diff options
Diffstat (limited to 'fs/ceph/caps.c')
-rw-r--r-- | fs/ceph/caps.c | 330 |
1 files changed, 199 insertions, 131 deletions
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index b9460b6fb76f..fb023f9fafcb 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -492,7 +492,7 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc, struct ceph_mount_options *opt = mdsc->fsc->mount_options; ci->i_hold_caps_max = round_jiffies(jiffies + opt->caps_wanted_delay_max * HZ); - dout("__cap_set_timeouts %p %lu\n", &ci->vfs_inode, + dout("__cap_set_timeouts %p %lu\n", &ci->netfs.inode, ci->i_hold_caps_max - jiffies); } @@ -507,7 +507,7 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc, static void __cap_delay_requeue(struct ceph_mds_client *mdsc, struct ceph_inode_info *ci) { - dout("__cap_delay_requeue %p flags 0x%lx at %lu\n", &ci->vfs_inode, + dout("__cap_delay_requeue %p flags 0x%lx at %lu\n", &ci->netfs.inode, ci->i_ceph_flags, ci->i_hold_caps_max); if (!mdsc->stopping) { spin_lock(&mdsc->cap_delay_lock); @@ -531,7 +531,7 @@ no_change: static void __cap_delay_requeue_front(struct ceph_mds_client *mdsc, struct ceph_inode_info *ci) { - dout("__cap_delay_requeue_front %p\n", &ci->vfs_inode); + dout("__cap_delay_requeue_front %p\n", &ci->netfs.inode); spin_lock(&mdsc->cap_delay_lock); ci->i_ceph_flags |= CEPH_I_FLUSH; if (!list_empty(&ci->i_cap_delay_list)) @@ -548,7 +548,7 @@ static void __cap_delay_requeue_front(struct ceph_mds_client *mdsc, static void __cap_delay_cancel(struct ceph_mds_client *mdsc, struct ceph_inode_info *ci) { - dout("__cap_delay_cancel %p\n", &ci->vfs_inode); + dout("__cap_delay_cancel %p\n", &ci->netfs.inode); if (list_empty(&ci->i_cap_delay_list)) return; spin_lock(&mdsc->cap_delay_lock); @@ -568,7 +568,7 @@ static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap, * Each time we receive FILE_CACHE anew, we increment * i_rdcache_gen. */ - if (S_ISREG(ci->vfs_inode.i_mode) && + if (S_ISREG(ci->netfs.inode.i_mode) && (issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) && (had & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0) { ci->i_rdcache_gen++; @@ -583,14 +583,14 @@ static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap, if ((issued & CEPH_CAP_FILE_SHARED) != (had & CEPH_CAP_FILE_SHARED)) { if (issued & CEPH_CAP_FILE_SHARED) atomic_inc(&ci->i_shared_gen); - if (S_ISDIR(ci->vfs_inode.i_mode)) { - dout(" marking %p NOT complete\n", &ci->vfs_inode); + if (S_ISDIR(ci->netfs.inode.i_mode)) { + dout(" marking %p NOT complete\n", &ci->netfs.inode); __ceph_dir_clear_complete(ci); } } /* Wipe saved layout if we're losing DIR_CREATE caps */ - if (S_ISDIR(ci->vfs_inode.i_mode) && (had & CEPH_CAP_DIR_CREATE) && + if (S_ISDIR(ci->netfs.inode.i_mode) && (had & CEPH_CAP_DIR_CREATE) && !(issued & CEPH_CAP_DIR_CREATE)) { ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns)); memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout)); @@ -602,8 +602,8 @@ static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap, * @ci: inode to be moved * @session: new auth caps session */ -static void change_auth_cap_ses(struct ceph_inode_info *ci, - struct ceph_mds_session *session) +void change_auth_cap_ses(struct ceph_inode_info *ci, + struct ceph_mds_session *session) { lockdep_assert_held(&ci->i_ceph_lock); @@ -754,6 +754,7 @@ void ceph_add_cap(struct inode *inode, cap->issue_seq = seq; cap->mseq = mseq; cap->cap_gen = gen; + wake_up_all(&ci->i_cap_wq); } /* @@ -771,7 +772,7 @@ static int __cap_is_valid(struct ceph_cap *cap) if (cap->cap_gen < gen || time_after_eq(jiffies, ttl)) { dout("__cap_is_valid %p cap %p issued %s " - "but STALE (gen %u vs %u)\n", &cap->ci->vfs_inode, + "but STALE (gen %u vs %u)\n", &cap->ci->netfs.inode, cap, ceph_cap_string(cap->issued), cap->cap_gen, gen); return 0; } @@ -797,7 +798,7 @@ int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented) if (!__cap_is_valid(cap)) continue; dout("__ceph_caps_issued %p cap %p issued %s\n", - &ci->vfs_inode, cap, ceph_cap_string(cap->issued)); + &ci->netfs.inode, cap, ceph_cap_string(cap->issued)); have |= cap->issued; if (implemented) *implemented |= cap->implemented; @@ -844,12 +845,12 @@ static void __touch_cap(struct ceph_cap *cap) spin_lock(&s->s_cap_lock); if (!s->s_cap_iterator) { - dout("__touch_cap %p cap %p mds%d\n", &cap->ci->vfs_inode, cap, + dout("__touch_cap %p cap %p mds%d\n", &cap->ci->netfs.inode, cap, s->s_mds); list_move_tail(&cap->session_caps, &s->s_caps); } else { dout("__touch_cap %p cap %p mds%d NOP, iterating over caps\n", - &cap->ci->vfs_inode, cap, s->s_mds); + &cap->ci->netfs.inode, cap, s->s_mds); } spin_unlock(&s->s_cap_lock); } @@ -867,7 +868,7 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch) if ((have & mask) == mask) { dout("__ceph_caps_issued_mask ino 0x%llx snap issued %s" - " (mask %s)\n", ceph_ino(&ci->vfs_inode), + " (mask %s)\n", ceph_ino(&ci->netfs.inode), ceph_cap_string(have), ceph_cap_string(mask)); return 1; @@ -879,7 +880,7 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch) continue; if ((cap->issued & mask) == mask) { dout("__ceph_caps_issued_mask ino 0x%llx cap %p issued %s" - " (mask %s)\n", ceph_ino(&ci->vfs_inode), cap, + " (mask %s)\n", ceph_ino(&ci->netfs.inode), cap, ceph_cap_string(cap->issued), ceph_cap_string(mask)); if (touch) @@ -891,7 +892,7 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch) have |= cap->issued; if ((have & mask) == mask) { dout("__ceph_caps_issued_mask ino 0x%llx combo issued %s" - " (mask %s)\n", ceph_ino(&ci->vfs_inode), + " (mask %s)\n", ceph_ino(&ci->netfs.inode), ceph_cap_string(cap->issued), ceph_cap_string(mask)); if (touch) { @@ -919,7 +920,7 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch) int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, int touch) { - struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb); + struct ceph_fs_client *fsc = ceph_sb_to_client(ci->netfs.inode.i_sb); int r; r = __ceph_caps_issued_mask(ci, mask, touch); @@ -950,7 +951,7 @@ int __ceph_caps_revoking_other(struct ceph_inode_info *ci, int ceph_caps_revoking(struct ceph_inode_info *ci, int mask) { - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; int ret; spin_lock(&ci->i_ceph_lock); @@ -969,8 +970,8 @@ int __ceph_caps_used(struct ceph_inode_info *ci) if (ci->i_rd_ref) used |= CEPH_CAP_FILE_RD; if (ci->i_rdcache_ref || - (S_ISREG(ci->vfs_inode.i_mode) && - ci->vfs_inode.i_data.nrpages)) + (S_ISREG(ci->netfs.inode.i_mode) && + ci->netfs.inode.i_data.nrpages)) used |= CEPH_CAP_FILE_CACHE; if (ci->i_wr_ref) used |= CEPH_CAP_FILE_WR; @@ -993,11 +994,11 @@ int __ceph_caps_file_wanted(struct ceph_inode_info *ci) const int WR_SHIFT = ffs(CEPH_FILE_MODE_WR); const int LAZY_SHIFT = ffs(CEPH_FILE_MODE_LAZY); struct ceph_mount_options *opt = - ceph_inode_to_client(&ci->vfs_inode)->mount_options; + ceph_inode_to_client(&ci->netfs.inode)->mount_options; unsigned long used_cutoff = jiffies - opt->caps_wanted_delay_max * HZ; unsigned long idle_cutoff = jiffies - opt->caps_wanted_delay_min * HZ; - if (S_ISDIR(ci->vfs_inode.i_mode)) { + if (S_ISDIR(ci->netfs.inode.i_mode)) { int want = 0; /* use used_cutoff here, to keep dir's wanted caps longer */ @@ -1050,7 +1051,7 @@ int __ceph_caps_file_wanted(struct ceph_inode_info *ci) int __ceph_caps_wanted(struct ceph_inode_info *ci) { int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); - if (S_ISDIR(ci->vfs_inode.i_mode)) { + if (S_ISDIR(ci->netfs.inode.i_mode)) { /* we want EXCL if holding caps of dir ops */ if (w & CEPH_CAP_ANY_DIR_OPS) w |= CEPH_CAP_FILE_EXCL; @@ -1116,9 +1117,9 @@ void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release) lockdep_assert_held(&ci->i_ceph_lock); - dout("__ceph_remove_cap %p from %p\n", cap, &ci->vfs_inode); + dout("__ceph_remove_cap %p from %p\n", cap, &ci->netfs.inode); - mdsc = ceph_inode_to_client(&ci->vfs_inode)->mdsc; + mdsc = ceph_inode_to_client(&ci->netfs.inode)->mdsc; /* remove from inode's cap rbtree, and clear auth cap */ rb_erase(&cap->ci_node, &ci->i_caps); @@ -1169,7 +1170,7 @@ void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release) * keep i_snap_realm. */ if (ci->i_wr_ref == 0 && ci->i_snap_realm) - ceph_change_snap_realm(&ci->vfs_inode, NULL); + ceph_change_snap_realm(&ci->netfs.inode, NULL); __cap_delay_cancel(mdsc, ci); } @@ -1188,11 +1189,11 @@ void ceph_remove_cap(struct ceph_cap *cap, bool queue_release) lockdep_assert_held(&ci->i_ceph_lock); - fsc = ceph_inode_to_client(&ci->vfs_inode); + fsc = ceph_inode_to_client(&ci->netfs.inode); WARN_ON_ONCE(ci->i_auth_cap == cap && !list_empty(&ci->i_dirty_item) && !fsc->blocklisted && - !ceph_inode_is_shutdown(&ci->vfs_inode)); + !ceph_inode_is_shutdown(&ci->netfs.inode)); __ceph_remove_cap(cap, queue_release); } @@ -1343,7 +1344,7 @@ static void __prep_cap(struct cap_msg_args *arg, struct ceph_cap *cap, int flushing, u64 flush_tid, u64 oldest_flush_tid) { struct ceph_inode_info *ci = cap->ci; - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; int held, revoking; lockdep_assert_held(&ci->i_ceph_lock); @@ -1440,7 +1441,7 @@ static void __prep_cap(struct cap_msg_args *arg, struct ceph_cap *cap, static void __send_cap(struct cap_msg_args *arg, struct ceph_inode_info *ci) { struct ceph_msg *msg; - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPS, CAP_MSG_SIZE, GFP_NOFS, false); if (!msg) { @@ -1528,7 +1529,7 @@ static void __ceph_flush_snaps(struct ceph_inode_info *ci, __releases(ci->i_ceph_lock) __acquires(ci->i_ceph_lock) { - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; struct ceph_mds_client *mdsc = session->s_mdsc; struct ceph_cap_snap *capsnap; u64 oldest_flush_tid = 0; @@ -1577,7 +1578,7 @@ static void __ceph_flush_snaps(struct ceph_inode_info *ci, while (first_tid <= last_tid) { struct ceph_cap *cap = ci->i_auth_cap; - struct ceph_cap_flush *cf; + struct ceph_cap_flush *cf = NULL, *iter; int ret; if (!(cap && cap->session == session)) { @@ -1587,8 +1588,9 @@ static void __ceph_flush_snaps(struct ceph_inode_info *ci, } ret = -ENOENT; - list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) { - if (cf->tid >= first_tid) { + list_for_each_entry(iter, &ci->i_cap_flush_list, i_list) { + if (iter->tid >= first_tid) { + cf = iter; ret = 0; break; } @@ -1621,7 +1623,7 @@ static void __ceph_flush_snaps(struct ceph_inode_info *ci, void ceph_flush_snaps(struct ceph_inode_info *ci, struct ceph_mds_session **psession) { - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc; struct ceph_mds_session *session = NULL; int mds; @@ -1681,8 +1683,8 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, struct ceph_cap_flush **pcf) { struct ceph_mds_client *mdsc = - ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc; - struct inode *inode = &ci->vfs_inode; + ceph_sb_to_client(ci->netfs.inode.i_sb)->mdsc; + struct inode *inode = &ci->netfs.inode; int was = ci->i_dirty_caps; int dirty = 0; @@ -1695,7 +1697,7 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, return 0; } - dout("__mark_dirty_caps %p %s dirty %s -> %s\n", &ci->vfs_inode, + dout("__mark_dirty_caps %p %s dirty %s -> %s\n", &ci->netfs.inode, ceph_cap_string(mask), ceph_cap_string(was), ceph_cap_string(was | mask)); ci->i_dirty_caps |= mask; @@ -1711,7 +1713,7 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, ci->i_snap_realm->cached_context); } dout(" inode %p now dirty snapc %p auth cap %p\n", - &ci->vfs_inode, ci->i_head_snapc, ci->i_auth_cap); + &ci->netfs.inode, ci->i_head_snapc, ci->i_auth_cap); BUG_ON(!list_empty(&ci->i_dirty_item)); spin_lock(&mdsc->cap_dirty_lock); list_add(&ci->i_dirty_item, &session->s_cap_dirty); @@ -1856,7 +1858,7 @@ static int try_nonblocking_invalidate(struct inode *inode) u32 invalidating_gen = ci->i_rdcache_gen; spin_unlock(&ci->i_ceph_lock); - ceph_fscache_invalidate(inode); + ceph_fscache_invalidate(inode, false); invalidate_mapping_pages(&inode->i_data, 0, -1); spin_lock(&ci->i_ceph_lock); @@ -1874,7 +1876,7 @@ static int try_nonblocking_invalidate(struct inode *inode) bool __ceph_should_report_size(struct ceph_inode_info *ci) { - loff_t size = i_size_read(&ci->vfs_inode); + loff_t size = i_size_read(&ci->netfs.inode); /* mds will adjust max size according to the reported size */ if (ci->i_flushing_caps & CEPH_CAP_FILE_WR) return false; @@ -1899,7 +1901,7 @@ bool __ceph_should_report_size(struct ceph_inode_info *ci) void ceph_check_caps(struct ceph_inode_info *ci, int flags, struct ceph_mds_session *session) { - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb); struct ceph_cap *cap; u64 flush_tid, oldest_flush_tid; @@ -1910,11 +1912,19 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, struct rb_node *p; bool queue_invalidate = false; bool tried_invalidate = false; + bool queue_writeback = false; if (session) ceph_get_mds_session(session); spin_lock(&ci->i_ceph_lock); + if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE) { + /* Don't send messages until we get async create reply */ + spin_unlock(&ci->i_ceph_lock); + ceph_put_mds_session(session); + return; + } + if (ci->i_ceph_flags & CEPH_I_FLUSH) flags |= CHECK_CAPS_FLUSH; retry: @@ -1969,14 +1979,15 @@ retry: } dout("check_caps %llx.%llx file_want %s used %s dirty %s flushing %s" - " issued %s revoking %s retain %s %s%s\n", ceph_vinop(inode), + " issued %s revoking %s retain %s %s%s%s\n", ceph_vinop(inode), ceph_cap_string(file_wanted), ceph_cap_string(used), ceph_cap_string(ci->i_dirty_caps), ceph_cap_string(ci->i_flushing_caps), ceph_cap_string(issued), ceph_cap_string(revoking), ceph_cap_string(retain), (flags & CHECK_CAPS_AUTHONLY) ? " AUTHONLY" : "", - (flags & CHECK_CAPS_FLUSH) ? " FLUSH" : ""); + (flags & CHECK_CAPS_FLUSH) ? " FLUSH" : "", + (flags & CHECK_CAPS_NOINVAL) ? " NOINVAL" : ""); /* * If we no longer need to hold onto old our caps, and we may @@ -2055,10 +2066,27 @@ retry: } /* completed revocation? going down and there are no caps? */ - if (revoking && (revoking & cap_used) == 0) { - dout("completed revocation of %s\n", - ceph_cap_string(cap->implemented & ~cap->issued)); - goto ack; + if (revoking) { + if ((revoking & cap_used) == 0) { + dout("completed revocation of %s\n", + ceph_cap_string(cap->implemented & ~cap->issued)); + goto ack; + } + + /* + * If the "i_wrbuffer_ref" was increased by mmap or generic + * cache write just before the ceph_check_caps() is called, + * the Fb capability revoking will fail this time. Then we + * must wait for the BDI's delayed work to flush the dirty + * pages and to release the "i_wrbuffer_ref", which will cost + * at most 5 seconds. That means the MDS needs to wait at + * most 5 seconds to finished the Fb capability's revocation. + * + * Let's queue a writeback for it. + */ + if (S_ISREG(inode->i_mode) && ci->i_wrbuffer_ref && + (revoking & CEPH_CAP_FILE_BUFFER)) + queue_writeback = true; } /* want more caps from mds? */ @@ -2128,6 +2156,8 @@ ack: spin_unlock(&ci->i_ceph_lock); ceph_put_mds_session(session); + if (queue_writeback) + ceph_queue_writeback(inode); if (queue_invalidate) ceph_queue_invalidate(inode); } @@ -2211,13 +2241,14 @@ static int caps_are_flushed(struct inode *inode, u64 flush_tid) } /* - * wait for any unsafe requests to complete. + * flush the mdlog and wait for any unsafe requests to complete. */ -static int unsafe_request_wait(struct inode *inode) +static int flush_mdlog_and_wait_inode_unsafe_requests(struct inode *inode) { struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc; struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_mds_request *req1 = NULL, *req2 = NULL; + unsigned int max_sessions; int ret, err = 0; spin_lock(&ci->i_unsafe_lock); @@ -2236,36 +2267,46 @@ static int unsafe_request_wait(struct inode *inode) spin_unlock(&ci->i_unsafe_lock); /* + * The mdsc->max_sessions is unlikely to be changed + * mostly, here we will retry it by reallocating the + * sessions array memory to get rid of the mdsc->mutex + * lock. + */ +retry: + max_sessions = mdsc->max_sessions; + + /* * Trigger to flush the journal logs in all the relevant MDSes * manually, or in the worst case we must wait at most 5 seconds * to wait the journal logs to be flushed by the MDSes periodically. */ - if (req1 || req2) { + if ((req1 || req2) && likely(max_sessions)) { struct ceph_mds_session **sessions = NULL; struct ceph_mds_session *s; struct ceph_mds_request *req; - unsigned int max; int i; - /* - * The mdsc->max_sessions is unlikely to be changed - * mostly, here we will retry it by reallocating the - * sessions arrary memory to get rid of the mdsc->mutex - * lock. - */ -retry: - max = mdsc->max_sessions; - sessions = krealloc(sessions, max * sizeof(s), __GFP_ZERO); - if (!sessions) - return -ENOMEM; + sessions = kcalloc(max_sessions, sizeof(s), GFP_KERNEL); + if (!sessions) { + err = -ENOMEM; + goto out; + } spin_lock(&ci->i_unsafe_lock); if (req1) { list_for_each_entry(req, &ci->i_unsafe_dirops, r_unsafe_dir_item) { s = req->r_session; - if (unlikely(s->s_mds >= max)) { + if (!s) + continue; + if (unlikely(s->s_mds >= max_sessions)) { spin_unlock(&ci->i_unsafe_lock); + for (i = 0; i < max_sessions; i++) { + s = sessions[i]; + if (s) + ceph_put_mds_session(s); + } + kfree(sessions); goto retry; } if (!sessions[s->s_mds]) { @@ -2278,8 +2319,16 @@ retry: list_for_each_entry(req, &ci->i_unsafe_iops, r_unsafe_target_item) { s = req->r_session; - if (unlikely(s->s_mds >= max)) { + if (!s) + continue; + if (unlikely(s->s_mds >= max_sessions)) { spin_unlock(&ci->i_unsafe_lock); + for (i = 0; i < max_sessions; i++) { + s = sessions[i]; + if (s) + ceph_put_mds_session(s); + } + kfree(sessions); goto retry; } if (!sessions[s->s_mds]) { @@ -2300,7 +2349,7 @@ retry: spin_unlock(&ci->i_ceph_lock); /* send flush mdlog request to MDSes */ - for (i = 0; i < max; i++) { + for (i = 0; i < max_sessions; i++) { s = sessions[i]; if (s) { send_flush_mdlog(s); @@ -2310,22 +2359,26 @@ retry: kfree(sessions); } - dout("unsafe_request_wait %p wait on tid %llu %llu\n", + dout("%s %p wait on tid %llu %llu\n", __func__, inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL); if (req1) { ret = !wait_for_completion_timeout(&req1->r_safe_completion, ceph_timeout_jiffies(req1->r_timeout)); if (ret) err = -EIO; - ceph_mdsc_put_request(req1); } if (req2) { ret = !wait_for_completion_timeout(&req2->r_safe_completion, ceph_timeout_jiffies(req2->r_timeout)); if (ret) err = -EIO; - ceph_mdsc_put_request(req2); } + +out: + if (req1) + ceph_mdsc_put_request(req1); + if (req2) + ceph_mdsc_put_request(req2); return err; } @@ -2350,7 +2403,7 @@ int ceph_fsync(struct file *file, loff_t start, loff_t end, int datasync) dirty = try_flush_caps(inode, &flush_tid); dout("fsync dirty caps are %s\n", ceph_cap_string(dirty)); - err = unsafe_request_wait(inode); + err = flush_mdlog_and_wait_inode_unsafe_requests(inode); /* * only wait on non-file metadata writeback (the mds @@ -2388,7 +2441,11 @@ int ceph_write_inode(struct inode *inode, struct writeback_control *wbc) int wait = (wbc->sync_mode == WB_SYNC_ALL && !wbc->for_sync); dout("write_inode %p wait=%d\n", inode, wait); + ceph_fscache_unpin_writeback(inode, wbc); if (wait) { + err = ceph_wait_on_async_create(inode); + if (err) + return err; dirty = try_flush_caps(inode, &flush_tid); if (dirty) err = wait_event_interruptible(ci->i_cap_wq, @@ -2412,13 +2469,17 @@ static void __kick_flushing_caps(struct ceph_mds_client *mdsc, __releases(ci->i_ceph_lock) __acquires(ci->i_ceph_lock) { - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; struct ceph_cap *cap; struct ceph_cap_flush *cf; int ret; u64 first_tid = 0; u64 last_snap_flush = 0; + /* Don't do anything until create reply comes in */ + if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE) + return; + ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH; list_for_each_entry_reverse(cf, &ci->i_cap_flush_list, i_list) { @@ -2501,7 +2562,7 @@ void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, cap = ci->i_auth_cap; if (!(cap && cap->session == session)) { pr_err("%p auth cap %p not mds%d ???\n", - &ci->vfs_inode, cap, session->s_mds); + &ci->netfs.inode, cap, session->s_mds); spin_unlock(&ci->i_ceph_lock); continue; } @@ -2551,7 +2612,7 @@ void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, cap = ci->i_auth_cap; if (!(cap && cap->session == session)) { pr_err("%p auth cap %p not mds%d ???\n", - &ci->vfs_inode, cap, session->s_mds); + &ci->netfs.inode, cap, session->s_mds); spin_unlock(&ci->i_ceph_lock); continue; } @@ -2571,7 +2632,7 @@ void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, lockdep_assert_held(&ci->i_ceph_lock); - dout("%s %p flushing %s\n", __func__, &ci->vfs_inode, + dout("%s %p flushing %s\n", __func__, &ci->netfs.inode, ceph_cap_string(ci->i_flushing_caps)); if (!list_empty(&ci->i_cap_flush_list)) { @@ -2614,10 +2675,10 @@ void ceph_take_cap_refs(struct ceph_inode_info *ci, int got, } if (got & CEPH_CAP_FILE_BUFFER) { if (ci->i_wb_ref == 0) - ihold(&ci->vfs_inode); + ihold(&ci->netfs.inode); ci->i_wb_ref++; dout("%s %p wb %d -> %d (?)\n", __func__, - &ci->vfs_inode, ci->i_wb_ref-1, ci->i_wb_ref); + &ci->netfs.inode, ci->i_wb_ref-1, ci->i_wb_ref); } } @@ -2699,13 +2760,17 @@ again: * on transition from wanted -> needed caps. This is needed * for WRBUFFER|WR -> WR to avoid a new WR sync write from * going before a prior buffered writeback happens. + * + * For RDCACHE|RD -> RD, there is not need to wait and we can + * just exclude the revoking caps and force to sync read. */ int not = want & ~(have & need); int revoking = implemented & ~have; + int exclude = revoking & not; dout("get_cap_refs %p have %s but not %s (revoking %s)\n", inode, ceph_cap_string(have), ceph_cap_string(not), ceph_cap_string(revoking)); - if ((revoking & not) == 0) { + if (!exclude || !(exclude & CEPH_CAP_FILE_BUFFER)) { if (!snap_rwsem_locked && !ci->i_head_snapc && (need & CEPH_CAP_FILE_WR)) { @@ -2727,7 +2792,7 @@ again: snap_rwsem_locked = true; } if ((have & want) == want) - *got = need | want; + *got = need | (want & ~exclude); else *got = need; ceph_take_cap_refs(ci, *got, true); @@ -2945,8 +3010,8 @@ int ceph_get_caps(struct file *filp, int need, int want, loff_t endoff, int *got return ret; } - if (S_ISREG(ci->vfs_inode.i_mode) && - ci->i_inline_version != CEPH_INLINE_NONE && + if (S_ISREG(ci->netfs.inode.i_mode) && + ceph_has_inline_data(ci) && (_got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) && i_size_read(inode) > 0) { struct page *page = @@ -3035,7 +3100,7 @@ enum put_cap_refs_mode { static void __ceph_put_cap_refs(struct ceph_inode_info *ci, int had, enum put_cap_refs_mode mode) { - struct inode *inode = &ci->vfs_inode; + struct inode *inode = &ci->netfs.inode; int last = 0, put = 0, flushsnaps = 0, wake = 0; bool check_flushsnaps = false; @@ -3143,11 +3208,10 @@ void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, int had) void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, struct ceph_snap_context *snapc) { - struct inode *inode = &ci->vfs_inode; - struct ceph_cap_snap *capsnap = NULL; + struct inode *inode = &ci->netfs.inode; + struct ceph_cap_snap *capsnap = NULL, *iter; int put = 0; bool last = false; - bool found = false; bool flush_snaps = false; bool complete_capsnap = false; @@ -3174,14 +3238,14 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head, last ? " LAST" : ""); } else { - list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) { - if (capsnap->context == snapc) { - found = true; + list_for_each_entry(iter, &ci->i_cap_snaps, ci_item) { + if (iter->context == snapc) { + capsnap = iter; break; } } - if (!found) { + if (!capsnap) { /* * The capsnap should already be removed when removing * auth cap in the case of a forced unmount. @@ -3375,8 +3439,7 @@ static void handle_cap_grant(struct inode *inode, if ((newcaps & CEPH_CAP_LINK_SHARED) && (extra_info->issued & CEPH_CAP_LINK_EXCL) == 0) { set_nlink(inode, le32_to_cpu(grant->nlink)); - if (inode->i_nlink == 0 && - (newcaps & (CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL))) + if (inode->i_nlink == 0) deleted_inode = true; } @@ -3492,6 +3555,9 @@ static void handle_cap_grant(struct inode *inode, check_caps = 1; /* check auth cap only */ else check_caps = 2; /* check all caps */ + /* If there is new caps, try to wake up the waiters */ + if (~cap->issued & newcaps) + wake = true; cap->issued = newcaps; cap->implemented |= newcaps; } else if (cap->issued == newcaps) { @@ -3521,24 +3587,23 @@ static void handle_cap_grant(struct inode *inode, fill_inline = true; } - if (ci->i_auth_cap == cap && - le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) { - if (newcaps & ~extra_info->issued) - wake = true; + if (le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) { + if (ci->i_auth_cap == cap) { + if (newcaps & ~extra_info->issued) + wake = true; - if (ci->i_requested_max_size > max_size || - !(le32_to_cpu(grant->wanted) & CEPH_CAP_ANY_FILE_WR)) { - /* re-request max_size if necessary */ - ci->i_requested_max_size = 0; - wake = true; - } + if (ci->i_requested_max_size > max_size || + !(le32_to_cpu(grant->wanted) & CEPH_CAP_ANY_FILE_WR)) { + /* re-request max_size if necessary */ + ci->i_requested_max_size = 0; + wake = true; + } - ceph_kick_flushing_inode_caps(session, ci); - spin_unlock(&ci->i_ceph_lock); + ceph_kick_flushing_inode_caps(session, ci); + } up_read(&session->s_mdsc->snap_rwsem); - } else { - spin_unlock(&ci->i_ceph_lock); } + spin_unlock(&ci->i_ceph_lock); if (fill_inline) ceph_fill_inline_data(inode, NULL, extra_info->inline_data, @@ -3641,7 +3706,7 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid, session->s_mds, &list_first_entry(&session->s_cap_flushing, struct ceph_inode_info, - i_flushing_item)->vfs_inode); + i_flushing_item)->netfs.inode); } } mdsc->num_cap_flushing--; @@ -3732,8 +3797,7 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid, struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc; u64 follows = le64_to_cpu(m->snap_follows); - struct ceph_cap_snap *capsnap; - bool flushed = false; + struct ceph_cap_snap *capsnap = NULL, *iter; bool wake_ci = false; bool wake_mdsc = false; @@ -3741,26 +3805,26 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid, inode, ci, session->s_mds, follows); spin_lock(&ci->i_ceph_lock); - list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) { - if (capsnap->follows == follows) { - if (capsnap->cap_flush.tid != flush_tid) { + list_for_each_entry(iter, &ci->i_cap_snaps, ci_item) { + if (iter->follows == follows) { + if (iter->cap_flush.tid != flush_tid) { dout(" cap_snap %p follows %lld tid %lld !=" - " %lld\n", capsnap, follows, - flush_tid, capsnap->cap_flush.tid); + " %lld\n", iter, follows, + flush_tid, iter->cap_flush.tid); break; } - flushed = true; + capsnap = iter; break; } else { dout(" skipping cap_snap %p follows %lld\n", - capsnap, capsnap->follows); + iter, iter->follows); } } - if (flushed) + if (capsnap) ceph_remove_capsnap(inode, capsnap, &wake_ci, &wake_mdsc); spin_unlock(&ci->i_ceph_lock); - if (flushed) { + if (capsnap) { ceph_put_snap_context(capsnap->context); ceph_put_cap_snap(capsnap); if (wake_ci) @@ -3837,6 +3901,7 @@ static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex, dout("handle_cap_export inode %p ci %p mds%d mseq %d target %d\n", inode, ci, mds, mseq, target); retry: + down_read(&mdsc->snap_rwsem); spin_lock(&ci->i_ceph_lock); cap = __get_cap_for_mds(ci, mds); if (!cap || cap->cap_id != le64_to_cpu(ex->cap_id)) @@ -3900,6 +3965,7 @@ retry: } spin_unlock(&ci->i_ceph_lock); + up_read(&mdsc->snap_rwsem); mutex_unlock(&session->s_mutex); /* open target session */ @@ -3925,6 +3991,7 @@ retry: out_unlock: spin_unlock(&ci->i_ceph_lock); + up_read(&mdsc->snap_rwsem); mutex_unlock(&session->s_mutex); if (tsession) { mutex_unlock(&tsession->s_mutex); @@ -4133,7 +4200,6 @@ void ceph_handle_caps(struct ceph_mds_session *session, /* lookup ino */ inode = ceph_find_inode(mdsc->fsc->sb, vino); - ci = ceph_inode(inode); dout(" op %s ino %llx.%llx inode %p\n", ceph_cap_op_name(op), vino.ino, vino.snap, inode); @@ -4159,6 +4225,7 @@ void ceph_handle_caps(struct ceph_mds_session *session, } goto flush_cap_releases; } + ci = ceph_inode(inode); /* these will work even if we don't have a cap yet */ switch (op) { @@ -4286,7 +4353,7 @@ unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc) break; list_del_init(&ci->i_cap_delay_list); - inode = igrab(&ci->vfs_inode); + inode = igrab(&ci->netfs.inode); if (inode) { spin_unlock(&mdsc->cap_delay_lock); dout("check_delayed_caps on %p\n", inode); @@ -4314,10 +4381,11 @@ static void flush_dirty_session_caps(struct ceph_mds_session *s) while (!list_empty(&s->s_cap_dirty)) { ci = list_first_entry(&s->s_cap_dirty, struct ceph_inode_info, i_dirty_item); - inode = &ci->vfs_inode; + inode = &ci->netfs.inode; ihold(inode); dout("flush_dirty_caps %llx.%llx\n", ceph_vinop(inode)); spin_unlock(&mdsc->cap_dirty_lock); + ceph_wait_on_async_create(inode); ceph_check_caps(ci, CHECK_CAPS_FLUSH, NULL); iput(inode); spin_lock(&mdsc->cap_dirty_lock); @@ -4348,9 +4416,9 @@ void __ceph_touch_fmode(struct ceph_inode_info *ci, void ceph_get_fmode(struct ceph_inode_info *ci, int fmode, int count) { - struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->vfs_inode.i_sb); + struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->netfs.inode.i_sb); int bits = (fmode << 1) | 1; - bool is_opened = false; + bool already_opened = false; int i; if (count == 1) @@ -4358,19 +4426,19 @@ void ceph_get_fmode(struct ceph_inode_info *ci, int fmode, int count) spin_lock(&ci->i_ceph_lock); for (i = 0; i < CEPH_FILE_MODE_BITS; i++) { - if (bits & (1 << i)) - ci->i_nr_by_mode[i] += count; - /* - * If any of the mode ref is larger than 1, + * If any of the mode ref is larger than 0, * that means it has been already opened by * others. Just skip checking the PIN ref. */ - if (i && ci->i_nr_by_mode[i] > 1) - is_opened = true; + if (i && ci->i_nr_by_mode[i]) + already_opened = true; + + if (bits & (1 << i)) + ci->i_nr_by_mode[i] += count; } - if (!is_opened) + if (!already_opened) percpu_counter_inc(&mdsc->metric.opened_inodes); spin_unlock(&ci->i_ceph_lock); } @@ -4382,7 +4450,7 @@ void ceph_get_fmode(struct ceph_inode_info *ci, int fmode, int count) */ void ceph_put_fmode(struct ceph_inode_info *ci, int fmode, int count) { - struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->vfs_inode.i_sb); + struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->netfs.inode.i_sb); int bits = (fmode << 1) | 1; bool is_closed = true; int i; @@ -4597,7 +4665,7 @@ int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invali lockdep_assert_held(&ci->i_ceph_lock); dout("removing cap %p, ci is %p, inode is %p\n", - cap, ci, &ci->vfs_inode); + cap, ci, &ci->netfs.inode); is_auth = (cap == ci->i_auth_cap); __ceph_remove_cap(cap, false); |