3112 ztest does not honor ZFS_DEBUG

3113 ztest should use watchpoints to protect frozen arc bufs
3114 some leaked nvlists in zfsdev_ioctl
3115 poll(2) returns prematurely in presence of spurious wakeups
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Matt Amdur <Matt.Amdur@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Christopher Siden <chris.siden@delphix.com>
Approved by: Eric Schrock <eric.schrock@delphix.com>

usr/src/cmd/ztest/ztest.c

@@ -5835,6 +5835,8 @@ main(int argc, char **argv)
 
 	(void) setvbuf(stdout, NULL, _IOLBF, 0);
 
+	dprintf_setup(&argc, argv);
+
 	if (!ischild) {
 		process_options(argc, argv);
 

usr/src/lib/libzpool/common/sys/zfs_context.h

@@ -61,6 +61,7 @@ extern "C" {
 #include <atomic.h>
 #include <dirent.h>
 #include <time.h>
+#include <procfs.h>
 #include <libsysevent.h>
 #include <sys/note.h>
 #include <sys/types.h>

usr/src/uts/common/fs/zfs/arc.c

@@ -135,6 +135,12 @@
 #include <sys/kstat.h>
 #include <zfs_fletcher.h>
 
+#ifndef _KERNEL
+/* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */
+boolean_t arc_watch = B_FALSE;
+int arc_procfd;
+#endif
+
 static kmutex_t		arc_reclaim_thr_lock;
 static kcondvar_t	arc_reclaim_thr_cv;	/* used to signal reclaim thr */
 static uint8_t		arc_thread_exit;
@@ -474,6 +480,7 @@ static void arc_get_data_buf(arc_buf_t *buf);
 static void arc_access(arc_buf_hdr_t *buf, kmutex_t *hash_lock);
 static int arc_evict_needed(arc_buf_contents_t type);
 static void arc_evict_ghost(arc_state_t *state, uint64_t spa, int64_t bytes);
+static void arc_buf_watch(arc_buf_t *buf);
 
 static boolean_t l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *ab);
 
@@ -949,6 +956,50 @@ arc_cksum_compute(arc_buf_t *buf, boolean_t force)
 	fletcher_2_native(buf->b_data, buf->b_hdr->b_size,
 	    buf->b_hdr->b_freeze_cksum);
 	mutex_exit(&buf->b_hdr->b_freeze_lock);
+	arc_buf_watch(buf);
+}
+
+#ifndef _KERNEL
+typedef struct procctl {
+	long cmd;
+	prwatch_t prwatch;
+} procctl_t;
+#endif
+
+/* ARGSUSED */
+static void
+arc_buf_unwatch(arc_buf_t *buf)
+{
+#ifndef _KERNEL
+	if (arc_watch) {
+		int result;
+		procctl_t ctl;
+		ctl.cmd = PCWATCH;
+		ctl.prwatch.pr_vaddr = (uintptr_t)buf->b_data;
+		ctl.prwatch.pr_size = 0;
+		ctl.prwatch.pr_wflags = 0;
+		result = write(arc_procfd, &ctl, sizeof (ctl));
+		ASSERT3U(result, ==, sizeof (ctl));
+	}
+#endif
+}
+
+/* ARGSUSED */
+static void
+arc_buf_watch(arc_buf_t *buf)
+{
+#ifndef _KERNEL
+	if (arc_watch) {
+		int result;
+		procctl_t ctl;
+		ctl.cmd = PCWATCH;
+		ctl.prwatch.pr_vaddr = (uintptr_t)buf->b_data;
+		ctl.prwatch.pr_size = buf->b_hdr->b_size;
+		ctl.prwatch.pr_wflags = WA_WRITE;
+		result = write(arc_procfd, &ctl, sizeof (ctl));
+		ASSERT3U(result, ==, sizeof (ctl));
+	}
+#endif
 }
 
 void
@@ -975,6 +1026,8 @@ arc_buf_thaw(arc_buf_t *buf)
 	}
 
 	mutex_exit(&buf->b_hdr->b_freeze_lock);
+
+	arc_buf_unwatch(buf);
 }
 
 void
@@ -992,6 +1045,7 @@ arc_buf_freeze(arc_buf_t *buf)
 	    buf->b_hdr->b_state == arc_anon);
 	arc_cksum_compute(buf, B_FALSE);
 	mutex_exit(hash_lock);
+
 }
 
 static void
@@ -1348,21 +1402,22 @@ arc_buf_add_ref(arc_buf_t *buf, void* tag)
  * the buffer is placed on l2arc_free_on_write to be freed later.
  */
 static void
-arc_buf_data_free(arc_buf_hdr_t *hdr, void (*free_func)(void *, size_t),
-    void *data, size_t size)
+arc_buf_data_free(arc_buf_t *buf, void (*free_func)(void *, size_t))
 {
+	arc_buf_hdr_t *hdr = buf->b_hdr;
+
 	if (HDR_L2_WRITING(hdr)) {
 		l2arc_data_free_t *df;
 		df = kmem_alloc(sizeof (l2arc_data_free_t), KM_SLEEP);
-		df->l2df_data = data;
-		df->l2df_size = size;
+		df->l2df_data = buf->b_data;
+		df->l2df_size = hdr->b_size;
 		df->l2df_func = free_func;
 		mutex_enter(&l2arc_free_on_write_mtx);
 		list_insert_head(l2arc_free_on_write, df);
 		mutex_exit(&l2arc_free_on_write_mtx);
 		ARCSTAT_BUMP(arcstat_l2_free_on_write);
 	} else {
-		free_func(data, size);
+		free_func(buf->b_data, hdr->b_size);
 	}
 }
 
@@ -1378,16 +1433,15 @@ arc_buf_destroy(arc_buf_t *buf, boolean_t recycle, boolean_t all)
 		arc_buf_contents_t type = buf->b_hdr->b_type;
 
 		arc_cksum_verify(buf);
+		arc_buf_unwatch(buf);
 
 		if (!recycle) {
 			if (type == ARC_BUFC_METADATA) {
-				arc_buf_data_free(buf->b_hdr, zio_buf_free,
-				    buf->b_data, size);
+				arc_buf_data_free(buf, zio_buf_free);
 				arc_space_return(size, ARC_SPACE_DATA);
 			} else {
 				ASSERT(type == ARC_BUFC_DATA);
-				arc_buf_data_free(buf->b_hdr,
-				    zio_data_buf_free, buf->b_data, size);
+				arc_buf_data_free(buf, zio_data_buf_free);
 				ARCSTAT_INCR(arcstat_data_size, -size);
 				atomic_add_64(&arc_size, -size);
 			}
@@ -2556,6 +2610,7 @@ arc_read_done(zio_t *zio)
 	}
 
 	arc_cksum_compute(buf, B_FALSE);
+	arc_buf_watch(buf);
 
 	if (hash_lock && zio->io_error == 0 && hdr->b_state == arc_anon) {
 		/*
@@ -3113,6 +3168,7 @@ arc_release(arc_buf_t *buf, void *tag)
 		}
 		hdr->b_datacnt -= 1;
 		arc_cksum_verify(buf);
+		arc_buf_unwatch(buf);
 
 		mutex_exit(hash_lock);
 

usr/src/uts/common/fs/zfs/dsl_dataset.c

@@ -2302,7 +2302,6 @@ dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv)
 			}
 		}
 	}
-
 }
 
 void

usr/src/uts/common/fs/zfs/dsl_synctask.c

@@ -230,12 +230,7 @@ dsl_sync_task_do_nowait(dsl_pool_t *dp,
     dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
     void *arg1, void *arg2, int blocks_modified, dmu_tx_t *tx)
 {
-	dsl_sync_task_group_t *dstg;
-
-	if (!spa_writeable(dp->dp_spa))
-		return;
-
-	dstg = dsl_sync_task_group_create(dp);
+	dsl_sync_task_group_t *dstg = dsl_sync_task_group_create(dp);
 	dsl_sync_task_create(dstg, checkfunc, syncfunc,
 	    arg1, arg2, blocks_modified);
 	dsl_sync_task_group_nowait(dstg, tx);

usr/src/uts/common/fs/zfs/spa_history.c

@@ -303,7 +303,7 @@ spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
 	dmu_tx_t *tx;
 	nvlist_t *nvarg;
 
-	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY)
+	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
 		return (EINVAL);
 
 	tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
@@ -439,8 +439,9 @@ log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
 	/*
 	 * If this is part of creating a pool, not everything is
 	 * initialized yet, so don't bother logging the internal events.
+	 * Likewise if the pool is not writeable.
 	 */
-	if (tx->tx_txg == TXG_INITIAL) {
+	if (tx->tx_txg == TXG_INITIAL || !spa_writeable(spa)) {
 		fnvlist_free(nvl);
 		return;
 	}

usr/src/uts/common/fs/zfs/spa_misc.c

@@ -1600,6 +1600,18 @@ spa_init(int mode)
 
 	spa_mode_global = mode;
 
+#ifndef _KERNEL
+	if (spa_mode_global != FREAD && dprintf_find_string("watch")) {
+		arc_procfd = open("/proc/self/ctl", O_WRONLY);
+		if (arc_procfd == -1) {
+			perror("could not enable watchpoints: "
+			    "opening /proc/self/ctl failed: ");
+		} else {
+			arc_watch = B_TRUE;
+		}
+	}
+#endif
+
 	refcount_init();
 	unique_init();
 	zio_init();

usr/src/uts/common/fs/zfs/sys/arc.h

@@ -20,6 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012 by Delphix. All rights reserved.
  */
 
 #ifndef	_SYS_ARC_H
@@ -135,6 +136,11 @@ void l2arc_fini(void);
 void l2arc_start(void);
 void l2arc_stop(void);
 
+#ifndef _KERNEL
+extern boolean_t arc_watch;
+extern int arc_procfd;
+#endif
+
 #ifdef	__cplusplus
 }
 #endif

usr/src/uts/common/fs/zfs/sys/zfs_debug.h

@@ -20,6 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012 by Delphix. All rights reserved.
  */
 
 #ifndef _SYS_ZFS_DEBUG_H
@@ -75,6 +76,10 @@ extern void zfs_dbgmsg_init(void);
 extern void zfs_dbgmsg_fini(void);
 extern void zfs_dbgmsg(const char *fmt, ...);
 
+#ifndef _KERNEL
+extern int dprintf_find_string(const char *string);
+#endif
+
 #ifdef	__cplusplus
 }
 #endif

usr/src/uts/common/fs/zfs/zio.c

@@ -125,11 +125,21 @@ zio_init(void)
 		while (p2 & (p2 - 1))
 			p2 &= p2 - 1;
 
+#ifndef _KERNEL
+		/*
+		 * If we are using watchpoints, put each buffer on its own page,
+		 * to eliminate the performance overhead of trapping to the
+		 * kernel when modifying a non-watched buffer that shares the
+		 * page with a watched buffer.
+		 */
+		if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
+			continue;
+#endif
 		if (size <= 4 * SPA_MINBLOCKSIZE) {
 			align = SPA_MINBLOCKSIZE;
-		} else if (P2PHASE(size, PAGESIZE) == 0) {
+		} else if (IS_P2ALIGNED(size, PAGESIZE)) {
 			align = PAGESIZE;
-		} else if (P2PHASE(size, p2 >> 2) == 0) {
+		} else if (IS_P2ALIGNED(size, p2 >> 2)) {
 			align = p2 >> 2;
 		}