[54/67] afs: Add O_DIRECT read support

Message ID	163456949634.2614702.977164796285038310.stgit@warthog.procyon.org.uk (mailing list archive)
State	New, archived
Headers	show Return-Path: <linux-cifs-owner@kernel.org> Organization: Red Hat UK Ltd. Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SI4 1TE, United Kingdom. Registered in England and Wales under Company Registration No. 3798903 Subject: [PATCH 54/67] afs: Add O_DIRECT read support From: David Howells <dhowells@redhat.com> To: linux-cachefs@redhat.com Cc: Marc Dionne <marc.dionne@auristor.com>, linux-afs@lists.infradead.org, dhowells@redhat.com, Trond Myklebust <trondmy@hammerspace.com>, Anna Schumaker <anna.schumaker@netapp.com>, Steve French <sfrench@samba.org>, Dominique Martinet <asmadeus@codewreck.org>, Jeff Layton <jlayton@redhat.com>, Matthew Wilcox <willy@infradead.org>, Alexander Viro <viro@zeniv.linux.org.uk>, Omar Sandoval <osandov@osandov.com>, Linus Torvalds <torvalds@linux-foundation.org>, linux-afs@lists.infradead.org, linux-nfs@vger.kernel.org, linux-cifs@vger.kernel.org, ceph-devel@vger.kernel.org, v9fs-developer@lists.sourceforge.net, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org Date: Mon, 18 Oct 2021 16:04:56 +0100 Message-ID: <163456949634.2614702.977164796285038310.stgit@warthog.procyon.org.uk> In-Reply-To: <163456861570.2614702.14754548462706508617.stgit@warthog.procyon.org.uk> References: <163456861570.2614702.14754548462706508617.stgit@warthog.procyon.org.uk> User-Agent: StGit/0.23 MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Precedence: bulk
Series	fscache: Rewrite index API and management system \| expand [00/67] fscache: Rewrite index API and management system [01/67] mm: Stop filemap_read() from grabbing a superfluous page [02/67] vfs: Provide S_KERNEL_FILE inode flag [03/67] vfs, fscache: Force ->write_inode() to occur if cookie pinned for writeback [04/67] afs: Handle len being extending over page end in write_begin/write_end [05/67] afs: Fix afs_write_end() to handle len > page size [06/67] nfs, cifs, ceph, 9p: Disable use of fscache prior to its rewrite [07/67] fscache: Remove the netfs data from the cookie [08/67] fscache: Remove struct fscache_cookie_def [09/67] fscache: Remove store_limit* from struct fscache_object [10/67] fscache: Remove fscache_check_consistency() [11/67] fscache: Remove fscache_attr_changed() [12/67] fscache: Remove obsolete stats [13/67] fscache: Remove old I/O tracepoints [14/67] fscache: Temporarily disable fscache_invalidate() [15/67] fscache: Disable fscache_begin_operation() [16/67] fscache: Remove the I/O operation manager [17/67] fscache: Rename fscache_cookie_{get,put,see}() [18/67] cachefiles: Remove tree of active files and use S_CACHE_FILE inode flag [19/67] cachefiles: Don't set an xattr on the root of the cache [20/67] cachefiles: Remove some redundant checks on unsigned values [21/67] cachefiles: Prevent inode from going away when burying a dentry [22/67] cachefiles: Simplify the pathwalk and save the filename for an object [23/67] cachefiles: trace: Improve the lookup tracepoint [24/67] cachefiles: Remove separate backer dentry from cachefiles_object [25/67] cachefiles: Fold fscache_object into cachefiles_object [26/67] cachefiles: Change to storing file* rather than dentry* [27/67] cachefiles: trace: Log coherency checks [28/67] cachefiles: Trace truncations [29/67] cachefiles: Trace read and write operations [30/67] cachefiles: Round the cachefile size up to DIO block size [31/67] cachefiles: Don't use XATTR_ flags with vfs_setxattr() [33/67] cachefiles: Trace decisions in cachefiles_prepare_read() [34/67] cachefiles: Make cachefiles_write_prepare() check for space [35/67] fscache: Automatically close a file that's been unused for a while [36/67] fscache: Add stats for the cookie commit LRU [37/67] fscache: Move fscache_update_cookie() complete inline [38/67] fscache: Remove more obsolete stats [39/67] fscache: Note the object size during invalidation [40/67] vfs, fscache: Force ->write_inode() to occur if cookie pinned for writeback [41/67] afs: Render cache cookie key as big endian [42/67] cachefiles: Use tmpfile/link [43/67] fscache: Rewrite invalidation [44/67] fscache: disable cookie when doing an invalidation for DIO write [45/67] cachefiles: Simplify the file lookup/creation/check code [46/67] fscache: Provide resize operation [47/67] cachefiles: Put more information in the xattr attached to the cache file [48/67] fscache: Implement "will_modify" parameter on fscache_use_cookie() [49/67] fscache: Add support for writing to the cache [50/67] fscache: Make fscache_clear_page_bits() conditional on cookie [51/67] fscache: Make fscache_write_to_cache() conditional on cookie [52/67] afs: Copy local writes to the cache when writing to the server [53/67] afs: Invoke fscache_resize_cookie() when handling ATTR_SIZE for setattr [54/67] afs: Add O_DIRECT read support [55/67] afs: Skip truncation on the server of data we haven't written yet [56/67] afs: Make afs_write_begin() return the THP subpage [57/67] cachefiles, afs: Drive FSCACHE_COOKIE_NO_DATA_TO_READ [58/67] NFS: Convert fscache_acquire_cookie and fscache_relinquish_cookie [59/67] NFS: Convert fscache_enable_cookie and fscache_disable_cookie [60/67] NFS: Convert fscache invalidation and update aux_data and i_size [61/67] nfs: Convert to new fscache volume/cookie API [62/67] 9p: Use fscache indexing rewrite and reenable caching [63/67] 9p: Copy local writes to the cache when writing to the server [64/67] netfs: Display the netfs inode number in the netfs_read tracepoint [65/67] cachefiles: Add tracepoints to log errors from ops on the backing fs [66/67] cachefiles: Add error injection support [67/67] cifs: Support fscache indexing rewrite (untested)

diff --git a/fs/afs/file.c b/fs/afs/file.c index 4af7ef607cf6..5db1e7d29ad5 100644 --- a/fs/afs/file.c +++ b/fs/afs/file.c @@ -28,6 +28,7 @@ static ssize_t afs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter); static void afs_vm_open(struct vm_area_struct *area); static void afs_vm_close(struct vm_area_struct *area); static vm_fault_t afs_vm_map_pages(struct vm_fault *vmf, pgoff_t start_pgoff, pgoff_t end_pgoff); +static ssize_t afs_direct_IO(struct kiocb *iocb, struct iov_iter *iter); const struct file_operations afs_file_operations = { .open = afs_open, @@ -56,6 +57,7 @@ const struct address_space_operations afs_fs_aops = { .launder_page = afs_launder_page, .releasepage = afs_releasepage, .invalidatepage = afs_invalidatepage, + .direct_IO = afs_direct_IO, .write_begin = afs_write_begin, .write_end = afs_write_end, .writepage = afs_writepage, @@ -601,3 +603,60 @@ static ssize_t afs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter) return generic_file_read_iter(iocb, iter); } + +/* + * Direct file read operation for an AFS file. + * + * TODO: To support AIO, the pages in the iterator have to be copied and + * refs taken on them. Then -EIOCBQUEUED needs to be returned. + * iocb->ki_complete must then be called upon completion of the operation. + */ +static ssize_t afs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter) +{ + struct file *file = iocb->ki_filp; + struct afs_vnode *vnode = AFS_FS_I(file_inode(file)); + struct afs_read *req; + ssize_t ret, transferred; + + _enter("%llx,%zx", iocb->ki_pos, iov_iter_count(iter)); + + req = afs_alloc_read(GFP_KERNEL); + if (!req) + return -ENOMEM; + + req->vnode = vnode; + req->key = key_get(afs_file_key(file)); + req->pos = iocb->ki_pos; + req->len = iov_iter_count(iter); + req->iter = iter; + + task_io_account_read(req->len); + + // TODO nfs_start_io_direct(inode); + ret = afs_fetch_data(vnode, req); + if (ret == 0) + transferred = req->actual_len; + afs_put_read(req); + + // TODO nfs_end_io_direct(inode); + + if (ret == 0) + ret = transferred; + + BUG_ON(ret == -EIOCBQUEUED); // TODO + //if (iocb->ki_complete) + // iocb->ki_complete(iocb, ret, 0); // only if ret == -EIOCBQUEUED + + _leave(" = %zu", ret); + return ret; +} + +/* + * Do direct I/O. + */ +static ssize_t afs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) +{ + if (iov_iter_rw(iter) == READ) + return afs_file_direct_read(iocb, iter); + return afs_file_direct_write(iocb, iter); +} diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index 4943413d9c5f..290d4d391a0e 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -439,7 +439,7 @@ static void afs_fs_fetch_data64(struct afs_operation *op) bp[3] = htonl(vp->fid.unique); bp[4] = htonl(upper_32_bits(req->pos)); bp[5] = htonl(lower_32_bits(req->pos)); - bp[6] = 0; + bp[6] = htonl(upper_32_bits(req->len)); bp[7] = htonl(lower_32_bits(req->len)); trace_afs_make_fs_call(call, &vp->fid); @@ -1057,6 +1057,7 @@ static void afs_fs_store_data64(struct afs_operation *op) struct afs_vnode_param *vp = &op->file[0]; struct afs_call *call; __be32 *bp; + u32 mask = 0; _enter(",%x,{%llx:%llu},,", key_serial(op->key), vp->fid.vid, vp->fid.vnode); @@ -1069,6 +1070,9 @@ static void afs_fs_store_data64(struct afs_operation *op) call->write_iter = op->store.write_iter; + if (op->flags & AFS_OPERATION_SET_MTIME) + mask |= AFS_SET_MTIME; + /* marshall the parameters */ bp = call->request; *bp++ = htonl(FSSTOREDATA64); @@ -1076,8 +1080,8 @@ static void afs_fs_store_data64(struct afs_operation *op) *bp++ = htonl(vp->fid.vnode); *bp++ = htonl(vp->fid.unique); - *bp++ = htonl(AFS_SET_MTIME); /* mask */ - *bp++ = htonl(op->mtime.tv_sec); /* mtime */ + *bp++ = htonl(mask); + *bp++ = htonl(op->mtime.tv_sec); *bp++ = 0; /* owner */ *bp++ = 0; /* group */ *bp++ = 0; /* unix mode */ @@ -1102,6 +1106,7 @@ void afs_fs_store_data(struct afs_operation *op) struct afs_vnode_param *vp = &op->file[0]; struct afs_call *call; __be32 *bp; + u32 mask = 0; _enter(",%x,{%llx:%llu},,", key_serial(op->key), vp->fid.vid, vp->fid.vnode); @@ -1122,6 +1127,9 @@ void afs_fs_store_data(struct afs_operation *op) call->write_iter = op->store.write_iter; + if (op->flags & AFS_OPERATION_SET_MTIME) + mask |= AFS_SET_MTIME; + /* marshall the parameters */ bp = call->request; *bp++ = htonl(FSSTOREDATA); @@ -1129,8 +1137,8 @@ void afs_fs_store_data(struct afs_operation *op) *bp++ = htonl(vp->fid.vnode); *bp++ = htonl(vp->fid.unique); - *bp++ = htonl(AFS_SET_MTIME); /* mask */ - *bp++ = htonl(op->mtime.tv_sec); /* mtime */ + *bp++ = htonl(mask); + *bp++ = htonl(op->mtime.tv_sec); *bp++ = 0; /* owner */ *bp++ = 0; /* group */ *bp++ = 0; /* unix mode */ diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 07d34291bf4f..d3fb9b3845bc 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -863,6 +863,7 @@ struct afs_operation { #define AFS_OPERATION_TRIED_ALL 0x0400 /* Set if we've tried all the fileservers */ #define AFS_OPERATION_RETRY_SERVER 0x0800 /* Set if we should retry the current server */ #define AFS_OPERATION_DIR_CONFLICT 0x1000 /* Set if we detected a 3rd-party dir change */ +#define AFS_OPERATION_SET_MTIME 0x2000 /* Set if we should try to store the mtime */ }; /* @@ -1538,6 +1539,7 @@ extern int afs_fsync(struct file *, loff_t, loff_t, int); extern vm_fault_t afs_page_mkwrite(struct vm_fault *vmf); extern void afs_prune_wb_keys(struct afs_vnode *); extern int afs_launder_page(struct page *); +extern ssize_t afs_file_direct_write(struct kiocb *, struct iov_iter *); /* * xattr.c diff --git a/fs/afs/write.c b/fs/afs/write.c index a540da6ad423..281f0e93e2c6 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -406,7 +406,7 @@ static int afs_store_data(struct afs_vnode *vnode, struct iov_iter *iter, loff_t op->store.i_size = max(pos + size, i_size); op->store.laundering = laundering; op->mtime = vnode->vfs_inode.i_mtime; - op->flags |= AFS_OPERATION_UNINTR; + op->flags |= AFS_OPERATION_SET_MTIME | AFS_OPERATION_UNINTR; op->ops = &afs_store_data_operation; try_next_key: @@ -1039,3 +1039,73 @@ static void afs_write_to_cache(struct afs_vnode *vnode, vnode->vfs_inode.i_mapping, start, len, i_size, afs_write_to_cache_done, vnode); } + +static void afs_dio_store_data_success(struct afs_operation *op) +{ + struct afs_vnode *vnode = op->file[0].vnode; + + op->ctime = op->file[0].scb.status.mtime_client; + afs_vnode_commit_status(op, &op->file[0]); + if (op->error == 0) { + afs_stat_v(vnode, n_stores); + atomic_long_add(op->store.size, &afs_v2net(vnode)->n_store_bytes); + } +} + +static const struct afs_operation_ops afs_dio_store_data_operation = { + .issue_afs_rpc = afs_fs_store_data, + .issue_yfs_rpc = yfs_fs_store_data, + .success = afs_dio_store_data_success, +}; + +/* + * Direct file write operation for an AFS file. + * + * TODO: To support AIO, the pages in the iterator have to be copied and + * refs taken on them. Then -EIOCBQUEUED needs to be returned. + * iocb->ki_complete must then be called upon completion of the operation. + */ +ssize_t afs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) +{ + struct file *file = iocb->ki_filp; + struct afs_vnode *vnode = AFS_FS_I(file_inode(file)); + struct afs_operation *op; + loff_t size = iov_iter_count(iter), i_size; + ssize_t ret; + + _enter("%s{%llx:%llu.%u},%llx,%llx", + vnode->volume->name, + vnode->fid.vid, + vnode->fid.vnode, + vnode->fid.unique, + size, iocb->ki_pos); + + op = afs_alloc_operation(afs_file_key(file), vnode->volume); + if (IS_ERR(op)) + return -ENOMEM; + + i_size = i_size_read(&vnode->vfs_inode); + + afs_op_set_vnode(op, 0, vnode); + op->file[0].dv_delta = 1; + op->file[0].set_size = true; + op->store.write_iter = iter; + op->store.pos = iocb->ki_pos; + op->store.size = size; + op->store.i_size = max(iocb->ki_pos + size, i_size); + op->ops = &afs_dio_store_data_operation; + + //if (!is_sync_kiocb(iocb)) { + + ret = afs_do_sync_operation(op); + if (ret == 0) + ret = size; + + afs_invalidate_cache(vnode, FSCACHE_INVAL_DIO_WRITE); + + //if (iocb->ki_complete) + // iocb->ki_complete(iocb, ret, 0); // only if ret == -EIOCBQUEUED + + _leave(" = %zd", ret); + return ret; +} diff --git a/fs/afs/yfsclient.c b/fs/afs/yfsclient.c index 2b35cba8ad62..a8c4e230002d 100644 --- a/fs/afs/yfsclient.c +++ b/fs/afs/yfsclient.c @@ -95,12 +95,16 @@ static __be32 *xdr_encode_YFSStoreStatus_mode(__be32 *bp, mode_t mode) return bp + xdr_size(x); } -static __be32 *xdr_encode_YFSStoreStatus_mtime(__be32 *bp, const struct timespec64 *t) +static __be32 *xdr_encode_YFSStoreStatus_mtime(__be32 *bp, struct afs_operation *op) { struct yfs_xdr_YFSStoreStatus *x = (void *)bp; - s64 mtime = linux_to_yfs_time(t); + s64 mtime = linux_to_yfs_time(&op->mtime); + u32 mask = 0; - x->mask = htonl(AFS_SET_MTIME); + if (op->flags & AFS_OPERATION_SET_MTIME) + mask |= AFS_SET_MTIME; + + x->mask = htonl(mask); x->mode = htonl(0); x->mtime_client = u64_to_xdr(mtime); x->owner = u64_to_xdr(0); @@ -1103,7 +1107,7 @@ void yfs_fs_store_data(struct afs_operation *op) bp = xdr_encode_u32(bp, YFSSTOREDATA64); bp = xdr_encode_u32(bp, 0); /* RPC flags */ bp = xdr_encode_YFSFid(bp, &vp->fid); - bp = xdr_encode_YFSStoreStatus_mtime(bp, &op->mtime); + bp = xdr_encode_YFSStoreStatus_mtime(bp, op); bp = xdr_encode_u64(bp, op->store.pos); bp = xdr_encode_u64(bp, op->store.size); bp = xdr_encode_u64(bp, op->store.i_size);

[54/67] afs: Add O_DIRECT read support

Commit Message

Patch