Print this page
NEX-7298 powertop dumps core when -c, -d or -t flags are used
Reviewed by: Gordon Ross <gordon.ross@nexenta.com>
NEX-6725 zpool offlining/onlining first disk in a mirror causes checksum error
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
Reviewed by: Roman Strashkin <roman.strashkin@nexenta.com>
Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
NEX-5795 Rename 'wrc' as 'wbc' in the source and in the tech docs
Reviewed by: Alex Aizman <alex.aizman@nexenta.com>
Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
NEX-4683 WRC: Special block pointer must know that it is special
Reviewed by: Alex Aizman <alex.aizman@nexenta.com>
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
Reviewed by: Saso Kiselkov <saso.kiselkov@nexenta.com>
NEX-4245 WRC: Code cleanup and refactoring to simplify merge with upstream
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
Reviewed by: Alex Aizman <alex.aizman@nexenta.com>
NEX-4091 WRC: Scrub sometimes reports checksum error
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
Reviewed by: Saso Kiselkov <saso.kiselkov@nexenta.com>
NEX-3558 KRRP Integration
NEX-3508 CLONE - Port NEX-2946 Add UNMAP/TRIM functionality to ZFS and illumos
Reviewed by: Josef Sipek <josef.sipek@nexenta.com>
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
Conflicts:
    usr/src/uts/common/io/scsi/targets/sd.c
    usr/src/uts/common/sys/scsi/targets/sddef.h
NEX-1007 added checks for NULL vdev in mirror_map
NEX-801 If a block pointer is corrupt read or write may crash
If block pointer is corrupt in such a way that vdev id of one of the
ditto blocks is wrong (out of range), zio_vdev_io_start or zio_vdev_io_done
may trip over it and crash.
This changeset takes care of this by claiming that an invalid vdev is
neither readable, nor writeable.
OS-80 support for vdev and CoS properties for the new I/O scheduler
OS-95 lint warning introduced by OS-61
Moved closed ZFS files to open repo, changed Makefiles accordingly
Removed unneeded weak symbols
re #12585 rb4049 ZFS++ work port - refactoring to improve separation of open/closed code, bug fixes, performance improvements - open code
Bug 11205: add missing libzfs_closed_stubs.c to fix opensource-only build.
ZFS plus work: special vdevs, cos, cos/vdev properties

@@ -22,10 +22,11 @@
  * Copyright 2010 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 /*
+ * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2012, 2015 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
 #include <sys/spa.h>

@@ -33,10 +34,11 @@
 #include <sys/dsl_pool.h>
 #include <sys/dsl_scan.h>
 #include <sys/vdev_impl.h>
 #include <sys/zio.h>
 #include <sys/abd.h>
+#include <sys/wbc.h>
 #include <sys/fs/zfs.h>
 
 /*
  * Virtual device vector for mirroring.
  */

@@ -46,10 +48,13 @@
         uint64_t        mc_offset;
         int             mc_error;
         uint8_t         mc_tried;
         uint8_t         mc_skipped;
         uint8_t         mc_speculative;
+        int             mc_index;       /* index in mirror_map_t */
+        avl_node_t      mc_node;        /* used for sorting based on weight */
+        int64_t         mc_weight;      /* thread-local copy of vdev_weight */
 } mirror_child_t;
 
 typedef struct mirror_map {
         int             mm_children;
         int             mm_resilvering;

@@ -82,36 +87,13 @@
         int c, d;
 
         if (vd == NULL) {
                 dva_t *dva = zio->io_bp->blk_dva;
                 spa_t *spa = zio->io_spa;
-                dva_t dva_copy[SPA_DVAS_PER_BP];
 
                 c = BP_GET_NDVAS(zio->io_bp);
 
-                /*
-                 * If we do not trust the pool config, some DVAs might be
-                 * invalid or point to vdevs that do not exist. We skip them.
-                 */
-                if (!spa_trust_config(spa)) {
-                        ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);
-                        int j = 0;
-                        for (int i = 0; i < c; i++) {
-                                if (zfs_dva_valid(spa, &dva[i], zio->io_bp))
-                                        dva_copy[j++] = dva[i];
-                        }
-                        if (j == 0) {
-                                zio->io_vsd = NULL;
-                                zio->io_error = ENXIO;
-                                return (NULL);
-                        }
-                        if (j < c) {
-                                dva = dva_copy;
-                                c = j;
-                        }
-                }
-
                 mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP);
                 mm->mm_children = c;
                 mm->mm_resilvering = B_FALSE;
                 mm->mm_preferred = spa_get_random(c);
                 mm->mm_root = B_TRUE;

@@ -131,10 +113,13 @@
                 for (c = 0; c < mm->mm_children; c++) {
                         mc = &mm->mm_child[c];
 
                         mc->mc_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[c]));
                         mc->mc_offset = DVA_GET_OFFSET(&dva[c]);
+                        mc->mc_index = c;
+                        mc->mc_weight = (mc->mc_vd != NULL ?
+                            mc->mc_vd->vdev_weight : 0);
                 }
         } else {
                 int replacing;
 
                 c = vd->vdev_children;

@@ -185,10 +170,13 @@
 
                 for (c = 0; c < mm->mm_children; c++) {
                         mc = &mm->mm_child[c];
                         mc->mc_vd = vd->vdev_child[c];
                         mc->mc_offset = zio->io_offset;
+                        mc->mc_index = c;
+                        mc->mc_weight = (mc->mc_vd != NULL ?
+                            mc->mc_vd->vdev_weight : 0);
                 }
         }
 
         zio->io_vsd = mm;
         zio->io_vsd_ops = &vdev_mirror_vsd_ops;

@@ -222,13 +210,10 @@
                 *max_asize = MIN(*max_asize - 1, cvd->vdev_max_asize - 1) + 1;
                 *ashift = MAX(*ashift, cvd->vdev_ashift);
         }
 
         if (numerrors == vd->vdev_children) {
-                if (vdev_children_are_offline(vd))
-                        vd->vdev_stat.vs_aux = VDEV_AUX_CHILDREN_OFFLINE;
-                else
                         vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
                 return (lasterror);
         }
 
         return (0);

@@ -274,54 +259,141 @@
         mc->mc_error = zio->io_error;
         mc->mc_tried = 1;
         mc->mc_skipped = 0;
 }
 
+static int
+vdev_weight_compar(const void *mc_a, const void *mc_b)
+{
+        const mirror_child_t *a = mc_a, *b = mc_b;
+
+        /*
+         * 1) if a's weight is less than b's, a goes right in the tree
+         * 2) if a's weight is greater than b's, a goes left
+         * 3) if a's and b's weights are equal, lower map index goes left
+         * 4) if weight and map index are equal, it's the same object
+         */
+        if (a->mc_weight < b->mc_weight)
+                return (1);
+        if (a->mc_weight > b->mc_weight)
+                return (-1);
+        if (a->mc_index > b->mc_index)
+                return (1);
+        if (a->mc_index < b->mc_index)
+                return (-1);
+        ASSERT3P(a->mc_vd, ==, b->mc_vd);
+        return (0);
+}
+
+static boolean_t
+child_select_mc(mirror_child_t *mc, uint64_t txg)
+{
+        if (mc->mc_tried || mc->mc_skipped)
+                return (B_FALSE);
+        if (!vdev_readable(mc->mc_vd)) {
+                mc->mc_error = SET_ERROR(ENXIO);
+                mc->mc_tried = 1;       /* don't even try */
+                mc->mc_skipped = 1;
+                return (B_FALSE);
+        }
+        if (!vdev_dtl_contains(mc->mc_vd, DTL_MISSING, txg, 1)) {
+                mc->mc_weight--;
+                mc->mc_vd->vdev_weight = mc->mc_weight;
+                return (B_TRUE);
+        }
+        mc->mc_error = SET_ERROR(ESTALE);
+        mc->mc_skipped = 1;
+        mc->mc_speculative = 1;
+        return (B_FALSE);
+}
+
+static void
+child_select_cleanup(mirror_map_t *mm, avl_tree_t *vdevs_by_weight)
+{
+        for (int i = 0; i < mm->mm_children; i++)
+                avl_remove(vdevs_by_weight, &mm->mm_child[i]);
+        avl_destroy(vdevs_by_weight);
+}
+
 /*
  * Try to find a child whose DTL doesn't contain the block we want to read.
  * If we can't, try the read on any vdev we haven't already tried.
  */
 static int
 vdev_mirror_child_select(zio_t *zio)
 {
         mirror_map_t *mm = zio->io_vsd;
-        mirror_child_t *mc;
         uint64_t txg = zio->io_txg;
-        int i, c;
+        /*
+         * Look at the weights of the vdevs in the mirror; the weights help
+         * decide which vdev to read from; the highest-weight suitable child
+         * index is returned, and its weight is decremented in order to avoid
+         * creating "hot" devices; once all the vdevs' weights are zero, the
+         * weights are set back to the ones configured in vdev props
+         */
+        int64_t max_weight = 0;
 
         ASSERT(zio->io_bp == NULL || BP_PHYSICAL_BIRTH(zio->io_bp) == txg);
 
+        for (int c = 0; c < mm->mm_children; c++) {
+                mirror_child_t *mc = &mm->mm_child[c];
+                if (mc->mc_vd == NULL)
+                        continue;
+                max_weight = MAX(max_weight, mc->mc_weight);
+        }
+
         /*
-         * Try to find a child whose DTL doesn't contain the block to read.
-         * If a child is known to be completely inaccessible (indicated by
-         * vdev_readable() returning B_FALSE), don't even try.
+         * Recalculate weights
          */
-        for (i = 0, c = mm->mm_preferred; i < mm->mm_children; i++, c++) {
-                if (c >= mm->mm_children)
-                        c = 0;
-                mc = &mm->mm_child[c];
-                if (mc->mc_tried || mc->mc_skipped)
+        if (max_weight == 0) {
+                for (int c = 0; c < mm->mm_children; c++) {
+                        mirror_child_t *mc = &mm->mm_child[c];
+                        if (mc->mc_vd == NULL)
                         continue;
-                if (!vdev_readable(mc->mc_vd)) {
-                        mc->mc_error = SET_ERROR(ENXIO);
-                        mc->mc_tried = 1;       /* don't even try */
-                        mc->mc_skipped = 1;
-                        continue;
+                        mc->mc_weight =
+                            vdev_queue_get_prop_uint64(&mc->mc_vd->vdev_queue,
+                            VDEV_PROP_PREFERRED_READ) + 1;
+                        mc->mc_vd->vdev_weight = mc->mc_weight;
                 }
-                if (!vdev_dtl_contains(mc->mc_vd, DTL_MISSING, txg, 1))
-                        return (c);
-                mc->mc_error = SET_ERROR(ESTALE);
-                mc->mc_skipped = 1;
-                mc->mc_speculative = 1;
         }
 
+        if (mm->mm_children > 1) {
+                avl_tree_t vdevs_by_weight;
+
+                avl_create(&vdevs_by_weight, vdev_weight_compar,
+                    sizeof (mirror_child_t), offsetof(mirror_child_t, mc_node));
+
         /*
+                 * Sort the weighted list
+                 */
+                for (int i = 0; i < mm->mm_children; i++)
+                        avl_add(&vdevs_by_weight, &mm->mm_child[i]);
+
+                /*
+                 * Try to find a child whose DTL doesn't contain the block to
+                 * read. If a child is known to be completely inaccessible
+                 * (vdev_readable() returning B_FALSE), don't even try.
+                 */
+                for (mirror_child_t *mc = avl_first(&vdevs_by_weight);
+                    mc != NULL; mc = AVL_NEXT(&vdevs_by_weight, mc)) {
+                        if (child_select_mc(mc, txg)) {
+                                child_select_cleanup(mm, &vdevs_by_weight);
+                                return (mc->mc_index);
+                        }
+                }
+                child_select_cleanup(mm, &vdevs_by_weight);
+        } else {
+                if (child_select_mc(&mm->mm_child[0], txg))
+                        return (0);
+        }
+
+        /*
          * Every device is either missing or has this txg in its DTL.
          * Look for any child we haven't already tried before giving up.
          */
-        for (c = 0; c < mm->mm_children; c++)
-                if (!mm->mm_child[c].mc_tried)
+        for (int c = 0; c < mm->mm_children; c++)
+                if (!mm->mm_child[c].mc_tried && mm->mm_child[c].mc_vd != NULL)
                         return (c);
 
         /*
          * Every child failed.  There's no place left to look.
          */

@@ -332,30 +404,52 @@
 vdev_mirror_io_start(zio_t *zio)
 {
         mirror_map_t *mm;
         mirror_child_t *mc;
         int c, children;
+        boolean_t spec_case = B_FALSE;
+        spa_t *spa = zio->io_spa;
 
         mm = vdev_mirror_map_alloc(zio);
 
-        if (mm == NULL) {
-                ASSERT(!spa_trust_config(zio->io_spa));
-                ASSERT(zio->io_type == ZIO_TYPE_READ);
-                zio_execute(zio);
-                return;
-        }
+        if (zio->io_child_type != ZIO_CHILD_VDEV &&
+            BP_IS_SPECIAL(zio->io_bp))
+                spec_case = B_TRUE;
 
         if (zio->io_type == ZIO_TYPE_READ) {
                 if ((zio->io_flags & ZIO_FLAG_SCRUB) && !mm->mm_resilvering) {
+                        int target = 0;
+
                         /*
+                         * Scrub of special BPs should take into
+                         * account the state of WBC-Window
+                         */
+                        if (spec_case) {
+                                target = wbc_select_dva(
+                                    spa_get_wbc_data(spa), zio);
+                        }
+
+                        /*
                          * For scrubbing reads we need to allocate a read
                          * buffer for each child and issue reads to all
                          * children.  If any child succeeds, it will copy its
                          * data into zio->io_data in vdev_mirror_scrub_done.
                          */
                         for (c = 0; c < mm->mm_children; c++) {
                                 mc = &mm->mm_child[c];
+                                if (mc->mc_vd == NULL) {
+                                        /*
+                                         * Invalid vdev id in blkptr caused
+                                         * mc_vd to be NULL here.
+                                         * Just skip this vdev.
+                                         */
+                                        continue;
+                                }
+
+                                if (spec_case && c != target)
+                                        continue;
+
                                 zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
                                     mc->mc_vd, mc->mc_offset,
                                     abd_alloc_sametype(zio->io_abd,
                                     zio->io_size), zio->io_size,
                                     zio->io_type, zio->io_priority, 0,

@@ -365,11 +459,16 @@
                         return;
                 }
                 /*
                  * For normal reads just pick one child.
                  */
+
+                if (spec_case)
+                        c = wbc_select_dva(spa_get_wbc_data(spa), zio);
+                else
                 c = vdev_mirror_child_select(zio);
+
                 children = (c >= 0);
         } else {
                 ASSERT(zio->io_type == ZIO_TYPE_WRITE);
 
                 /*

@@ -377,17 +476,33 @@
                  */
                 c = 0;
                 children = mm->mm_children;
         }
 
-        while (children--) {
+        for (; children--; c++) {
                 mc = &mm->mm_child[c];
+                if (mc->mc_vd == NULL) {
+                        /*
+                         * Invalid vdev in blkptr caused mc_vd to be NULL here.
+                         * Just skip this vdev.
+                         */
+                        continue;
+                }
+
+                if (spec_case) {
+                        if (zio->io_type == ZIO_TYPE_WRITE &&
+                            !vdev_is_special(mc->mc_vd))
+                                continue;
+                }
+
                 zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
                     mc->mc_vd, mc->mc_offset, zio->io_abd, zio->io_size,
                     zio->io_type, zio->io_priority, 0,
                     vdev_mirror_child_done, mc));
-                c++;
+
+                if (spec_case)
+                        break;
         }
 
         zio_execute(zio);
 }
 

@@ -412,13 +527,10 @@
         mirror_child_t *mc;
         int c;
         int good_copies = 0;
         int unexpected_errors = 0;
 
-        if (mm == NULL)
-                return;
-
         for (c = 0; c < mm->mm_children; c++) {
                 mc = &mm->mm_child[c];
 
                 if (mc->mc_error) {
                         if (!mc->mc_skipped)

@@ -520,23 +632,17 @@
 }
 
 static void
 vdev_mirror_state_change(vdev_t *vd, int faulted, int degraded)
 {
-        if (faulted == vd->vdev_children) {
-                if (vdev_children_are_offline(vd)) {
-                        vdev_set_state(vd, B_FALSE, VDEV_STATE_OFFLINE,
-                            VDEV_AUX_CHILDREN_OFFLINE);
-                } else {
+        if (faulted == vd->vdev_children)
                         vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
                             VDEV_AUX_NO_REPLICAS);
-                }
-        } else if (degraded + faulted != 0) {
+        else if (degraded + faulted != 0)
                 vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
-        } else {
+        else
                 vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
-        }
 }
 
 vdev_ops_t vdev_mirror_ops = {
         vdev_mirror_open,
         vdev_mirror_close,