From: Sage Weil <sage@newdream.net>
Date: Fri, 23 Jan 2009 22:50:13 +0000 (-0800)
Subject: kclient: do sync writepages if FILE_WRBUFFER is being revoked
X-Git-Tag: v0.7~326
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=c4760afe685156680f7e904f3ef9f540883a2b41;p=ceph.git

kclient: do sync writepages if FILE_WRBUFFER is being revoked

This doesn't capture the case where we already started writeout
and _then_ the MDS tries to revoke, but that's probably not worth
the complexity.
---

diff --git a/src/kernel/addr.c b/src/kernel/addr.c
index eeb3cbf07d5d9..5eb529b72e931 100644
--- a/src/kernel/addr.c
+++ b/src/kernel/addr.c
@@ -545,6 +545,9 @@ static int ceph_writepages_start(struct address_space *mapping,
 	struct ceph_osd_request *req = NULL;
 	int do_sync = atomic_read(&ci->i_want_sync_writeout);
 
+	if (ceph_caps_revoking(ci) & CEPH_CAP_FILE_WRBUFFER)
+		do_sync = 1;
+
 	client = ceph_inode_to_client(inode);
 	if (client->mount_state == CEPH_MOUNT_SHUTDOWN) {
 		dout(1, "writepage on forced umount\n");
diff --git a/src/kernel/caps.c b/src/kernel/caps.c
index da1ae54dcb912..e277b091aac4b 100644
--- a/src/kernel/caps.c
+++ b/src/kernel/caps.c
@@ -276,6 +276,38 @@ retry:
 	return 0;
 }
 
+/*
+ * Return true if cap has not timed out and belongs to the current
+ * generation of the MDS session.
+ */
+static int __cap_is_valid(struct ceph_cap *cap)
+{
+	unsigned long ttl;
+	u32 gen;
+
+	spin_lock(&cap->session->s_cap_lock);
+	gen = cap->session->s_cap_gen;
+	ttl = cap->session->s_cap_ttl;
+	spin_unlock(&cap->session->s_cap_lock);
+
+	if (cap->gen < gen || time_after_eq(jiffies, ttl)) {
+		dout(30, "__cap_is_valid %p cap %p issued %s "
+		     "but STALE (gen %u vs %u)\n", &cap->ci->vfs_inode,
+		     cap, ceph_cap_string(cap->issued), cap->gen, gen);
+		return 0;
+	}
+
+	if (time_after_eq(jiffies, cap->expires) &&
+	    (cap->issued & ~CEPH_CAP_EXPIREABLE) == 0) {
+		dout(30, "__cap_is_valid %p cap %p issued %s "
+		     "but readonly and expired\n", &cap->ci->vfs_inode,
+		     cap, ceph_cap_string(cap->issued));
+		return 0;
+	}
+
+	return 1;
+}
+
 /*
  * Return set of valid cap bits issued to us.  Note that caps time
  * out, and may be invalidated in bulk if the client session times out
@@ -285,35 +317,14 @@ int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented)
 {
 	int have = ci->i_snap_caps;
 	struct ceph_cap *cap;
-	u32 gen;
-	unsigned long ttl;
 	struct rb_node *p;
 
 	if (implemented)
 		*implemented = 0;
 	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
 		cap = rb_entry(p, struct ceph_cap, ci_node);
-
-		spin_lock(&cap->session->s_cap_lock);
-		gen = cap->session->s_cap_gen;
-		ttl = cap->session->s_cap_ttl;
-		spin_unlock(&cap->session->s_cap_lock);
-
-		if (cap->gen < gen || time_after_eq(jiffies, ttl)) {
-			dout(30, "__ceph_caps_issued %p cap %p issued %s "
-			     "but STALE (gen %u vs %u)\n", &ci->vfs_inode,
-			     cap, ceph_cap_string(cap->issued), cap->gen, gen);
-			continue;
-		}
-
-		if (time_after_eq(jiffies, cap->expires) &&
-		    (cap->issued & ~CEPH_CAP_EXPIREABLE) == 0) {
-			dout(30, "__ceph_caps_issued %p cap %p issued %s "
-			     "but readonly and expired\n", &ci->vfs_inode,
-			     cap, ceph_cap_string(cap->issued));
+		if (!__cap_is_valid(cap))
 			continue;
-		}
-
 		dout(30, "__ceph_caps_issued %p cap %p issued %s\n",
 		     &ci->vfs_inode, cap, ceph_cap_string(cap->issued));
 		have |= cap->issued;
@@ -336,6 +347,8 @@ int __ceph_caps_dirty(struct ceph_inode_info *ci)
 
 	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
 		cap = rb_entry(p, struct ceph_cap, ci_node);
+		if (!__cap_is_valid(cap))
+			continue;
 		dout(30, "__ceph_caps_dirty %p cap %p flushing %s\n",
 		     inode, cap, ceph_cap_string(cap->flushing));
 		dirty |= cap->flushing;
@@ -345,6 +358,28 @@ int __ceph_caps_dirty(struct ceph_inode_info *ci)
 	return dirty;
 }
 
+/*
+ * Return mask of caps currently being revoked by an MDS.
+ */
+int ceph_caps_revoking(struct ceph_inode_info *ci)
+{
+	struct inode *inode = &ci->vfs_inode;
+	struct ceph_cap *cap;
+	struct rb_node *p;
+	int revoking = 0;
+
+	spin_lock(&inode->i_lock);
+	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
+		cap = rb_entry(p, struct ceph_cap, ci_node);
+		if (__cap_is_valid(cap))
+			revoking |= cap->implemented & ~cap->issued;
+	}
+	spin_unlock(&inode->i_lock);
+	dout(30, "ceph_caps_revoking %p revoking %s\n", inode,
+	     ceph_cap_string(revoking));
+	return revoking;
+}
+
 /*
  * called under i_lock
  */
diff --git a/src/kernel/super.h b/src/kernel/super.h
index b6087e99cc088..79946318c78a7 100644
--- a/src/kernel/super.h
+++ b/src/kernel/super.h
@@ -435,6 +435,7 @@ static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 }
 
 extern int __ceph_caps_dirty(struct ceph_inode_info *ci);
+extern int ceph_caps_revoking(struct ceph_inode_info *ci);
 
 static inline int __ceph_caps_used(struct ceph_inode_info *ci)
 {