Return-Path: Received: from mail-it0-f66.google.com ([209.85.214.66]:35604 "EHLO mail-it0-f66.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932422AbcGFWa4 (ORCPT ); Wed, 6 Jul 2016 18:30:56 -0400 Received: by mail-it0-f66.google.com with SMTP id g4so900487ith.2 for ; Wed, 06 Jul 2016 15:30:56 -0700 (PDT) From: Trond Myklebust To: linux-nfs@vger.kernel.org Subject: [PATCH v4 19/28] NFS: Do not serialise O_DIRECT reads and writes Date: Wed, 6 Jul 2016 18:29:56 -0400 Message-Id: <1467844205-76852-20-git-send-email-trond.myklebust@primarydata.com> In-Reply-To: <1467844205-76852-19-git-send-email-trond.myklebust@primarydata.com> References: <1467844205-76852-1-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-2-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-3-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-4-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-5-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-6-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-7-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-8-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-9-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-10-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-11-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-12-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-13-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-14-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-15-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-16-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-17-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-18-git-send-email-trond.myklebust@primarydata.com> <1467844205-76852-19-git-send-email-trond.myklebust@primarydata.com> Sender: linux-nfs-owner@vger.kernel.org List-ID: Allow dio requests to be scheduled in parallel, but ensuring that they do not conflict with buffered I/O. Signed-off-by: Trond Myklebust --- fs/nfs/Makefile | 2 +- fs/nfs/direct.c | 41 +++----------- fs/nfs/file.c | 12 ++-- fs/nfs/internal.h | 8 +++ fs/nfs/io.c | 147 +++++++++++++++++++++++++++++++++++++++++++++++++ include/linux/nfs_fs.h | 1 + 6 files changed, 174 insertions(+), 37 deletions(-) create mode 100644 fs/nfs/io.c diff --git a/fs/nfs/Makefile b/fs/nfs/Makefile index 8664417955a2..6abdda209642 100644 --- a/fs/nfs/Makefile +++ b/fs/nfs/Makefile @@ -6,7 +6,7 @@ obj-$(CONFIG_NFS_FS) += nfs.o CFLAGS_nfstrace.o += -I$(src) nfs-y := client.o dir.o file.o getroot.o inode.o super.o \ - direct.o pagelist.o read.o symlink.o unlink.o \ + io.o direct.o pagelist.o read.o symlink.o unlink.o \ write.o namespace.o mount_clnt.o nfstrace.o nfs-$(CONFIG_ROOT_NFS) += nfsroot.o nfs-$(CONFIG_SYSCTL) += sysctl.o diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 0169eca8eb42..6d0e88096440 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -578,17 +578,12 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter) if (!count) goto out; - inode_lock(inode); - result = nfs_sync_mapping(mapping); - if (result) - goto out_unlock; - task_io_account_read(count); result = -ENOMEM; dreq = nfs_direct_req_alloc(); if (dreq == NULL) - goto out_unlock; + goto out; dreq->inode = inode; dreq->bytes_left = dreq->max_count = count; @@ -603,10 +598,12 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter) if (!is_sync_kiocb(iocb)) dreq->iocb = iocb; + nfs_start_io_direct(inode); + NFS_I(inode)->read_io += count; result = nfs_direct_read_schedule_iovec(dreq, iter, iocb->ki_pos); - inode_unlock(inode); + nfs_end_io_direct(inode); if (!result) { result = nfs_direct_wait(dreq); @@ -614,13 +611,8 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter) iocb->ki_pos += result; } - nfs_direct_req_release(dreq); - return result; - out_release: nfs_direct_req_release(dreq); -out_unlock: - inode_unlock(inode); out: return result; } @@ -1008,25 +1000,12 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) pos = iocb->ki_pos; end = (pos + iov_iter_count(iter) - 1) >> PAGE_SHIFT; - inode_lock(inode); - - result = nfs_sync_mapping(mapping); - if (result) - goto out_unlock; - - if (mapping->nrpages) { - result = invalidate_inode_pages2_range(mapping, - pos >> PAGE_SHIFT, end); - if (result) - goto out_unlock; - } - task_io_account_write(count); result = -ENOMEM; dreq = nfs_direct_req_alloc(); if (!dreq) - goto out_unlock; + goto out; dreq->inode = inode; dreq->bytes_left = dreq->max_count = count; @@ -1041,6 +1020,8 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) if (!is_sync_kiocb(iocb)) dreq->iocb = iocb; + nfs_start_io_direct(inode); + result = nfs_direct_write_schedule_iovec(dreq, iter, pos); if (mapping->nrpages) { @@ -1048,7 +1029,7 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) pos >> PAGE_SHIFT, end); } - inode_unlock(inode); + nfs_end_io_direct(inode); if (!result) { result = nfs_direct_wait(dreq); @@ -1058,13 +1039,9 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) generic_write_sync(iocb, result); } } - nfs_direct_req_release(dreq); - return result; - out_release: nfs_direct_req_release(dreq); -out_unlock: - inode_unlock(inode); +out: return result; } diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 46cf0afe3c0f..9f8da9e1b23f 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -170,12 +170,14 @@ nfs_file_read(struct kiocb *iocb, struct iov_iter *to) iocb->ki_filp, iov_iter_count(to), (unsigned long) iocb->ki_pos); - result = nfs_revalidate_mapping_protected(inode, iocb->ki_filp->f_mapping); + nfs_start_io_read(inode); + result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping); if (!result) { result = generic_file_read_iter(iocb, to); if (result > 0) nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result); } + nfs_end_io_read(inode); return result; } EXPORT_SYMBOL_GPL(nfs_file_read); @@ -191,12 +193,14 @@ nfs_file_splice_read(struct file *filp, loff_t *ppos, dprintk("NFS: splice_read(%pD2, %lu@%Lu)\n", filp, (unsigned long) count, (unsigned long long) *ppos); - res = nfs_revalidate_mapping_protected(inode, filp->f_mapping); + nfs_start_io_read(inode); + res = nfs_revalidate_mapping(inode, filp->f_mapping); if (!res) { res = generic_file_splice_read(filp, ppos, pipe, count, flags); if (res > 0) nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, res); } + nfs_end_io_read(inode); return res; } EXPORT_SYMBOL_GPL(nfs_file_splice_read); @@ -645,14 +649,14 @@ ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) goto out; } - inode_lock(inode); + nfs_start_io_write(inode); result = generic_write_checks(iocb, from); if (result > 0) { current->backing_dev_info = inode_to_bdi(inode); result = generic_perform_write(file, from, iocb->ki_pos); current->backing_dev_info = NULL; } - inode_unlock(inode); + nfs_end_io_write(inode); if (result <= 0) goto out; diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 0eb5c924886d..159b64ede82a 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -411,6 +411,14 @@ extern void __exit unregister_nfs_fs(void); extern bool nfs_sb_active(struct super_block *sb); extern void nfs_sb_deactive(struct super_block *sb); +/* io.c */ +extern void nfs_start_io_read(struct inode *inode); +extern void nfs_end_io_read(struct inode *inode); +extern void nfs_start_io_write(struct inode *inode); +extern void nfs_end_io_write(struct inode *inode); +extern void nfs_start_io_direct(struct inode *inode); +extern void nfs_end_io_direct(struct inode *inode); + /* namespace.c */ #define NFS_PATH_CANONICAL 1 extern char *nfs_path(char **p, struct dentry *dentry, diff --git a/fs/nfs/io.c b/fs/nfs/io.c new file mode 100644 index 000000000000..1fc5d1ce327e --- /dev/null +++ b/fs/nfs/io.c @@ -0,0 +1,147 @@ +/* + * Copyright (c) 2016 Trond Myklebust + * + * I/O and data path helper functionality. + */ + +#include +#include +#include +#include +#include +#include + +#include "internal.h" + +/* Call with exclusively locked inode->i_rwsem */ +static void nfs_block_o_direct(struct nfs_inode *nfsi, struct inode *inode) +{ + if (test_bit(NFS_INO_ODIRECT, &nfsi->flags)) { + clear_bit(NFS_INO_ODIRECT, &nfsi->flags); + inode_dio_wait(inode); + } +} + +/** + * nfs_start_io_read - declare the file is being used for buffered reads + * @inode - file inode + * + * Declare that a buffered read operation is about to start, and ensure + * that we block all direct I/O. + * On exit, the function ensures that the NFS_INO_ODIRECT flag is unset, + * and holds a shared lock on inode->i_rwsem to ensure that the flag + * cannot be changed. + * In practice, this means that buffered read operations are allowed to + * execute in parallel, thanks to the shared lock, whereas direct I/O + * operations need to wait to grab an exclusive lock in order to set + * NFS_INO_ODIRECT. + * Note that buffered writes and truncates both take a write lock on + * inode->i_rwsem, meaning that those are serialised w.r.t. the reads. + */ +void +nfs_start_io_read(struct inode *inode) +{ + struct nfs_inode *nfsi = NFS_I(inode); + /* Be an optimist! */ + down_read(&inode->i_rwsem); + if (test_bit(NFS_INO_ODIRECT, &nfsi->flags) == 0) + return; + up_read(&inode->i_rwsem); + /* Slow path.... */ + down_write(&inode->i_rwsem); + nfs_block_o_direct(nfsi, inode); + downgrade_write(&inode->i_rwsem); +} + +/** + * nfs_end_io_read - declare that the buffered read operation is done + * @inode - file inode + * + * Declare that a buffered read operation is done, and release the shared + * lock on inode->i_rwsem. + */ +void +nfs_end_io_read(struct inode *inode) +{ + up_read(&inode->i_rwsem); +} + +/** + * nfs_start_io_write - declare the file is being used for buffered writes + * @inode - file inode + * + * Declare that a buffered read operation is about to start, and ensure + * that we block all direct I/O. + */ +void +nfs_start_io_write(struct inode *inode) +{ + down_write(&inode->i_rwsem); + nfs_block_o_direct(NFS_I(inode), inode); +} + +/** + * nfs_end_io_write - declare that the buffered write operation is done + * @inode - file inode + * + * Declare that a buffered write operation is done, and release the + * lock on inode->i_rwsem. + */ +void +nfs_end_io_write(struct inode *inode) +{ + up_write(&inode->i_rwsem); +} + +/* Call with exclusively locked inode->i_rwsem */ +static void nfs_block_buffered(struct nfs_inode *nfsi, struct inode *inode) +{ + if (!test_bit(NFS_INO_ODIRECT, &nfsi->flags)) { + set_bit(NFS_INO_ODIRECT, &nfsi->flags); + nfs_wb_all(inode); + } +} + +/** + * nfs_end_io_direct - declare the file is being used for direct i/o + * @inode - file inode + * + * Declare that a direct I/O operation is about to start, and ensure + * that we block all buffered I/O. + * On exit, the function ensures that the NFS_INO_ODIRECT flag is set, + * and holds a shared lock on inode->i_rwsem to ensure that the flag + * cannot be changed. + * In practice, this means that direct I/O operations are allowed to + * execute in parallel, thanks to the shared lock, whereas buffered I/O + * operations need to wait to grab an exclusive lock in order to clear + * NFS_INO_ODIRECT. + * Note that buffered writes and truncates both take a write lock on + * inode->i_rwsem, meaning that those are serialised w.r.t. O_DIRECT. + */ +void +nfs_start_io_direct(struct inode *inode) +{ + struct nfs_inode *nfsi = NFS_I(inode); + /* Be an optimist! */ + down_read(&inode->i_rwsem); + if (test_bit(NFS_INO_ODIRECT, &nfsi->flags) != 0) + return; + up_read(&inode->i_rwsem); + /* Slow path.... */ + down_write(&inode->i_rwsem); + nfs_block_buffered(nfsi, inode); + downgrade_write(&inode->i_rwsem); +} + +/** + * nfs_end_io_direct - declare that the direct i/o operation is done + * @inode - file inode + * + * Declare that a direct I/O operation is done, and release the shared + * lock on inode->i_rwsem. + */ +void +nfs_end_io_direct(struct inode *inode) +{ + up_read(&inode->i_rwsem); +} diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 120dd04b553c..225d17d35277 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -210,6 +210,7 @@ struct nfs_inode { #define NFS_INO_LAYOUTCOMMIT (9) /* layoutcommit required */ #define NFS_INO_LAYOUTCOMMITTING (10) /* layoutcommit inflight */ #define NFS_INO_LAYOUTSTATS (11) /* layoutstats inflight */ +#define NFS_INO_ODIRECT (12) /* I/O setting is O_DIRECT */ static inline struct nfs_inode *NFS_I(const struct inode *inode) { -- 2.7.4