk_status = wdata->pnfs_error; + wdata->task.tk_status = wdata->header->pnfs_error; wdata->verf.committed = NFS_FILE_SYNC; INIT_WORK(&wdata->task.u.tk_work, bl_write_cleanup); schedule_work(&wdata->task.u.tk_work); @@ -540,6 +546,7 @@ check_page: static enum pnfs_try_status bl_write_pagelist(struct nfs_write_data *wdata, int sync) { + struct nfs_pgio_header *header = wdata->header; int i, ret, npg_zero, pg_index, last = 0; struct bio *bio = NULL; struct pnfs_block_extent *be = NULL, *cow_read = NULL; @@ -552,7 +559,7 @@ bl_write_pagelist(struct nfs_write_data *wdata, int sync) pgoff_t index; u64 temp; int npg_per_block = - NFS_SERVER(wdata->inode)->pnfs_blksize >> PAGE_CACHE_SHIFT; + NFS_SERVER(header->inode)->pnfs_blksize >> PAGE_CACHE_SHIFT; dprintk("%s enter, %Zu@%lld\n", __func__, count, offset); /* At this point, wdata->pages is a (sequential) list of nfs_pages. @@ -566,7 +573,7 @@ bl_write_pagelist(struct nfs_write_data *wdata, int sync) /* At this point, have to be more careful with error handling */ isect = (sector_t) ((offset & (long)PAGE_CACHE_MASK) >> SECTOR_SHIFT); - be = bl_find_get_extent(BLK_LSEG2EXT(wdata->lseg), isect, &cow_read); + be = bl_find_get_extent(BLK_LSEG2EXT(header->lseg), isect, &cow_read); if (!be || !is_writable(be, isect)) { dprintk("%s no matching extents!\n", __func__); goto out_mds; @@ -597,10 +604,10 @@ fill_invalid_ext: dprintk("%s zero %dth page: index %lu isect %llu\n", __func__, npg_zero, index, (unsigned long long)isect); - page = bl_find_get_zeroing_page(wdata->inode, index, + page = bl_find_get_zeroing_page(header->inode, index, cow_read); if (unlikely(IS_ERR(page))) { - wdata->pnfs_error = PTR_ERR(page); + header->pnfs_error = PTR_ERR(page); goto out; } else if (page == NULL) goto next_page; @@ -612,7 +619,7 @@ fill_invalid_ext: __func__, ret); end_page_writeback(page); page_cache_release(page); - wdata->pnfs_error = ret; + header->pnfs_error = ret; goto out; } if (likely(!bl_push_one_short_extent(be->be_inval))) @@ -620,11 +627,11 @@ fill_invalid_ext: else { end_page_writeback(page); page_cache_release(page); - wdata->pnfs_error = -ENOMEM; + header->pnfs_error = -ENOMEM; goto out; } /* FIXME: This should be done in bi_end_io */ - mark_extents_written(BLK_LSEG2EXT(wdata->lseg), + mark_extents_written(BLK_LSEG2EXT(header->lseg), page->index << PAGE_CACHE_SHIFT, PAGE_CACHE_SIZE); @@ -632,7 +639,7 @@ fill_invalid_ext: isect, page, be, bl_end_io_write_zero, par); if (IS_ERR(bio)) { - wdata->pnfs_error = PTR_ERR(bio); + header->pnfs_error = PTR_ERR(bio); bio = NULL; goto out; } @@ -653,10 +660,10 @@ next_page: bl_put_extent(be); bio = bl_submit_bio(WRITE, bio); /* Get the next one */ - be = bl_find_get_extent(BLK_LSEG2EXT(wdata->lseg), + be = bl_find_get_extent(BLK_LSEG2EXT(header->lseg), isect, NULL); if (!be || !is_writable(be, isect)) { - wdata->pnfs_error = -EINVAL; + header->pnfs_error = -EINVAL; goto out; } if (be->be_state == PNFS_BLOCK_INVALID_DATA) { @@ -664,7 +671,7 @@ next_page: be->be_inval))) par->bse_count++; else { - wdata->pnfs_error = -ENOMEM; + header->pnfs_error = -ENOMEM; goto out; } } @@ -677,7 +684,7 @@ next_page: if (unlikely(ret)) { dprintk("%s bl_mark_sectors_init fail %d\n", __func__, ret); - wdata->pnfs_error = ret; + header->pnfs_error = ret; goto out; } } @@ -685,7 +692,7 @@ next_page: isect, pages[i], be, bl_end_io_write, par); if (IS_ERR(bio)) { - wdata->pnfs_error = PTR_ERR(bio); + header->pnfs_error = PTR_ERR(bio); bio = NULL; goto out; } diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index fb7fbaa79c20..56176af1436f 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -242,7 +242,7 @@ static void nfs_direct_read_release(void *calldata) { struct nfs_read_data *data = calldata; - struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req; + struct nfs_direct_req *dreq = (struct nfs_direct_req *)data->header->req; int status = data->task.tk_status; spin_lock(&dreq->lock); @@ -269,6 +269,15 @@ static const struct rpc_call_ops nfs_read_direct_ops = { .rpc_release = nfs_direct_read_release, }; +static void nfs_direct_readhdr_release(struct nfs_read_header *rhdr) +{ + struct nfs_read_data *data = &rhdr->rpc_data; + + if (data->pagevec != data->page_array) + kfree(data->pagevec); + nfs_readhdr_free(&rhdr->header); +} + /* * For each rsize'd chunk of the user's buffer, dispatch an NFS READ * operation. If nfs_readdata_alloc() or get_user_pages() fails, @@ -301,6 +310,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, ssize_t started = 0; do { + struct nfs_read_header *rhdr; struct nfs_read_data *data; size_t bytes; @@ -308,23 +318,24 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, bytes = min(rsize,count); result = -ENOMEM; - data = nfs_readdata_alloc(nfs_page_array_len(pgbase, bytes)); - if (unlikely(!data)) + rhdr = nfs_readhdr_alloc(nfs_page_array_len(pgbase, bytes)); + if (unlikely(!rhdr)) break; + data = &rhdr->rpc_data; down_read(¤t->mm->mmap_sem); result = get_user_pages(current, current->mm, user_addr, data->npages, 1, 0, data->pagevec, NULL); up_read(¤t->mm->mmap_sem); if (result < 0) { - nfs_readdata_free(data); + nfs_direct_readhdr_release(rhdr); break; } if ((unsigned)result < data->npages) { bytes = result * PAGE_SIZE; if (bytes <= pgbase) { nfs_direct_release_pages(data->pagevec, result); - nfs_readdata_free(data); + nfs_direct_readhdr_release(rhdr); break; } bytes -= pgbase; @@ -333,9 +344,9 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, get_dreq(dreq); - data->req = (struct nfs_page *) dreq; - data->inode = inode; - data->cred = msg.rpc_cred; + rhdr->header.req = (struct nfs_page *) dreq; + rhdr->header.inode = inode; + rhdr->header.cred = msg.rpc_cred; data->args.fh = NFS_FH(inode); data->args.context = get_nfs_open_context(ctx); data->args.lock_context = dreq->l_ctx; @@ -447,13 +458,23 @@ out: return result; } +static void nfs_direct_writehdr_release(struct nfs_write_header *whdr) +{ + struct nfs_write_data *data = &whdr->rpc_data; + + if (data->pagevec != data->page_array) + kfree(data->pagevec); + nfs_writehdr_free(&whdr->header); +} + static void nfs_direct_free_writedata(struct nfs_direct_req *dreq) { while (!list_empty(&dreq->rewrite_list)) { - struct nfs_write_data *data = list_entry(dreq->rewrite_list.next, struct nfs_write_data, pages); - list_del(&data->pages); - nfs_direct_release_pages(data->pagevec, data->npages); - nfs_writedata_free(data); + struct nfs_pgio_header *hdr = list_entry(dreq->rewrite_list.next, struct nfs_pgio_header, pages); + struct nfs_write_header *whdr = container_of(hdr, struct nfs_write_header, header); + list_del(&hdr->pages); + nfs_direct_release_pages(whdr->rpc_data.pagevec, whdr->rpc_data.npages); + nfs_direct_writehdr_release(whdr); } } @@ -463,6 +484,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) struct inode *inode = dreq->inode; struct list_head *p; struct nfs_write_data *data; + struct nfs_pgio_header *hdr; struct rpc_task *task; struct rpc_message msg = { .rpc_cred = dreq->ctx->cred, @@ -479,7 +501,8 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) get_dreq(dreq); list_for_each(p, &dreq->rewrite_list) { - data = list_entry(p, struct nfs_write_data, pages); + hdr = list_entry(p, struct nfs_pgio_header, pages); + data = &(container_of(hdr, struct nfs_write_header, header))->rpc_data; get_dreq(dreq); @@ -652,7 +675,8 @@ static void nfs_direct_write_result(struct rpc_task *task, void *calldata) static void nfs_direct_write_release(void *calldata) { struct nfs_write_data *data = calldata; - struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req; + struct nfs_pgio_header *hdr = data->header; + struct nfs_direct_req *dreq = (struct nfs_direct_req *) hdr->req; int status = data->task.tk_status; spin_lock(&dreq->lock); @@ -684,7 +708,7 @@ out_unlock: spin_unlock(&dreq->lock); if (put_dreq(dreq)) - nfs_direct_write_complete(dreq, data->inode); + nfs_direct_write_complete(dreq, hdr->inode); } static const struct rpc_call_ops nfs_write_direct_ops = { @@ -725,6 +749,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, ssize_t started = 0; do { + struct nfs_write_header *whdr; struct nfs_write_data *data; size_t bytes; @@ -732,23 +757,25 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, bytes = min(wsize,count); result = -ENOMEM; - data = nfs_writedata_alloc(nfs_page_array_len(pgbase, bytes)); - if (unlikely(!data)) + whdr = nfs_writehdr_alloc(nfs_page_array_len(pgbase, bytes)); + if (unlikely(!whdr)) break; + data = &whdr->rpc_data; + down_read(¤t->mm->mmap_sem); result = get_user_pages(current, current->mm, user_addr, data->npages, 0, 0, data->pagevec, NULL); up_read(¤t->mm->mmap_sem); if (result < 0) { - nfs_writedata_free(data); + nfs_direct_writehdr_release(whdr); break; } if ((unsigned)result < data->npages) { bytes = result * PAGE_SIZE; if (bytes <= pgbase) { nfs_direct_release_pages(data->pagevec, result); - nfs_writedata_free(data); + nfs_direct_writehdr_release(whdr); break; } bytes -= pgbase; @@ -757,11 +784,11 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, get_dreq(dreq); - list_move_tail(&data->pages, &dreq->rewrite_list); + list_move_tail(&whdr->header.pages, &dreq->rewrite_list); - data->req = (struct nfs_page *) dreq; - data->inode = inode; - data->cred = msg.rpc_cred; + whdr->header.req = (struct nfs_page *) dreq; + whdr->header.inode = inode; + whdr->header.cred = msg.rpc_cred; data->args.fh = NFS_FH(inode); data->args.context = ctx; data->args.lock_context = dreq->l_ctx; diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 650127fd24bb..7dc9be1a6e1a 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -296,6 +296,8 @@ extern int nfs4_get_rootfh(struct nfs_server *server, struct nfs_fh *mntfh); struct nfs_pageio_descriptor; /* read.c */ +extern struct nfs_read_header *nfs_readhdr_alloc(unsigned int npages); +extern void nfs_readhdr_free(struct nfs_pgio_header *hdr); extern int nfs_initiate_read(struct rpc_clnt *clnt, struct nfs_read_data *data, const struct rpc_call_ops *call_ops); @@ -309,6 +311,8 @@ extern void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio); extern void nfs_readdata_release(struct nfs_read_data *rdata); /* write.c */ +extern struct nfs_write_header *nfs_writehdr_alloc(unsigned int npages); +extern void nfs_writehdr_free(struct nfs_pgio_header *hdr); extern int nfs_generic_flush(struct nfs_pageio_descriptor *desc, struct list_head *head); extern void nfs_pageio_init_write_mds(struct nfs_pageio_descriptor *pgio, diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index b1daca7f0f7b..56dcefc2f3f7 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -811,11 +811,13 @@ nfs3_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, static int nfs3_read_done(struct rpc_task *task, struct nfs_read_data *data) { - if (nfs3_async_handle_jukebox(task, data->inode)) + struct inode *inode = data->header->inode; + + if (nfs3_async_handle_jukebox(task, inode)) return -EAGAIN; - nfs_invalidate_atime(data->inode); - nfs_refresh_inode(data->inode, &data->fattr); + nfs_invalidate_atime(inode); + nfs_refresh_inode(inode, &data->fattr); return 0; } @@ -831,10 +833,12 @@ static void nfs3_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_da static int nfs3_write_done(struct rpc_task *task, struct nfs_write_data *data) { - if (nfs3_async_handle_jukebox(task, data->inode)) + struct inode *inode = data->header->inode; + + if (nfs3_async_handle_jukebox(task, inode)) return -EAGAIN; if (task->tk_status >= 0) - nfs_post_op_update_inode_force_wcc(data->inode, data->res.fattr); + nfs_post_op_update_inode_force_wcc(inode, data->res.fattr); return 0; } diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index c536328557cb..ad1d68013a5b 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -148,6 +148,7 @@ wait_on_recovery: static int filelayout_read_done_cb(struct rpc_task *task, struct nfs_read_data *data) { + struct nfs_pgio_header *hdr = data->header; int reset = 0; dprintk("%s DS read\n", __func__); @@ -157,7 +158,7 @@ static int filelayout_read_done_cb(struct rpc_task *task, dprintk("%s calling restart ds_clp %p ds_clp->cl_session %p\n", __func__, data->ds_clp, data->ds_clp->cl_session); if (reset) { - pnfs_set_lo_fail(data->lseg); + pnfs_set_lo_fail(hdr->lseg); nfs4_reset_read(task, data); } rpc_restart_call_prepare(task); @@ -175,13 +176,15 @@ static int filelayout_read_done_cb(struct rpc_task *task, static void filelayout_set_layoutcommit(struct nfs_write_data *wdata) { - if (FILELAYOUT_LSEG(wdata->lseg)->commit_through_mds || + struct nfs_pgio_header *hdr = wdata->header; + + if (FILELAYOUT_LSEG(hdr->lseg)->commit_through_mds || wdata->res.verf->committed == NFS_FILE_SYNC) return; pnfs_set_layoutcommit(wdata); - dprintk("%s ionde %lu pls_end_pos %lu\n", __func__, wdata->inode->i_ino, - (unsigned long) NFS_I(wdata->inode)->layout->plh_lwb); + dprintk("%s ionde %lu pls_end_pos %lu\n", __func__, hdr->inode->i_ino, + (unsigned long) NFS_I(hdr->inode)->layout->plh_lwb); } /* @@ -210,27 +213,28 @@ static void filelayout_read_call_done(struct rpc_task *task, void *data) dprintk("--> %s task->tk_status %d\n", __func__, task->tk_status); /* Note this may cause RPC to be resent */ - rdata->mds_ops->rpc_call_done(task, data); + rdata->header->mds_ops->rpc_call_done(task, data); } static void filelayout_read_count_stats(struct rpc_task *task, void *data) { struct nfs_read_data *rdata = data; - rpc_count_iostats(task, NFS_SERVER(rdata->inode)->client->cl_metrics); + rpc_count_iostats(task, NFS_SERVER(rdata->header->inode)->client->cl_metrics); } static void filelayout_read_release(void *data) { struct nfs_read_data *rdata = data; - put_lseg(rdata->lseg); - rdata->mds_ops->rpc_release(data); + put_lseg(rdata->header->lseg); + rdata->header->mds_ops->rpc_release(data); } static int filelayout_write_done_cb(struct rpc_task *task, struct nfs_write_data *data) { + struct nfs_pgio_header *hdr = data->header; int reset = 0; if (filelayout_async_handle_error(task, data->args.context->state, @@ -238,7 +242,7 @@ static int filelayout_write_done_cb(struct rpc_task *task, dprintk("%s calling restart ds_clp %p ds_clp->cl_session %p\n", __func__, data->ds_clp, data->ds_clp->cl_session); if (reset) { - pnfs_set_lo_fail(data->lseg); + pnfs_set_lo_fail(hdr->lseg); nfs4_reset_write(task, data); } rpc_restart_call_prepare(task); @@ -297,22 +301,22 @@ static void filelayout_write_call_done(struct rpc_task *task, void *data) struct nfs_write_data *wdata = data; /* Note this may cause RPC to be resent */ - wdata->mds_ops->rpc_call_done(task, data); + wdata->header->mds_ops->rpc_call_done(task, data); } static void filelayout_write_count_stats(struct rpc_task *task, void *data) { struct nfs_write_data *wdata = data; - rpc_count_iostats(task, NFS_SERVER(wdata->inode)->client->cl_metrics); + rpc_count_iostats(task, NFS_SERVER(wdata->header->inode)->client->cl_metrics); } static void filelayout_write_release(void *data) { struct nfs_write_data *wdata = data; - put_lseg(wdata->lseg); - wdata->mds_ops->rpc_release(data); + put_lseg(wdata->header->lseg); + wdata->header->mds_ops->rpc_release(data); } static void filelayout_commit_prepare(struct rpc_task *task, void *data) @@ -377,7 +381,8 @@ static const struct rpc_call_ops filelayout_commit_call_ops = { static enum pnfs_try_status filelayout_read_pagelist(struct nfs_read_data *data) { - struct pnfs_layout_segment *lseg = data->lseg; + struct nfs_pgio_header *hdr = data->header; + struct pnfs_layout_segment *lseg = hdr->lseg; struct nfs4_pnfs_ds *ds; loff_t offset = data->args.offset; u32 j, idx; @@ -385,7 +390,7 @@ filelayout_read_pagelist(struct nfs_read_data *data) int status; dprintk("--> %s ino %lu pgbase %u req %Zu@%llu\n", - __func__, data->inode->i_ino, + __func__, hdr->inode->i_ino, data->args.pgbase, (size_t)data->args.count, offset); if (test_bit(NFS_DEVICEID_INVALID, &FILELAYOUT_DEVID_NODE(lseg)->flags)) @@ -423,7 +428,8 @@ filelayout_read_pagelist(struct nfs_read_data *data) static enum pnfs_try_status filelayout_write_pagelist(struct nfs_write_data *data, int sync) { - struct pnfs_layout_segment *lseg = data->lseg; + struct nfs_pgio_header *hdr = data->header; + struct pnfs_layout_segment *lseg = hdr->lseg; struct nfs4_pnfs_ds *ds; loff_t offset = data->args.offset; u32 j, idx; @@ -445,7 +451,7 @@ filelayout_write_pagelist(struct nfs_write_data *data, int sync) return PNFS_NOT_ATTEMPTED; } dprintk("%s ino %lu sync %d req %Zu@%llu DS: %s\n", __func__, - data->inode->i_ino, sync, (size_t) data->args.count, offset, + hdr->inode->i_ino, sync, (size_t) data->args.count, offset, ds->ds_remotestr); data->write_done_cb = filelayout_write_done_cb; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index cc04b6e409ed..5375862075de 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3336,12 +3336,12 @@ static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, void __nfs4_read_done_cb(struct nfs_read_data *data) { - nfs_invalidate_atime(data->inode); + nfs_invalidate_atime(data->header->inode); } static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data) { - struct nfs_server *server = NFS_SERVER(data->inode); + struct nfs_server *server = NFS_SERVER(data->header->inode); if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) { rpc_restart_call_prepare(task); @@ -3376,7 +3376,7 @@ static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data) { - if (nfs4_setup_sequence(NFS_SERVER(data->inode), + if (nfs4_setup_sequence(NFS_SERVER(data->header->inode), &data->args.seq_args, &data->res.seq_res, task)) @@ -3387,22 +3387,25 @@ static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_da /* Reset the the nfs_read_data to send the read to the MDS. */ void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data) { + struct nfs_pgio_header *hdr = data->header; + struct inode *inode = hdr->inode; + dprintk("%s Reset task for i/o through\n", __func__); - put_lseg(data->lseg); - data->lseg = NULL; + put_lseg(hdr->lseg); + hdr->lseg = NULL; + data->ds_clp = NULL; /* offsets will differ in the dense stripe case */ data->args.offset = data->mds_offset; - data->ds_clp = NULL; - data->args.fh = NFS_FH(data->inode); + data->args.fh = NFS_FH(inode); data->read_done_cb = nfs4_read_done_cb; - task->tk_ops = data->mds_ops; - rpc_task_reset_client(task, NFS_CLIENT(data->inode)); + task->tk_ops = hdr->mds_ops; + rpc_task_reset_client(task, NFS_CLIENT(inode)); } EXPORT_SYMBOL_GPL(nfs4_reset_read); static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data) { - struct inode *inode = data->inode; + struct inode *inode = data->header->inode; if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) { rpc_restart_call_prepare(task); @@ -3426,25 +3429,28 @@ static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data) /* Reset the the nfs_write_data to send the write to the MDS. */ void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data) { + struct nfs_pgio_header *hdr = data->header; + struct inode *inode = hdr->inode; + dprintk("%s Reset task for i/o through\n", __func__); - put_lseg(data->lseg); - data->lseg = NULL; - data->ds_clp = NULL; + put_lseg(hdr->lseg); + hdr->lseg = NULL; + data->ds_clp = NULL; data->write_done_cb = nfs4_write_done_cb; - data->args.fh = NFS_FH(data->inode); + data->args.fh = NFS_FH(inode); data->args.bitmask = data->res.server->cache_consistency_bitmask; data->args.offset = data->mds_offset; data->res.fattr = &data->fattr; - task->tk_ops = data->mds_ops; - rpc_task_reset_client(task, NFS_CLIENT(data->inode)); + task->tk_ops = hdr->mds_ops; + rpc_task_reset_client(task, NFS_CLIENT(inode)); } EXPORT_SYMBOL_GPL(nfs4_reset_write); static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg) { - struct nfs_server *server = NFS_SERVER(data->inode); + struct nfs_server *server = NFS_SERVER(data->header->inode); - if (data->lseg) { + if (data->header->lseg) { data->args.bitmask = NULL; data->res.fattr = NULL; } else @@ -3460,7 +3466,7 @@ static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_messag static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data) { - if (nfs4_setup_sequence(NFS_SERVER(data->inode), + if (nfs4_setup_sequence(NFS_SERVER(data->header->inode), &data->args.seq_args, &data->res.seq_res, task)) diff --git a/fs/nfs/objlayout/objio_osd.c b/fs/nfs/objlayout/objio_osd.c index 4bff4a3dab46..fbf4874ec252 100644 --- a/fs/nfs/objlayout/objio_osd.c +++ b/fs/nfs/objlayout/objio_osd.c @@ -440,11 +440,12 @@ static void _read_done(struct ore_io_state *ios, void *private) int objio_read_pagelist(struct nfs_read_data *rdata) { + struct nfs_pgio_header *hdr = rdata->header; struct objio_state *objios; int ret; - ret = objio_alloc_io_state(NFS_I(rdata->inode)->layout, true, - rdata->lseg, rdata->args.pages, rdata->args.pgbase, + ret = objio_alloc_io_state(NFS_I(hdr->inode)->layout, true, + hdr->lseg, rdata->args.pages, rdata->args.pgbase, rdata->args.offset, rdata->args.count, rdata, GFP_KERNEL, &objios); if (unlikely(ret)) @@ -483,12 +484,12 @@ static struct page *__r4w_get_page(void *priv, u64 offset, bool *uptodate) { struct objio_state *objios = priv; struct nfs_write_data *wdata = objios->oir.rpcdata; + struct address_space *mapping = wdata->header->inode->i_mapping; pgoff_t index = offset / PAGE_SIZE; - struct page *page = find_get_page(wdata->inode->i_mapping, index); + struct page *page = find_get_page(mapping, index); if (!page) { - page = find_or_create_page(wdata->inode->i_mapping, - index, GFP_NOFS); + page = find_or_create_page(mapping, index, GFP_NOFS); if (unlikely(!page)) { dprintk("%s: grab_cache_page Failed index=0x%lx\n", __func__, index); @@ -518,11 +519,12 @@ static const struct _ore_r4w_op _r4w_op = { int objio_write_pagelist(struct nfs_write_data *wdata, int how) { + struct nfs_pgio_header *hdr = wdata->header; struct objio_state *objios; int ret; - ret = objio_alloc_io_state(NFS_I(wdata->inode)->layout, false, - wdata->lseg, wdata->args.pages, wdata->args.pgbase, + ret = objio_alloc_io_state(NFS_I(hdr->inode)->layout, false, + hdr->lseg, wdata->args.pages, wdata->args.pgbase, wdata->args.offset, wdata->args.count, wdata, GFP_NOFS, &objios); if (unlikely(ret)) diff --git a/fs/nfs/objlayout/objlayout.c b/fs/nfs/objlayout/objlayout.c index 595c5fc21a19..874613545301 100644 --- a/fs/nfs/objlayout/objlayout.c +++ b/fs/nfs/objlayout/objlayout.c @@ -258,7 +258,7 @@ objlayout_read_done(struct objlayout_io_res *oir, ssize_t status, bool sync) if (status >= 0) rdata->res.count = status; else - rdata->pnfs_error = status; + rdata->header->pnfs_error = status; objlayout_iodone(oir); /* must not use oir after this point */ @@ -279,12 +279,14 @@ objlayout_read_done(struct objlayout_io_res *oir, ssize_t status, bool sync) enum pnfs_try_status objlayout_read_pagelist(struct nfs_read_data *rdata) { + struct nfs_pgio_header *hdr = rdata->header; + struct inode *inode = hdr->inode; loff_t offset = rdata->args.offset; size_t count = rdata->args.count; int err; loff_t eof; - eof = i_size_read(rdata->inode); + eof = i_size_read(inode); if (unlikely(offset + count > eof)) { if (offset >= eof) { err = 0; @@ -297,17 +299,17 @@ objlayout_read_pagelist(struct nfs_read_data *rdata) } rdata->res.eof = (offset + count) >= eof; - _fix_verify_io_params(rdata->lseg, &rdata->args.pages, + _fix_verify_io_params(hdr->lseg, &rdata->args.pages, &rdata->args.pgbase, rdata->args.offset, rdata->args.count); dprintk("%s: inode(%lx) offset 0x%llx count 0x%Zx eof=%d\n", - __func__, rdata->inode->i_ino, offset, count, rdata->res.eof); + __func__, inode->i_ino, offset, count, rdata->res.eof); err = objio_read_pagelist(rdata); out: if (unlikely(err)) { - rdata->pnfs_error = err; + hdr->pnfs_error = err; dprintk("%s: Returned Error %d\n", __func__, err); return PNFS_NOT_ATTEMPTED; } @@ -340,7 +342,7 @@ objlayout_write_done(struct objlayout_io_res *oir, ssize_t status, bool sync) wdata->res.count = status; wdata->verf.committed = oir->committed; } else { - wdata->pnfs_error = status; + wdata->header->pnfs_error = status; } objlayout_iodone(oir); /* must not use oir after this point */ @@ -363,15 +365,16 @@ enum pnfs_try_status objlayout_write_pagelist(struct nfs_write_data *wdata, int how) { + struct nfs_pgio_header *hdr = wdata->header; int err; - _fix_verify_io_params(wdata->lseg, &wdata->args.pages, + _fix_verify_io_params(hdr->lseg, &wdata->args.pages, &wdata->args.pgbase, wdata->args.offset, wdata->args.count); err = objio_write_pagelist(wdata, how); if (unlikely(err)) { - wdata->pnfs_error = err; + hdr->pnfs_error = err; dprintk("%s: Returned Error %d\n", __func__, err); return PNFS_NOT_ATTEMPTED; } diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 9c4d14a17d49..d705da427e6d 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1191,13 +1191,15 @@ static int pnfs_write_done_resend_to_mds(struct inode *inode, struct list_head * static void pnfs_ld_handle_write_error(struct nfs_write_data *data) { - dprintk("pnfs write error = %d\n", data->pnfs_error); - if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & + struct nfs_pgio_header *hdr = data->header; + + dprintk("pnfs write error = %d\n", hdr->pnfs_error); + if (NFS_SERVER(hdr->inode)->pnfs_curr_ld->flags & PNFS_LAYOUTRET_ON_ERROR) { - clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(data->inode)->flags); - pnfs_return_layout(data->inode); + clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(hdr->inode)->flags); + pnfs_return_layout(hdr->inode); } - data->task.tk_status = pnfs_write_done_resend_to_mds(data->inode, &data->pages); + data->task.tk_status = pnfs_write_done_resend_to_mds(hdr->inode, &hdr->pages); } /* @@ -1205,13 +1207,15 @@ static void pnfs_ld_handle_write_error(struct nfs_write_data *data) */ void pnfs_ld_write_done(struct nfs_write_data *data) { - if (likely(!data->pnfs_error)) { + struct nfs_pgio_header *hdr = data->header; + + if (!hdr->pnfs_error) { pnfs_set_layoutcommit(data); - data->mds_ops->rpc_call_done(&data->task, data); + hdr->mds_ops->rpc_call_done(&data->task, data); } else pnfs_ld_handle_write_error(data); - put_lseg(data->lseg); - data->mds_ops->rpc_release(data); + put_lseg(hdr->lseg); + hdr->mds_ops->rpc_release(data); } EXPORT_SYMBOL_GPL(pnfs_ld_write_done); @@ -1219,12 +1223,14 @@ static void pnfs_write_through_mds(struct nfs_pageio_descriptor *desc, struct nfs_write_data *data) { - list_splice_tail_init(&data->pages, &desc->pg_list); - if (data->req && list_empty(&data->req->wb_list)) - nfs_list_add_request(data->req, &desc->pg_list); + struct nfs_pgio_header *hdr = data->header; + + list_splice_tail_init(&hdr->pages, &desc->pg_list); + if (hdr->req && list_empty(&hdr->req->wb_list)) + nfs_list_add_request(hdr->req, &desc->pg_list); nfs_pageio_reset_write_mds(desc); desc->pg_recoalesce = 1; - put_lseg(data->lseg); + put_lseg(hdr->lseg); nfs_writedata_release(data); } @@ -1234,20 +1240,21 @@ pnfs_try_to_write_data(struct nfs_write_data *wdata, struct pnfs_layout_segment *lseg, int how) { - struct inode *inode = wdata->inode; + struct nfs_pgio_header *hdr = wdata->header; + struct inode *inode = hdr->inode; enum pnfs_try_status trypnfs; struct nfs_server *nfss = NFS_SERVER(inode); - wdata->mds_ops = call_ops; - wdata->lseg = get_lseg(lseg); + hdr->mds_ops = call_ops; + hdr->lseg = get_lseg(lseg); dprintk("%s: Writing ino:%lu %u@%llu (how %d)\n", __func__, inode->i_ino, wdata->args.count, wdata->args.offset, how); trypnfs = nfss->pnfs_curr_ld->write_pagelist(wdata, how); if (trypnfs == PNFS_NOT_ATTEMPTED) { - put_lseg(wdata->lseg); - wdata->lseg = NULL; + put_lseg(hdr->lseg); + hdr->lseg = NULL; } else nfs_inc_stats(inode, NFSIOS_PNFS_WRITE); @@ -1318,13 +1325,15 @@ static int pnfs_read_done_resend_to_mds(struct inode *inode, struct list_head *h static void pnfs_ld_handle_read_error(struct nfs_read_data *data) { - dprintk("pnfs read error = %d\n", data->pnfs_error); - if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & + struct nfs_pgio_header *hdr = data->header; + + dprintk("pnfs read error = %d\n", hdr->pnfs_error); + if (NFS_SERVER(hdr->inode)->pnfs_curr_ld->flags & PNFS_LAYOUTRET_ON_ERROR) { - clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(data->inode)->flags); - pnfs_return_layout(data->inode); + clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(hdr->inode)->flags); + pnfs_return_layout(hdr->inode); } - data->task.tk_status = pnfs_read_done_resend_to_mds(data->inode, &data->pages); + data->task.tk_status = pnfs_read_done_resend_to_mds(hdr->inode, &hdr->pages); } /* @@ -1332,13 +1341,15 @@ static void pnfs_ld_handle_read_error(struct nfs_read_data *data) */ void pnfs_ld_read_done(struct nfs_read_data *data) { - if (likely(!data->pnfs_error)) { + struct nfs_pgio_header *hdr = data->header; + + if (likely(!hdr->pnfs_error)) { __nfs4_read_done_cb(data); - data->mds_ops->rpc_call_done(&data->task, data); + hdr->mds_ops->rpc_call_done(&data->task, data); } else pnfs_ld_handle_read_error(data); - put_lseg(data->lseg); - data->mds_ops->rpc_release(data); + put_lseg(hdr->lseg); + hdr->mds_ops->rpc_release(data); } EXPORT_SYMBOL_GPL(pnfs_ld_read_done); @@ -1346,9 +1357,11 @@ static void pnfs_read_through_mds(struct nfs_pageio_descriptor *desc, struct nfs_read_data *data) { - list_splice_tail_init(&data->pages, &desc->pg_list); - if (data->req && list_empty(&data->req->wb_list)) - nfs_list_add_request(data->req, &desc->pg_list); + struct nfs_pgio_header *hdr = data->header; + + list_splice_tail_init(&hdr->pages, &desc->pg_list); + if (hdr->req && list_empty(&hdr->req->wb_list)) + nfs_list_add_request(hdr->req, &desc->pg_list); nfs_pageio_reset_read_mds(desc); desc->pg_recoalesce = 1; nfs_readdata_release(data); @@ -1362,20 +1375,21 @@ pnfs_try_to_read_data(struct nfs_read_data *rdata, const struct rpc_call_ops *call_ops, struct pnfs_layout_segment *lseg) { - struct inode *inode = rdata->inode; + struct nfs_pgio_header *hdr = rdata->header; + struct inode *inode = hdr->inode; struct nfs_server *nfss = NFS_SERVER(inode); enum pnfs_try_status trypnfs; - rdata->mds_ops = call_ops; - rdata->lseg = get_lseg(lseg); + hdr->mds_ops = call_ops; + hdr->lseg = get_lseg(lseg); dprintk("%s: Reading ino:%lu %u@%llu\n", __func__, inode->i_ino, rdata->args.count, rdata->args.offset); trypnfs = nfss->pnfs_curr_ld->read_pagelist(rdata); if (trypnfs == PNFS_NOT_ATTEMPTED) { - put_lseg(rdata->lseg); - rdata->lseg = NULL; + put_lseg(hdr->lseg); + hdr->lseg = NULL; } else { nfs_inc_stats(inode, NFSIOS_PNFS_READ); } @@ -1450,30 +1464,32 @@ EXPORT_SYMBOL_GPL(pnfs_set_lo_fail); void pnfs_set_layoutcommit(struct nfs_write_data *wdata) { - struct nfs_inode *nfsi = NFS_I(wdata->inode); + struct nfs_pgio_header *hdr = wdata->header; + struct inode *inode = hdr->inode; + struct nfs_inode *nfsi = NFS_I(inode); loff_t end_pos = wdata->mds_offset + wdata->res.count; bool mark_as_dirty = false; - spin_lock(&nfsi->vfs_inode.i_lock); + spin_lock(&inode->i_lock); if (!test_and_set_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) { mark_as_dirty = true; dprintk("%s: Set layoutcommit for inode %lu ", - __func__, wdata->inode->i_ino); + __func__, inode->i_ino); } - if (!test_and_set_bit(NFS_LSEG_LAYOUTCOMMIT, &wdata->lseg->pls_flags)) { + if (!test_and_set_bit(NFS_LSEG_LAYOUTCOMMIT, &hdr->lseg->pls_flags)) { /* references matched in nfs4_layoutcommit_release */ - get_lseg(wdata->lseg); + get_lseg(hdr->lseg); } if (end_pos > nfsi->layout->plh_lwb) nfsi->layout->plh_lwb = end_pos; - spin_unlock(&nfsi->vfs_inode.i_lock); + spin_unlock(&inode->i_lock); dprintk("%s: lseg %p end_pos %llu\n", - __func__, wdata->lseg, nfsi->layout->plh_lwb); + __func__, hdr->lseg, nfsi->layout->plh_lwb); /* if pnfs_layoutcommit_inode() runs between inode locks, the next one * will be a noop because NFS_INO_LAYOUTCOMMIT will not be set */ if (mark_as_dirty) - mark_inode_dirty_sync(wdata->inode); + mark_inode_dirty_sync(inode); } EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit); diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index bf80503200f5..22ee70586875 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -641,12 +641,14 @@ nfs_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, static int nfs_read_done(struct rpc_task *task, struct nfs_read_data *data) { + struct inode *inode = data->header->inode; + if (nfs_async_handle_expired_key(task)) return -EAGAIN; - nfs_invalidate_atime(data->inode); + nfs_invalidate_atime(inode); if (task->tk_status >= 0) { - nfs_refresh_inode(data->inode, data->res.fattr); + nfs_refresh_inode(inode, data->res.fattr); /* Emulate the eof flag, which isn't normally needed in NFSv2 * as it is guaranteed to always return the file attributes */ @@ -668,11 +670,13 @@ static void nfs_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_dat static int nfs_write_done(struct rpc_task *task, struct nfs_write_data *data) { + struct inode *inode = data->header->inode; + if (nfs_async_handle_expired_key(task)) return -EAGAIN; if (task->tk_status >= 0) - nfs_post_op_update_inode_force_wcc(data->inode, data->res.fattr); + nfs_post_op_update_inode_force_wcc(inode, data->res.fattr); return 0; } diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 4ddba6706347..d6d46823d9e4 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -35,19 +35,24 @@ static const struct rpc_call_ops nfs_read_full_ops; static struct kmem_cache *nfs_rdata_cachep; -struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount) +struct nfs_read_header *nfs_readhdr_alloc(unsigned int pagecount) { - struct nfs_read_data *p; + struct nfs_read_header *p; p = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); if (p) { - INIT_LIST_HEAD(&p->pages); - p->npages = pagecount; - if (pagecount <= ARRAY_SIZE(p->page_array)) - p->pagevec = p->page_array; + struct nfs_pgio_header *hdr = &p->header; + struct nfs_read_data *data = &p->rpc_data; + + INIT_LIST_HEAD(&hdr->pages); + INIT_LIST_HEAD(&data->list); + data->npages = pagecount; + data->header = hdr; + if (pagecount <= ARRAY_SIZE(data->page_array)) + data->pagevec = data->page_array; else { - p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); - if (!p->pagevec) { + data->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); + if (!data->pagevec) { kmem_cache_free(nfs_rdata_cachep, p); p = NULL; } @@ -56,17 +61,19 @@ struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount) return p; } -void nfs_readdata_free(struct nfs_read_data *p) +void nfs_readhdr_free(struct nfs_pgio_header *hdr) { - if (p && (p->pagevec != &p->page_array[0])) - kfree(p->pagevec); - kmem_cache_free(nfs_rdata_cachep, p); + struct nfs_read_header *rhdr = container_of(hdr, struct nfs_read_header, header); + + kmem_cache_free(nfs_rdata_cachep, rhdr); } void nfs_readdata_release(struct nfs_read_data *rdata) { put_nfs_open_context(rdata->args.context); - nfs_readdata_free(rdata); + if (rdata->pagevec != rdata->page_array) + kfree(rdata->pagevec); + nfs_readhdr_free(rdata->header); } static @@ -173,13 +180,13 @@ int nfs_initiate_read(struct rpc_clnt *clnt, struct nfs_read_data *data, const struct rpc_call_ops *call_ops) { - struct inode *inode = data->inode; + struct inode *inode = data->header->inode; int swap_flags = IS_SWAPFILE(inode) ? NFS_RPC_SWAPFLAGS : 0; struct rpc_task *task; struct rpc_message msg = { .rpc_argp = &data->args, .rpc_resp = &data->res, - .rpc_cred = data->cred, + .rpc_cred = data->header->cred, }; struct rpc_task_setup task_setup_data = { .task = &data->task, @@ -216,11 +223,11 @@ EXPORT_SYMBOL_GPL(nfs_initiate_read); static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, unsigned int count, unsigned int offset) { - struct inode *inode = req->wb_context->dentry->d_inode; + struct inode *inode = data->header->inode; - data->req = req; - data->inode = inode; - data->cred = req->wb_context->cred; + data->header->req = req; + data->header->inode = inode; + data->header->cred = req->wb_context->cred; data->args.fh = NFS_FH(inode); data->args.offset = req_offset(req) + offset; @@ -239,7 +246,7 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, static int nfs_do_read(struct nfs_read_data *data, const struct rpc_call_ops *call_ops) { - struct inode *inode = data->args.context->dentry->d_inode; + struct inode *inode = data->header->inode; return nfs_initiate_read(NFS_CLIENT(inode), data, call_ops); } @@ -293,6 +300,7 @@ static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head { struct nfs_page *req = nfs_list_entry(desc->pg_list.next); struct page *page = req->wb_page; + struct nfs_read_header *rhdr; struct nfs_read_data *data; size_t rsize = desc->pg_bsize, nbytes; unsigned int offset; @@ -306,9 +314,10 @@ static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head do { size_t len = min(nbytes,rsize); - data = nfs_readdata_alloc(1); - if (!data) + rhdr = nfs_readhdr_alloc(1); + if (!rhdr) goto out_bad; + data = &rhdr->rpc_data; data->pagevec[0] = page; nfs_read_rpcsetup(req, data, len, offset); list_add(&data->list, res); @@ -333,26 +342,28 @@ static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, struct list_head * { struct nfs_page *req; struct page **pages; + struct nfs_read_header *rhdr; struct nfs_read_data *data; struct list_head *head = &desc->pg_list; int ret = 0; - data = nfs_readdata_alloc(nfs_page_array_len(desc->pg_base, - desc->pg_count)); - if (!data) { + rhdr = nfs_readhdr_alloc(nfs_page_array_len(desc->pg_base, + desc->pg_count)); + if (!rhdr) { nfs_async_read_error(head); ret = -ENOMEM; goto out; } + data = &rhdr->rpc_data; pages = data->pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_list_add_request(req, &data->pages); + nfs_list_add_request(req, &rhdr->header.pages); *pages++ = req->wb_page; } - req = nfs_list_entry(data->pages.next); + req = nfs_list_entry(rhdr->header.pages.next); nfs_read_rpcsetup(req, data, desc->pg_count, 0); list_add(&data->list, res); @@ -390,20 +401,21 @@ static const struct nfs_pageio_ops nfs_pageio_read_ops = { */ int nfs_readpage_result(struct rpc_task *task, struct nfs_read_data *data) { + struct inode *inode = data->header->inode; int status; dprintk("NFS: %s: %5u, (status %d)\n", __func__, task->tk_pid, task->tk_status); - status = NFS_PROTO(data->inode)->read_done(task, data); + status = NFS_PROTO(inode)->read_done(task, data); if (status != 0) return status; - nfs_add_stats(data->inode, NFSIOS_SERVERREADBYTES, data->res.count); + nfs_add_stats(inode, NFSIOS_SERVERREADBYTES, data->res.count); if (task->tk_status == -ESTALE) { - set_bit(NFS_INO_STALE, &NFS_I(data->inode)->flags); - nfs_mark_for_revalidate(data->inode); + set_bit(NFS_INO_STALE, &NFS_I(inode)->flags); + nfs_mark_for_revalidate(inode); } return 0; } @@ -417,7 +429,7 @@ static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data return; /* This is a short read! */ - nfs_inc_stats(data->inode, NFSIOS_SHORTREAD); + nfs_inc_stats(data->header->inode, NFSIOS_SHORTREAD); /* Has the server at least made some progress? */ if (resp->count == 0) return; @@ -449,7 +461,7 @@ static void nfs_readpage_result_partial(struct rpc_task *task, void *calldata) static void nfs_readpage_release_partial(void *calldata) { struct nfs_read_data *data = calldata; - struct nfs_page *req = data->req; + struct nfs_page *req = data->header->req; struct page *page = req->wb_page; int status = data->task.tk_status; @@ -461,13 +473,13 @@ static void nfs_readpage_release_partial(void *calldata) SetPageUptodate(page); nfs_readpage_release(req); } - nfs_readdata_release(calldata); + nfs_readdata_release(data); } void nfs_read_prepare(struct rpc_task *task, void *calldata) { struct nfs_read_data *data = calldata; - NFS_PROTO(data->inode)->read_rpc_prepare(task, data); + NFS_PROTO(data->header->inode)->read_rpc_prepare(task, data); } static const struct rpc_call_ops nfs_read_partial_ops = { @@ -524,9 +536,10 @@ static void nfs_readpage_result_full(struct rpc_task *task, void *calldata) static void nfs_readpage_release_full(void *calldata) { struct nfs_read_data *data = calldata; + struct nfs_pgio_header *hdr = data->header; - while (!list_empty(&data->pages)) { - struct nfs_page *req = nfs_list_entry(data->pages.next); + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); nfs_list_remove_request(req); nfs_readpage_release(req); @@ -685,7 +698,7 @@ out: int __init nfs_init_readpagecache(void) { nfs_rdata_cachep = kmem_cache_create("nfs_read_data", - sizeof(struct nfs_read_data), + sizeof(struct nfs_read_header), 0, SLAB_HWCACHE_ALIGN, NULL); if (nfs_rdata_cachep == NULL) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 76735dd8c9a7..dbb5c0a613b8 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -69,19 +69,24 @@ void nfs_commit_free(struct nfs_commit_data *p) } EXPORT_SYMBOL_GPL(nfs_commit_free); -struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount) +struct nfs_write_header *nfs_writehdr_alloc(unsigned int pagecount) { - struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, GFP_NOFS); + struct nfs_write_header *p = mempool_alloc(nfs_wdata_mempool, GFP_NOFS); if (p) { + struct nfs_pgio_header *hdr = &p->header; + struct nfs_write_data *data = &p->rpc_data; + memset(p, 0, sizeof(*p)); - INIT_LIST_HEAD(&p->pages); - p->npages = pagecount; - if (pagecount <= ARRAY_SIZE(p->page_array)) - p->pagevec = p->page_array; + INIT_LIST_HEAD(&hdr->pages); + INIT_LIST_HEAD(&data->list); + data->npages = pagecount; + data->header = hdr; + if (pagecount <= ARRAY_SIZE(data->page_array)) + data->pagevec = data->page_array; else { - p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS); - if (!p->pagevec) { + data->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS); + if (!data->pagevec) { mempool_free(p, nfs_wdata_mempool); p = NULL; } @@ -90,17 +95,18 @@ struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount) return p; } -void nfs_writedata_free(struct nfs_write_data *p) +void nfs_writehdr_free(struct nfs_pgio_header *hdr) { - if (p && (p->pagevec != &p->page_array[0])) - kfree(p->pagevec); - mempool_free(p, nfs_wdata_mempool); + struct nfs_write_header *whdr = container_of(hdr, struct nfs_write_header, header); + mempool_free(whdr, nfs_wdata_mempool); } void nfs_writedata_release(struct nfs_write_data *wdata) { put_nfs_open_context(wdata->args.context); - nfs_writedata_free(wdata); + if (wdata->pagevec != wdata->page_array) + kfree(wdata->pagevec); + nfs_writehdr_free(wdata->header); } static void nfs_context_set_write_error(struct nfs_open_context *ctx, int error) @@ -507,9 +513,8 @@ static inline int nfs_write_need_commit(struct nfs_write_data *data) { if (data->verf.committed == NFS_DATA_SYNC) - return data->lseg == NULL; - else - return data->verf.committed != NFS_FILE_SYNC; + return data->header->lseg == NULL; + return data->verf.committed != NFS_FILE_SYNC; } static inline @@ -517,7 +522,7 @@ int nfs_reschedule_unstable_write(struct nfs_page *req, struct nfs_write_data *data) { if (test_and_clear_bit(PG_NEED_COMMIT, &req->wb_flags)) { - nfs_mark_request_commit(req, data->lseg); + nfs_mark_request_commit(req, data->header->lseg); return 1; } if (test_and_clear_bit(PG_NEED_RESCHED, &req->wb_flags)) { @@ -841,13 +846,13 @@ int nfs_initiate_write(struct rpc_clnt *clnt, const struct rpc_call_ops *call_ops, int how) { - struct inode *inode = data->inode; + struct inode *inode = data->header->inode; int priority = flush_task_priority(how); struct rpc_task *task; struct rpc_message msg = { .rpc_argp = &data->args, .rpc_resp = &data->res, - .rpc_cred = data->cred, + .rpc_cred = data->header->cred, }; struct rpc_task_setup task_setup_data = { .rpc_client = clnt, @@ -896,14 +901,15 @@ static void nfs_write_rpcsetup(struct nfs_page *req, unsigned int count, unsigned int offset, int how) { + struct nfs_pgio_header *hdr = data->header; struct inode *inode = req->wb_context->dentry->d_inode; /* Set up the RPC argument and reply structs * NB: take care not to mess about with data->commit et al. */ - data->req = req; - data->inode = inode = req->wb_context->dentry->d_inode; - data->cred = req->wb_context->cred; + hdr->req = req; + hdr->inode = inode = req->wb_context->dentry->d_inode; + hdr->cred = req->wb_context->cred; data->args.fh = NFS_FH(inode); data->args.offset = req_offset(req) + offset; @@ -935,7 +941,7 @@ static int nfs_do_write(struct nfs_write_data *data, const struct rpc_call_ops *call_ops, int how) { - struct inode *inode = data->args.context->dentry->d_inode; + struct inode *inode = data->header->inode; return nfs_initiate_write(NFS_CLIENT(inode), data, call_ops, how); } @@ -981,6 +987,7 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, struct list_head { struct nfs_page *req = nfs_list_entry(desc->pg_list.next); struct page *page = req->wb_page; + struct nfs_write_header *whdr; struct nfs_write_data *data; size_t wsize = desc->pg_bsize, nbytes; unsigned int offset; @@ -1000,9 +1007,10 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, struct list_head do { size_t len = min(nbytes, wsize); - data = nfs_writedata_alloc(1); - if (!data) + whdr = nfs_writehdr_alloc(1); + if (!whdr) goto out_bad; + data = &whdr->rpc_data; data->pagevec[0] = page; nfs_write_rpcsetup(req, data, len, offset, desc->pg_ioflags); list_add(&data->list, res); @@ -1036,13 +1044,14 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc, struct list_head *r { struct nfs_page *req; struct page **pages; + struct nfs_write_header *whdr; struct nfs_write_data *data; struct list_head *head = &desc->pg_list; int ret = 0; - data = nfs_writedata_alloc(nfs_page_array_len(desc->pg_base, - desc->pg_count)); - if (!data) { + whdr = nfs_writehdr_alloc(nfs_page_array_len(desc->pg_base, + desc->pg_count)); + if (!whdr) { while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); @@ -1051,14 +1060,15 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc, struct list_head *r ret = -ENOMEM; goto out; } + data = &whdr->rpc_data; pages = data->pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_list_add_request(req, &data->pages); + nfs_list_add_request(req, &whdr->header.pages); *pages++ = req->wb_page; } - req = nfs_list_entry(data->pages.next); + req = nfs_list_entry(whdr->header.pages.next); if ((desc->pg_ioflags & FLUSH_COND_STABLE) && (desc->pg_moreio || NFS_I(desc->pg_inode)->ncommit)) @@ -1126,10 +1136,11 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata) dprintk("NFS: %5u write(%s/%lld %d@%lld)", task->tk_pid, - data->req->wb_context->dentry->d_inode->i_sb->s_id, + data->header->inode->i_sb->s_id, (long long) - NFS_FILEID(data->req->wb_context->dentry->d_inode), - data->req->wb_bytes, (long long)req_offset(data->req)); + NFS_FILEID(data->header->inode), + data->header->req->wb_bytes, + (long long)req_offset(data->header->req)); nfs_writeback_done(task, data); } @@ -1137,7 +1148,7 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata) static void nfs_writeback_release_partial(void *calldata) { struct nfs_write_data *data = calldata; - struct nfs_page *req = data->req; + struct nfs_page *req = data->header->req; struct page *page = req->wb_page; int status = data->task.tk_status; @@ -1169,13 +1180,13 @@ static void nfs_writeback_release_partial(void *calldata) out: if (atomic_dec_and_test(&req->wb_complete)) nfs_writepage_release(req, data); - nfs_writedata_release(calldata); + nfs_writedata_release(data); } void nfs_write_prepare(struct rpc_task *task, void *calldata) { struct nfs_write_data *data = calldata; - NFS_PROTO(data->inode)->write_rpc_prepare(task, data); + NFS_PROTO(data->header->inode)->write_rpc_prepare(task, data); } void nfs_commit_prepare(struct rpc_task *task, void *calldata) @@ -1208,11 +1219,12 @@ static void nfs_writeback_done_full(struct rpc_task *task, void *calldata) static void nfs_writeback_release_full(void *calldata) { struct nfs_write_data *data = calldata; + struct nfs_pgio_header *hdr = data->header; int status = data->task.tk_status; /* Update attributes as result of writeback. */ - while (!list_empty(&data->pages)) { - struct nfs_page *req = nfs_list_entry(data->pages.next); + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); struct page *page = req->wb_page; nfs_list_remove_request(req); @@ -1233,7 +1245,7 @@ static void nfs_writeback_release_full(void *calldata) if (nfs_write_need_commit(data)) { memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf)); - nfs_mark_request_commit(req, data->lseg); + nfs_mark_request_commit(req, hdr->lseg); dprintk(" marked for commit\n"); goto next; } @@ -1244,7 +1256,7 @@ remove_request: nfs_unlock_request(req); nfs_end_page_writeback(page); } - nfs_writedata_release(calldata); + nfs_writedata_release(data); } static const struct rpc_call_ops nfs_write_full_ops = { @@ -1261,6 +1273,7 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) { struct nfs_writeargs *argp = &data->args; struct nfs_writeres *resp = &data->res; + struct inode *inode = data->header->inode; int status; dprintk("NFS: %5u nfs_writeback_done (status %d)\n", @@ -1273,10 +1286,10 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) * another writer had changed the file, but some applications * depend on tighter cache coherency when writing. */ - status = NFS_PROTO(data->inode)->write_done(task, data); + status = NFS_PROTO(inode)->write_done(task, data); if (status != 0) return; - nfs_add_stats(data->inode, NFSIOS_SERVERWRITTENBYTES, resp->count); + nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, resp->count); #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) if (resp->verf->committed < argp->stable && task->tk_status >= 0) { @@ -1294,7 +1307,7 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) if (time_before(complain, jiffies)) { dprintk("NFS: faulty NFS server %s:" " (committed = %d) != (stable = %d)\n", - NFS_SERVER(data->inode)->nfs_client->cl_hostname, + NFS_SERVER(inode)->nfs_client->cl_hostname, resp->verf->committed, argp->stable); complain = jiffies + 300 * HZ; } @@ -1304,7 +1317,7 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) if (task->tk_status >= 0 && resp->count < argp->count) { static unsigned long complain; - nfs_inc_stats(data->inode, NFSIOS_SHORTWRITE); + nfs_inc_stats(inode, NFSIOS_SHORTWRITE); /* Has the server at least made some progress? */ if (resp->count != 0) { @@ -1333,7 +1346,6 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) /* Can't do anything about it except throw an error. */ task->tk_status = -EIO; } - return; } @@ -1745,7 +1757,7 @@ int nfs_migrate_page(struct address_space *mapping, struct page *newpage, int __init nfs_init_writepagecache(void) { nfs_wdata_cachep = kmem_cache_create("nfs_write_data", - sizeof(struct nfs_write_data), + sizeof(struct nfs_write_header), 0, SLAB_HWCACHE_ALIGN, NULL); if (nfs_wdata_cachep == NULL) diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index d5d68f322bf0..8d3a2b804201 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -568,12 +568,6 @@ nfs_have_writebacks(struct inode *inode) return NFS_I(inode)->npages != 0; } -/* - * Allocate nfs_write_data structures - */ -extern struct nfs_write_data *nfs_writedata_alloc(unsigned int npages); -extern void nfs_writedata_free(struct nfs_write_data *); - /* * linux/fs/nfs/read.c */ @@ -584,12 +578,6 @@ extern int nfs_readpage_result(struct rpc_task *, struct nfs_read_data *); extern int nfs_readpage_async(struct nfs_open_context *, struct inode *, struct page *); -/* - * Allocate nfs_read_data structures - */ -extern struct nfs_read_data *nfs_readdata_alloc(unsigned int npages); -extern void nfs_readdata_free(struct nfs_read_data *); - /* * linux/fs/nfs3proc.c */ diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 8fb036a0d489..fee324175391 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1168,52 +1168,58 @@ struct nfs_page; #define NFS_PAGEVEC_SIZE (8U) struct nfs_read_data { + struct nfs_pgio_header *header; + struct list_head list; struct rpc_task task; - struct inode *inode; - struct rpc_cred *cred; struct nfs_fattr fattr; /* fattr storage */ - struct list_head pages; /* Coalesced read requests */ - struct list_head list; /* lists of struct nfs_read_data */ - struct nfs_page *req; /* multi ops per nfs_page */ struct page **pagevec; unsigned int npages; /* Max length of pagevec */ struct nfs_readargs args; struct nfs_readres res; unsigned long timestamp; /* For lease renewal */ - struct pnfs_layout_segment *lseg; - struct nfs_client *ds_clp; /* pNFS data server */ - const struct rpc_call_ops *mds_ops; int (*read_done_cb) (struct rpc_task *task, struct nfs_read_data *data); __u64 mds_offset; - int pnfs_error; struct page *page_array[NFS_PAGEVEC_SIZE]; + struct nfs_client *ds_clp; /* pNFS data server */ +}; + +struct nfs_pgio_header { + struct inode *inode; + struct rpc_cred *cred; + struct list_head pages; + struct nfs_page *req; + struct pnfs_layout_segment *lseg; + const struct rpc_call_ops *mds_ops; + int pnfs_error; +}; + +struct nfs_read_header { + struct nfs_pgio_header header; + struct nfs_read_data rpc_data; }; struct nfs_direct_req; struct nfs_write_data { + struct nfs_pgio_header *header; + struct list_head list; struct rpc_task task; - struct inode *inode; - struct rpc_cred *cred; struct nfs_fattr fattr; struct nfs_writeverf verf; - struct list_head pages; /* Coalesced requests we wish to flush */ - struct list_head list; /* lists of struct nfs_write_data */ - struct nfs_page *req; /* multi ops per nfs_page */ struct page **pagevec; unsigned int npages; /* Max length of pagevec */ struct nfs_writeargs args; /* argument struct */ struct nfs_writeres res; /* result struct */ - struct pnfs_layout_segment *lseg; - struct nfs_client *ds_clp; /* pNFS data server */ - const struct rpc_call_ops *mds_ops; - int (*write_done_cb) (struct rpc_task *task, struct nfs_write_data *data); -#ifdef CONFIG_NFS_V4 unsigned long timestamp; /* For lease renewal */ -#endif + int (*write_done_cb) (struct rpc_task *task, struct nfs_write_data *data); __u64 mds_offset; /* Filelayout dense stripe */ - int pnfs_error; struct page *page_array[NFS_PAGEVEC_SIZE]; + struct nfs_client *ds_clp; /* pNFS data server */ +}; + +struct nfs_write_header { + struct nfs_pgio_header header; + struct nfs_write_data rpc_data; }; struct nfs_commit_data { -- cgit v1.2.3 From 30dd374f6fc1b202db3a1b57b61afff1326bad92 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:45 -0400 Subject: NFS: create struct nfs_page_array Both nfs_read_data and nfs_write_data devote several fields which can be combined into a single shared struct. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/blocklayout.c | 11 ++++++----- fs/nfs/direct.c | 40 +++++++++++++++++++++++----------------- fs/nfs/internal.h | 1 + fs/nfs/pagelist.c | 13 +++++++++++++ fs/nfs/read.c | 22 ++++++++-------------- fs/nfs/write.c | 22 ++++++++-------------- include/linux/nfs_xdr.h | 14 ++++++++------ 7 files changed, 67 insertions(+), 56 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c index 7a482517f4c6..7ae8a608956f 100644 --- a/fs/nfs/blocklayout/blocklayout.c +++ b/fs/nfs/blocklayout/blocklayout.c @@ -242,7 +242,7 @@ bl_read_pagelist(struct nfs_read_data *rdata) int pg_index = rdata->args.pgbase >> PAGE_CACHE_SHIFT; dprintk("%s enter nr_pages %u offset %lld count %u\n", __func__, - rdata->npages, f_offset, (unsigned int)rdata->args.count); + rdata->pages.npages, f_offset, (unsigned int)rdata->args.count); par = alloc_parallel(rdata); if (!par) @@ -252,7 +252,7 @@ bl_read_pagelist(struct nfs_read_data *rdata) isect = (sector_t) (f_offset >> SECTOR_SHIFT); /* Code assumes extents are page-aligned */ - for (i = pg_index; i < rdata->npages; i++) { + for (i = pg_index; i < rdata->pages.npages; i++) { if (!extent_length) { /* We've used up the previous extent */ bl_put_extent(be); @@ -285,7 +285,8 @@ bl_read_pagelist(struct nfs_read_data *rdata) struct pnfs_block_extent *be_read; be_read = (hole && cow_read) ? cow_read : be; - bio = bl_add_page_to_bio(bio, rdata->npages - i, READ, + bio = bl_add_page_to_bio(bio, rdata->pages.npages - i, + READ, isect, pages[i], be_read, bl_end_io_read, par); if (IS_ERR(bio)) { @@ -654,7 +655,7 @@ next_page: /* Middle pages */ pg_index = wdata->args.pgbase >> PAGE_CACHE_SHIFT; - for (i = pg_index; i < wdata->npages; i++) { + for (i = pg_index; i < wdata->pages.npages; i++) { if (!extent_length) { /* We've used up the previous extent */ bl_put_extent(be); @@ -688,7 +689,7 @@ next_page: goto out; } } - bio = bl_add_page_to_bio(bio, wdata->npages - i, WRITE, + bio = bl_add_page_to_bio(bio, wdata->pages.npages - i, WRITE, isect, pages[i], be, bl_end_io_write, par); if (IS_ERR(bio)) { diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 56176af1436f..0faba4cb531d 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -252,11 +252,11 @@ static void nfs_direct_read_release(void *calldata) } else { dreq->count += data->res.count; spin_unlock(&dreq->lock); - nfs_direct_dirty_pages(data->pagevec, + nfs_direct_dirty_pages(data->pages.pagevec, data->args.pgbase, data->res.count); } - nfs_direct_release_pages(data->pagevec, data->npages); + nfs_direct_release_pages(data->pages.pagevec, data->pages.npages); if (put_dreq(dreq)) nfs_direct_complete(dreq); @@ -273,8 +273,8 @@ static void nfs_direct_readhdr_release(struct nfs_read_header *rhdr) { struct nfs_read_data *data = &rhdr->rpc_data; - if (data->pagevec != data->page_array) - kfree(data->pagevec); + if (data->pages.pagevec != data->pages.page_array) + kfree(data->pages.pagevec); nfs_readhdr_free(&rhdr->header); } @@ -312,6 +312,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, do { struct nfs_read_header *rhdr; struct nfs_read_data *data; + struct nfs_page_array *pages; size_t bytes; pgbase = user_addr & ~PAGE_MASK; @@ -322,24 +323,25 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, if (unlikely(!rhdr)) break; data = &rhdr->rpc_data; + pages = &data->pages; down_read(¤t->mm->mmap_sem); result = get_user_pages(current, current->mm, user_addr, - data->npages, 1, 0, data->pagevec, NULL); + pages->npages, 1, 0, pages->pagevec, NULL); up_read(¤t->mm->mmap_sem); if (result < 0) { nfs_direct_readhdr_release(rhdr); break; } - if ((unsigned)result < data->npages) { + if ((unsigned)result < pages->npages) { bytes = result * PAGE_SIZE; if (bytes <= pgbase) { - nfs_direct_release_pages(data->pagevec, result); + nfs_direct_release_pages(pages->pagevec, result); nfs_direct_readhdr_release(rhdr); break; } bytes -= pgbase; - data->npages = result; + pages->npages = result; } get_dreq(dreq); @@ -352,7 +354,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, data->args.lock_context = dreq->l_ctx; data->args.offset = pos; data->args.pgbase = pgbase; - data->args.pages = data->pagevec; + data->args.pages = pages->pagevec; data->args.count = bytes; data->res.fattr = &data->fattr; data->res.eof = 0; @@ -462,8 +464,8 @@ static void nfs_direct_writehdr_release(struct nfs_write_header *whdr) { struct nfs_write_data *data = &whdr->rpc_data; - if (data->pagevec != data->page_array) - kfree(data->pagevec); + if (data->pages.pagevec != data->pages.page_array) + kfree(data->pages.pagevec); nfs_writehdr_free(&whdr->header); } @@ -472,8 +474,10 @@ static void nfs_direct_free_writedata(struct nfs_direct_req *dreq) while (!list_empty(&dreq->rewrite_list)) { struct nfs_pgio_header *hdr = list_entry(dreq->rewrite_list.next, struct nfs_pgio_header, pages); struct nfs_write_header *whdr = container_of(hdr, struct nfs_write_header, header); + struct nfs_page_array *p = &whdr->rpc_data.pages; + list_del(&hdr->pages); - nfs_direct_release_pages(whdr->rpc_data.pagevec, whdr->rpc_data.npages); + nfs_direct_release_pages(p->pagevec, p->npages); nfs_direct_writehdr_release(whdr); } } @@ -751,6 +755,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, do { struct nfs_write_header *whdr; struct nfs_write_data *data; + struct nfs_page_array *pages; size_t bytes; pgbase = user_addr & ~PAGE_MASK; @@ -762,24 +767,25 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, break; data = &whdr->rpc_data; + pages = &data->pages; down_read(¤t->mm->mmap_sem); result = get_user_pages(current, current->mm, user_addr, - data->npages, 0, 0, data->pagevec, NULL); + pages->npages, 0, 0, pages->pagevec, NULL); up_read(¤t->mm->mmap_sem); if (result < 0) { nfs_direct_writehdr_release(whdr); break; } - if ((unsigned)result < data->npages) { + if ((unsigned)result < pages->npages) { bytes = result * PAGE_SIZE; if (bytes <= pgbase) { - nfs_direct_release_pages(data->pagevec, result); + nfs_direct_release_pages(pages->pagevec, result); nfs_direct_writehdr_release(whdr); break; } bytes -= pgbase; - data->npages = result; + pages->npages = result; } get_dreq(dreq); @@ -794,7 +800,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, data->args.lock_context = dreq->l_ctx; data->args.offset = pos; data->args.pgbase = pgbase; - data->args.pages = data->pagevec; + data->args.pages = pages->pagevec; data->args.count = bytes; data->args.stable = sync; data->res.fattr = &data->fattr; diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 7dc9be1a6e1a..5c3d77fda560 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -210,6 +210,7 @@ extern void nfs_destroy_writepagecache(void); extern int __init nfs_init_directcache(void); extern void nfs_destroy_directcache(void); +extern bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount); /* nfs2xdr.c */ extern int nfs_stat_to_errno(enum nfs_stat); diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index d21fceaa9f62..d349bd4c48db 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -26,6 +26,19 @@ static struct kmem_cache *nfs_page_cachep; +bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount) +{ + p->npages = pagecount; + if (pagecount <= ARRAY_SIZE(p->page_array)) + p->pagevec = p->page_array; + else { + p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); + if (!p->pagevec) + p->npages = 0; + } + return p->pagevec != NULL; +} + static inline struct nfs_page * nfs_page_alloc(void) { diff --git a/fs/nfs/read.c b/fs/nfs/read.c index d6d46823d9e4..f6ab30b5a462 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -46,16 +46,10 @@ struct nfs_read_header *nfs_readhdr_alloc(unsigned int pagecount) INIT_LIST_HEAD(&hdr->pages); INIT_LIST_HEAD(&data->list); - data->npages = pagecount; data->header = hdr; - if (pagecount <= ARRAY_SIZE(data->page_array)) - data->pagevec = data->page_array; - else { - data->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); - if (!data->pagevec) { - kmem_cache_free(nfs_rdata_cachep, p); - p = NULL; - } + if (!nfs_pgarray_set(&data->pages, pagecount)) { + kmem_cache_free(nfs_rdata_cachep, p); + p = NULL; } } return p; @@ -71,8 +65,8 @@ void nfs_readhdr_free(struct nfs_pgio_header *hdr) void nfs_readdata_release(struct nfs_read_data *rdata) { put_nfs_open_context(rdata->args.context); - if (rdata->pagevec != rdata->page_array) - kfree(rdata->pagevec); + if (rdata->pages.pagevec != rdata->pages.page_array) + kfree(rdata->pages.pagevec); nfs_readhdr_free(rdata->header); } @@ -232,7 +226,7 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, data->args.fh = NFS_FH(inode); data->args.offset = req_offset(req) + offset; data->args.pgbase = req->wb_pgbase + offset; - data->args.pages = data->pagevec; + data->args.pages = data->pages.pagevec; data->args.count = count; data->args.context = get_nfs_open_context(req->wb_context); data->args.lock_context = req->wb_lock_context; @@ -318,7 +312,7 @@ static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head if (!rhdr) goto out_bad; data = &rhdr->rpc_data; - data->pagevec[0] = page; + data->pages.pagevec[0] = page; nfs_read_rpcsetup(req, data, len, offset); list_add(&data->list, res); requests++; @@ -356,7 +350,7 @@ static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, struct list_head * } data = &rhdr->rpc_data; - pages = data->pagevec; + pages = data->pages.pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index dbb5c0a613b8..2efae049b4f0 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -80,16 +80,10 @@ struct nfs_write_header *nfs_writehdr_alloc(unsigned int pagecount) memset(p, 0, sizeof(*p)); INIT_LIST_HEAD(&hdr->pages); INIT_LIST_HEAD(&data->list); - data->npages = pagecount; data->header = hdr; - if (pagecount <= ARRAY_SIZE(data->page_array)) - data->pagevec = data->page_array; - else { - data->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS); - if (!data->pagevec) { - mempool_free(p, nfs_wdata_mempool); - p = NULL; - } + if (!nfs_pgarray_set(&data->pages, pagecount)) { + mempool_free(p, nfs_wdata_mempool); + p = NULL; } } return p; @@ -104,8 +98,8 @@ void nfs_writehdr_free(struct nfs_pgio_header *hdr) void nfs_writedata_release(struct nfs_write_data *wdata) { put_nfs_open_context(wdata->args.context); - if (wdata->pagevec != wdata->page_array) - kfree(wdata->pagevec); + if (wdata->pages.pagevec != wdata->pages.page_array) + kfree(wdata->pages.pagevec); nfs_writehdr_free(wdata->header); } @@ -916,7 +910,7 @@ static void nfs_write_rpcsetup(struct nfs_page *req, /* pnfs_set_layoutcommit needs this */ data->mds_offset = data->args.offset; data->args.pgbase = req->wb_pgbase + offset; - data->args.pages = data->pagevec; + data->args.pages = data->pages.pagevec; data->args.count = count; data->args.context = get_nfs_open_context(req->wb_context); data->args.lock_context = req->wb_lock_context; @@ -1011,7 +1005,7 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, struct list_head if (!whdr) goto out_bad; data = &whdr->rpc_data; - data->pagevec[0] = page; + data->pages.pagevec[0] = page; nfs_write_rpcsetup(req, data, len, offset, desc->pg_ioflags); list_add(&data->list, res); requests++; @@ -1061,7 +1055,7 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc, struct list_head *r goto out; } data = &whdr->rpc_data; - pages = data->pagevec; + pages = data->pages.pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index fee324175391..e34beaf86e9c 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1167,19 +1167,23 @@ struct nfs_page; #define NFS_PAGEVEC_SIZE (8U) +struct nfs_page_array { + struct page **pagevec; + unsigned int npages; /* Max length of pagevec */ + struct page *page_array[NFS_PAGEVEC_SIZE]; +}; + struct nfs_read_data { struct nfs_pgio_header *header; struct list_head list; struct rpc_task task; struct nfs_fattr fattr; /* fattr storage */ - struct page **pagevec; - unsigned int npages; /* Max length of pagevec */ struct nfs_readargs args; struct nfs_readres res; unsigned long timestamp; /* For lease renewal */ int (*read_done_cb) (struct rpc_task *task, struct nfs_read_data *data); __u64 mds_offset; - struct page *page_array[NFS_PAGEVEC_SIZE]; + struct nfs_page_array pages; struct nfs_client *ds_clp; /* pNFS data server */ }; @@ -1206,14 +1210,12 @@ struct nfs_write_data { struct rpc_task task; struct nfs_fattr fattr; struct nfs_writeverf verf; - struct page **pagevec; - unsigned int npages; /* Max length of pagevec */ struct nfs_writeargs args; /* argument struct */ struct nfs_writeres res; /* result struct */ unsigned long timestamp; /* For lease renewal */ int (*write_done_cb) (struct rpc_task *task, struct nfs_write_data *data); __u64 mds_offset; /* Filelayout dense stripe */ - struct page *page_array[NFS_PAGEVEC_SIZE]; + struct nfs_page_array pages; struct nfs_client *ds_clp; /* pNFS data server */ }; -- cgit v1.2.3 From 4db6e0b74c0f6dfc2f9c0690e8df512e3b635983 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:46 -0400 Subject: NFS: merge _full and _partial read rpc_ops Decouple nfs_pgio_header and nfs_read_data, and have (possibly multiple) nfs_read_datas each take a refcount on nfs_pgio_header. For the moment keeps nfs_read_header as a way to preallocate a single nfs_read_data with the nfs_pgio_header. The code doesn't need this, and would be prettier without, but given the amount of churn I am already introducing I didn't want to play with tuning new mempools. This also fixes bug in pnfs_ld_handle_read_error. In the case of desc->pg_bsize < PAGE_CACHE_SIZE, the pages list was empty, causing replay attempt to do nothing. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 10 +- fs/nfs/internal.h | 15 ++- fs/nfs/nfs4filelayout.c | 1 - fs/nfs/nfs4proc.c | 2 - fs/nfs/pagelist.c | 24 ++++ fs/nfs/pnfs.c | 55 +++++--- fs/nfs/read.c | 338 ++++++++++++++++++++++------------------------- include/linux/nfs_page.h | 1 - include/linux/nfs_xdr.h | 16 +++ 9 files changed, 252 insertions(+), 210 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 0faba4cb531d..90b00ce42cbe 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -319,10 +319,16 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, bytes = min(rsize,count); result = -ENOMEM; - rhdr = nfs_readhdr_alloc(nfs_page_array_len(pgbase, bytes)); + rhdr = nfs_readhdr_alloc(); if (unlikely(!rhdr)) break; - data = &rhdr->rpc_data; + data = nfs_readdata_alloc(&rhdr->header, nfs_page_array_len(pgbase, bytes)); + if (!data) { + nfs_readhdr_free(&rhdr->header); + break; + } + data->header = &rhdr->header; + atomic_inc(&data->header->refcnt); pages = &data->pages; down_read(¤t->mm->mmap_sem); diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 5c3d77fda560..33af5e51c0bb 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -200,6 +200,7 @@ struct vfsmount *nfs_do_refmount(struct rpc_clnt *client, struct dentry *dentry) extern struct svc_version nfs4_callback_version1; extern struct svc_version nfs4_callback_version4; +struct nfs_pageio_descriptor; /* pagelist.c */ extern int __init nfs_init_nfspagecache(void); extern void nfs_destroy_nfspagecache(void); @@ -211,6 +212,10 @@ extern void nfs_destroy_writepagecache(void); extern int __init nfs_init_directcache(void); extern void nfs_destroy_directcache(void); extern bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount); +extern void nfs_pgheader_init(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr, + void (*release)(struct nfs_pgio_header *hdr)); +void nfs_set_pgio_error(struct nfs_pgio_header *hdr, int error, loff_t pos); /* nfs2xdr.c */ extern int nfs_stat_to_errno(enum nfs_stat); @@ -295,17 +300,19 @@ extern struct dentry *nfs4_get_root(struct super_block *, struct nfs_fh *, extern int nfs4_get_rootfh(struct nfs_server *server, struct nfs_fh *mntfh); #endif -struct nfs_pageio_descriptor; /* read.c */ -extern struct nfs_read_header *nfs_readhdr_alloc(unsigned int npages); +extern void nfs_async_read_error(struct list_head *head); +extern struct nfs_read_header *nfs_readhdr_alloc(void); extern void nfs_readhdr_free(struct nfs_pgio_header *hdr); +extern void nfs_read_completion(struct nfs_pgio_header *hdr); +extern struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount); extern int nfs_initiate_read(struct rpc_clnt *clnt, struct nfs_read_data *data, const struct rpc_call_ops *call_ops); extern void nfs_read_prepare(struct rpc_task *task, void *calldata); extern int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, - struct list_head *head); - + struct nfs_pgio_header *hdr); extern void nfs_pageio_init_read_mds(struct nfs_pageio_descriptor *pgio, struct inode *inode); extern void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio); diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index ad1d68013a5b..333e765f3ac2 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -227,7 +227,6 @@ static void filelayout_read_release(void *data) { struct nfs_read_data *rdata = data; - put_lseg(rdata->header->lseg); rdata->header->mds_ops->rpc_release(data); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 5375862075de..ce31ab22bc55 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3391,8 +3391,6 @@ void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data) struct inode *inode = hdr->inode; dprintk("%s Reset task for i/o through\n", __func__); - put_lseg(hdr->lseg); - hdr->lseg = NULL; data->ds_clp = NULL; /* offsets will differ in the dense stripe case */ data->args.offset = data->mds_offset; diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index d349bd4c48db..cd4c038135a7 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -39,6 +39,30 @@ bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount) return p->pagevec != NULL; } +void nfs_pgheader_init(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr, + void (*release)(struct nfs_pgio_header *hdr)) +{ + hdr->req = nfs_list_entry(desc->pg_list.next); + hdr->inode = desc->pg_inode; + hdr->cred = hdr->req->wb_context->cred; + hdr->io_start = req_offset(hdr->req); + hdr->good_bytes = desc->pg_count; + hdr->release = release; +} + +void nfs_set_pgio_error(struct nfs_pgio_header *hdr, int error, loff_t pos) +{ + spin_lock(&hdr->lock); + if (pos < hdr->io_start + hdr->good_bytes) { + set_bit(NFS_IOHDR_ERROR, &hdr->flags); + clear_bit(NFS_IOHDR_EOF, &hdr->flags); + hdr->good_bytes = pos - hdr->io_start; + hdr->error = error; + } + spin_unlock(&hdr->lock); +} + static inline struct nfs_page * nfs_page_alloc(void) { diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index d705da427e6d..d1a91dbe7654 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1333,7 +1333,9 @@ static void pnfs_ld_handle_read_error(struct nfs_read_data *data) clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(hdr->inode)->flags); pnfs_return_layout(hdr->inode); } - data->task.tk_status = pnfs_read_done_resend_to_mds(hdr->inode, &hdr->pages); + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) + data->task.tk_status = pnfs_read_done_resend_to_mds(hdr->inode, + &hdr->pages); } /* @@ -1348,7 +1350,6 @@ void pnfs_ld_read_done(struct nfs_read_data *data) hdr->mds_ops->rpc_call_done(&data->task, data); } else pnfs_ld_handle_read_error(data); - put_lseg(hdr->lseg); hdr->mds_ops->rpc_release(data); } EXPORT_SYMBOL_GPL(pnfs_ld_read_done); @@ -1359,11 +1360,11 @@ pnfs_read_through_mds(struct nfs_pageio_descriptor *desc, { struct nfs_pgio_header *hdr = data->header; - list_splice_tail_init(&hdr->pages, &desc->pg_list); - if (hdr->req && list_empty(&hdr->req->wb_list)) - nfs_list_add_request(hdr->req, &desc->pg_list); - nfs_pageio_reset_read_mds(desc); - desc->pg_recoalesce = 1; + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) { + list_splice_tail_init(&hdr->pages, &desc->pg_list); + nfs_pageio_reset_read_mds(desc); + desc->pg_recoalesce = 1; + } nfs_readdata_release(data); } @@ -1381,18 +1382,13 @@ pnfs_try_to_read_data(struct nfs_read_data *rdata, enum pnfs_try_status trypnfs; hdr->mds_ops = call_ops; - hdr->lseg = get_lseg(lseg); dprintk("%s: Reading ino:%lu %u@%llu\n", __func__, inode->i_ino, rdata->args.count, rdata->args.offset); trypnfs = nfss->pnfs_curr_ld->read_pagelist(rdata); - if (trypnfs == PNFS_NOT_ATTEMPTED) { - put_lseg(hdr->lseg); - hdr->lseg = NULL; - } else { + if (trypnfs != PNFS_NOT_ATTEMPTED) nfs_inc_stats(inode, NFSIOS_PNFS_READ); - } dprintk("%s End (trypnfs:%d)\n", __func__, trypnfs); return trypnfs; } @@ -1408,7 +1404,7 @@ pnfs_do_multiple_reads(struct nfs_pageio_descriptor *desc, struct list_head *hea while (!list_empty(head)) { enum pnfs_try_status trypnfs; - data = list_entry(head->next, struct nfs_read_data, list); + data = list_first_entry(head, struct nfs_read_data, list); list_del_init(&data->list); trypnfs = pnfs_try_to_read_data(data, call_ops, lseg); @@ -1418,20 +1414,41 @@ pnfs_do_multiple_reads(struct nfs_pageio_descriptor *desc, struct list_head *hea put_lseg(lseg); } +static void pnfs_readhdr_free(struct nfs_pgio_header *hdr) +{ + put_lseg(hdr->lseg); + nfs_readhdr_free(hdr); +} + int pnfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_read_header *rhdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_pagein(desc, &head); - if (ret != 0) { + rhdr = nfs_readhdr_alloc(); + if (!rhdr) { + nfs_async_read_error(&desc->pg_list); + ret = -ENOMEM; put_lseg(desc->pg_lseg); desc->pg_lseg = NULL; return ret; } - pnfs_do_multiple_reads(desc, &head); - return 0; + hdr = &rhdr->header; + nfs_pgheader_init(desc, hdr, pnfs_readhdr_free); + hdr->lseg = get_lseg(desc->pg_lseg); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_pagein(desc, hdr); + if (ret != 0) { + put_lseg(desc->pg_lseg); + desc->pg_lseg = NULL; + set_bit(NFS_IOHDR_REDO, &hdr->flags); + } else + pnfs_do_multiple_reads(desc, &hdr->rpc_list); + if (atomic_dec_and_test(&hdr->refcnt)) + nfs_read_completion(hdr); + return ret; } EXPORT_SYMBOL_GPL(pnfs_generic_pg_readpages); diff --git a/fs/nfs/read.c b/fs/nfs/read.c index f6ab30b5a462..c9633b2501bd 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -30,29 +30,49 @@ #define NFSDBG_FACILITY NFSDBG_PAGECACHE static const struct nfs_pageio_ops nfs_pageio_read_ops; -static const struct rpc_call_ops nfs_read_partial_ops; -static const struct rpc_call_ops nfs_read_full_ops; +static const struct rpc_call_ops nfs_read_common_ops; static struct kmem_cache *nfs_rdata_cachep; -struct nfs_read_header *nfs_readhdr_alloc(unsigned int pagecount) +struct nfs_read_header *nfs_readhdr_alloc() { - struct nfs_read_header *p; + struct nfs_read_header *rhdr; - p = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); - if (p) { - struct nfs_pgio_header *hdr = &p->header; - struct nfs_read_data *data = &p->rpc_data; + rhdr = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); + if (rhdr) { + struct nfs_pgio_header *hdr = &rhdr->header; INIT_LIST_HEAD(&hdr->pages); - INIT_LIST_HEAD(&data->list); + INIT_LIST_HEAD(&hdr->rpc_list); + spin_lock_init(&hdr->lock); + atomic_set(&hdr->refcnt, 0); + } + return rhdr; +} + +struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount) +{ + struct nfs_read_data *data, *prealloc; + + prealloc = &container_of(hdr, struct nfs_read_header, header)->rpc_data; + if (prealloc->header == NULL) + data = prealloc; + else + data = kzalloc(sizeof(*data), GFP_KERNEL); + if (!data) + goto out; + + if (nfs_pgarray_set(&data->pages, pagecount)) { data->header = hdr; - if (!nfs_pgarray_set(&data->pages, pagecount)) { - kmem_cache_free(nfs_rdata_cachep, p); - p = NULL; - } + atomic_inc(&hdr->refcnt); + } else { + if (data != prealloc) + kfree(data); + data = NULL; } - return p; +out: + return data; } void nfs_readhdr_free(struct nfs_pgio_header *hdr) @@ -64,10 +84,18 @@ void nfs_readhdr_free(struct nfs_pgio_header *hdr) void nfs_readdata_release(struct nfs_read_data *rdata) { + struct nfs_pgio_header *hdr = rdata->header; + struct nfs_read_header *read_header = container_of(hdr, struct nfs_read_header, header); + put_nfs_open_context(rdata->args.context); if (rdata->pages.pagevec != rdata->pages.page_array) kfree(rdata->pages.pagevec); - nfs_readhdr_free(rdata->header); + if (rdata != &read_header->rpc_data) + kfree(rdata); + else + rdata->header = NULL; + if (atomic_dec_and_test(&hdr->refcnt)) + nfs_read_completion(hdr); } static @@ -79,35 +107,6 @@ int nfs_return_empty_page(struct page *page) return 0; } -static void nfs_readpage_truncate_uninitialised_page(struct nfs_read_data *data) -{ - unsigned int remainder = data->args.count - data->res.count; - unsigned int base = data->args.pgbase + data->res.count; - unsigned int pglen; - struct page **pages; - - if (data->res.eof == 0 || remainder == 0) - return; - /* - * Note: "remainder" can never be negative, since we check for - * this in the XDR code. - */ - pages = &data->args.pages[base >> PAGE_CACHE_SHIFT]; - base &= ~PAGE_CACHE_MASK; - pglen = PAGE_CACHE_SIZE - base; - for (;;) { - if (remainder <= pglen) { - zero_user(*pages, base, remainder); - break; - } - zero_user(*pages, base, pglen); - pages++; - remainder -= pglen; - pglen = PAGE_CACHE_SIZE; - base = 0; - } -} - void nfs_pageio_init_read_mds(struct nfs_pageio_descriptor *pgio, struct inode *inode) { @@ -170,6 +169,46 @@ static void nfs_readpage_release(struct nfs_page *req) nfs_release_request(req); } +/* Note io was page aligned */ +void nfs_read_completion(struct nfs_pgio_header *hdr) +{ + unsigned long bytes = 0; + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out; + if (!test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { + if (bytes > hdr->good_bytes) + zero_user(page, 0, PAGE_SIZE); + else if (hdr->good_bytes - bytes < PAGE_SIZE) + zero_user_segment(page, + hdr->good_bytes & ~PAGE_MASK, + PAGE_SIZE); + } + SetPageUptodate(page); + nfs_list_remove_request(req); + nfs_readpage_release(req); + bytes += PAGE_SIZE; + } + } else { + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + + bytes += req->wb_bytes; + if (bytes <= hdr->good_bytes) + SetPageUptodate(req->wb_page); + nfs_list_remove_request(req); + nfs_readpage_release(req); + } + } +out: + hdr->release(hdr); +} + int nfs_initiate_read(struct rpc_clnt *clnt, struct nfs_read_data *data, const struct rpc_call_ops *call_ops) @@ -214,16 +253,12 @@ EXPORT_SYMBOL_GPL(nfs_initiate_read); /* * Set up the NFS read request struct */ -static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, +static void nfs_read_rpcsetup(struct nfs_read_data *data, unsigned int count, unsigned int offset) { - struct inode *inode = data->header->inode; - - data->header->req = req; - data->header->inode = inode; - data->header->cred = req->wb_context->cred; + struct nfs_page *req = data->header->req; - data->args.fh = NFS_FH(inode); + data->args.fh = NFS_FH(data->header->inode); data->args.offset = req_offset(req) + offset; data->args.pgbase = req->wb_pgbase + offset; data->args.pages = data->pages.pagevec; @@ -255,7 +290,7 @@ nfs_do_multiple_reads(struct list_head *head, while (!list_empty(head)) { int ret2; - data = list_entry(head->next, struct nfs_read_data, list); + data = list_first_entry(head, struct nfs_read_data, list); list_del_init(&data->list); ret2 = nfs_do_read(data, call_ops); @@ -265,7 +300,7 @@ nfs_do_multiple_reads(struct list_head *head, return ret; } -static void +void nfs_async_read_error(struct list_head *head) { struct nfs_page *req; @@ -290,11 +325,11 @@ nfs_async_read_error(struct list_head *head) * won't see the new data until our attribute cache is updated. This is more * or less conventional NFS client behavior. */ -static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head *res) +static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { - struct nfs_page *req = nfs_list_entry(desc->pg_list.next); + struct nfs_page *req = hdr->req; struct page *page = req->wb_page; - struct nfs_read_header *rhdr; struct nfs_read_data *data; size_t rsize = desc->pg_bsize, nbytes; unsigned int offset; @@ -302,85 +337,97 @@ static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head int ret = 0; nfs_list_remove_request(req); + nfs_list_add_request(req, &hdr->pages); offset = 0; nbytes = desc->pg_count; do { size_t len = min(nbytes,rsize); - rhdr = nfs_readhdr_alloc(1); - if (!rhdr) + data = nfs_readdata_alloc(hdr, 1); + if (!data) goto out_bad; - data = &rhdr->rpc_data; data->pages.pagevec[0] = page; - nfs_read_rpcsetup(req, data, len, offset); - list_add(&data->list, res); + nfs_read_rpcsetup(data, len, offset); + list_add(&data->list, &hdr->rpc_list); requests++; nbytes -= len; offset += len; } while(nbytes != 0); - atomic_set(&req->wb_complete, requests); - desc->pg_rpc_callops = &nfs_read_partial_ops; + desc->pg_rpc_callops = &nfs_read_common_ops; return ret; out_bad: - while (!list_empty(res)) { - data = list_entry(res->next, struct nfs_read_data, list); + while (!list_empty(&hdr->rpc_list)) { + data = list_first_entry(&hdr->rpc_list, struct nfs_read_data, list); list_del(&data->list); nfs_readdata_release(data); } - nfs_readpage_release(req); + nfs_async_read_error(&hdr->pages); return -ENOMEM; } -static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, struct list_head *res) +static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { struct nfs_page *req; struct page **pages; - struct nfs_read_header *rhdr; - struct nfs_read_data *data; + struct nfs_read_data *data; struct list_head *head = &desc->pg_list; int ret = 0; - rhdr = nfs_readhdr_alloc(nfs_page_array_len(desc->pg_base, - desc->pg_count)); - if (!rhdr) { + data = nfs_readdata_alloc(hdr, nfs_page_array_len(desc->pg_base, + desc->pg_count)); + if (!data) { nfs_async_read_error(head); ret = -ENOMEM; goto out; } - data = &rhdr->rpc_data; pages = data->pages.pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_list_add_request(req, &rhdr->header.pages); + nfs_list_add_request(req, &hdr->pages); *pages++ = req->wb_page; } - req = nfs_list_entry(rhdr->header.pages.next); - nfs_read_rpcsetup(req, data, desc->pg_count, 0); - list_add(&data->list, res); - desc->pg_rpc_callops = &nfs_read_full_ops; + nfs_read_rpcsetup(data, desc->pg_count, 0); + list_add(&data->list, &hdr->rpc_list); + desc->pg_rpc_callops = &nfs_read_common_ops; out: return ret; } -int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, struct list_head *head) +int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { if (desc->pg_bsize < PAGE_CACHE_SIZE) - return nfs_pagein_multi(desc, head); - return nfs_pagein_one(desc, head); + return nfs_pagein_multi(desc, hdr); + return nfs_pagein_one(desc, hdr); } static int nfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_read_header *rhdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_pagein(desc, &head); + rhdr = nfs_readhdr_alloc(); + if (!rhdr) { + nfs_async_read_error(&desc->pg_list); + return -ENOMEM; + } + hdr = &rhdr->header; + nfs_pgheader_init(desc, hdr, nfs_readhdr_free); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_pagein(desc, hdr); if (ret == 0) - ret = nfs_do_multiple_reads(&head, desc->pg_rpc_callops); + ret = nfs_do_multiple_reads(&hdr->rpc_list, + desc->pg_rpc_callops); + else + set_bit(NFS_IOHDR_REDO, &hdr->flags); + if (atomic_dec_and_test(&hdr->refcnt)) + nfs_read_completion(hdr); return ret; } @@ -419,15 +466,13 @@ static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data struct nfs_readargs *argp = &data->args; struct nfs_readres *resp = &data->res; - if (resp->eof || resp->count == argp->count) - return; - /* This is a short read! */ nfs_inc_stats(data->header->inode, NFSIOS_SHORTREAD); /* Has the server at least made some progress? */ - if (resp->count == 0) + if (resp->count == 0) { + nfs_set_pgio_error(data->header, -EIO, argp->offset); return; - + } /* Yes, so retry the read at the end of the data */ data->mds_offset += resp->count; argp->offset += resp->count; @@ -436,38 +481,34 @@ static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data rpc_restart_call_prepare(task); } -/* - * Handle a read reply that fills part of a page. - */ -static void nfs_readpage_result_partial(struct rpc_task *task, void *calldata) +static void nfs_readpage_result_common(struct rpc_task *task, void *calldata) { struct nfs_read_data *data = calldata; - + struct nfs_pgio_header *hdr = data->header; + + /* Note the only returns of nfs_readpage_result are 0 and -EAGAIN */ if (nfs_readpage_result(task, data) != 0) return; if (task->tk_status < 0) - return; - - nfs_readpage_truncate_uninitialised_page(data); - nfs_readpage_retry(task, data); + nfs_set_pgio_error(hdr, task->tk_status, data->args.offset); + else if (data->res.eof) { + loff_t bound; + + bound = data->args.offset + data->res.count; + spin_lock(&hdr->lock); + if (bound < hdr->io_start + hdr->good_bytes) { + set_bit(NFS_IOHDR_EOF, &hdr->flags); + clear_bit(NFS_IOHDR_ERROR, &hdr->flags); + hdr->good_bytes = bound - hdr->io_start; + } + spin_unlock(&hdr->lock); + } else if (data->res.count != data->args.count) + nfs_readpage_retry(task, data); } -static void nfs_readpage_release_partial(void *calldata) +static void nfs_readpage_release_common(void *calldata) { - struct nfs_read_data *data = calldata; - struct nfs_page *req = data->header->req; - struct page *page = req->wb_page; - int status = data->task.tk_status; - - if (status < 0) - set_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags); - - if (atomic_dec_and_test(&req->wb_complete)) { - if (!test_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags)) - SetPageUptodate(page); - nfs_readpage_release(req); - } - nfs_readdata_release(data); + nfs_readdata_release(calldata); } void nfs_read_prepare(struct rpc_task *task, void *calldata) @@ -476,75 +517,10 @@ void nfs_read_prepare(struct rpc_task *task, void *calldata) NFS_PROTO(data->header->inode)->read_rpc_prepare(task, data); } -static const struct rpc_call_ops nfs_read_partial_ops = { - .rpc_call_prepare = nfs_read_prepare, - .rpc_call_done = nfs_readpage_result_partial, - .rpc_release = nfs_readpage_release_partial, -}; - -static void nfs_readpage_set_pages_uptodate(struct nfs_read_data *data) -{ - unsigned int count = data->res.count; - unsigned int base = data->args.pgbase; - struct page **pages; - - if (data->res.eof) - count = data->args.count; - if (unlikely(count == 0)) - return; - pages = &data->args.pages[base >> PAGE_CACHE_SHIFT]; - base &= ~PAGE_CACHE_MASK; - count += base; - for (;count >= PAGE_CACHE_SIZE; count -= PAGE_CACHE_SIZE, pages++) - SetPageUptodate(*pages); - if (count == 0) - return; - /* Was this a short read? */ - if (data->res.eof || data->res.count == data->args.count) - SetPageUptodate(*pages); -} - -/* - * This is the callback from RPC telling us whether a reply was - * received or some error occurred (timeout or socket shutdown). - */ -static void nfs_readpage_result_full(struct rpc_task *task, void *calldata) -{ - struct nfs_read_data *data = calldata; - - if (nfs_readpage_result(task, data) != 0) - return; - if (task->tk_status < 0) - return; - /* - * Note: nfs_readpage_retry may change the values of - * data->args. In the multi-page case, we therefore need - * to ensure that we call nfs_readpage_set_pages_uptodate() - * first. - */ - nfs_readpage_truncate_uninitialised_page(data); - nfs_readpage_set_pages_uptodate(data); - nfs_readpage_retry(task, data); -} - -static void nfs_readpage_release_full(void *calldata) -{ - struct nfs_read_data *data = calldata; - struct nfs_pgio_header *hdr = data->header; - - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); - - nfs_list_remove_request(req); - nfs_readpage_release(req); - } - nfs_readdata_release(calldata); -} - -static const struct rpc_call_ops nfs_read_full_ops = { +static const struct rpc_call_ops nfs_read_common_ops = { .rpc_call_prepare = nfs_read_prepare, - .rpc_call_done = nfs_readpage_result_full, - .rpc_release = nfs_readpage_release_full, + .rpc_call_done = nfs_readpage_result_common, + .rpc_release = nfs_readpage_release_common, }; /* diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index eac30d6bec17..5c520344d8ad 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -27,7 +27,6 @@ enum { PG_CLEAN, PG_NEED_COMMIT, PG_NEED_RESCHED, - PG_PARTIAL_READ_FAILED, PG_COMMIT_TO_DS, }; diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index e34beaf86e9c..164862148ba0 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1187,14 +1187,30 @@ struct nfs_read_data { struct nfs_client *ds_clp; /* pNFS data server */ }; +/* used as flag bits in nfs_pgio_header */ +enum { + NFS_IOHDR_ERROR = 0, + NFS_IOHDR_EOF, + NFS_IOHDR_REDO, +}; + struct nfs_pgio_header { struct inode *inode; struct rpc_cred *cred; struct list_head pages; + struct list_head rpc_list; + atomic_t refcnt; struct nfs_page *req; struct pnfs_layout_segment *lseg; + loff_t io_start; const struct rpc_call_ops *mds_ops; + void (*release) (struct nfs_pgio_header *hdr); + spinlock_t lock; + /* fields protected by lock */ int pnfs_error; + int error; /* merge with pnfs_error */ + unsigned long good_bytes; /* boundary of good data */ + unsigned long flags; }; struct nfs_read_header { -- cgit v1.2.3 From 6c75dc0d498caa402fb17b1bf769835a9db875c8 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:47 -0400 Subject: NFS: merge _full and _partial write rpc_ops Decouple nfs_pgio_header and nfs_write_data, and have (possibly multiple) nfs_write_datas each take a refcount on nfs_pgio_header. For the moment keeps nfs_write_header as a way to preallocate a single nfs_write_data with the nfs_pgio_header. The code doesn't need this, and would be prettier without, but given the amount of churn I am already introducing I didn't want to play with tuning new mempools. This also fixes bug in pnfs_ld_handle_write_error. In the case of desc->pg_bsize < PAGE_CACHE_SIZE, the pages list was empty, causing replay attempt to do nothing. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 10 +- fs/nfs/internal.h | 8 +- fs/nfs/nfs4filelayout.c | 1 - fs/nfs/nfs4proc.c | 4 +- fs/nfs/pnfs.c | 58 +++++--- fs/nfs/write.c | 383 ++++++++++++++++++++++-------------------------- include/linux/nfs_xdr.h | 2 + 7 files changed, 227 insertions(+), 239 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 90b00ce42cbe..22a40c408449 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -768,11 +768,17 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, bytes = min(wsize,count); result = -ENOMEM; - whdr = nfs_writehdr_alloc(nfs_page_array_len(pgbase, bytes)); + whdr = nfs_writehdr_alloc(); if (unlikely(!whdr)) break; - data = &whdr->rpc_data; + data = nfs_writedata_alloc(&whdr->header, nfs_page_array_len(pgbase, bytes)); + if (!data) { + nfs_writehdr_free(&whdr->header); + break; + } + data->header = &whdr->header; + atomic_inc(&data->header->refcnt); pages = &data->pages; down_read(¤t->mm->mmap_sem); diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 33af5e51c0bb..16bc9c47c83e 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -319,10 +319,14 @@ extern void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio); extern void nfs_readdata_release(struct nfs_read_data *rdata); /* write.c */ -extern struct nfs_write_header *nfs_writehdr_alloc(unsigned int npages); +extern void nfs_async_write_error(struct list_head *head); +extern struct nfs_write_header *nfs_writehdr_alloc(void); extern void nfs_writehdr_free(struct nfs_pgio_header *hdr); +extern struct nfs_write_data *nfs_writedata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount); +extern void nfs_write_completion(struct nfs_pgio_header *hdr); extern int nfs_generic_flush(struct nfs_pageio_descriptor *desc, - struct list_head *head); + struct nfs_pgio_header *hdr); extern void nfs_pageio_init_write_mds(struct nfs_pageio_descriptor *pgio, struct inode *inode, int ioflags); extern void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio); diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 333e765f3ac2..02d8170ce0f3 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -314,7 +314,6 @@ static void filelayout_write_release(void *data) { struct nfs_write_data *wdata = data; - put_lseg(wdata->header->lseg); wdata->header->mds_ops->rpc_release(data); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index ce31ab22bc55..87af80d28a82 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3431,8 +3431,6 @@ void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data) struct inode *inode = hdr->inode; dprintk("%s Reset task for i/o through\n", __func__); - put_lseg(hdr->lseg); - hdr->lseg = NULL; data->ds_clp = NULL; data->write_done_cb = nfs4_write_done_cb; data->args.fh = NFS_FH(inode); @@ -3448,7 +3446,7 @@ static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_messag { struct nfs_server *server = NFS_SERVER(data->header->inode); - if (data->header->lseg) { + if (data->ds_clp) { data->args.bitmask = NULL; data->res.fattr = NULL; } else diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index d1a91dbe7654..d515f00614cd 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1199,7 +1199,9 @@ static void pnfs_ld_handle_write_error(struct nfs_write_data *data) clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(hdr->inode)->flags); pnfs_return_layout(hdr->inode); } - data->task.tk_status = pnfs_write_done_resend_to_mds(hdr->inode, &hdr->pages); + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) + data->task.tk_status = pnfs_write_done_resend_to_mds(hdr->inode, + &hdr->pages); } /* @@ -1214,7 +1216,6 @@ void pnfs_ld_write_done(struct nfs_write_data *data) hdr->mds_ops->rpc_call_done(&data->task, data); } else pnfs_ld_handle_write_error(data); - put_lseg(hdr->lseg); hdr->mds_ops->rpc_release(data); } EXPORT_SYMBOL_GPL(pnfs_ld_write_done); @@ -1225,12 +1226,11 @@ pnfs_write_through_mds(struct nfs_pageio_descriptor *desc, { struct nfs_pgio_header *hdr = data->header; - list_splice_tail_init(&hdr->pages, &desc->pg_list); - if (hdr->req && list_empty(&hdr->req->wb_list)) - nfs_list_add_request(hdr->req, &desc->pg_list); - nfs_pageio_reset_write_mds(desc); - desc->pg_recoalesce = 1; - put_lseg(hdr->lseg); + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) { + list_splice_tail_init(&hdr->pages, &desc->pg_list); + nfs_pageio_reset_write_mds(desc); + desc->pg_recoalesce = 1; + } nfs_writedata_release(data); } @@ -1246,18 +1246,12 @@ pnfs_try_to_write_data(struct nfs_write_data *wdata, struct nfs_server *nfss = NFS_SERVER(inode); hdr->mds_ops = call_ops; - hdr->lseg = get_lseg(lseg); dprintk("%s: Writing ino:%lu %u@%llu (how %d)\n", __func__, inode->i_ino, wdata->args.count, wdata->args.offset, how); - trypnfs = nfss->pnfs_curr_ld->write_pagelist(wdata, how); - if (trypnfs == PNFS_NOT_ATTEMPTED) { - put_lseg(hdr->lseg); - hdr->lseg = NULL; - } else + if (trypnfs != PNFS_NOT_ATTEMPTED) nfs_inc_stats(inode, NFSIOS_PNFS_WRITE); - dprintk("%s End (trypnfs:%d)\n", __func__, trypnfs); return trypnfs; } @@ -1273,7 +1267,7 @@ pnfs_do_multiple_writes(struct nfs_pageio_descriptor *desc, struct list_head *he while (!list_empty(head)) { enum pnfs_try_status trypnfs; - data = list_entry(head->next, struct nfs_write_data, list); + data = list_first_entry(head, struct nfs_write_data, list); list_del_init(&data->list); trypnfs = pnfs_try_to_write_data(data, call_ops, lseg, how); @@ -1283,20 +1277,40 @@ pnfs_do_multiple_writes(struct nfs_pageio_descriptor *desc, struct list_head *he put_lseg(lseg); } +static void pnfs_writehdr_free(struct nfs_pgio_header *hdr) +{ + put_lseg(hdr->lseg); + nfs_writehdr_free(hdr); +} + int pnfs_generic_pg_writepages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_write_header *whdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_flush(desc, &head); - if (ret != 0) { + whdr = nfs_writehdr_alloc(); + if (!whdr) { + nfs_async_write_error(&desc->pg_list); put_lseg(desc->pg_lseg); desc->pg_lseg = NULL; - return ret; + return -ENOMEM; } - pnfs_do_multiple_writes(desc, &head, desc->pg_ioflags); - return 0; + hdr = &whdr->header; + nfs_pgheader_init(desc, hdr, pnfs_writehdr_free); + hdr->lseg = get_lseg(desc->pg_lseg); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_flush(desc, hdr); + if (ret != 0) { + put_lseg(desc->pg_lseg); + desc->pg_lseg = NULL; + set_bit(NFS_IOHDR_REDO, &hdr->flags); + } else + pnfs_do_multiple_writes(desc, &hdr->rpc_list, desc->pg_ioflags); + if (atomic_dec_and_test(&hdr->refcnt)) + nfs_write_completion(hdr); + return ret; } EXPORT_SYMBOL_GPL(pnfs_generic_pg_writepages); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 2efae049b4f0..076075eb676c 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -42,8 +42,7 @@ static void nfs_pageio_init_write(struct nfs_pageio_descriptor *desc, struct inode *inode, int ioflags); static void nfs_redirty_request(struct nfs_page *req); -static const struct rpc_call_ops nfs_write_partial_ops; -static const struct rpc_call_ops nfs_write_full_ops; +static const struct rpc_call_ops nfs_write_common_ops; static const struct rpc_call_ops nfs_commit_ops; static struct kmem_cache *nfs_wdata_cachep; @@ -69,26 +68,47 @@ void nfs_commit_free(struct nfs_commit_data *p) } EXPORT_SYMBOL_GPL(nfs_commit_free); -struct nfs_write_header *nfs_writehdr_alloc(unsigned int pagecount) +struct nfs_write_header *nfs_writehdr_alloc(void) { struct nfs_write_header *p = mempool_alloc(nfs_wdata_mempool, GFP_NOFS); if (p) { struct nfs_pgio_header *hdr = &p->header; - struct nfs_write_data *data = &p->rpc_data; memset(p, 0, sizeof(*p)); INIT_LIST_HEAD(&hdr->pages); - INIT_LIST_HEAD(&data->list); - data->header = hdr; - if (!nfs_pgarray_set(&data->pages, pagecount)) { - mempool_free(p, nfs_wdata_mempool); - p = NULL; - } + INIT_LIST_HEAD(&hdr->rpc_list); + spin_lock_init(&hdr->lock); + atomic_set(&hdr->refcnt, 0); } return p; } +struct nfs_write_data *nfs_writedata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount) +{ + struct nfs_write_data *data, *prealloc; + + prealloc = &container_of(hdr, struct nfs_write_header, header)->rpc_data; + if (prealloc->header == NULL) + data = prealloc; + else + data = kzalloc(sizeof(*data), GFP_KERNEL); + if (!data) + goto out; + + if (nfs_pgarray_set(&data->pages, pagecount)) { + data->header = hdr; + atomic_inc(&hdr->refcnt); + } else { + if (data != prealloc) + kfree(data); + data = NULL; + } +out: + return data; +} + void nfs_writehdr_free(struct nfs_pgio_header *hdr) { struct nfs_write_header *whdr = container_of(hdr, struct nfs_write_header, header); @@ -97,10 +117,18 @@ void nfs_writehdr_free(struct nfs_pgio_header *hdr) void nfs_writedata_release(struct nfs_write_data *wdata) { + struct nfs_pgio_header *hdr = wdata->header; + struct nfs_write_header *write_header = container_of(hdr, struct nfs_write_header, header); + put_nfs_open_context(wdata->args.context); if (wdata->pages.pagevec != wdata->pages.page_array) kfree(wdata->pages.pagevec); - nfs_writehdr_free(wdata->header); + if (wdata != &write_header->rpc_data) + kfree(wdata); + else + wdata->header = NULL; + if (atomic_dec_and_test(&hdr->refcnt)) + nfs_write_completion(hdr); } static void nfs_context_set_write_error(struct nfs_open_context *ctx, int error) @@ -511,20 +539,6 @@ int nfs_write_need_commit(struct nfs_write_data *data) return data->verf.committed != NFS_FILE_SYNC; } -static inline -int nfs_reschedule_unstable_write(struct nfs_page *req, - struct nfs_write_data *data) -{ - if (test_and_clear_bit(PG_NEED_COMMIT, &req->wb_flags)) { - nfs_mark_request_commit(req, data->header->lseg); - return 1; - } - if (test_and_clear_bit(PG_NEED_RESCHED, &req->wb_flags)) { - nfs_mark_request_dirty(req); - return 1; - } - return 0; -} #else static void nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg) @@ -542,13 +556,43 @@ int nfs_write_need_commit(struct nfs_write_data *data) return 0; } -static inline -int nfs_reschedule_unstable_write(struct nfs_page *req, - struct nfs_write_data *data) +#endif + +void nfs_write_completion(struct nfs_pgio_header *hdr) { - return 0; + unsigned long bytes = 0; + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out; + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + bytes += req->wb_bytes; + nfs_list_remove_request(req); + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && + (hdr->good_bytes < bytes)) { + nfs_set_pageerror(page); + nfs_context_set_write_error(req->wb_context, hdr->error); + goto remove_req; + } + if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) { + nfs_mark_request_dirty(req); + goto next; + } + if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) { + nfs_mark_request_commit(req, hdr->lseg); + goto next; + } +remove_req: + nfs_inode_remove_request(req); +next: + nfs_unlock_request(req); + nfs_end_page_writeback(page); + } +out: + hdr->release(hdr); } -#endif #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) static int @@ -813,17 +857,6 @@ int nfs_updatepage(struct file *file, struct page *page, return status; } -static void nfs_writepage_release(struct nfs_page *req, - struct nfs_write_data *data) -{ - struct page *page = req->wb_page; - - if (PageError(req->wb_page) || !nfs_reschedule_unstable_write(req, data)) - nfs_inode_remove_request(req); - nfs_unlock_request(req); - nfs_end_page_writeback(page); -} - static int flush_task_priority(int how) { switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) { @@ -890,22 +923,16 @@ EXPORT_SYMBOL_GPL(nfs_initiate_write); /* * Set up the argument/result storage required for the RPC call. */ -static void nfs_write_rpcsetup(struct nfs_page *req, - struct nfs_write_data *data, +static void nfs_write_rpcsetup(struct nfs_write_data *data, unsigned int count, unsigned int offset, int how) { - struct nfs_pgio_header *hdr = data->header; - struct inode *inode = req->wb_context->dentry->d_inode; + struct nfs_page *req = data->header->req; /* Set up the RPC argument and reply structs * NB: take care not to mess about with data->commit et al. */ - hdr->req = req; - hdr->inode = inode = req->wb_context->dentry->d_inode; - hdr->cred = req->wb_context->cred; - - data->args.fh = NFS_FH(inode); + data->args.fh = NFS_FH(data->header->inode); data->args.offset = req_offset(req) + offset; /* pnfs_set_layoutcommit needs this */ data->mds_offset = data->args.offset; @@ -919,7 +946,7 @@ static void nfs_write_rpcsetup(struct nfs_page *req, case 0: break; case FLUSH_COND_STABLE: - if (nfs_need_commit(NFS_I(inode))) + if (nfs_need_commit(NFS_I(data->header->inode))) break; default: data->args.stable = NFS_FILE_SYNC; @@ -950,7 +977,7 @@ static int nfs_do_multiple_writes(struct list_head *head, while (!list_empty(head)) { int ret2; - data = list_entry(head->next, struct nfs_write_data, list); + data = list_first_entry(head, struct nfs_write_data, list); list_del_init(&data->list); ret2 = nfs_do_write(data, call_ops, how); @@ -973,15 +1000,26 @@ static void nfs_redirty_request(struct nfs_page *req) nfs_end_page_writeback(page); } +void nfs_async_write_error(struct list_head *head) +{ + struct nfs_page *req; + + while (!list_empty(head)) { + req = nfs_list_entry(head->next); + nfs_list_remove_request(req); + nfs_redirty_request(req); + } +} + /* * Generate multiple small requests to write out a single * contiguous dirty area on one page. */ -static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, struct list_head *res) +static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { - struct nfs_page *req = nfs_list_entry(desc->pg_list.next); + struct nfs_page *req = hdr->req; struct page *page = req->wb_page; - struct nfs_write_header *whdr; struct nfs_write_data *data; size_t wsize = desc->pg_bsize, nbytes; unsigned int offset; @@ -989,6 +1027,7 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, struct list_head int ret = 0; nfs_list_remove_request(req); + nfs_list_add_request(req, &hdr->pages); if ((desc->pg_ioflags & FLUSH_COND_STABLE) && (desc->pg_moreio || NFS_I(desc->pg_inode)->ncommit || @@ -1001,28 +1040,27 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, struct list_head do { size_t len = min(nbytes, wsize); - whdr = nfs_writehdr_alloc(1); - if (!whdr) + data = nfs_writedata_alloc(hdr, 1); + if (!data) goto out_bad; - data = &whdr->rpc_data; data->pages.pagevec[0] = page; - nfs_write_rpcsetup(req, data, len, offset, desc->pg_ioflags); - list_add(&data->list, res); + nfs_write_rpcsetup(data, len, offset, desc->pg_ioflags); + list_add(&data->list, &hdr->rpc_list); requests++; nbytes -= len; offset += len; } while (nbytes != 0); atomic_set(&req->wb_complete, requests); - desc->pg_rpc_callops = &nfs_write_partial_ops; + desc->pg_rpc_callops = &nfs_write_common_ops; return ret; out_bad: - while (!list_empty(res)) { - data = list_entry(res->next, struct nfs_write_data, list); + while (!list_empty(&hdr->rpc_list)) { + data = list_first_entry(&hdr->rpc_list, struct nfs_write_data, list); list_del(&data->list); nfs_writedata_release(data); } - nfs_redirty_request(req); + nfs_async_write_error(&hdr->pages); return -ENOMEM; } @@ -1034,64 +1072,74 @@ out_bad: * This is the case if nfs_updatepage detects a conflicting request * that has been written but not committed. */ -static int nfs_flush_one(struct nfs_pageio_descriptor *desc, struct list_head *res) +static int nfs_flush_one(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { struct nfs_page *req; struct page **pages; - struct nfs_write_header *whdr; struct nfs_write_data *data; struct list_head *head = &desc->pg_list; int ret = 0; - whdr = nfs_writehdr_alloc(nfs_page_array_len(desc->pg_base, - desc->pg_count)); - if (!whdr) { - while (!list_empty(head)) { - req = nfs_list_entry(head->next); - nfs_list_remove_request(req); - nfs_redirty_request(req); - } + data = nfs_writedata_alloc(hdr, nfs_page_array_len(desc->pg_base, + desc->pg_count)); + if (!data) { + nfs_async_write_error(head); ret = -ENOMEM; goto out; } - data = &whdr->rpc_data; + pages = data->pages.pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_list_add_request(req, &whdr->header.pages); + nfs_list_add_request(req, &hdr->pages); *pages++ = req->wb_page; } - req = nfs_list_entry(whdr->header.pages.next); if ((desc->pg_ioflags & FLUSH_COND_STABLE) && (desc->pg_moreio || NFS_I(desc->pg_inode)->ncommit)) desc->pg_ioflags &= ~FLUSH_COND_STABLE; /* Set up the argument struct */ - nfs_write_rpcsetup(req, data, desc->pg_count, 0, desc->pg_ioflags); - list_add(&data->list, res); - desc->pg_rpc_callops = &nfs_write_full_ops; + nfs_write_rpcsetup(data, desc->pg_count, 0, desc->pg_ioflags); + list_add(&data->list, &hdr->rpc_list); + desc->pg_rpc_callops = &nfs_write_common_ops; out: return ret; } -int nfs_generic_flush(struct nfs_pageio_descriptor *desc, struct list_head *head) +int nfs_generic_flush(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { if (desc->pg_bsize < PAGE_CACHE_SIZE) - return nfs_flush_multi(desc, head); - return nfs_flush_one(desc, head); + return nfs_flush_multi(desc, hdr); + return nfs_flush_one(desc, hdr); } static int nfs_generic_pg_writepages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_write_header *whdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_flush(desc, &head); + whdr = nfs_writehdr_alloc(); + if (!whdr) { + nfs_async_write_error(&desc->pg_list); + return -ENOMEM; + } + hdr = &whdr->header; + nfs_pgheader_init(desc, hdr, nfs_writehdr_free); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_flush(desc, hdr); if (ret == 0) - ret = nfs_do_multiple_writes(&head, desc->pg_rpc_callops, - desc->pg_ioflags); + ret = nfs_do_multiple_writes(&hdr->rpc_list, + desc->pg_rpc_callops, + desc->pg_ioflags); + else + set_bit(NFS_IOHDR_REDO, &hdr->flags); + if (atomic_dec_and_test(&hdr->refcnt)) + nfs_write_completion(hdr); return ret; } @@ -1121,62 +1169,6 @@ static void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, nfs_pageio_init_write_mds(pgio, inode, ioflags); } -/* - * Handle a write reply that flushed part of a page. - */ -static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata) -{ - struct nfs_write_data *data = calldata; - - dprintk("NFS: %5u write(%s/%lld %d@%lld)", - task->tk_pid, - data->header->inode->i_sb->s_id, - (long long) - NFS_FILEID(data->header->inode), - data->header->req->wb_bytes, - (long long)req_offset(data->header->req)); - - nfs_writeback_done(task, data); -} - -static void nfs_writeback_release_partial(void *calldata) -{ - struct nfs_write_data *data = calldata; - struct nfs_page *req = data->header->req; - struct page *page = req->wb_page; - int status = data->task.tk_status; - - if (status < 0) { - nfs_set_pageerror(page); - nfs_context_set_write_error(req->wb_context, status); - dprintk(", error = %d\n", status); - goto out; - } - - if (nfs_write_need_commit(data)) { - struct inode *inode = page->mapping->host; - - spin_lock(&inode->i_lock); - if (test_bit(PG_NEED_RESCHED, &req->wb_flags)) { - /* Do nothing we need to resend the writes */ - } else if (!test_and_set_bit(PG_NEED_COMMIT, &req->wb_flags)) { - memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf)); - dprintk(" defer commit\n"); - } else if (memcmp(&req->wb_verf, &data->verf, sizeof(req->wb_verf))) { - set_bit(PG_NEED_RESCHED, &req->wb_flags); - clear_bit(PG_NEED_COMMIT, &req->wb_flags); - dprintk(" server reboot detected\n"); - } - spin_unlock(&inode->i_lock); - } else - dprintk(" OK\n"); - -out: - if (atomic_dec_and_test(&req->wb_complete)) - nfs_writepage_release(req, data); - nfs_writedata_release(data); -} - void nfs_write_prepare(struct rpc_task *task, void *calldata) { struct nfs_write_data *data = calldata; @@ -1190,12 +1182,6 @@ void nfs_commit_prepare(struct rpc_task *task, void *calldata) NFS_PROTO(data->inode)->commit_rpc_prepare(task, data); } -static const struct rpc_call_ops nfs_write_partial_ops = { - .rpc_call_prepare = nfs_write_prepare, - .rpc_call_done = nfs_writeback_done_partial, - .rpc_release = nfs_writeback_release_partial, -}; - /* * Handle a write reply that flushes a whole page. * @@ -1203,60 +1189,37 @@ static const struct rpc_call_ops nfs_write_partial_ops = { * writebacks since the page->count is kept > 1 for as long * as the page has a write request pending. */ -static void nfs_writeback_done_full(struct rpc_task *task, void *calldata) +static void nfs_writeback_done_common(struct rpc_task *task, void *calldata) { struct nfs_write_data *data = calldata; nfs_writeback_done(task, data); } -static void nfs_writeback_release_full(void *calldata) +static void nfs_writeback_release_common(void *calldata) { struct nfs_write_data *data = calldata; struct nfs_pgio_header *hdr = data->header; int status = data->task.tk_status; + struct nfs_page *req = hdr->req; - /* Update attributes as result of writeback. */ - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); - struct page *page = req->wb_page; - - nfs_list_remove_request(req); - - dprintk("NFS: %5u write (%s/%lld %d@%lld)", - data->task.tk_pid, - req->wb_context->dentry->d_inode->i_sb->s_id, - (long long)NFS_FILEID(req->wb_context->dentry->d_inode), - req->wb_bytes, - (long long)req_offset(req)); - - if (status < 0) { - nfs_set_pageerror(page); - nfs_context_set_write_error(req->wb_context, status); - dprintk(", error = %d\n", status); - goto remove_request; - } - - if (nfs_write_need_commit(data)) { + if ((status >= 0) && nfs_write_need_commit(data)) { + spin_lock(&hdr->lock); + if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) + ; /* Do nothing */ + else if (!test_and_set_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf)); - nfs_mark_request_commit(req, hdr->lseg); - dprintk(" marked for commit\n"); - goto next; - } - dprintk(" OK\n"); -remove_request: - nfs_inode_remove_request(req); - next: - nfs_unlock_request(req); - nfs_end_page_writeback(page); + else if (memcmp(&req->wb_verf, &data->verf, sizeof(req->wb_verf))) + set_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags); + spin_unlock(&hdr->lock); } nfs_writedata_release(data); } -static const struct rpc_call_ops nfs_write_full_ops = { +static const struct rpc_call_ops nfs_write_common_ops = { .rpc_call_prepare = nfs_write_prepare, - .rpc_call_done = nfs_writeback_done_full, - .rpc_release = nfs_writeback_release_full, + .rpc_call_done = nfs_writeback_done_common, + .rpc_release = nfs_writeback_release_common, }; @@ -1307,38 +1270,40 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) } } #endif - /* Is this a short write? */ - if (task->tk_status >= 0 && resp->count < argp->count) { + if (task->tk_status < 0) + nfs_set_pgio_error(data->header, task->tk_status, argp->offset); + else if (resp->count < argp->count) { static unsigned long complain; + /* This a short write! */ nfs_inc_stats(inode, NFSIOS_SHORTWRITE); /* Has the server at least made some progress? */ - if (resp->count != 0) { - /* Was this an NFSv2 write or an NFSv3 stable write? */ - if (resp->verf->committed != NFS_UNSTABLE) { - /* Resend from where the server left off */ - data->mds_offset += resp->count; - argp->offset += resp->count; - argp->pgbase += resp->count; - argp->count -= resp->count; - } else { - /* Resend as a stable write in order to avoid - * headaches in the case of a server crash. - */ - argp->stable = NFS_FILE_SYNC; + if (resp->count == 0) { + if (time_before(complain, jiffies)) { + printk(KERN_WARNING + "NFS: Server wrote zero bytes, expected %u.\n", + argp->count); + complain = jiffies + 300 * HZ; } - rpc_restart_call_prepare(task); + nfs_set_pgio_error(data->header, -EIO, argp->offset); + task->tk_status = -EIO; return; } - if (time_before(complain, jiffies)) { - printk(KERN_WARNING - "NFS: Server wrote zero bytes, expected %u.\n", - argp->count); - complain = jiffies + 300 * HZ; + /* Was this an NFSv2 write or an NFSv3 stable write? */ + if (resp->verf->committed != NFS_UNSTABLE) { + /* Resend from where the server left off */ + data->mds_offset += resp->count; + argp->offset += resp->count; + argp->pgbase += resp->count; + argp->count -= resp->count; + } else { + /* Resend as a stable write in order to avoid + * headaches in the case of a server crash. + */ + argp->stable = NFS_FILE_SYNC; } - /* Can't do anything about it except throw an error. */ - task->tk_status = -EIO; + rpc_restart_call_prepare(task); } } diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 164862148ba0..0d17db7973de 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1192,6 +1192,8 @@ enum { NFS_IOHDR_ERROR = 0, NFS_IOHDR_EOF, NFS_IOHDR_REDO, + NFS_IOHDR_NEED_COMMIT, + NFS_IOHDR_NEED_RESCHED, }; struct nfs_pgio_header { -- cgit v1.2.3 From 584aa810b6240d88c28113a90c5029449814a3b5 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:51 -0400 Subject: NFS: rewrite directio read to use async coalesce code This also has the advantage that it allows directio to use pnfs. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 255 +++++++++++++++++++++++------------------------ fs/nfs/internal.h | 5 +- fs/nfs/pagelist.c | 7 +- fs/nfs/read.c | 10 +- include/linux/nfs_page.h | 1 + include/linux/nfs_xdr.h | 4 +- 6 files changed, 138 insertions(+), 144 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 22a40c408449..4ba9a2c839bb 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -124,22 +124,6 @@ ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_ return -EINVAL; } -static void nfs_direct_dirty_pages(struct page **pages, unsigned int pgbase, size_t count) -{ - unsigned int npages; - unsigned int i; - - if (count == 0) - return; - pages += (pgbase >> PAGE_SHIFT); - npages = (count + (pgbase & ~PAGE_MASK) + PAGE_SIZE - 1) >> PAGE_SHIFT; - for (i = 0; i < npages; i++) { - struct page *page = pages[i]; - if (!PageCompound(page)) - set_page_dirty(page); - } -} - static void nfs_direct_release_pages(struct page **pages, unsigned int npages) { unsigned int i; @@ -226,58 +210,92 @@ static void nfs_direct_complete(struct nfs_direct_req *dreq) nfs_direct_req_release(dreq); } -/* - * We must hold a reference to all the pages in this direct read request - * until the RPCs complete. This could be long *after* we are woken up in - * nfs_direct_wait (for instance, if someone hits ^C on a slow server). - */ -static void nfs_direct_read_result(struct rpc_task *task, void *calldata) +void nfs_direct_readpage_release(struct nfs_page *req) { - struct nfs_read_data *data = calldata; - - nfs_readpage_result(task, data); + dprintk("NFS: direct read done (%s/%lld %d@%lld)\n", + req->wb_context->dentry->d_inode->i_sb->s_id, + (long long)NFS_FILEID(req->wb_context->dentry->d_inode), + req->wb_bytes, + (long long)req_offset(req)); + nfs_release_request(req); } -static void nfs_direct_read_release(void *calldata) +static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) { + unsigned long bytes = 0; + struct nfs_direct_req *dreq = hdr->dreq; - struct nfs_read_data *data = calldata; - struct nfs_direct_req *dreq = (struct nfs_direct_req *)data->header->req; - int status = data->task.tk_status; + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out_put; spin_lock(&dreq->lock); - if (unlikely(status < 0)) { - dreq->error = status; - spin_unlock(&dreq->lock); + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0)) + dreq->error = hdr->error; + else + dreq->count += hdr->good_bytes; + spin_unlock(&dreq->lock); + + if (!test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { + if (bytes > hdr->good_bytes) + zero_user(page, 0, PAGE_SIZE); + else if (hdr->good_bytes - bytes < PAGE_SIZE) + zero_user_segment(page, + hdr->good_bytes & ~PAGE_MASK, + PAGE_SIZE); + } + bytes += req->wb_bytes; + nfs_list_remove_request(req); + nfs_direct_readpage_release(req); + if (!PageCompound(page)) + set_page_dirty(page); + page_cache_release(page); + } } else { - dreq->count += data->res.count; - spin_unlock(&dreq->lock); - nfs_direct_dirty_pages(data->pages.pagevec, - data->args.pgbase, - data->res.count); + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + + if (bytes < hdr->good_bytes) + if (!PageCompound(req->wb_page)) + set_page_dirty(req->wb_page); + bytes += req->wb_bytes; + page_cache_release(req->wb_page); + nfs_list_remove_request(req); + nfs_direct_readpage_release(req); + } } - nfs_direct_release_pages(data->pages.pagevec, data->pages.npages); - +out_put: if (put_dreq(dreq)) nfs_direct_complete(dreq); - nfs_readdata_release(data); + hdr->release(hdr); } -static const struct rpc_call_ops nfs_read_direct_ops = { - .rpc_call_prepare = nfs_read_prepare, - .rpc_call_done = nfs_direct_read_result, - .rpc_release = nfs_direct_read_release, -}; - -static void nfs_direct_readhdr_release(struct nfs_read_header *rhdr) +static void nfs_sync_pgio_error(struct list_head *head) { - struct nfs_read_data *data = &rhdr->rpc_data; + struct nfs_page *req; - if (data->pages.pagevec != data->pages.page_array) - kfree(data->pages.pagevec); - nfs_readhdr_free(&rhdr->header); + while (!list_empty(head)) { + req = nfs_list_entry(head->next); + nfs_list_remove_request(req); + nfs_release_request(req); + } } +static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr) +{ + get_dreq(hdr->dreq); +} + +static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = { + .error_cleanup = nfs_sync_pgio_error, + .init_hdr = nfs_direct_pgio_init, + .completion = nfs_direct_read_completion, +}; + /* * For each rsize'd chunk of the user's buffer, dispatch an NFS READ * operation. If nfs_readdata_alloc() or get_user_pages() fails, @@ -285,118 +303,85 @@ static void nfs_direct_readhdr_release(struct nfs_read_header *rhdr) * handled automatically by nfs_direct_read_result(). Otherwise, if * no requests have been sent, just return an error. */ -static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, +static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *desc, const struct iovec *iov, loff_t pos) { + struct nfs_direct_req *dreq = desc->pg_dreq; struct nfs_open_context *ctx = dreq->ctx; struct inode *inode = ctx->dentry->d_inode; unsigned long user_addr = (unsigned long)iov->iov_base; size_t count = iov->iov_len; size_t rsize = NFS_SERVER(inode)->rsize; - struct rpc_task *task; - struct rpc_message msg = { - .rpc_cred = ctx->cred, - }; - struct rpc_task_setup task_setup_data = { - .rpc_client = NFS_CLIENT(inode), - .rpc_message = &msg, - .callback_ops = &nfs_read_direct_ops, - .workqueue = nfsiod_workqueue, - .flags = RPC_TASK_ASYNC, - }; unsigned int pgbase; int result; ssize_t started = 0; + struct page **pagevec = NULL; + unsigned int npages; do { - struct nfs_read_header *rhdr; - struct nfs_read_data *data; - struct nfs_page_array *pages; size_t bytes; + int i; pgbase = user_addr & ~PAGE_MASK; - bytes = min(rsize,count); + bytes = min(max(rsize, PAGE_SIZE), count); result = -ENOMEM; - rhdr = nfs_readhdr_alloc(); - if (unlikely(!rhdr)) - break; - data = nfs_readdata_alloc(&rhdr->header, nfs_page_array_len(pgbase, bytes)); - if (!data) { - nfs_readhdr_free(&rhdr->header); + npages = nfs_page_array_len(pgbase, bytes); + if (!pagevec) + pagevec = kmalloc(npages * sizeof(struct page *), + GFP_KERNEL); + if (!pagevec) break; - } - data->header = &rhdr->header; - atomic_inc(&data->header->refcnt); - pages = &data->pages; - down_read(¤t->mm->mmap_sem); result = get_user_pages(current, current->mm, user_addr, - pages->npages, 1, 0, pages->pagevec, NULL); + npages, 1, 0, pagevec, NULL); up_read(¤t->mm->mmap_sem); - if (result < 0) { - nfs_direct_readhdr_release(rhdr); + if (result < 0) break; - } - if ((unsigned)result < pages->npages) { + if ((unsigned)result < npages) { bytes = result * PAGE_SIZE; if (bytes <= pgbase) { - nfs_direct_release_pages(pages->pagevec, result); - nfs_direct_readhdr_release(rhdr); + nfs_direct_release_pages(pagevec, result); break; } bytes -= pgbase; - pages->npages = result; + npages = result; } - get_dreq(dreq); - - rhdr->header.req = (struct nfs_page *) dreq; - rhdr->header.inode = inode; - rhdr->header.cred = msg.rpc_cred; - data->args.fh = NFS_FH(inode); - data->args.context = get_nfs_open_context(ctx); - data->args.lock_context = dreq->l_ctx; - data->args.offset = pos; - data->args.pgbase = pgbase; - data->args.pages = pages->pagevec; - data->args.count = bytes; - data->res.fattr = &data->fattr; - data->res.eof = 0; - data->res.count = bytes; - nfs_fattr_init(&data->fattr); - msg.rpc_argp = &data->args; - msg.rpc_resp = &data->res; - - task_setup_data.task = &data->task; - task_setup_data.callback_data = data; - NFS_PROTO(inode)->read_setup(data, &msg); - - task = rpc_run_task(&task_setup_data); - if (IS_ERR(task)) - break; - - dprintk("NFS: %5u initiated direct read call " - "(req %s/%Ld, %zu bytes @ offset %Lu)\n", - task->tk_pid, - inode->i_sb->s_id, - (long long)NFS_FILEID(inode), - bytes, - (unsigned long long)data->args.offset); - rpc_put_task(task); - - started += bytes; - user_addr += bytes; - pos += bytes; - /* FIXME: Remove this unnecessary math from final patch */ - pgbase += bytes; - pgbase &= ~PAGE_MASK; - BUG_ON(pgbase != (user_addr & ~PAGE_MASK)); - - count -= bytes; + for (i = 0; i < npages; i++) { + struct nfs_page *req; + unsigned int req_len = min(bytes, PAGE_SIZE - pgbase); + /* XXX do we need to do the eof zeroing found in async_filler? */ + req = nfs_create_request(dreq->ctx, dreq->inode, + pagevec[i], + pgbase, req_len); + if (IS_ERR(req)) { + nfs_direct_release_pages(pagevec + i, + npages - i); + result = PTR_ERR(req); + break; + } + req->wb_index = pos >> PAGE_SHIFT; + req->wb_offset = pos & ~PAGE_MASK; + if (!nfs_pageio_add_request(desc, req)) { + result = desc->pg_error; + nfs_release_request(req); + nfs_direct_release_pages(pagevec + i, + npages - i); + break; + } + pgbase = 0; + bytes -= req_len; + started += req_len; + user_addr += req_len; + pos += req_len; + count -= req_len; + } } while (count != 0); + kfree(pagevec); + if (started) return started; return result < 0 ? (ssize_t) result : -EFAULT; @@ -407,15 +392,19 @@ static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, unsigned long nr_segs, loff_t pos) { + struct nfs_pageio_descriptor desc; ssize_t result = -EINVAL; size_t requested_bytes = 0; unsigned long seg; + nfs_pageio_init_read(&desc, dreq->inode, + &nfs_direct_read_completion_ops); get_dreq(dreq); + desc.pg_dreq = dreq; for (seg = 0; seg < nr_segs; seg++) { const struct iovec *vec = &iov[seg]; - result = nfs_direct_read_schedule_segment(dreq, vec, pos); + result = nfs_direct_read_schedule_segment(&desc, vec, pos); if (result < 0) break; requested_bytes += result; @@ -424,6 +413,8 @@ static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, pos += vec->iov_len; } + nfs_pageio_complete(&desc); + /* * If no bytes were started, return the error, and let the * generic layer handle the completion. diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 3ef8fcda1a5f..cd5d4a300bc9 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -304,8 +304,9 @@ struct nfs_pgio_completion_ops; /* read.c */ extern struct nfs_read_header *nfs_readhdr_alloc(void); extern void nfs_readhdr_free(struct nfs_pgio_header *hdr); -extern struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr, - unsigned int pagecount); +extern void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, + struct inode *inode, + const struct nfs_pgio_completion_ops *compl_ops); extern int nfs_initiate_read(struct rpc_clnt *clnt, struct nfs_read_data *data, const struct rpc_call_ops *call_ops); diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 638ca7f5a1e4..33a21ca9b84b 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -48,8 +48,11 @@ void nfs_pgheader_init(struct nfs_pageio_descriptor *desc, hdr->cred = hdr->req->wb_context->cred; hdr->io_start = req_offset(hdr->req); hdr->good_bytes = desc->pg_count; + hdr->dreq = desc->pg_dreq; hdr->release = release; hdr->completion_ops = desc->pg_completion_ops; + if (hdr->completion_ops->init_hdr) + hdr->completion_ops->init_hdr(hdr); } void nfs_set_pgio_error(struct nfs_pgio_header *hdr, int error, loff_t pos) @@ -116,9 +119,6 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode, req->wb_page = page; req->wb_index = page->index; page_cache_get(page); - BUG_ON(PagePrivate(page)); - BUG_ON(!PageLocked(page)); - BUG_ON(page->mapping->host != inode); req->wb_offset = offset; req->wb_pgbase = offset; req->wb_bytes = count; @@ -257,6 +257,7 @@ void nfs_pageio_init(struct nfs_pageio_descriptor *desc, desc->pg_ioflags = io_flags; desc->pg_error = 0; desc->pg_lseg = NULL; + desc->pg_dreq = NULL; } /** diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 5e78af162039..35e2dcebffe6 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -51,8 +51,8 @@ struct nfs_read_header *nfs_readhdr_alloc() return rhdr; } -struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr, - unsigned int pagecount) +static struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount) { struct nfs_read_data *data, *prealloc; @@ -123,9 +123,9 @@ void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio) } EXPORT_SYMBOL_GPL(nfs_pageio_reset_read_mds); -static void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, - struct inode *inode, - const struct nfs_pgio_completion_ops *compl_ops) +void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, + struct inode *inode, + const struct nfs_pgio_completion_ops *compl_ops) { if (!pnfs_pageio_init_read(pgio, inode, compl_ops)) nfs_pageio_init_read_mds(pgio, inode, compl_ops); diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index 0a5b63f16116..f9ee9eba7f88 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -68,6 +68,7 @@ struct nfs_pageio_descriptor { const struct rpc_call_ops *pg_rpc_callops; const struct nfs_pgio_completion_ops *pg_completion_ops; struct pnfs_layout_segment *pg_lseg; + struct nfs_direct_req *pg_dreq; }; #define NFS_WBACK_BUSY(req) (test_bit(PG_BUSY,&(req)->wb_flags)) diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 6fa1d2278c9d..38687b87ca9b 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1208,6 +1208,7 @@ struct nfs_pgio_header { const struct rpc_call_ops *mds_ops; void (*release) (struct nfs_pgio_header *hdr); const struct nfs_pgio_completion_ops *completion_ops; + struct nfs_direct_req *dreq; spinlock_t lock; /* fields protected by lock */ int pnfs_error; @@ -1221,8 +1222,6 @@ struct nfs_read_header { struct nfs_read_data rpc_data; }; -struct nfs_direct_req; - struct nfs_write_data { struct nfs_pgio_header *header; struct list_head list; @@ -1264,6 +1263,7 @@ struct nfs_commit_data { struct nfs_pgio_completion_ops { void (*error_cleanup)(struct list_head *head); + void (*init_hdr)(struct nfs_pgio_header *hdr); void (*completion)(struct nfs_pgio_header *hdr); }; -- cgit v1.2.3 From 1763da1234cba663b849476d451bdccac5147859 Mon Sep 17 00:00:00 2001 From: Fred Isaman Date: Fri, 20 Apr 2012 14:47:57 -0400 Subject: NFS: rewrite directio write to use async coalesce code This also has the advantage that it allows directio to use pnfs. Signed-off-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 527 +++++++++++++++++++++--------------------------- fs/nfs/internal.h | 18 +- fs/nfs/nfs4filelayout.c | 44 +++- fs/nfs/pnfs.h | 17 ++ fs/nfs/write.c | 36 ++-- 5 files changed, 319 insertions(+), 323 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 4ba9a2c839bb..d44de2f83952 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -56,6 +56,7 @@ #include "internal.h" #include "iostat.h" +#include "pnfs.h" #define NFSDBG_FACILITY NFSDBG_VFS @@ -81,16 +82,19 @@ struct nfs_direct_req { struct completion completion; /* wait for i/o completion */ /* commit state */ - struct list_head rewrite_list; /* saved nfs_write_data structs */ - struct nfs_commit_data *commit_data; /* special write_data for commits */ + struct nfs_mds_commit_info mds_cinfo; /* Storage for cinfo */ + struct pnfs_ds_commit_info ds_cinfo; /* Storage for cinfo */ + struct work_struct work; int flags; #define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */ #define NFS_ODIRECT_RESCHED_WRITES (2) /* write verification failed */ struct nfs_writeverf verf; /* unstable write verifier */ }; +static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops; +static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops; static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode); -static const struct rpc_call_ops nfs_write_direct_ops; +static void nfs_direct_write_schedule_work(struct work_struct *work); static inline void get_dreq(struct nfs_direct_req *dreq) { @@ -131,6 +135,16 @@ static void nfs_direct_release_pages(struct page **pages, unsigned int npages) page_cache_release(pages[i]); } +void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, + struct nfs_direct_req *dreq) +{ + cinfo->lock = &dreq->lock; + cinfo->mds = &dreq->mds_cinfo; + cinfo->ds = &dreq->ds_cinfo; + cinfo->dreq = dreq; + cinfo->completion_ops = &nfs_direct_commit_completion_ops; +} + static inline struct nfs_direct_req *nfs_direct_req_alloc(void) { struct nfs_direct_req *dreq; @@ -142,7 +156,11 @@ static inline struct nfs_direct_req *nfs_direct_req_alloc(void) kref_init(&dreq->kref); kref_get(&dreq->kref); init_completion(&dreq->completion); - INIT_LIST_HEAD(&dreq->rewrite_list); + dreq->mds_cinfo.ncommit = 0; + atomic_set(&dreq->mds_cinfo.rpcs_out, 0); + INIT_LIST_HEAD(&dreq->mds_cinfo.list); + INIT_WORK(&dreq->work, nfs_direct_write_schedule_work); + memset(&dreq->ds_cinfo, 0, sizeof(dreq->ds_cinfo)); dreq->iocb = NULL; dreq->ctx = NULL; dreq->l_ctx = NULL; @@ -457,112 +475,60 @@ out: return result; } -static void nfs_direct_writehdr_release(struct nfs_write_header *whdr) -{ - struct nfs_write_data *data = &whdr->rpc_data; - - if (data->pages.pagevec != data->pages.page_array) - kfree(data->pages.pagevec); - nfs_writehdr_free(&whdr->header); -} - -static void nfs_direct_free_writedata(struct nfs_direct_req *dreq) -{ - while (!list_empty(&dreq->rewrite_list)) { - struct nfs_pgio_header *hdr = list_entry(dreq->rewrite_list.next, struct nfs_pgio_header, pages); - struct nfs_write_header *whdr = container_of(hdr, struct nfs_write_header, header); - struct nfs_page_array *p = &whdr->rpc_data.pages; - - list_del(&hdr->pages); - nfs_direct_release_pages(p->pagevec, p->npages); - nfs_direct_writehdr_release(whdr); - } -} - #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) { - struct inode *inode = dreq->inode; - struct list_head *p; - struct nfs_write_data *data; - struct nfs_pgio_header *hdr; - struct rpc_task *task; - struct rpc_message msg = { - .rpc_cred = dreq->ctx->cred, - }; - struct rpc_task_setup task_setup_data = { - .rpc_client = NFS_CLIENT(inode), - .rpc_message = &msg, - .callback_ops = &nfs_write_direct_ops, - .workqueue = nfsiod_workqueue, - .flags = RPC_TASK_ASYNC, - }; + struct nfs_pageio_descriptor desc; + struct nfs_page *req, *tmp; + LIST_HEAD(reqs); + struct nfs_commit_info cinfo; + LIST_HEAD(failed); + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo); + spin_lock(cinfo.lock); + nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0); + spin_unlock(cinfo.lock); dreq->count = 0; get_dreq(dreq); - list_for_each(p, &dreq->rewrite_list) { - hdr = list_entry(p, struct nfs_pgio_header, pages); - data = &(container_of(hdr, struct nfs_write_header, header))->rpc_data; - - get_dreq(dreq); - - /* Use stable writes */ - data->args.stable = NFS_FILE_SYNC; - - /* - * Reset data->res. - */ - nfs_fattr_init(&data->fattr); - data->res.count = data->args.count; - memset(&data->verf, 0, sizeof(data->verf)); - - /* - * Reuse data->task; data->args should not have changed - * since the original request was sent. - */ - task_setup_data.task = &data->task; - task_setup_data.callback_data = data; - msg.rpc_argp = &data->args; - msg.rpc_resp = &data->res; - NFS_PROTO(inode)->write_setup(data, &msg); - - /* - * We're called via an RPC callback, so BKL is already held. - */ - task = rpc_run_task(&task_setup_data); - if (!IS_ERR(task)) - rpc_put_task(task); - - dprintk("NFS: %5u rescheduled direct write call (req %s/%Ld, %u bytes @ offset %Lu)\n", - data->task.tk_pid, - inode->i_sb->s_id, - (long long)NFS_FILEID(inode), - data->args.count, - (unsigned long long)data->args.offset); - } + nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, + &nfs_direct_write_completion_ops); + desc.pg_dreq = dreq; - if (put_dreq(dreq)) - nfs_direct_write_complete(dreq, inode); -} + list_for_each_entry_safe(req, tmp, &reqs, wb_list) { + if (!nfs_pageio_add_request(&desc, req)) { + nfs_list_add_request(req, &failed); + spin_lock(cinfo.lock); + dreq->flags = 0; + dreq->error = -EIO; + spin_unlock(cinfo.lock); + } + } + nfs_pageio_complete(&desc); -static void nfs_direct_commit_result(struct rpc_task *task, void *calldata) -{ - struct nfs_commit_data *data = calldata; + while (!list_empty(&failed)) { + page_cache_release(req->wb_page); + nfs_release_request(req); + nfs_unlock_request(req); + } - /* Call the NFS version-specific code */ - NFS_PROTO(data->inode)->commit_done(task, data); + if (put_dreq(dreq)) + nfs_direct_write_complete(dreq, dreq->inode); } -static void nfs_direct_commit_release(void *calldata) +static void nfs_direct_commit_complete(struct nfs_commit_data *data) { - struct nfs_commit_data *data = calldata; struct nfs_direct_req *dreq = data->dreq; + struct nfs_commit_info cinfo; + struct nfs_page *req; int status = data->task.tk_status; + nfs_init_cinfo_from_dreq(&cinfo, dreq); if (status < 0) { dprintk("NFS: %5u commit failed with error %d.\n", - data->task.tk_pid, status); + data->task.tk_pid, status); dreq->flags = NFS_ODIRECT_RESCHED_WRITES; } else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) { dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid); @@ -570,59 +536,49 @@ static void nfs_direct_commit_release(void *calldata) } dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status); - nfs_direct_write_complete(dreq, data->inode); - nfs_commit_free(data); + while (!list_empty(&data->pages)) { + req = nfs_list_entry(data->pages.next); + nfs_list_remove_request(req); + if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) { + /* Note the rewrite will go through mds */ + nfs_mark_request_commit(req, NULL, &cinfo); + } else { + page_cache_release(req->wb_page); + nfs_release_request(req); + } + nfs_unlock_request(req); + } + + if (atomic_dec_and_test(&cinfo.mds->rpcs_out)) + nfs_direct_write_complete(dreq, data->inode); } -static const struct rpc_call_ops nfs_commit_direct_ops = { - .rpc_call_prepare = nfs_commit_prepare, - .rpc_call_done = nfs_direct_commit_result, - .rpc_release = nfs_direct_commit_release, +static void nfs_direct_error_cleanup(struct nfs_inode *nfsi) +{ + /* There is no lock to clear */ +} + +static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = { + .completion = nfs_direct_commit_complete, + .error_cleanup = nfs_direct_error_cleanup, }; static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq) { - struct nfs_commit_data *data = dreq->commit_data; - struct rpc_task *task; - struct rpc_message msg = { - .rpc_argp = &data->args, - .rpc_resp = &data->res, - .rpc_cred = dreq->ctx->cred, - }; - struct rpc_task_setup task_setup_data = { - .task = &data->task, - .rpc_client = NFS_CLIENT(dreq->inode), - .rpc_message = &msg, - .callback_ops = &nfs_commit_direct_ops, - .callback_data = data, - .workqueue = nfsiod_workqueue, - .flags = RPC_TASK_ASYNC, - }; - - data->inode = dreq->inode; - data->cred = msg.rpc_cred; - - data->args.fh = NFS_FH(data->inode); - data->args.offset = 0; - data->args.count = 0; - data->res.fattr = &data->fattr; - data->res.verf = &data->verf; - nfs_fattr_init(&data->fattr); - - NFS_PROTO(data->inode)->commit_setup(data, &msg); - - /* Note: task.tk_ops->rpc_release will free dreq->commit_data */ - dreq->commit_data = NULL; - - dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid); - - task = rpc_run_task(&task_setup_data); - if (!IS_ERR(task)) - rpc_put_task(task); + int res; + struct nfs_commit_info cinfo; + LIST_HEAD(mds_list); + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + nfs_scan_commit(dreq->inode, &mds_list, &cinfo); + res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo); + if (res < 0) /* res == -ENOMEM */ + nfs_direct_write_reschedule(dreq); } -static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) +static void nfs_direct_write_schedule_work(struct work_struct *work) { + struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work); int flags = dreq->flags; dreq->flags = 0; @@ -634,90 +590,29 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode nfs_direct_write_reschedule(dreq); break; default: - if (dreq->commit_data != NULL) - nfs_commit_free(dreq->commit_data); - nfs_direct_free_writedata(dreq); - nfs_zap_mapping(inode, inode->i_mapping); + nfs_zap_mapping(dreq->inode, dreq->inode->i_mapping); nfs_direct_complete(dreq); } } -static void nfs_alloc_commit_data(struct nfs_direct_req *dreq) +static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) { - dreq->commit_data = nfs_commitdata_alloc(); - if (dreq->commit_data != NULL) - dreq->commit_data->dreq = dreq; + schedule_work(&dreq->work); /* Calls nfs_direct_write_schedule_work */ } + #else -static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq) -{ - dreq->commit_data = NULL; -} static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) { - nfs_direct_free_writedata(dreq); nfs_zap_mapping(inode, inode->i_mapping); nfs_direct_complete(dreq); } #endif -static void nfs_direct_write_result(struct rpc_task *task, void *calldata) -{ - struct nfs_write_data *data = calldata; - - nfs_writeback_done(task, data); -} - /* * NB: Return the value of the first error return code. Subsequent * errors after the first one are ignored. */ -static void nfs_direct_write_release(void *calldata) -{ - struct nfs_write_data *data = calldata; - struct nfs_pgio_header *hdr = data->header; - struct nfs_direct_req *dreq = (struct nfs_direct_req *) hdr->req; - int status = data->task.tk_status; - - spin_lock(&dreq->lock); - - if (unlikely(status < 0)) { - /* An error has occurred, so we should not commit */ - dreq->flags = 0; - dreq->error = status; - } - if (unlikely(dreq->error != 0)) - goto out_unlock; - - dreq->count += data->res.count; - - if (data->res.verf->committed != NFS_FILE_SYNC) { - switch (dreq->flags) { - case 0: - memcpy(&dreq->verf, &data->verf, sizeof(dreq->verf)); - dreq->flags = NFS_ODIRECT_DO_COMMIT; - break; - case NFS_ODIRECT_DO_COMMIT: - if (memcmp(&dreq->verf, &data->verf, sizeof(dreq->verf))) { - dprintk("NFS: %5u write verify failed\n", data->task.tk_pid); - dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - } - } - } -out_unlock: - spin_unlock(&dreq->lock); - - if (put_dreq(dreq)) - nfs_direct_write_complete(dreq, hdr->inode); -} - -static const struct rpc_call_ops nfs_write_direct_ops = { - .rpc_call_prepare = nfs_write_prepare, - .rpc_call_done = nfs_direct_write_result, - .rpc_release = nfs_direct_write_release, -}; - /* * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE * operation. If nfs_writedata_alloc() or get_user_pages() fails, @@ -725,143 +620,181 @@ static const struct rpc_call_ops nfs_write_direct_ops = { * handled automatically by nfs_direct_write_result(). Otherwise, if * no requests have been sent, just return an error. */ -static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, +static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *desc, const struct iovec *iov, - loff_t pos, int sync) + loff_t pos) { + struct nfs_direct_req *dreq = desc->pg_dreq; struct nfs_open_context *ctx = dreq->ctx; struct inode *inode = ctx->dentry->d_inode; unsigned long user_addr = (unsigned long)iov->iov_base; size_t count = iov->iov_len; - struct rpc_task *task; - struct rpc_message msg = { - .rpc_cred = ctx->cred, - }; - struct rpc_task_setup task_setup_data = { - .rpc_client = NFS_CLIENT(inode), - .rpc_message = &msg, - .callback_ops = &nfs_write_direct_ops, - .workqueue = nfsiod_workqueue, - .flags = RPC_TASK_ASYNC, - }; size_t wsize = NFS_SERVER(inode)->wsize; unsigned int pgbase; int result; ssize_t started = 0; + struct page **pagevec = NULL; + unsigned int npages; do { - struct nfs_write_header *whdr; - struct nfs_write_data *data; - struct nfs_page_array *pages; size_t bytes; + int i; pgbase = user_addr & ~PAGE_MASK; - bytes = min(wsize,count); + bytes = min(max(wsize, PAGE_SIZE), count); result = -ENOMEM; - whdr = nfs_writehdr_alloc(); - if (unlikely(!whdr)) + npages = nfs_page_array_len(pgbase, bytes); + if (!pagevec) + pagevec = kmalloc(npages * sizeof(struct page *), GFP_KERNEL); + if (!pagevec) break; - data = nfs_writedata_alloc(&whdr->header, nfs_page_array_len(pgbase, bytes)); - if (!data) { - nfs_writehdr_free(&whdr->header); - break; - } - data->header = &whdr->header; - atomic_inc(&data->header->refcnt); - pages = &data->pages; - down_read(¤t->mm->mmap_sem); result = get_user_pages(current, current->mm, user_addr, - pages->npages, 0, 0, pages->pagevec, NULL); + npages, 0, 0, pagevec, NULL); up_read(¤t->mm->mmap_sem); - if (result < 0) { - nfs_direct_writehdr_release(whdr); + if (result < 0) break; - } - if ((unsigned)result < pages->npages) { + + if ((unsigned)result < npages) { bytes = result * PAGE_SIZE; if (bytes <= pgbase) { - nfs_direct_release_pages(pages->pagevec, result); - nfs_direct_writehdr_release(whdr); + nfs_direct_release_pages(pagevec, result); break; } bytes -= pgbase; - pages->npages = result; + npages = result; } - get_dreq(dreq); - - list_move_tail(&whdr->header.pages, &dreq->rewrite_list); - - whdr->header.req = (struct nfs_page *) dreq; - whdr->header.inode = inode; - whdr->header.cred = msg.rpc_cred; - data->args.fh = NFS_FH(inode); - data->args.context = ctx; - data->args.lock_context = dreq->l_ctx; - data->args.offset = pos; - data->args.pgbase = pgbase; - data->args.pages = pages->pagevec; - data->args.count = bytes; - data->args.stable = sync; - data->res.fattr = &data->fattr; - data->res.count = bytes; - data->res.verf = &data->verf; - nfs_fattr_init(&data->fattr); - - task_setup_data.task = &data->task; - task_setup_data.callback_data = data; - msg.rpc_argp = &data->args; - msg.rpc_resp = &data->res; - NFS_PROTO(inode)->write_setup(data, &msg); - - task = rpc_run_task(&task_setup_data); - if (IS_ERR(task)) - break; + for (i = 0; i < npages; i++) { + struct nfs_page *req; + unsigned int req_len = min(bytes, PAGE_SIZE - pgbase); - dprintk("NFS: %5u initiated direct write call " - "(req %s/%Ld, %zu bytes @ offset %Lu)\n", - task->tk_pid, - inode->i_sb->s_id, - (long long)NFS_FILEID(inode), - bytes, - (unsigned long long)data->args.offset); - rpc_put_task(task); - - started += bytes; - user_addr += bytes; - pos += bytes; - - /* FIXME: Remove this useless math from the final patch */ - pgbase += bytes; - pgbase &= ~PAGE_MASK; - BUG_ON(pgbase != (user_addr & ~PAGE_MASK)); - - count -= bytes; + req = nfs_create_request(dreq->ctx, dreq->inode, + pagevec[i], + pgbase, req_len); + if (IS_ERR(req)) { + nfs_direct_release_pages(pagevec + i, + npages - i); + result = PTR_ERR(req); + break; + } + nfs_lock_request(req); + req->wb_index = pos >> PAGE_SHIFT; + req->wb_offset = pos & ~PAGE_MASK; + if (!nfs_pageio_add_request(desc, req)) { + result = desc->pg_error; + nfs_unlock_request(req); + nfs_release_request(req); + nfs_direct_release_pages(pagevec + i, + npages - i); + } + pgbase = 0; + bytes -= req_len; + started += req_len; + user_addr += req_len; + pos += req_len; + count -= req_len; + } } while (count != 0); + kfree(pagevec); + if (started) return started; return result < 0 ? (ssize_t) result : -EFAULT; } +static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) +{ + struct nfs_direct_req *dreq = hdr->dreq; + struct nfs_commit_info cinfo; + int bit = -1; + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out_put; + + nfs_init_cinfo_from_dreq(&cinfo, dreq); + + spin_lock(&dreq->lock); + + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { + dreq->flags = 0; + dreq->error = hdr->error; + } + if (dreq->error != 0) + bit = NFS_IOHDR_ERROR; + else { + dreq->count += hdr->good_bytes; + if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) { + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + bit = NFS_IOHDR_NEED_RESCHED; + } else if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) { + if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) + bit = NFS_IOHDR_NEED_RESCHED; + else if (dreq->flags == 0) { + memcpy(&dreq->verf, &req->wb_verf, + sizeof(dreq->verf)); + bit = NFS_IOHDR_NEED_COMMIT; + dreq->flags = NFS_ODIRECT_DO_COMMIT; + } else if (dreq->flags == NFS_ODIRECT_DO_COMMIT) { + if (memcmp(&dreq->verf, &req->wb_verf, sizeof(dreq->verf))) { + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + bit = NFS_IOHDR_NEED_RESCHED; + } else + bit = NFS_IOHDR_NEED_COMMIT; + } + } + } + spin_unlock(&dreq->lock); + + while (!list_empty(&hdr->pages)) { + req = nfs_list_entry(hdr->pages.next); + nfs_list_remove_request(req); + switch (bit) { + case NFS_IOHDR_NEED_RESCHED: + case NFS_IOHDR_NEED_COMMIT: + nfs_mark_request_commit(req, hdr->lseg, &cinfo); + break; + default: + page_cache_release(req->wb_page); + nfs_release_request(req); + } + nfs_unlock_request(req); + } + +out_put: + if (put_dreq(dreq)) + nfs_direct_write_complete(dreq, hdr->inode); + hdr->release(hdr); +} + +static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = { + .error_cleanup = nfs_sync_pgio_error, + .init_hdr = nfs_direct_pgio_init, + .completion = nfs_direct_write_completion, +}; + static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, const struct iovec *iov, unsigned long nr_segs, - loff_t pos, int sync) + loff_t pos) { + struct nfs_pageio_descriptor desc; ssize_t result = 0; size_t requested_bytes = 0; unsigned long seg; + nfs_pageio_init_write(&desc, dreq->inode, FLUSH_COND_STABLE, + &nfs_direct_write_completion_ops); + desc.pg_dreq = dreq; get_dreq(dreq); for (seg = 0; seg < nr_segs; seg++) { const struct iovec *vec = &iov[seg]; - result = nfs_direct_write_schedule_segment(dreq, vec, - pos, sync); + result = nfs_direct_write_schedule_segment(&desc, vec, pos); if (result < 0) break; requested_bytes += result; @@ -869,6 +802,7 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, break; pos += vec->iov_len; } + nfs_pageio_complete(&desc); /* * If no bytes were started, return the error, and let the @@ -891,16 +825,10 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov, ssize_t result = -ENOMEM; struct inode *inode = iocb->ki_filp->f_mapping->host; struct nfs_direct_req *dreq; - size_t wsize = NFS_SERVER(inode)->wsize; - int sync = NFS_UNSTABLE; dreq = nfs_direct_req_alloc(); if (!dreq) goto out; - nfs_alloc_commit_data(dreq); - - if (dreq->commit_data == NULL || count <= wsize) - sync = NFS_FILE_SYNC; dreq->inode = inode; dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); @@ -910,7 +838,7 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov, if (!is_sync_kiocb(iocb)) dreq->iocb = iocb; - result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, sync); + result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos); if (!result) result = nfs_direct_wait(dreq); out_release: @@ -1030,10 +958,15 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov, task_io_account_write(count); retval = nfs_direct_write(iocb, iov, nr_segs, pos, count); + if (retval > 0) { + struct inode *inode = mapping->host; - if (retval > 0) iocb->ki_pos = pos + retval; - + spin_lock(&inode->i_lock); + if (i_size_read(inode) < iocb->ki_pos) + i_size_write(inode, iocb->ki_pos); + spin_unlock(&inode->i_lock); + } out: return retval; } diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 137f5cd71433..d68810f61869 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -320,10 +320,11 @@ extern void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio); extern void nfs_readdata_release(struct nfs_read_data *rdata); /* write.c */ +extern void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, + struct inode *inode, int ioflags, + const struct nfs_pgio_completion_ops *compl_ops); extern struct nfs_write_header *nfs_writehdr_alloc(void); extern void nfs_writehdr_free(struct nfs_pgio_header *hdr); -extern struct nfs_write_data *nfs_writedata_alloc(struct nfs_pgio_header *hdr, - unsigned int pagecount); extern int nfs_generic_flush(struct nfs_pageio_descriptor *desc, struct nfs_pgio_header *hdr); extern void nfs_pageio_init_write_mds(struct nfs_pageio_descriptor *pgio, @@ -346,6 +347,15 @@ extern void nfs_init_commit(struct nfs_commit_data *data, struct list_head *head, struct pnfs_layout_segment *lseg, struct nfs_commit_info *cinfo); +int nfs_scan_commit_list(struct list_head *src, struct list_head *dst, + struct nfs_commit_info *cinfo, int max); +int nfs_scan_commit(struct inode *inode, struct list_head *dst, + struct nfs_commit_info *cinfo); +void nfs_mark_request_commit(struct nfs_page *req, + struct pnfs_layout_segment *lseg, + struct nfs_commit_info *cinfo); +int nfs_generic_commit_list(struct inode *inode, struct list_head *head, + int how, struct nfs_commit_info *cinfo); void nfs_retry_commit(struct list_head *page_list, struct pnfs_layout_segment *lseg, struct nfs_commit_info *cinfo); @@ -365,6 +375,10 @@ extern int nfs_migrate_page(struct address_space *, #define nfs_migrate_page NULL #endif +/* direct.c */ +void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, + struct nfs_direct_req *dreq); + /* nfs4proc.c */ extern void __nfs4_read_done_cb(struct nfs_read_data *); extern void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data); diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 26d1da486761..806a55f513d9 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -996,12 +996,9 @@ static int filelayout_initiate_commit(struct nfs_commit_data *data, int how) } static int -filelayout_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, - struct nfs_commit_info *cinfo, - int max) +transfer_commit_list(struct list_head *src, struct list_head *dst, + struct nfs_commit_info *cinfo, int max) { - struct list_head *src = &bucket->written; - struct list_head *dst = &bucket->committing; struct nfs_page *req, *tmp; int ret = 0; @@ -1014,9 +1011,22 @@ filelayout_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, clear_bit(PG_COMMIT_TO_DS, &req->wb_flags); nfs_list_add_request(req, dst); ret++; - if (ret == max) + if ((ret == max) && !cinfo->dreq) break; } + return ret; +} + +static int +filelayout_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, + struct nfs_commit_info *cinfo, + int max) +{ + struct list_head *src = &bucket->written; + struct list_head *dst = &bucket->committing; + int ret; + + ret = transfer_commit_list(src, dst, cinfo, max); if (ret) { cinfo->ds->nwritten -= ret; cinfo->ds->ncommitting += ret; @@ -1046,6 +1056,27 @@ static int filelayout_scan_commit_lists(struct nfs_commit_info *cinfo, return rv; } +/* Pull everything off the committing lists and dump into @dst */ +static void filelayout_recover_commit_reqs(struct list_head *dst, + struct nfs_commit_info *cinfo) +{ + struct pnfs_commit_bucket *b; + int i; + + /* NOTE cinfo->lock is NOT held, relying on fact that this is + * only called on single thread per dreq. + * Can't take the lock because need to do put_lseg + */ + for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) { + if (transfer_commit_list(&b->written, dst, cinfo, 0)) { + BUG_ON(!list_empty(&b->written)); + put_lseg(b->wlseg); + b->wlseg = NULL; + } + } + cinfo->ds->nwritten = 0; +} + static unsigned int alloc_ds_commits(struct nfs_commit_info *cinfo, struct list_head *list) { @@ -1170,6 +1201,7 @@ static struct pnfs_layoutdriver_type filelayout_type = { .mark_request_commit = filelayout_mark_request_commit, .clear_request_commit = filelayout_clear_request_commit, .scan_commit_lists = filelayout_scan_commit_lists, + .recover_commit_reqs = filelayout_recover_commit_reqs, .commit_pagelist = filelayout_commit_pagelist, .read_pagelist = filelayout_read_pagelist, .write_pagelist = filelayout_write_pagelist, diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 4cd8760c2f89..8efbee769ba7 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -102,6 +102,8 @@ struct pnfs_layoutdriver_type { struct nfs_commit_info *cinfo); int (*scan_commit_lists) (struct nfs_commit_info *cinfo, int max); + void (*recover_commit_reqs) (struct list_head *list, + struct nfs_commit_info *cinfo); int (*commit_pagelist)(struct inode *inode, struct list_head *mds_pages, int how, @@ -323,6 +325,15 @@ pnfs_scan_commit_lists(struct inode *inode, struct nfs_commit_info *cinfo, return NFS_SERVER(inode)->pnfs_curr_ld->scan_commit_lists(cinfo, max); } +static inline void +pnfs_recover_commit_reqs(struct inode *inode, struct list_head *list, + struct nfs_commit_info *cinfo) +{ + if (cinfo->ds == NULL || cinfo->ds->nwritten == 0) + return; + NFS_SERVER(inode)->pnfs_curr_ld->recover_commit_reqs(list, cinfo); +} + /* Should the pNFS client commit and return the layout upon a setattr */ static inline bool pnfs_ld_layoutret_on_setattr(struct inode *inode) @@ -456,6 +467,12 @@ pnfs_scan_commit_lists(struct inode *inode, struct nfs_commit_info *cinfo, return 0; } +static inline void +pnfs_recover_commit_reqs(struct inode *inode, struct list_head *list, + struct nfs_commit_info *cinfo) +{ +} + static inline int pnfs_layoutcommit_inode(struct inode *inode, bool sync) { return 0; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 56db9e7fa47a..fec214bfa702 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -39,9 +39,6 @@ /* * Local function declarations */ -static void nfs_pageio_init_write(struct nfs_pageio_descriptor *desc, - struct inode *inode, int ioflags, - const struct nfs_pgio_completion_ops *compl_ops); static void nfs_redirty_request(struct nfs_page *req); static const struct rpc_call_ops nfs_write_common_ops; static const struct rpc_call_ops nfs_commit_ops; @@ -87,8 +84,8 @@ struct nfs_write_header *nfs_writehdr_alloc(void) return p; } -struct nfs_write_data *nfs_writedata_alloc(struct nfs_pgio_header *hdr, - unsigned int pagecount) +static struct nfs_write_data *nfs_writedata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount) { struct nfs_write_data *data, *prealloc; @@ -518,14 +515,17 @@ void nfs_init_cinfo(struct nfs_commit_info *cinfo, struct inode *inode, struct nfs_direct_req *dreq) { - nfs_init_cinfo_from_inode(cinfo, inode); + if (dreq) + nfs_init_cinfo_from_dreq(cinfo, dreq); + else + nfs_init_cinfo_from_inode(cinfo, inode); } EXPORT_SYMBOL_GPL(nfs_init_cinfo); /* * Add a request to the inode's commit list. */ -static void +void nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg, struct nfs_commit_info *cinfo) { @@ -567,7 +567,7 @@ int nfs_write_need_commit(struct nfs_write_data *data) } #else -static void +void nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg, struct nfs_commit_info *cinfo) { @@ -632,7 +632,7 @@ nfs_reqs_to_commit(struct nfs_commit_info *cinfo) } /* cinfo->lock held by caller */ -static int +int nfs_scan_commit_list(struct list_head *src, struct list_head *dst, struct nfs_commit_info *cinfo, int max) { @@ -647,7 +647,7 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst, nfs_request_remove_commit_list(req, cinfo); nfs_list_add_request(req, dst); ret++; - if (ret == max) + if ((ret == max) && !cinfo->dreq) break; } return ret; @@ -662,7 +662,7 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst, * Moves requests from the inode's 'commit' request list. * The requests are *not* checked to ensure that they form a contiguous set. */ -static int +int nfs_scan_commit(struct inode *inode, struct list_head *dst, struct nfs_commit_info *cinfo) { @@ -686,8 +686,8 @@ static unsigned long nfs_reqs_to_commit(struct nfs_commit_info *cinfo) return 0; } -static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, - struct nfs_commit_info *cinfo) +int nfs_scan_commit(struct inode *inode, struct list_head *dst, + struct nfs_commit_info *cinfo) { return 0; } @@ -1202,9 +1202,9 @@ void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio) } EXPORT_SYMBOL_GPL(nfs_pageio_reset_write_mds); -static void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, - struct inode *inode, int ioflags, - const struct nfs_pgio_completion_ops *compl_ops) +void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, + struct inode *inode, int ioflags, + const struct nfs_pgio_completion_ops *compl_ops) { if (!pnfs_pageio_init_write(pgio, inode, ioflags, compl_ops)) nfs_pageio_init_write_mds(pgio, inode, ioflags, compl_ops); @@ -1568,8 +1568,8 @@ static const struct nfs_commit_completion_ops nfs_commit_completion_ops = { .error_cleanup = nfs_commit_clear_lock, }; -static int nfs_generic_commit_list(struct inode *inode, struct list_head *head, - int how, struct nfs_commit_info *cinfo) +int nfs_generic_commit_list(struct inode *inode, struct list_head *head, + int how, struct nfs_commit_info *cinfo) { int status; -- cgit v1.2.3 From 24fc9211f4d48c04882a52e42b21c9b4abc4f9bf Mon Sep 17 00:00:00 2001 From: Bryan Schumaker Date: Mon, 30 Apr 2012 13:27:11 -0400 Subject: NFS: Define nfs_direct_write_schedule_work() when v3 and v4 are disabled v2 doesn't have commits, so this function can be a no-op. Signed-off-by: Bryan Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index d44de2f83952..e83545c4d36f 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -601,6 +601,9 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode } #else +static void nfs_direct_write_schedule_work(struct work_struct *work) +{ +} static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) { -- cgit v1.2.3 From 71e8cc00c63e8518ce86b4079355fc9086a4869d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 30 Apr 2012 13:22:54 -0400 Subject: NFS: Ensure that we break out of read/write_schedule_segment on error Currently we do break out of the for() loop, but we also need to break out of the enclosing do {} while()... Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index e83545c4d36f..f30d5c26a763 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -396,7 +396,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de pos += req_len; count -= req_len; } - } while (count != 0); + } while (count != 0 && result >= 0); kfree(pagevec); @@ -692,6 +692,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d nfs_release_request(req); nfs_direct_release_pages(pagevec + i, npages - i); + break; } pgbase = 0; bytes -= req_len; @@ -700,7 +701,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d pos += req_len; count -= req_len; } - } while (count != 0); + } while (count != 0 && result >= 0); kfree(pagevec); -- cgit v1.2.3 From 3e9e0ca3f19e911ce13c2e6c9858fcb41a37496c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 30 Apr 2012 13:40:06 -0400 Subject: NFS: O_DIRECT pgio_completion_ops error_cleanup must unlock the request Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index f30d5c26a763..af02bde7741e 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -292,7 +292,7 @@ out_put: hdr->release(hdr); } -static void nfs_sync_pgio_error(struct list_head *head) +static void nfs_read_sync_pgio_error(struct list_head *head) { struct nfs_page *req; @@ -309,7 +309,7 @@ static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr) } static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = { - .error_cleanup = nfs_sync_pgio_error, + .error_cleanup = nfs_read_sync_pgio_error, .init_hdr = nfs_direct_pgio_init, .completion = nfs_direct_read_completion, }; @@ -775,8 +775,20 @@ out_put: hdr->release(hdr); } +static void nfs_write_sync_pgio_error(struct list_head *head) +{ + struct nfs_page *req; + + while (!list_empty(head)) { + req = nfs_list_entry(head->next); + nfs_list_remove_request(req); + nfs_release_request(req); + nfs_unlock_request(req); + } +} + static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = { - .error_cleanup = nfs_sync_pgio_error, + .error_cleanup = nfs_write_sync_pgio_error, .init_hdr = nfs_direct_pgio_init, .completion = nfs_direct_write_completion, }; -- cgit v1.2.3 From 6d74743b088d116e31fe1b73f47e782ee2016b94 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 30 Apr 2012 13:27:31 -0400 Subject: NFS: Simplify O_DIRECT page referencing The O_DIRECT code shouldn't need to hold 2 references to each page. The reference held by the struct nfs_page should suffice. Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 22 ++++++---------------- 1 file changed, 6 insertions(+), 16 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index af02bde7741e..78d1ead8bc30 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -268,10 +268,9 @@ static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) } bytes += req->wb_bytes; nfs_list_remove_request(req); - nfs_direct_readpage_release(req); if (!PageCompound(page)) set_page_dirty(page); - page_cache_release(page); + nfs_direct_readpage_release(req); } } else { while (!list_empty(&hdr->pages)) { @@ -281,7 +280,6 @@ static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) if (!PageCompound(req->wb_page)) set_page_dirty(req->wb_page); bytes += req->wb_bytes; - page_cache_release(req->wb_page); nfs_list_remove_request(req); nfs_direct_readpage_release(req); } @@ -375,8 +373,6 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de pagevec[i], pgbase, req_len); if (IS_ERR(req)) { - nfs_direct_release_pages(pagevec + i, - npages - i); result = PTR_ERR(req); break; } @@ -385,8 +381,6 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de if (!nfs_pageio_add_request(desc, req)) { result = desc->pg_error; nfs_release_request(req); - nfs_direct_release_pages(pagevec + i, - npages - i); break; } pgbase = 0; @@ -396,6 +390,8 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de pos += req_len; count -= req_len; } + /* The nfs_page now hold references to these pages */ + nfs_direct_release_pages(pagevec, npages); } while (count != 0 && result >= 0); kfree(pagevec); @@ -509,7 +505,6 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) nfs_pageio_complete(&desc); while (!list_empty(&failed)) { - page_cache_release(req->wb_page); nfs_release_request(req); nfs_unlock_request(req); } @@ -542,10 +537,8 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) { /* Note the rewrite will go through mds */ nfs_mark_request_commit(req, NULL, &cinfo); - } else { - page_cache_release(req->wb_page); + } else nfs_release_request(req); - } nfs_unlock_request(req); } @@ -678,8 +671,6 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d pagevec[i], pgbase, req_len); if (IS_ERR(req)) { - nfs_direct_release_pages(pagevec + i, - npages - i); result = PTR_ERR(req); break; } @@ -690,8 +681,6 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d result = desc->pg_error; nfs_unlock_request(req); nfs_release_request(req); - nfs_direct_release_pages(pagevec + i, - npages - i); break; } pgbase = 0; @@ -701,6 +690,8 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d pos += req_len; count -= req_len; } + /* The nfs_page now hold references to these pages */ + nfs_direct_release_pages(pagevec, npages); } while (count != 0 && result >= 0); kfree(pagevec); @@ -763,7 +754,6 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) nfs_mark_request_commit(req, hdr->lseg, &cinfo); break; default: - page_cache_release(req->wb_page); nfs_release_request(req); } nfs_unlock_request(req); -- cgit v1.2.3 From 292f3eeef00a20fa0ef4feec62792ad0065760a0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 30 Apr 2012 18:31:49 -0400 Subject: NFS: Use kmem_cache_zalloc() in nfs_direct_req_alloc Simplify the initialisation of O_DIRECT requests. Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 12 +----------- 1 file changed, 1 insertion(+), 11 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 78d1ead8bc30..f17e4695c842 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -149,26 +149,16 @@ static inline struct nfs_direct_req *nfs_direct_req_alloc(void) { struct nfs_direct_req *dreq; - dreq = kmem_cache_alloc(nfs_direct_cachep, GFP_KERNEL); + dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL); if (!dreq) return NULL; kref_init(&dreq->kref); kref_get(&dreq->kref); init_completion(&dreq->completion); - dreq->mds_cinfo.ncommit = 0; - atomic_set(&dreq->mds_cinfo.rpcs_out, 0); INIT_LIST_HEAD(&dreq->mds_cinfo.list); INIT_WORK(&dreq->work, nfs_direct_write_schedule_work); - memset(&dreq->ds_cinfo, 0, sizeof(dreq->ds_cinfo)); - dreq->iocb = NULL; - dreq->ctx = NULL; - dreq->l_ctx = NULL; spin_lock_init(&dreq->lock); - atomic_set(&dreq->io_count, 0); - dreq->count = 0; - dreq->error = 0; - dreq->flags = 0; return dreq; } -- cgit v1.2.3 From 4bd8b010136afa0df9122a08bad361686bda0a1d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 1 May 2012 12:49:58 -0400 Subject: NFS: Simplify the nfs_read_completion functions Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 48 ++++++++++++++++++++---------------------------- fs/nfs/read.c | 44 ++++++++++++++++++-------------------------- 2 files changed, 38 insertions(+), 54 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index f17e4695c842..aab3016c437c 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -243,36 +243,28 @@ static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) dreq->count += hdr->good_bytes; spin_unlock(&dreq->lock); - if (!test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); - struct page *page = req->wb_page; - - if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { - if (bytes > hdr->good_bytes) - zero_user(page, 0, PAGE_SIZE); - else if (hdr->good_bytes - bytes < PAGE_SIZE) - zero_user_segment(page, - hdr->good_bytes & ~PAGE_MASK, - PAGE_SIZE); - } - bytes += req->wb_bytes; - nfs_list_remove_request(req); - if (!PageCompound(page)) - set_page_dirty(page); - nfs_direct_readpage_release(req); + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { + if (bytes > hdr->good_bytes) + zero_user(page, 0, PAGE_SIZE); + else if (hdr->good_bytes - bytes < PAGE_SIZE) + zero_user_segment(page, + hdr->good_bytes & ~PAGE_MASK, + PAGE_SIZE); } - } else { - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); - - if (bytes < hdr->good_bytes) - if (!PageCompound(req->wb_page)) - set_page_dirty(req->wb_page); - bytes += req->wb_bytes; - nfs_list_remove_request(req); - nfs_direct_readpage_release(req); + if (!PageCompound(page)) { + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { + if (bytes < hdr->good_bytes) + set_page_dirty(page); + } else + set_page_dirty(page); } + bytes += req->wb_bytes; + nfs_list_remove_request(req); + nfs_direct_readpage_release(req); } out_put: if (put_dreq(dreq)) diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 1961a192f1cb..37c9eb2930b0 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -179,34 +179,26 @@ static void nfs_read_completion(struct nfs_pgio_header *hdr) if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) goto out; - if (!test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); - struct page *page = req->wb_page; - - if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { - if (bytes > hdr->good_bytes) - zero_user(page, 0, PAGE_SIZE); - else if (hdr->good_bytes - bytes < PAGE_SIZE) - zero_user_segment(page, - hdr->good_bytes & ~PAGE_MASK, - PAGE_SIZE); - } - SetPageUptodate(page); - nfs_list_remove_request(req); - nfs_readpage_release(req); - bytes += PAGE_SIZE; + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { + if (bytes > hdr->good_bytes) + zero_user(page, 0, PAGE_SIZE); + else if (hdr->good_bytes - bytes < PAGE_SIZE) + zero_user_segment(page, + hdr->good_bytes & ~PAGE_MASK, + PAGE_SIZE); } - } else { - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); - - bytes += req->wb_bytes; + bytes += req->wb_bytes; + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { if (bytes <= hdr->good_bytes) - SetPageUptodate(req->wb_page); - nfs_list_remove_request(req); - nfs_readpage_release(req); - } + SetPageUptodate(page); + } else + SetPageUptodate(page); + nfs_list_remove_request(req); + nfs_readpage_release(req); } out: hdr->release(hdr); -- cgit v1.2.3 From bf5fc4028ef751904a114ffc4b5d2cd9f0233142 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 4 May 2012 13:47:16 -0400 Subject: NFS: Fix O_DIRECT compile warnings Fix the following compile warnings: fs/nfs/direct.c: In function 'nfs_direct_read_schedule_segment': fs/nfs/direct.c:325:11: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c:325:11: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c:325:11: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c:352:27: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c: In function 'nfs_direct_write_schedule_segment': fs/nfs/direct.c:622:11: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c:622:11: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c:622:11: warning: comparison of distinct pointer types lacks a cast [enabled by default] fs/nfs/direct.c:650:27: warning: comparison of distinct pointer types lacks a cast [enabled by default] Reported-by: Stephen Rothwell Signed-off-by: Trond Myklebust Cc: Fred Isaman Cc: Stephen Rothwell --- fs/nfs/direct.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index aab3016c437c..dca9c8143009 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -322,7 +322,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de int i; pgbase = user_addr & ~PAGE_MASK; - bytes = min(max(rsize, PAGE_SIZE), count); + bytes = min(max_t(size_t, rsize, PAGE_SIZE), count); result = -ENOMEM; npages = nfs_page_array_len(pgbase, bytes); @@ -349,7 +349,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de for (i = 0; i < npages; i++) { struct nfs_page *req; - unsigned int req_len = min(bytes, PAGE_SIZE - pgbase); + unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); /* XXX do we need to do the eof zeroing found in async_filler? */ req = nfs_create_request(dreq->ctx, dreq->inode, pagevec[i], @@ -619,7 +619,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d int i; pgbase = user_addr & ~PAGE_MASK; - bytes = min(max(wsize, PAGE_SIZE), count); + bytes = min(max_t(size_t, wsize, PAGE_SIZE), count); result = -ENOMEM; npages = nfs_page_array_len(pgbase, bytes); @@ -647,7 +647,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d for (i = 0; i < npages; i++) { struct nfs_page *req; - unsigned int req_len = min(bytes, PAGE_SIZE - pgbase); + unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); req = nfs_create_request(dreq->ctx, dreq->inode, pagevec[i], -- cgit v1.2.3 From 1385b8117325e79f74c1e7d1cbf45c789deb85c5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 4 May 2012 13:54:24 -0400 Subject: NFS: Fix sparse warnings Fix the following sparse warnings: fs/nfs/direct.c:221:6: warning: symbol 'nfs_direct_readpage_release' was not declared. Should it be static? fs/nfs/read.c:38:43: warning: non-ANSI function declaration of function 'nfs_readhdr_alloc' fs/nfs/objlayout/objio_osd.c:214:5: warning: symbol '__alloc_objio_seg' was not declared. Should it be static? Reported-by: Dan Carpenter Signed-off-by: Trond Myklebust Cc: Fred Isaman Cc: Boaz Harrosh --- fs/nfs/direct.c | 2 +- fs/nfs/objlayout/objio_osd.c | 2 +- fs/nfs/read.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index dca9c8143009..257d0091148b 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -218,7 +218,7 @@ static void nfs_direct_complete(struct nfs_direct_req *dreq) nfs_direct_req_release(dreq); } -void nfs_direct_readpage_release(struct nfs_page *req) +static void nfs_direct_readpage_release(struct nfs_page *req) { dprintk("NFS: direct read done (%s/%lld %d@%lld)\n", req->wb_context->dentry->d_inode->i_sb->s_id, diff --git a/fs/nfs/objlayout/objio_osd.c b/fs/nfs/objlayout/objio_osd.c index fbf4874ec252..b47277baebab 100644 --- a/fs/nfs/objlayout/objio_osd.c +++ b/fs/nfs/objlayout/objio_osd.c @@ -211,7 +211,7 @@ static void copy_single_comp(struct ore_components *oc, unsigned c, memcpy(ocomp->cred, src_comp->oc_cap.cred, sizeof(ocomp->cred)); } -int __alloc_objio_seg(unsigned numdevs, gfp_t gfp_flags, +static int __alloc_objio_seg(unsigned numdevs, gfp_t gfp_flags, struct objio_segment **pseg) { /* This is the in memory structure of the objio_segment diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 37c9eb2930b0..f23cf255bd39 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -35,7 +35,7 @@ static const struct nfs_pgio_completion_ops nfs_async_read_completion_ops; static struct kmem_cache *nfs_rdata_cachep; -struct nfs_read_header *nfs_readhdr_alloc() +struct nfs_read_header *nfs_readhdr_alloc(void) { struct nfs_read_header *rhdr; -- cgit v1.2.3 From 7ad84aa9448571678c243f0c5ef383fbe5b50f4f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 9 May 2012 13:19:15 -0400 Subject: NFS: Clean up - simplify nfs_lock_request() We only have two places where we need to grab a reference when trying to lock the nfs_page. We're better off making that explicit. Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 1 + fs/nfs/write.c | 11 ++++++----- include/linux/nfs_page.h | 14 ++------------ 3 files changed, 9 insertions(+), 17 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 257d0091148b..465ea84a2874 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -657,6 +657,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d break; } nfs_lock_request(req); + kref_get(&req->wb_kref); req->wb_index = pos >> PAGE_SHIFT; req->wb_offset = pos & ~PAGE_MASK; if (!nfs_pageio_add_request(desc, req)) { diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 83823294ffd2..553f7ef1079b 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -260,10 +260,10 @@ static struct nfs_page *nfs_find_and_lock_request(struct page *page, bool nonblo req = nfs_page_find_request_locked(page); if (req == NULL) break; - if (nfs_lock_request_dontget(req)) + if (nfs_lock_request(req)) break; /* Note: If we hold the page lock, as is the case in nfs_writepage, - * then the call to nfs_lock_request_dontget() will always + * then the call to nfs_lock_request() will always * succeed provided that someone hasn't already marked the * request as dirty (in which case we don't care). */ @@ -406,7 +406,7 @@ static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req) struct nfs_inode *nfsi = NFS_I(inode); /* Lock the request! */ - nfs_lock_request_dontget(req); + nfs_lock_request(req); spin_lock(&inode->i_lock); if (!nfsi->npages && nfs_have_delegation(inode, FMODE_WRITE)) @@ -651,6 +651,7 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst, list_for_each_entry_safe(req, tmp, src, wb_list) { if (!nfs_lock_request(req)) continue; + kref_get(&req->wb_kref); if (cond_resched_lock(cinfo->lock)) list_safe_reset_next(req, tmp, wb_list); nfs_request_remove_commit_list(req, cinfo); @@ -741,7 +742,7 @@ static struct nfs_page *nfs_try_to_update_request(struct inode *inode, || end < req->wb_offset) goto out_flushme; - if (nfs_lock_request_dontget(req)) + if (nfs_lock_request(req)) break; /* The request is locked, so wait and then retry */ @@ -1717,7 +1718,7 @@ int nfs_wb_page_cancel(struct inode *inode, struct page *page) req = nfs_page_find_request(page); if (req == NULL) break; - if (nfs_lock_request_dontget(req)) { + if (nfs_lock_request(req)) { nfs_clear_request_commit(req); nfs_inode_remove_request(req); /* diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index ef7504215446..263f30a5e10d 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -99,24 +99,14 @@ extern void nfs_unlock_request(struct nfs_page *req); extern void nfs_unlock_request_dont_release(struct nfs_page *req); /* - * Lock the page of an asynchronous request without getting a new reference + * Lock the page of an asynchronous request */ -static inline int -nfs_lock_request_dontget(struct nfs_page *req) -{ - return !test_and_set_bit(PG_BUSY, &req->wb_flags); -} - static inline int nfs_lock_request(struct nfs_page *req) { - if (test_and_set_bit(PG_BUSY, &req->wb_flags)) - return 0; - kref_get(&req->wb_kref); - return 1; + return !test_and_set_bit(PG_BUSY, &req->wb_flags); } - /** * nfs_list_add_request - Insert a request into a list * @req: request -- cgit v1.2.3 From 1d1afcbc294cc7c788eb5c7b6b98e8d63caf002c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 9 May 2012 14:04:55 -0400 Subject: NFS: Clean up - Rename nfs_unlock_request and nfs_unlock_request_dont_release Function rename to ensure that the functionality of nfs_unlock_request() mirrors that of nfs_lock_request(). Then let nfs_unlock_and_release_request() do the work of what used to be called nfs_unlock_request()... Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 10 +++++----- fs/nfs/pagelist.c | 11 ++++++----- fs/nfs/write.c | 12 ++++++------ include/linux/nfs_page.h | 2 +- 4 files changed, 18 insertions(+), 17 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 465ea84a2874..845e20196803 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -488,7 +488,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) while (!list_empty(&failed)) { nfs_release_request(req); - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); } if (put_dreq(dreq)) @@ -521,7 +521,7 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) nfs_mark_request_commit(req, NULL, &cinfo); } else nfs_release_request(req); - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); } if (atomic_dec_and_test(&cinfo.mds->rpcs_out)) @@ -662,7 +662,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d req->wb_offset = pos & ~PAGE_MASK; if (!nfs_pageio_add_request(desc, req)) { result = desc->pg_error; - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); nfs_release_request(req); break; } @@ -739,7 +739,7 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) default: nfs_release_request(req); } - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); } out_put: @@ -756,7 +756,7 @@ static void nfs_write_sync_pgio_error(struct list_head *head) req = nfs_list_entry(head->next); nfs_list_remove_request(req); nfs_release_request(req); - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); } } diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 69146f386989..aed913c833f4 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -128,10 +128,10 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode, } /** - * nfs_unlock_request_dont_release - Unlock request and wake up sleepers. + * nfs_unlock_request - Unlock request and wake up sleepers. * @req: */ -void nfs_unlock_request_dont_release(struct nfs_page *req) +void nfs_unlock_request(struct nfs_page *req) { if (!NFS_WBACK_BUSY(req)) { printk(KERN_ERR "NFS: Invalid unlock attempted\n"); @@ -144,11 +144,12 @@ void nfs_unlock_request_dont_release(struct nfs_page *req) } /** - * nfs_unlock_request - Unlock request and release the nfs_page + * nfs_unlock_and_release_request - Unlock request and release the nfs_page + * @req: */ -void nfs_unlock_request(struct nfs_page *req) +void nfs_unlock_and_release_request(struct nfs_page *req) { - nfs_unlock_request_dont_release(req); + nfs_unlock_request(req); nfs_release_request(req); } diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 553f7ef1079b..8ffd7d5ed58b 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -625,7 +625,7 @@ static void nfs_write_completion(struct nfs_pgio_header *hdr) remove_req: nfs_inode_remove_request(req); next: - nfs_unlock_request_dont_release(req); + nfs_unlock_request(req); nfs_end_page_writeback(req->wb_page); nfs_release_request(req); } @@ -812,7 +812,7 @@ static int nfs_writepage_setup(struct nfs_open_context *ctx, struct page *page, nfs_grow_file(page, offset, count); nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes); nfs_mark_request_dirty(req); - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); return 0; } @@ -1039,7 +1039,7 @@ static int nfs_do_multiple_writes(struct list_head *head, static void nfs_redirty_request(struct nfs_page *req) { nfs_mark_request_dirty(req); - nfs_unlock_request_dont_release(req); + nfs_unlock_request(req); nfs_end_page_writeback(req->wb_page); nfs_release_request(req); } @@ -1477,7 +1477,7 @@ void nfs_retry_commit(struct list_head *page_list, dec_bdi_stat(req->wb_page->mapping->backing_dev_info, BDI_RECLAIMABLE); } - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); } } EXPORT_SYMBOL_GPL(nfs_retry_commit); @@ -1555,7 +1555,7 @@ static void nfs_commit_release_pages(struct nfs_commit_data *data) dprintk(" mismatch\n"); nfs_mark_request_dirty(req); next: - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); } nfs_init_cinfo(&cinfo, data->inode, data->dreq); if (atomic_dec_and_test(&cinfo.mds->rpcs_out)) @@ -1726,7 +1726,7 @@ int nfs_wb_page_cancel(struct inode *inode, struct page *page) * page as being dirty */ cancel_dirty_page(page, PAGE_CACHE_SIZE); - nfs_unlock_request(req); + nfs_unlock_and_release_request(req); break; } ret = nfs_wait_on_request(req); diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index 263f30a5e10d..88d166b555e8 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -96,7 +96,7 @@ extern bool nfs_generic_pg_test(struct nfs_pageio_descriptor *desc, struct nfs_page *req); extern int nfs_wait_on_request(struct nfs_page *); extern void nfs_unlock_request(struct nfs_page *req); -extern void nfs_unlock_request_dont_release(struct nfs_page *req); +extern void nfs_unlock_and_release_request(struct nfs_page *req); /* * Lock the page of an asynchronous request -- cgit v1.2.3 From 0427708657750bdc03af3491a0297cab5e7efabf Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 9 May 2012 13:54:53 -0400 Subject: NFS: Clean up - Simplify reference counting in fs/nfs/direct.c Signed-off-by: Trond Myklebust Cc: Fred Isaman --- fs/nfs/direct.c | 15 ++++----------- 1 file changed, 4 insertions(+), 11 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 845e20196803..c47a46eaf905 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -486,10 +486,8 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) } nfs_pageio_complete(&desc); - while (!list_empty(&failed)) { - nfs_release_request(req); + while (!list_empty(&failed)) nfs_unlock_and_release_request(req); - } if (put_dreq(dreq)) nfs_direct_write_complete(dreq, dreq->inode); @@ -518,9 +516,9 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) nfs_list_remove_request(req); if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) { /* Note the rewrite will go through mds */ + kref_get(&req->wb_kref); nfs_mark_request_commit(req, NULL, &cinfo); - } else - nfs_release_request(req); + } nfs_unlock_and_release_request(req); } @@ -657,13 +655,11 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *d break; } nfs_lock_request(req); - kref_get(&req->wb_kref); req->wb_index = pos >> PAGE_SHIFT; req->wb_offset = pos & ~PAGE_MASK; if (!nfs_pageio_add_request(desc, req)) { result = desc->pg_error; nfs_unlock_and_release_request(req); - nfs_release_request(req); break; } pgbase = 0; @@ -734,10 +730,8 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) switch (bit) { case NFS_IOHDR_NEED_RESCHED: case NFS_IOHDR_NEED_COMMIT: + kref_get(&req->wb_kref); nfs_mark_request_commit(req, hdr->lseg, &cinfo); - break; - default: - nfs_release_request(req); } nfs_unlock_and_release_request(req); } @@ -755,7 +749,6 @@ static void nfs_write_sync_pgio_error(struct list_head *head) while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_release_request(req); nfs_unlock_and_release_request(req); } } -- cgit v1.2.3 From 2701d086dbfca03b2d28b25c6dc11dd78d0e26ad Mon Sep 17 00:00:00 2001 From: Andy Adamson Date: Thu, 24 May 2012 13:13:24 -0400 Subject: NFSv4.1 add nfs_inode book keeping for mdsthreshold Keep track of the number of bytes read or written via buffered, direct, and mem-mapped i/o for use by mdsthreshold size_io hints. Signed-off-by: Andy Adamson Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 2 ++ fs/nfs/file.c | 1 + fs/nfs/inode.c | 2 ++ fs/nfs/pnfs.c | 3 +++ fs/nfs/read.c | 2 ++ include/linux/nfs_fs.h | 3 +++ 6 files changed, 13 insertions(+) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index c47a46eaf905..23d170bc44f4 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -447,6 +447,7 @@ static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov, result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos); if (!result) result = nfs_direct_wait(dreq); + NFS_I(inode)->read_io += result; out_release: nfs_direct_req_release(dreq); out: @@ -785,6 +786,7 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, pos += vec->iov_len; } nfs_pageio_complete(&desc); + NFS_I(dreq->inode)->write_io += desc.pg_bytes_written; /* * If no bytes were started, return the error, and let the diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 8eda8a6644c3..56311ca5f9f8 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -424,6 +424,7 @@ static int nfs_write_end(struct file *file, struct address_space *mapping, if (status < 0) return status; + NFS_I(mapping->host)->write_io += copied; return copied; } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 889f7e5e92e1..a6f5fbbe9b9c 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -323,6 +323,8 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr) inode->i_gid = -2; inode->i_blocks = 0; memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf)); + nfsi->write_io = 0; + nfsi->read_io = 0; nfsi->read_cache_jiffies = fattr->time_start; nfsi->attr_gencount = fattr->gencount; diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index cbcb6aea58a3..6620606f2687 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -395,6 +395,9 @@ mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, dprintk("%s:Begin lo %p\n", __func__, lo); if (list_empty(&lo->plh_segs)) { + /* Reset MDS Threshold I/O counters */ + NFS_I(lo->plh_inode)->write_io = 0; + NFS_I(lo->plh_inode)->read_io = 0; if (!test_and_set_bit(NFS_LAYOUT_DESTROYED, &lo->plh_flags)) put_layout_hdr_locked(lo); return 0; diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 2cfdd7785411..86ced7836214 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -152,6 +152,7 @@ int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode, nfs_pageio_init_read(&pgio, inode, &nfs_async_read_completion_ops); nfs_pageio_add_request(&pgio, new); nfs_pageio_complete(&pgio); + NFS_I(inode)->read_io += pgio.pg_bytes_written; return 0; } @@ -656,6 +657,7 @@ int nfs_readpages(struct file *filp, struct address_space *mapping, ret = read_cache_pages(mapping, pages, readpage_async_filler, &desc); nfs_pageio_complete(&pgio); + NFS_I(inode)->read_io += pgio.pg_bytes_written; npages = (pgio.pg_bytes_written + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; nfs_add_stats(inode, NFSIOS_READPAGES, npages); read_complete: diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index ce910cb7d761..b23cfc120edb 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -202,6 +202,9 @@ struct nfs_inode { /* pNFS layout information */ struct pnfs_layout_hdr *layout; #endif /* CONFIG_NFS_V4*/ + /* how many bytes have been written/read and how many bytes queued up */ + __u64 write_io; + __u64 read_io; #ifdef CONFIG_NFS_FSCACHE struct fscache_cookie *fscache; #endif -- cgit v1.2.3 From 1d59d61f606547f0712aa6971f91f71154071c99 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 31 May 2012 12:22:33 -0400 Subject: NFS: Ensure that setattr and getattr wait for O_DIRECT write completion Use the same mechanism as the block devices are using, but move the helper functions from fs/direct-io.c into fs/inode.c to remove the dependency on CONFIG_BLOCK. Signed-off-by: Trond Myklebust Cc: Christoph Hellwig Cc: Al Viro Cc: Fred Isaman Signed-off-by: Linus Torvalds --- fs/direct-io.c | 44 -------------------------------------------- fs/inode.c | 47 +++++++++++++++++++++++++++++++++++++++++++++++ fs/nfs/direct.c | 15 ++++++++++++--- fs/nfs/inode.c | 5 ++++- fs/nfs/internal.h | 4 ++++ include/linux/fs.h | 9 +++------ 6 files changed, 70 insertions(+), 54 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/direct-io.c b/fs/direct-io.c index f4aadd15b613..0c85fae37666 100644 --- a/fs/direct-io.c +++ b/fs/direct-io.c @@ -145,50 +145,6 @@ struct dio { static struct kmem_cache *dio_cache __read_mostly; -static void __inode_dio_wait(struct inode *inode) -{ - wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP); - DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP); - - do { - prepare_to_wait(wq, &q.wait, TASK_UNINTERRUPTIBLE); - if (atomic_read(&inode->i_dio_count)) - schedule(); - } while (atomic_read(&inode->i_dio_count)); - finish_wait(wq, &q.wait); -} - -/** - * inode_dio_wait - wait for outstanding DIO requests to finish - * @inode: inode to wait for - * - * Waits for all pending direct I/O requests to finish so that we can - * proceed with a truncate or equivalent operation. - * - * Must be called under a lock that serializes taking new references - * to i_dio_count, usually by inode->i_mutex. - */ -void inode_dio_wait(struct inode *inode) -{ - if (atomic_read(&inode->i_dio_count)) - __inode_dio_wait(inode); -} -EXPORT_SYMBOL(inode_dio_wait); - -/* - * inode_dio_done - signal finish of a direct I/O requests - * @inode: inode the direct I/O happens on - * - * This is called once we've finished processing a direct I/O request, - * and is used to wake up callers waiting for direct I/O to be quiesced. - */ -void inode_dio_done(struct inode *inode) -{ - if (atomic_dec_and_test(&inode->i_dio_count)) - wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); -} -EXPORT_SYMBOL(inode_dio_done); - /* * How many pages are in the queue? */ diff --git a/fs/inode.c b/fs/inode.c index 6bc8761cc333..c474c1d7062b 100644 --- a/fs/inode.c +++ b/fs/inode.c @@ -1748,3 +1748,50 @@ bool inode_owner_or_capable(const struct inode *inode) return false; } EXPORT_SYMBOL(inode_owner_or_capable); + +/* + * Direct i/o helper functions + */ +static void __inode_dio_wait(struct inode *inode) +{ + wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP); + DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP); + + do { + prepare_to_wait(wq, &q.wait, TASK_UNINTERRUPTIBLE); + if (atomic_read(&inode->i_dio_count)) + schedule(); + } while (atomic_read(&inode->i_dio_count)); + finish_wait(wq, &q.wait); +} + +/** + * inode_dio_wait - wait for outstanding DIO requests to finish + * @inode: inode to wait for + * + * Waits for all pending direct I/O requests to finish so that we can + * proceed with a truncate or equivalent operation. + * + * Must be called under a lock that serializes taking new references + * to i_dio_count, usually by inode->i_mutex. + */ +void inode_dio_wait(struct inode *inode) +{ + if (atomic_read(&inode->i_dio_count)) + __inode_dio_wait(inode); +} +EXPORT_SYMBOL(inode_dio_wait); + +/* + * inode_dio_done - signal finish of a direct I/O requests + * @inode: inode the direct I/O happens on + * + * This is called once we've finished processing a direct I/O request, + * and is used to wake up callers waiting for direct I/O to be quiesced. + */ +void inode_dio_done(struct inode *inode) +{ + if (atomic_dec_and_test(&inode->i_dio_count)) + wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); +} +EXPORT_SYMBOL(inode_dio_done); diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 23d170bc44f4..ad2775d3e219 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -454,6 +454,12 @@ out: return result; } +static void nfs_inode_dio_write_done(struct inode *inode) +{ + nfs_zap_mapping(inode, inode->i_mapping); + inode_dio_done(inode); +} + #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) { @@ -564,7 +570,7 @@ static void nfs_direct_write_schedule_work(struct work_struct *work) nfs_direct_write_reschedule(dreq); break; default: - nfs_zap_mapping(dreq->inode, dreq->inode->i_mapping); + nfs_inode_dio_write_done(dreq->inode); nfs_direct_complete(dreq); } } @@ -581,7 +587,7 @@ static void nfs_direct_write_schedule_work(struct work_struct *work) static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) { - nfs_zap_mapping(inode, inode->i_mapping); + nfs_inode_dio_write_done(inode); nfs_direct_complete(dreq); } #endif @@ -766,14 +772,16 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, loff_t pos) { struct nfs_pageio_descriptor desc; + struct inode *inode = dreq->inode; ssize_t result = 0; size_t requested_bytes = 0; unsigned long seg; - nfs_pageio_init_write(&desc, dreq->inode, FLUSH_COND_STABLE, + nfs_pageio_init_write(&desc, inode, FLUSH_COND_STABLE, &nfs_direct_write_completion_ops); desc.pg_dreq = dreq; get_dreq(dreq); + atomic_inc(&inode->i_dio_count); for (seg = 0; seg < nr_segs; seg++) { const struct iovec *vec = &iov[seg]; @@ -793,6 +801,7 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, * generic layer handle the completion. */ if (requested_bytes == 0) { + inode_dio_done(inode); nfs_direct_req_release(dreq); return result < 0 ? result : -EIO; } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 2f6f78c4b42d..e605d695dbcb 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -418,8 +418,10 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) return 0; /* Write all dirty data */ - if (S_ISREG(inode->i_mode)) + if (S_ISREG(inode->i_mode)) { + nfs_inode_dio_wait(inode); nfs_wb_all(inode); + } fattr = nfs_alloc_fattr(); if (fattr == NULL) @@ -503,6 +505,7 @@ int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) /* Flush out writes to the server in order to update c/mtime. */ if (S_ISREG(inode->i_mode)) { + nfs_inode_dio_wait(inode); err = filemap_write_and_wait(inode->i_mapping); if (err) goto out; diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 1848a7275592..18f99ef71343 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -369,6 +369,10 @@ extern int nfs_migrate_page(struct address_space *, /* direct.c */ void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, struct nfs_direct_req *dreq); +static inline void nfs_inode_dio_wait(struct inode *inode) +{ + inode_dio_wait(inode); +} /* nfs4proc.c */ extern void __nfs4_read_done_cb(struct nfs_read_data *); diff --git a/include/linux/fs.h b/include/linux/fs.h index 038076b27ea4..598a5892ff2b 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -2453,8 +2453,6 @@ enum { }; void dio_end_io(struct bio *bio, int error); -void inode_dio_wait(struct inode *inode); -void inode_dio_done(struct inode *inode); ssize_t __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, struct block_device *bdev, const struct iovec *iov, loff_t offset, @@ -2469,12 +2467,11 @@ static inline ssize_t blockdev_direct_IO(int rw, struct kiocb *iocb, offset, nr_segs, get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES); } -#else -static inline void inode_dio_wait(struct inode *inode) -{ -} #endif +void inode_dio_wait(struct inode *inode); +void inode_dio_done(struct inode *inode); + extern const struct file_operations generic_ro_fops; #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) -- cgit v1.2.3