From e2354ea344c248ede9faa872c260d46985830c0f Mon Sep 17 00:00:00 2001 From: Heinz Mauelshagen Date: Fri, 24 Feb 2017 00:50:00 +0100 Subject: [PATCH] lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978 --- lib/activate/activate.c | 60 ++++++++++++++- lib/activate/activate.h | 2 + lib/activate/dev_manager.c | 19 +++-- lib/format1/import-extents.c | 10 +-- lib/format_pool/import_export.c | 10 +-- lib/format_text/export.c | 6 +- lib/format_text/flags.c | 5 +- lib/format_text/import_vsn1.c | 16 +++- lib/metadata/lv.c | 5 +- lib/metadata/lv_alloc.h | 4 +- lib/metadata/lv_manip.c | 76 ++++++++++++------- lib/metadata/merge.c | 22 +++--- lib/metadata/metadata-exported.h | 15 +++- lib/metadata/raid_manip.c | 12 +-- lib/metadata/segtype.c | 3 +- lib/metadata/segtype.h | 7 ++ lib/metadata/snapshot_manip.c | 4 +- lib/raid/raid.c | 126 +++++++++++++++++++++---------- libdm/libdevmapper.h | 20 +++-- libdm/libdm-common.h | 2 + libdm/libdm-deptree.c | 117 ++++++++++++++++++++-------- libdm/libdm-targets.c | 18 +++++ tools/lvconvert.c | 34 ++++++--- 23 files changed, 429 insertions(+), 164 deletions(-) diff --git a/lib/activate/activate.c b/lib/activate/activate.c index ea567359b..a3978ada9 100644 --- a/lib/activate/activate.c +++ b/lib/activate/activate.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2016 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -272,10 +272,18 @@ int lv_raid_percent(const struct logical_volume *lv, dm_percent_t *percent) { return 0; } +int lv_raid_data_offset(const struct logical_volume *lv, uint64_t *data_offset) +{ + return 0; +} int lv_raid_dev_health(const struct logical_volume *lv, char **dev_health) { return 0; } +int lv_raid_dev_count(const struct logical_volume *lv, uint32_t *dev_cnt) +{ + return 0; +} int lv_raid_mismatch_count(const struct logical_volume *lv, uint64_t *cnt) { return 0; @@ -984,6 +992,30 @@ int lv_raid_percent(const struct logical_volume *lv, dm_percent_t *percent) return lv_mirror_percent(lv->vg->cmd, lv, 0, percent, NULL); } +int lv_raid_data_offset(const struct logical_volume *lv, uint64_t *data_offset) +{ + int r; + struct dev_manager *dm; + struct dm_status_raid *status; + + if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0)) + return 0; + + log_debug_activation("Checking raid data offset and dev sectors for LV %s/%s", + lv->vg->name, lv->name); + if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1))) + return_0; + + if (!(r = dev_manager_raid_status(dm, lv, &status))) + stack; + + *data_offset = status->data_offset; + + dev_manager_destroy(dm); + + return r; +} + int lv_raid_dev_health(const struct logical_volume *lv, char **dev_health) { int r; @@ -1013,6 +1045,32 @@ int lv_raid_dev_health(const struct logical_volume *lv, char **dev_health) return r; } +int lv_raid_dev_count(const struct logical_volume *lv, uint32_t *dev_cnt) +{ + struct dev_manager *dm; + struct dm_status_raid *status; + + *dev_cnt = 0; + + if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0)) + return 0; + + log_debug_activation("Checking raid device count for LV %s/%s", + lv->vg->name, lv->name); + if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1))) + return_0; + + if (!dev_manager_raid_status(dm, lv, &status)) { + dev_manager_destroy(dm); + return_0; + } + *dev_cnt = status->dev_count; + + dev_manager_destroy(dm); + + return 1; +} + int lv_raid_mismatch_count(const struct logical_volume *lv, uint64_t *cnt) { struct dev_manager *dm; diff --git a/lib/activate/activate.h b/lib/activate/activate.h index 85c152171..09d25c5b3 100644 --- a/lib/activate/activate.h +++ b/lib/activate/activate.h @@ -168,6 +168,8 @@ int lv_snapshot_percent(const struct logical_volume *lv, dm_percent_t *percent); int lv_mirror_percent(struct cmd_context *cmd, const struct logical_volume *lv, int wait, dm_percent_t *percent, uint32_t *event_nr); int lv_raid_percent(const struct logical_volume *lv, dm_percent_t *percent); +int lv_raid_dev_count(const struct logical_volume *lv, uint32_t *dev_cnt); +int lv_raid_data_offset(const struct logical_volume *lv, uint64_t *data_offset); int lv_raid_dev_health(const struct logical_volume *lv, char **dev_health); int lv_raid_mismatch_count(const struct logical_volume *lv, uint64_t *cnt); int lv_raid_sync_action(const struct logical_volume *lv, char **sync_action); diff --git a/lib/activate/dev_manager.c b/lib/activate/dev_manager.c index cde026e16..49d9ad31b 100644 --- a/lib/activate/dev_manager.c +++ b/lib/activate/dev_manager.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2016 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -214,6 +214,14 @@ typedef enum { STATUS, /* DM_DEVICE_STATUS ioctl */ } info_type_t; +/* Return length of segment depending on type and reshape_len */ +static uint32_t _seg_len(const struct lv_segment *seg) +{ + uint32_t reshape_len = seg_is_raid(seg) ? ((seg->area_count - seg->segtype->parity_devs) * seg->reshape_len) : 0; + + return seg->len - reshape_len; +} + static int _info_run(const char *dlid, struct dm_info *dminfo, uint32_t *read_ahead, struct lv_seg_status *seg_status, @@ -250,7 +258,7 @@ static int _info_run(const char *dlid, struct dm_info *dminfo, if (seg_status && dminfo->exists) { start = length = seg_status->seg->lv->vg->extent_size; start *= seg_status->seg->le; - length *= seg_status->seg->len; + length *= _seg_len(seg_status->seg); do { target = dm_get_next_target(dmt, target, &target_start, @@ -2214,7 +2222,7 @@ static char *_add_error_or_zero_device(struct dev_manager *dm, struct dm_tree *d struct lv_segment *seg_i; struct dm_info info; int segno = -1, i = 0; - uint64_t size = (uint64_t) seg->len * seg->lv->vg->extent_size; + uint64_t size = (uint64_t) _seg_len(seg) * seg->lv->vg->extent_size; dm_list_iterate_items(seg_i, &seg->lv->segments) { if (seg == seg_i) { @@ -2500,7 +2508,7 @@ static int _add_target_to_dtree(struct dev_manager *dm, return seg->segtype->ops->add_target_line(dm, dm->mem, dm->cmd, &dm->target_state, seg, laopts, dnode, - extent_size * seg->len, + extent_size * _seg_len(seg), &dm->pvmove_mirror_count); } @@ -2693,7 +2701,7 @@ static int _add_segment_to_dtree(struct dev_manager *dm, /* Replace target and all its used devs with error mapping */ log_debug_activation("Using error for pending delete %s.", display_lvname(seg->lv)); - if (!dm_tree_node_add_error_target(dnode, (uint64_t)seg->lv->vg->extent_size * seg->len)) + if (!dm_tree_node_add_error_target(dnode, (uint64_t)seg->lv->vg->extent_size * _seg_len(seg))) return_0; } else if (!_add_target_to_dtree(dm, dnode, seg, laopts)) return_0; @@ -3165,7 +3173,6 @@ static int _tree_action(struct dev_manager *dm, const struct logical_volume *lv, log_error(INTERNAL_ERROR "_tree_action: Action %u not supported.", action); goto out; } - r = 1; out: diff --git a/lib/format1/import-extents.c b/lib/format1/import-extents.c index 3ab3ac443..4c259c126 100644 --- a/lib/format1/import-extents.c +++ b/lib/format1/import-extents.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -225,8 +225,8 @@ static int _read_linear(struct cmd_context *cmd, struct lv_map *lvm) while (le < lvm->lv->le_count) { len = _area_length(lvm, le); - if (!(seg = alloc_lv_segment(segtype, lvm->lv, le, len, 0, 0, - NULL, 1, len, 0, 0, 0, NULL))) { + if (!(seg = alloc_lv_segment(segtype, lvm->lv, le, len, 0, 0, 0, + NULL, 1, len, 0, 0, 0, 0, NULL))) { log_error("Failed to allocate linear segment."); return 0; } @@ -297,10 +297,10 @@ static int _read_stripes(struct cmd_context *cmd, struct lv_map *lvm) if (!(seg = alloc_lv_segment(segtype, lvm->lv, lvm->stripes * first_area_le, - lvm->stripes * area_len, + lvm->stripes * area_len, 0, 0, lvm->stripe_size, NULL, lvm->stripes, - area_len, 0, 0, 0, NULL))) { + area_len, 0, 0, 0, 0, NULL))) { log_error("Failed to allocate striped segment."); return 0; } diff --git a/lib/format_pool/import_export.c b/lib/format_pool/import_export.c index 2f0f2ebda..f4097a7ae 100644 --- a/lib/format_pool/import_export.c +++ b/lib/format_pool/import_export.c @@ -1,6 +1,6 @@ /* * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -192,9 +192,9 @@ static int _add_stripe_seg(struct dm_pool *mem, return_0; if (!(seg = alloc_lv_segment(segtype, lv, *le_cur, - area_len * usp->num_devs, 0, + area_len * usp->num_devs, 0, 0, usp->striping, NULL, usp->num_devs, - area_len, 0, 0, 0, NULL))) { + area_len, 0, 0, 0, 0, NULL))) { log_error("Unable to allocate striped lv_segment structure"); return 0; } @@ -232,8 +232,8 @@ static int _add_linear_seg(struct dm_pool *mem, area_len = (usp->devs[j].blocks) / POOL_PE_SIZE; if (!(seg = alloc_lv_segment(segtype, lv, *le_cur, - area_len, 0, usp->striping, - NULL, 1, area_len, + area_len, 0, 0, usp->striping, + NULL, 1, area_len, 0, POOL_PE_SIZE, 0, 0, NULL))) { log_error("Unable to allocate linear lv_segment " "structure"); diff --git a/lib/format_text/export.c b/lib/format_text/export.c index 199c185ab..899ff45cb 100644 --- a/lib/format_text/export.c +++ b/lib/format_text/export.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2009 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -583,8 +583,10 @@ static int _print_segment(struct formatter *f, struct volume_group *vg, outf(f, "start_extent = %u", seg->le); outsize(f, (uint64_t) seg->len * vg->extent_size, "extent_count = %u", seg->len); - outnl(f); + if (seg->reshape_len) + outsize(f, (uint64_t) seg->reshape_len * vg->extent_size, + "reshape_count = %u", seg->reshape_len); outf(f, "type = \"%s\"", seg->segtype->name); if (!_out_list(f, &seg->tags, "tags")) diff --git a/lib/format_text/flags.c b/lib/format_text/flags.c index 75b905473..716e63201 100644 --- a/lib/format_text/flags.c +++ b/lib/format_text/flags.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2013 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -61,6 +61,9 @@ static const struct flag _lv_flags[] = { {LOCKED, "LOCKED", STATUS_FLAG}, {LV_NOTSYNCED, "NOTSYNCED", STATUS_FLAG}, {LV_REBUILD, "REBUILD", STATUS_FLAG}, + {LV_RESHAPE_DELTA_DISKS_PLUS, "RESHAPE_DELTA_DISKS_PLUS", STATUS_FLAG}, + {LV_RESHAPE_DELTA_DISKS_MINUS, "RESHAPE_DELTA_DISKS_MINUS", STATUS_FLAG}, + {LV_REMOVE_AFTER_RESHAPE, "REMOVE_AFTER_RESHAPE", STATUS_FLAG}, {LV_WRITEMOSTLY, "WRITEMOSTLY", STATUS_FLAG}, {LV_ACTIVATION_SKIP, "ACTIVATION_SKIP", COMPATIBLE_FLAG}, {LV_ERROR_WHEN_FULL, "ERROR_WHEN_FULL", COMPATIBLE_FLAG}, diff --git a/lib/format_text/import_vsn1.c b/lib/format_text/import_vsn1.c index 00caf210f..e54500812 100644 --- a/lib/format_text/import_vsn1.c +++ b/lib/format_text/import_vsn1.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -354,7 +354,7 @@ static int _read_segment(struct logical_volume *lv, const struct dm_config_node struct lv_segment *seg; const struct dm_config_node *sn_child = sn->child; const struct dm_config_value *cv; - uint32_t start_extent, extent_count; + uint32_t area_extents, start_extent, extent_count, reshape_count, data_copies; struct segment_type *segtype; const char *segtype_str; @@ -375,6 +375,12 @@ static int _read_segment(struct logical_volume *lv, const struct dm_config_node return 0; } + if (!_read_int32(sn_child, "reshape_count", &reshape_count)) + reshape_count = 0; + + if (!_read_int32(sn_child, "data_copies", &data_copies)) + data_copies = 1; + segtype_str = SEG_TYPE_NAME_STRIPED; if (!dm_config_get_str(sn_child, "type", &segtype_str)) { @@ -389,9 +395,11 @@ static int _read_segment(struct logical_volume *lv, const struct dm_config_node !segtype->ops->text_import_area_count(sn_child, &area_count)) return_0; + area_extents = segtype->parity_devs ? + raid_rimage_extents(segtype, extent_count, area_count - segtype->parity_devs, data_copies) : extent_count; if (!(seg = alloc_lv_segment(segtype, lv, start_extent, - extent_count, 0, 0, NULL, area_count, - extent_count, 0, 0, 0, NULL))) { + extent_count, reshape_count, 0, 0, NULL, area_count, + area_extents, data_copies, 0, 0, 0, NULL))) { log_error("Segment allocation failed"); return 0; } diff --git a/lib/metadata/lv.c b/lib/metadata/lv.c index 85879025a..b54e39180 100644 --- a/lib/metadata/lv.c +++ b/lib/metadata/lv.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2016 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -1278,6 +1278,9 @@ char *lv_attr_dup_with_info_and_seg_status(struct dm_pool *mem, const struct lv_ repstr[8] = 'm'; /* RAID has 'm'ismatches */ } else if (lv->status & LV_WRITEMOSTLY) repstr[8] = 'w'; /* sub-LV has 'w'ritemostly */ + else if (lv->status & LV_REMOVE_AFTER_RESHAPE) + repstr[8] = 'R'; /* sub-LV got freed from raid set by reshaping + and has to be 'R'emoved */ } else if (lvdm->seg_status.type == SEG_STATUS_CACHE) { if (lvdm->seg_status.cache->fail) repstr[8] = 'F'; diff --git a/lib/metadata/lv_alloc.h b/lib/metadata/lv_alloc.h index f7bc71360..cf2c579c6 100644 --- a/lib/metadata/lv_alloc.h +++ b/lib/metadata/lv_alloc.h @@ -1,6 +1,6 @@ /* * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2012 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -21,11 +21,13 @@ struct lv_segment *alloc_lv_segment(const struct segment_type *segtype, struct logical_volume *lv, uint32_t le, uint32_t len, + uint32_t reshape_len, uint64_t status, uint32_t stripe_size, struct logical_volume *log_lv, uint32_t area_count, uint32_t area_len, + uint32_t data_copies, uint32_t chunk_size, uint32_t region_size, uint32_t extents_copied, diff --git a/lib/metadata/lv_manip.c b/lib/metadata/lv_manip.c index 78cc30e96..f3cad0977 100644 --- a/lib/metadata/lv_manip.c +++ b/lib/metadata/lv_manip.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2014 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -912,11 +912,13 @@ static uint32_t _round_to_stripe_boundary(struct volume_group *vg, uint32_t exte struct lv_segment *alloc_lv_segment(const struct segment_type *segtype, struct logical_volume *lv, uint32_t le, uint32_t len, + uint32_t reshape_len, uint64_t status, uint32_t stripe_size, struct logical_volume *log_lv, uint32_t area_count, uint32_t area_len, + uint32_t data_copies, uint32_t chunk_size, uint32_t region_size, uint32_t extents_copied, @@ -950,10 +952,12 @@ struct lv_segment *alloc_lv_segment(const struct segment_type *segtype, seg->lv = lv; seg->le = le; seg->len = len; + seg->reshape_len = reshape_len; seg->status = status; seg->stripe_size = stripe_size; seg->area_count = area_count; seg->area_len = area_len; + seg->data_copies = data_copies ? : 0; // lv_raid_data_copies(segtype, area_count); seg->chunk_size = chunk_size; seg->region_size = region_size; seg->extents_copied = extents_copied; @@ -1047,11 +1051,10 @@ static int _release_and_discard_lv_segment_area(struct lv_segment *seg, uint32_t if (lv_is_raid_image(lv)) { /* Calculate the amount of extents to reduce per rmate/rimage LV */ uint32_t rimage_extents; + struct lv_segment *seg1 = first_seg(lv); - /* FIXME: avoid extra seg_is_*() conditonals */ - area_reduction =_round_to_stripe_boundary(lv->vg, area_reduction, - (seg_is_raid1(seg) || seg_is_any_raid0(seg)) ? 0 : _raid_stripes_count(seg), 0); - rimage_extents = raid_rimage_extents(seg->segtype, area_reduction, seg_is_any_raid0(seg) ? 0 : _raid_stripes_count(seg), + /* FIXME: avoid extra seg_is_*() conditionals here */ + rimage_extents = raid_rimage_extents(seg1->segtype, area_reduction, seg_is_any_raid0(seg) ? 0 : _raid_stripes_count(seg), seg_is_raid10(seg) ? 1 :_raid_data_copies(seg)); if (!rimage_extents) return 0; @@ -1258,7 +1261,7 @@ static uint32_t _calc_area_multiple(const struct segment_type *segtype, * the 'stripes' argument will always need to * be given. */ - if (!strcmp(segtype->name, _lv_type_names[LV_TYPE_RAID10])) { + if (segtype_is_raid10(segtype)) { if (!stripes) return area_count / 2; return stripes; @@ -1278,16 +1281,17 @@ static uint32_t _calc_area_multiple(const struct segment_type *segtype, static int _lv_segment_reduce(struct lv_segment *seg, uint32_t reduction) { uint32_t area_reduction, s; + uint32_t areas = (seg->area_count / (seg_is_raid10(seg) ? seg->data_copies : 1)) - seg->segtype->parity_devs; /* Caller must ensure exact divisibility */ - if (seg_is_striped(seg)) { - if (reduction % seg->area_count) { + if (seg_is_striped(seg) || seg_is_striped_raid(seg)) { + if (reduction % areas) { log_error("Segment extent reduction %" PRIu32 " not divisible by #stripes %" PRIu32, reduction, seg->area_count); return 0; } - area_reduction = (reduction / seg->area_count); + area_reduction = reduction / areas; } else area_reduction = reduction; @@ -1296,7 +1300,11 @@ static int _lv_segment_reduce(struct lv_segment *seg, uint32_t reduction) return_0; seg->len -= reduction; - seg->area_len -= area_reduction; + + if (seg_is_raid(seg)) + seg->area_len = seg->len; + else + seg->area_len -= area_reduction; return 1; } @@ -1306,11 +1314,13 @@ static int _lv_segment_reduce(struct lv_segment *seg, uint32_t reduction) */ static int _lv_reduce(struct logical_volume *lv, uint32_t extents, int delete) { - struct lv_segment *seg; + struct lv_segment *seg = first_seg(lv);; uint32_t count = extents; uint32_t reduction; struct logical_volume *pool_lv; struct logical_volume *external_lv = NULL; + int is_raid10 = seg_is_any_raid10(seg) && seg->reshape_len; + uint32_t data_copies = seg->data_copies; if (lv_is_merging_origin(lv)) { log_debug_metadata("Dropping snapshot merge of %s to removed origin %s.", @@ -1373,7 +1383,15 @@ static int _lv_reduce(struct logical_volume *lv, uint32_t extents, int delete) count -= reduction; } - lv->le_count -= extents; + seg = first_seg(lv); + + if (is_raid10) { + lv->le_count -= extents * data_copies; + if (seg) + seg->len = seg->area_len = lv->le_count; + } else + lv->le_count -= extents; + lv->size = (uint64_t) lv->le_count * lv->vg->extent_size; if (!delete) @@ -1793,10 +1811,10 @@ static int _setup_alloced_segment(struct logical_volume *lv, uint64_t status, area_multiple = _calc_area_multiple(segtype, area_count, 0); extents = aa[0].len * area_multiple; - if (!(seg = alloc_lv_segment(segtype, lv, lv->le_count, extents, + if (!(seg = alloc_lv_segment(segtype, lv, lv->le_count, extents, 0, status, stripe_size, NULL, area_count, - aa[0].len, 0u, region_size, 0u, NULL))) { + aa[0].len, 0, 0u, region_size, 0u, NULL))) { log_error("Couldn't allocate new LV segment."); return 0; } @@ -3234,9 +3252,9 @@ int lv_add_virtual_segment(struct logical_volume *lv, uint64_t status, seg->area_len += extents; seg->len += extents; } else { - if (!(seg = alloc_lv_segment(segtype, lv, lv->le_count, extents, + if (!(seg = alloc_lv_segment(segtype, lv, lv->le_count, extents, 0, status, 0, NULL, 0, - extents, 0, 0, 0, NULL))) { + extents, 0, 0, 0, 0, NULL))) { log_error("Couldn't allocate new %s segment.", segtype->name); return 0; } @@ -3562,10 +3580,10 @@ static struct lv_segment *_convert_seg_to_mirror(struct lv_segment *seg, } if (!(newseg = alloc_lv_segment(get_segtype_from_string(seg->lv->vg->cmd, SEG_TYPE_NAME_MIRROR), - seg->lv, seg->le, seg->len, + seg->lv, seg->le, seg->len, 0, seg->status, seg->stripe_size, log_lv, - seg->area_count, seg->area_len, + seg->area_count, seg->area_len, 0, seg->chunk_size, region_size, seg->extents_copied, NULL))) { log_error("Couldn't allocate converted LV segment."); @@ -3667,8 +3685,8 @@ int lv_add_segmented_mirror_image(struct alloc_handle *ah, } if (!(new_seg = alloc_lv_segment(segtype, copy_lv, - seg->le, seg->len, PVMOVE, 0, - NULL, 1, seg->len, + seg->le, seg->len, 0, PVMOVE, 0, + NULL, 1, seg->len, 0, 0, 0, 0, NULL))) return_0; @@ -3863,9 +3881,9 @@ static int _lv_insert_empty_sublvs(struct logical_volume *lv, /* * First, create our top-level segment for our top-level LV */ - if (!(mapseg = alloc_lv_segment(segtype, lv, 0, 0, lv->status, + if (!(mapseg = alloc_lv_segment(segtype, lv, 0, 0, 0, lv->status, stripe_size, NULL, - devices, 0, 0, region_size, 0, NULL))) { + devices, 0, 0, 0, region_size, 0, NULL))) { log_error("Failed to create mapping segment for %s.", display_lvname(lv)); return 0; @@ -4063,8 +4081,11 @@ static int _lv_extend_layered_lv(struct alloc_handle *ah, lv_set_hidden(seg_metalv(seg, s)); } - seg->area_len += extents / area_multiple; seg->len += extents; + if (seg_is_raid(seg)) + seg->area_len = seg->len; + else + seg->area_len += extents / area_multiple; if (!_setup_lv_size(lv, lv->le_count + extents)) return_0; @@ -6309,7 +6330,6 @@ static int _lv_update_and_reload(struct logical_volume *lv, int origin_only) log_very_verbose("Updating logical volume %s on disk(s)%s.", display_lvname(lock_lv), origin_only ? " (origin only)": ""); - if (!vg_write(vg)) return_0; @@ -6776,8 +6796,8 @@ struct logical_volume *insert_layer_for_lv(struct cmd_context *cmd, return_NULL; /* allocate a new linear segment */ - if (!(mapseg = alloc_lv_segment(segtype, lv_where, 0, layer_lv->le_count, - status, 0, NULL, 1, layer_lv->le_count, + if (!(mapseg = alloc_lv_segment(segtype, lv_where, 0, layer_lv->le_count, 0, + status, 0, NULL, 1, layer_lv->le_count, 0, 0, 0, 0, NULL))) return_NULL; @@ -6833,8 +6853,8 @@ static int _extend_layer_lv_for_segment(struct logical_volume *layer_lv, /* allocate a new segment */ if (!(mapseg = alloc_lv_segment(segtype, layer_lv, layer_lv->le_count, - seg->area_len, status, 0, - NULL, 1, seg->area_len, 0, 0, 0, seg))) + seg->area_len, 0, status, 0, + NULL, 1, seg->area_len, 0, 0, 0, 0, seg))) return_0; /* map the new segment to the original underlying are */ diff --git a/lib/metadata/merge.c b/lib/metadata/merge.c index 63118182f..f82e1e70d 100644 --- a/lib/metadata/merge.c +++ b/lib/metadata/merge.c @@ -236,7 +236,7 @@ static void _check_raid_seg(struct lv_segment *seg, int *error_count) if (!seg->areas) raid_seg_error("zero areas"); - if (seg->extents_copied > seg->area_len) + if (seg->extents_copied > seg->len) raid_seg_error_val("extents_copied too large", seg->extents_copied); /* Default < 10, change once raid1 split shift and rename SubLVs works! */ @@ -475,7 +475,7 @@ int check_lv_segments(struct logical_volume *lv, int complete_vg) struct lv_segment *seg, *seg2; uint32_t le = 0; unsigned seg_count = 0, seg_found, external_lv_found = 0; - uint32_t area_multiplier, s; + uint32_t data_rimage_count, s; struct seg_list *sl; struct glv_list *glvl; int error_count = 0; @@ -498,13 +498,13 @@ int check_lv_segments(struct logical_volume *lv, int complete_vg) inc_error_count; } - area_multiplier = segtype_is_striped(seg->segtype) ? - seg->area_count : 1; - - if (seg->area_len * area_multiplier != seg->len) { - log_error("LV %s: segment %u has inconsistent " - "area_len %u", - lv->name, seg_count, seg->area_len); + data_rimage_count = seg->area_count - seg->segtype->parity_devs; + /* FIXME: raid varies seg->area_len? */ + if (seg->len != seg->area_len && + seg->len != seg->area_len * data_rimage_count) { + log_error("LV %s: segment %u with len=%u " + " has inconsistent area_len %u", + lv->name, seg_count, seg->len, seg->area_len); inc_error_count; } @@ -766,10 +766,10 @@ static int _lv_split_segment(struct logical_volume *lv, struct lv_segment *seg, /* Clone the existing segment */ if (!(split_seg = alloc_lv_segment(seg->segtype, - seg->lv, seg->le, seg->len, + seg->lv, seg->le, seg->len, seg->reshape_len, seg->status, seg->stripe_size, seg->log_lv, - seg->area_count, seg->area_len, + seg->area_count, seg->area_len, seg->data_copies, seg->chunk_size, seg->region_size, seg->extents_copied, seg->pvmove_source_seg))) { log_error("Couldn't allocate cloned LV segment."); diff --git a/lib/metadata/metadata-exported.h b/lib/metadata/metadata-exported.h index e4d2267d0..11fb24701 100644 --- a/lib/metadata/metadata-exported.h +++ b/lib/metadata/metadata-exported.h @@ -1,6 +1,6 @@ /* * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. - * Copyright (C) 2004-2016 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved. * * This file is part of LVM2. * @@ -137,7 +137,11 @@ e.g. to prohibit allocation of a RAID image on a PV already holing an image of the RAID set */ #define LOCKD_SANLOCK_LV UINT64_C(0x0080000000000000) /* LV - Internal use only */ -/* Next unused flag: UINT64_C(0x0100000000000000) */ +#define LV_RESHAPE_DELTA_DISKS_PLUS UINT64_C(0x0100000000000000) /* LV reshape flag delta disks plus image(s) */ +#define LV_RESHAPE_DELTA_DISKS_MINUS UINT64_C(0x0200000000000000) /* LV reshape flag delta disks minus image(s) */ + +#define LV_REMOVE_AFTER_RESHAPE UINT64_C(0x0400000000000000) /* LV needs to be removed after a shrinking reshape */ +/* Next unused flag: UINT64_C(0x0800000000000000) */ /* Format features flags */ #define FMT_SEGMENTS 0x00000001U /* Arbitrary segment params? */ @@ -446,6 +450,7 @@ struct lv_segment { const struct segment_type *segtype; uint32_t le; uint32_t len; + uint32_t reshape_len; /* For RAID: user hidden additional out of place reshaping length off area_len and len */ uint64_t status; @@ -454,6 +459,7 @@ struct lv_segment { uint32_t writebehind; /* For RAID (RAID1 only) */ uint32_t min_recovery_rate; /* For RAID */ uint32_t max_recovery_rate; /* For RAID */ + uint32_t data_offset; /* For RAID: data offset in sectors on each data component image */ uint32_t area_count; uint32_t area_len; uint32_t chunk_size; /* For snapshots/thin_pool. In sectors. */ @@ -464,6 +470,7 @@ struct lv_segment { struct logical_volume *cow; struct dm_list origin_list; uint32_t region_size; /* For mirrors, replicators - in sectors */ + uint32_t data_copies; /* For RAID: number of data copies (e.g. 3 for RAID 6 */ uint32_t extents_copied;/* Number of extents synced for raids/mirrors */ struct logical_volume *log_lv; struct lv_segment *pvmove_source_seg; @@ -1205,7 +1212,8 @@ struct logical_volume *first_replicator_dev(const struct logical_volume *lv); int lv_is_raid_with_tracking(const struct logical_volume *lv); uint32_t lv_raid_image_count(const struct logical_volume *lv); int lv_raid_change_image_count(struct logical_volume *lv, - uint32_t new_count, struct dm_list *allocate_pvs); + uint32_t new_count, + struct dm_list *allocate_pvs); int lv_raid_split(struct logical_volume *lv, const char *split_name, uint32_t new_count, struct dm_list *splittable_pvs); int lv_raid_split_and_track(struct logical_volume *lv, @@ -1233,6 +1241,7 @@ uint32_t raid_ensure_min_region_size(const struct logical_volume *lv, uint64_t r int lv_raid_change_region_size(struct logical_volume *lv, int yes, int force, uint32_t new_region_size); int lv_raid_in_sync(const struct logical_volume *lv); +uint32_t lv_raid_data_copies(const struct segment_type *segtype, uint32_t area_count); /* -- metadata/raid_manip.c */ /* ++ metadata/cache_manip.c */ diff --git a/lib/metadata/raid_manip.c b/lib/metadata/raid_manip.c index b6215a2dd..9ecc41068 100644 --- a/lib/metadata/raid_manip.c +++ b/lib/metadata/raid_manip.c @@ -1952,9 +1952,9 @@ static int _alloc_and_add_new_striped_segment(struct logical_volume *lv, /* Allocate a segment with seg->area_count areas */ if (!(new_seg = alloc_lv_segment(striped_segtype, lv, le, area_len * seg->area_count, - 0, + 0, 0, seg->stripe_size, NULL, seg->area_count, - area_len, seg->chunk_size, 0, 0, NULL))) + area_len, 0, seg->chunk_size, 0, 0, NULL))) return_0; dm_list_add(new_segments, &new_seg->list); @@ -2510,8 +2510,8 @@ static int _striped_to_raid0_move_segs_to_raid0_lvs(struct logical_volume *lv, if (!(seg_new = alloc_lv_segment(segtype, dlv, le, seg_from->area_len, status, - 0 /* stripe_size */, NULL, 1 /* area_count */, - seg_from->area_len, + 0, 0 /* stripe_size */, NULL, 1 /* area_count */, + seg_from->area_len, 0, 0 /* chunk_size */, 0 /* region_size */, 0, NULL))) return_0; @@ -2726,9 +2726,9 @@ static struct lv_segment *_convert_striped_to_raid0(struct logical_volume *lv, seg = first_seg(dm_list_item(dm_list_first(&data_lvs), struct lv_list)->lv); if (!(raid0_seg = alloc_lv_segment(segtype, lv, 0 /* le */, lv->le_count /* len */, - 0, + 0, 0, stripe_size, NULL /* log_lv */, - area_count, area_len, + area_count, area_len, 0, 0 /* chunk_size */, 0 /* seg->region_size */, 0u /* extents_copied */ , NULL /* pvmove_source_seg */))) { diff --git a/lib/metadata/segtype.c b/lib/metadata/segtype.c index b66ab0278..d0508ca35 100644 --- a/lib/metadata/segtype.c +++ b/lib/metadata/segtype.c @@ -43,7 +43,8 @@ struct segment_type *get_segtype_from_flag(struct cmd_context *cmd, uint64_t fla { struct segment_type *segtype; - dm_list_iterate_items(segtype, &cmd->segtypes) + /* Iterate backwards to provide aliases; e.g. raid5 instead of raid5_ls */ + dm_list_iterate_back_items(segtype, &cmd->segtypes) if (flag & segtype->flags) return segtype; diff --git a/lib/metadata/segtype.h b/lib/metadata/segtype.h index 921282449..bea714189 100644 --- a/lib/metadata/segtype.h +++ b/lib/metadata/segtype.h @@ -140,7 +140,11 @@ struct dev_manager; #define segtype_is_any_raid10(segtype) ((segtype)->flags & SEG_RAID10 ? 1 : 0) #define segtype_is_raid10(segtype) ((segtype)->flags & SEG_RAID10 ? 1 : 0) #define segtype_is_raid10_near(segtype) segtype_is_raid10(segtype) +/* FIXME: once raid10_offset supported */ +#define segtype_is_raid10_offset(segtype) 0 // ((segtype)->flags & SEG_RAID10_OFFSET ? 1 : 0) #define segtype_is_raid_with_meta(segtype) (segtype_is_raid(segtype) && !segtype_is_raid0(segtype)) +#define segtype_is_striped_raid(segtype) (segtype_is_raid(segtype) && !segtype_is_raid1(segtype)) +#define segtype_is_reshapable_raid(segtype) ((segtype_is_striped_raid(segtype) && !segtype_is_any_raid0(segtype)) || segtype_is_raid10_near(segtype) || segtype_is_raid10_offset(segtype)) #define segtype_is_snapshot(segtype) ((segtype)->flags & SEG_SNAPSHOT ? 1 : 0) #define segtype_is_striped(segtype) ((segtype)->flags & SEG_AREAS_STRIPED ? 1 : 0) #define segtype_is_thin(segtype) ((segtype)->flags & (SEG_THIN_POOL|SEG_THIN_VOLUME) ? 1 : 0) @@ -190,6 +194,8 @@ struct dev_manager; #define seg_is_raid10(seg) segtype_is_raid10((seg)->segtype) #define seg_is_raid10_near(seg) segtype_is_raid10_near((seg)->segtype) #define seg_is_raid_with_meta(seg) segtype_is_raid_with_meta((seg)->segtype) +#define seg_is_striped_raid(seg) segtype_is_striped_raid((seg)->segtype) +#define seg_is_reshapable_raid(seg) segtype_is_reshapable_raid((seg)->segtype) #define seg_is_replicator(seg) ((seg)->segtype->flags & SEG_REPLICATOR ? 1 : 0) #define seg_is_replicator_dev(seg) ((seg)->segtype->flags & SEG_REPLICATOR_DEV ? 1 : 0) #define seg_is_snapshot(seg) segtype_is_snapshot((seg)->segtype) @@ -280,6 +286,7 @@ struct segment_type *init_unknown_segtype(struct cmd_context *cmd, #define RAID_FEATURE_RAID0 (1U << 1) /* version 1.7 */ #define RAID_FEATURE_RESHAPING (1U << 2) /* version 1.8 */ #define RAID_FEATURE_RAID4 (1U << 3) /* ! version 1.8 or 1.9.0 */ +#define RAID_FEATURE_RESHAPE (1U << 4) /* version 1.10.2 */ #ifdef RAID_INTERNAL int init_raid_segtypes(struct cmd_context *cmd, struct segtype_library *seglib); diff --git a/lib/metadata/snapshot_manip.c b/lib/metadata/snapshot_manip.c index b5fb60c4a..57fbef93b 100644 --- a/lib/metadata/snapshot_manip.c +++ b/lib/metadata/snapshot_manip.c @@ -238,8 +238,8 @@ static struct lv_segment *_alloc_snapshot_seg(struct logical_volume *lv) return NULL; } - if (!(seg = alloc_lv_segment(segtype, lv, 0, lv->le_count, 0, 0, - NULL, 0, lv->le_count, 0, 0, 0, NULL))) { + if (!(seg = alloc_lv_segment(segtype, lv, 0, lv->le_count, 0, 0, 0, + NULL, 0, lv->le_count, 0, 0, 0, 0, NULL))) { log_error("Couldn't allocate new snapshot segment."); return NULL; } diff --git a/lib/raid/raid.c b/lib/raid/raid.c index c679207af..398e8b004 100644 --- a/lib/raid/raid.c +++ b/lib/raid/raid.c @@ -137,6 +137,7 @@ static int _raid_text_import(struct lv_segment *seg, } raid_attr_import[] = { { "region_size", &seg->region_size }, { "stripe_size", &seg->stripe_size }, + { "data_copies", &seg->data_copies }, { "writebehind", &seg->writebehind }, { "min_recovery_rate", &seg->min_recovery_rate }, { "max_recovery_rate", &seg->max_recovery_rate }, @@ -146,6 +147,10 @@ static int _raid_text_import(struct lv_segment *seg, for (i = 0; i < DM_ARRAY_SIZE(raid_attr_import); i++, aip++) { if (dm_config_has_node(sn, aip->name)) { if (!dm_config_get_uint32(sn, aip->name, aip->var)) { + if (!strcmp(aip->name, "data_copies")) { + *aip->var = 0; + continue; + } log_error("Couldn't read '%s' for segment %s of logical volume %s.", aip->name, dm_config_parent_name(sn), seg->lv->name); return 0; @@ -165,6 +170,9 @@ static int _raid_text_import(struct lv_segment *seg, return 0; } + if (seg->data_copies < 2) + seg->data_copies = 0; // lv_raid_data_copies(seg->segtype, seg->area_count); + if (seg_is_any_raid0(seg)) seg->area_len /= seg->area_count; @@ -183,18 +191,31 @@ static int _raid_text_export_raid0(const struct lv_segment *seg, struct formatte static int _raid_text_export_raid(const struct lv_segment *seg, struct formatter *f) { - outf(f, "device_count = %u", seg->area_count); + int raid0 = seg_is_any_raid0(seg); + + if (raid0) + outfc(f, (seg->area_count == 1) ? "# linear" : NULL, + "stripe_count = %u", seg->area_count); + + else { + outf(f, "device_count = %u", seg->area_count); + if (seg_is_any_raid10(seg) && seg->data_copies > 0) + outf(f, "data_copies = %" PRIu32, seg->data_copies); + if (seg->region_size) + outf(f, "region_size = %" PRIu32, seg->region_size); + } if (seg->stripe_size) outf(f, "stripe_size = %" PRIu32, seg->stripe_size); - if (seg->region_size) - outf(f, "region_size = %" PRIu32, seg->region_size); - if (seg->writebehind) - outf(f, "writebehind = %" PRIu32, seg->writebehind); - if (seg->min_recovery_rate) - outf(f, "min_recovery_rate = %" PRIu32, seg->min_recovery_rate); - if (seg->max_recovery_rate) - outf(f, "max_recovery_rate = %" PRIu32, seg->max_recovery_rate); + + if (!raid0) { + if (seg_is_raid1(seg) && seg->writebehind) + outf(f, "writebehind = %" PRIu32, seg->writebehind); + if (seg->min_recovery_rate) + outf(f, "min_recovery_rate = %" PRIu32, seg->min_recovery_rate); + if (seg->max_recovery_rate) + outf(f, "max_recovery_rate = %" PRIu32, seg->max_recovery_rate); + } return out_areas(f, seg, "raid"); } @@ -216,14 +237,16 @@ static int _raid_add_target_line(struct dev_manager *dm __attribute__((unused)), struct dm_tree_node *node, uint64_t len, uint32_t *pvmove_mirror_count __attribute__((unused))) { + int delta_disks = 0, delta_disks_minus = 0, delta_disks_plus = 0, data_offset = 0; uint32_t s; uint64_t flags = 0; - uint64_t rebuilds = 0; - uint64_t writemostly = 0; + uint64_t rebuilds[RAID_BITMAP_SIZE]; + uint64_t writemostly[RAID_BITMAP_SIZE]; struct dm_tree_node_raid_params params; - int raid0 = seg_is_any_raid0(seg); memset(¶ms, 0, sizeof(params)); + memset(&rebuilds, 0, sizeof(rebuilds)); + memset(&writemostly, 0, sizeof(writemostly)); if (!seg->area_count) { log_error(INTERNAL_ERROR "_raid_add_target_line called " @@ -232,63 +255,84 @@ static int _raid_add_target_line(struct dev_manager *dm __attribute__((unused)), } /* - * 64 device restriction imposed by kernel as well. It is - * not strictly a userspace limitation. + * 253 device restriction imposed by kernel due to MD and dm-raid bitfield limitation in superblock. + * It is not strictly a userspace limitation. */ - if (seg->area_count > 64) { - log_error("Unable to handle more than 64 devices in a " - "single RAID array"); + if (seg->area_count > DEFAULT_RAID_MAX_IMAGES) { + log_error("Unable to handle more than %u devices in a " + "single RAID array", DEFAULT_RAID_MAX_IMAGES); return 0; } - if (!raid0) { + if (!seg_is_any_raid0(seg)) { if (!seg->region_size) { - log_error("Missing region size for mirror segment."); + log_error("Missing region size for raid segment in %s.", + seg_lv(seg, 0)->name); return 0; } - for (s = 0; s < seg->area_count; s++) - if (seg_lv(seg, s)->status & LV_REBUILD) - rebuilds |= 1ULL << s; + for (s = 0; s < seg->area_count; s++) { + uint64_t status = seg_lv(seg, s)->status; + + if (status & LV_REBUILD) + rebuilds[s/64] |= 1ULL << (s%64); + + if (status & LV_RESHAPE_DELTA_DISKS_PLUS) { + delta_disks++; + delta_disks_plus++; + } else if (status & LV_RESHAPE_DELTA_DISKS_MINUS) { + delta_disks--; + delta_disks_minus++; + } + + if (delta_disks_plus && delta_disks_minus) { + log_error(INTERNAL_ERROR "Invalid request for delta disks minus and delta disks plus!"); + return 0; + } + + if (status & LV_WRITEMOSTLY) + writemostly[s/64] |= 1ULL << (s%64); + } - for (s = 0; s < seg->area_count; s++) - if (seg_lv(seg, s)->status & LV_WRITEMOSTLY) - writemostly |= 1ULL << s; + data_offset = seg->data_offset; if (mirror_in_sync()) flags = DM_NOSYNC; } params.raid_type = lvseg_name(seg); - params.stripe_size = seg->stripe_size; - params.flags = flags; - if (raid0) { - params.mirrors = 1; - params.stripes = seg->area_count; - } else if (seg->segtype->parity_devs) { + if (seg->segtype->parity_devs) { /* RAID 4/5/6 */ params.mirrors = 1; params.stripes = seg->area_count - seg->segtype->parity_devs; - } else if (seg_is_raid10(seg)) { - /* RAID 10 only supports 2 mirrors now */ - params.mirrors = 2; - params.stripes = seg->area_count / 2; + } else if (seg_is_any_raid0(seg)) { + params.mirrors = 1; + params.stripes = seg->area_count; + } else if (seg_is_any_raid10(seg)) { + params.data_copies = seg->data_copies; + params.stripes = seg->area_count; } else { /* RAID 1 */ - params.mirrors = seg->area_count; + params.mirrors = seg->data_copies; params.stripes = 1; params.writebehind = seg->writebehind; + memcpy(params.writemostly, writemostly, sizeof(params.writemostly)); } - if (!raid0) { + /* RAID 0 doesn't have a bitmap, thus no region_size, rebuilds etc. */ + if (!seg_is_any_raid0(seg)) { params.region_size = seg->region_size; - params.rebuilds = rebuilds; - params.writemostly = writemostly; + memcpy(params.rebuilds, rebuilds, sizeof(params.rebuilds)); params.min_recovery_rate = seg->min_recovery_rate; params.max_recovery_rate = seg->max_recovery_rate; + params.delta_disks = delta_disks; + params.data_offset = data_offset; } + params.stripe_size = seg->stripe_size; + params.flags = flags; + if (!dm_tree_node_add_raid_target_with_params(node, len, ¶ms)) return_0; @@ -450,6 +494,10 @@ static int _raid_target_present(struct cmd_context *cmd, else log_very_verbose("Target raid does not support %s.", SEG_TYPE_NAME_RAID4); + + if (maj > 1 || + (maj == 1 && (min > 10 || (min == 10 && patchlevel >= 2)))) + _raid_attrs |= RAID_FEATURE_RESHAPE; } if (attributes) diff --git a/libdm/libdevmapper.h b/libdm/libdevmapper.h index 9a1025202..7fccac560 100644 --- a/libdm/libdevmapper.h +++ b/libdm/libdevmapper.h @@ -331,6 +331,7 @@ struct dm_status_raid { char *dev_health; /* idle, frozen, resync, recover, check, repair */ char *sync_action; + uint64_t data_offset; /* RAID out-of-place reshaping */ }; int dm_get_status_raid(struct dm_pool *mem, const char *params, @@ -1719,7 +1720,7 @@ int dm_tree_node_add_raid_target(struct dm_tree_node *node, const char *raid_type, uint32_t region_size, uint32_t stripe_size, - uint64_t rebuilds, + uint64_t *rebuilds, uint64_t flags); /* @@ -1738,6 +1739,8 @@ int dm_tree_node_add_raid_target(struct dm_tree_node *node, */ #define DM_CACHE_METADATA_MAX_SECTORS DM_THIN_METADATA_MAX_SECTORS +#define RAID_BITMAP_SIZE 4 + struct dm_tree_node_raid_params { const char *raid_type; @@ -1746,18 +1749,23 @@ struct dm_tree_node_raid_params { uint32_t region_size; uint32_t stripe_size; + int delta_disks; /* +/- number of disks to add/remove (reshaping) */ + int data_offset; /* data offset to set (out-of-place reshaping) */ + /* * 'rebuilds' and 'writemostly' are bitfields that signify * which devices in the array are to be rebuilt or marked - * writemostly. By choosing a 'uint64_t', we limit ourself - * to RAID arrays with 64 devices. + * writemostly. The kernel supports up to 253 legs. + * We limit ourselvs by choosing a lower value + * for DEFAULT_RAID_MAX_IMAGES. */ - uint64_t rebuilds; - uint64_t writemostly; - uint32_t writebehind; /* I/Os (kernel default COUNTER_MAX / 2) */ + uint64_t rebuilds[RAID_BITMAP_SIZE]; + uint64_t writemostly[RAID_BITMAP_SIZE]; + uint32_t writebehind; /* I/Os (kernel default COUNTER_MAX / 2) */ uint32_t sync_daemon_sleep; /* ms (kernel default = 5sec) */ uint32_t max_recovery_rate; /* kB/sec/disk */ uint32_t min_recovery_rate; /* kB/sec/disk */ + uint32_t data_copies; /* RAID # of data copies */ uint32_t stripe_cache; /* sectors */ uint64_t flags; /* [no]sync */ diff --git a/libdm/libdm-common.h b/libdm/libdm-common.h index 4dc1870fc..a064db846 100644 --- a/libdm/libdm-common.h +++ b/libdm/libdm-common.h @@ -23,6 +23,8 @@ #define DEV_NAME(dmt) (dmt->mangled_dev_name ? : dmt->dev_name) #define DEV_UUID(DMT) (dmt->mangled_uuid ? : dmt->uuid) +#define RAID_BITMAP_SIZE 4 + int mangle_string(const char *str, const char *str_name, size_t len, char *buf, size_t buf_len, dm_string_mangling_t mode); diff --git a/libdm/libdm-deptree.c b/libdm/libdm-deptree.c index d658bf99d..a26cfcc0f 100644 --- a/libdm/libdm-deptree.c +++ b/libdm/libdm-deptree.c @@ -205,11 +205,14 @@ struct load_segment { struct dm_tree_node *replicator;/* Replicator-dev */ uint64_t rdevice_index; /* Replicator-dev */ - uint64_t rebuilds; /* raid */ - uint64_t writemostly; /* raid */ + int delta_disks; /* raid reshape number of disks */ + int data_offset; /* raid reshape data offset on disk to set */ + uint64_t rebuilds[RAID_BITMAP_SIZE]; /* raid */ + uint64_t writemostly[RAID_BITMAP_SIZE]; /* raid */ uint32_t writebehind; /* raid */ uint32_t max_recovery_rate; /* raid kB/sec/disk */ uint32_t min_recovery_rate; /* raid kB/sec/disk */ + uint32_t data_copies; /* raid10 data_copies */ struct dm_tree_node *metadata; /* Thin_pool + Cache */ struct dm_tree_node *pool; /* Thin_pool, Thin */ @@ -2353,16 +2356,21 @@ static int _mirror_emit_segment_line(struct dm_task *dmt, struct load_segment *s return 1; } -/* Is parameter non-zero? */ -#define PARAM_IS_SET(p) ((p) ? 1 : 0) +static int _2_if_value(unsigned p) +{ + return p ? 2 : 0; +} -/* Return number of bits assuming 4 * 64 bit size */ -static int _get_params_count(uint64_t bits) +/* Return number of bits passed in @bits assuming 2 * 64 bit size */ +static int _get_params_count(uint64_t *bits) { int r = 0; + int i = RAID_BITMAP_SIZE; - r += 2 * hweight32(bits & 0xFFFFFFFF); - r += 2 * hweight32(bits >> 32); + while (i--) { + r += 2 * hweight32(bits[i] & 0xFFFFFFFF); + r += 2 * hweight32(bits[i] >> 32); + } return r; } @@ -2373,32 +2381,60 @@ static int _raid_emit_segment_line(struct dm_task *dmt, uint32_t major, size_t paramsize) { uint32_t i; + uint32_t area_count = seg->area_count / 2; int param_count = 1; /* mandatory 'chunk size'/'stripe size' arg */ int pos = 0; - unsigned type = seg->type; + unsigned type; + + if (seg->area_count % 2) + return 0; if ((seg->flags & DM_NOSYNC) || (seg->flags & DM_FORCESYNC)) param_count++; - param_count += 2 * (PARAM_IS_SET(seg->region_size) + - PARAM_IS_SET(seg->writebehind) + - PARAM_IS_SET(seg->min_recovery_rate) + - PARAM_IS_SET(seg->max_recovery_rate)); + param_count += _2_if_value(seg->data_offset) + + _2_if_value(seg->delta_disks) + + _2_if_value(seg->region_size) + + _2_if_value(seg->writebehind) + + _2_if_value(seg->min_recovery_rate) + + _2_if_value(seg->max_recovery_rate) + + _2_if_value(seg->data_copies > 1); - /* rebuilds and writemostly are 64 bits */ + /* rebuilds and writemostly are BITMAP_SIZE * 64 bits */ param_count += _get_params_count(seg->rebuilds); param_count += _get_params_count(seg->writemostly); - if ((type == SEG_RAID1) && seg->stripe_size) - log_error("WARNING: Ignoring RAID1 stripe size"); + if ((seg->type == SEG_RAID1) && seg->stripe_size) + log_info("WARNING: Ignoring RAID1 stripe size"); /* Kernel only expects "raid0", not "raid0_meta" */ + type = seg->type; if (type == SEG_RAID0_META) type = SEG_RAID0; +#if 0 + /* Kernel only expects "raid10", not "raid10_{far,offset}" */ + else if (type == SEG_RAID10_FAR || + type == SEG_RAID10_OFFSET) { + param_count += 2; + type = SEG_RAID10_NEAR; + } +#endif - EMIT_PARAMS(pos, "%s %d %u", _dm_segtypes[type].target, + EMIT_PARAMS(pos, "%s %d %u", + // type == SEG_RAID10_NEAR ? "raid10" : _dm_segtypes[type].target, + type == SEG_RAID10 ? "raid10" : _dm_segtypes[type].target, param_count, seg->stripe_size); +#if 0 + if (seg->type == SEG_RAID10_FAR) + EMIT_PARAMS(pos, " raid10_format far"); + else if (seg->type == SEG_RAID10_OFFSET) + EMIT_PARAMS(pos, " raid10_format offset"); +#endif + + if (seg->data_copies > 1 && type == SEG_RAID10) + EMIT_PARAMS(pos, " raid10_copies %u", seg->data_copies); + if (seg->flags & DM_NOSYNC) EMIT_PARAMS(pos, " nosync"); else if (seg->flags & DM_FORCESYNC) @@ -2407,27 +2443,38 @@ static int _raid_emit_segment_line(struct dm_task *dmt, uint32_t major, if (seg->region_size) EMIT_PARAMS(pos, " region_size %u", seg->region_size); - for (i = 0; i < (seg->area_count / 2); i++) - if (seg->rebuilds & (1ULL << i)) - EMIT_PARAMS(pos, " rebuild %u", i); + /* If seg-data_offset == 1, kernel needs a zero offset to adjust to it */ + if (seg->data_offset) + EMIT_PARAMS(pos, " data_offset %d", seg->data_offset == 1 ? 0 : seg->data_offset); - if (seg->min_recovery_rate) - EMIT_PARAMS(pos, " min_recovery_rate %u", - seg->min_recovery_rate); + if (seg->delta_disks) + EMIT_PARAMS(pos, " delta_disks %d", seg->delta_disks); - if (seg->max_recovery_rate) - EMIT_PARAMS(pos, " max_recovery_rate %u", - seg->max_recovery_rate); + for (i = 0; i < area_count; i++) + if (seg->rebuilds[i/64] & (1ULL << (i%64))) + EMIT_PARAMS(pos, " rebuild %u", i); - for (i = 0; i < (seg->area_count / 2); i++) - if (seg->writemostly & (1ULL << i)) + for (i = 0; i < area_count; i++) + if (seg->writemostly[i/64] & (1ULL << (i%64))) EMIT_PARAMS(pos, " write_mostly %u", i); if (seg->writebehind) EMIT_PARAMS(pos, " max_write_behind %u", seg->writebehind); + /* + * Has to be before "min_recovery_rate" or the kernels + * check will fail when both set and min > previous max + */ + if (seg->max_recovery_rate) + EMIT_PARAMS(pos, " max_recovery_rate %u", + seg->max_recovery_rate); + + if (seg->min_recovery_rate) + EMIT_PARAMS(pos, " min_recovery_rate %u", + seg->min_recovery_rate); + /* Print number of metadata/data device pairs */ - EMIT_PARAMS(pos, " %u", seg->area_count/2); + EMIT_PARAMS(pos, " %u", area_count); if (_emit_areas_line(dmt, seg, params, paramsize, &pos) <= 0) return_0; @@ -3267,11 +3314,14 @@ int dm_tree_node_add_raid_target_with_params(struct dm_tree_node *node, seg->region_size = p->region_size; seg->stripe_size = p->stripe_size; seg->area_count = 0; - seg->rebuilds = p->rebuilds; - seg->writemostly = p->writemostly; + seg->delta_disks = p->delta_disks; + seg->data_offset = p->data_offset; + memcpy(seg->rebuilds, p->rebuilds, sizeof(seg->rebuilds)); + memcpy(seg->writemostly, p->writemostly, sizeof(seg->writemostly)); seg->writebehind = p->writebehind; seg->min_recovery_rate = p->min_recovery_rate; seg->max_recovery_rate = p->max_recovery_rate; + seg->data_copies = p->data_copies; seg->flags = p->flags; return 1; @@ -3282,17 +3332,18 @@ int dm_tree_node_add_raid_target(struct dm_tree_node *node, const char *raid_type, uint32_t region_size, uint32_t stripe_size, - uint64_t rebuilds, + uint64_t *rebuilds, uint64_t flags) { struct dm_tree_node_raid_params params = { .raid_type = raid_type, .region_size = region_size, .stripe_size = stripe_size, - .rebuilds = rebuilds, .flags = flags }; + memcpy(params.rebuilds, rebuilds, sizeof(params.rebuilds)); + return dm_tree_node_add_raid_target_with_params(node, size, ¶ms); } diff --git a/libdm/libdm-targets.c b/libdm/libdm-targets.c index c94e05753..6b8337234 100644 --- a/libdm/libdm-targets.c +++ b/libdm/libdm-targets.c @@ -89,6 +89,8 @@ static unsigned _count_fields(const char *p) * <#devs> * Versions 1.5.0+ (6 fields): * <#devs> + * Versions 1.9.0+ (7 fields): + * <#devs> */ int dm_get_status_raid(struct dm_pool *mem, const char *params, struct dm_status_raid **status) @@ -147,6 +149,22 @@ int dm_get_status_raid(struct dm_pool *mem, const char *params, if (sscanf(p, "%s %" PRIu64, s->sync_action, &s->mismatch_count) != 2) goto_bad; + if (num_fields < 7) + goto out; + + /* + * All pre-1.9.0 version parameters are read. Now we check + * for additional 1.9.0+ parameters (i.e. nr_fields at least 7). + * + * Note that data_offset will be 0 if the + * kernel returns a pre-1.9.0 status. + */ + msg_fields = ""; + if (!(p = _skip_fields(params, 6))) /* skip pre-1.9.0 params */ + goto bad; + if (sscanf(p, "%" PRIu64, &s->data_offset) != 1) + goto bad; + out: *status = s; diff --git a/tools/lvconvert.c b/tools/lvconvert.c index b9e0665c7..abec04a7e 100644 --- a/tools/lvconvert.c +++ b/tools/lvconvert.c @@ -1228,6 +1228,9 @@ static int _lvconvert_mirrors(struct cmd_context *cmd, static int _is_valid_raid_conversion(const struct segment_type *from_segtype, const struct segment_type *to_segtype) { + if (!from_segtype) + return 1; + if (from_segtype == to_segtype) return 1; @@ -1356,7 +1359,7 @@ static int _lvconvert_raid(struct logical_volume *lv, struct lvconvert_params *l DEFAULT_RAID1_MAX_IMAGES, lp->segtype->name, display_lvname(lv)); return 0; } - if (!lv_raid_change_image_count(lv, image_count, lp->pvh)) + if (!lv_raid_change_image_count(lv, image_count, /* lp->region_size, */ lp->pvh)) return_0; log_print_unless_silent("Logical volume %s successfully converted.", @@ -1365,10 +1368,13 @@ static int _lvconvert_raid(struct logical_volume *lv, struct lvconvert_params *l return 1; } goto try_new_takeover_or_reshape; - } else if (!*lp->type_str || seg->segtype == lp->segtype) { + } +#if 0 + } else if ((!*lp->type_str || seg->segtype == lp->segtype) && !lp->stripe_size_supplied) { log_error("Conversion operation not yet supported."); return 0; } +#endif if ((seg_is_linear(seg) || seg_is_striped(seg) || seg_is_mirrored(seg) || lv_is_raid(lv)) && (lp->type_str && lp->type_str[0])) { @@ -1390,10 +1396,14 @@ static int _lvconvert_raid(struct logical_volume *lv, struct lvconvert_params *l return 0; } + /* FIXME This needs changing globally. */ if (!arg_is_set(cmd, stripes_long_ARG)) lp->stripes = 0; + if (!arg_is_set(cmd, type_ARG)) + lp->segtype = NULL; - if (!lv_raid_convert(lv, lp->segtype, lp->yes, lp->force, lp->stripes, lp->stripe_size_supplied, lp->stripe_size, + if (!lv_raid_convert(lv, lp->segtype, + lp->yes, lp->force, lp->stripes, lp->stripe_size_supplied, lp->stripe_size, lp->region_size, lp->pvh)) return_0; @@ -1410,12 +1420,16 @@ try_new_takeover_or_reshape: /* FIXME This needs changing globally. */ if (!arg_is_set(cmd, stripes_long_ARG)) lp->stripes = 0; + if (!arg_is_set(cmd, type_ARG)) + lp->segtype = NULL; /* Only let raid4 through for now. */ - if (lp->type_str && lp->type_str[0] && lp->segtype != seg->segtype && - ((seg_is_raid4(seg) && seg_is_striped(lp) && lp->stripes > 1) || - (seg_is_striped(seg) && seg->area_count > 1 && seg_is_raid4(lp)))) { - if (!lv_raid_convert(lv, lp->segtype, lp->yes, lp->force, lp->stripes, lp->stripe_size_supplied, lp->stripe_size, + if (!lp->segtype || + (lp->type_str && lp->type_str[0] && lp->segtype != seg->segtype && + ((seg_is_raid4(seg) && seg_is_striped(lp) && lp->stripes > 1) || + (seg_is_striped(seg) && seg->area_count > 1 && seg_is_raid4(lp))))) { + if (!lv_raid_convert(lv, lp->segtype, + lp->yes, lp->force, lp->stripes, lp->stripe_size_supplied, lp->stripe_size, lp->region_size, lp->pvh)) return_0; @@ -1700,6 +1714,8 @@ static int _lvconvert_raid_types(struct cmd_context *cmd, struct logical_volume /* FIXME This is incomplete */ if (_mirror_or_raid_type_requested(cmd, lp->type_str) || _raid0_type_requested(lp->type_str) || _striped_type_requested(lp->type_str) || lp->mirrorlog || lp->corelog) { + if (!arg_is_set(cmd, type_ARG)) + lp->segtype = first_seg(lv)->segtype; /* FIXME Handle +/- adjustments too? */ if (!get_stripe_params(cmd, lp->segtype, &lp->stripes, &lp->stripe_size, &lp->stripes_supplied, &lp->stripe_size_supplied)) goto_out; @@ -2990,9 +3006,9 @@ static int _lvconvert_to_pool(struct cmd_context *cmd, } /* Allocate a new pool segment */ - if (!(seg = alloc_lv_segment(pool_segtype, pool_lv, 0, data_lv->le_count, + if (!(seg = alloc_lv_segment(pool_segtype, pool_lv, 0, data_lv->le_count, 0, pool_lv->status, 0, NULL, 1, - data_lv->le_count, 0, 0, 0, NULL))) + data_lv->le_count, 0, 0, 0, 0, NULL))) return_0; /* Add the new segment to the layer LV */ -- 2.43.5