Merge tag 'ceph-for-6.5-rc1' of https://github.com/ceph/ceph-client

Pull ceph updates from Ilya Dryomov:
 "A bunch of CephFS fixups from Xiubo, mostly around dropping caps,
  along with a fix for a regression in the readahead handling code which
  sneaked in with the switch to netfs helpers"

* tag 'ceph-for-6.5-rc1' of https://github.com/ceph/ceph-client:
  ceph: don't let check_caps skip sending responses for revoke msgs
  ceph: issue a cap release immediately if no cap exists
  ceph: trigger to flush the buffer when making snapshot
  ceph: fix blindly expanding the readahead windows
  ceph: add a dedicated private data for netfs rreq
  ceph: voluntarily drop Xx caps for requests those touch parent mtime
  ceph: try to dump the msgs when decoding fails
  ceph: only send metrics when the MDS rank is ready

8 files changed, 154 insertions, 40 deletions

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index 6bb251a4d613e..59cbfb80edbda 100644
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -187,16 +187,42 @@ static void ceph_netfs_expand_readahead(struct netfs_io_request *rreq)
 	struct inode *inode = rreq->inode;
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	struct ceph_file_layout *lo = &ci->i_layout;
+	unsigned long max_pages = inode->i_sb->s_bdi->ra_pages;
+	loff_t end = rreq->start + rreq->len, new_end;
+	struct ceph_netfs_request_data *priv = rreq->netfs_priv;
+	unsigned long max_len;
 	u32 blockoff;
-	u64 blockno;
 
-	/* Expand the start downward */
-	blockno = div_u64_rem(rreq->start, lo->stripe_unit, &blockoff);
-	rreq->start = blockno * lo->stripe_unit;
-	rreq->len += blockoff;
+	if (priv) {
+		/* Readahead is disabled by posix_fadvise POSIX_FADV_RANDOM */
+		if (priv->file_ra_disabled)
+			max_pages = 0;
+		else
+			max_pages = priv->file_ra_pages;
+
+	}
 
-	/* Now, round up the length to the next block */
-	rreq->len = roundup(rreq->len, lo->stripe_unit);
+	/* Readahead is disabled */
+	if (!max_pages)
+		return;
+
+	max_len = max_pages << PAGE_SHIFT;
+
+	/*
+	 * Try to expand the length forward by rounding up it to the next
+	 * block, but do not exceed the file size, unless the original
+	 * request already exceeds it.
+	 */
+	new_end = min(round_up(end, lo->stripe_unit), rreq->i_size);
+	if (new_end > end && new_end <= rreq->start + max_len)
+		rreq->len = new_end - rreq->start;
+
+	/* Try to expand the start downward */
+	div_u64_rem(rreq->start, lo->stripe_unit, &blockoff);
+	if (rreq->len + blockoff <= max_len) {
+		rreq->start -= blockoff;
+		rreq->len += blockoff;
+	}
 }
 
 static bool ceph_netfs_clamp_length(struct netfs_io_subrequest *subreq)
@@ -362,18 +388,28 @@ static int ceph_init_request(struct netfs_io_request *rreq, struct file *file)
 {
 	struct inode *inode = rreq->inode;
 	int got = 0, want = CEPH_CAP_FILE_CACHE;
+	struct ceph_netfs_request_data *priv;
 	int ret = 0;
 
 	if (rreq->origin != NETFS_READAHEAD)
 		return 0;
 
+	priv = kzalloc(sizeof(*priv), GFP_NOFS);
+	if (!priv)
+		return -ENOMEM;
+
 	if (file) {
 		struct ceph_rw_context *rw_ctx;
 		struct ceph_file_info *fi = file->private_data;
 
+		priv->file_ra_pages = file->f_ra.ra_pages;
+		priv->file_ra_disabled = file->f_mode & FMODE_RANDOM;
+
 		rw_ctx = ceph_find_rw_context(fi);
-		if (rw_ctx)
+		if (rw_ctx) {
+			rreq->netfs_priv = priv;
 			return 0;
+		}
 	}
 
 	/*
@@ -383,27 +419,40 @@ static int ceph_init_request(struct netfs_io_request *rreq, struct file *file)
 	ret = ceph_try_get_caps(inode, CEPH_CAP_FILE_RD, want, true, &got);
 	if (ret < 0) {
 		dout("start_read %p, error getting cap\n", inode);
-		return ret;
+		goto out;
 	}
 
 	if (!(got & want)) {
 		dout("start_read %p, no cache cap\n", inode);
-		return -EACCES;
+		ret = -EACCES;
+		goto out;
+	}
+	if (ret == 0) {
+		ret = -EACCES;
+		goto out;
 	}
-	if (ret == 0)
-		return -EACCES;
 
-	rreq->netfs_priv = (void *)(uintptr_t)got;
-	return 0;
+	priv->caps = got;
+	rreq->netfs_priv = priv;
+
+out:
+	if (ret < 0)
+		kfree(priv);
+
+	return ret;
 }
 
 static void ceph_netfs_free_request(struct netfs_io_request *rreq)
 {
-	struct ceph_inode_info *ci = ceph_inode(rreq->inode);
-	int got = (uintptr_t)rreq->netfs_priv;
+	struct ceph_netfs_request_data *priv = rreq->netfs_priv;
+
+	if (!priv)
+		return;
 
-	if (got)
-		ceph_put_cap_refs(ci, got);
+	if (priv->caps)
+		ceph_put_cap_refs(ceph_inode(rreq->inode), priv->caps);
+	kfree(priv);
+	rreq->netfs_priv = NULL;
 }
 
 const struct netfs_request_ops ceph_netfs_ops = {
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index 2321e5ddb664d..e2bb0d0072da5 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -3109,6 +3109,12 @@ static void __ceph_put_cap_refs(struct ceph_inode_info *ci, int had,
 	}
 	if (had & CEPH_CAP_FILE_WR) {
 		if (--ci->i_wr_ref == 0) {
+			/*
+			 * The Fb caps will always be took and released
+			 * together with the Fw caps.
+			 */
+			WARN_ON_ONCE(ci->i_wb_ref);
+
 			last++;
 			check_flushsnaps = true;
 			if (ci->i_wrbuffer_ref_head == 0 &&
@@ -3560,6 +3566,15 @@ static void handle_cap_grant(struct inode *inode,
 	}
 	BUG_ON(cap->issued & ~cap->implemented);
 
+	/* don't let check_caps skip sending a response to MDS for revoke msgs */
+	if (le32_to_cpu(grant->op) == CEPH_CAP_OP_REVOKE) {
+		cap->mds_wanted = 0;
+		if (cap == ci->i_auth_cap)
+			check_caps = 1; /* check auth cap only */
+		else
+			check_caps = 2; /* check all caps */
+	}
+
 	if (extra_info->inline_version > 0 &&
 	    extra_info->inline_version >= ci->i_inline_version) {
 		ci->i_inline_version = extra_info->inline_version;
@@ -4086,6 +4101,7 @@ void ceph_handle_caps(struct ceph_mds_session *session,
 	struct cap_extra_info extra_info = {};
 	bool queue_trunc;
 	bool close_sessions = false;
+	bool do_cap_release = false;
 
 	dout("handle_caps from mds%d\n", session->s_mds);
 
@@ -4192,17 +4208,14 @@ void ceph_handle_caps(struct ceph_mds_session *session,
 	if (!inode) {
 		dout(" i don't have ino %llx\n", vino.ino);
 
-		if (op == CEPH_CAP_OP_IMPORT) {
-			cap = ceph_get_cap(mdsc, NULL);
-			cap->cap_ino = vino.ino;
-			cap->queue_release = 1;
-			cap->cap_id = le64_to_cpu(h->cap_id);
-			cap->mseq = mseq;
-			cap->seq = seq;
-			cap->issue_seq = seq;
-			spin_lock(&session->s_cap_lock);
-			__ceph_queue_cap_release(session, cap);
-			spin_unlock(&session->s_cap_lock);
+		switch (op) {
+		case CEPH_CAP_OP_IMPORT:
+		case CEPH_CAP_OP_REVOKE:
+		case CEPH_CAP_OP_GRANT:
+			do_cap_release = true;
+			break;
+		default:
+			break;
 		}
 		goto flush_cap_releases;
 	}
@@ -4252,6 +4265,14 @@ void ceph_handle_caps(struct ceph_mds_session *session,
 		     inode, ceph_ino(inode), ceph_snap(inode),
 		     session->s_mds);
 		spin_unlock(&ci->i_ceph_lock);
+		switch (op) {
+		case CEPH_CAP_OP_REVOKE:
+		case CEPH_CAP_OP_GRANT:
+			do_cap_release = true;
+			break;
+		default:
+			break;
+		}
 		goto flush_cap_releases;
 	}
 
@@ -4302,6 +4323,18 @@ flush_cap_releases:
 	 * along for the mds (who clearly thinks we still have this
 	 * cap).
 	 */
+	if (do_cap_release) {
+		cap = ceph_get_cap(mdsc, NULL);
+		cap->cap_ino = vino.ino;
+		cap->queue_release = 1;
+		cap->cap_id = le64_to_cpu(h->cap_id);
+		cap->mseq = mseq;
+		cap->seq = seq;
+		cap->issue_seq = seq;
+		spin_lock(&session->s_cap_lock);
+		__ceph_queue_cap_release(session, cap);
+		spin_unlock(&session->s_cap_lock);
+	}
 	ceph_flush_cap_releases(mdsc, session);
 	goto done;
 
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index cb67ac821f0ee..4a2b39d9a61a2 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -886,7 +886,8 @@ static int ceph_mknod(struct mnt_idmap *idmap, struct inode *dir,
 	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_args.mknod.mode = cpu_to_le32(mode);
 	req->r_args.mknod.rdev = cpu_to_le32(rdev);
-	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL |
+			     CEPH_CAP_XATTR_EXCL;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	if (as_ctx.pagelist) {
 		req->r_pagelist = as_ctx.pagelist;
@@ -953,7 +954,8 @@ static int ceph_symlink(struct mnt_idmap *idmap, struct inode *dir,
 	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
-	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL |
+			     CEPH_CAP_XATTR_EXCL;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	if (as_ctx.pagelist) {
 		req->r_pagelist = as_ctx.pagelist;
@@ -1022,7 +1024,8 @@ static int ceph_mkdir(struct mnt_idmap *idmap, struct inode *dir,
 	ihold(dir);
 	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_args.mkdir.mode = cpu_to_le32(mode);
-	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL |
+			     CEPH_CAP_XATTR_EXCL;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	if (as_ctx.pagelist) {
 		req->r_pagelist = as_ctx.pagelist;
@@ -1079,7 +1082,7 @@ static int ceph_link(struct dentry *old_dentry, struct inode *dir,
 	req->r_parent = dir;
 	ihold(dir);
 	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
-	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_XATTR_EXCL;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	/* release LINK_SHARED on source inode (mds will lock it) */
 	req->r_old_inode_drop = CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL;
@@ -1218,7 +1221,7 @@ retry:
 	req->r_num_caps = 2;
 	req->r_parent = dir;
 	ihold(dir);
-	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_XATTR_EXCL;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	req->r_inode_drop = ceph_drop_caps_for_unlink(inode);
 
@@ -1320,9 +1323,9 @@ static int ceph_rename(struct mnt_idmap *idmap, struct inode *old_dir,
 	req->r_parent = new_dir;
 	ihold(new_dir);
 	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
-	req->r_old_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_old_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_XATTR_EXCL;
 	req->r_old_dentry_unless = CEPH_CAP_FILE_EXCL;
-	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_XATTR_EXCL;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	/* release LINK_RDCACHE on source inode (mds will lock it) */
 	req->r_old_inode_drop = CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL;
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index b1925232dc082..63efe5389783c 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -791,7 +791,8 @@ retry:
 	if (flags & O_CREAT) {
 		struct ceph_file_layout lo;
 
-		req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL;
+		req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL |
+				     CEPH_CAP_XATTR_EXCL;
 		req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 		if (as_ctx.pagelist) {
 			req->r_pagelist = as_ctx.pagelist;
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 4c0f22acf53d2..66048a86c480c 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -645,6 +645,7 @@ bad:
 	err = -EIO;
 out_bad:
 	pr_err("mds parse_reply err %d\n", err);
+	ceph_msg_dump(msg);
 	return err;
 }
 
@@ -3538,6 +3539,7 @@ static void handle_forward(struct ceph_mds_client *mdsc,
 
 bad:
 	pr_err("mdsc_handle_forward decode error err=%d\n", err);
+	ceph_msg_dump(msg);
 }
 
 static int __decode_session_metadata(void **p, void *end,
@@ -5258,6 +5260,7 @@ void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
 bad:
 	pr_err("error decoding fsmap %d. Shutting down mount.\n", err);
 	ceph_umount_begin(mdsc->fsc->sb);
+	ceph_msg_dump(msg);
 err_out:
 	mutex_lock(&mdsc->mutex);
 	mdsc->mdsmap_err = err;
@@ -5326,6 +5329,7 @@ bad_unlock:
 bad:
 	pr_err("error decoding mdsmap %d. Shutting down mount.\n", err);
 	ceph_umount_begin(mdsc->fsc->sb);
+	ceph_msg_dump(msg);
 	return;
 }
 
diff --git a/fs/ceph/metric.c b/fs/ceph/metric.c
index c47347d2e84e3..cce78d769f551 100644
--- a/fs/ceph/metric.c
+++ b/fs/ceph/metric.c
@@ -36,6 +36,14 @@ static bool ceph_mdsc_send_metrics(struct ceph_mds_client *mdsc,
 	s32 items = 0;
 	s32 len;
 
+	/* Do not send the metrics until the MDS rank is ready */
+	mutex_lock(&mdsc->mutex);
+	if (ceph_mdsmap_get_state(mdsc->mdsmap, s->s_mds) != CEPH_MDS_STATE_ACTIVE) {
+		mutex_unlock(&mdsc->mutex);
+		return false;
+	}
+	mutex_unlock(&mdsc->mutex);
+
 	len = sizeof(*head) + sizeof(*cap) + sizeof(*read) + sizeof(*write)
 	      + sizeof(*meta) + sizeof(*dlease) + sizeof(*files)
 	      + sizeof(*icaps) + sizeof(*inodes) + sizeof(*rsize)
diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
index 2e73ba62bd7aa..343d738448dcd 100644
--- a/fs/ceph/snap.c
+++ b/fs/ceph/snap.c
@@ -675,14 +675,17 @@ int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 		return 0;
 	}
 
-	/* Fb cap still in use, delay it */
-	if (ci->i_wb_ref) {
+	/*
+	 * Defer flushing the capsnap if the dirty buffer not flushed yet.
+	 * And trigger to flush the buffer immediately.
+	 */
+	if (ci->i_wrbuffer_ref) {
 		dout("%s %p %llx.%llx cap_snap %p snapc %p %llu %s s=%llu "
 		     "used WRBUFFER, delaying\n", __func__, inode,
 		     ceph_vinop(inode), capsnap, capsnap->context,
 		     capsnap->context->seq, ceph_cap_string(capsnap->dirty),
 		     capsnap->size);
-		capsnap->writing = 1;
+		ceph_queue_writeback(inode);
 		return 0;
 	}
 
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index d24bf0db52346..3bfddf34d488b 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -451,6 +451,19 @@ struct ceph_inode_info {
 	unsigned long  i_work_mask;
 };
 
+struct ceph_netfs_request_data {
+	int caps;
+
+	/*
+	 * Maximum size of a file readahead request.
+	 * The fadvise could update the bdi's default ra_pages.
+	 */
+	unsigned int file_ra_pages;
+
+	/* Set it if fadvise disables file readahead entirely */
+	bool file_ra_disabled;
+};
+
 static inline struct ceph_inode_info *
 ceph_inode(const struct inode *inode)
 {