TRIM/UNMAP/DISCARD support for vdevs (2)

cmd/ztest/ztest.c

@@ -6038,6 +6038,9 @@ main(int argc, char **argv)
 	/* Override location of zpool.cache */
 	VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache",
 	    ztest_opts.zo_dir) != -1);
+
+	/* Make sure TRIM zeroes data so that we can test it */
+	zfs_trim_zero = 1;
 
 	ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t),
 	    UMEM_NOFAIL);

config/kernel-max-discard-sectors.m4

@@ -0,0 +1,20 @@
+dnl #
+dnl # 2.6.32 API change
+dnl # max_discard_sectors is available.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_MAX_DISCARD_SECTORS], [
+	AC_MSG_CHECKING([whether ql->max_discard_sectors is available])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/blkdev.h>
+	],[
+		struct queue_limits ql __attribute__ ((unused));
+
+		ql.max_discard_sectors = 0;
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_MAX_DISCARD_SECTORS, 1,
+		          [ql->max_discard_sectors is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel.m4

@@ -39,6 +39,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_GET_GENDISK
 	ZFS_AC_KERNEL_RQ_IS_SYNC
 	ZFS_AC_KERNEL_RQ_FOR_EACH_SEGMENT
+	ZFS_AC_KERNEL_MAX_DISCARD_SECTORS
 	ZFS_AC_KERNEL_DISCARD_GRANULARITY
 	ZFS_AC_KERNEL_CONST_XATTR_HANDLER
 	ZFS_AC_KERNEL_XATTR_HANDLER_GET

include/linux/blkdev_compat.h

@@ -432,6 +432,37 @@ bio_set_flags_failfast(struct block_device *bdev, int *flags)
 # define VDEV_REQ_DISCARD		REQ_DISCARD
 #endif
 
+/*
+ * 2.6.32 API change
+ * blk_queue_discard is now available.
+ */
+#ifdef REQ_DISCARD
+#ifndef HAVE_BLK_QUEUE_DISCARD
+static inline unsigned long
+blk_queue_discard(struct request_queue *q)
+{
+	return !!q->prepare_discard_fn;
+}
+#endif
+#endif
+
+/*
+ * 2.6.32 API change
+ * On 2.6.32, maximum discard request size lies in:
+ *      request_queue.limits.max_discard_sectors
+ * Before 2.6.32, it lies in:
+ *      request_queue.max_hw_sectors
+ */
+static inline unsigned int
+blk_queue_max_discard_sectors_get(struct request_queue *q)
+{
+#ifdef HAVE_MAX_DISCARD_SECTORS
+	return q->limits.max_discard_sectors;
+#else
+	return q->max_hw_sectors;
+#endif
+}
+
 /*
  * 2.6.33 API change
  * Discard granularity and alignment restrictions may now be set.  For
@@ -443,8 +474,14 @@ blk_queue_discard_granularity(struct request_queue *q, unsigned int dg)
 {
 	q->limits.discard_granularity = dg;
 }
+static inline unsigned int
+blk_queue_discard_granularity_get(struct request_queue *q)
+{
+	return q->limits.discard_granularity;
+}
 #else
 #define blk_queue_discard_granularity(x, dg)	((void)0)
+#define blk_queue_discard_granularity_get(x) (0)
 #endif /* HAVE_DISCARD_GRANULARITY */
 
 /*

include/sys/Makefile.am

@@ -36,6 +36,7 @@ COMMON_H = \
 	$(top_srcdir)/include/sys/space_map.h \
 	$(top_srcdir)/include/sys/spa.h \
 	$(top_srcdir)/include/sys/spa_impl.h \
+	$(top_srcdir)/include/sys/trim_map.h \
 	$(top_srcdir)/include/sys/txg.h \
 	$(top_srcdir)/include/sys/txg_impl.h \
 	$(top_srcdir)/include/sys/u8_textprep_data.h \

include/sys/arc.h

@@ -143,7 +143,7 @@ void arc_fini(void);
  */
 
 void l2arc_add_vdev(spa_t *spa, vdev_t *vd);
-void l2arc_remove_vdev(vdev_t *vd);
+void l2arc_remove_vdev(vdev_t *vd, int permanent);
 boolean_t l2arc_vdev_present(vdev_t *vd);
 void l2arc_init(void);
 void l2arc_fini(void);

include/sys/spa_impl.h

@@ -217,6 +217,9 @@ struct spa {
 	spa_proc_state_t spa_proc_state;	/* see definition */
 	proc_t		*spa_proc;		/* "zpool-poolname" process */
 	uint64_t	spa_did;		/* if procp != p0, did of t1 */
+	kthread_t	*spa_trim_thread;	/* thread sending TRIM I/Os */
+	kmutex_t	spa_trim_lock;		/* protects spa_trim_cv */
+	kcondvar_t	spa_trim_cv;		/* used to notify TRIM thread */
 	boolean_t	spa_autoreplace;	/* autoreplace set in open */
 	int		spa_vdev_locks;		/* locks grabbed */
 	uint64_t	spa_creation_version;	/* version at pool creation */

include/sys/trim_map.h

@@ -0,0 +1,51 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright (c) 2012 Pawel Jakub Dawidek <[email protected]>.
+ * All rights reserved.
+ */
+
+#ifndef _SYS_TRIM_MAP_H
+#define	_SYS_TRIM_MAP_H
+
+#include <sys/avl.h>
+#include <sys/list.h>
+#include <sys/spa.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+extern void trim_map_create(vdev_t *vd);
+extern void trim_map_destroy(vdev_t *vd);
+extern void trim_map_free(vdev_t *vd, uint64_t offset, uint64_t size, uint64_t txg);
+extern boolean_t trim_map_write_start(zio_t *zio);
+extern void trim_map_write_done(zio_t *zio);
+
+extern void trim_thread_create(spa_t *spa);
+extern void trim_thread_destroy(spa_t *spa);
+extern void trim_thread_wakeup(spa_t *spa);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _SYS_TRIM_MAP_H */

include/sys/vdev.h

@@ -45,6 +45,7 @@ typedef enum vdev_dtl_type {
 } vdev_dtl_type_t;
 
 extern int zfs_nocacheflush;
+extern int zfs_trim_zero;
 
 extern int vdev_open(vdev_t *);
 extern void vdev_open_children(vdev_t *);

include/sys/vdev_impl.h

@@ -189,6 +189,7 @@ struct vdev {
 	uint64_t	vdev_unspare;	/* unspare when resilvering done */
 	hrtime_t	vdev_last_try;	/* last reopen time		*/
 	boolean_t	vdev_nowritecache; /* true if flushwritecache failed */
+	boolean_t	vdev_notrim;	/* true if trim failed 		*/
 	boolean_t	vdev_checkremove; /* temporary online test	*/
 	boolean_t	vdev_forcefault; /* force online fault		*/
 	boolean_t	vdev_splitting;	/* split or repair in progress  */
@@ -204,6 +205,7 @@ struct vdev {
 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache vdevs		*/
 	zio_t		*vdev_probe_zio; /* root of current probe	*/
 	vdev_aux_t	vdev_label_aux;	/* on-disk aux state		*/
+	struct trim_map *vdev_trimmap;
 
 	/*
 	 * For DTrace to work in userland (libzpool) context, these fields must
@@ -326,6 +328,8 @@ extern void vdev_set_min_asize(vdev_t *vd);
  */
 extern int zfs_vdev_cache_size;
 
+extern int zfs_notrim;
+
 #ifdef	__cplusplus
 }
 #endif

include/sys/zfs_context.h

@@ -493,13 +493,19 @@ typedef struct vsecattr {
 
 #define	CRCREAT		0
 
+#define	F_FREESP        11      /* Free file space */
+
 extern int fop_getattr(vnode_t *vp, vattr_t *vap);
+extern int fop_space(vnode_t *vp, int cmd, struct flock *bfp,
+    int flag, off_t offset);
 
 #define	VOP_CLOSE(vp, f, c, o, cr, ct)	vn_close(vp)
 #define	VOP_PUTPAGE(vp, of, sz, fl, cr, ct)	0
 #define	VOP_GETATTR(vp, vap, fl, cr, ct)  fop_getattr((vp), (vap));
 
 #define	VOP_FSYNC(vp, f, cr, ct)	fsync((vp)->v_fd)
+#define	VOP_SPACE(vp, cmd, bfp, flag, offset, cr, ct) \
+    fop_space((vp), (cmd), (bfp), (flag), (offset))
 
 #define	VN_RELE(vp)	vn_close(vp)
 

include/sys/zio.h

@@ -140,7 +140,8 @@ enum zio_compress {
 #define	ZIO_PRIORITY_RESILVER		(zio_priority_table[9])
 #define	ZIO_PRIORITY_SCRUB		(zio_priority_table[10])
 #define	ZIO_PRIORITY_DDT_PREFETCH	(zio_priority_table[11])
-#define	ZIO_PRIORITY_TABLE_SIZE		12
+#define	ZIO_PRIORITY_TRIM		(zio_priority_table[12])
+#define	ZIO_PRIORITY_TABLE_SIZE		13
 
 #define	ZIO_PIPELINE_CONTINUE		0x100
 #define	ZIO_PIPELINE_STOP		0x101
@@ -357,6 +358,39 @@ typedef struct zio_link {
 	list_node_t	zl_child_node;
 } zio_link_t;
 
+/*
+ * Used for TRIM kstat.
+ */
+typedef struct zio_trim_stats {
+	/*
+	 * Number of bytes successfully TRIMmed.
+	 */
+	kstat_named_t zio_trim_bytes;
+
+	/*
+	 * Number of successful TRIM requests.
+	 */
+	kstat_named_t zio_trim_success;
+
+	/*
+	 * Number of TRIM requests that failed because TRIM is not
+	 * supported.
+	 */
+	kstat_named_t zio_trim_unsupported;
+
+	/*
+	 * Number of TRIM requests that failed for other reasons.
+	 */
+	kstat_named_t zio_trim_failed;
+} zio_trim_stats_t;
+
+extern zio_trim_stats_t zio_trim_stats;
+
+#define ZIO_TRIM_STAT_INCR(stat, val) \
+	atomic_add_64(&zio_trim_stats.stat.value.ui64, (val));
+#define ZIO_TRIM_STAT_BUMP(stat) \
+	ZIO_TRIM_STAT_INCR(stat, 1);
+
 struct zio {
 	/* Core information about this I/O */
 	zbookmark_t	io_bookmark;
@@ -429,6 +463,9 @@ struct zio {
 
 	/* Taskq dispatching state */
 	taskq_ent_t	io_tqent;
+
+	avl_node_t	io_trim_node;
+	list_node_t	io_trim_link;
 };
 
 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
@@ -459,7 +496,8 @@ extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
     zio_done_func_t *done, void *private, enum zio_flag flags);
 
 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
-    zio_done_func_t *done, void *private, int priority, enum zio_flag flags);
+    uint64_t offset, uint64_t size, zio_done_func_t *done, 
+    void *private, int priority, enum zio_flag flags);
 
 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
     uint64_t size, void *data, int checksum,
@@ -472,12 +510,14 @@ extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
     boolean_t labels);
 
 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
-    const blkptr_t *bp, enum zio_flag flags);
+    const blkptr_t *bp, uint64_t size, enum zio_flag flags);
 
 extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp,
     blkptr_t *old_bp, uint64_t size, boolean_t use_slog);
 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
 extern void zio_flush(zio_t *zio, vdev_t *vd);
+extern zio_t *zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd,
+    uint64_t offset, uint64_t size, enum zio_flag flags);
 extern void zio_shrink(zio_t *zio, uint64_t size);
 
 extern int zio_wait(zio_t *zio);

include/sys/zio_impl.h

@@ -60,9 +60,9 @@ enum zio_stage {
 
 	ZIO_STAGE_READY			= 1 << 15,	/* RWFCI */
 
-	ZIO_STAGE_VDEV_IO_START		= 1 << 16,	/* RW--I */
-	ZIO_STAGE_VDEV_IO_DONE		= 1 << 17,	/* RW--I */
-	ZIO_STAGE_VDEV_IO_ASSESS	= 1 << 18,	/* RW--I */
+	ZIO_STAGE_VDEV_IO_START		= 1 << 16,	/* RWF-I */
+	ZIO_STAGE_VDEV_IO_DONE		= 1 << 17,	/* RWF-- */
+	ZIO_STAGE_VDEV_IO_ASSESS	= 1 << 18,	/* RWF-I */
 
 	ZIO_STAGE_CHECKSUM_VERIFY	= 1 << 19,	/* R---- */
 
@@ -143,7 +143,9 @@ enum zio_stage {
 #define	ZIO_FREE_PIPELINE			\
 	(ZIO_INTERLOCK_STAGES |			\
 	ZIO_STAGE_FREE_BP_INIT |		\
-	ZIO_STAGE_DVA_FREE)
+	ZIO_STAGE_DVA_FREE |			\
+	ZIO_STAGE_VDEV_IO_START |		\
+	ZIO_STAGE_VDEV_IO_ASSESS)
 
 #define	ZIO_DDT_FREE_PIPELINE			\
 	(ZIO_INTERLOCK_STAGES |			\

lib/libspl/include/sys/dkio.h

@@ -188,6 +188,8 @@ struct dk_geom {
  */
 #define	DKIOCFLUSHWRITECACHE	(DKIOC|34)	/* flush cache to phys medium */
 
+#define DKIOCTRIM		(DKIOC|35)	/* TRIM a block */
+
 struct dk_callback {
 	void (*dkc_callback)(void *dkc_cookie, int error);
 	void *dkc_cookie;

lib/libspl/include/sys/time.h

@@ -30,6 +30,10 @@
 #include_next <sys/time.h>
 #include <sys/types.h>
 
+#ifndef TIME_MAX
+#define TIME_MAX LLONG_MAX
+#endif
+
 #ifndef SEC
 #define SEC		1
 #endif
@@ -50,6 +54,10 @@
 #define NSEC_PER_USEC	1000L
 #endif
 
+#ifndef NSEC_PER_SEC
+#define NSEC_PER_SEC 1000000000L
+#endif
+
 extern hrtime_t gethrtime(void);
 extern void gethrestime(timestruc_t *);
 

lib/libzpool/Makefile.am

@@ -59,6 +59,7 @@ libzpool_la_SOURCES = \
 	$(top_srcdir)/module/zfs/spa_history.c \
 	$(top_srcdir)/module/zfs/spa_misc.c \
 	$(top_srcdir)/module/zfs/space_map.c \
+	$(top_srcdir)/module/zfs/trim_map.c \
 	$(top_srcdir)/module/zfs/txg.c \
 	$(top_srcdir)/module/zfs/uberblock.c \
 	$(top_srcdir)/module/zfs/unique.c \