From 954998b60caa8f2a3bf3abe490de6f08d283687a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:37 -0400 Subject: NFS: Fix error handling for O_DIRECT write scheduling If we fail to schedule a request for transmission, there are 2 possibilities: 1) Either we hit a fatal error, and we just want to drop the remaining requests on the floor. 2) We were asked to try again, in which case we should allow the outstanding RPC calls to complete, so that we can recoalesce requests and try again. Fixes: d600ad1f2bdb ("NFS41: pop some layoutget errors to application") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 62 ++++++++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 46 insertions(+), 16 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 47d892a1d363..ee88f0a6e7b8 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -528,10 +528,9 @@ nfs_direct_write_scan_commit_list(struct inode *inode, static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) { struct nfs_pageio_descriptor desc; - struct nfs_page *req, *tmp; + struct nfs_page *req; LIST_HEAD(reqs); struct nfs_commit_info cinfo; - LIST_HEAD(failed); nfs_init_cinfo_from_dreq(&cinfo, dreq); nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); @@ -549,27 +548,36 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) &nfs_direct_write_completion_ops); desc.pg_dreq = dreq; - list_for_each_entry_safe(req, tmp, &reqs, wb_list) { + while (!list_empty(&reqs)) { + req = nfs_list_entry(reqs.next); /* Bump the transmission count */ req->wb_nio++; if (!nfs_pageio_add_request(&desc, req)) { - nfs_list_move_request(req, &failed); spin_lock(&cinfo.inode->i_lock); - dreq->flags = 0; - if (desc.pg_error < 0) + if (dreq->error < 0) { + desc.pg_error = dreq->error; + } else if (desc.pg_error != -EAGAIN) { + dreq->flags = 0; + if (!desc.pg_error) + desc.pg_error = -EIO; dreq->error = desc.pg_error; - else - dreq->error = -EIO; + } else + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; spin_unlock(&cinfo.inode->i_lock); + break; } nfs_release_request(req); } nfs_pageio_complete(&desc); - while (!list_empty(&failed)) { - req = nfs_list_entry(failed.next); + while (!list_empty(&reqs)) { + req = nfs_list_entry(reqs.next); nfs_list_remove_request(req); nfs_unlock_and_release_request(req); + if (desc.pg_error == -EAGAIN) + nfs_mark_request_commit(req, NULL, &cinfo, 0); + else + nfs_release_request(req); } if (put_dreq(dreq)) @@ -794,9 +802,11 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, { struct nfs_pageio_descriptor desc; struct inode *inode = dreq->inode; + struct nfs_commit_info cinfo; ssize_t result = 0; size_t requested_bytes = 0; size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE); + bool defer = false; trace_nfs_direct_write_schedule_iovec(dreq); @@ -837,17 +847,37 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, break; } - nfs_lock_request(req); - if (!nfs_pageio_add_request(&desc, req)) { - result = desc.pg_error; - nfs_unlock_and_release_request(req); - break; - } pgbase = 0; bytes -= req_len; requested_bytes += req_len; pos += req_len; dreq->bytes_left -= req_len; + + if (defer) { + nfs_mark_request_commit(req, NULL, &cinfo, 0); + continue; + } + + nfs_lock_request(req); + if (nfs_pageio_add_request(&desc, req)) + continue; + + /* Exit on hard errors */ + if (desc.pg_error < 0 && desc.pg_error != -EAGAIN) { + result = desc.pg_error; + nfs_unlock_and_release_request(req); + break; + } + + /* If the error is soft, defer remaining requests */ + nfs_init_cinfo_from_dreq(&cinfo, dreq); + spin_lock(&cinfo.inode->i_lock); + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; + spin_unlock(&cinfo.inode->i_lock); + nfs_unlock_request(req); + nfs_mark_request_commit(req, NULL, &cinfo, 0); + desc.pg_error = 0; + defer = true; } nfs_direct_release_pages(pagevec, npages); kvfree(pagevec); -- cgit v1.2.3 From 7c6339322ce0c6128acbe36aacc1eeb986dd7bf1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:38 -0400 Subject: NFS: Fix O_DIRECT locking issues The dreq fields are protected by the dreq->lock. Fixes: 954998b60caa ("NFS: Fix error handling for O_DIRECT write scheduling") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index ee88f0a6e7b8..e8a1645857dd 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -553,7 +553,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) /* Bump the transmission count */ req->wb_nio++; if (!nfs_pageio_add_request(&desc, req)) { - spin_lock(&cinfo.inode->i_lock); + spin_lock(&dreq->lock); if (dreq->error < 0) { desc.pg_error = dreq->error; } else if (desc.pg_error != -EAGAIN) { @@ -563,7 +563,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) dreq->error = desc.pg_error; } else dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - spin_unlock(&cinfo.inode->i_lock); + spin_unlock(&dreq->lock); break; } nfs_release_request(req); @@ -871,9 +871,9 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, /* If the error is soft, defer remaining requests */ nfs_init_cinfo_from_dreq(&cinfo, dreq); - spin_lock(&cinfo.inode->i_lock); + spin_lock(&dreq->lock); dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - spin_unlock(&cinfo.inode->i_lock); + spin_unlock(&dreq->lock); nfs_unlock_request(req); nfs_mark_request_commit(req, NULL, &cinfo, 0); desc.pg_error = 0; -- cgit v1.2.3 From 8982f7aff39fb526aba4441fff2525fcedd5e1a3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:39 -0400 Subject: NFS: More O_DIRECT accounting fixes for error paths If we hit a fatal error when retransmitting, we do need to record the removal of the request from the count of written bytes. Fixes: 031d73ed768a ("NFS: Fix O_DIRECT accounting of number of bytes read/written") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 47 +++++++++++++++++++++++++++++++---------------- 1 file changed, 31 insertions(+), 16 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index e8a1645857dd..a53e50123499 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -93,12 +93,10 @@ nfs_direct_handle_truncated(struct nfs_direct_req *dreq, dreq->max_count = dreq_len; if (dreq->count > dreq_len) dreq->count = dreq_len; - - if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) - dreq->error = hdr->error; - else /* Clear outstanding error if this is EOF */ - dreq->error = 0; } + + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && !dreq->error) + dreq->error = hdr->error; } static void @@ -120,6 +118,18 @@ nfs_direct_count_bytes(struct nfs_direct_req *dreq, dreq->count = dreq_len; } +static void nfs_direct_truncate_request(struct nfs_direct_req *dreq, + struct nfs_page *req) +{ + loff_t offs = req_offset(req); + size_t req_start = (size_t)(offs - dreq->io_start); + + if (req_start < dreq->max_count) + dreq->max_count = req_start; + if (req_start < dreq->count) + dreq->count = req_start; +} + /** * nfs_swap_rw - NFS address space operation for swap I/O * @iocb: target I/O control block @@ -537,10 +547,6 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) nfs_direct_join_group(&reqs, dreq->inode); - dreq->count = 0; - dreq->max_count = 0; - list_for_each_entry(req, &reqs, wb_list) - dreq->max_count += req->wb_bytes; nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo); get_dreq(dreq); @@ -574,10 +580,14 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) req = nfs_list_entry(reqs.next); nfs_list_remove_request(req); nfs_unlock_and_release_request(req); - if (desc.pg_error == -EAGAIN) + if (desc.pg_error == -EAGAIN) { nfs_mark_request_commit(req, NULL, &cinfo, 0); - else + } else { + spin_lock(&dreq->lock); + nfs_direct_truncate_request(dreq, req); + spin_unlock(&dreq->lock); nfs_release_request(req); + } } if (put_dreq(dreq)) @@ -597,8 +607,6 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) if (status < 0) { /* Errors in commit are fatal */ dreq->error = status; - dreq->max_count = 0; - dreq->count = 0; dreq->flags = NFS_ODIRECT_DONE; } else { status = dreq->error; @@ -609,7 +617,12 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) while (!list_empty(&data->pages)) { req = nfs_list_entry(data->pages.next); nfs_list_remove_request(req); - if (status >= 0 && !nfs_write_match_verf(verf, req)) { + if (status < 0) { + spin_lock(&dreq->lock); + nfs_direct_truncate_request(dreq, req); + spin_unlock(&dreq->lock); + nfs_release_request(req); + } else if (!nfs_write_match_verf(verf, req)) { dreq->flags = NFS_ODIRECT_RESCHED_WRITES; /* * Despite the reboot, the write was successful, @@ -617,7 +630,7 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data) */ req->wb_nio = 0; nfs_mark_request_commit(req, NULL, &cinfo, 0); - } else /* Error or match */ + } else nfs_release_request(req); nfs_unlock_and_release_request(req); } @@ -670,6 +683,7 @@ static void nfs_direct_write_clear_reqs(struct nfs_direct_req *dreq) while (!list_empty(&reqs)) { req = nfs_list_entry(reqs.next); nfs_list_remove_request(req); + nfs_direct_truncate_request(dreq, req); nfs_release_request(req); nfs_unlock_and_release_request(req); } @@ -719,7 +733,8 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) } nfs_direct_count_bytes(dreq, hdr); - if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags)) { + if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags) && + !test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { if (!dreq->flags) dreq->flags = NFS_ODIRECT_DO_COMMIT; flags = dreq->flags; -- cgit v1.2.3 From b193a78ddb5ee7dba074d3f28dc050069ba083c0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:40 -0400 Subject: NFS: Use the correct commit info in nfs_join_page_group() Ensure that nfs_clear_request_commit() updates the correct counters when it removes them from the commit list. Fixes: ed5d588fe47f ("NFS: Try to join page groups before an O_DIRECT retransmission") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index a53e50123499..3391c8b97da5 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -498,7 +498,9 @@ static void nfs_direct_add_page_head(struct list_head *list, kref_get(&head->wb_kref); } -static void nfs_direct_join_group(struct list_head *list, struct inode *inode) +static void nfs_direct_join_group(struct list_head *list, + struct nfs_commit_info *cinfo, + struct inode *inode) { struct nfs_page *req, *subreq; @@ -520,7 +522,7 @@ static void nfs_direct_join_group(struct list_head *list, struct inode *inode) nfs_release_request(subreq); } } while ((subreq = subreq->wb_this_page) != req); - nfs_join_page_group(req, inode); + nfs_join_page_group(req, cinfo, inode); } } @@ -545,7 +547,7 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) nfs_init_cinfo_from_dreq(&cinfo, dreq); nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); - nfs_direct_join_group(&reqs, dreq->inode); + nfs_direct_join_group(&reqs, &cinfo, dreq->inode); nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo); get_dreq(dreq); -- cgit v1.2.3 From b11243f720ee5f9376861099019c8542969b6318 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Sep 2023 12:34:41 -0400 Subject: NFS: More fixes for nfs_direct_write_reschedule_io() Ensure that all requests are put back onto the commit list so that they can be rescheduled. Fixes: 4daaeba93822 ("NFS: Fix nfs_direct_write_reschedule_io()") Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/direct.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) (limited to 'fs/nfs/direct.c') diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 3391c8b97da5..f6c74f424691 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -780,18 +780,23 @@ static void nfs_write_sync_pgio_error(struct list_head *head, int error) static void nfs_direct_write_reschedule_io(struct nfs_pgio_header *hdr) { struct nfs_direct_req *dreq = hdr->dreq; + struct nfs_page *req; + struct nfs_commit_info cinfo; trace_nfs_direct_write_reschedule_io(dreq); + nfs_init_cinfo_from_dreq(&cinfo, dreq); spin_lock(&dreq->lock); - if (dreq->error == 0) { + if (dreq->error == 0) dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - /* fake unstable write to let common nfs resend pages */ - hdr->verf.committed = NFS_UNSTABLE; - hdr->good_bytes = hdr->args.offset + hdr->args.count - - hdr->io_start; - } + set_bit(NFS_IOHDR_REDO, &hdr->flags); spin_unlock(&dreq->lock); + while (!list_empty(&hdr->pages)) { + req = nfs_list_entry(hdr->pages.next); + nfs_list_remove_request(req); + nfs_unlock_request(req); + nfs_mark_request_commit(req, NULL, &cinfo, 0); + } } static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = { -- cgit v1.2.3