[XFS] Add I/O path tracing code, useful in diagnosing that last unwritten extent problem.

SGI Modid: xfs-linux:xfs-kern:166497a

fs/xfs/linux/xfs_aops.c

@@ -54,9 +54,54 @@
 #include "xfs_iomap.h"
 #include <linux/mpage.h>
 
+STATIC void count_page_state(struct page *, int *, int *, int *);
 STATIC void convert_page(struct inode *, struct page *,
 			xfs_iomap_t *, void *, int, int);
 
+#if defined(XFS_RW_TRACE)
+void
+xfs_page_trace(
+	int		tag,
+	struct inode	*inode,
+	struct page	*page,
+	int		mask)
+{
+	xfs_inode_t	*ip;
+	bhv_desc_t	*bdp;
+	vnode_t		*vp = LINVFS_GET_VP(inode);
+	loff_t		isize = i_size_read(inode);
+	loff_t		offset = page->index << PAGE_CACHE_SHIFT;
+	int		delalloc, unmapped, unwritten;
+
+	count_page_state(page, &delalloc, &unmapped, &unwritten);
+
+	bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
+	ip = XFS_BHVTOI(bdp);
+	if (!ip->i_rwtrace)
+		return;
+
+	ktrace_enter(ip->i_rwtrace,
+		(void *)((unsigned long)tag),
+		(void *)ip,
+		(void *)inode,
+		(void *)page,
+		(void *)((unsigned long)mask),
+		(void *)((unsigned long)((ip->i_d.di_size >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(ip->i_d.di_size & 0xffffffff)),
+		(void *)((unsigned long)((isize >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(isize & 0xffffffff)),
+		(void *)((unsigned long)((offset >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(offset & 0xffffffff)),
+		(void *)((unsigned long)delalloc),
+		(void *)((unsigned long)unmapped),
+		(void *)((unsigned long)unwritten),
+		(void *)NULL,
+		(void *)NULL);
+}
+#else
+#define xfs_page_trace(tag, inode, page, mask)
+#endif
+
 void
 linvfs_unwritten_done(
 	struct buffer_head	*bh,
@@ -1085,13 +1130,16 @@ linvfs_writepage(
 	int			delalloc, unmapped, unwritten;
 	struct inode		*inode = page->mapping->host;
 
+	xfs_page_trace(XFS_WRITEPAGE_ENTER, inode, page, 0);
+
 	/*
 	 * We need a transaction if:
 	 *  1. There are delalloc buffers on the page
-	 *  2. The page is upto date and we have unmapped buffers
-	 *  3. The page is upto date and we have no buffers
+	 *  2. The page is uptodate and we have unmapped buffers
+	 *  3. The page is uptodate and we have no buffers
 	 *  4. There are unwritten buffers on the page
 	 */
+
 	if (!page_has_buffers(page)) {
 		unmapped = 1;
 		need_trans = 1;
@@ -1166,6 +1214,8 @@ linvfs_release_page(
 	struct inode		*inode = page->mapping->host;
 	int			dirty, delalloc, unmapped, unwritten;
 
+	xfs_page_trace(XFS_RELEASEPAGE_ENTER, inode, page, gfp_mask);
+
 	count_page_state(page, &delalloc, &unmapped, &unwritten);
 	if (!delalloc && !unwritten)
 		goto free_buffers;

fs/xfs/linux/xfs_lrw.c

@@ -283,7 +283,6 @@ xfs_read(
 	ip = XFS_BHVTOI(bdp);
 	vp = BHV_TO_VNODE(bdp);
 	mp = ip->i_mount;
-	vn_trace_entry(vp, "xfs_read", (inst_t *)__return_address);
 
 	XFS_STATS_INC(xs_read_calls);
 
@@ -345,6 +344,8 @@ xfs_read(
 		}
 	}
 
+	xfs_rw_enter_trace(XFS_READ_ENTER, &ip->i_iocore,
+				iovp, segs, *offset, ioflags);
 	ret = __generic_file_aio_read(iocb, iovp, segs, offset);
 	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
 
@@ -377,7 +378,6 @@ xfs_sendfile(
 	ip = XFS_BHVTOI(bdp);
 	vp = BHV_TO_VNODE(bdp);
 	mp = ip->i_mount;
-	vn_trace_entry(vp, "xfs_sendfile", (inst_t *)__return_address);
 
 	XFS_STATS_INC(xs_read_calls);
 
@@ -405,6 +405,8 @@ xfs_sendfile(
 			return -error;
 		}
 	}
+	xfs_rw_enter_trace(XFS_SENDFILE_ENTER, &ip->i_iocore,
+				target, count, *offset, ioflags);
 	ret = generic_file_sendfile(filp, offset, count, actor, target);
 	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
 
@@ -658,7 +660,6 @@ xfs_write(
 	XFS_STATS_INC(xs_write_calls);
 
 	vp = BHV_TO_VNODE(bdp);
-	vn_trace_entry(vp, "xfs_write", (inst_t *)__return_address);
 	xip = XFS_BHVTOI(bdp);
 
 	/* START copy & waste from filemap.c */
@@ -678,7 +679,7 @@ xfs_write(
 	if (size == 0)
 		return 0;
 
-	io = &(xip->i_iocore);
+	io = &xip->i_iocore;
 	mp = io->io_mount;
 
 	xfs_check_frozen(mp, bdp, XFS_FREEZE_WRITE);

fs/xfs/linux/xfs_lrw.h

@@ -45,9 +45,7 @@ struct xfs_iomap;
 /*
  * Defines for the trace mechanisms in xfs_lrw.c.
  */
-#define	XFS_RW_KTRACE_SIZE	64
-#define	XFS_STRAT_KTRACE_SIZE	64
-#define	XFS_STRAT_GTRACE_SIZE	512
+#define	XFS_RW_KTRACE_SIZE	128
 
 #define	XFS_READ_ENTER		1
 #define	XFS_WRITE_ENTER		2
@@ -69,6 +67,12 @@ struct xfs_iomap;
 #define	XFS_INVAL_CACHED	18
 #define	XFS_DIORD_ENTER		19
 #define	XFS_DIOWR_ENTER		20
+#define	XFS_SENDFILE_ENTER	21
+#define	XFS_WRITEPAGE_ENTER	22
+#define	XFS_RELEASEPAGE_ENTER	23
+#define	XFS_IOMAP_ALLOC_ENTER	24
+#define	XFS_IOMAP_ALLOC_MAP	25
+#define	XFS_IOMAP_UNWRITTEN	26
 extern void xfs_rw_enter_trace(int, struct xfs_iocore *,
 			const struct iovec *, size_t, loff_t, int);
 extern void xfs_inval_cached_trace(struct xfs_iocore *,

fs/xfs/xfs_iomap.c

 /*
- * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -69,6 +69,76 @@
 #include "xfs_utils.h"
 #include "xfs_iomap.h"
 
+#if defined(XFS_RW_TRACE)
+void
+xfs_iomap_enter_trace(
+	int		tag,
+	xfs_iocore_t	*io,
+	xfs_off_t	offset,
+	ssize_t		count)
+{
+	xfs_inode_t	*ip = XFS_IO_INODE(io);
+
+	if (!ip->i_rwtrace)
+		return;
+
+	ktrace_enter(ip->i_rwtrace,
+		(void *)((unsigned long)tag),
+		(void *)ip,
+		(void *)((unsigned long)((ip->i_d.di_size >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(ip->i_d.di_size & 0xffffffff)),
+		(void *)((unsigned long)((offset >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(offset & 0xffffffff)),
+		(void *)((unsigned long)count),
+		(void *)((unsigned long)((io->io_new_size >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(io->io_new_size & 0xffffffff)),
+		(void *)NULL,
+		(void *)NULL,
+		(void *)NULL,
+		(void *)NULL,
+		(void *)NULL,
+		(void *)NULL,
+		(void *)NULL);
+}
+
+void
+xfs_iomap_map_trace(
+	int		tag,
+	xfs_iocore_t	*io,
+	xfs_off_t	offset,
+	ssize_t		count,
+	xfs_iomap_t	*iomapp,
+	xfs_bmbt_irec_t	*imapp,
+	int		flags)
+{
+	xfs_inode_t	*ip = XFS_IO_INODE(io);
+
+	if (!ip->i_rwtrace)
+		return;
+
+	ktrace_enter(ip->i_rwtrace,
+		(void *)((unsigned long)tag),
+		(void *)ip,
+		(void *)((unsigned long)((ip->i_d.di_size >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(ip->i_d.di_size & 0xffffffff)),
+		(void *)((unsigned long)((offset >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(offset & 0xffffffff)),
+		(void *)((unsigned long)count),
+		(void *)((unsigned long)flags),
+		(void *)((unsigned long)((iomapp->iomap_offset >> 32) & 0xffffffff)),
+		(void *)((unsigned long)(iomapp->iomap_offset & 0xffffffff)),
+		(void *)((unsigned long)(iomapp->iomap_delta)),
+		(void *)((unsigned long)(iomapp->iomap_bsize)),
+		(void *)((unsigned long)(iomapp->iomap_bn)),
+		(void *)(__psint_t)(imapp->br_startoff),
+		(void *)((unsigned long)(imapp->br_blockcount)),
+		(void *)(__psint_t)(imapp->br_startblock));
+}
+#else
+#define xfs_iomap_enter_trace(tag, io, offset, count)
+#define xfs_iomap_map_trace(tag, io, offset, count, iomapp, imapp, flags)
+#endif
+
 #define XFS_WRITEIO_ALIGN(mp,off)	(((off) >> mp->m_writeio_log) \
 						<< mp->m_writeio_log)
 #define XFS_STRAT_WRITE_IMAPS	2
@@ -149,17 +219,20 @@ xfs_iomap(
 		(BMAPI_READ | BMAPI_WRITE | BMAPI_ALLOCATE |
 		 BMAPI_UNWRITTEN | BMAPI_DEVICE)) {
 	case BMAPI_READ:
+		xfs_iomap_enter_trace(XFS_IOMAP_READ_ENTER, io, offset, count);
 		lockmode = XFS_LCK_MAP_SHARED(mp, io);
 		bmapi_flags = XFS_BMAPI_ENTIRE;
 		if (flags & BMAPI_IGNSTATE)
 			bmapi_flags |= XFS_BMAPI_IGSTATE;
 		break;
 	case BMAPI_WRITE:
+		xfs_iomap_enter_trace(XFS_IOMAP_WRITE_ENTER, io, offset, count);
 		lockmode = XFS_ILOCK_EXCL|XFS_EXTSIZE_WR;
 		bmapi_flags = 0;
 		XFS_ILOCK(mp, io, lockmode);
 		break;
 	case BMAPI_ALLOCATE:
+		xfs_iomap_enter_trace(XFS_IOMAP_ALLOC_ENTER, io, offset, count);
 		lockmode = XFS_ILOCK_SHARED|XFS_EXTSIZE_RD;
 		bmapi_flags = XFS_BMAPI_ENTIRE;
 		/* Attempt non-blocking lock */
@@ -201,8 +274,11 @@ xfs_iomap(
 	switch (flags & (BMAPI_WRITE|BMAPI_ALLOCATE|BMAPI_UNWRITTEN)) {
 	case BMAPI_WRITE:
 		/* If we found an extent, return it */
-		if (nimaps && (imap.br_startblock != HOLESTARTBLOCK))
+		if (nimaps && (imap.br_startblock != HOLESTARTBLOCK)) {
+			xfs_iomap_map_trace(XFS_IOMAP_WRITE_MAP, io,
+					offset, count, iomapp, &imap, flags);
 			break;
+		}
 
 		if (flags & (BMAPI_DIRECT|BMAPI_MMAP)) {
 			error = XFS_IOMAP_WRITE_DIRECT(mp, io, offset,
@@ -211,6 +287,10 @@ xfs_iomap(
 			error = XFS_IOMAP_WRITE_DELAY(mp, io, offset, count,
 					flags, &imap, &nimaps);
 		}
+		if (!error) {
+			xfs_iomap_map_trace(XFS_IOMAP_ALLOC_MAP, io,
+					offset, count, iomapp, &imap, flags);
+		}
 		iomap_flags = IOMAP_NEW;
 		break;
 	case BMAPI_ALLOCATE:
@@ -218,8 +298,11 @@ xfs_iomap(
 		XFS_IUNLOCK(mp, io, lockmode);
 		lockmode = 0;
 
-		if (nimaps && !ISNULLSTARTBLOCK(imap.br_startblock))
+		if (nimaps && !ISNULLSTARTBLOCK(imap.br_startblock)) {
+			xfs_iomap_map_trace(XFS_IOMAP_WRITE_MAP, io,
+					offset, count, iomapp, &imap, flags);
 			break;
+		}
 
 		error = XFS_IOMAP_WRITE_ALLOCATE(mp, io, &imap, &nimaps);
 		break;
@@ -309,7 +392,6 @@ xfs_iomap_write_direct(
 	 * Make sure that the dquots are there. This doesn't hold
 	 * the ilock across a disk read.
 	 */
-
 	error = XFS_QM_DQATTACH(ip->i_mount, ip, XFS_QMOPT_ILOCKED);
 	if (error)
 		return XFS_ERROR(error);
@@ -540,8 +622,9 @@ xfs_iomap_write_delay(
 	 * If bmapi returned us nothing, and if we didn't get back EDQUOT,
 	 * then we must have run out of space.
 	 */
-
 	if (nimaps == 0) {
+		xfs_iomap_enter_trace(XFS_IOMAP_WRITE_NOSPACE,
+					io, offset, count);
 		if (xfs_flush_space(ip, &fsynced, &ioflag))
 			return XFS_ERROR(ENOSPC);
 
@@ -584,7 +667,6 @@ xfs_iomap_write_allocate(
 	/*
 	 * Make sure that the dquots are there.
 	 */
-
 	if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
 		return XFS_ERROR(error);
 
@@ -612,7 +694,6 @@ xfs_iomap_write_allocate(
 					XFS_WRITE_LOG_RES(mp),
 					0, XFS_TRANS_PERM_LOG_RES,
 					XFS_WRITE_LOG_COUNT);
-
 			if (error == ENOSPC) {
 				error = xfs_trans_reserve(tp, 0,
 						XFS_WRITE_LOG_RES(mp),
@@ -653,19 +734,16 @@ xfs_iomap_write_allocate(
 			error = xfs_bmapi(tp, ip, map_start_fsb, count_fsb,
 					XFS_BMAPI_WRITE, &first_block, 1,
 					imap, &nimaps, &free_list);
-
 			if (error)
 				goto trans_cancel;
 
 			error = xfs_bmap_finish(&tp, &free_list,
 					first_block, &committed);
-
 			if (error)
 				goto trans_cancel;
 
 			error = xfs_trans_commit(tp,
 					XFS_TRANS_RELEASE_LOG_RES, NULL);
-
 			if (error)
 				goto error0;
 
@@ -725,6 +803,9 @@ xfs_iomap_write_unwritten(
 	xfs_fsblock_t	firstfsb;
 	xfs_bmap_free_t	free_list;
 
+	xfs_iomap_enter_trace(XFS_IOMAP_UNWRITTEN,
+				&ip->i_iocore, offset, count);
+
 	offset_fsb = XFS_B_TO_FSBT(mp, offset);
 	count_fsb = XFS_B_TO_FSB(mp, count);