From 9a4bf31d05a801e2358d96f69b39fb8ce2c69dd8 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Thu, 10 Dec 2015 10:41:58 -0500 Subject: nfs: add new tracepoint for pnfs_update_layout pnfs_update_layout is really the "nexus" of layout handling. If it returns NULL then we end up going through the MDS. This patch adds some tracepoints to that function that allow us to determine the cause when we end up going through the MDS unexpectedly. Signed-off-by: Jeff Layton Signed-off-by: Trond Myklebust --- include/linux/nfs4.h | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'include') diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index e7e7853..0e30f2c 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -592,4 +592,18 @@ enum data_content4 { NFS4_CONTENT_HOLE = 1, }; +enum pnfs_update_layout_reason { + PNFS_UPDATE_LAYOUT_UNKNOWN = 0, + PNFS_UPDATE_LAYOUT_NO_PNFS, + PNFS_UPDATE_LAYOUT_RD_ZEROLEN, + PNFS_UPDATE_LAYOUT_MDSTHRESH, + PNFS_UPDATE_LAYOUT_NOMEM, + PNFS_UPDATE_LAYOUT_BULK_RECALL, + PNFS_UPDATE_LAYOUT_IO_TEST_FAIL, + PNFS_UPDATE_LAYOUT_FOUND_CACHED, + PNFS_UPDATE_LAYOUT_RETURN, + PNFS_UPDATE_LAYOUT_BLOCKED, + PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET, +}; + #endif -- cgit v1.1 From 99ade3c71b1e40e7174d6527709399a87f3d05e0 Mon Sep 17 00:00:00 2001 From: Andrew Elble Date: Wed, 2 Dec 2015 09:39:51 -0500 Subject: nfs: machine credential support for additional operations Allow LAYOUTRETURN and DELEGRETURN to use machine credentials if the server supports it. Add request for OPEN_DOWNGRADE as the close path also uses that. Signed-off-by: Andrew Elble Signed-off-by: Trond Myklebust --- include/linux/nfs_fs_sb.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 2469ab0..7fcc13c 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -102,6 +102,7 @@ struct nfs_client { #define NFS_SP4_MACH_CRED_STATEID 4 /* TEST_STATEID and FREE_STATEID */ #define NFS_SP4_MACH_CRED_WRITE 5 /* WRITE */ #define NFS_SP4_MACH_CRED_COMMIT 6 /* COMMIT */ +#define NFS_SP4_MACH_CRED_PNFS_CLEANUP 7 /* LAYOUTRETURN */ #endif /* CONFIG_NFS_V4 */ /* Our own IP address, as a null-terminated string. -- cgit v1.1 From d6c843b96e1cb5199147e3281a724e3c0b69a9ab Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Sat, 5 Dec 2015 16:20:43 +0800 Subject: nfs: only remove page from mapping if launder_page fails Instead of dropping pages when write fails, only do it when we get fatal failure in launder_page write back. Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- include/linux/nfs_fs.h | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index c0e9614..b88fc46 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -517,13 +517,25 @@ extern int nfs_updatepage(struct file *, struct page *, unsigned int, unsigned */ extern int nfs_sync_inode(struct inode *inode); extern int nfs_wb_all(struct inode *inode); -extern int nfs_wb_page(struct inode *inode, struct page* page); +extern int nfs_wb_single_page(struct inode *inode, struct page *page, bool launder); extern int nfs_wb_page_cancel(struct inode *inode, struct page* page); extern int nfs_commit_inode(struct inode *, int); extern struct nfs_commit_data *nfs_commitdata_alloc(void); extern void nfs_commit_free(struct nfs_commit_data *data); static inline int +nfs_wb_launder_page(struct inode *inode, struct page *page) +{ + return nfs_wb_single_page(inode, page, true); +} + +static inline int +nfs_wb_page(struct inode *inode, struct page *page) +{ + return nfs_wb_single_page(inode, page, false); +} + +static inline int nfs_have_writebacks(struct inode *inode) { return NFS_I(inode)->nrequests != 0; -- cgit v1.1 From 37e9ed22b1552fa94ee7db2901a5e7d8bdf60b15 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 22 Dec 2015 12:30:24 -0500 Subject: pNFS: Add flag to track if we've called nfs4_ff_layout_stat_io_start_read/write Signed-off-by: Trond Myklebust --- include/linux/nfs_xdr.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include') diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 11bbae4..7b30ac0 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1375,6 +1375,7 @@ enum { NFS_IOHDR_ERROR = 0, NFS_IOHDR_EOF, NFS_IOHDR_REDO, + NFS_IOHDR_STAT, }; struct nfs_pgio_header { @@ -1454,6 +1455,7 @@ struct nfs_commit_data { const struct rpc_call_ops *mds_ops; const struct nfs_commit_completion_ops *completion_ops; int (*commit_done_cb) (struct rpc_task *task, struct nfs_commit_data *data); + unsigned long flags; }; struct nfs_pgio_completion_ops { -- cgit v1.1 From dc602dd706cb64036132a7903ead1c67d9a7bcb9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 31 Dec 2015 11:44:06 -0500 Subject: NFS/pNFS: Fix up pNFS write reschedule layering violations and bugs The flexfiles layout in particular, seems to want to poke around in the O_DIRECT flags when retransmitting. This patch sets up an interface to allow it to call back into O_DIRECT to handle retransmission correctly. It also fixes a potential bug whereby we could change the behaviour of O_DIRECT if an error is already pending. Signed-off-by: Trond Myklebust --- include/linux/nfs_xdr.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 11bbae4..e89dbb1 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1460,6 +1460,7 @@ struct nfs_pgio_completion_ops { void (*error_cleanup)(struct list_head *head); void (*init_hdr)(struct nfs_pgio_header *hdr); void (*completion)(struct nfs_pgio_header *hdr); + void (*reschedule_io)(struct nfs_pgio_header *hdr); }; struct nfs_unlinkdata { -- cgit v1.1 From af7cf057933f01dc7f33ddfb5e436ad598ed17ad Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 29 Sep 2015 20:34:05 -0400 Subject: NFS: Allow multiple commit requests in flight per file Allow synchronous RPC calls to wait for pending RPC calls to finish, but also allow asynchronous ones to just fire off another commit. With this patch, the xfstests generic/074 test completes in 226s instead of 242s Signed-off-by: Trond Myklebust --- include/linux/nfs_fs.h | 1 - include/linux/nfs_xdr.h | 1 - 2 files changed, 2 deletions(-) (limited to 'include') diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index c0e9614..ebf0bd7 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -216,7 +216,6 @@ struct nfs_inode { #define NFS_INO_FLUSHING (4) /* inode is flushing out data */ #define NFS_INO_FSCACHE (5) /* inode can be cached by FS-Cache */ #define NFS_INO_FSCACHE_LOCK (6) /* FS-Cache cookie management lock */ -#define NFS_INO_COMMIT (7) /* inode is committing unstable writes */ #define NFS_INO_LAYOUTCOMMIT (9) /* layoutcommit required */ #define NFS_INO_LAYOUTCOMMITTING (10) /* layoutcommit inflight */ #define NFS_INO_LAYOUTSTATS (11) /* layoutstats inflight */ diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index e89dbb1..a8905b7 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1423,7 +1423,6 @@ struct nfs_mds_commit_info { struct nfs_commit_data; struct nfs_inode; struct nfs_commit_completion_ops { - void (*error_cleanup) (struct nfs_inode *nfsi); void (*completion) (struct nfs_commit_data *data); }; -- cgit v1.1 From b20135d0b2431900a3a5395970ffb7e4f3767c8b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 31 Dec 2015 09:28:06 -0500 Subject: NFSv4.1/pNFS: Don't queue up a new commit if the layout segment is invalid If the layout segment is invalid, then we should not be adding more write requests to the commit list. Instead, those writes should be replayed after requesting a new layout. Signed-off-by: Trond Myklebust --- include/linux/nfs_xdr.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include') diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index a8905b7..bee3e60 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1420,10 +1420,12 @@ struct nfs_mds_commit_info { struct list_head list; }; +struct nfs_commit_info; struct nfs_commit_data; struct nfs_inode; struct nfs_commit_completion_ops { void (*completion) (struct nfs_commit_data *data); + void (*resched_write) (struct nfs_commit_info *, struct nfs_page *); }; struct nfs_commit_info { -- cgit v1.1 From 210c7c1750fdf769647d1d526c9ea34c412c9eee Mon Sep 17 00:00:00 2001 From: Benjamin Coddington Date: Wed, 6 Jan 2016 10:40:18 -0500 Subject: NFS: Use wait_on_atomic_t() for unlock after readahead The use of wait_on_atomic_t() for waiting on I/O to complete before unlocking allows us to git rid of the NFS_IO_INPROGRESS flag, and thus the nfs_iocounter's flags member, and finally the nfs_iocounter altogether. The count of I/O is moved to the lock context, and the counter increment/decrement functions become simple enough to open-code. Signed-off-by: Benjamin Coddington [Trond: Fix up conflict with existing function nfs_wait_atomic_killable()] Signed-off-by: Trond Myklebust --- include/linux/nfs_fs.h | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'include') diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 9eee972..196aace 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -60,18 +60,12 @@ struct nfs_lockowner { pid_t l_pid; }; -#define NFS_IO_INPROGRESS 0 -struct nfs_io_counter { - unsigned long flags; - atomic_t io_count; -}; - struct nfs_lock_context { atomic_t count; struct list_head list; struct nfs_open_context *open_context; struct nfs_lockowner lockowner; - struct nfs_io_counter io_count; + atomic_t io_count; }; struct nfs4_state; -- cgit v1.1