Restrict lvm1 partial mode.
PATCH 1/1 Restrict lvm1 partial mode.
Current lvm1 allocation code seems to not properly map segments on missing PVs.
For now disable this functionality.
(It never worked and previous commit just introduced segfault here.)
So the partial mode in lvm1 can only process missing PVs with no LV segments only.
Also do not use random PV UUID for missing part but use fixed string derived from VG UUID (to not confuse clvmd tests).
--- lib/format1/format1.c | 28 ++++++++++++++++++++++++++-- 1 files changed, 26 insertions(+), 2 deletions(-)
lib/format1/format1.c +26 -2
@@ -118,6 +118,28 @@ static int _fix_partial_vg(struct volume_group *vg, struct dm_list *pvs)struct disk_list *dl;struct dm_list *pvh;struct pv_list *pvl; +struct lv_list *ll; +struct lv_segment *seg; + +/* +* FIXME: code should remap missing segments to error segment. +* Also current mapping code allocates 1 segment per missing extent. +* For now bail out completely - allocated structures are not complete +*/ + dm_list_iterate_items(ll, &vg->lvs) + dm_list_iterate_items(seg, &ll->lv->segments) { + +/* area_count is always 1 here, s == 0 */ +if (seg_type (seg,0 ) !=AREA_PV ) +continue ; + +if (seg_pv (seg,0 )) +continue ; + +log_error ("Partial mode support for missing lvm1 PVs and " +"partially available LVs is currently not implemented." ); +return 0 ; + } dm_list_iterate(pvh, pvs) { dl = dm_list_item(pvh, struct disk_list);
@@ -129,8 +151,10 @@ static int _fix_partial_vg(struct volume_group *vg, struct dm_list *pvs) !(pvl->pv = dm_pool_zalloc(vg->vgmem, sizeof(*pvl->pv)))) return_0; -if (!id_create (&pvl->pv->id)) - goto_out; +/* Use vg uuid with replaced first chars to "missing" as missing PV UUID */ +memcpy (&pvl->pv->id.uuid, vg->id.uuid,sizeof (pvl->pv->id.uuid)); +memcpy (&pvl->pv->id.uuid,"missing" ,7 ); +if (!(pvl->pv->vg_name =dm_pool_strdup (vg->vgmem, vg->name))) goto_out;memcpy (&pvl->pv->vgid, &vg->id,sizeof (vg->id));
Signature · 1 line
1.7.1
PATCH 1/1 Restrict lvm1 partial mode.
Quoted 20 lines · Milan Broz
lib/format1/format1.c +12
-0@@ -118,6 +118,28 @@ static int _fix_partial_vg(struct volume_group *vg, struct dm_list *pvs)struct disk_list *dl;struct dm_list *pvh;struct pv_list *pvl; +struct lv_list *ll; +struct lv_segment *seg; + +/* +* FIXME: code should remap missing segments to error segment. +* Also current mapping code allocates 1 segment per missing extent. +* For now bail out completely - allocated structures are not complete +*/ + dm_list_iterate_items(ll, &vg->lvs) + dm_list_iterate_items(seg, &ll->lv->segments) { + +/* area_count is always 1 here, s == 0 */
lib/format1/format1.c static int _fix_partial_vg(struct volume_group *vg, struct dm_list *pvs) · line 142
Quoted 10 lines
lib/format1/format1.c +10
-0+if (seg_type (seg,0 ) !=AREA_PV ) +continue ; + +if (seg_pv (seg,0 )) +continue ; + +log_error ("Partial mode support for missing lvm1 PVs and " +"partially available LVs is currently not implemented." ); +return 0 ; + }
OK.
lib/format1/format1.c static int _fix_partial_vg(struct volume_group *vg, struct dm_list *pvs) · line 156
Quoted 11 lines
lib/format1/format1.c +3
-2dm_list_iterate(pvh, pvs) { dl = dm_list_item(pvh, struct disk_list);@@ -129,8 +151,10 @@ static int _fix_partial_vg(struct volume_group *vg, struct dm_list *pvs) !(pvl->pv = dm_pool_zalloc(vg->vgmem, sizeof(*pvl->pv)))) return_0; -if (!id_create (&pvl->pv->id)) - goto_out; +/* Use vg uuid with replaced first chars to "missing" as missing PV UUID */ +memcpy (&pvl->pv->id.uuid, vg->id.uuid,sizeof (pvl->pv->id.uuid)); +memcpy (&pvl->pv->id.uuid,"missing" ,7 );
Does what the comment says... would the result be a bit less confusing though if it said
memcpy(&pvl->pv->id.uuid, "missing_", 8);
instead? I know, a nitpick.
Other than that,
Yours,
Petr.
PS: Would it be possible to have automated tests for LVM1 stuff as well? The below seems to work (i.e. new LVM1 VGs can still be created), so we could just add couple tests for missing PV behaviour with LVM1...
aux prepare_devs 4
pvcreate -M1 $devs
vgcreate -M1 $vg $devs
vgs -o +vg_fmt | grep lvm1
PATCH 1/1 Restrict lvm1 partial mode.
Quoted 3 lines · Petr Rockai
PS: Would it be possible to have automated tests for LVM1 stuff as well? The below seems to work (i.e. new LVM1 VGs can still be created), so we could just add couple tests for missing PV behaviour with LVM1...
It depends if we want to spend resources for supporting lvm1 format forever.
It is very rarely used today and I can remember only troubles with it:-)
Milan
PATCH 1/1 Restrict lvm1 partial mode.
In reply to Petr RockaiQuoted 6 lines · Petr Rockai
Does what the comment says... would the result be a bit less confusing though if it said
memcpy(&pvl->pv->id.uuid, "missing_", 8);instead? I know, a nitpick.
negative. Because we do not allow "_" in UUID :-)
m.