SEEK_HOLE should not block on txg_wait_synced()

Force flushing of txg's can be painfully slow when competing for disk
IO, since this is a process meant to execute asynchronously. Optimize
this path via allowing data/hole seeking if the file is clean, but if
dirty fall back to old logic. This is a compromise to disabling the
feature entirely.

Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Debabrata Banerjee <dbanerje@akamai.com>
Closes openzfs#4306
Closes openzfs#5962

man/man5/zfs-module-parameters.5

@@ -1427,6 +1427,20 @@ Enable NOP writes
 Use \fB1\fR for yes (default) and \fB0\fR to disable.
 .RE
 
+.sp
+.ne 2
+.na
+\fBzfs_dmu_offset_next_sync\fR (int)
+.ad
+.RS 12n
+Enable forcing txg sync to find holes. When enabled forces ZFS to act
+like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which
+when a dnode is dirty causes txg's to be synced so that this data can be
+found.
+.sp
+Use \fB1\fR for yes and \fB0\fR to disable (default).
+.RE
+
 .sp
 .ne 2
 .na

module/zfs/dmu.c

@@ -67,6 +67,11 @@ int zfs_nopwrite_enabled = 1;
  */
 unsigned long zfs_per_txg_dirty_frees_percent = 30;
 
+/*
+ * Enable/disable forcing txg sync when dirty in dmu_offset_next.
+ */
+int zfs_dmu_offset_next_sync = 0;
+
 const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
 	{	DMU_BSWAP_UINT8,	TRUE,	"unallocated"		},
 	{	DMU_BSWAP_ZAP,		TRUE,	"object directory"	},
@@ -1989,32 +1994,56 @@ dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp,
 	zp->zp_nopwrite = nopwrite;
 }
 
+/*
+ * This function is only called from zfs_holey_common() for zpl_llseek()
+ * in order to determine the location of holes.  In order to accurately
+ * report holes all dirty data must be synced to disk.  This causes extremely
+ * poor performance when seeking for holes in a dirty file.  As a compromise,
+ * only provide hole data when the dnode is clean.  When a dnode is dirty
+ * report the dnode as having no holes which is always a safe thing to do.
+ */
 int
 dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off)
 {
 	dnode_t *dn;
 	int i, err;
+	boolean_t clean = B_TRUE;
 
 	err = dnode_hold(os, object, FTAG, &dn);
 	if (err)
 		return (err);
+
 	/*
-	 * Sync any current changes before
-	 * we go trundling through the block pointers.
+	 * Check if dnode is dirty
 	 */
-	for (i = 0; i < TXG_SIZE; i++) {
-		if (list_link_active(&dn->dn_dirty_link[i]))
-			break;
+	if (dn->dn_dirtyctx != DN_UNDIRTIED) {
+		for (i = 0; i < TXG_SIZE; i++) {
+			if (!list_is_empty(&dn->dn_dirty_records[i])) {
+				clean = B_FALSE;
+				break;
+			}
+		}
 	}
-	if (i != TXG_SIZE) {
+
+	/*
+	 * If compatibility option is on, sync any current changes before
+	 * we go trundling through the block pointers.
+	 */
+	if (!clean && zfs_dmu_offset_next_sync) {
+		clean = B_TRUE;
 		dnode_rele(dn, FTAG);
 		txg_wait_synced(dmu_objset_pool(os), 0);
 		err = dnode_hold(os, object, FTAG, &dn);
 		if (err)
 			return (err);
 	}
 
-	err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
+	if (clean)
+		err = dnode_next_offset(dn,
+		    (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
+	else
+		err = SET_ERROR(EBUSY);
+
 	dnode_rele(dn, FTAG);
 
 	return (err);
@@ -2238,5 +2267,11 @@ MODULE_PARM_DESC(zfs_nopwrite_enabled, "Enable NOP writes");
 module_param(zfs_per_txg_dirty_frees_percent, ulong, 0644);
 MODULE_PARM_DESC(zfs_per_txg_dirty_frees_percent,
 	"percentage of dirtied blocks from frees in one TXG");
+
+module_param(zfs_dmu_offset_next_sync, int, 0644);
+MODULE_PARM_DESC(zfs_dmu_offset_next_sync,
+	"Enable forcing txg sync to find holes");
+
 /* END CSTYLED */
+
 #endif

module/zfs/zfs_vnops.c

@@ -278,6 +278,10 @@ zfs_holey_common(struct inode *ip, int cmd, loff_t *off)
 	if (error == ESRCH)
 		return (SET_ERROR(ENXIO));
 
+	/* file was dirty, so fall back to using file_sz logic */
+	if (error == EBUSY)
+		error = 0;
+
 	/*
 	 * We could find a hole that begins after the logical end-of-file,
 	 * because dmu_offset_next() only works on whole blocks.  If the