From 6c93df5db628e710697c43bc1bd78a786549a548 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Fri, 15 Apr 2016 13:56:12 +0800 Subject: ceph: don't call truncate_pagecache in ceph_writepages_start truncate_pagecache() may decrease inode's reference. This can cause deadlock if inode's last reference is dropped and iput_final() wants to evict the inode. (evict() calls inode_wait_for_writeback(), which waits for ceph_writepages_start() to return). The fix is use work thead to truncate dirty pages. Also add 'forced umount' check to ceph_update_writeable_page(), which prevents new pages getting dirty. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index edfade037738..b906e02cddad 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1623,10 +1623,21 @@ static void ceph_invalidate_work(struct work_struct *work) struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info, i_pg_inv_work); struct inode *inode = &ci->vfs_inode; + struct ceph_fs_client *fsc = ceph_inode_to_client(inode); u32 orig_gen; int check = 0; mutex_lock(&ci->i_truncate_mutex); + + if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) { + pr_warn_ratelimited("invalidate_pages %p %lld forced umount\n", + inode, ceph_ino(inode)); + mapping_set_error(inode->i_mapping, -EIO); + truncate_pagecache(inode, 0); + mutex_unlock(&ci->i_truncate_mutex); + goto out; + } + spin_lock(&ci->i_ceph_lock); dout("invalidate_pages %p gen %d revoking %d\n", inode, ci->i_rdcache_gen, ci->i_rdcache_revoking); -- cgit v1.2.3 From a78600e7c4fb47fb5ef34265456b731fde27a9c3 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 27 Apr 2016 17:32:34 +0800 Subject: ceph: simplify 'offset in frag' don't distinguish leftmost frag from other frags. always use 2 as first entry's offset. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index b906e02cddad..51ce5ce59f04 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1405,10 +1405,7 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, dout("readdir_prepopulate got new frag %x -> %x\n", frag, le32_to_cpu(rinfo->dir_dir->frag)); frag = le32_to_cpu(rinfo->dir_dir->frag); - if (ceph_frag_is_leftmost(frag)) - req->r_readdir_offset = 2; - else - req->r_readdir_offset = 0; + req->r_readdir_offset = 2; } if (le32_to_cpu(rinfo->head->op) == CEPH_MDS_OP_LSSNAP) { -- cgit v1.2.3 From 2a5beea3f1b6544d6c72ea220e860a2eda2f9104 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Thu, 28 Apr 2016 09:37:39 +0800 Subject: ceph: define struct for dir entry in readdir reply This avoids defining multiple arrays for entries in readdir reply Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 51ce5ce59f04..40d081d7028e 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1308,12 +1308,13 @@ static int readdir_prepopulate_inodes_only(struct ceph_mds_request *req, int i, err = 0; for (i = 0; i < rinfo->dir_nr; i++) { + struct ceph_mds_reply_dir_entry *rde = rinfo->dir_entries + i; struct ceph_vino vino; struct inode *in; int rc; - vino.ino = le64_to_cpu(rinfo->dir_in[i].in->ino); - vino.snap = le64_to_cpu(rinfo->dir_in[i].in->snapid); + vino.ino = le64_to_cpu(rde->inode.in->ino); + vino.snap = le64_to_cpu(rde->inode.in->snapid); in = ceph_get_inode(req->r_dentry->d_sb, vino); if (IS_ERR(in)) { @@ -1321,7 +1322,7 @@ static int readdir_prepopulate_inodes_only(struct ceph_mds_request *req, dout("new_inode badness got %d\n", err); continue; } - rc = fill_inode(in, NULL, &rinfo->dir_in[i], NULL, session, + rc = fill_inode(in, NULL, &rde->inode, NULL, session, req->r_request_started, -1, &req->r_caps_reservation); if (rc < 0) { @@ -1433,14 +1434,15 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, /* FIXME: release caps/leases if error occurs */ for (i = 0; i < rinfo->dir_nr; i++) { + struct ceph_mds_reply_dir_entry *rde = rinfo->dir_entries + i; struct ceph_vino vino; - dname.name = rinfo->dir_dname[i]; - dname.len = rinfo->dir_dname_len[i]; + dname.name = rde->name; + dname.len = rde->name_len; dname.hash = full_name_hash(dname.name, dname.len); - vino.ino = le64_to_cpu(rinfo->dir_in[i].in->ino); - vino.snap = le64_to_cpu(rinfo->dir_in[i].in->snapid); + vino.ino = le64_to_cpu(rde->inode.in->ino); + vino.snap = le64_to_cpu(rde->inode.in->snapid); retry_lookup: dn = d_lookup(parent, &dname); @@ -1486,7 +1488,7 @@ retry_lookup: } } - ret = fill_inode(in, NULL, &rinfo->dir_in[i], NULL, session, + ret = fill_inode(in, NULL, &rde->inode, NULL, session, req->r_request_started, -1, &req->r_caps_reservation); if (ret < 0) { @@ -1522,8 +1524,7 @@ retry_lookup: di = dn->d_fsdata; di->offset = ceph_make_fpos(frag, i + req->r_readdir_offset); - update_dentry_lease(dn, rinfo->dir_dlease[i], - req->r_session, + update_dentry_lease(dn, rde->lease, req->r_session, req->r_request_started); if (err == 0 && skipped == 0 && cache_ctl.index >= 0) { -- cgit v1.2.3 From 8974eebd38737c9534d81c4131c5fdb1fe24d3e9 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Thu, 28 Apr 2016 15:17:40 +0800 Subject: ceph: record 'offset' for each entry of readdir result This is preparation for using hash value as dentry 'offset' Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 40d081d7028e..b53c95903aeb 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1523,6 +1523,7 @@ retry_lookup: di = dn->d_fsdata; di->offset = ceph_make_fpos(frag, i + req->r_readdir_offset); + rde->offset = di->offset; update_dentry_lease(dn, rde->lease, req->r_session, req->r_request_started); -- cgit v1.2.3 From f3c4ebe65ea149ec892f94474233cfebe9cbe299 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Fri, 29 Apr 2016 11:27:30 +0800 Subject: ceph: using hash value to compose dentry offset If MDS sorts dentries in dirfrag in hash order, we use hash value to compose dentry offset. dentry offset is: (0xff << 52) | ((24 bits hash) << 28) | (the nth entry hash hash collision) This offset is stable across directory fragmentation. This alos means there is no need to reset readdir offset if directory get fragmented in the middle of readdir. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 31 +++++++++++++++++++++++++------ 1 file changed, 25 insertions(+), 6 deletions(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index b53c95903aeb..f51b6fd5f570 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1387,6 +1387,7 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, struct ceph_mds_session *session) { struct dentry *parent = req->r_dentry; + struct ceph_inode_info *ci = ceph_inode(d_inode(parent)); struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info; struct qstr dname; struct dentry *dn; @@ -1394,19 +1395,27 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, int err = 0, skipped = 0, ret, i; struct inode *snapdir = NULL; struct ceph_mds_request_head *rhead = req->r_request->front.iov_base; - struct ceph_dentry_info *di; u32 frag = le32_to_cpu(rhead->args.readdir.frag); + u32 last_hash = 0; + u32 fpos_offset; struct ceph_readdir_cache_control cache_ctl = {}; if (req->r_aborted) return readdir_prepopulate_inodes_only(req, session); + if (rinfo->hash_order && req->r_path2) { + last_hash = ceph_str_hash(ci->i_dir_layout.dl_dir_hash, + req->r_path2, strlen(req->r_path2)); + last_hash = ceph_frag_value(last_hash); + } + if (rinfo->dir_dir && le32_to_cpu(rinfo->dir_dir->frag) != frag) { dout("readdir_prepopulate got new frag %x -> %x\n", frag, le32_to_cpu(rinfo->dir_dir->frag)); frag = le32_to_cpu(rinfo->dir_dir->frag); - req->r_readdir_offset = 2; + if (!rinfo->hash_order) + req->r_readdir_offset = 2; } if (le32_to_cpu(rinfo->head->op) == CEPH_MDS_OP_LSSNAP) { @@ -1424,13 +1433,13 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, if (ceph_frag_is_leftmost(frag) && req->r_readdir_offset == 2) { /* note dir version at start of readdir so we can tell * if any dentries get dropped */ - struct ceph_inode_info *ci = ceph_inode(d_inode(parent)); req->r_dir_release_cnt = atomic64_read(&ci->i_release_count); req->r_dir_ordered_cnt = atomic64_read(&ci->i_ordered_count); req->r_readdir_cache_idx = 0; } cache_ctl.index = req->r_readdir_cache_idx; + fpos_offset = req->r_readdir_offset; /* FIXME: release caps/leases if error occurs */ for (i = 0; i < rinfo->dir_nr; i++) { @@ -1444,6 +1453,18 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, vino.ino = le64_to_cpu(rde->inode.in->ino); vino.snap = le64_to_cpu(rde->inode.in->snapid); + if (rinfo->hash_order) { + u32 hash = ceph_str_hash(ci->i_dir_layout.dl_dir_hash, + rde->name, rde->name_len); + hash = ceph_frag_value(hash); + if (hash != last_hash) + fpos_offset = 2; + last_hash = hash; + rde->offset = ceph_make_fpos(hash, fpos_offset++, true); + } else { + rde->offset = ceph_make_fpos(frag, fpos_offset++, false); + } + retry_lookup: dn = d_lookup(parent, &dname); dout("d_lookup on parent=%p name=%.*s got %p\n", @@ -1521,9 +1542,7 @@ retry_lookup: dn = realdn; } - di = dn->d_fsdata; - di->offset = ceph_make_fpos(frag, i + req->r_readdir_offset); - rde->offset = di->offset; + ceph_dentry(dn)->offset = rde->offset; update_dentry_lease(dn, rde->lease, req->r_session, req->r_request_started); -- cgit v1.2.3 From 209ae762a64397557be2bc48490fd968912e9a71 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Fri, 29 Apr 2016 23:40:23 +0800 Subject: ceph: fix inode reference leak Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index f51b6fd5f570..4889aaa72acc 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1328,8 +1328,8 @@ static int readdir_prepopulate_inodes_only(struct ceph_mds_request *req, if (rc < 0) { pr_err("fill_inode badness on %p got %d\n", in, rc); err = rc; - continue; } + iput(in); } return err; -- cgit v1.2.3 From a407846ef7c6e7905ab0d1ebc0c86fbebb065b87 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Tue, 3 May 2016 20:55:50 +0800 Subject: ceph: don't assume frag tree splits in mds reply are sorted The algorithm that updates i_fragtree relies on that the frag tree splits in mds reply are of the same order of i_fragtree. This is not true because current MDS encodes frag tree splits in ascending order of (unsigned)frag_t. But nodes in i_fragtree are sorted according to ceph_frag_compare(). The fix is sort the frag tree splits first, then updates i_fragtree. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 4889aaa72acc..d91eb6b49259 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -10,6 +10,7 @@ #include #include #include +#include #include "super.h" #include "mds_client.h" @@ -299,6 +300,13 @@ out: return err; } +static int frag_tree_split_cmp(const void *l, const void *r) +{ + struct ceph_frag_tree_split *ls = (struct ceph_frag_tree_split*)l; + struct ceph_frag_tree_split *rs = (struct ceph_frag_tree_split*)r; + return ceph_frag_compare(ls->frag, rs->frag); +} + static int ceph_fill_fragtree(struct inode *inode, struct ceph_frag_tree_head *fragtree, struct ceph_mds_reply_dirfrag *dirinfo) @@ -331,6 +339,11 @@ static int ceph_fill_fragtree(struct inode *inode, if (!update) goto out_unlock; + if (nsplits > 1) { + sort(fragtree->splits, nsplits, sizeof(fragtree->splits[0]), + frag_tree_split_cmp, NULL); + } + dout("fill_fragtree %llx.%llx\n", ceph_vinop(inode)); rb_node = rb_first(&ci->i_fragtree); for (i = 0; i < nsplits; i++) { -- cgit v1.2.3 From 421721195a68f46a8218c664154076b5b06f5f51 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Tue, 3 May 2016 22:33:20 +0800 Subject: ceph: fix dir_auth check in ceph_fill_dirfrag() -1 is CDIR_AUTH_PARENT, it means dir's auth mds is the same as inode's auth mds Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index d91eb6b49259..562f57d15f46 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -254,6 +254,9 @@ static int ceph_fill_dirfrag(struct inode *inode, diri_auth = ci->i_auth_cap->mds; spin_unlock(&ci->i_ceph_lock); + if (mds == -1) /* CDIR_AUTH_PARENT */ + mds = diri_auth; + mutex_lock(&ci->i_fragtree_mutex); if (ndist == 0 && mds == diri_auth) { /* no delegation info needed. */ -- cgit v1.2.3 From a4b7431f39438f415cef1fa28502ec58016686f7 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 4 May 2016 11:05:10 +0800 Subject: ceph: keep leaf frag when updating fragtree Nodes in i_fragtree are sorted according to ceph_compare_frag(). It means frag node in i_fragtree always follow its direct parent node. To check if a leaf node is valid, we just need to check if it's child of previous split node. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 28 +++++++++++++++++++++++----- 1 file changed, 23 insertions(+), 5 deletions(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 562f57d15f46..09713f32ea88 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -310,12 +310,21 @@ static int frag_tree_split_cmp(const void *l, const void *r) return ceph_frag_compare(ls->frag, rs->frag); } +static bool is_frag_child(u32 f, struct ceph_inode_frag *frag) +{ + if (!frag) + return f == ceph_frag_make(0, 0); + if (ceph_frag_bits(f) != ceph_frag_bits(frag->frag) + frag->split_by) + return false; + return ceph_frag_contains_value(frag->frag, ceph_frag_value(f)); +} + static int ceph_fill_fragtree(struct inode *inode, struct ceph_frag_tree_head *fragtree, struct ceph_mds_reply_dirfrag *dirinfo) { struct ceph_inode_info *ci = ceph_inode(inode); - struct ceph_inode_frag *frag; + struct ceph_inode_frag *frag, *prev_frag = NULL; struct rb_node *rb_node; int i; u32 id, nsplits; @@ -362,8 +371,12 @@ static int ceph_fill_fragtree(struct inode *inode, break; } rb_node = rb_next(rb_node); - rb_erase(&frag->node, &ci->i_fragtree); - kfree(frag); + /* delete stale split/leaf node */ + if (frag->split_by > 0 || + !is_frag_child(frag->frag, prev_frag)) { + rb_erase(&frag->node, &ci->i_fragtree); + kfree(frag); + } frag = NULL; } if (!frag) { @@ -373,12 +386,17 @@ static int ceph_fill_fragtree(struct inode *inode, } frag->split_by = le32_to_cpu(fragtree->splits[i].by); dout(" frag %x split by %d\n", frag->frag, frag->split_by); + prev_frag = frag; } while (rb_node) { frag = rb_entry(rb_node, struct ceph_inode_frag, node); rb_node = rb_next(rb_node); - rb_erase(&frag->node, &ci->i_fragtree); - kfree(frag); + /* delete stale split/leaf node */ + if (frag->split_by > 0 || + !is_frag_child(frag->frag, prev_frag)) { + rb_erase(&frag->node, &ci->i_fragtree); + kfree(frag); + } } out_unlock: mutex_unlock(&ci->i_fragtree_mutex); -- cgit v1.2.3 From 1b1bc16d66a7c7af3b4f30d1cf5a363168b217f4 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 4 May 2016 11:40:30 +0800 Subject: ceph: improve fragtree change detection check if number of splits in i_fragtree is equal to number of splits in mds reply Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 24 ++++++++++++++++++++---- 1 file changed, 20 insertions(+), 4 deletions(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 09713f32ea88..7ba8e1c2557b 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -326,13 +326,15 @@ static int ceph_fill_fragtree(struct inode *inode, struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_inode_frag *frag, *prev_frag = NULL; struct rb_node *rb_node; - int i; - u32 id, nsplits; + unsigned i, split_by, nsplits; + u32 id; bool update = false; mutex_lock(&ci->i_fragtree_mutex); nsplits = le32_to_cpu(fragtree->nsplits); - if (nsplits) { + if (nsplits != ci->i_fragtree_nsplits) { + update = true; + } else if (nsplits) { i = prandom_u32() % nsplits; id = le32_to_cpu(fragtree->splits[i].frag); if (!__ceph_find_frag(ci, id)) @@ -360,6 +362,13 @@ static int ceph_fill_fragtree(struct inode *inode, rb_node = rb_first(&ci->i_fragtree); for (i = 0; i < nsplits; i++) { id = le32_to_cpu(fragtree->splits[i].frag); + split_by = le32_to_cpu(fragtree->splits[i].by); + if (split_by == 0 || ceph_frag_bits(id) + split_by > 24) { + pr_err("fill_fragtree %llx.%llx invalid split %d/%u, " + "frag %x split by %d\n", ceph_vinop(inode), + i, nsplits, id, split_by); + continue; + } frag = NULL; while (rb_node) { frag = rb_entry(rb_node, struct ceph_inode_frag, node); @@ -375,6 +384,8 @@ static int ceph_fill_fragtree(struct inode *inode, if (frag->split_by > 0 || !is_frag_child(frag->frag, prev_frag)) { rb_erase(&frag->node, &ci->i_fragtree); + if (frag->split_by > 0) + ci->i_fragtree_nsplits--; kfree(frag); } frag = NULL; @@ -384,7 +395,9 @@ static int ceph_fill_fragtree(struct inode *inode, if (IS_ERR(frag)) continue; } - frag->split_by = le32_to_cpu(fragtree->splits[i].by); + if (frag->split_by == 0) + ci->i_fragtree_nsplits++; + frag->split_by = split_by; dout(" frag %x split by %d\n", frag->frag, frag->split_by); prev_frag = frag; } @@ -395,6 +408,8 @@ static int ceph_fill_fragtree(struct inode *inode, if (frag->split_by > 0 || !is_frag_child(frag->frag, prev_frag)) { rb_erase(&frag->node, &ci->i_fragtree); + if (frag->split_by > 0) + ci->i_fragtree_nsplits--; kfree(frag); } } @@ -546,6 +561,7 @@ void ceph_destroy_inode(struct inode *inode) rb_erase(n, &ci->i_fragtree); kfree(frag); } + ci->i_fragtree_nsplits = 0; __ceph_destroy_xattrs(ci); if (ci->i_xattrs.blob) -- cgit v1.2.3 From 224a7542b8fdde3cc7c600f8b0870c5541a9f678 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Thu, 5 May 2016 16:40:17 +0800 Subject: ceph: tolerate bad i_size for symlink inode A mds bug can cause symlink's size to be truncated to zero. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 22 +++++++++++++++------- 1 file changed, 15 insertions(+), 7 deletions(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 7ba8e1c2557b..89d08155986d 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -582,6 +582,11 @@ int ceph_drop_inode(struct inode *inode) return 1; } +static inline blkcnt_t calc_inode_blocks(u64 size) +{ + return (size + (1<<9) - 1) >> 9; +} + /* * Helpers to fill in size, ctime, mtime, and atime. We have to be * careful because either the client or MDS may have more up to date @@ -604,7 +609,7 @@ int ceph_fill_file_size(struct inode *inode, int issued, size = 0; } i_size_write(inode, size); - inode->i_blocks = (size + (1<<9) - 1) >> 9; + inode->i_blocks = calc_inode_blocks(size); ci->i_reported_size = size; if (truncate_seq != ci->i_truncate_seq) { dout("truncate_seq %u -> %u\n", @@ -863,9 +868,13 @@ static int fill_inode(struct inode *inode, struct page *locked_page, spin_unlock(&ci->i_ceph_lock); - err = -EINVAL; - if (WARN_ON(symlen != i_size_read(inode))) - goto out; + if (symlen != i_size_read(inode)) { + pr_err("fill_inode %llx.%llx BAD symlink " + "size %lld\n", ceph_vinop(inode), + i_size_read(inode)); + i_size_write(inode, symlen); + inode->i_blocks = calc_inode_blocks(symlen); + } err = -ENOMEM; sym = kstrndup(iinfo->symlink, symlen, GFP_NOFS); @@ -1629,7 +1638,7 @@ int ceph_inode_set_size(struct inode *inode, loff_t size) spin_lock(&ci->i_ceph_lock); dout("set_size %p %llu -> %llu\n", inode, inode->i_size, size); i_size_write(inode, size); - inode->i_blocks = (size + (1 << 9) - 1) >> 9; + inode->i_blocks = calc_inode_blocks(size); /* tell the MDS if we are approaching max_size */ if ((size << 1) >= ci->i_max_size && @@ -2002,8 +2011,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) if ((issued & CEPH_CAP_FILE_EXCL) && attr->ia_size > inode->i_size) { i_size_write(inode, attr->ia_size); - inode->i_blocks = - (attr->ia_size + (1 << 9) - 1) >> 9; + inode->i_blocks = calc_inode_blocks(attr->ia_size); inode->i_ctime = attr->ia_ctime; ci->i_reported_size = attr->ia_size; dirtied |= CEPH_CAP_FILE_EXCL; -- cgit v1.2.3 From 9abd4db713704aac146395e079224ddd716e9b95 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 18 May 2016 20:58:26 +0800 Subject: ceph: don't use truncate_pagecache() to invalidate read cache truncate_pagecache() drops dirty pages, it's dangerous to use it to invalidate read cache. Besides, we shouldn't start invalidating read cache while there are buffer writers. Because buffer writers may add dirty pages later. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/ceph/inode.c') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 89d08155986d..07495ba61fe3 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1728,7 +1728,9 @@ static void ceph_invalidate_work(struct work_struct *work) orig_gen = ci->i_rdcache_gen; spin_unlock(&ci->i_ceph_lock); - truncate_pagecache(inode, 0); + if (invalidate_inode_pages2(inode->i_mapping) < 0) { + pr_err("invalidate_pages %p fails\n", inode); + } spin_lock(&ci->i_ceph_lock); if (orig_gen == ci->i_rdcache_gen && -- cgit v1.2.3