From 1b1bc66bb4ebd737fb750b0e1afc753098f4a1b2 Mon Sep 17 00:00:00 2001 From: Weston Andros Adamson Date: Fri, 1 Apr 2016 11:42:28 -0400 Subject: pnfs: set NFS_IOHDR_REDO in pnfs_read_resend_pnfs Like other resend paths, mark the (old) hdr as NFS_IOHDR_REDO. This ensures the hdr completion function will not count the (old) hdr as good bytes. Also, vector the error back through the hdr->task.tk_status like other retry calls. This fixes a bug with the FlexFiles layout where libaio was reporting more bytes read than requested. Signed-off-by: Weston Andros Adamson Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 14 ++++++++------ fs/nfs/pnfs.h | 2 +- 2 files changed, 9 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 89a5ef4df08a..5b404d926e08 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2143,12 +2143,15 @@ pnfs_try_to_read_data(struct nfs_pgio_header *hdr, } /* Resend all requests through pnfs. */ -int pnfs_read_resend_pnfs(struct nfs_pgio_header *hdr) +void pnfs_read_resend_pnfs(struct nfs_pgio_header *hdr) { struct nfs_pageio_descriptor pgio; - nfs_pageio_init_read(&pgio, hdr->inode, false, hdr->completion_ops); - return nfs_pageio_resend(&pgio, hdr); + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) { + nfs_pageio_init_read(&pgio, hdr->inode, false, + hdr->completion_ops); + hdr->task.tk_status = nfs_pageio_resend(&pgio, hdr); + } } EXPORT_SYMBOL_GPL(pnfs_read_resend_pnfs); @@ -2158,12 +2161,11 @@ pnfs_do_read(struct nfs_pageio_descriptor *desc, struct nfs_pgio_header *hdr) const struct rpc_call_ops *call_ops = desc->pg_rpc_callops; struct pnfs_layout_segment *lseg = desc->pg_lseg; enum pnfs_try_status trypnfs; - int err = 0; trypnfs = pnfs_try_to_read_data(hdr, call_ops, lseg); if (trypnfs == PNFS_TRY_AGAIN) - err = pnfs_read_resend_pnfs(hdr); - if (trypnfs == PNFS_NOT_ATTEMPTED || err) + pnfs_read_resend_pnfs(hdr); + if (trypnfs == PNFS_NOT_ATTEMPTED || hdr->task.tk_status) pnfs_read_through_mds(desc, hdr); } diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 1ac1db5f6dad..7222d3a35439 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -282,7 +282,7 @@ int _pnfs_return_layout(struct inode *); int pnfs_commit_and_return_layout(struct inode *); void pnfs_ld_write_done(struct nfs_pgio_header *); void pnfs_ld_read_done(struct nfs_pgio_header *); -int pnfs_read_resend_pnfs(struct nfs_pgio_header *); +void pnfs_read_resend_pnfs(struct nfs_pgio_header *); struct pnfs_layout_segment *pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx, loff_t pos, -- cgit v1.2.3-59-g8ed1b From ed3743a6d4f309f5278c87a69e3125382b78455c Mon Sep 17 00:00:00 2001 From: Weston Andros Adamson Date: Fri, 1 Apr 2016 11:42:29 -0400 Subject: nfs: add debug to directio "good_bytes" counting This will pop a warning if we count too many good bytes. Signed-off-by: Weston Andros Adamson Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index c93826e4a8c6..f79d98ae4c10 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -87,6 +87,7 @@ struct nfs_direct_req { int mirror_count; ssize_t count, /* bytes actually processed */ + max_count, /* max expected count */ bytes_left, /* bytes left to be sent */ io_start, /* start of IO */ error; /* any reported error */ @@ -123,6 +124,8 @@ nfs_direct_good_bytes(struct nfs_direct_req *dreq, struct nfs_pgio_header *hdr) int i; ssize_t count; + WARN_ON_ONCE(dreq->count >= dreq->max_count); + if (dreq->mirror_count == 1) { dreq->mirrors[hdr->pgio_mirror_idx].count += hdr->good_bytes; dreq->count += hdr->good_bytes; @@ -593,7 +596,7 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter, goto out_unlock; dreq->inode = inode; - dreq->bytes_left = count; + dreq->bytes_left = dreq->max_count = count; dreq->io_start = pos; dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); l_ctx = nfs_get_lock_context(dreq->ctx); @@ -1026,7 +1029,7 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) goto out_unlock; dreq->inode = inode; - dreq->bytes_left = iov_iter_count(iter); + dreq->bytes_left = dreq->max_count = iov_iter_count(iter); dreq->io_start = pos; dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); l_ctx = nfs_get_lock_context(dreq->ctx); -- cgit v1.2.3-59-g8ed1b From fe238e601d2519f259103ab65caea3b077ed7b39 Mon Sep 17 00:00:00 2001 From: Dave Wysochanski Date: Fri, 1 Apr 2016 13:45:09 -0400 Subject: NFS: Save struct inode * inside nfs_commit_info to clarify usage of i_lock Commit ea2cf22 created nfs_commit_info and saved &inode->i_lock inside this NFS specific structure. This obscures the usage of i_lock. Instead, save struct inode * so later it's clear the spinlock taken is i_lock. Should be no functional change. Signed-off-by: Dave Wysochanski Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 10 +++++----- fs/nfs/filelayout/filelayout.c | 4 ++-- fs/nfs/flexfilelayout/flexfilelayout.c | 4 ++-- fs/nfs/pnfs_nfs.c | 32 ++++++++++++++++---------------- fs/nfs/write.c | 16 ++++++++-------- include/linux/nfs_xdr.h | 2 +- 6 files changed, 34 insertions(+), 34 deletions(-) (limited to 'fs') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index f79d98ae4c10..7f03163b5364 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -278,7 +278,7 @@ static void nfs_direct_release_pages(struct page **pages, unsigned int npages) void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, struct nfs_direct_req *dreq) { - cinfo->lock = &dreq->inode->i_lock; + cinfo->inode = dreq->inode; cinfo->mds = &dreq->mds_cinfo; cinfo->ds = &dreq->ds_cinfo; cinfo->dreq = dreq; @@ -635,13 +635,13 @@ nfs_direct_write_scan_commit_list(struct inode *inode, struct list_head *list, struct nfs_commit_info *cinfo) { - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); #ifdef CONFIG_NFS_V4_1 if (cinfo->ds != NULL && cinfo->ds->nwritten != 0) NFS_SERVER(inode)->pnfs_curr_ld->recover_commit_reqs(list, cinfo); #endif nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0); - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); } static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) @@ -676,13 +676,13 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) if (!nfs_pageio_add_request(&desc, req)) { nfs_list_remove_request(req); nfs_list_add_request(req, &failed); - spin_lock(cinfo.lock); + spin_lock(&cinfo.inode->i_lock); dreq->flags = 0; if (desc.pg_error < 0) dreq->error = desc.pg_error; else dreq->error = -EIO; - spin_unlock(cinfo.lock); + spin_unlock(&cinfo.inode->i_lock); } nfs_release_request(req); } diff --git a/fs/nfs/filelayout/filelayout.c b/fs/nfs/filelayout/filelayout.c index 3384dc8e6683..3e50057eeadf 100644 --- a/fs/nfs/filelayout/filelayout.c +++ b/fs/nfs/filelayout/filelayout.c @@ -795,7 +795,7 @@ filelayout_alloc_commit_info(struct pnfs_layout_segment *lseg, buckets[i].direct_verf.committed = NFS_INVALID_STABLE_HOW; } - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); if (cinfo->ds->nbuckets >= size) goto out; for (i = 0; i < cinfo->ds->nbuckets; i++) { @@ -811,7 +811,7 @@ filelayout_alloc_commit_info(struct pnfs_layout_segment *lseg, swap(cinfo->ds->buckets, buckets); cinfo->ds->nbuckets = size; out: - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); kfree(buckets); return 0; } diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 0cb1abd535e3..3b398f7b4637 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -745,7 +745,7 @@ ff_layout_alloc_commit_info(struct pnfs_layout_segment *lseg, else { int i; - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); if (cinfo->ds->nbuckets != 0) kfree(buckets); else { @@ -759,7 +759,7 @@ ff_layout_alloc_commit_info(struct pnfs_layout_segment *lseg, NFS_INVALID_STABLE_HOW; } } - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); return 0; } } diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index 4aaed890048f..d2a7c9f7aa94 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -61,7 +61,7 @@ EXPORT_SYMBOL_GPL(pnfs_generic_commit_release); /* The generic layer is about to remove the req from the commit list. * If this will make the bucket empty, it will need to put the lseg reference. - * Note this must be called holding the inode (/cinfo) lock + * Note this must be called holding i_lock */ void pnfs_generic_clear_request_commit(struct nfs_page *req, @@ -98,7 +98,7 @@ pnfs_generic_transfer_commit_list(struct list_head *src, struct list_head *dst, if (!nfs_lock_request(req)) continue; kref_get(&req->wb_kref); - if (cond_resched_lock(cinfo->lock)) + if (cond_resched_lock(&cinfo->inode->i_lock)) list_safe_reset_next(req, tmp, wb_list); nfs_request_remove_commit_list(req, cinfo); clear_bit(PG_COMMIT_TO_DS, &req->wb_flags); @@ -119,7 +119,7 @@ pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, struct list_head *dst = &bucket->committing; int ret; - lockdep_assert_held(cinfo->lock); + lockdep_assert_held(&cinfo->inode->i_lock); ret = pnfs_generic_transfer_commit_list(src, dst, cinfo, max); if (ret) { cinfo->ds->nwritten -= ret; @@ -142,7 +142,7 @@ int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo, { int i, rv = 0, cnt; - lockdep_assert_held(cinfo->lock); + lockdep_assert_held(&cinfo->inode->i_lock); for (i = 0; i < cinfo->ds->nbuckets && max != 0; i++) { cnt = pnfs_generic_scan_ds_commit_list(&cinfo->ds->buckets[i], cinfo, max); @@ -161,16 +161,16 @@ void pnfs_generic_recover_commit_reqs(struct list_head *dst, struct pnfs_layout_segment *freeme; int i; - lockdep_assert_held(cinfo->lock); + lockdep_assert_held(&cinfo->inode->i_lock); restart: for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) { if (pnfs_generic_transfer_commit_list(&b->written, dst, cinfo, 0)) { freeme = b->wlseg; b->wlseg = NULL; - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); pnfs_put_lseg(freeme); - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); goto restart; } } @@ -186,7 +186,7 @@ static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx) LIST_HEAD(pages); int i; - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); for (i = idx; i < fl_cinfo->nbuckets; i++) { bucket = &fl_cinfo->buckets[i]; if (list_empty(&bucket->committing)) @@ -194,12 +194,12 @@ static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx) freeme = bucket->clseg; bucket->clseg = NULL; list_splice_init(&bucket->committing, &pages); - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); nfs_retry_commit(&pages, freeme, cinfo, i); pnfs_put_lseg(freeme); - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); } - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); } static unsigned int @@ -238,11 +238,11 @@ void pnfs_fetch_commit_bucket_list(struct list_head *pages, struct pnfs_commit_bucket *bucket; bucket = &cinfo->ds->buckets[data->ds_commit_index]; - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); list_splice_init(&bucket->committing, pages); data->lseg = bucket->clseg; bucket->clseg = NULL; - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); } @@ -874,12 +874,12 @@ pnfs_layout_mark_request_commit(struct nfs_page *req, struct list_head *list; struct pnfs_commit_bucket *buckets; - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); buckets = cinfo->ds->buckets; list = &buckets[ds_commit_idx].written; if (list_empty(list)) { if (!pnfs_is_valid_lseg(lseg)) { - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); cinfo->completion_ops->resched_write(cinfo, req); return; } @@ -896,7 +896,7 @@ pnfs_layout_mark_request_commit(struct nfs_page *req, cinfo->ds->nwritten++; nfs_request_add_commit_list_locked(req, list, cinfo); - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); nfs_mark_page_unstable(req->wb_page, cinfo); } EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 5f4fd53e5764..9283a96b9fb8 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -804,7 +804,7 @@ nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi, * number of outstanding requests requiring a commit as well as * the MM page stats. * - * The caller must hold the cinfo->lock, and the nfs_page lock. + * The caller must hold cinfo->inode->i_lock, and the nfs_page lock. */ void nfs_request_add_commit_list_locked(struct nfs_page *req, struct list_head *dst, @@ -832,9 +832,9 @@ EXPORT_SYMBOL_GPL(nfs_request_add_commit_list_locked); void nfs_request_add_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo) { - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); nfs_request_add_commit_list_locked(req, &cinfo->mds->list, cinfo); - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); nfs_mark_page_unstable(req->wb_page, cinfo); } EXPORT_SYMBOL_GPL(nfs_request_add_commit_list); @@ -864,7 +864,7 @@ EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list); static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo, struct inode *inode) { - cinfo->lock = &inode->i_lock; + cinfo->inode = inode; cinfo->mds = &NFS_I(inode)->commit_info; cinfo->ds = pnfs_get_ds_info(inode); cinfo->dreq = NULL; @@ -967,7 +967,7 @@ nfs_reqs_to_commit(struct nfs_commit_info *cinfo) return cinfo->mds->ncommit; } -/* cinfo->lock held by caller */ +/* cinfo->inode->i_lock held by caller */ int nfs_scan_commit_list(struct list_head *src, struct list_head *dst, struct nfs_commit_info *cinfo, int max) @@ -979,7 +979,7 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst, if (!nfs_lock_request(req)) continue; kref_get(&req->wb_kref); - if (cond_resched_lock(cinfo->lock)) + if (cond_resched_lock(&cinfo->inode->i_lock)) list_safe_reset_next(req, tmp, wb_list); nfs_request_remove_commit_list(req, cinfo); nfs_list_add_request(req, dst); @@ -1005,7 +1005,7 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, { int ret = 0; - spin_lock(cinfo->lock); + spin_lock(&cinfo->inode->i_lock); if (cinfo->mds->ncommit > 0) { const int max = INT_MAX; @@ -1013,7 +1013,7 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, cinfo, max); ret += pnfs_scan_commit_lists(inode, cinfo, max - ret); } - spin_unlock(cinfo->lock); + spin_unlock(&cinfo->inode->i_lock); return ret; } diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index d320906cf13e..cb9982d8f38f 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1431,7 +1431,7 @@ struct nfs_commit_completion_ops { }; struct nfs_commit_info { - spinlock_t *lock; /* inode->i_lock */ + struct inode *inode; /* Needed for inode->i_lock */ struct nfs_mds_commit_info *mds; struct pnfs_ds_commit_info *ds; struct nfs_direct_req *dreq; /* O_DIRECT request */ -- cgit v1.2.3-59-g8ed1b From 547a637630c61b9e1dae9abce2b44ce7076244af Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 21 Apr 2016 20:51:57 -0400 Subject: nfs: don't call nfs4_ff_layout_prepare_ds from ff_layout_get_ds_cred All the callers already call that function before calling into here, so it ends up being a no-op anyway. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index add0e5a70bd6..a0dbf94d15ae 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -433,16 +433,12 @@ ff_layout_get_ds_cred(struct pnfs_layout_segment *lseg, u32 ds_idx, struct rpc_cred *mdscred) { struct nfs4_ff_layout_mirror *mirror = FF_LAYOUT_COMP(lseg, ds_idx); - struct rpc_cred *cred = ERR_PTR(-EINVAL); - - if (!nfs4_ff_layout_prepare_ds(lseg, ds_idx, true)) - goto out; + struct rpc_cred *cred; if (mirror && mirror->cred) cred = mirror->cred; else cred = mdscred; -out: return cred; } -- cgit v1.2.3-59-g8ed1b From 57f3f4c0cd50e90aa92eec20d9c309dd67c594a5 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 21 Apr 2016 20:51:58 -0400 Subject: nfs: have ff_layout_get_ds_cred take a reference to the cred In later patches, we're going to want to allow the creds to be updated when we get a new layout with updated creds. Have this function take a reference to the cred that is later put once the call has been dispatched. Also, prepare for this change by ensuring we follow RCU rules when getting a reference to the cred as well. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 15 +++++++++------ fs/nfs/flexfilelayout/flexfilelayoutdev.c | 30 ++++++++++++++++++++++++++---- 2 files changed, 35 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 3b398f7b4637..1f46043c181e 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1712,7 +1712,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr) goto out_failed; ds_cred = ff_layout_get_ds_cred(lseg, idx, hdr->cred); - if (IS_ERR(ds_cred)) + if (!ds_cred) goto out_failed; vers = nfs4_ff_layout_ds_version(lseg, idx); @@ -1737,7 +1737,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr) vers == 3 ? &ff_layout_read_call_ops_v3 : &ff_layout_read_call_ops_v4, 0, RPC_TASK_SOFTCONN); - + put_rpccred(ds_cred); return PNFS_ATTEMPTED; out_failed: @@ -1769,7 +1769,7 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync) return PNFS_NOT_ATTEMPTED; ds_cred = ff_layout_get_ds_cred(lseg, idx, hdr->cred); - if (IS_ERR(ds_cred)) + if (!ds_cred) return PNFS_NOT_ATTEMPTED; vers = nfs4_ff_layout_ds_version(lseg, idx); @@ -1798,6 +1798,7 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync) vers == 3 ? &ff_layout_write_call_ops_v3 : &ff_layout_write_call_ops_v4, sync, RPC_TASK_SOFTCONN); + put_rpccred(ds_cred); return PNFS_ATTEMPTED; } @@ -1824,7 +1825,7 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how) struct rpc_clnt *ds_clnt; struct rpc_cred *ds_cred; u32 idx; - int vers; + int vers, ret; struct nfs_fh *fh; idx = calc_ds_index_from_commit(lseg, data->ds_commit_index); @@ -1838,7 +1839,7 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how) goto out_err; ds_cred = ff_layout_get_ds_cred(lseg, idx, data->cred); - if (IS_ERR(ds_cred)) + if (!ds_cred) goto out_err; vers = nfs4_ff_layout_ds_version(lseg, idx); @@ -1854,10 +1855,12 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how) if (fh) data->args.fh = fh; - return nfs_initiate_commit(ds_clnt, data, ds->ds_clp->rpc_ops, + ret = nfs_initiate_commit(ds_clnt, data, ds->ds_clp->rpc_ops, vers == 3 ? &ff_layout_commit_call_ops_v3 : &ff_layout_commit_call_ops_v4, how, RPC_TASK_SOFTCONN); + put_rpccred(ds_cred); + return ret; out_err: pnfs_generic_prepare_to_resend_writes(data); pnfs_generic_commit_release(data); diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index a0dbf94d15ae..baee22929174 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -338,6 +338,25 @@ static int ff_layout_update_mirror_cred(struct nfs4_ff_layout_mirror *mirror, return 0; } +static struct rpc_cred * +ff_layout_get_mirror_cred(struct nfs4_ff_layout_mirror *mirror, u32 iomode) +{ + struct rpc_cred *cred, **pcred; + + pcred = &mirror->cred; + + rcu_read_lock(); + do { + cred = rcu_dereference(*pcred); + if (!cred) + break; + + cred = get_rpccred_rcu(cred); + } while(!cred); + rcu_read_unlock(); + return cred; +} + struct nfs_fh * nfs4_ff_layout_select_ds_fh(struct pnfs_layout_segment *lseg, u32 mirror_idx) { @@ -435,10 +454,13 @@ ff_layout_get_ds_cred(struct pnfs_layout_segment *lseg, u32 ds_idx, struct nfs4_ff_layout_mirror *mirror = FF_LAYOUT_COMP(lseg, ds_idx); struct rpc_cred *cred; - if (mirror && mirror->cred) - cred = mirror->cred; - else - cred = mdscred; + if (mirror) { + cred = ff_layout_get_mirror_cred(mirror, lseg->pls_range.iomode); + if (!cred) + cred = get_rpccred(mdscred); + } else { + cred = get_rpccred(mdscred); + } return cred; } -- cgit v1.2.3-59-g8ed1b From 90a0be00e9d29d9baac27c0c63a5e71ca1334001 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 21 Apr 2016 20:51:59 -0400 Subject: nfs: get a reference to the credential in ff_layout_alloc_lseg We're just as likely to have allocation problems here as we would if we delay looking up the credential like we currently do. Fix the code to get a rpc_cred reference early, as soon as the mirror is set up. This allows us to eliminate the mirror early if there is a problem getting an rpc credential. This also allows us to drop the uid/gid from the layout_mirror struct as well. In the event that we find an existing mirror where this one would go, we swap in the new creds unconditionally, and drop the reference to the old one. Note that the old ff_layout_update_mirror_cred function wouldn't set this pointer unless the DS version was 3, but we don't know what the DS version is at this point. I'm a little unclear on why it did that as you still need creds to talk to v4 servers as well. I have the code set it regardless of the DS version here. Also note the change to using generic creds instead of calling lookup_cred directly. With that change, we also need to populate the group_info pointer in the acred as some functions expect that to never be NULL. Instead of allocating one every time however, we can allocate one when the module is loaded and share it since the group_info is refcounted. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 41 ++++++++++++++++++++++----- fs/nfs/flexfilelayout/flexfilelayout.h | 2 -- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 47 ++----------------------------- 3 files changed, 36 insertions(+), 54 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 1f46043c181e..ee87c0f9fd55 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -26,6 +26,8 @@ #define FF_LAYOUT_POLL_RETRY_MAX (15*HZ) +static struct group_info *ff_zero_group; + static struct pnfs_layout_hdr * ff_layout_alloc_layout_hdr(struct inode *inode, gfp_t gfp_flags) { @@ -407,8 +409,9 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, struct nfs4_ff_layout_mirror *mirror; struct nfs4_deviceid devid; struct nfs4_deviceid_node *idnode; - u32 ds_count; - u32 fh_count; + struct auth_cred acred = { .group_info = ff_zero_group }; + struct rpc_cred *cred; + u32 ds_count, fh_count, id; int j; rc = -EIO; @@ -484,24 +487,39 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, fls->mirror_array[i]->fh_versions_cnt = fh_count; /* user */ - rc = decode_name(&stream, &fls->mirror_array[i]->uid); + rc = decode_name(&stream, &id); if (rc) goto out_err_free; + acred.uid = make_kuid(&init_user_ns, id); + /* group */ - rc = decode_name(&stream, &fls->mirror_array[i]->gid); + rc = decode_name(&stream, &id); if (rc) goto out_err_free; + acred.gid = make_kgid(&init_user_ns, id); + + /* find the cred for it */ + cred = rpc_lookup_generic_cred(&acred, 0, gfp_flags); + if (IS_ERR(cred)) { + rc = PTR_ERR(cred); + goto out_err_free; + } + + rcu_assign_pointer(fls->mirror_array[i]->cred, cred); + mirror = ff_layout_add_mirror(lh, fls->mirror_array[i]); if (mirror != fls->mirror_array[i]) { + /* swap cred ptrs so free_mirror will clean up old */ + fls->mirror_array[i]->cred = xchg(&mirror->cred, cred); ff_layout_free_mirror(fls->mirror_array[i]); fls->mirror_array[i] = mirror; } - dprintk("%s: uid %d gid %d\n", __func__, - fls->mirror_array[i]->uid, - fls->mirror_array[i]->gid); + dprintk("%s: uid %u gid %u\n", __func__, + from_kuid(&init_user_ns, acred.uid), + from_kgid(&init_user_ns, acred.gid)); } p = xdr_inline_decode(&stream, 4); @@ -2226,6 +2244,11 @@ static int __init nfs4flexfilelayout_init(void) { printk(KERN_INFO "%s: NFSv4 Flexfile Layout Driver Registering...\n", __func__); + if (!ff_zero_group) { + ff_zero_group = groups_alloc(0); + if (!ff_zero_group) + return -ENOMEM; + } return pnfs_register_layoutdriver(&flexfilelayout_type); } @@ -2234,6 +2257,10 @@ static void __exit nfs4flexfilelayout_exit(void) printk(KERN_INFO "%s: NFSv4 Flexfile Layout Driver Unregistering...\n", __func__); pnfs_unregister_layoutdriver(&flexfilelayout_type); + if (ff_zero_group) { + put_group_info(ff_zero_group); + ff_zero_group = NULL; + } } MODULE_ALIAS("nfs-layouttype4-4"); diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index dd353bb7dc0a..c29fc853ce74 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -76,8 +76,6 @@ struct nfs4_ff_layout_mirror { u32 fh_versions_cnt; struct nfs_fh *fh_versions; nfs4_stateid stateid; - u32 uid; - u32 gid; struct rpc_cred *cred; atomic_t ref; spinlock_t lock; diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index baee22929174..6ddd8a5c5ae0 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -302,42 +302,6 @@ int ff_layout_track_ds_error(struct nfs4_flexfile_layout *flo, return 0; } -/* currently we only support AUTH_NONE and AUTH_SYS */ -static rpc_authflavor_t -nfs4_ff_layout_choose_authflavor(struct nfs4_ff_layout_mirror *mirror) -{ - if (mirror->uid == (u32)-1) - return RPC_AUTH_NULL; - return RPC_AUTH_UNIX; -} - -/* fetch cred for NFSv3 DS */ -static int ff_layout_update_mirror_cred(struct nfs4_ff_layout_mirror *mirror, - struct nfs4_pnfs_ds *ds) -{ - if (ds->ds_clp && !mirror->cred && - mirror->mirror_ds->ds_versions[0].version == 3) { - struct rpc_auth *auth = ds->ds_clp->cl_rpcclient->cl_auth; - struct rpc_cred *cred; - struct auth_cred acred = { - .uid = make_kuid(&init_user_ns, mirror->uid), - .gid = make_kgid(&init_user_ns, mirror->gid), - }; - - /* AUTH_NULL ignores acred */ - cred = auth->au_ops->lookup_cred(auth, &acred, 0); - if (IS_ERR(cred)) { - dprintk("%s: lookup_cred failed with %ld\n", - __func__, PTR_ERR(cred)); - return PTR_ERR(cred); - } else { - if (cmpxchg(&mirror->cred, NULL, cred)) - put_rpccred(cred); - } - } - return 0; -} - static struct rpc_cred * ff_layout_get_mirror_cred(struct nfs4_ff_layout_mirror *mirror, u32 iomode) { @@ -386,7 +350,6 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, struct inode *ino = lseg->pls_layout->plh_inode; struct nfs_server *s = NFS_SERVER(ino); unsigned int max_payload; - rpc_authflavor_t flavor; if (!ff_layout_mirror_valid(lseg, mirror)) { pr_err_ratelimited("NFS: %s: No data server for offset index %d\n", @@ -402,9 +365,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, /* matching smp_wmb() in _nfs4_pnfs_v3/4_ds_connect */ smp_rmb(); if (ds->ds_clp) - goto out_update_creds; - - flavor = nfs4_ff_layout_choose_authflavor(mirror); + goto out; /* FIXME: For now we assume the server sent only one version of NFS * to use for the DS. @@ -413,7 +374,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, dataserver_retrans, mirror->mirror_ds->ds_versions[0].version, mirror->mirror_ds->ds_versions[0].minor_version, - flavor); + RPC_AUTH_UNIX); /* connect success, check rsize/wsize limit */ if (ds->ds_clp) { @@ -438,11 +399,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, } else pnfs_error_mark_layout_for_return(ino, lseg); ds = NULL; - goto out; } -out_update_creds: - if (ff_layout_update_mirror_cred(mirror, ds)) - ds = NULL; out: return ds; } -- cgit v1.2.3-59-g8ed1b From 3064b6861d00b7124558cdf79f9387f948361be3 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 21 Apr 2016 20:52:00 -0400 Subject: nfs: have flexfiles mirror keep creds for both ro and rw layouts A mirror can be shared between multiple layouts, even with different iomodes. That makes stats gathering simpler, but it causes a problem when we get different creds in READ vs. RW layouts. The current code drops the newer credentials onto the floor when this occurs. That's problematic when you fetch a READ layout first, and then a RW. If the READ layout doesn't have the correct creds to do a write, then writes will fail. We could just overwrite the READ credentials with the RW ones, but that would break the ability for the server to fence the layout for reads if things go awry. We need to be able to revert to the earlier READ creds if the RW layout is returned afterward. The simplest fix is to just keep two sets of creds per mirror. One for READ layouts and one for RW, and then use the appropriate set depending on the iomode of the layout segment. Also fix up some RCU nits that sparse found. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 30 +++++++++++++++++++++++------- fs/nfs/flexfilelayout/flexfilelayout.h | 3 ++- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 7 +++++-- 3 files changed, 30 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index ee87c0f9fd55..60d690dbc947 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -213,10 +213,16 @@ static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags) static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror) { + struct rpc_cred *cred; + ff_layout_remove_mirror(mirror); kfree(mirror->fh_versions); - if (mirror->cred) - put_rpccred(mirror->cred); + cred = rcu_access_pointer(mirror->ro_cred); + if (cred) + put_rpccred(cred); + cred = rcu_access_pointer(mirror->rw_cred); + if (cred) + put_rpccred(cred); nfs4_ff_layout_put_deviceid(mirror->mirror_ds); kfree(mirror); } @@ -410,7 +416,7 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, struct nfs4_deviceid devid; struct nfs4_deviceid_node *idnode; struct auth_cred acred = { .group_info = ff_zero_group }; - struct rpc_cred *cred; + struct rpc_cred __rcu *cred; u32 ds_count, fh_count, id; int j; @@ -501,23 +507,33 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, acred.gid = make_kgid(&init_user_ns, id); /* find the cred for it */ - cred = rpc_lookup_generic_cred(&acred, 0, gfp_flags); + rcu_assign_pointer(cred, rpc_lookup_generic_cred(&acred, 0, gfp_flags)); if (IS_ERR(cred)) { rc = PTR_ERR(cred); goto out_err_free; } - rcu_assign_pointer(fls->mirror_array[i]->cred, cred); + if (lgr->range.iomode == IOMODE_READ) + rcu_assign_pointer(fls->mirror_array[i]->ro_cred, cred); + else + rcu_assign_pointer(fls->mirror_array[i]->rw_cred, cred); mirror = ff_layout_add_mirror(lh, fls->mirror_array[i]); if (mirror != fls->mirror_array[i]) { /* swap cred ptrs so free_mirror will clean up old */ - fls->mirror_array[i]->cred = xchg(&mirror->cred, cred); + if (lgr->range.iomode == IOMODE_READ) { + cred = xchg(&mirror->ro_cred, cred); + rcu_assign_pointer(fls->mirror_array[i]->ro_cred, cred); + } else { + cred = xchg(&mirror->rw_cred, cred); + rcu_assign_pointer(fls->mirror_array[i]->rw_cred, cred); + } ff_layout_free_mirror(fls->mirror_array[i]); fls->mirror_array[i] = mirror; } - dprintk("%s: uid %u gid %u\n", __func__, + dprintk("%s: iomode %s uid %u gid %u\n", __func__, + lgr->range.iomode == IOMODE_READ ? "READ" : "RW", from_kuid(&init_user_ns, acred.uid), from_kgid(&init_user_ns, acred.gid)); } diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index c29fc853ce74..1318c77aeb35 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -76,7 +76,8 @@ struct nfs4_ff_layout_mirror { u32 fh_versions_cnt; struct nfs_fh *fh_versions; nfs4_stateid stateid; - struct rpc_cred *cred; + struct rpc_cred __rcu *ro_cred; + struct rpc_cred __rcu *rw_cred; atomic_t ref; spinlock_t lock; struct nfs4_ff_layoutstat read_stat; diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index 6ddd8a5c5ae0..56296f3df19c 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -305,9 +305,12 @@ int ff_layout_track_ds_error(struct nfs4_flexfile_layout *flo, static struct rpc_cred * ff_layout_get_mirror_cred(struct nfs4_ff_layout_mirror *mirror, u32 iomode) { - struct rpc_cred *cred, **pcred; + struct rpc_cred *cred, __rcu **pcred; - pcred = &mirror->cred; + if (iomode == IOMODE_READ) + pcred = &mirror->ro_cred; + else + pcred = &mirror->rw_cred; rcu_read_lock(); do { -- cgit v1.2.3-59-g8ed1b From 11476e9dec39d90fe1e9bf12abc6f3efe35a073d Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 11 Apr 2016 16:20:22 -0400 Subject: NFS: Fix an LOCK/OPEN race when unlinking an open file At Connectathon 2016, we found that recent upstream Linux clients would occasionally send a LOCK operation with a zero stateid. This appeared to happen in close proximity to another thread returning a delegation before unlinking the same file while it remained open. Earlier, the client received a write delegation on this file and returned the open stateid. Now, as it is getting ready to unlink the file, it returns the write delegation. But there is still an open file descriptor on that file, so the client must OPEN the file again before it returns the delegation. Since commit 24311f884189 ('NFSv4: Recovery of recalled read delegations is broken'), nfs_open_delegation_recall() clears the NFS_DELEGATED_STATE flag _before_ it sends the OPEN. This allows a racing LOCK on the same inode to be put on the wire before the OPEN operation has returned a valid open stateid. To eliminate this race, serialize delegation return with the acquisition of a file lock on the same file. Adopt the same approach as is used in the unlock path. This patch also eliminates a similar race seen when sending a LOCK operation at the same time as returning a delegation on the same file. Fixes: 24311f884189 ('NFSv4: Recovery of recalled read ... ') Signed-off-by: Chuck Lever [Anna: Add sentence about LOCK / delegation race] Signed-off-by: Anna Schumaker --- fs/nfs/nfs4proc.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 327b8c34d360..bc2676c95e1b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -6054,6 +6054,7 @@ static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *reques static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request) { struct nfs_inode *nfsi = NFS_I(state->inode); + struct nfs4_state_owner *sp = state->owner; unsigned char fl_flags = request->fl_flags; int status = -ENOLCK; @@ -6068,6 +6069,7 @@ static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock status = do_vfs_lock(state->inode, request); if (status < 0) goto out; + mutex_lock(&sp->so_delegreturn_mutex); down_read(&nfsi->rwsem); if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { /* Yes: cache locks! */ @@ -6075,9 +6077,11 @@ static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock request->fl_flags = fl_flags & ~FL_SLEEP; status = do_vfs_lock(state->inode, request); up_read(&nfsi->rwsem); + mutex_unlock(&sp->so_delegreturn_mutex); goto out; } up_read(&nfsi->rwsem); + mutex_unlock(&sp->so_delegreturn_mutex); status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW); out: request->fl_flags = fl_flags; -- cgit v1.2.3-59-g8ed1b From 7e3fcf61abde92a895533b1c6932ad17e073a49c Mon Sep 17 00:00:00 2001 From: "J. Bruce Fields" Date: Tue, 3 May 2016 15:16:02 -0400 Subject: nfs: don't share mounts between network namespaces There's no guarantee that an IP address in a different network namespace actually represents the same endpoint. Also, if we allow unprivileged nfs mounts some day then this might allow an unprivileged user in another network namespace to misdirect somebody else's nfs mounts. If sharing between containers is really what's wanted then that could still be arranged explicitly, for example with bind mounts. Reported-by: "Eric W. Biederman" Signed-off-by: J. Bruce Fields Signed-off-by: Anna Schumaker --- fs/nfs/super.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs') diff --git a/fs/nfs/super.c b/fs/nfs/super.c index f1268280244e..ff537a205188 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -2408,6 +2408,11 @@ static int nfs_compare_super_address(struct nfs_server *server1, struct nfs_server *server2) { struct sockaddr *sap1, *sap2; + struct rpc_xprt *xprt1 = server1->client->cl_xprt; + struct rpc_xprt *xprt2 = server2->client->cl_xprt; + + if (!net_eq(xprt1->xprt_net, xprt2->xprt_net)) + return 0; sap1 = (struct sockaddr *)&server1->nfs_client->cl_addr; sap2 = (struct sockaddr *)&server2->nfs_client->cl_addr; -- cgit v1.2.3-59-g8ed1b From c2985d001d2fb77357aeae675545893b61c50044 Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Tue, 10 May 2016 16:57:41 -0400 Subject: Fixing oops in callback path Commit 80f9642724af5 ("NFSv4.x: Enforce the ca_maxreponsesize_cached on the back channel") causes an oops when it receives a callback with cachethis=yes. [ 109.667378] BUG: unable to handle kernel NULL pointer dereference at 00000000000002c8 [ 109.669476] IP: [] nfs4_callback_compound+0x4f8/0x690 [nfsv4] [ 109.671216] PGD 0 [ 109.671736] Oops: 0000 [#1] SMP [ 109.705427] CPU: 1 PID: 3579 Comm: nfsv4.1-svc Not tainted 4.5.0-rc1+ #1 [ 109.706987] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/20/2014 [ 109.709468] task: ffff8800b4408000 ti: ffff88008448c000 task.ti: ffff88008448c000 [ 109.711207] RIP: 0010:[] [] nfs4_callback_compound+0x4f8/0x690 [nfsv4] [ 109.713521] RSP: 0018:ffff88008448fca0 EFLAGS: 00010286 [ 109.714762] RAX: ffff880081ee202c RBX: ffff8800b7b5b600 RCX: 0000000000000001 [ 109.716427] RDX: 0000000000000008 RSI: 0000000000000008 RDI: 0000000000000000 [ 109.718091] RBP: ffff88008448fda8 R08: 0000000000000000 R09: 000000000b000000 [ 109.719757] R10: ffff880137786000 R11: ffff8800b7b5b600 R12: 0000000001000000 [ 109.721415] R13: 0000000000000002 R14: 0000000053270000 R15: 000000000000000b [ 109.723061] FS: 0000000000000000(0000) GS:ffff880139640000(0000) knlGS:0000000000000000 [ 109.724931] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 109.726278] CR2: 00000000000002c8 CR3: 0000000034d50000 CR4: 00000000001406e0 [ 109.727972] Stack: [ 109.728465] ffff880081ee202c ffff880081ee201c 000000008448fcc0 ffff8800baccb800 [ 109.730349] ffff8800baccc800 ffffffffa08d0380 0000000000000000 0000000000000000 [ 109.732211] ffff8800b7b5b600 0000000000000001 ffffffff81d073c0 ffff880081ee3090 [ 109.734056] Call Trace: [ 109.734657] [] svc_process_common+0x5c4/0x6c0 [sunrpc] [ 109.736267] [] bc_svc_process+0x1fc/0x360 [sunrpc] [ 109.737775] [] nfs41_callback_svc+0x10c/0x1d0 [nfsv4] [ 109.739335] [] ? prepare_to_wait_event+0xf0/0xf0 [ 109.740799] [] ? nfs4_callback_svc+0x50/0x50 [nfsv4] [ 109.742349] [] kthread+0xd8/0xf0 [ 109.743495] [] ? kthread_park+0x60/0x60 [ 109.744776] [] ret_from_fork+0x3f/0x70 [ 109.746037] [] ? kthread_park+0x60/0x60 [ 109.747324] Code: cc 45 31 f6 48 8b 85 00 ff ff ff 44 89 30 48 8b 85 f8 fe ff ff 44 89 20 48 8b 9d 38 ff ff ff 48 8b bd 30 ff ff ff 48 85 db 74 4c <4c> 8b af c8 02 00 00 4d 8d a5 08 02 00 00 49 81 c5 98 02 00 00 [ 109.754361] RIP [] nfs4_callback_compound+0x4f8/0x690 [nfsv4] [ 109.756123] RSP [ 109.756951] CR2: 00000000000002c8 [ 109.757738] ---[ end trace 2b8555511ab5dfb4 ]--- [ 109.758819] Kernel panic - not syncing: Fatal exception [ 109.760126] Kernel Offset: disabled [ 118.938934] ---[ end Kernel panic - not syncing: Fatal exception It doesn't unlock the table nor does it set the cps->clp pointer which is later needed by nfs4_cb_free_slot(). Fixes: 80f9642724af5 ("NFSv4.x: Enforce the ca_maxresponsesize_cached ...") CC: stable@vger.kernel.org Signed-off-by: Olga Kornievskaia Signed-off-by: Anna Schumaker --- fs/nfs/callback_proc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 618ced381a14..7c9fbf504f07 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -500,8 +500,10 @@ __be32 nfs4_callback_sequence(struct cb_sequenceargs *args, cps->slot = slot; /* The ca_maxresponsesize_cached is 0 with no DRC */ - if (args->csa_cachethis != 0) - return htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE); + if (args->csa_cachethis != 0) { + status = htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE); + goto out_unlock; + } /* * Check for pending referring calls. If a match is found, a -- cgit v1.2.3-59-g8ed1b From 67911c8f18b595040f5c2df0550c69a7c9397891 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Tue, 19 Jan 2016 14:37:30 -0500 Subject: NFS: Add nfs_commit_file() Copy will use this to set up a commit request for a generic range. I don't want to allocate a new pagecache entry for the file, so I needed to change parts of the commit path to handle requests with a null wb_page. Signed-off-by: Anna Schumaker --- fs/nfs/internal.h | 1 + fs/nfs/pagelist.c | 6 ++++-- fs/nfs/write.c | 41 +++++++++++++++++++++++++++++++++++++---- 3 files changed, 42 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index f1d1d2c472e9..5154fa65a2f2 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -477,6 +477,7 @@ void nfs_mark_request_commit(struct nfs_page *req, u32 ds_commit_idx); int nfs_write_need_commit(struct nfs_pgio_header *); void nfs_writeback_update_inode(struct nfs_pgio_header *hdr); +int nfs_commit_file(struct file *file, struct nfs_write_verifier *verf); 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, diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 1f6db4231057..174dd4cf5747 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -341,8 +341,10 @@ nfs_create_request(struct nfs_open_context *ctx, struct page *page, * long write-back delay. This will be adjusted in * update_nfs_request below if the region is not locked. */ req->wb_page = page; - req->wb_index = page_file_index(page); - get_page(page); + if (page) { + req->wb_index = page_file_index(page); + get_page(page); + } req->wb_offset = offset; req->wb_pgbase = offset; req->wb_bytes = count; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 9283a96b9fb8..e18a97aab275 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -737,7 +737,7 @@ static void nfs_inode_remove_request(struct nfs_page *req) head = req->wb_head; spin_lock(&inode->i_lock); - if (likely(!PageSwapCache(head->wb_page))) { + if (likely(head->wb_page && !PageSwapCache(head->wb_page))) { set_page_private(head->wb_page, 0); ClearPagePrivate(head->wb_page); smp_mb__after_atomic(); @@ -759,7 +759,8 @@ static void nfs_inode_remove_request(struct nfs_page *req) static void nfs_mark_request_dirty(struct nfs_page *req) { - __set_page_dirty_nobuffers(req->wb_page); + if (req->wb_page) + __set_page_dirty_nobuffers(req->wb_page); } /* @@ -835,7 +836,8 @@ nfs_request_add_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo) spin_lock(&cinfo->inode->i_lock); nfs_request_add_commit_list_locked(req, &cinfo->mds->list, cinfo); spin_unlock(&cinfo->inode->i_lock); - nfs_mark_page_unstable(req->wb_page, cinfo); + if (req->wb_page) + nfs_mark_page_unstable(req->wb_page, cinfo); } EXPORT_SYMBOL_GPL(nfs_request_add_commit_list); @@ -1724,6 +1726,36 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how, return -ENOMEM; } +int nfs_commit_file(struct file *file, struct nfs_write_verifier *verf) +{ + struct inode *inode = file_inode(file); + struct nfs_open_context *open; + struct nfs_commit_info cinfo; + struct nfs_page *req; + int ret; + + open = get_nfs_open_context(nfs_file_open_context(file)); + req = nfs_create_request(open, NULL, NULL, 0, i_size_read(inode)); + if (!req) { + ret = -ENOMEM; + goto out_put; + } + + nfs_init_cinfo_from_inode(&cinfo, inode); + + memcpy(&req->wb_verf, verf, sizeof(struct nfs_write_verifier)); + nfs_request_add_commit_list(req, &cinfo); + ret = nfs_commit_inode(inode, FLUSH_SYNC); + if (ret > 0) + ret = 0; + + nfs_free_request(req); +out_put: + put_nfs_open_context(open); + return ret; +} +EXPORT_SYMBOL_GPL(nfs_commit_file); + /* * COMMIT call returned */ @@ -1748,7 +1780,8 @@ static void nfs_commit_release_pages(struct nfs_commit_data *data) while (!list_empty(&data->pages)) { req = nfs_list_entry(data->pages.next); nfs_list_remove_request(req); - nfs_clear_page_commit(req->wb_page); + if (req->wb_page) + nfs_clear_page_commit(req->wb_page); dprintk("NFS: commit (%s/%llu %d@%lld)", req->wb_context->dentry->d_sb->s_id, -- cgit v1.2.3-59-g8ed1b From 2e72448b07dc3ff1b7593e9bfff91db182262857 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Tue, 21 May 2013 16:53:03 -0400 Subject: NFS: Add COPY nfs operation This adds the copy_range file_ops function pointer used by the sys_copy_range() function call. This patch only implements sync copies, so if an async copy happens we decode the stateid and ignore it. Signed-off-by: Anna Schumaker --- fs/nfs/nfs42.h | 1 + fs/nfs/nfs42proc.c | 105 +++++++++++++++++++++++++++++++++ fs/nfs/nfs42xdr.c | 146 ++++++++++++++++++++++++++++++++++++++++++++++ fs/nfs/nfs4file.c | 23 ++++++++ fs/nfs/nfs4proc.c | 1 + fs/nfs/nfs4xdr.c | 1 + include/linux/nfs4.h | 1 + include/linux/nfs_fs_sb.h | 1 + include/linux/nfs_xdr.h | 26 +++++++++ 9 files changed, 305 insertions(+) (limited to 'fs') diff --git a/fs/nfs/nfs42.h b/fs/nfs/nfs42.h index b587ccd31083..b6cd15314bab 100644 --- a/fs/nfs/nfs42.h +++ b/fs/nfs/nfs42.h @@ -13,6 +13,7 @@ /* nfs4.2proc.c */ int nfs42_proc_allocate(struct file *, loff_t, loff_t); +ssize_t nfs42_proc_copy(struct file *, loff_t, struct file *, loff_t, size_t); int nfs42_proc_deallocate(struct file *, loff_t, loff_t); loff_t nfs42_proc_llseek(struct file *, loff_t, int); int nfs42_proc_layoutstats_generic(struct nfs_server *, diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index dff83460e5a6..579ee20e4120 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -126,6 +126,111 @@ int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len) return err; } +static ssize_t _nfs42_proc_copy(struct file *src, loff_t pos_src, + struct nfs_lock_context *src_lock, + struct file *dst, loff_t pos_dst, + struct nfs_lock_context *dst_lock, + size_t count) +{ + struct nfs42_copy_args args = { + .src_fh = NFS_FH(file_inode(src)), + .src_pos = pos_src, + .dst_fh = NFS_FH(file_inode(dst)), + .dst_pos = pos_dst, + .count = count, + }; + struct nfs42_copy_res res; + struct rpc_message msg = { + .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY], + .rpc_argp = &args, + .rpc_resp = &res, + }; + struct inode *dst_inode = file_inode(dst); + struct nfs_server *server = NFS_SERVER(dst_inode); + int status; + + status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context, + src_lock, FMODE_READ); + if (status) + return status; + + status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context, + dst_lock, FMODE_WRITE); + if (status) + return status; + + status = nfs4_call_sync(server->client, server, &msg, + &args.seq_args, &res.seq_res, 0); + if (status == -ENOTSUPP) + server->caps &= ~NFS_CAP_COPY; + if (status) + return status; + + if (res.write_res.verifier.committed != NFS_FILE_SYNC) { + status = nfs_commit_file(dst, &res.write_res.verifier.verifier); + if (status) + return status; + } + + truncate_pagecache_range(dst_inode, pos_dst, + pos_dst + res.write_res.count); + + return res.write_res.count; +} + +ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src, + struct file *dst, loff_t pos_dst, + size_t count) +{ + struct nfs_server *server = NFS_SERVER(file_inode(dst)); + struct nfs_lock_context *src_lock; + struct nfs_lock_context *dst_lock; + struct nfs4_exception src_exception = { }; + struct nfs4_exception dst_exception = { }; + ssize_t err, err2; + + if (!nfs_server_capable(file_inode(dst), NFS_CAP_COPY)) + return -EOPNOTSUPP; + + src_lock = nfs_get_lock_context(nfs_file_open_context(src)); + if (IS_ERR(src_lock)) + return PTR_ERR(src_lock); + + src_exception.inode = file_inode(src); + src_exception.state = src_lock->open_context->state; + + dst_lock = nfs_get_lock_context(nfs_file_open_context(dst)); + if (IS_ERR(dst_lock)) { + err = PTR_ERR(dst_lock); + goto out_put_src_lock; + } + + dst_exception.inode = file_inode(dst); + dst_exception.state = dst_lock->open_context->state; + + do { + mutex_lock(&file_inode(dst)->i_mutex); + err = _nfs42_proc_copy(src, pos_src, src_lock, + dst, pos_dst, dst_lock, count); + mutex_unlock(&file_inode(dst)->i_mutex); + + if (err == -ENOTSUPP) { + err = -EOPNOTSUPP; + break; + } + + err2 = nfs4_handle_exception(server, err, &src_exception); + err = nfs4_handle_exception(server, err, &dst_exception); + if (!err) + err = err2; + } while (src_exception.retry || dst_exception.retry); + + nfs_put_lock_context(dst_lock); +out_put_src_lock: + nfs_put_lock_context(src_lock); + return err; +} + static loff_t _nfs42_proc_llseek(struct file *filep, struct nfs_lock_context *lock, loff_t offset, int whence) { diff --git a/fs/nfs/nfs42xdr.c b/fs/nfs/nfs42xdr.c index 0ca482a51e53..6dc6f2aea0d6 100644 --- a/fs/nfs/nfs42xdr.c +++ b/fs/nfs/nfs42xdr.c @@ -9,9 +9,22 @@ #define encode_fallocate_maxsz (encode_stateid_maxsz + \ 2 /* offset */ + \ 2 /* length */) +#define NFS42_WRITE_RES_SIZE (1 /* wr_callback_id size */ +\ + XDR_QUADLEN(NFS4_STATEID_SIZE) + \ + 2 /* wr_count */ + \ + 1 /* wr_committed */ + \ + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) #define encode_allocate_maxsz (op_encode_hdr_maxsz + \ encode_fallocate_maxsz) #define decode_allocate_maxsz (op_decode_hdr_maxsz) +#define encode_copy_maxsz (op_encode_hdr_maxsz + \ + XDR_QUADLEN(NFS4_STATEID_SIZE) + \ + XDR_QUADLEN(NFS4_STATEID_SIZE) + \ + 2 + 2 + 2 + 1 + 1 + 1) +#define decode_copy_maxsz (op_decode_hdr_maxsz + \ + NFS42_WRITE_RES_SIZE + \ + 1 /* cr_consecutive */ + \ + 1 /* cr_synchronous */) #define encode_deallocate_maxsz (op_encode_hdr_maxsz + \ encode_fallocate_maxsz) #define decode_deallocate_maxsz (op_decode_hdr_maxsz) @@ -49,6 +62,16 @@ decode_putfh_maxsz + \ decode_allocate_maxsz + \ decode_getattr_maxsz) +#define NFS4_enc_copy_sz (compound_encode_hdr_maxsz + \ + encode_putfh_maxsz + \ + encode_savefh_maxsz + \ + encode_putfh_maxsz + \ + encode_copy_maxsz) +#define NFS4_dec_copy_sz (compound_decode_hdr_maxsz + \ + decode_putfh_maxsz + \ + decode_savefh_maxsz + \ + decode_putfh_maxsz + \ + decode_copy_maxsz) #define NFS4_enc_deallocate_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_deallocate_maxsz + \ @@ -102,6 +125,23 @@ static void encode_allocate(struct xdr_stream *xdr, encode_fallocate(xdr, args); } +static void encode_copy(struct xdr_stream *xdr, + struct nfs42_copy_args *args, + struct compound_hdr *hdr) +{ + encode_op_hdr(xdr, OP_COPY, decode_copy_maxsz, hdr); + encode_nfs4_stateid(xdr, &args->src_stateid); + encode_nfs4_stateid(xdr, &args->dst_stateid); + + encode_uint64(xdr, args->src_pos); + encode_uint64(xdr, args->dst_pos); + encode_uint64(xdr, args->count); + + encode_uint32(xdr, 1); /* consecutive = true */ + encode_uint32(xdr, 1); /* synchronous = true */ + encode_uint32(xdr, 0); /* src server list */ +} + static void encode_deallocate(struct xdr_stream *xdr, struct nfs42_falloc_args *args, struct compound_hdr *hdr) @@ -181,6 +221,26 @@ static void nfs4_xdr_enc_allocate(struct rpc_rqst *req, encode_nops(&hdr); } +/* + * Encode COPY request + */ +static void nfs4_xdr_enc_copy(struct rpc_rqst *req, + struct xdr_stream *xdr, + struct nfs42_copy_args *args) +{ + struct compound_hdr hdr = { + .minorversion = nfs4_xdr_minorversion(&args->seq_args), + }; + + encode_compound_hdr(xdr, req, &hdr); + encode_sequence(xdr, &args->seq_args, &hdr); + encode_putfh(xdr, args->src_fh, &hdr); + encode_savefh(xdr, &hdr); + encode_putfh(xdr, args->dst_fh, &hdr); + encode_copy(xdr, args, &hdr); + encode_nops(&hdr); +} + /* * Encode DEALLOCATE request */ @@ -266,6 +326,62 @@ static int decode_allocate(struct xdr_stream *xdr, struct nfs42_falloc_res *res) return decode_op_hdr(xdr, OP_ALLOCATE); } +static int decode_write_response(struct xdr_stream *xdr, + struct nfs42_write_res *res) +{ + __be32 *p; + int stateids; + + p = xdr_inline_decode(xdr, 4 + 8 + 4); + if (unlikely(!p)) + goto out_overflow; + + stateids = be32_to_cpup(p++); + p = xdr_decode_hyper(p, &res->count); + res->verifier.committed = be32_to_cpup(p); + return decode_verifier(xdr, &res->verifier.verifier); + +out_overflow: + print_overflow_msg(__func__, xdr); + return -EIO; +} + +static int decode_copy_requirements(struct xdr_stream *xdr, + struct nfs42_copy_res *res) { + __be32 *p; + + p = xdr_inline_decode(xdr, 4 + 4); + if (unlikely(!p)) + goto out_overflow; + + res->consecutive = be32_to_cpup(p++); + res->synchronous = be32_to_cpup(p++); + return 0; +out_overflow: + print_overflow_msg(__func__, xdr); + return -EIO; +} + +static int decode_copy(struct xdr_stream *xdr, struct nfs42_copy_res *res) +{ + int status; + + status = decode_op_hdr(xdr, OP_COPY); + if (status == NFS4ERR_OFFLOAD_NO_REQS) { + status = decode_copy_requirements(xdr, res); + if (status) + return status; + return NFS4ERR_OFFLOAD_NO_REQS; + } else if (status) + return status; + + status = decode_write_response(xdr, &res->write_res); + if (status) + return status; + + return decode_copy_requirements(xdr, res); +} + static int decode_deallocate(struct xdr_stream *xdr, struct nfs42_falloc_res *res) { return decode_op_hdr(xdr, OP_DEALLOCATE); @@ -330,6 +446,36 @@ out: return status; } +/* + * Decode COPY response + */ +static int nfs4_xdr_dec_copy(struct rpc_rqst *rqstp, + struct xdr_stream *xdr, + struct nfs42_copy_res *res) +{ + struct compound_hdr hdr; + int status; + + status = decode_compound_hdr(xdr, &hdr); + if (status) + goto out; + status = decode_sequence(xdr, &res->seq_res, rqstp); + if (status) + goto out; + status = decode_putfh(xdr); + if (status) + goto out; + status = decode_savefh(xdr); + if (status) + goto out; + status = decode_putfh(xdr); + if (status) + goto out; + status = decode_copy(xdr, res); +out: + return status; +} + /* * Decode DEALLOCATE request */ diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c index d0390516467c..014b0e41ace5 100644 --- a/fs/nfs/nfs4file.c +++ b/fs/nfs/nfs4file.c @@ -129,6 +129,28 @@ nfs4_file_flush(struct file *file, fl_owner_t id) } #ifdef CONFIG_NFS_V4_2 +static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in, + struct file *file_out, loff_t pos_out, + size_t count, unsigned int flags) +{ + struct inode *in_inode = file_inode(file_in); + struct inode *out_inode = file_inode(file_out); + int ret; + + if (in_inode == out_inode) + return -EINVAL; + + /* flush any pending writes */ + ret = nfs_sync_inode(in_inode); + if (ret) + return ret; + ret = nfs_sync_inode(out_inode); + if (ret) + return ret; + + return nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count); +} + static loff_t nfs4_file_llseek(struct file *filep, loff_t offset, int whence) { loff_t ret; @@ -243,6 +265,7 @@ const struct file_operations nfs4_file_operations = { .check_flags = nfs_check_flags, .setlease = simple_nosetlease, #ifdef CONFIG_NFS_V4_2 + .copy_file_range = nfs4_copy_file_range, .llseek = nfs4_file_llseek, .fallocate = nfs42_fallocate, .clone_file_range = nfs42_clone_file_range, diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index bc2676c95e1b..4e83385ea6a9 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -8797,6 +8797,7 @@ static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = { | NFS_CAP_STATEID_NFSV41 | NFS_CAP_ATOMIC_OPEN_V1 | NFS_CAP_ALLOCATE + | NFS_CAP_COPY | NFS_CAP_DEALLOCATE | NFS_CAP_SEEK | NFS_CAP_LAYOUTSTATS diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 88474a4fc669..d1c96fc62c51 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -7515,6 +7515,7 @@ struct rpc_procinfo nfs4_procedures[] = { PROC(DEALLOCATE, enc_deallocate, dec_deallocate), PROC(LAYOUTSTATS, enc_layoutstats, dec_layoutstats), PROC(CLONE, enc_clone, dec_clone), + PROC(COPY, enc_copy, dec_copy), #endif /* CONFIG_NFS_V4_2 */ }; diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index 011433478a14..722509482e1a 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -504,6 +504,7 @@ enum { NFSPROC4_CLNT_DEALLOCATE, NFSPROC4_CLNT_LAYOUTSTATS, NFSPROC4_CLNT_CLONE, + NFSPROC4_CLNT_COPY, }; /* nfs41 types */ diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 7fcc13c8cf1f..14a762d2734d 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -246,5 +246,6 @@ struct nfs_server { #define NFS_CAP_DEALLOCATE (1U << 21) #define NFS_CAP_LAYOUTSTATS (1U << 22) #define NFS_CAP_CLONE (1U << 23) +#define NFS_CAP_COPY (1U << 24) #endif diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index cb9982d8f38f..e70ed54dad94 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1343,6 +1343,32 @@ struct nfs42_falloc_res { const struct nfs_server *falloc_server; }; +struct nfs42_copy_args { + struct nfs4_sequence_args seq_args; + + struct nfs_fh *src_fh; + nfs4_stateid src_stateid; + u64 src_pos; + + struct nfs_fh *dst_fh; + nfs4_stateid dst_stateid; + u64 dst_pos; + + u64 count; +}; + +struct nfs42_write_res { + u64 count; + struct nfs_writeverf verifier; +}; + +struct nfs42_copy_res { + struct nfs4_sequence_res seq_res; + struct nfs42_write_res write_res; + bool consecutive; + bool synchronous; +}; + struct nfs42_seek_args { struct nfs4_sequence_args seq_args; -- cgit v1.2.3-59-g8ed1b From a1d1c4f11adf8a0036cd40a643543047fc952eb6 Mon Sep 17 00:00:00 2001 From: Tigran Mkrtchyan Date: Thu, 12 May 2016 11:16:38 +0200 Subject: nfs4: client: do not send empty SETATTR after OPEN_CREATE OPEN_CREATE with EXCLUSIVE4_1 sends initial file permission. Ignoring fact, that server have indicated that file mod is set, client will send yet another SETATTR request, but, as mode is already set, new SETATTR will be empty. This is not a problem, nevertheless an extra roundtrip and slow open on high latency networks. This change is aims to skip extra setattr after open if there are no attributes to be set. Signed-off-by: Tigran Mkrtchyan Signed-off-by: Anna Schumaker --- fs/nfs/nfs4proc.c | 34 +++++++++++++++++++++++++--------- 1 file changed, 25 insertions(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 4e83385ea6a9..9795725d708c 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -74,6 +74,17 @@ #define NFS4_POLL_RETRY_MIN (HZ/10) #define NFS4_POLL_RETRY_MAX (15*HZ) +/* file attributes which can be mapped to nfs attributes */ +#define NFS4_VALID_ATTRS (ATTR_MODE \ + | ATTR_UID \ + | ATTR_GID \ + | ATTR_SIZE \ + | ATTR_ATIME \ + | ATTR_MTIME \ + | ATTR_CTIME \ + | ATTR_ATIME_SET \ + | ATTR_MTIME_SET) + struct nfs4_opendata; static int _nfs4_proc_open(struct nfs4_opendata *data); static int _nfs4_recover_proc_open(struct nfs4_opendata *data); @@ -2558,15 +2569,20 @@ static int _nfs4_do_open(struct inode *dir, if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) && (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) { nfs4_exclusive_attrset(opendata, sattr, &label); - - nfs_fattr_init(opendata->o_res.f_attr); - status = nfs4_do_setattr(state->inode, cred, - opendata->o_res.f_attr, sattr, - state, label, olabel); - if (status == 0) { - nfs_setattr_update_inode(state->inode, sattr, - opendata->o_res.f_attr); - nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel); + /* + * send create attributes which was not set by open + * with an extra setattr. + */ + if (sattr->ia_valid & NFS4_VALID_ATTRS) { + nfs_fattr_init(opendata->o_res.f_attr); + status = nfs4_do_setattr(state->inode, cred, + opendata->o_res.f_attr, sattr, + state, label, olabel); + if (status == 0) { + nfs_setattr_update_inode(state->inode, sattr, + opendata->o_res.f_attr); + nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel); + } } } if (opened && opendata->file_created) -- cgit v1.2.3-59-g8ed1b From 181342c5ebe8cc31f75b80ace18ae8a89a0c145a Mon Sep 17 00:00:00 2001 From: Shirley Ma Date: Mon, 2 May 2016 14:40:23 -0400 Subject: xprtrdma: Add rdma6 option to support NFS/RDMA IPv6 RFC 5666: The "rdma" netid is to be used when IPv4 addressing is employed by the underlying transport, and "rdma6" for IPv6 addressing. Add mount -o proto=rdma6 option to support NFS/RDMA IPv6 addressing. Changes from v2: - Integrated comments from Chuck Level, Anna Schumaker, Trodt Myklebust - Add a little more to the patch description to describe NFS/RDMA IPv6 suggested by Chuck Level and Anna Schumaker - Removed duplicated rdma6 define - Remove Opt_xprt_rdma mountfamily since it doesn't support Signed-off-by: Shirley Ma Reviewed-by: Chuck Lever Signed-off-by: Anna Schumaker --- fs/nfs/super.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs') diff --git a/fs/nfs/super.c b/fs/nfs/super.c index ff537a205188..2137e0202f25 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -191,6 +191,7 @@ static const match_table_t nfs_mount_option_tokens = { enum { Opt_xprt_udp, Opt_xprt_udp6, Opt_xprt_tcp, Opt_xprt_tcp6, Opt_xprt_rdma, + Opt_xprt_rdma6, Opt_xprt_err }; @@ -201,6 +202,7 @@ static const match_table_t nfs_xprt_protocol_tokens = { { Opt_xprt_tcp, "tcp" }, { Opt_xprt_tcp6, "tcp6" }, { Opt_xprt_rdma, "rdma" }, + { Opt_xprt_rdma6, "rdma6" }, { Opt_xprt_err, NULL } }; @@ -1456,6 +1458,8 @@ static int nfs_parse_mount_options(char *raw, mnt->flags |= NFS_MOUNT_TCP; mnt->nfs_server.protocol = XPRT_TRANSPORT_TCP; break; + case Opt_xprt_rdma6: + protofamily = AF_INET6; case Opt_xprt_rdma: /* vector side protocols to TCP */ mnt->flags |= NFS_MOUNT_TCP; -- cgit v1.2.3-59-g8ed1b From 6b26cc8c8ead3636a18bfd9489984983f4ddd6f4 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 2 May 2016 14:40:40 -0400 Subject: sunrpc: Advertise maximum backchannel payload size RPC-over-RDMA transports have a limit on how large a backward direction (backchannel) RPC message can be. Ensure that the NFSv4.x CREATE_SESSION operation advertises this limit to servers. Signed-off-by: Chuck Lever Tested-by: Steve Wise Signed-off-by: Anna Schumaker --- fs/nfs/nfs4proc.c | 10 ++++++---- include/linux/sunrpc/clnt.h | 1 + include/linux/sunrpc/xprt.h | 1 + net/sunrpc/clnt.c | 17 +++++++++++++++++ net/sunrpc/xprtrdma/backchannel.c | 16 ++++++++++++++++ net/sunrpc/xprtrdma/transport.c | 1 + net/sunrpc/xprtrdma/xprt_rdma.h | 1 + net/sunrpc/xprtsock.c | 6 ++++++ 8 files changed, 49 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 9795725d708c..196e41e12621 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -7371,9 +7371,11 @@ int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo) * always set csa_cachethis to FALSE because the current implementation * of the back channel DRC only supports caching the CB_SEQUENCE operation. */ -static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args) +static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args, + struct rpc_clnt *clnt) { unsigned int max_rqst_sz, max_resp_sz; + unsigned int max_bc_payload = rpc_max_bc_payload(clnt); max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead; max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead; @@ -7391,8 +7393,8 @@ static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args) args->fc_attrs.max_ops, args->fc_attrs.max_reqs); /* Back channel attributes */ - args->bc_attrs.max_rqst_sz = PAGE_SIZE; - args->bc_attrs.max_resp_sz = PAGE_SIZE; + args->bc_attrs.max_rqst_sz = max_bc_payload; + args->bc_attrs.max_resp_sz = max_bc_payload; args->bc_attrs.max_resp_sz_cached = 0; args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS; args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS; @@ -7496,7 +7498,7 @@ static int _nfs4_proc_create_session(struct nfs_client *clp, }; int status; - nfs4_init_channel_attrs(&args); + nfs4_init_channel_attrs(&args, clp->cl_rpcclient); args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN); status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT); diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 9a7ddbaf116e..19c659d1c0f8 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -176,6 +176,7 @@ void rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int); int rpc_protocol(struct rpc_clnt *); struct net * rpc_net_ns(struct rpc_clnt *); size_t rpc_max_payload(struct rpc_clnt *); +size_t rpc_max_bc_payload(struct rpc_clnt *); unsigned long rpc_get_timeout(struct rpc_clnt *clnt); void rpc_force_rebind(struct rpc_clnt *); size_t rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t); diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h index fb0d212e0d3a..5aa3834619a8 100644 --- a/include/linux/sunrpc/xprt.h +++ b/include/linux/sunrpc/xprt.h @@ -142,6 +142,7 @@ struct rpc_xprt_ops { int (*bc_setup)(struct rpc_xprt *xprt, unsigned int min_reqs); int (*bc_up)(struct svc_serv *serv, struct net *net); + size_t (*bc_maxpayload)(struct rpc_xprt *xprt); void (*bc_free_rqst)(struct rpc_rqst *rqst); void (*bc_destroy)(struct rpc_xprt *xprt, unsigned int max_reqs); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 7e0c9bf22df8..06b4df9faaa1 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -1413,6 +1413,23 @@ size_t rpc_max_payload(struct rpc_clnt *clnt) } EXPORT_SYMBOL_GPL(rpc_max_payload); +/** + * rpc_max_bc_payload - Get maximum backchannel payload size, in bytes + * @clnt: RPC client to query + */ +size_t rpc_max_bc_payload(struct rpc_clnt *clnt) +{ + struct rpc_xprt *xprt; + size_t ret; + + rcu_read_lock(); + xprt = rcu_dereference(clnt->cl_xprt); + ret = xprt->ops->bc_maxpayload(xprt); + rcu_read_unlock(); + return ret; +} +EXPORT_SYMBOL_GPL(rpc_max_bc_payload); + /** * rpc_get_timeout - Get timeout for transport in units of HZ * @clnt: RPC client to query diff --git a/net/sunrpc/xprtrdma/backchannel.c b/net/sunrpc/xprtrdma/backchannel.c index 2dcd7640eeb5..87762d976b63 100644 --- a/net/sunrpc/xprtrdma/backchannel.c +++ b/net/sunrpc/xprtrdma/backchannel.c @@ -191,6 +191,22 @@ int xprt_rdma_bc_up(struct svc_serv *serv, struct net *net) return 0; } +/** + * xprt_rdma_bc_maxpayload - Return maximum backchannel message size + * @xprt: transport + * + * Returns maximum size, in bytes, of a backchannel message + */ +size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *xprt) +{ + struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); + struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data; + size_t maxmsg; + + maxmsg = min_t(unsigned int, cdata->inline_rsize, cdata->inline_wsize); + return maxmsg - RPCRDMA_HDRLEN_MIN; +} + /** * rpcrdma_bc_marshal_reply - Send backwards direction reply * @rqst: buffer containing RPC reply data diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c index b1b009f10ea3..9954342924df 100644 --- a/net/sunrpc/xprtrdma/transport.c +++ b/net/sunrpc/xprtrdma/transport.c @@ -707,6 +707,7 @@ static struct rpc_xprt_ops xprt_rdma_procs = { #if defined(CONFIG_SUNRPC_BACKCHANNEL) .bc_setup = xprt_rdma_bc_setup, .bc_up = xprt_rdma_bc_up, + .bc_maxpayload = xprt_rdma_bc_maxpayload, .bc_free_rqst = xprt_rdma_bc_free_rqst, .bc_destroy = xprt_rdma_bc_destroy, #endif diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index 2ebc743cb96f..7723e5faff4d 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -534,6 +534,7 @@ void xprt_rdma_cleanup(void); #if defined(CONFIG_SUNRPC_BACKCHANNEL) int xprt_rdma_bc_setup(struct rpc_xprt *, unsigned int); int xprt_rdma_bc_up(struct svc_serv *, struct net *); +size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *); int rpcrdma_bc_post_recv(struct rpcrdma_xprt *, unsigned int); void rpcrdma_bc_receive_call(struct rpcrdma_xprt *, struct rpcrdma_rep *); int rpcrdma_bc_marshal_reply(struct rpc_rqst *); diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 65e759569e48..f1faf6b9aaff 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -1365,6 +1365,11 @@ static int xs_tcp_bc_up(struct svc_serv *serv, struct net *net) return ret; return 0; } + +static size_t xs_tcp_bc_maxpayload(struct rpc_xprt *xprt) +{ + return PAGE_SIZE; +} #else static inline int _xs_tcp_read_data(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) @@ -2660,6 +2665,7 @@ static struct rpc_xprt_ops xs_tcp_ops = { #ifdef CONFIG_SUNRPC_BACKCHANNEL .bc_setup = xprt_setup_bc, .bc_up = xs_tcp_bc_up, + .bc_maxpayload = xs_tcp_bc_maxpayload, .bc_free_rqst = xprt_free_bc_rqst, .bc_destroy = xprt_destroy_bc, #endif -- cgit v1.2.3-59-g8ed1b From f538d0ba5be39c52addff01c7ec31f0b90e57dfc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 16 May 2016 14:41:14 -0400 Subject: pNFS: Fix a leaked layoutstats flag Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 5b404d926e08..faea98281c75 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2407,7 +2407,7 @@ pnfs_report_layoutstat(struct inode *inode, gfp_t gfp_flags) spin_lock(&inode->i_lock); if (!NFS_I(inode)->layout) { spin_unlock(&inode->i_lock); - goto out; + goto out_clear_layoutstats; } hdr = NFS_I(inode)->layout; pnfs_get_layout_hdr(hdr); @@ -2436,6 +2436,7 @@ out_free: kfree(data); out_put: pnfs_put_layout_hdr(hdr); +out_clear_layoutstats: smp_mb__before_atomic(); clear_bit(NFS_INO_LAYOUTSTATS, &nfsi->flags); smp_mb__after_atomic(); -- cgit v1.2.3-59-g8ed1b From 93b717fd81bf6b9a73c3702e9b079b4de8148b34 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 16 May 2016 17:42:43 -0400 Subject: NFSv4: Label stateids with the type In order to more easily distinguish what kind of stateid we are dealing with, introduce a type that can be used to label the stateid structure. The label will be useful both for debugging, but also when dealing with operations like SETATTR, READ and WRITE that can take several different types of stateid as arguments. Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/callback_xdr.c | 17 ++++++++++--- fs/nfs/flexfilelayout/flexfilelayout.c | 7 +++--- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 3 ++- fs/nfs/nfs4_fs.h | 7 ++++-- fs/nfs/nfs4proc.c | 3 +++ fs/nfs/nfs4state.c | 5 +++- fs/nfs/nfs4xdr.c | 42 ++++++++++++++++++++++++------- include/linux/nfs4.h | 25 ++++++++++++++---- 8 files changed, 85 insertions(+), 24 deletions(-) (limited to 'fs') diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c index 976c90608e56..d81f96aacd51 100644 --- a/fs/nfs/callback_xdr.c +++ b/fs/nfs/callback_xdr.c @@ -146,10 +146,16 @@ static __be32 decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) p = read_buf(xdr, NFS4_STATEID_SIZE); if (unlikely(p == NULL)) return htonl(NFS4ERR_RESOURCE); - memcpy(stateid, p, NFS4_STATEID_SIZE); + memcpy(stateid->data, p, NFS4_STATEID_SIZE); return 0; } +static __be32 decode_delegation_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +{ + stateid->type = NFS4_DELEGATION_STATEID_TYPE; + return decode_stateid(xdr, stateid); +} + static __be32 decode_compound_hdr_arg(struct xdr_stream *xdr, struct cb_compound_hdr_arg *hdr) { __be32 *p; @@ -211,7 +217,7 @@ static __be32 decode_recall_args(struct svc_rqst *rqstp, struct xdr_stream *xdr, __be32 *p; __be32 status; - status = decode_stateid(xdr, &args->stateid); + status = decode_delegation_stateid(xdr, &args->stateid); if (unlikely(status != 0)) goto out; p = read_buf(xdr, 4); @@ -227,6 +233,11 @@ out: } #if defined(CONFIG_NFS_V4_1) +static __be32 decode_layout_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +{ + stateid->type = NFS4_LAYOUT_STATEID_TYPE; + return decode_stateid(xdr, stateid); +} static __be32 decode_layoutrecall_args(struct svc_rqst *rqstp, struct xdr_stream *xdr, @@ -263,7 +274,7 @@ static __be32 decode_layoutrecall_args(struct svc_rqst *rqstp, } p = xdr_decode_hyper(p, &args->cbl_range.offset); p = xdr_decode_hyper(p, &args->cbl_range.length); - status = decode_stateid(xdr, &args->cbl_stateid); + status = decode_layout_stateid(xdr, &args->cbl_stateid); if (unlikely(status != 0)) goto out; } else if (args->cbl_recall_type == RETURN_FSID) { diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 60d690dbc947..53b6391e0eba 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -55,14 +55,15 @@ ff_layout_free_layout_hdr(struct pnfs_layout_hdr *lo) kfree(FF_LAYOUT_FROM_HDR(lo)); } -static int decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +static int decode_pnfs_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) { __be32 *p; p = xdr_inline_decode(xdr, NFS4_STATEID_SIZE); if (unlikely(p == NULL)) return -ENOBUFS; - memcpy(stateid, p, NFS4_STATEID_SIZE); + stateid->type = NFS4_PNFS_DS_STATEID_TYPE; + memcpy(stateid->data, p, NFS4_STATEID_SIZE); dprintk("%s: stateid id= [%x%x%x%x]\n", __func__, p[0], p[1], p[2], p[3]); return 0; @@ -465,7 +466,7 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, fls->mirror_array[i]->efficiency = be32_to_cpup(p); /* stateid */ - rc = decode_stateid(&stream, &fls->mirror_array[i]->stateid); + rc = decode_pnfs_stateid(&stream, &fls->mirror_array[i]->stateid); if (rc) goto out_err_free; diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index 56296f3df19c..eeef89359ad2 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -228,7 +228,8 @@ ff_ds_error_match(const struct nfs4_ff_layout_ds_err *e1, return e1->opnum < e2->opnum ? -1 : 1; if (e1->status != e2->status) return e1->status < e2->status ? -1 : 1; - ret = memcmp(&e1->stateid, &e2->stateid, sizeof(e1->stateid)); + ret = memcmp(e1->stateid.data, e2->stateid.data, + sizeof(e1->stateid.data)); if (ret != 0) return ret; ret = memcmp(&e1->deviceid, &e2->deviceid, sizeof(e1->deviceid)); diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 4afdee420d25..b5d9f345c9f2 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -496,12 +496,15 @@ extern struct svc_version nfs4_callback_version4; static inline void nfs4_stateid_copy(nfs4_stateid *dst, const nfs4_stateid *src) { - memcpy(dst, src, sizeof(*dst)); + memcpy(dst->data, src->data, sizeof(dst->data)); + dst->type = src->type; } static inline bool nfs4_stateid_match(const nfs4_stateid *dst, const nfs4_stateid *src) { - return memcmp(dst, src, sizeof(*dst)) == 0; + if (dst->type != src->type) + return false; + return memcmp(dst->data, src->data, sizeof(dst->data)) == 0; } static inline bool nfs4_stateid_match_other(const nfs4_stateid *dst, const nfs4_stateid *src) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 196e41e12621..2516467ff17f 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -8675,6 +8675,9 @@ nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp) static bool nfs41_match_stateid(const nfs4_stateid *s1, const nfs4_stateid *s2) { + if (s1->type != s2->type) + return false; + if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0) return false; diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index d854693a15b0..d630f9cca0f1 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -65,7 +65,10 @@ #define OPENOWNER_POOL_SIZE 8 -const nfs4_stateid zero_stateid; +const nfs4_stateid zero_stateid = { + .data = { 0 }, + .type = NFS4_SPECIAL_STATEID_TYPE, +}; static DEFINE_MUTEX(nfs_clid_init_mutex); int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred) diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index d1c96fc62c51..661e753fe1c9 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -4270,6 +4270,24 @@ static int decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) return decode_opaque_fixed(xdr, stateid, NFS4_STATEID_SIZE); } +static int decode_open_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +{ + stateid->type = NFS4_OPEN_STATEID_TYPE; + return decode_stateid(xdr, stateid); +} + +static int decode_lock_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +{ + stateid->type = NFS4_LOCK_STATEID_TYPE; + return decode_stateid(xdr, stateid); +} + +static int decode_delegation_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +{ + stateid->type = NFS4_DELEGATION_STATEID_TYPE; + return decode_stateid(xdr, stateid); +} + static int decode_close(struct xdr_stream *xdr, struct nfs_closeres *res) { int status; @@ -4278,7 +4296,7 @@ static int decode_close(struct xdr_stream *xdr, struct nfs_closeres *res) if (status != -EIO) nfs_increment_open_seqid(status, res->seqid); if (!status) - status = decode_stateid(xdr, &res->stateid); + status = decode_open_stateid(xdr, &res->stateid); return status; } @@ -4937,7 +4955,7 @@ static int decode_lock(struct xdr_stream *xdr, struct nfs_lock_res *res) if (status == -EIO) goto out; if (status == 0) { - status = decode_stateid(xdr, &res->stateid); + status = decode_lock_stateid(xdr, &res->stateid); if (unlikely(status)) goto out; } else if (status == -NFS4ERR_DENIED) @@ -4966,7 +4984,7 @@ static int decode_locku(struct xdr_stream *xdr, struct nfs_locku_res *res) if (status != -EIO) nfs_increment_lock_seqid(status, res->seqid); if (status == 0) - status = decode_stateid(xdr, &res->stateid); + status = decode_lock_stateid(xdr, &res->stateid); return status; } @@ -5016,7 +5034,7 @@ static int decode_rw_delegation(struct xdr_stream *xdr, __be32 *p; int status; - status = decode_stateid(xdr, &res->delegation); + status = decode_delegation_stateid(xdr, &res->delegation); if (unlikely(status)) return status; p = xdr_inline_decode(xdr, 4); @@ -5096,7 +5114,7 @@ static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res) nfs_increment_open_seqid(status, res->seqid); if (status) return status; - status = decode_stateid(xdr, &res->stateid); + status = decode_open_stateid(xdr, &res->stateid); if (unlikely(status)) return status; @@ -5136,7 +5154,7 @@ static int decode_open_confirm(struct xdr_stream *xdr, struct nfs_open_confirmre if (status != -EIO) nfs_increment_open_seqid(status, res->seqid); if (!status) - status = decode_stateid(xdr, &res->stateid); + status = decode_open_stateid(xdr, &res->stateid); return status; } @@ -5148,7 +5166,7 @@ static int decode_open_downgrade(struct xdr_stream *xdr, struct nfs_closeres *re if (status != -EIO) nfs_increment_open_seqid(status, res->seqid); if (!status) - status = decode_stateid(xdr, &res->stateid); + status = decode_open_stateid(xdr, &res->stateid); return status; } @@ -5838,6 +5856,12 @@ out_overflow: } #if defined(CONFIG_NFS_V4_1) +static int decode_layout_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid) +{ + stateid->type = NFS4_LAYOUT_STATEID_TYPE; + return decode_stateid(xdr, stateid); +} + static int decode_getdeviceinfo(struct xdr_stream *xdr, struct nfs4_getdeviceinfo_res *res) { @@ -5919,7 +5943,7 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req, if (unlikely(!p)) goto out_overflow; res->return_on_close = be32_to_cpup(p); - decode_stateid(xdr, &res->stateid); + decode_layout_stateid(xdr, &res->stateid); p = xdr_inline_decode(xdr, 4); if (unlikely(!p)) goto out_overflow; @@ -5985,7 +6009,7 @@ static int decode_layoutreturn(struct xdr_stream *xdr, goto out_overflow; res->lrs_present = be32_to_cpup(p); if (res->lrs_present) - status = decode_stateid(xdr, &res->stateid); + status = decode_layout_stateid(xdr, &res->stateid); return status; out_overflow: print_overflow_msg(__func__, xdr); diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index 722509482e1a..e1692c96cbc8 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -50,12 +50,27 @@ struct nfs4_label { typedef struct { char data[NFS4_VERIFIER_SIZE]; } nfs4_verifier; -struct nfs_stateid4 { - __be32 seqid; - char other[NFS4_STATEID_OTHER_SIZE]; -} __attribute__ ((packed)); +struct nfs4_stateid_struct { + union { + char data[NFS4_STATEID_SIZE]; + struct { + __be32 seqid; + char other[NFS4_STATEID_OTHER_SIZE]; + } __attribute__ ((packed)); + }; + + enum { + NFS4_INVALID_STATEID_TYPE = 0, + NFS4_SPECIAL_STATEID_TYPE, + NFS4_OPEN_STATEID_TYPE, + NFS4_LOCK_STATEID_TYPE, + NFS4_DELEGATION_STATEID_TYPE, + NFS4_LAYOUT_STATEID_TYPE, + NFS4_PNFS_DS_STATEID_TYPE, + } type; +}; -typedef struct nfs_stateid4 nfs4_stateid; +typedef struct nfs4_stateid_struct nfs4_stateid; enum nfs_opnum4 { OP_ACCESS = 3, -- cgit v1.2.3-59-g8ed1b From abf4e13cc1e16ed83ec8363d44f76149034b2851 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 16 May 2016 17:42:44 -0400 Subject: NFSv4: Use the right stateid for delegations in setattr, read and write When we're using a delegation to represent our open state, we should ensure that we use the stateid that was used to create that delegation. Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/delegation.c | 9 ++++++--- fs/nfs/delegation.h | 2 +- fs/nfs/nfs4_fs.h | 5 +++-- fs/nfs/nfs4proc.c | 13 +++++++++---- fs/nfs/nfs4state.c | 13 +++++++++---- 5 files changed, 28 insertions(+), 14 deletions(-) (limited to 'fs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 5166adcfc0fb..322c2585bc34 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -875,15 +875,16 @@ int nfs_delegations_present(struct nfs_client *clp) /** * nfs4_copy_delegation_stateid - Copy inode's state ID information - * @dst: stateid data structure to fill in * @inode: inode to check * @flags: delegation type requirement + * @dst: stateid data structure to fill in + * @cred: optional argument to retrieve credential * * Returns "true" and fills in "dst->data" * if inode had a delegation, * otherwise "false" is returned. */ -bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode, - fmode_t flags) +bool nfs4_copy_delegation_stateid(struct inode *inode, fmode_t flags, + nfs4_stateid *dst, struct rpc_cred **cred) { struct nfs_inode *nfsi = NFS_I(inode); struct nfs_delegation *delegation; @@ -896,6 +897,8 @@ bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode, if (ret) { nfs4_stateid_copy(dst, &delegation->stateid); nfs_mark_delegation_referenced(delegation); + if (cred) + *cred = get_rpccred(delegation->cred); } rcu_read_unlock(); return ret; diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index 333063e032f0..64724d252a79 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -56,7 +56,7 @@ void nfs_delegation_reap_unclaimed(struct nfs_client *clp); int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync); int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid, fmode_t type); int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid); -bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode, fmode_t flags); +bool nfs4_copy_delegation_stateid(struct inode *inode, fmode_t flags, nfs4_stateid *dst, struct rpc_cred **cred); void nfs_mark_delegation_referenced(struct nfs_delegation *delegation); int nfs4_have_delegation(struct inode *inode, fmode_t flags); diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index b5d9f345c9f2..768456fa1b17 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -438,8 +438,9 @@ extern void nfs41_handle_server_scope(struct nfs_client *, struct nfs41_server_scope **); extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp); extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl); -extern int nfs4_select_rw_stateid(nfs4_stateid *, struct nfs4_state *, - fmode_t, const struct nfs_lockowner *); +extern int nfs4_select_rw_stateid(struct nfs4_state *, fmode_t, + const struct nfs_lockowner *, nfs4_stateid *, + struct rpc_cred **); extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask); extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 2516467ff17f..c9b66085d392 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2692,6 +2692,7 @@ static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred, .rpc_resp = &res, .rpc_cred = cred, }; + struct rpc_cred *delegation_cred = NULL; unsigned long timestamp = jiffies; fmode_t fmode; bool truncate; @@ -2707,7 +2708,7 @@ static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred, truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false; fmode = truncate ? FMODE_WRITE : FMODE_READ; - if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) { + if (nfs4_copy_delegation_stateid(inode, fmode, &arg.stateid, &delegation_cred)) { /* Use that stateid */ } else if (truncate && state != NULL) { struct nfs_lockowner lockowner = { @@ -2716,13 +2717,17 @@ static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred, }; if (!nfs4_valid_open_stateid(state)) return -EBADF; - if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE, - &lockowner) == -EIO) + if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner, + &arg.stateid, &delegation_cred) == -EIO) return -EBADF; } else nfs4_stateid_copy(&arg.stateid, &zero_stateid); + if (delegation_cred) + msg.rpc_cred = delegation_cred; status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1); + + put_rpccred(delegation_cred); if (status == 0 && state != NULL) renew_lease(server, timestamp); trace_nfs4_setattr(inode, &arg.stateid, status); @@ -4301,7 +4306,7 @@ int nfs4_set_rw_stateid(nfs4_stateid *stateid, if (l_ctx != NULL) lockowner = &l_ctx->lockowner; - return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner); + return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL); } EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid); diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index d630f9cca0f1..5075592df145 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -988,15 +988,20 @@ static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state) * Byte-range lock aware utility to initialize the stateid of read/write * requests. */ -int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state, - fmode_t fmode, const struct nfs_lockowner *lockowner) +int nfs4_select_rw_stateid(struct nfs4_state *state, + fmode_t fmode, const struct nfs_lockowner *lockowner, + nfs4_stateid *dst, struct rpc_cred **cred) { - int ret = nfs4_copy_lock_stateid(dst, state, lockowner); + int ret; + + if (cred != NULL) + *cred = NULL; + ret = nfs4_copy_lock_stateid(dst, state, lockowner); if (ret == -EIO) /* A lost lock - don't even consider delegations */ goto out; /* returns true if delegation stateid found and copied */ - if (nfs4_copy_delegation_stateid(dst, state->inode, fmode)) { + if (nfs4_copy_delegation_stateid(state->inode, fmode, dst, cred)) { ret = 0; goto out; } -- cgit v1.2.3-59-g8ed1b From cca588d6c835e21fd7a3e02a964ce5fe09fdad26 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 16 May 2016 17:42:45 -0400 Subject: NFS: Reclaim writes via writepage are opportunistic No need to make them a priority any more, or to make them succeed. Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/write.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index e18a97aab275..4dac51ba1f7e 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -245,8 +245,7 @@ static void nfs_mark_uptodate(struct nfs_page *req) static int wb_priority(struct writeback_control *wbc) { int ret = 0; - if (wbc->for_reclaim) - return FLUSH_HIGHPRI | FLUSH_COND_STABLE; + if (wbc->sync_mode == WB_SYNC_ALL) ret = FLUSH_COND_STABLE; return ret; -- cgit v1.2.3-59-g8ed1b From 75bf47ebf6b5349d875fdffe7d64b0fc4e936170 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 17 May 2016 12:28:36 -0400 Subject: pNFS/flexfile: Fix erroneous fall back to read/write through the MDS This patch fixes a problem whereby the pNFS client falls back to doing reads and writes through the metadata server even when the layout flag FF_FLAGS_NO_IO_THRU_MDS is set. Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 23 ++++++----------------- 1 file changed, 6 insertions(+), 17 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 53b6391e0eba..b6e0e32b6053 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1295,7 +1295,7 @@ ff_layout_set_layoutcommit(struct nfs_pgio_header *hdr) } static bool -ff_layout_reset_to_mds(struct pnfs_layout_segment *lseg, int idx) +ff_layout_device_unavailable(struct pnfs_layout_segment *lseg, int idx) { /* No mirroring for now */ struct nfs4_deviceid_node *node = FF_LAYOUT_DEVID_NODE(lseg, idx); @@ -1332,16 +1332,10 @@ static int ff_layout_read_prepare_common(struct rpc_task *task, rpc_exit(task, -EIO); return -EIO; } - if (ff_layout_reset_to_mds(hdr->lseg, hdr->pgio_mirror_idx)) { - dprintk("%s task %u reset io to MDS\n", __func__, task->tk_pid); - if (ff_layout_has_available_ds(hdr->lseg)) - pnfs_read_resend_pnfs(hdr); - else - ff_layout_reset_read(hdr); - rpc_exit(task, 0); + if (ff_layout_device_unavailable(hdr->lseg, hdr->pgio_mirror_idx)) { + rpc_exit(task, -EHOSTDOWN); return -EAGAIN; } - hdr->pgio_done_cb = ff_layout_read_done_cb; ff_layout_read_record_layoutstats_start(task, hdr); return 0; @@ -1531,14 +1525,8 @@ static int ff_layout_write_prepare_common(struct rpc_task *task, return -EIO; } - if (ff_layout_reset_to_mds(hdr->lseg, hdr->pgio_mirror_idx)) { - bool retry_pnfs; - - retry_pnfs = ff_layout_has_available_ds(hdr->lseg); - dprintk("%s task %u reset io to %s\n", __func__, - task->tk_pid, retry_pnfs ? "pNFS" : "MDS"); - ff_layout_reset_write(hdr, retry_pnfs); - rpc_exit(task, 0); + if (ff_layout_device_unavailable(hdr->lseg, hdr->pgio_mirror_idx)) { + rpc_exit(task, -EHOSTDOWN); return -EAGAIN; } @@ -1755,6 +1743,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr) dprintk("%s USE DS: %s cl_count %d vers %d\n", __func__, ds->ds_remotestr, atomic_read(&ds->ds_clp->cl_count), vers); + hdr->pgio_done_cb = ff_layout_read_done_cb; atomic_inc(&ds->ds_clp->cl_count); hdr->ds_clp = ds->ds_clp; fh = nfs4_ff_layout_select_ds_fh(lseg, idx); -- cgit v1.2.3-59-g8ed1b From 3b13b4b311a508a5899070936a873e30fcc96a2f Mon Sep 17 00:00:00 2001 From: Tom Haynes Date: Tue, 17 May 2016 12:28:37 -0400 Subject: pNFS/flexfiles: When checking for available DSes, conditionally check for MDS io Whenever we check to see if we have the needed number of DSes for the action, we may also have to check to see whether IO is allowed to go to the MDS or not. [jlayton: fix merge conflict due to lack of localio patches here] Signed-off-by: Tom Haynes Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 5 ++--- fs/nfs/flexfilelayout/flexfilelayout.h | 1 + fs/nfs/flexfilelayout/flexfilelayoutdev.c | 6 ++++++ 3 files changed, 9 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index b6e0e32b6053..a835b75f85ab 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1102,8 +1102,7 @@ static int ff_layout_async_handle_error_v4(struct rpc_task *task, rpc_wake_up(&tbl->slot_tbl_waitq); /* fall through */ default: - if (ff_layout_no_fallback_to_mds(lseg) || - ff_layout_has_available_ds(lseg)) + if (ff_layout_avoid_mds_available_ds(lseg)) return -NFS4ERR_RESET_TO_PNFS; reset: dprintk("%s Retry through MDS. Error %d\n", __func__, @@ -1765,7 +1764,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr) return PNFS_ATTEMPTED; out_failed: - if (ff_layout_has_available_ds(lseg)) + if (ff_layout_avoid_mds_available_ds(lseg)) return PNFS_TRY_AGAIN; return PNFS_NOT_ATTEMPTED; } diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index 1318c77aeb35..b54058122647 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -191,4 +191,5 @@ nfs4_ff_find_or_create_ds_client(struct pnfs_layout_segment *lseg, struct rpc_cred *ff_layout_get_ds_cred(struct pnfs_layout_segment *lseg, u32 ds_idx, struct rpc_cred *mdscred); bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg); +bool ff_layout_avoid_mds_available_ds(struct pnfs_layout_segment *lseg); #endif /* FS_NFS_NFS4FLEXFILELAYOUT_H */ diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index eeef89359ad2..8242ab327fee 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -541,6 +541,12 @@ bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg) return ff_rw_layout_has_available_ds(lseg); } +bool ff_layout_avoid_mds_available_ds(struct pnfs_layout_segment *lseg) +{ + return ff_layout_no_fallback_to_mds(lseg) || + ff_layout_has_available_ds(lseg); +} + module_param(dataserver_retrans, uint, 0644); MODULE_PARM_DESC(dataserver_retrans, "The number of times the NFSv4.1 client " "retries a request before it attempts further " -- cgit v1.2.3-59-g8ed1b From 446ca2195303c550b1cec7dcf9a478910dc84672 Mon Sep 17 00:00:00 2001 From: Tom Haynes Date: Tue, 17 May 2016 12:28:38 -0400 Subject: pNFS/flexfiles: When initing reads or writes, we might have to retry connecting to DSes If we are initializing reads or writes and can not connect to a DS, then check whether or not IO is allowed through the MDS. If it is allowed, reset to the MDS. Else, fail the layout segment and force a retry of a new layout segment. Signed-off-by: Tom Haynes Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 29 +++++++++++++++++++++++++---- 1 file changed, 25 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index a835b75f85ab..2911ceda8cd9 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -848,8 +848,13 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, goto out_mds; ds = ff_layout_choose_best_ds_for_read(pgio->pg_lseg, 0, &ds_idx); - if (!ds) - goto out_mds; + if (!ds) { + if (ff_layout_no_fallback_to_mds(pgio->pg_lseg)) + goto out_pnfs; + else + goto out_mds; + } + mirror = FF_LAYOUT_COMP(pgio->pg_lseg, ds_idx); pgio->pg_mirror_idx = ds_idx; @@ -863,6 +868,12 @@ out_mds: pnfs_put_lseg(pgio->pg_lseg); pgio->pg_lseg = NULL; nfs_pageio_reset_read_mds(pgio); + return; + +out_pnfs: + pnfs_set_lo_fail(pgio->pg_lseg); + pnfs_put_lseg(pgio->pg_lseg); + pgio->pg_lseg = NULL; } static void @@ -905,8 +916,12 @@ ff_layout_pg_init_write(struct nfs_pageio_descriptor *pgio, for (i = 0; i < pgio->pg_mirror_count; i++) { ds = nfs4_ff_layout_prepare_ds(pgio->pg_lseg, i, true); - if (!ds) - goto out_mds; + if (!ds) { + if (ff_layout_no_fallback_to_mds(pgio->pg_lseg)) + goto out_pnfs; + else + goto out_mds; + } pgm = &pgio->pg_mirrors[i]; mirror = FF_LAYOUT_COMP(pgio->pg_lseg, i); pgm->pg_bsize = mirror->mirror_ds->ds_versions[0].wsize; @@ -918,6 +933,12 @@ out_mds: pnfs_put_lseg(pgio->pg_lseg); pgio->pg_lseg = NULL; nfs_pageio_reset_write_mds(pgio); + return; + +out_pnfs: + pnfs_set_lo_fail(pgio->pg_lseg); + pnfs_put_lseg(pgio->pg_lseg); + pgio->pg_lseg = NULL; } static unsigned int -- cgit v1.2.3-59-g8ed1b From ee26bdd680aa10df628499fe95a7b88345574248 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:39 -0400 Subject: pnfs: don't merge new ff lsegs with ones that have LAYOUTRETURN bit set Otherwise, we'll end up returning layouts that we've just received if the client issues a new LAYOUTGET prior to the LAYOUTRETURN. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 2911ceda8cd9..3bac2094c175 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -299,6 +299,8 @@ ff_lseg_merge(struct pnfs_layout_segment *new, { u64 new_end, old_end; + if (test_bit(NFS_LSEG_LAYOUTRETURN, &old->pls_flags)) + return false; if (new->pls_range.iomode != old->pls_range.iomode) return false; old_end = pnfs_calc_offset_end(old->pls_range.offset, @@ -319,8 +321,6 @@ ff_lseg_merge(struct pnfs_layout_segment *new, new_end); if (test_bit(NFS_LSEG_ROC, &old->pls_flags)) set_bit(NFS_LSEG_ROC, &new->pls_flags); - if (test_bit(NFS_LSEG_LAYOUTRETURN, &old->pls_flags)) - set_bit(NFS_LSEG_LAYOUTRETURN, &new->pls_flags); return true; } -- cgit v1.2.3-59-g8ed1b From 6675528380f8b7fffa82047211d8352dc6797321 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:40 -0400 Subject: pnfs: record sequence in pnfs_layout_segment when it's created In later patches, we're going to teach the client to be more selective about how it returns layouts. This means keeping a record of what the stateid's seqid was at the time that the server handed out a layout segment. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 1 + fs/nfs/pnfs.h | 1 + 2 files changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index faea98281c75..c09ced88bd38 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1697,6 +1697,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) init_lseg(lo, lseg); lseg->pls_range = res->range; + lseg->pls_seq = be32_to_cpu(res->stateid.seqid); spin_lock(&ino->i_lock); if (pnfs_layoutgets_blocked(lo)) { diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 7222d3a35439..361fa5494aa5 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -64,6 +64,7 @@ struct pnfs_layout_segment { struct list_head pls_lc_list; struct pnfs_layout_range pls_range; atomic_t pls_refcount; + u32 pls_seq; unsigned long pls_flags; struct pnfs_layout_hdr *pls_layout; struct work_struct pls_work; -- cgit v1.2.3-59-g8ed1b From 3982a6a2d0e64441c938f4e7f93985aff37d8fb3 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:41 -0400 Subject: pnfs: keep track of the return sequence number in pnfs_layout_hdr When we want to selectively do a LAYOUTRETURN, we need to specify a stateid that represents most recent layout acquisition that is to be returned. When we mark a layout stateid to be returned, we update the return sequence number in the layout header with that value, if it's newer than the existing one. Then, when we go to do a LAYOUTRETURN on layout header put, we overwrite the seqid in the stateid with the saved one, and then zero it out. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 11 ++++++++--- fs/nfs/pnfs.h | 1 + 2 files changed, 9 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index c09ced88bd38..3ad6bd377d35 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -899,6 +899,7 @@ pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo) if (test_and_set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) return false; lo->plh_return_iomode = 0; + lo->plh_return_seq = 0; pnfs_get_layout_hdr(lo); clear_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags); return true; @@ -969,6 +970,7 @@ static void pnfs_layoutreturn_before_put_layout_hdr(struct pnfs_layout_hdr *lo) bool send; nfs4_stateid_copy(&stateid, &lo->plh_stateid); + stateid.seqid = cpu_to_be32(lo->plh_return_seq); iomode = lo->plh_return_iomode; send = pnfs_prepare_layoutreturn(lo); spin_unlock(&inode->i_lock); @@ -1747,7 +1749,8 @@ out_forget_reply: } static void -pnfs_set_plh_return_iomode(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode) +pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode, + u32 seq) { if (lo->plh_return_iomode == iomode) return; @@ -1755,6 +1758,8 @@ pnfs_set_plh_return_iomode(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode) iomode = IOMODE_ANY; lo->plh_return_iomode = iomode; set_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags); + if (!lo->plh_return_seq || pnfs_seqid_is_newer(seq, lo->plh_return_seq)) + lo->plh_return_seq = seq; } /** @@ -1793,7 +1798,7 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, continue; remaining++; set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags); - pnfs_set_plh_return_iomode(lo, return_range->iomode); + pnfs_set_plh_return_info(lo, return_range->iomode, lseg->pls_seq); } return remaining; } @@ -1811,7 +1816,7 @@ void pnfs_error_mark_layout_for_return(struct inode *inode, bool return_now = false; spin_lock(&inode->i_lock); - pnfs_set_plh_return_iomode(lo, range.iomode); + pnfs_set_plh_return_info(lo, range.iomode, lseg->pls_seq); /* * mark all matching lsegs so that we are sure to have no live * segments at hand when sending layoutreturn. See pnfs_put_lseg() diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 361fa5494aa5..3476c9850678 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -195,6 +195,7 @@ struct pnfs_layout_hdr { unsigned long plh_flags; nfs4_stateid plh_stateid; u32 plh_barrier; /* ignore lower seqids */ + u32 plh_return_seq; enum pnfs_iomode plh_return_iomode; loff_t plh_lwb; /* last write byte for layoutcommit */ struct rpc_cred *plh_lc_cred; /* layoutcommit cred */ -- cgit v1.2.3-59-g8ed1b From 6d597e1750126a2c6fd99e4b2cc176955bb9972e Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:42 -0400 Subject: pnfs: only tear down lsegs that precede seqid in LAYOUTRETURN args LAYOUTRETURN is "special" in that servers and clients are expected to work with old stateids. When the client sends a LAYOUTRETURN with an old stateid in it then the server is expected to only tear down layout segments that were present when that seqid was current. Ensure that the client handles its accounting accordingly. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/callback_proc.c | 3 ++- fs/nfs/nfs42proc.c | 2 +- fs/nfs/nfs4proc.c | 5 ++-- fs/nfs/pnfs.c | 64 +++++++++++++++++++++++++++++++++----------------- fs/nfs/pnfs.h | 6 +++-- 5 files changed, 52 insertions(+), 28 deletions(-) (limited to 'fs') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 7c9fbf504f07..aaa2e8d3df6f 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -217,7 +217,8 @@ static u32 initiate_file_draining(struct nfs_client *clp, } if (pnfs_mark_matching_lsegs_return(lo, &free_me_list, - &args->cbl_range)) { + &args->cbl_range, + be32_to_cpu(args->cbl_stateid.seqid))) { rv = NFS4_OK; goto unlock; } diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 579ee20e4120..20c44d1209dc 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -337,7 +337,7 @@ nfs42_layoutstat_done(struct rpc_task *task, void *calldata) * with the current stateid. */ set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags); - pnfs_mark_matching_lsegs_invalid(lo, &head, NULL); + pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0); spin_unlock(&inode->i_lock); pnfs_free_lseg_list(&head); } else diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index c9b66085d392..2a29f5d12aeb 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -7953,7 +7953,7 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata) * with the current stateid. */ set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags); - pnfs_mark_matching_lsegs_invalid(lo, &head, NULL); + pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0); spin_unlock(&inode->i_lock); pnfs_free_lseg_list(&head); } else @@ -8145,7 +8145,8 @@ static void nfs4_layoutreturn_release(void *calldata) dprintk("--> %s\n", __func__); spin_lock(&lo->plh_inode->i_lock); - pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range); + pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range, + be32_to_cpu(lrp->args.stateid.seqid)); pnfs_mark_layout_returned_if_empty(lo); if (lrp->res.lrs_present) pnfs_set_layout_stateid(lo, &lrp->res.stateid, true); diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 3ad6bd377d35..21ff53fe7a9e 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -270,7 +270,7 @@ pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo, }; set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags); - return pnfs_mark_matching_lsegs_invalid(lo, lseg_list, &range); + return pnfs_mark_matching_lsegs_invalid(lo, lseg_list, &range, 0); } static int @@ -308,7 +308,7 @@ pnfs_layout_io_set_failed(struct pnfs_layout_hdr *lo, u32 iomode) spin_lock(&inode->i_lock); pnfs_layout_set_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); - pnfs_mark_matching_lsegs_invalid(lo, &head, &range); + pnfs_mark_matching_lsegs_invalid(lo, &head, &range, 0); spin_unlock(&inode->i_lock); pnfs_free_lseg_list(&head); dprintk("%s Setting layout IOMODE_%s fail bit\n", __func__, @@ -522,13 +522,35 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg, return rv; } -/* Returns count of number of matching invalid lsegs remaining in list - * after call. +/* + * Compare 2 layout stateid sequence ids, to see which is newer, + * taking into account wraparound issues. + */ +static bool pnfs_seqid_is_newer(u32 s1, u32 s2) +{ + return (s32)(s1 - s2) > 0; +} + +/** + * pnfs_mark_matching_lsegs_invalid - tear down lsegs or mark them for later + * @lo: layout header containing the lsegs + * @tmp_list: list head where doomed lsegs should go + * @recall_range: optional recall range argument to match (may be NULL) + * @seq: only invalidate lsegs obtained prior to this sequence (may be 0) + * + * Walk the list of lsegs in the layout header, and tear down any that should + * be destroyed. If "recall_range" is specified then the segment must match + * that range. If "seq" is non-zero, then only match segments that were handed + * out at or before that sequence. + * + * Returns number of matching invalid lsegs remaining in list after scanning + * it and purging them. */ int pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, - const struct pnfs_layout_range *recall_range) + const struct pnfs_layout_range *recall_range, + u32 seq) { struct pnfs_layout_segment *lseg, *next; int remaining = 0; @@ -540,10 +562,12 @@ pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list) if (!recall_range || should_free_lseg(&lseg->pls_range, recall_range)) { - dprintk("%s: freeing lseg %p iomode %d " + if (seq && pnfs_seqid_is_newer(lseg->pls_seq, seq)) + continue; + dprintk("%s: freeing lseg %p iomode %d seq %u" "offset %llu length %llu\n", __func__, - lseg, lseg->pls_range.iomode, lseg->pls_range.offset, - lseg->pls_range.length); + lseg, lseg->pls_range.iomode, lseg->pls_seq, + lseg->pls_range.offset, lseg->pls_range.length); if (!mark_lseg_invalid(lseg, tmp_list)) remaining++; } @@ -730,15 +754,6 @@ pnfs_destroy_all_layouts(struct nfs_client *clp) pnfs_destroy_layouts_byclid(clp, false); } -/* - * Compare 2 layout stateid sequence ids, to see which is newer, - * taking into account wraparound issues. - */ -static bool pnfs_seqid_is_newer(u32 s1, u32 s2) -{ - return (s32)(s1 - s2) > 0; -} - /* update lo->plh_stateid with new if is more recent */ void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, const nfs4_stateid *new, @@ -1014,7 +1029,7 @@ _pnfs_return_layout(struct inode *ino) pnfs_get_layout_hdr(lo); empty = list_empty(&lo->plh_segs); pnfs_clear_layoutcommit(ino, &tmp_list); - pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL); + pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL, 0); if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) { struct pnfs_layout_range range = { @@ -1721,7 +1736,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) * inode invalid, and don't bother validating the stateid * sequence number. */ - pnfs_mark_matching_lsegs_invalid(lo, &free_me, NULL); + pnfs_mark_matching_lsegs_invalid(lo, &free_me, NULL, 0); nfs4_stateid_copy(&lo->plh_stateid, &res->stateid); lo->plh_barrier = be32_to_cpu(res->stateid.seqid); @@ -1775,7 +1790,8 @@ pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode, int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, - const struct pnfs_layout_range *return_range) + const struct pnfs_layout_range *return_range, + u32 seq) { struct pnfs_layout_segment *lseg, *next; int remaining = 0; @@ -1798,8 +1814,11 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, continue; remaining++; set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags); - pnfs_set_plh_return_info(lo, return_range->iomode, lseg->pls_seq); } + + if (remaining) + pnfs_set_plh_return_info(lo, return_range->iomode, seq); + return remaining; } @@ -1822,7 +1841,8 @@ void pnfs_error_mark_layout_for_return(struct inode *inode, * segments at hand when sending layoutreturn. See pnfs_put_lseg() * for how it works. */ - if (!pnfs_mark_matching_lsegs_return(lo, &free_me, &range)) { + if (!pnfs_mark_matching_lsegs_return(lo, &free_me, + &range, lseg->pls_seq)) { nfs4_stateid stateid; enum pnfs_iomode iomode = lo->plh_return_iomode; diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 3476c9850678..971068b58647 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -266,10 +266,12 @@ int pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct nfs4_state *open_state); int pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, - const struct pnfs_layout_range *recall_range); + const struct pnfs_layout_range *recall_range, + u32 seq); int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, - const struct pnfs_layout_range *recall_range); + const struct pnfs_layout_range *recall_range, + u32 seq); bool pnfs_roc(struct inode *ino); void pnfs_roc_release(struct inode *ino); void pnfs_roc_set_barrier(struct inode *ino, u32 barrier); -- cgit v1.2.3-59-g8ed1b From 094069f1d96f691637b5c335f5eb79566d2457ff Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:43 -0400 Subject: flexfiles: remove pointless setting of NFS_LAYOUT_RETURN_REQUESTED Setting just the NFS_LAYOUT_RETURN_REQUESTED flag doesn't do anything, unless there are lsegs that are also being marked for return. At the point where that happens this flag is also set, so these set_bit calls don't do anything useful. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 2 -- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 8 +------- 2 files changed, 1 insertion(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 3bac2094c175..40bccb2c06d1 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1270,8 +1270,6 @@ static int ff_layout_read_done_cb(struct rpc_task *task, hdr->pgio_mirror_idx + 1, &hdr->pgio_mirror_idx)) goto out_eagain; - set_bit(NFS_LAYOUT_RETURN_REQUESTED, - &hdr->lseg->pls_layout->plh_flags); pnfs_read_resend_pnfs(hdr); return task->tk_status; case -NFS4ERR_RESET_TO_MDS: diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index 8242ab327fee..bb598923b2ef 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -394,13 +394,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, mirror, lseg->pls_range.offset, lseg->pls_range.length, NFS4ERR_NXIO, OP_ILLEGAL, GFP_NOIO); - if (!fail_return) { - if (ff_layout_has_available_ds(lseg)) - set_bit(NFS_LAYOUT_RETURN_REQUESTED, - &lseg->pls_layout->plh_flags); - else - pnfs_error_mark_layout_for_return(ino, lseg); - } else + if (fail_return || !ff_layout_has_available_ds(lseg)) pnfs_error_mark_layout_for_return(ino, lseg); ds = NULL; } -- cgit v1.2.3-59-g8ed1b From 95e2b7e95d43c5c95620b12355a90713268376d3 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:44 -0400 Subject: flexfiles: add kerneldoc header to nfs4_ff_layout_prepare_ds Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index bb598923b2ef..35d84d0c0f21 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -343,7 +343,23 @@ out: return fh; } -/* Upon return, either ds is connected, or ds is NULL */ +/** + * nfs4_ff_layout_prepare_ds - prepare a DS connection for an RPC call + * @lseg: the layout segment we're operating on + * @ds_idx: index of the DS to use + * @fail_return: return layout on connect failure? + * + * Try to prepare a DS connection to accept an RPC call. This involves + * selecting a mirror to use and connecting the client to it if it's not + * already connected. + * + * Since we only need a single functioning mirror to satisfy a read, we don't + * want to return the layout if there is one. For writes though, any down + * mirror should result in a LAYOUTRETURN. @fail_return is how we distinguish + * between the two cases. + * + * Returns a pointer to a connected DS object on success or NULL on failure. + */ struct nfs4_pnfs_ds * nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, bool fail_return) -- cgit v1.2.3-59-g8ed1b From d03ab29dbbe905c6c7c5abd78e02547fa954ec07 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:45 -0400 Subject: pnfs: fix bad error handling in send_layoutget Currently, the code will clear the fail bit if we get back a fatal error. I don't think that's correct -- we want to clear that bit if we do not get a fatal error. Fixes: 0bcbf039f6 (nfs: handle request add failure properly) Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 21ff53fe7a9e..e3e9973b5f76 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -876,11 +876,16 @@ send_layoutget(struct pnfs_layout_hdr *lo, lseg = nfs4_proc_layoutget(lgp, gfp_flags); } while (lseg == ERR_PTR(-EAGAIN)); - if (IS_ERR(lseg) && !nfs_error_is_fatal(PTR_ERR(lseg))) - lseg = NULL; - else + if (IS_ERR(lseg)) { + if (!nfs_error_is_fatal(PTR_ERR(lseg))) { + pnfs_layout_clear_fail_bit(lo, + pnfs_iomode_to_fail_bit(range->iomode)); + lseg = NULL; + } + } else { pnfs_layout_clear_fail_bit(lo, pnfs_iomode_to_fail_bit(range->iomode)); + } return lseg; } -- cgit v1.2.3-59-g8ed1b From 83026d80a16ea6a4e4f06a994fc7831b1d8d6375 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:46 -0400 Subject: pnfs: lift retry logic from send_layoutget to pnfs_update_layout If we get back something like NFS4ERR_OLD_STATEID, that will be translated into -EAGAIN, and the do/while loop in send_layoutget will drive the call again. This is not quite what we want, I think. An error like that is a sign that something has changed. That something could have been a concurrent LAYOUTGET that would give us a usable lseg. Lift the retry logic into pnfs_update_layout instead. That allows us to redo the layout search, and may spare us from having to issue an RPC. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 72 +++++++++++++++++++++++++++++------------------------------ 1 file changed, 36 insertions(+), 36 deletions(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index e3e9973b5f76..2a609af845fe 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -839,7 +839,6 @@ send_layoutget(struct pnfs_layout_hdr *lo, struct inode *ino = lo->plh_inode; struct nfs_server *server = NFS_SERVER(ino); struct nfs4_layoutget *lgp; - struct pnfs_layout_segment *lseg; loff_t i_size; dprintk("--> %s\n", __func__); @@ -849,45 +848,30 @@ send_layoutget(struct pnfs_layout_hdr *lo, * store in lseg. If we race with a concurrent seqid morphing * op, then re-send the LAYOUTGET. */ - do { - lgp = kzalloc(sizeof(*lgp), gfp_flags); - if (lgp == NULL) - return NULL; - - i_size = i_size_read(ino); - - lgp->args.minlength = PAGE_SIZE; - if (lgp->args.minlength > range->length) - lgp->args.minlength = range->length; - if (range->iomode == IOMODE_READ) { - if (range->offset >= i_size) - lgp->args.minlength = 0; - else if (i_size - range->offset < lgp->args.minlength) - lgp->args.minlength = i_size - range->offset; - } - lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE; - pnfs_copy_range(&lgp->args.range, range); - lgp->args.type = server->pnfs_curr_ld->id; - lgp->args.inode = ino; - lgp->args.ctx = get_nfs_open_context(ctx); - lgp->gfp_flags = gfp_flags; - lgp->cred = lo->plh_lc_cred; + lgp = kzalloc(sizeof(*lgp), gfp_flags); + if (lgp == NULL) + return ERR_PTR(-ENOMEM); - lseg = nfs4_proc_layoutget(lgp, gfp_flags); - } while (lseg == ERR_PTR(-EAGAIN)); + i_size = i_size_read(ino); - if (IS_ERR(lseg)) { - if (!nfs_error_is_fatal(PTR_ERR(lseg))) { - pnfs_layout_clear_fail_bit(lo, - pnfs_iomode_to_fail_bit(range->iomode)); - lseg = NULL; - } - } else { - pnfs_layout_clear_fail_bit(lo, - pnfs_iomode_to_fail_bit(range->iomode)); + lgp->args.minlength = PAGE_SIZE; + if (lgp->args.minlength > range->length) + lgp->args.minlength = range->length; + if (range->iomode == IOMODE_READ) { + if (range->offset >= i_size) + lgp->args.minlength = 0; + else if (i_size - range->offset < lgp->args.minlength) + lgp->args.minlength = i_size - range->offset; } + lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE; + pnfs_copy_range(&lgp->args.range, range); + lgp->args.type = server->pnfs_curr_ld->id; + lgp->args.inode = ino; + lgp->args.ctx = get_nfs_open_context(ctx); + lgp->gfp_flags = gfp_flags; + lgp->cred = lo->plh_lc_cred; - return lseg; + return nfs4_proc_layoutget(lgp, gfp_flags); } static void pnfs_clear_layoutcommit(struct inode *inode, @@ -1649,6 +1633,22 @@ lookup_again: arg.length = PAGE_ALIGN(arg.length); lseg = send_layoutget(lo, ctx, &arg, gfp_flags); + if (IS_ERR(lseg)) { + if (lseg == ERR_PTR(-EAGAIN)) { + if (first) + pnfs_clear_first_layoutget(lo); + pnfs_put_layout_hdr(lo); + goto lookup_again; + } + + if (!nfs_error_is_fatal(PTR_ERR(lseg))) { + pnfs_layout_clear_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); + lseg = NULL; + } + } else { + pnfs_layout_clear_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); + } + atomic_dec(&lo->plh_outstanding); trace_pnfs_update_layout(ino, pos, count, iomode, lo, PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET); -- cgit v1.2.3-59-g8ed1b From 183d9e7b112aaed0d19c16ffcf0f8c3a86dc71e0 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:47 -0400 Subject: pnfs: rework LAYOUTGET retry handling There are several problems in the way a stateid is selected for a LAYOUTGET operation: We pick a stateid to use in the RPC prepare op, but that makes it difficult to serialize LAYOUTGETs that use the open stateid. That serialization is done in pnfs_update_layout, which occurs well before the rpc_prepare operation. Between those two events, the i_lock is dropped and reacquired. pnfs_update_layout can find that the list has lsegs in it and not do any serialization, but then later pnfs_choose_layoutget_stateid ends up choosing the open stateid. This patch changes the client to select the stateid to use in the LAYOUTGET earlier, when we're searching for a usable layout segment. This way we can do it all while holding the i_lock the first time, and ensure that we serialize any LAYOUTGET call that uses a non-layout stateid. This also means a rework of how LAYOUTGET replies are handled, as we must now get the latest stateid if we want to retransmit in response to a retryable error. Most of those errors boil down to the fact that the layout state has changed in some fashion. Thus, what we really want to do is to re-search for a layout when it fails with a retryable error, so that we can avoid reissuing the RPC at all if possible. While the LAYOUTGET RPC is async, the initiating thread always waits for it to complete, so it's effectively synchronous anyway. Currently, when we need to retry a LAYOUTGET because of an error, we drive that retry via the rpc state machine. This means that once the call has been submitted, it runs until it completes. So, we must move the error handling for this RPC out of the rpc_call_done operation and into the caller. In order to handle errors like NFS4ERR_DELAY properly, we must also pass a pointer to the sliding timeout, which is now moved to the stack in pnfs_update_layout. The complicating errors are -NFS4ERR_RECALLCONFLICT and -NFS4ERR_LAYOUTTRYLATER, as those involve a timeout after which we give up and return NULL back to the caller. So, there is some special handling for those errors to ensure that the layers driving the retries can handle that appropriately. Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/nfs4proc.c | 115 ++++++++++++++++---------------------- fs/nfs/nfs4trace.h | 10 +++- fs/nfs/pnfs.c | 144 +++++++++++++++++++++++++----------------------- fs/nfs/pnfs.h | 6 +- include/linux/errno.h | 1 + include/linux/nfs4.h | 2 + include/linux/nfs_xdr.h | 2 - 7 files changed, 136 insertions(+), 144 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 2a29f5d12aeb..62d67f040906 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -427,6 +427,7 @@ static int nfs4_do_handle_exception(struct nfs_server *server, case -NFS4ERR_DELAY: nfs_inc_server_stats(server, NFSIOS_DELAY); case -NFS4ERR_GRACE: + case -NFS4ERR_RECALLCONFLICT: exception->delay = 1; return 0; @@ -7847,40 +7848,34 @@ nfs4_layoutget_prepare(struct rpc_task *task, void *calldata) struct nfs4_layoutget *lgp = calldata; struct nfs_server *server = NFS_SERVER(lgp->args.inode); struct nfs4_session *session = nfs4_get_session(server); - int ret; dprintk("--> %s\n", __func__); - /* Note the is a race here, where a CB_LAYOUTRECALL can come in - * right now covering the LAYOUTGET we are about to send. - * However, that is not so catastrophic, and there seems - * to be no way to prevent it completely. - */ - if (nfs41_setup_sequence(session, &lgp->args.seq_args, - &lgp->res.seq_res, task)) - return; - ret = pnfs_choose_layoutget_stateid(&lgp->args.stateid, - NFS_I(lgp->args.inode)->layout, - &lgp->args.range, - lgp->args.ctx->state); - if (ret < 0) - rpc_exit(task, ret); + nfs41_setup_sequence(session, &lgp->args.seq_args, + &lgp->res.seq_res, task); + dprintk("<-- %s\n", __func__); } static void nfs4_layoutget_done(struct rpc_task *task, void *calldata) { struct nfs4_layoutget *lgp = calldata; + + dprintk("--> %s\n", __func__); + nfs41_sequence_done(task, &lgp->res.seq_res); + dprintk("<-- %s\n", __func__); +} + +static int +nfs4_layoutget_handle_exception(struct rpc_task *task, + struct nfs4_layoutget *lgp, struct nfs4_exception *exception) +{ struct inode *inode = lgp->args.inode; struct nfs_server *server = NFS_SERVER(inode); struct pnfs_layout_hdr *lo; - struct nfs4_state *state = NULL; - unsigned long timeo, now, giveup; + int status = task->tk_status; dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status); - if (!nfs41_sequence_done(task, &lgp->res.seq_res)) - goto out; - - switch (task->tk_status) { + switch (status) { case 0: goto out; @@ -7890,57 +7885,43 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata) * retry go inband. */ case -NFS4ERR_LAYOUTUNAVAILABLE: - task->tk_status = -ENODATA; + status = -ENODATA; goto out; /* * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3). */ case -NFS4ERR_BADLAYOUT: - goto out_overflow; + status = -EOVERFLOW; + goto out; /* * NFS4ERR_LAYOUTTRYLATER is a conflict with another client * (or clients) writing to the same RAID stripe except when * the minlength argument is 0 (see RFC5661 section 18.43.3). + * + * Treat it like we would RECALLCONFLICT -- we retry for a little + * while, and then eventually give up. */ case -NFS4ERR_LAYOUTTRYLATER: - if (lgp->args.minlength == 0) - goto out_overflow; - /* - * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall - * existing layout before getting a new one). - */ - case -NFS4ERR_RECALLCONFLICT: - timeo = rpc_get_timeout(task->tk_client); - giveup = lgp->args.timestamp + timeo; - now = jiffies; - if (time_after(giveup, now)) { - unsigned long delay; - - /* Delay for: - * - Not less then NFS4_POLL_RETRY_MIN. - * - One last time a jiffie before we give up - * - exponential backoff (time_now minus start_attempt) - */ - delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN, - min((giveup - now - 1), - now - lgp->args.timestamp)); - - dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n", - __func__, delay); - rpc_delay(task, delay); - /* Do not call nfs4_async_handle_error() */ - goto out_restart; + if (lgp->args.minlength == 0) { + status = -EOVERFLOW; + goto out; } - break; + /* Fallthrough */ + case -NFS4ERR_RECALLCONFLICT: + nfs4_handle_exception(server, -NFS4ERR_RECALLCONFLICT, + exception); + status = -ERECALLCONFLICT; + goto out; case -NFS4ERR_EXPIRED: case -NFS4ERR_BAD_STATEID: + exception->timeout = 0; spin_lock(&inode->i_lock); if (nfs4_stateid_match(&lgp->args.stateid, &lgp->args.ctx->state->stateid)) { spin_unlock(&inode->i_lock); /* If the open stateid was bad, then recover it. */ - state = lgp->args.ctx->state; + exception->state = lgp->args.ctx->state; break; } lo = NFS_I(inode)->layout; @@ -7958,20 +7939,16 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata) pnfs_free_lseg_list(&head); } else spin_unlock(&inode->i_lock); - goto out_restart; + status = -EAGAIN; + goto out; } - if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN) - goto out_restart; + + status = nfs4_handle_exception(server, status, exception); + if (exception->retry) + status = -EAGAIN; out: dprintk("<-- %s\n", __func__); - return; -out_restart: - task->tk_status = 0; - rpc_restart_call_prepare(task); - return; -out_overflow: - task->tk_status = -EOVERFLOW; - goto out; + return status; } static size_t max_response_pages(struct nfs_server *server) @@ -8040,7 +8017,7 @@ static const struct rpc_call_ops nfs4_layoutget_call_ops = { }; struct pnfs_layout_segment * -nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags) +nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags) { struct inode *inode = lgp->args.inode; struct nfs_server *server = NFS_SERVER(inode); @@ -8060,6 +8037,7 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags) .flags = RPC_TASK_ASYNC, }; struct pnfs_layout_segment *lseg = NULL; + struct nfs4_exception exception = { .timeout = *timeout }; int status = 0; dprintk("--> %s\n", __func__); @@ -8073,7 +8051,6 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags) return ERR_PTR(-ENOMEM); } lgp->args.layout.pglen = max_pages * PAGE_SIZE; - lgp->args.timestamp = jiffies; lgp->res.layoutp = &lgp->args.layout; lgp->res.seq_res.sr_slot = NULL; @@ -8083,13 +8060,17 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags) if (IS_ERR(task)) return ERR_CAST(task); status = nfs4_wait_for_completion_rpc_task(task); - if (status == 0) - status = task->tk_status; + if (status == 0) { + status = nfs4_layoutget_handle_exception(task, lgp, &exception); + *timeout = exception.timeout; + } + trace_nfs4_layoutget(lgp->args.ctx, &lgp->args.range, &lgp->res.range, &lgp->res.stateid, status); + /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */ if (status == 0 && lgp->res.layoutp->len) lseg = pnfs_layout_process(lgp); diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 2c8d05dae5b1..9c150b153782 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -1520,6 +1520,8 @@ DEFINE_NFS4_INODE_EVENT(nfs4_layoutreturn_on_close); { PNFS_UPDATE_LAYOUT_FOUND_CACHED, "found cached" }, \ { PNFS_UPDATE_LAYOUT_RETURN, "layoutreturn" }, \ { PNFS_UPDATE_LAYOUT_BLOCKED, "layouts blocked" }, \ + { PNFS_UPDATE_LAYOUT_INVALID_OPEN, "invalid open" }, \ + { PNFS_UPDATE_LAYOUT_RETRY, "retrying" }, \ { PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET, "sent layoutget" }) TRACE_EVENT(pnfs_update_layout, @@ -1528,9 +1530,10 @@ TRACE_EVENT(pnfs_update_layout, u64 count, enum pnfs_iomode iomode, struct pnfs_layout_hdr *lo, + struct pnfs_layout_segment *lseg, enum pnfs_update_layout_reason reason ), - TP_ARGS(inode, pos, count, iomode, lo, reason), + TP_ARGS(inode, pos, count, iomode, lo, lseg, reason), TP_STRUCT__entry( __field(dev_t, dev) __field(u64, fileid) @@ -1540,6 +1543,7 @@ TRACE_EVENT(pnfs_update_layout, __field(enum pnfs_iomode, iomode) __field(int, layoutstateid_seq) __field(u32, layoutstateid_hash) + __field(long, lseg) __field(enum pnfs_update_layout_reason, reason) ), TP_fast_assign( @@ -1559,11 +1563,12 @@ TRACE_EVENT(pnfs_update_layout, __entry->layoutstateid_seq = 0; __entry->layoutstateid_hash = 0; } + __entry->lseg = (long)lseg; ), TP_printk( "fileid=%02x:%02x:%llu fhandle=0x%08x " "iomode=%s pos=%llu count=%llu " - "layoutstateid=%d:0x%08x (%s)", + "layoutstateid=%d:0x%08x lseg=0x%lx (%s)", MAJOR(__entry->dev), MINOR(__entry->dev), (unsigned long long)__entry->fileid, __entry->fhandle, @@ -1571,6 +1576,7 @@ TRACE_EVENT(pnfs_update_layout, (unsigned long long)__entry->pos, (unsigned long long)__entry->count, __entry->layoutstateid_seq, __entry->layoutstateid_hash, + __entry->lseg, show_pnfs_update_layout_reason(__entry->reason) ) ); diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 2a609af845fe..46339a7fb191 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -796,45 +796,18 @@ pnfs_layoutgets_blocked(const struct pnfs_layout_hdr *lo) test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); } -int -pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo, - const struct pnfs_layout_range *range, - struct nfs4_state *open_state) -{ - int status = 0; - - dprintk("--> %s\n", __func__); - spin_lock(&lo->plh_inode->i_lock); - if (pnfs_layoutgets_blocked(lo)) { - status = -EAGAIN; - } else if (!nfs4_valid_open_stateid(open_state)) { - status = -EBADF; - } else if (list_empty(&lo->plh_segs) || - test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags)) { - int seq; - - do { - seq = read_seqbegin(&open_state->seqlock); - nfs4_stateid_copy(dst, &open_state->stateid); - } while (read_seqretry(&open_state->seqlock, seq)); - } else - nfs4_stateid_copy(dst, &lo->plh_stateid); - spin_unlock(&lo->plh_inode->i_lock); - dprintk("<-- %s\n", __func__); - return status; -} - /* -* Get layout from server. -* for now, assume that whole file layouts are requested. -* arg->offset: 0 -* arg->length: all ones -*/ + * Get layout from server. + * for now, assume that whole file layouts are requested. + * arg->offset: 0 + * arg->length: all ones + */ static struct pnfs_layout_segment * send_layoutget(struct pnfs_layout_hdr *lo, struct nfs_open_context *ctx, + nfs4_stateid *stateid, const struct pnfs_layout_range *range, - gfp_t gfp_flags) + long *timeout, gfp_t gfp_flags) { struct inode *ino = lo->plh_inode; struct nfs_server *server = NFS_SERVER(ino); @@ -868,10 +841,11 @@ send_layoutget(struct pnfs_layout_hdr *lo, lgp->args.type = server->pnfs_curr_ld->id; lgp->args.inode = ino; lgp->args.ctx = get_nfs_open_context(ctx); + nfs4_stateid_copy(&lgp->args.stateid, stateid); lgp->gfp_flags = gfp_flags; lgp->cred = lo->plh_lc_cred; - return nfs4_proc_layoutget(lgp, gfp_flags); + return nfs4_proc_layoutget(lgp, timeout, gfp_flags); } static void pnfs_clear_layoutcommit(struct inode *inode, @@ -1511,27 +1485,30 @@ pnfs_update_layout(struct inode *ino, .offset = pos, .length = count, }; - unsigned pg_offset; + unsigned pg_offset, seq; struct nfs_server *server = NFS_SERVER(ino); struct nfs_client *clp = server->nfs_client; - struct pnfs_layout_hdr *lo; + struct pnfs_layout_hdr *lo = NULL; struct pnfs_layout_segment *lseg = NULL; + nfs4_stateid stateid; + long timeout = 0; + unsigned long giveup = jiffies + rpc_get_timeout(server->client); bool first; if (!pnfs_enabled_sb(NFS_SERVER(ino))) { - trace_pnfs_update_layout(ino, pos, count, iomode, NULL, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_NO_PNFS); goto out; } if (iomode == IOMODE_READ && i_size_read(ino) == 0) { - trace_pnfs_update_layout(ino, pos, count, iomode, NULL, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_RD_ZEROLEN); goto out; } if (pnfs_within_mdsthreshold(ctx, ino, iomode)) { - trace_pnfs_update_layout(ino, pos, count, iomode, NULL, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_MDSTHRESH); goto out; } @@ -1542,14 +1519,14 @@ lookup_again: lo = pnfs_find_alloc_layout(ino, ctx, gfp_flags); if (lo == NULL) { spin_unlock(&ino->i_lock); - trace_pnfs_update_layout(ino, pos, count, iomode, NULL, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_NOMEM); goto out; } /* Do we even need to bother with this? */ if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) { - trace_pnfs_update_layout(ino, pos, count, iomode, lo, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_BULK_RECALL); dprintk("%s matches recall, use MDS\n", __func__); goto out_unlock; @@ -1557,14 +1534,34 @@ lookup_again: /* if LAYOUTGET already failed once we don't try again */ if (pnfs_layout_io_test_failed(lo, iomode)) { - trace_pnfs_update_layout(ino, pos, count, iomode, lo, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_IO_TEST_FAIL); goto out_unlock; } - first = list_empty(&lo->plh_segs); - if (first) { - /* The first layoutget for the file. Need to serialize per + lseg = pnfs_find_lseg(lo, &arg); + if (lseg) { + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, + PNFS_UPDATE_LAYOUT_FOUND_CACHED); + goto out_unlock; + } + + if (!nfs4_valid_open_stateid(ctx->state)) { + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, + PNFS_UPDATE_LAYOUT_INVALID_OPEN); + goto out_unlock; + } + + /* + * Choose a stateid for the LAYOUTGET. If we don't have a layout + * stateid, or it has been invalidated, then we must use the open + * stateid. + */ + if (lo->plh_stateid.seqid == 0 || + test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags)) { + + /* + * The first layoutget for the file. Need to serialize per * RFC 5661 Errata 3208. */ if (test_and_set_bit(NFS_LAYOUT_FIRST_LAYOUTGET, @@ -1573,18 +1570,17 @@ lookup_again: wait_on_bit(&lo->plh_flags, NFS_LAYOUT_FIRST_LAYOUTGET, TASK_UNINTERRUPTIBLE); pnfs_put_layout_hdr(lo); + dprintk("%s retrying\n", __func__); goto lookup_again; } + + first = true; + do { + seq = read_seqbegin(&ctx->state->seqlock); + nfs4_stateid_copy(&stateid, &ctx->state->stateid); + } while (read_seqretry(&ctx->state->seqlock, seq)); } else { - /* Check to see if the layout for the given range - * already exists - */ - lseg = pnfs_find_lseg(lo, &arg); - if (lseg) { - trace_pnfs_update_layout(ino, pos, count, iomode, lo, - PNFS_UPDATE_LAYOUT_FOUND_CACHED); - goto out_unlock; - } + nfs4_stateid_copy(&stateid, &lo->plh_stateid); } /* @@ -1599,15 +1595,17 @@ lookup_again: pnfs_clear_first_layoutget(lo); pnfs_put_layout_hdr(lo); dprintk("%s retrying\n", __func__); + trace_pnfs_update_layout(ino, pos, count, iomode, lo, + lseg, PNFS_UPDATE_LAYOUT_RETRY); goto lookup_again; } - trace_pnfs_update_layout(ino, pos, count, iomode, lo, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_RETURN); goto out_put_layout_hdr; } if (pnfs_layoutgets_blocked(lo)) { - trace_pnfs_update_layout(ino, pos, count, iomode, lo, + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_BLOCKED); goto out_unlock; } @@ -1632,26 +1630,36 @@ lookup_again: if (arg.length != NFS4_MAX_UINT64) arg.length = PAGE_ALIGN(arg.length); - lseg = send_layoutget(lo, ctx, &arg, gfp_flags); + lseg = send_layoutget(lo, ctx, &stateid, &arg, &timeout, gfp_flags); + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, + PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET); if (IS_ERR(lseg)) { - if (lseg == ERR_PTR(-EAGAIN)) { + switch(PTR_ERR(lseg)) { + case -ERECALLCONFLICT: + if (time_after(jiffies, giveup)) + lseg = NULL; + /* Fallthrough */ + case -EAGAIN: + pnfs_put_layout_hdr(lo); if (first) pnfs_clear_first_layoutget(lo); - pnfs_put_layout_hdr(lo); - goto lookup_again; - } - - if (!nfs_error_is_fatal(PTR_ERR(lseg))) { - pnfs_layout_clear_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); - lseg = NULL; + if (lseg) { + trace_pnfs_update_layout(ino, pos, count, + iomode, lo, lseg, PNFS_UPDATE_LAYOUT_RETRY); + goto lookup_again; + } + /* Fallthrough */ + default: + if (!nfs_error_is_fatal(PTR_ERR(lseg))) { + pnfs_layout_clear_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); + lseg = NULL; + } } } else { pnfs_layout_clear_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); } atomic_dec(&lo->plh_outstanding); - trace_pnfs_update_layout(ino, pos, count, iomode, lo, - PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET); out_put_layout_hdr: if (first) pnfs_clear_first_layoutget(lo); diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 971068b58647..f9f3331bef49 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -228,7 +228,7 @@ extern void pnfs_unregister_layoutdriver(struct pnfs_layoutdriver_type *); extern int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *dev, struct rpc_cred *cred); -extern struct pnfs_layout_segment* nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags); +extern struct pnfs_layout_segment* nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags); extern int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync); /* pnfs.c */ @@ -260,10 +260,6 @@ void pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo); void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, const nfs4_stateid *new, bool update_barrier); -int pnfs_choose_layoutget_stateid(nfs4_stateid *dst, - struct pnfs_layout_hdr *lo, - const struct pnfs_layout_range *range, - struct nfs4_state *open_state); int pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, const struct pnfs_layout_range *recall_range, diff --git a/include/linux/errno.h b/include/linux/errno.h index 89627b9187f9..7ce9fb1b7d28 100644 --- a/include/linux/errno.h +++ b/include/linux/errno.h @@ -28,5 +28,6 @@ #define EBADTYPE 527 /* Type not supported by server */ #define EJUKEBOX 528 /* Request initiated, but will not complete before timeout */ #define EIOCBQUEUED 529 /* iocb queued, will get completion event */ +#define ERECALLCONFLICT 530 /* conflict with recalled state */ #endif diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index e1692c96cbc8..bfed6b367350 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -637,7 +637,9 @@ enum pnfs_update_layout_reason { PNFS_UPDATE_LAYOUT_IO_TEST_FAIL, PNFS_UPDATE_LAYOUT_FOUND_CACHED, PNFS_UPDATE_LAYOUT_RETURN, + PNFS_UPDATE_LAYOUT_RETRY, PNFS_UPDATE_LAYOUT_BLOCKED, + PNFS_UPDATE_LAYOUT_INVALID_OPEN, PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET, }; diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index e70ed54dad94..ccb2928a0e64 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -233,7 +233,6 @@ struct nfs4_layoutget_args { struct inode *inode; struct nfs_open_context *ctx; nfs4_stateid stateid; - unsigned long timestamp; struct nfs4_layoutdriver_data layout; }; @@ -251,7 +250,6 @@ struct nfs4_layoutget { struct nfs4_layoutget_res res; struct rpc_cred *cred; gfp_t gfp_flags; - long timeout; }; struct nfs4_getdeviceinfo_args { -- cgit v1.2.3-59-g8ed1b From 1b3c6d07e29515064aca8a9f86efaea7da4da027 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 17 May 2016 12:28:48 -0400 Subject: pnfs: make pnfs_layout_process more robust It can return NULL if layoutgets are blocked currently. Fix it to return -EAGAIN in that case, so we can properly handle it in pnfs_update_layout. Also, clean up and simplify the error handling -- eliminate "status" and just use "lseg". Signed-off-by: Jeff Layton Signed-off-by: Anna Schumaker --- fs/nfs/pnfs.c | 27 +++++++++++---------------- 1 file changed, 11 insertions(+), 16 deletions(-) (limited to 'fs') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 46339a7fb191..79ae3049608d 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1708,21 +1708,19 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) struct pnfs_layout_segment *lseg; struct inode *ino = lo->plh_inode; LIST_HEAD(free_me); - int status = -EINVAL; if (!pnfs_sanity_check_layout_range(&res->range)) - goto out; + return ERR_PTR(-EINVAL); /* Inject layout blob into I/O device driver */ lseg = NFS_SERVER(ino)->pnfs_curr_ld->alloc_lseg(lo, res, lgp->gfp_flags); - if (!lseg || IS_ERR(lseg)) { + if (IS_ERR_OR_NULL(lseg)) { if (!lseg) - status = -ENOMEM; - else - status = PTR_ERR(lseg); - dprintk("%s: Could not allocate layout: error %d\n", - __func__, status); - goto out; + lseg = ERR_PTR(-ENOMEM); + + dprintk("%s: Could not allocate layout: error %ld\n", + __func__, PTR_ERR(lseg)); + return lseg; } init_lseg(lo, lseg); @@ -1732,15 +1730,14 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) spin_lock(&ino->i_lock); if (pnfs_layoutgets_blocked(lo)) { dprintk("%s forget reply due to state\n", __func__); - goto out_forget_reply; + goto out_forget; } if (nfs4_stateid_match_other(&lo->plh_stateid, &res->stateid)) { /* existing state ID, make sure the sequence number matches. */ if (pnfs_layout_stateid_blocked(lo, &res->stateid)) { dprintk("%s forget reply due to sequence\n", __func__); - status = -EAGAIN; - goto out_forget_reply; + goto out_forget; } pnfs_set_layout_stateid(lo, &res->stateid, false); } else { @@ -1766,14 +1763,12 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&free_me); return lseg; -out: - return ERR_PTR(status); -out_forget_reply: +out_forget: spin_unlock(&ino->i_lock); lseg->pls_layout = lo; NFS_SERVER(ino)->pnfs_curr_ld->free_lseg(lseg); - goto out; + return ERR_PTR(-EAGAIN); } static void -- cgit v1.2.3-59-g8ed1b From 2997bfd0492d33d230b1475909a0211157a734bc Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Mon, 23 May 2016 13:21:01 +0300 Subject: NFS: checking for NULL instead of IS_ERR() in nfs_commit_file() nfs_create_request() doesn't return NULL, it returns error pointers. Fixes: 67911c8f18b5 ('NFS: Add nfs_commit_file()') Signed-off-by: Dan Carpenter Signed-off-by: Anna Schumaker --- fs/nfs/write.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 4dac51ba1f7e..54a4eb6bc462 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1735,8 +1735,8 @@ int nfs_commit_file(struct file *file, struct nfs_write_verifier *verf) open = get_nfs_open_context(nfs_file_open_context(file)); req = nfs_create_request(open, NULL, NULL, 0, i_size_read(inode)); - if (!req) { - ret = -ENOMEM; + if (IS_ERR(req)) { + ret = PTR_ERR(req); goto out_put; } -- cgit v1.2.3-59-g8ed1b From ade8febde0271513360bac44883dbebad44276c3 Mon Sep 17 00:00:00 2001 From: Weston Andros Adamson Date: Wed, 25 May 2016 10:07:23 -0400 Subject: nfs: avoid race that crashes nfs_init_commit Since the patch "NFS: Allow multiple commit requests in flight per file" we can run multiple simultaneous commits on the same inode. This introduced a race over collecting pages to commit that made it possible to call nfs_init_commit() with an empty list - which causes crashes like the one below. The fix is to catch this race and avoid calling nfs_init_commit and initiate_commit when there is no work to do. Here is the crash: [600522.076832] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040 [600522.078475] IP: [] nfs_init_commit+0x22/0x130 [nfs] [600522.078745] PGD 4272b1067 PUD 4272cb067 PMD 0 [600522.078972] Oops: 0000 [#1] SMP [600522.079204] Modules linked in: nfsv3 nfs_layout_flexfiles rpcsec_gss_krb5 nfsv4 dns_resolver nfs fscache dcdbas ip6t_rpfilter ip6t_REJECT nf_reject_ipv6 xt_conntrack ebtable_nat ebtable_broute bridge stp llc ebtable_filter ebtables ip6table_nat nf_conntrack_ipv6 nf_defrag_ipv6 nf_nat_ipv6 ip6table_mangle ip6table_security ip6table_raw ip6table_filter ip6_tables iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_mangle iptable_security iptable_raw vmw_vsock_vmci_transport vsock bonding ipmi_devintf ipmi_msghandler coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel ppdev vmw_balloon parport_pc parport acpi_cpufreq vmw_vmci i2c_piix4 shpchp nfsd auth_rpcgss nfs_acl lockd grace sunrpc xfs libcrc32c vmwgfx drm_kms_helper ttm drm crc32c_intel serio_raw vmxnet3 [600522.081380] vmw_pvscsi ata_generic pata_acpi [600522.081809] CPU: 3 PID: 15667 Comm: /usr/bin/python Not tainted 4.1.9-100.pd.88.el7.x86_64 #1 [600522.082281] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/30/2014 [600522.082814] task: ffff8800bbbfa780 ti: ffff88042ae84000 task.ti: ffff88042ae84000 [600522.083378] RIP: 0010:[] [] nfs_init_commit+0x22/0x130 [nfs] [600522.083973] RSP: 0018:ffff88042ae87438 EFLAGS: 00010246 [600522.084571] RAX: 0000000000000000 RBX: ffff880003485e40 RCX: ffff88042ae87588 [600522.085188] RDX: 0000000000000000 RSI: ffff88042ae874b0 RDI: ffff880003485e40 [600522.085756] RBP: ffff88042ae87448 R08: ffff880003486010 R09: ffff88042ae874b0 [600522.086332] R10: 0000000000000000 R11: 0000000000000005 R12: ffff88042ae872d0 [600522.086905] R13: ffff88042ae874b0 R14: ffff880003485e40 R15: ffff88042704c840 [600522.087484] FS: 00007f4728ff2740(0000) GS:ffff88043fd80000(0000) knlGS:0000000000000000 [600522.088070] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [600522.088663] CR2: 0000000000000040 CR3: 000000042b6aa000 CR4: 00000000001406e0 [600522.089327] Stack: [600522.089926] 0000000000000001 ffff88042ae87588 ffff88042ae874f8 ffffffffa04f09fa [600522.090549] 0000000000017840 0000000000017840 ffff88042ae87588 ffff8803258d9930 [600522.091169] ffff88042ae87578 ffffffffa0563d80 0000000000000000 ffff88042704c840 [600522.091789] Call Trace: [600522.092420] [] pnfs_generic_commit_pagelist+0x1da/0x320 [nfsv4] [600522.093052] [] ? ff_layout_commit_prepare_v3+0x30/0x30 [nfs_layout_flexfiles] [600522.093696] [] ff_layout_commit_pagelist+0x15/0x20 [nfs_layout_flexfiles] [600522.094359] [] nfs_generic_commit_list+0xe8/0x120 [nfs] [600522.095032] [] nfs_commit_inode+0xba/0x110 [nfs] [600522.095719] [] nfs_release_page+0x44/0xd0 [nfs] [600522.096410] [] try_to_release_page+0x32/0x50 [600522.097109] [] shrink_page_list+0x961/0xb30 [600522.097812] [] shrink_inactive_list+0x1cd/0x550 [600522.098530] [] shrink_lruvec+0x635/0x840 [600522.099250] [] shrink_zone+0xf0/0x2f0 [600522.099974] [] do_try_to_free_pages+0x192/0x470 [600522.100709] [] try_to_free_pages+0xda/0x170 [600522.101464] [] __alloc_pages_nodemask+0x588/0x970 [600522.102235] [] alloc_pages_vma+0xb5/0x230 [600522.103000] [] ? cpumask_any_but+0x39/0x50 [600522.103774] [] wp_page_copy.isra.55+0x95/0x490 [600522.104558] [] ? __wake_up+0x48/0x60 [600522.105357] [] do_wp_page+0xab/0x4f0 [600522.106137] [] ? release_task+0x36b/0x470 [600522.106902] [] ? eventfd_ctx_read+0x67/0x1c0 [600522.107659] [] handle_mm_fault+0xc78/0x1900 [600522.108431] [] __do_page_fault+0x181/0x420 [600522.109173] [] ? __audit_syscall_exit+0x1e6/0x280 [600522.109893] [] do_page_fault+0x30/0x80 [600522.110594] [] ? syscall_trace_leave+0xc6/0x120 [600522.111288] [] page_fault+0x28/0x30 [600522.111947] Code: 5d c3 0f 1f 80 00 00 00 00 0f 1f 44 00 00 55 4c 8d 87 d0 01 00 00 48 89 e5 53 48 89 fb 48 83 ec 08 4c 8b 0e 49 8b 41 18 4c 39 ce <48> 8b 40 40 4c 8b 50 30 74 24 48 8b 87 d0 01 00 00 48 8b 7e 08 [600522.113343] RIP [] nfs_init_commit+0x22/0x130 [nfs] [600522.114003] RSP [600522.114636] CR2: 0000000000000040 Fixes: af7cf057 (NFS: Allow multiple commit requests in flight per file) CC: stable@vger.kernel.org Signed-off-by: Weston Andros Adamson Signed-off-by: Anna Schumaker --- fs/nfs/pnfs_nfs.c | 28 ++++++++++++++++++++++++++++ fs/nfs/write.c | 4 ++++ 2 files changed, 32 insertions(+) (limited to 'fs') diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index d2a7c9f7aa94..0dfc476da3e1 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -246,6 +246,23 @@ void pnfs_fetch_commit_bucket_list(struct list_head *pages, } +/* Helper function for pnfs_generic_commit_pagelist to catch an empty + * page list. This can happen when two commits race. */ +static bool +pnfs_generic_commit_cancel_empty_pagelist(struct list_head *pages, + struct nfs_commit_data *data, + struct nfs_commit_info *cinfo) +{ + if (list_empty(pages)) { + if (atomic_dec_and_test(&cinfo->mds->rpcs_out)) + wake_up_atomic_t(&cinfo->mds->rpcs_out); + nfs_commitdata_release(data); + return true; + } + + return false; +} + /* This follows nfs_commit_list pretty closely */ int pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, @@ -280,6 +297,11 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, list_for_each_entry_safe(data, tmp, &list, pages) { list_del_init(&data->pages); if (data->ds_commit_index < 0) { + /* another commit raced with us */ + if (pnfs_generic_commit_cancel_empty_pagelist(mds_pages, + data, cinfo)) + continue; + nfs_init_commit(data, mds_pages, NULL, cinfo); nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(data->inode), @@ -288,6 +310,12 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, LIST_HEAD(pages); pnfs_fetch_commit_bucket_list(&pages, data, cinfo); + + /* another commit raced with us */ + if (pnfs_generic_commit_cancel_empty_pagelist(&pages, + data, cinfo)) + continue; + nfs_init_commit(data, &pages, data->lseg, cinfo); initiate_commit(data, how); } diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 54a4eb6bc462..e1c74d3db64d 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1710,6 +1710,10 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how, { struct nfs_commit_data *data; + /* another commit raced with us */ + if (list_empty(head)) + return 0; + data = nfs_commitdata_alloc(); if (!data) -- cgit v1.2.3-59-g8ed1b From fb1084e3320555eca0fb86c9eaad5429ee761282 Mon Sep 17 00:00:00 2001 From: Tom Haynes Date: Wed, 25 May 2016 07:31:12 -0700 Subject: nfs/flexfiles: Helper function to detect FF_FLAGS_NO_READ_IO The mds can inform the client not to use the IOMODE_RW layout segment for doing READs. I.e., it is basically a IOMODE_WRITE layout segment. It would do this to not interfere with the WRITEs. Signed-off-by: Tom Haynes Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.h | 11 ++++++++++- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 6 ++++++ 2 files changed, 16 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index b54058122647..1bcdb15d0c41 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -10,7 +10,8 @@ #define FS_NFS_NFS4FLEXFILELAYOUT_H #define FF_FLAGS_NO_LAYOUTCOMMIT 1 -#define FF_FLAGS_NO_IO_THRU_MDS 2 +#define FF_FLAGS_NO_IO_THRU_MDS 2 +#define FF_FLAGS_NO_READ_IO 4 #include "../pnfs.h" @@ -152,6 +153,12 @@ ff_layout_no_fallback_to_mds(struct pnfs_layout_segment *lseg) return FF_LAYOUT_LSEG(lseg)->flags & FF_FLAGS_NO_IO_THRU_MDS; } +static inline bool +ff_layout_no_read_on_rw(struct pnfs_layout_segment *lseg) +{ + return FF_LAYOUT_LSEG(lseg)->flags & FF_FLAGS_NO_READ_IO; +} + static inline bool ff_layout_test_devid_unavailable(struct nfs4_deviceid_node *node) { @@ -192,4 +199,6 @@ struct rpc_cred *ff_layout_get_ds_cred(struct pnfs_layout_segment *lseg, u32 ds_idx, struct rpc_cred *mdscred); bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg); bool ff_layout_avoid_mds_available_ds(struct pnfs_layout_segment *lseg); +bool ff_layout_avoid_read_on_rw(struct pnfs_layout_segment *lseg); + #endif /* FS_NFS_NFS4FLEXFILELAYOUT_H */ diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index 35d84d0c0f21..0aa36be71fce 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -557,6 +557,12 @@ bool ff_layout_avoid_mds_available_ds(struct pnfs_layout_segment *lseg) ff_layout_has_available_ds(lseg); } +bool ff_layout_avoid_read_on_rw(struct pnfs_layout_segment *lseg) +{ + return lseg->pls_range.iomode == IOMODE_RW && + ff_layout_no_read_on_rw(lseg); +} + module_param(dataserver_retrans, uint, 0644); MODULE_PARM_DESC(dataserver_retrans, "The number of times the NFSv4.1 client " "retries a request before it attempts further " -- cgit v1.2.3-59-g8ed1b From 602c4cd452d93355166daf8a662c84a5f85d2856 Mon Sep 17 00:00:00 2001 From: Tom Haynes Date: Wed, 25 May 2016 07:31:13 -0700 Subject: nfs/flexfiles: Use the layout segment for reading unless it a IOMODE_RW and reading is disabled Signed-off-by: Tom Haynes Signed-off-by: Anna Schumaker --- fs/nfs/flexfilelayout/flexfilelayout.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 40bccb2c06d1..a7aeb7407252 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -830,7 +830,8 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, int ds_idx; /* Use full layout for now */ - if (!pgio->pg_lseg) { + if (!pgio->pg_lseg || ff_layout_avoid_read_on_rw(pgio->pg_lseg)) { + pnfs_put_lseg(pgio->pg_lseg); pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode, req->wb_context, 0, @@ -840,9 +841,9 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); pgio->pg_lseg = NULL; - return; } } + /* If no lseg, fall back to read through mds */ if (pgio->pg_lseg == NULL) goto out_mds; -- cgit v1.2.3-59-g8ed1b From c7d73af2d249f0323f5cdb171a59497ce80011fb Mon Sep 17 00:00:00 2001 From: Tom Haynes Date: Wed, 25 May 2016 07:31:14 -0700 Subject: pnfs: pnfs_update_layout needs to consider if strict iomode checking is on As flexfiles has FF_FLAGS_NO_READ_IO, there is a need to generically support enforcing that a IOMODE_RW segment will not allow READ I/O. Signed-off-by: Tom Haynes Signed-off-by: Anna Schumaker --- fs/nfs/filelayout/filelayout.c | 2 ++ fs/nfs/flexfilelayout/flexfilelayout.c | 49 +++++++++++++++++++++++++--------- fs/nfs/pnfs.c | 34 ++++++++++++++--------- fs/nfs/pnfs.h | 1 + 4 files changed, 61 insertions(+), 25 deletions(-) (limited to 'fs') diff --git a/fs/nfs/filelayout/filelayout.c b/fs/nfs/filelayout/filelayout.c index 3e50057eeadf..aa59757389dc 100644 --- a/fs/nfs/filelayout/filelayout.c +++ b/fs/nfs/filelayout/filelayout.c @@ -890,6 +890,7 @@ filelayout_pg_init_read(struct nfs_pageio_descriptor *pgio, 0, NFS4_MAX_UINT64, IOMODE_READ, + false, GFP_KERNEL); if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); @@ -915,6 +916,7 @@ filelayout_pg_init_write(struct nfs_pageio_descriptor *pgio, 0, NFS4_MAX_UINT64, IOMODE_RW, + false, GFP_NOFS); if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index a7aeb7407252..0e8018bc9880 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -820,6 +820,36 @@ ff_layout_choose_best_ds_for_read(struct pnfs_layout_segment *lseg, return NULL; } +static void +ff_layout_pg_get_read(struct nfs_pageio_descriptor *pgio, + struct nfs_page *req, + bool strict_iomode) +{ +retry_strict: + pnfs_put_lseg(pgio->pg_lseg); + pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode, + req->wb_context, + 0, + NFS4_MAX_UINT64, + IOMODE_READ, + strict_iomode, + GFP_KERNEL); + if (IS_ERR(pgio->pg_lseg)) { + pgio->pg_error = PTR_ERR(pgio->pg_lseg); + pgio->pg_lseg = NULL; + } + + /* If we don't have checking, do get a IOMODE_RW + * segment, and the server wants to avoid READs + * there, then retry! + */ + if (pgio->pg_lseg && !strict_iomode && + ff_layout_avoid_read_on_rw(pgio->pg_lseg)) { + strict_iomode = true; + goto retry_strict; + } +} + static void ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *req) @@ -830,19 +860,10 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, int ds_idx; /* Use full layout for now */ - if (!pgio->pg_lseg || ff_layout_avoid_read_on_rw(pgio->pg_lseg)) { - pnfs_put_lseg(pgio->pg_lseg); - pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode, - req->wb_context, - 0, - NFS4_MAX_UINT64, - IOMODE_READ, - GFP_KERNEL); - if (IS_ERR(pgio->pg_lseg)) { - pgio->pg_error = PTR_ERR(pgio->pg_lseg); - pgio->pg_lseg = NULL; - } - } + if (!pgio->pg_lseg) + ff_layout_pg_get_read(pgio, req, false); + else if (ff_layout_avoid_read_on_rw(pgio->pg_lseg)) + ff_layout_pg_get_read(pgio, req, true); /* If no lseg, fall back to read through mds */ if (pgio->pg_lseg == NULL) @@ -894,6 +915,7 @@ ff_layout_pg_init_write(struct nfs_pageio_descriptor *pgio, 0, NFS4_MAX_UINT64, IOMODE_RW, + false, GFP_NOFS); if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); @@ -952,6 +974,7 @@ ff_layout_pg_get_mirror_count_write(struct nfs_pageio_descriptor *pgio, 0, NFS4_MAX_UINT64, IOMODE_RW, + false, GFP_NOFS); if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 79ae3049608d..0c7e0d45a4de 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1321,23 +1321,28 @@ out_existing: /* * iomode matching rules: - * iomode lseg match - * ----- ----- ----- - * ANY READ true - * ANY RW true - * RW READ false - * RW RW true - * READ READ true - * READ RW true + * iomode lseg strict match + * iomode + * ----- ----- ------ ----- + * ANY READ N/A true + * ANY RW N/A true + * RW READ N/A false + * RW RW N/A true + * READ READ N/A true + * READ RW true false + * READ RW false true */ static bool pnfs_lseg_range_match(const struct pnfs_layout_range *ls_range, - const struct pnfs_layout_range *range) + const struct pnfs_layout_range *range, + bool strict_iomode) { struct pnfs_layout_range range1; if ((range->iomode == IOMODE_RW && ls_range->iomode != IOMODE_RW) || + (range->iomode != ls_range->iomode && + strict_iomode == true) || !pnfs_lseg_range_intersecting(ls_range, range)) return 0; @@ -1352,7 +1357,8 @@ pnfs_lseg_range_match(const struct pnfs_layout_range *ls_range, */ static struct pnfs_layout_segment * pnfs_find_lseg(struct pnfs_layout_hdr *lo, - struct pnfs_layout_range *range) + struct pnfs_layout_range *range, + bool strict_iomode) { struct pnfs_layout_segment *lseg, *ret = NULL; @@ -1361,7 +1367,8 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo, list_for_each_entry(lseg, &lo->plh_segs, pls_list) { if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags) && !test_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags) && - pnfs_lseg_range_match(&lseg->pls_range, range)) { + pnfs_lseg_range_match(&lseg->pls_range, range, + strict_iomode)) { ret = pnfs_get_lseg(lseg); break; } @@ -1478,6 +1485,7 @@ pnfs_update_layout(struct inode *ino, loff_t pos, u64 count, enum pnfs_iomode iomode, + bool strict_iomode, gfp_t gfp_flags) { struct pnfs_layout_range arg = { @@ -1539,7 +1547,7 @@ lookup_again: goto out_unlock; } - lseg = pnfs_find_lseg(lo, &arg); + lseg = pnfs_find_lseg(lo, &arg, strict_iomode); if (lseg) { trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_FOUND_CACHED); @@ -1883,6 +1891,7 @@ pnfs_generic_pg_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *r req_offset(req), rd_size, IOMODE_READ, + false, GFP_KERNEL); if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); @@ -1907,6 +1916,7 @@ pnfs_generic_pg_init_write(struct nfs_pageio_descriptor *pgio, req_offset(req), wb_size, IOMODE_RW, + false, GFP_NOFS); if (IS_ERR(pgio->pg_lseg)) { pgio->pg_error = PTR_ERR(pgio->pg_lseg); diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index f9f3331bef49..b21bd0bee784 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -288,6 +288,7 @@ struct pnfs_layout_segment *pnfs_update_layout(struct inode *ino, loff_t pos, u64 count, enum pnfs_iomode iomode, + bool strict_iomode, gfp_t gfp_flags); void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo); -- cgit v1.2.3-59-g8ed1b