rescrubbing only previously detected error block

Signed-off-by: TulsiJain <tulsi.jain@delphix.com>

Author: TulsiJain

.gitignore

@@ -62,4 +62,7 @@ cscope.*
 *.orig
 *.log
 venv
+# Ignore Eclipse files
+/.cproject
+/.project
 

cmd/zpool/zpool_main.c

@@ -100,6 +100,7 @@ static int zpool_do_split(int, char **);
 
 static int zpool_do_initialize(int, char **);
 static int zpool_do_scrub(int, char **);
+static int zpool_do_rescrub(int, char **);
 static int zpool_do_resilver(int, char **);
 static int zpool_do_trim(int, char **);
 
@@ -157,6 +158,7 @@ typedef enum {
 	HELP_REMOVE,
 	HELP_INITIALIZE,
 	HELP_SCRUB,
+	HELP_RESCRUB,
 	HELP_RESILVER,
 	HELP_TRIM,
 	HELP_STATUS,
@@ -296,6 +298,7 @@ static zpool_command_t command_table[] = {
 	{ "initialize",	zpool_do_initialize,	HELP_INITIALIZE		},
 	{ "resilver",	zpool_do_resilver,	HELP_RESILVER		},
 	{ "scrub",	zpool_do_scrub,		HELP_SCRUB		},
+	{ "rescrub",	zpool_do_rescrub,	HELP_RESCRUB		},
 	{ "trim",	zpool_do_trim,		HELP_TRIM		},
 	{ NULL },
 	{ "import",	zpool_do_import,	HELP_IMPORT		},
@@ -382,6 +385,8 @@ get_usage(zpool_help_t idx)
 		    "[<device> ...]\n"));
 	case HELP_SCRUB:
 		return (gettext("\tscrub [-s | -p] <pool> ...\n"));
+	case HELP_RESCRUB:
+		return (gettext("\trescrub [-s | -p] <pool> ...\n"));
 	case HELP_RESILVER:
 		return (gettext("\tresilver <pool> ...\n"));
 	case HELP_TRIM:
@@ -6703,6 +6708,26 @@ scrub_callback(zpool_handle_t *zhp, void *data)
 	return (err != 0);
 }
 
+int
+rescrub_callback(zpool_handle_t *zhp, void *data)
+{
+	scrub_cbdata_t *cb = data;
+	int err;
+
+	/*
+	 * Ignore faulted pools.
+	 */
+	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
+		(void) fprintf(stderr, gettext("cannot rescrub '%s': pool is "
+		    "currently unavailable\n"), zpool_get_name(zhp));
+		return (1);
+	}
+
+	err = zpool_rescrub(zhp, cb->cb_type, cb->cb_scrub_cmd);
+
+	return (err != 0);
+}
+
 /*
  * zpool scrub [-s | -p] <pool> ...
  *
@@ -6754,6 +6779,62 @@ zpool_do_scrub(int argc, char **argv)
 	return (for_each_pool(argc, argv, B_TRUE, NULL, scrub_callback, &cb));
 }
 
+/*
+ * zpool rescrub [-s | -p] <pool> ...
+ *
+ *	-s	Stop.  Stops any in-progress rescrub.
+ *	-p	Pause. Pause in-progress rescrub.
+ */
+int
+zpool_do_rescrub(int argc, char **argv)
+{
+	int c;
+	scrub_cbdata_t cb;
+
+	cb.cb_type = POOL_RESCRUB;
+	cb.cb_scrub_cmd = POOL_RESCRUB_NORMAL;
+
+	/* check options */
+	while ((c = getopt(argc, argv, "sp")) != -1) {
+		switch (c) {
+		case 's':
+			if (cb.cb_scrub_cmd != POOL_RESCRUB_NORMAL) {
+				(void) fprintf(stderr, gettext("invalid option"
+				    " combination: -s and -p are mutually"
+				    " exclusive\n"));
+				usage(B_FALSE);
+			}
+			cb.cb_scrub_cmd = POOL_RESCRUB_STOP;
+			break;
+		case 'p':
+			if (cb.cb_scrub_cmd != POOL_RESCRUB_NORMAL) {
+				(void) fprintf(stderr, gettext("invalid option"
+				    " combination: -s and -p are mutually"
+				    " exclusive\n"));
+				usage(B_FALSE);
+			}
+			cb.cb_scrub_cmd = POOL_RESCRUB_PAUSE;
+			break;
+		case '?':
+			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
+			    optopt);
+			usage(B_FALSE);
+		}
+	}
+
+	cb.cb_argc = argc;
+	cb.cb_argv = argv;
+	argc -= optind;
+	argv += optind;
+
+	if (argc < 1) {
+		(void) fprintf(stderr, gettext("missing pool name argument\n"));
+		usage(B_FALSE);
+	}
+
+	return (for_each_pool(argc, argv, B_TRUE, NULL, rescrub_callback, &cb));
+}
+
 /*
  * zpool resilver <pool> ...
  *

include/libzfs.h

@@ -125,7 +125,10 @@ typedef enum zfs_error {
 	EZFS_THREADCREATEFAILED, /* thread create failed */
 	EZFS_POSTSPLIT_ONLINE,	/* onlining a disk after splitting it */
 	EZFS_SCRUBBING,		/* currently scrubbing */
+	EZFS_RESCRUBBING,	/* currently rescrubbing */
+	EZFS_RESCRUB_PAUSED,	/* rescrub currently paused */
 	EZFS_NO_SCRUB,		/* no active scrub */
+	EZFS_NO_RESCRUB,	/* no active rescrub */
 	EZFS_DIFF,		/* general failure of zfs diff */
 	EZFS_DIFFDATA,		/* bad zfs diff data */
 	EZFS_POOLREADONLY,	/* pool is in read-only mode */
@@ -273,6 +276,7 @@ typedef struct trimflags {
  * Functions to manipulate pool and vdev state
  */
 extern int zpool_scan(zpool_handle_t *, pool_scan_func_t, pool_scrub_cmd_t);
+extern int zpool_rescrub(zpool_handle_t *, pool_scan_func_t, pool_scrub_cmd_t);
 extern int zpool_initialize(zpool_handle_t *, pool_initialize_func_t,
     nvlist_t *);
 extern int zpool_trim(zpool_handle_t *, pool_trim_func_t, nvlist_t *,

include/sys/dmu.h

@@ -373,6 +373,7 @@ typedef struct dmu_buf {
 #define	DMU_POOL_DDT_STATS		"DDT-statistics"
 #define	DMU_POOL_CREATION_VERSION	"creation_version"
 #define	DMU_POOL_SCAN			"scan"
+#define	DMU_POOL_RESCRUB		"rescrub"
 #define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
 #define	DMU_POOL_BPTREE_OBJ		"bptree_obj"
 #define	DMU_POOL_EMPTY_BPOBJ		"empty_bpobj"

include/sys/dsl_scan.h

@@ -29,6 +29,7 @@
 
 #include <sys/zfs_context.h>
 #include <sys/zio.h>
+#include <sys/zap.h>
 #include <sys/ddt.h>
 #include <sys/bplist.h>
 
@@ -67,7 +68,9 @@ typedef struct dsl_scan_phys {
 	uint64_t scn_flags; /* dsl_scan_flags_t */
 } dsl_scan_phys_t;
 
+
 #define	SCAN_PHYS_NUMINTS (sizeof (dsl_scan_phys_t) / sizeof (uint64_t))
+#define	RESCRUB_BLOCK_IN_TXG 10000
 
 typedef enum dsl_scan_flags {
 	DSF_VISIT_DS_AGAIN = 1<<0,
@@ -76,6 +79,19 @@ typedef enum dsl_scan_flags {
 
 #define	DSL_SCAN_FLAGS_MASK (DSF_VISIT_DS_AGAIN)
 
+typedef struct dsl_rescrub_phys {
+	uint64_t rescrub_func; /* pool_scan_func_t */
+	uint64_t rescrub_state; /* dsl_scan_state_t */
+	zap_cursor_t cursor; /* zap cursor */
+	uint64_t rescrub_start_time;
+	uint64_t rescrub_end_time;
+	uint64_t rescrub_to_examine; /* total blocks to be rescrubbed */
+	uint64_t rescrub_examined; /* blocks remaining so far */
+	uint64_t rescrub_errors;	/* rescrub I/O error count */
+	uint8_t rescrub_paused_flags; /* flag for paused */
+} dsl_rescrub_phys_t;
+
+
 /*
  * Every pool will have one dsl_scan_t and this structure will contain
  * in-memory information about the scan and a pointer to the on-disk
@@ -153,6 +169,9 @@ typedef struct dsl_scan {
 	dsl_scan_phys_t scn_phys_cached;
 	avl_tree_t scn_queue;		/* queue of datasets to scan */
 	uint64_t scn_bytes_pending;	/* outstanding data to issue */
+
+	/* members needed for syncing rescrub status to disk */
+	dsl_rescrub_phys_t rescrub_phys; /* on disk representation of rescrub */
 } dsl_scan_t;
 
 typedef struct dsl_scan_io_queue dsl_scan_io_queue_t;
@@ -166,6 +185,11 @@ int dsl_scan_cancel(struct dsl_pool *);
 int dsl_scan(struct dsl_pool *, pool_scan_func_t);
 boolean_t dsl_scan_scrubbing(const struct dsl_pool *dp);
 int dsl_scrub_set_pause_resume(const struct dsl_pool *dp, pool_scrub_cmd_t cmd);
+int dsl_rescrub_set_pause_resume(const struct dsl_pool *dp,
+    pool_scrub_cmd_t cmd);
+int dsl_rescrub(struct dsl_pool *, pool_scan_func_t);
+void dsl_rescrub_sync(struct dsl_pool *, dmu_tx_t *);
+int dsl_rescrub_cancel(struct dsl_pool *);
 void dsl_resilver_restart(struct dsl_pool *, uint64_t txg);
 boolean_t dsl_scan_resilvering(struct dsl_pool *dp);
 boolean_t dsl_dataset_unstable(struct dsl_dataset *ds);

include/sys/fs/zfs.h

@@ -899,6 +899,7 @@ typedef enum pool_scan_func {
 	POOL_SCAN_NONE,
 	POOL_SCAN_SCRUB,
 	POOL_SCAN_RESILVER,
+	POOL_RESCRUB,
 	POOL_SCAN_FUNCS
 } pool_scan_func_t;
 
@@ -908,6 +909,9 @@ typedef enum pool_scan_func {
 typedef enum pool_scrub_cmd {
 	POOL_SCRUB_NORMAL = 0,
 	POOL_SCRUB_PAUSE,
+	POOL_RESCRUB_NORMAL,
+	POOL_RESCRUB_PAUSE,
+	POOL_RESCRUB_STOP,
 	POOL_SCRUB_FLAGS_END
 } pool_scrub_cmd_t;
 
@@ -983,6 +987,7 @@ typedef enum dsl_scan_state {
 	DSS_SCANNING,
 	DSS_FINISHED,
 	DSS_CANCELED,
+	DSS_RESCRUBING,
 	DSS_NUM_STATES
 } dsl_scan_state_t;
 
@@ -1277,6 +1282,7 @@ typedef enum zfs_ioc {
 	ZFS_IOC_POOL_TRIM,			/* 0x5a50 */
 	ZFS_IOC_REDACT,				/* 0x5a51 */
 	ZFS_IOC_GET_BOOKMARK_PROPS,		/* 0x5a52 */
+	ZFS_IOC_POOL_RESCRUB,			/* 0x5a53 */
 
 	/*
 	 * Linux - 3/64 numbers reserved.

include/sys/spa.h

@@ -858,6 +858,11 @@ extern int spa_scan(spa_t *spa, pool_scan_func_t func);
 extern int spa_scan_stop(spa_t *spa);
 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
 
+/* rescrubing */
+extern int spa_rescrub(spa_t *spa, pool_scan_func_t func);
+extern int spa_rescrub_stop(spa_t *spa);
+extern int spa_rescrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
+
 /* spa syncing */
 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
 extern void spa_sync_allpools(void);
@@ -1167,11 +1172,13 @@ extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
 extern uint64_t spa_get_errlog_size(spa_t *spa);
+extern uint64_t spa_get_last_errlog_size(spa_t *spa);
 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
 extern void spa_errlog_rotate(spa_t *spa);
 extern void spa_errlog_drain(spa_t *spa);
 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
+// extern void name_to_bookmark(char *buf, zbookmark_phys_t *zb);
 
 /* vdev cache */
 extern void vdev_cache_stat_init(void);

include/sys/spa_impl.h

@@ -284,6 +284,8 @@ struct spa {
 	uint64_t	spa_scan_pass_exam;	/* examined bytes per pass */
 	uint64_t	spa_scan_pass_issued;	/* issued bytes per pass */
 
+	uint64_t	spa_scan_pass_rescrub_pause; /* rescrub pause time */
+	uint64_t	spa_scan_pass_rescrub_spent_paused; /* total paused */
 	/*
 	 * We are in the middle of a resilver, and another resilver
 	 * is needed once this one completes. This is set iff any

include/sys/sysevent/eventdefs.h

@@ -123,6 +123,11 @@ extern "C" {
 #define	ESC_ZFS_TRIM_CANCEL		"trim_cancel"
 #define	ESC_ZFS_TRIM_RESUME		"trim_resume"
 #define	ESC_ZFS_TRIM_SUSPEND		"trim_suspend"
+#define	ESC_ZFS_RESCRUB_START		"rescrub_start"
+#define	ESC_ZFS_RESCRUB_FINISH		"rescrub_finish"
+#define	ESC_ZFS_RESCRUB_ABORT		"rescrub_abort"
+#define	ESC_ZFS_RESCRUB_RESUME		"rescrub_resume"
+#define	ESC_ZFS_RESCRUB_PAUSED		"rescrub_paused"
 
 /*
  * datalink subclass definitions.

lib/libzfs/libzfs_pool.c

@@ -2364,6 +2364,7 @@ zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
 		pool_scan_stat_t *ps = NULL;
 		uint_t psc;
 
+		// needs some improvement because of rescrub
 		verify(nvlist_lookup_nvlist(zhp->zpool_config,
 		    ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
 		(void) nvlist_lookup_uint64_array(nvroot,
@@ -2385,6 +2386,83 @@ zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
 	}
 }
 
+
+/*
+ * rescrub the pool.
+ */
+int
+zpool_rescrub(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
+{
+	zfs_cmd_t zc = {"\0"};
+	char msg[1024];
+	int err;
+	libzfs_handle_t *hdl = zhp->zpool_hdl;
+
+	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
+	zc.zc_cookie = func;
+	zc.zc_flags = cmd;
+
+	if (zfs_ioctl(hdl, ZFS_IOC_POOL_RESCRUB, &zc) == 0)
+		return (0);
+
+	err = errno;
+
+	assert(func == POOL_RESCRUB);
+
+	/*
+	 * ECANCELED is returned when either we resumed a paused rescrub or
+	 * has no disk errorlog.
+	 */
+	if (err == ECANCELED && cmd == POOL_RESCRUB_NORMAL)
+		return (0);
+
+	if (err == ENOENT && cmd == POOL_RESCRUB_NORMAL)
+		return (0);
+
+	if (cmd == POOL_RESCRUB_PAUSE) {
+		(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
+		    "cannot pause rescrubbing %s"), zc.zc_name);
+	} else if (cmd == POOL_RESCRUB_NORMAL) {
+		(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
+		    "cannot rescrub %s"), zc.zc_name);
+	} else {
+		assert(cmd == POOL_RESCRUB_STOP);
+		(void) snprintf(msg, sizeof (msg),
+		    dgettext(TEXT_DOMAIN, "cannot cancel rescrubbing "
+		    "%s"), zc.zc_name);
+	}
+
+	if (err == EBUSY) {
+		nvlist_t *nvroot;
+		pool_scan_stat_t *ps = NULL;
+		uint_t psc;
+
+		// needs improvement working on it
+		verify(nvlist_lookup_nvlist(zhp->zpool_config,
+		    ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
+		(void) nvlist_lookup_uint64_array(nvroot,
+		    ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
+		if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
+			if (cmd == POOL_SCRUB_PAUSE)
+				return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
+			else
+				return (zfs_error(hdl, EZFS_SCRUBBING, msg));
+		} else if (ps && ps->pss_func == POOL_RESCRUB) {
+			if (cmd == POOL_RESCRUB_PAUSE)
+				return (zfs_error(hdl, EZFS_RESCRUB_PAUSED,
+				    msg));
+			else
+				return (zfs_error(hdl, EZFS_RESCRUBBING, msg));
+		} else {
+			return (zfs_error(hdl, EZFS_RESILVERING, msg));
+		}
+	} else if (err == ENOENT) {
+		return (zfs_error(hdl, EZFS_NO_RESCRUB, msg));
+	} else {
+		return (zpool_standard_error(hdl, err, msg));
+	}
+}
+
 /*
  * Find a vdev that matches the search criteria specified. We use the
  * the nvpair name to determine how we should look for the device.