afs: Rewrite writeback handling

Rewrite the writeback handling to make the writeback records refcounted
separately from the completion management so that a ref can be taken on one
without preventing completion from happening.
Signed-off-by: David Howells <dhowells@redhat.com>

fs/afs/file.c

@@ -23,7 +23,6 @@ static int afs_readpage(struct file *file, struct page *page);
 static void afs_invalidatepage(struct page *page, unsigned int offset,
 			       unsigned int length);
 static int afs_releasepage(struct page *page, gfp_t gfp_flags);
-static int afs_launder_page(struct page *page);
 
 static int afs_readpages(struct file *filp, struct address_space *mapping,
 			 struct list_head *pages, unsigned nr_pages);
@@ -463,13 +462,23 @@ static int afs_readpages(struct file *file, struct address_space *mapping,
 }
 
 /*
- * write back a dirty page
+ * Try to remove a page from any writeback it may be entertaining.
  */
-static int afs_launder_page(struct page *page)
+static bool afs_remove_wb_from_page(struct afs_vnode *vnode, struct page *page,
+				    enum afs_writeback_trace why)
 {
-	_enter("{%lu}", page->index);
+	struct afs_writeback *wb;
+	bool cleared = true;
+
+	spin_lock(&vnode->writeback_lock);
+	wb = afs_get_writeback(vnode, (struct afs_writeback *)page_private(page), why);
+	spin_unlock(&vnode->writeback_lock);
+	if (wb) {
+		cleared = afs_writeback_remove_page(vnode, wb, page);
+		afs_put_writeback(vnode, wb, 1);
+	}
 
-	return 0;
+	return cleared;
 }
 
 /*
@@ -480,7 +489,6 @@ static int afs_launder_page(struct page *page)
 static void afs_invalidatepage(struct page *page, unsigned int offset,
 			       unsigned int length)
 {
-	struct afs_writeback *wb = (struct afs_writeback *) page_private(page);
 	struct afs_vnode *vnode = AFS_FS_I(page->mapping->host);
 
 	_enter("{%lu},%u,%u", page->index, offset, length);
@@ -496,15 +504,8 @@ static void afs_invalidatepage(struct page *page, unsigned int offset,
 		}
 #endif
 
-		if (PagePrivate(page)) {
-			if (wb && !PageWriteback(page)) {
-				set_page_private(page, 0);
-				afs_put_writeback(vnode, wb);
-			}
-
-			if (!page_private(page))
-				ClearPagePrivate(page);
-		}
+		afs_remove_wb_from_page(vnode, page,
+					afs_writeback_trace_invalidate_page);
 	}
 
 	_leave("");
@@ -516,8 +517,8 @@ static void afs_invalidatepage(struct page *page, unsigned int offset,
  */
 static int afs_releasepage(struct page *page, gfp_t gfp_flags)
 {
-	struct afs_writeback *wb = (struct afs_writeback *) page_private(page);
 	struct afs_vnode *vnode = AFS_FS_I(page->mapping->host);
+	bool cleared = true;
 
 	_enter("{{%x:%u}[%lu],%lx},%x",
 	       vnode->fid.vid, vnode->fid.vnode, page->index, page->flags,
@@ -532,15 +533,8 @@ static int afs_releasepage(struct page *page, gfp_t gfp_flags)
 	}
 #endif
 
-	if (PagePrivate(page)) {
-		if (wb) {
-			set_page_private(page, 0);
-			afs_put_writeback(vnode, wb);
-		}
-		ClearPagePrivate(page);
-	}
-
-	/* indicate that the page can be released */
-	_leave(" = T");
-	return 1;
+	cleared = afs_remove_wb_from_page(vnode, page,
+					  afs_writeback_trace_release_page);
+	_leave(" = %d", cleared);
+	return cleared;
 }

fs/afs/inode.c

@@ -474,10 +474,8 @@ int afs_setattr(struct dentry *dentry, struct iattr *attr)
 	}
 
 	/* flush any dirty data outstanding on a regular file */
-	if (S_ISREG(vnode->vfs_inode.i_mode)) {
-		filemap_write_and_wait(vnode->vfs_inode.i_mapping);
+	if (S_ISREG(vnode->vfs_inode.i_mode))
 		afs_writeback_all(vnode);
-	}
 
 	if (attr->ia_valid & ATTR_FILE) {
 		key = attr->ia_file->private_data;

fs/afs/internal.h

@@ -156,7 +156,8 @@ struct afs_writeback {
 	unsigned		offset_first;	/* offset into first page of start of write */
 	unsigned		to_last;	/* offset into last page of end of write */
 	int			num_conflicts;	/* count of conflicting writes in list */
-	int			usage;
+	atomic_t		usage;
+	int			nr_pages;	/* Number of pages contributing */
 	bool			conflicts;	/* T if has dependent conflicts */
 	enum {
 		AFS_WBACK_SYNCING,		/* synchronisation being performed */
@@ -411,6 +412,8 @@ struct afs_interface {
 	unsigned	mtu;		/* MTU of interface */
 };
 
+#include <trace/events/afs.h>
+
 /*****************************************************************************/
 /*
  * cache.c
@@ -709,7 +712,13 @@ extern int afs_volume_release_fileserver(struct afs_vnode *,
  * write.c
  */
 extern int afs_set_page_dirty(struct page *);
-extern void afs_put_writeback(struct afs_vnode *, struct afs_writeback *);
+extern struct afs_writeback *afs_get_writeback(struct afs_vnode *,
+					       struct afs_writeback *,
+					       enum afs_writeback_trace);
+extern void afs_put_writeback(struct afs_vnode *, struct afs_writeback *,
+			      unsigned);
+extern bool afs_writeback_remove_page(struct afs_vnode *,
+				      struct afs_writeback *, struct page *);
 extern int afs_write_begin(struct file *file, struct address_space *mapping,
 			loff_t pos, unsigned len, unsigned flags,
 			struct page **pagep, void **fsdata);
@@ -723,14 +732,12 @@ extern ssize_t afs_file_write(struct kiocb *, struct iov_iter *);
 extern int afs_writeback_all(struct afs_vnode *);
 extern int afs_flush(struct file *, fl_owner_t);
 extern int afs_fsync(struct file *, loff_t, loff_t, int);
-
+extern int afs_launder_page(struct page *);
 
 /*****************************************************************************/
 /*
  * debug tracing
  */
-#include <trace/events/afs.h>
-
 extern unsigned afs_debug;
 
 #define dbgprintk(FMT,...) \

fs/afs/rxrpc.c

@@ -330,7 +330,7 @@ static int afs_send_pages(struct afs_call *call, struct msghdr *msg)
 		first += nr;
 	} while (first <= last);
 
-	trace_afs_sent_pages(call, first, last, ret);
+	trace_afs_sent_pages(call, call->first, last, first, ret);
 	return ret;
 }
 

include/trace/events/afs.h

@@ -60,6 +60,26 @@ enum afs_vl_operation {
 	afs_VL_Probe		= 514,	/* AFS Probe Volume Location Service operation ID */
 };
 
+enum afs_writeback_trace {
+	afs_writeback_trace_alloc,
+	afs_writeback_trace_append,
+	afs_writeback_trace_conflict,
+	afs_writeback_trace_discard,
+	afs_writeback_trace_flush,
+	afs_writeback_trace_fsync,
+	afs_writeback_trace_invalidate_page,
+	afs_writeback_trace_new,
+	afs_writeback_trace_no_wake,
+	afs_writeback_trace_put,
+	afs_writeback_trace_release_page,
+	afs_writeback_trace_subsume,
+	afs_writeback_trace_unlink,
+	afs_writeback_trace_wake,
+	afs_writeback_trace_write,
+	afs_writeback_trace_writepages,
+	afs_writeback_trace_written,
+};
+
 #endif /* end __AFS_DECLARE_TRACE_ENUMS_ONCE_ONLY */
 
 /*
@@ -100,6 +120,25 @@ enum afs_vl_operation {
 	EM(afs_VL_GetEntryByName,		"VL.GetEntryByName") \
 	E_(afs_VL_Probe,			"VL.Probe")
 
+#define afs_writeback_traces \
+	EM(afs_writeback_trace_alloc,		"Alloc ") \
+	EM(afs_writeback_trace_append,		"Append") \
+	EM(afs_writeback_trace_conflict,	"Conflc") \
+	EM(afs_writeback_trace_discard,		"Discrd") \
+	EM(afs_writeback_trace_flush,		"Flush ") \
+	EM(afs_writeback_trace_fsync,		"Fsync ") \
+	EM(afs_writeback_trace_invalidate_page,	"InvlPg") \
+	EM(afs_writeback_trace_new,		"New   ") \
+	EM(afs_writeback_trace_no_wake,		"NoWake") \
+	EM(afs_writeback_trace_put,		"Put   ") \
+	EM(afs_writeback_trace_release_page,	"RelsPg") \
+	EM(afs_writeback_trace_subsume,		"Subsum") \
+	EM(afs_writeback_trace_unlink,		"Unlink") \
+	EM(afs_writeback_trace_wake,		"Wake  ") \
+	EM(afs_writeback_trace_write,		"Write ") \
+	EM(afs_writeback_trace_writepages,	"Wpages") \
+	E_(afs_writeback_trace_written,		"Writtn")
+
 
 /*
  * Export enum symbols via userspace.
@@ -112,6 +151,7 @@ enum afs_vl_operation {
 afs_call_traces;
 afs_fs_operations;
 afs_vl_operations;
+afs_writeback_traces;
 
 /*
  * Now redefine the EM() and E_() macros to map the enums to the strings that
@@ -344,14 +384,16 @@ TRACE_EVENT(afs_send_pages,
 	    );
 
 TRACE_EVENT(afs_sent_pages,
-	    TP_PROTO(struct afs_call *call, pgoff_t first, pgoff_t last, int ret),
+	    TP_PROTO(struct afs_call *call, pgoff_t first, pgoff_t last,
+		     pgoff_t cursor, int ret),
 
-	    TP_ARGS(call, first, last, ret),
+	    TP_ARGS(call, first, last, cursor, ret),
 
 	    TP_STRUCT__entry(
 		    __field(struct afs_call *,		call		)
 		    __field(pgoff_t,			first		)
 		    __field(pgoff_t,			last		)
+		    __field(pgoff_t,			cursor		)
 		    __field(int,			ret		)
 			     ),
 
@@ -359,13 +401,69 @@ TRACE_EVENT(afs_sent_pages,
 		    __entry->call = call;
 		    __entry->first = first;
 		    __entry->last = last;
+		    __entry->cursor = cursor;
 		    __entry->ret = ret;
 			   ),
 
-	    TP_printk(" c=%p %lx-%lx r=%d",
+	    TP_printk(" c=%p %lx-%lx c=%lx r=%d",
 		      __entry->call,
 		      __entry->first, __entry->last,
-		      __entry->ret)
+		      __entry->cursor, __entry->ret)
+	    );
+
+TRACE_EVENT(afs_write_begin,
+	    TP_PROTO(struct afs_vnode *vnode, pgoff_t index,
+		     unsigned from, unsigned to, unsigned flags),
+
+	    TP_ARGS(vnode, index, from, to, flags),
+
+	    TP_STRUCT__entry(
+		    __field(struct afs_vnode *,		vnode		)
+		    __field(pgoff_t,			index		)
+		    __field(unsigned,			from		)
+		    __field(unsigned,			to		)
+		    __field(unsigned,			flags		)
+			     ),
+
+	    TP_fast_assign(
+		    __entry->vnode = vnode;
+		    __entry->index = index;
+		    __entry->from = from;
+		    __entry->to = to;
+		    __entry->flags = flags;
+			   ),
+
+	    TP_printk("vn=%p pg=%lu %u-%u fl=%u",
+		      __entry->vnode, __entry->index,
+		      __entry->from, __entry->to, __entry->flags)
+	    );
+
+TRACE_EVENT(afs_writeback,
+	    TP_PROTO(struct afs_vnode *vnode, struct afs_writeback *wb,
+		     enum afs_writeback_trace op, int usage, int delta),
+
+	    TP_ARGS(vnode, wb, op, usage, delta),
+
+	    TP_STRUCT__entry(
+		    __field(struct afs_vnode *,		vnode		)
+		    __field(struct afs_writeback *,	wb		)
+		    __field(int,			op		)
+		    __field(int,			usage		)
+		    __field(int,			delta		)
+			     ),
+
+	    TP_fast_assign(
+		    __entry->vnode = vnode;
+		    __entry->wb = wb;
+		    __entry->op = op;
+		    __entry->usage = usage;
+		    __entry->delta = delta;
+			   ),
+
+	    TP_printk("vn=%p wb=%p %s u=%d d=%d",
+		      __entry->vnode, __entry->wb,
+		      __print_symbolic(__entry->op, afs_writeback_traces),
+		      __entry->usage, __entry->delta)
 	    );
 
 #endif /* _TRACE_AFS_H */