Avoid scanning all pvs in the system if pvcreating on a device with mdas.
PATCH 1/1 Avoid scanning all pvs in the system if pvcreating on a device with mdas.
Other areas of the code check whether there are mdas on an orphan pv before doing the expensive scan of the system. This patch adds this check to pvcreate as well, and so will avoid the unnecessary scan if pvcreate on a device that is an orphan PV.
--- lib/metadata/metadata.c | 7 +++++-- 1 files changed, 5 insertions(+), 2 deletions(-)
lib/metadata/metadata.c +5 -2
@@ -1254,11 +1254,14 @@ static int pvcreate_check(struct cmd_context *cmd, const char *name,struct device *dev;uint64_t md_superblock, swap_signature;int wipe_md, wipe_swap; +struct dm_list mdas; + +dm_list_init (&mdas);/* FIXME Check partition type is LVM unless --force is given */ /* Is there a pv here already? */ - pv =pv_read (cmd, name,NULL ,NULL ,0 ,0 ); + pv =pv_read (cmd, name, &mdas,NULL ,0 ,0 );/* * If a PV has no MDAs it may appear to be an orphan until the
@@ -1266,7 +1269,7 @@ static int pvcreate_check(struct cmd_context *cmd, const char *name,* this means checking every VG by scanning every PV on the * system. */ -if (pv &&is_orphan (pv)) { +if (pv &&is_orphan (pv) && !dm_list_size (&mdas)) {if (!scan_vgs_for_pvs (cmd)) return_0; pv =pv_read (cmd, name,NULL ,NULL ,0 ,0 );
Signature · 1 line
1.6.0.6
PATCH 1/1 Avoid scanning all pvs in the system if pvcreating on a device with mdas.
Hi,
Quoted 4 lines · Dave Wysochanski
Other areas of the code check whether there are mdas on an orphan pv before doing the expensive scan of the system. This patch adds this check to pvcreate as well, and so will avoid the unnecessary scan if pvcreate on a device that is an orphan PV.
Looks OK to me. As far as I can tell, it should never happen that the metadata is out-of-date and we overwrite a non-orphan PV mistakenly.
This would require that the metadata on the PV claim this is an orphan but a newer copy of the metadata elsewhere claims this is part of a VG. That would mean that vgextend (or similar) failed to update the metadata on the new PV, which would presumably lead to overall vgextend failure and no new metadata on the pre-existing PVs either. So this should be safe.
Quoted 21 lines
lib/metadata/metadata.c +4
-1@@ -1254,11 +1254,14 @@ static int pvcreate_check(struct cmd_context *cmd, const char *name,struct device *dev;uint64_t md_superblock, swap_signature;int wipe_md, wipe_swap; +struct dm_list mdas; + +dm_list_init (&mdas);/* FIXME Check partition type is LVM unless --force is given */ /* Is there a pv here already? */ - pv =pv_read (cmd, name,NULL ,NULL ,0 ,0 ); + pv =pv_read (cmd, name, &mdas,NULL ,0 ,0 );/* * If a PV has no MDAs it may appear to be an orphan until the
lib/metadata/metadata.c static int pvcreate_check(struct cmd_context *cmd, const char *name, · line 1275
Quoted 8 lines
lib/metadata/metadata.c +1
-1@@ -1266,7 +1269,7 @@ static int pvcreate_check(struct cmd_context *cmd, const char *name,* this means checking every VG by scanning every PV on the * system. */ -if (pv &&is_orphan (pv)) { +if (pv &&is_orphan (pv) && !dm_list_size (&mdas)) {if (!scan_vgs_for_pvs (cmd)) return_0; pv =pv_read (cmd, name,NULL ,NULL ,0 ,0 );
ACK.
Yours,
Petr.
PATCH 1/1 Avoid scanning all pvs in the system if pvcreating on a device with mdas.
Quoted 10 lines · Petr Rockai
Hi,
Quoted 4 lines · Dave Wysochanski
Other areas of the code check whether there are mdas on an orphan pv before doing the expensive scan of the system. This patch adds this check to pvcreate as well, and so will avoid the unnecessary scan if pvcreate on a device that is an orphan PV.
Looks OK to me. As far as I can tell, it should never happen that the metadata is out-of-date and we overwrite a non-orphan PV mistakenly.
Quoted 7 lines
This would require that the metadata on the PV claim this is an orphan but a newer copy of the metadata elsewhere claims this is part of a VG. That would mean that vgextend (or similar) failed to update the metadata on the new PV, which would presumably lead to overall vgextend failure and no new metadata on the pre-existing PVs either. So this should be safe.
Ok - turns out I should have searched more carefully as there are more instances where we could save the scanning.
I'll work on an updated patch.
PATCH 1/1 Avoid scanning all pvs in the system if operating on a device with mdas.
When we pv_read() a device that has an orphan vgname, we might need to scan the system to be sure this is true. However, if the PV has mdas, there's no way possible for it to have an orphan vgname unless it is a true orphan. Some areas of the code were optimized to take advantage of this fact, while others were not (we would still do the expensive scan if a device had mdas but had an orphan VG).
This patch unifies the code so that every place we are operating on such a PV, we skip the expensive scan if there are mdas.
--- lib/metadata/metadata.c | 13 +++++++++---- tools/toollib.c | 8 ++++++-- 2 files changed, 15 insertions(+), 6 deletions(-)
lib/metadata/metadata.c +9 -4
@@ -1251,11 +1251,14 @@ static int pvcreate_check(struct cmd_context *cmd, const char *name,struct device *dev;uint64_t md_superblock, swap_signature;int wipe_md, wipe_swap; +struct dm_list mdas; + +dm_list_init (&mdas);/* FIXME Check partition type is LVM unless --force is given */ /* Is there a pv here already? */ - pv =pv_read (cmd, name,NULL ,NULL ,0 ,0 ); + pv =pv_read (cmd, name, &mdas,NULL ,0 ,0 );/* * If a PV has no MDAs it may appear to be an orphan until the
@@ -1263,7 +1266,7 @@ static int pvcreate_check(struct cmd_context *cmd, const char *name,* this means checking every VG by scanning every PV on the * system. */ -if (pv &&is_orphan (pv)) { +if (pv &&is_orphan (pv) && !dm_list_size (&mdas)) {if (!scan_vgs_for_pvs (cmd)) return_0; pv =pv_read (cmd, name,NULL ,NULL ,0 ,0 );
@@ -1747,14 +1750,16 @@ struct physical_volume *find_pv_by_name(struct cmd_context *cmd,static struct physical_volume *_find_pv_by_name (struct cmd_context *cmd,const char *pv_name) { +struct dm_list mdas;struct physical_volume *pv; -if (!(pv =_pv_read (cmd, cmd->mem, pv_name,NULL ,NULL ,1 ,0 ))) { +dm_list_init (&mdas); +if (!(pv =_pv_read (cmd, cmd->mem, pv_name, &mdas,NULL ,1 ,0 ))) {log_error ("Physical volume %s not found" , pv_name);return NULL ; } -if (is_orphan_vg (pv->vg_name)) { +if (is_orphan_vg (pv->vg_name) && !dm_list_size (&mdas)) {/* If a PV has no MDAs - need to search all VGs for it */ if (!scan_vgs_for_pvs (cmd)) return_NULL;
tools/toollib.c +6 -2
@@ -642,6 +642,7 @@ int process_each_pv(struct cmd_context *cmd, int argc, char **argv,struct str_list *sll;char *tagname;int scanned =0 ; +struct dm_list mdas;dm_list_init (&tags);
@@ -682,7 +683,9 @@ int process_each_pv(struct cmd_context *cmd, int argc, char **argv, } pv = pvl->pv; }else { -if (!(pv =pv_read (cmd, argv[opt],NULL , + +dm_list_init (&mdas); +if (!(pv =pv_read (cmd, argv[opt], &mdas,NULL ,1 , scan_label_only))) {log_error ("Failed to read physical " "volume \"%s\"" , argv[opt]);
@@ -697,7 +700,8 @@ int process_each_pv(struct cmd_context *cmd, int argc, char **argv,* means checking every VG by scanning every * PV on the system. */ -if (!scanned &&is_orphan (pv)) { +if (!scanned &&is_orphan (pv) && + !dm_list_size (&mdas)) {if (!scan_label_only && !scan_vgs_for_pvs (cmd)) { stack;
Signature · 1 line
1.6.0.6
PATCH 1/1 Avoid scanning all pvs in the system if operating on a device with mdas.
Quoted 8 lines · Dave Wysochanski
When we
pv_read()a device that has an orphan vgname, we might need to scan the system to be sure this is true. However, if the PV has mdas, there's no way possible for it to have an orphan vgname unless it is a true orphan. Some areas of the code were optimized to take advantage of this fact, while others were not (we would still do the expensive scan if a device had mdas but had an orphan VG).
This patch unifies the code so that every place we are operating on such a PV, we skip the expensive scan if there are mdas.
Makes sense.
Ack.
Alasdair
PATCH 1/1 Avoid scanning all pvs in the system if pvcreating on a device with mdas.
In reply to Petr RockaiQuoted 8 lines · Peter Rockai
Looks OK to me. As far as I can tell, it should never happen that the metadata is out-of-date and we overwrite a non-orphan PV mistakenly. This would require that the metadata on the PV claim this is an orphan but a newer copy of the metadata elsewhere claims this is part of a VG. That would mean that vgextend (or similar) failed to update the metadata on the new PV, which would presumably lead to overall vgextend failure and no new metadata on the pre-existing PVs either. So this should be safe.
There is actually a failure mode when this can happen.
The VG metadata is considered committed when the *first* vg_commit occurs. If the process dies at that point, the next vg_read of the VG does automatic recovery which completes writing the correct VG metadata to the remaining devices including the newly-added one.
So the optimisation in this patch assumes that if a process doing a vgextend died at exactly the wrong point, there would be a vg_read before this pv* code. With a reboot or any sort of display command a sysadmin would want to run after failure to check the state of the system, this would happen.
So I'm not particularly concerned about that failure mode, and the logic may change anyway when mdas are allowed to be empty.
Alasdair