* [PATCH 0/3] dm thin metadata: handful of improvements
@ 2019-04-15 21:17 Mike Snitzer
2019-04-15 21:17 ` [PATCH 1/3] dm space map common: zero entire ll_disk Mike Snitzer
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Mike Snitzer @ 2019-04-15 21:17 UTC (permalink / raw)
To: dm-devel; +Cc: ejt
All are working well in my testing but I'd appreciate further review
from others.
Thanks,
Mike
Mike Snitzer (3):
dm space map common: zero entire ll_disk
dm thin metadata: do not write metadata if no changes occurred
dm thin metadata: check __commit_transaction()'s return
drivers/md/dm-thin-metadata.c | 63 +++++++++++++------
.../md/persistent-data/dm-space-map-common.c | 2 +
2 files changed, 47 insertions(+), 18 deletions(-)
--
2.18.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 1/3] dm space map common: zero entire ll_disk
2019-04-15 21:17 [PATCH 0/3] dm thin metadata: handful of improvements Mike Snitzer
@ 2019-04-15 21:17 ` Mike Snitzer
2019-04-15 21:17 ` [PATCH 2/3] dm thin metadata: do not write metadata if no changes occurred Mike Snitzer
2019-04-15 21:17 ` [PATCH 3/3] dm thin metadata: check __commit_transaction()'s return Mike Snitzer
2 siblings, 0 replies; 6+ messages in thread
From: Mike Snitzer @ 2019-04-15 21:17 UTC (permalink / raw)
To: dm-devel; +Cc: ejt
Otherwise, memory that is allocated (and potentially not previously
zeroed) will get written to disk as part of the space maps.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
---
drivers/md/persistent-data/dm-space-map-common.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/md/persistent-data/dm-space-map-common.c b/drivers/md/persistent-data/dm-space-map-common.c
index 0a3b8ae4a29c..b8a62188f6be 100644
--- a/drivers/md/persistent-data/dm-space-map-common.c
+++ b/drivers/md/persistent-data/dm-space-map-common.c
@@ -190,6 +190,8 @@ static int sm_find_free(void *addr, unsigned begin, unsigned end,
static int sm_ll_init(struct ll_disk *ll, struct dm_transaction_manager *tm)
{
+ memset(ll, 0, sizeof(struct ll_disk));
+
ll->tm = tm;
ll->bitmap_info.tm = tm;
--
2.18.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 2/3] dm thin metadata: do not write metadata if no changes occurred
2019-04-15 21:17 [PATCH 0/3] dm thin metadata: handful of improvements Mike Snitzer
2019-04-15 21:17 ` [PATCH 1/3] dm space map common: zero entire ll_disk Mike Snitzer
@ 2019-04-15 21:17 ` Mike Snitzer
2019-04-16 2:39 ` Mike Snitzer
2019-04-15 21:17 ` [PATCH 3/3] dm thin metadata: check __commit_transaction()'s return Mike Snitzer
2 siblings, 1 reply; 6+ messages in thread
From: Mike Snitzer @ 2019-04-15 21:17 UTC (permalink / raw)
To: dm-devel; +Cc: ejt
Otherwise, just loading a thin-pool and then removing it will cause its
metadata to be changed (e.g. superblock written) -- even without any
thin devices being made active or metadata snapshot being manipulated.
Add 'changed' flag to struct dm_pool_metadata. Set both td->changed and
pmd->changed using __set_thin_changed() -- this assures pmd->changed is
always set when td->changed is. For methods that don't involve changing
a td, set pmd->changed directly.
Rename __delete_device() to __delete_thin() -- improves symmetry with
__create_thin().
Update dm_pool_changed_this_transaction() to simply return pmd->changed.
Fix dm_pool_commit_metadata() to open the next transaction if the return
from __commit_transaction() is 0. Not seeing why the early return ever
made since for a return of 0 given that dm-io's async_io(), as used by
bufio, always returns 0.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
---
drivers/md/dm-thin-metadata.c | 56 ++++++++++++++++++++++++-----------
1 file changed, 39 insertions(+), 17 deletions(-)
diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c
index ed3caceaed07..f4e8327abc75 100644
--- a/drivers/md/dm-thin-metadata.c
+++ b/drivers/md/dm-thin-metadata.c
@@ -201,6 +201,12 @@ struct dm_pool_metadata {
*/
bool fail_io:1;
+ /*
+ * Set if metadata has changed, particularly important for methods that
+ * alter metadata without the ability to reflect the change in td->changed.
+ */
+ bool changed:1;
+
/*
* Reading the space map roots can fail, so we read it into these
* buffers before the superblock is locked and updated.
@@ -769,7 +775,7 @@ static int __write_changed_details(struct dm_pool_metadata *pmd)
return r;
if (td->open_count)
- td->changed = 0;
+ td->changed = false;
else {
list_del(&td->list);
kfree(td);
@@ -779,6 +785,14 @@ static int __write_changed_details(struct dm_pool_metadata *pmd)
return 0;
}
+static void __set_thin_changed(struct dm_thin_device *td)
+{
+ struct dm_pool_metadata *pmd = td->pmd;
+
+ td->changed = true;
+ pmd->changed = true;
+}
+
static int __commit_transaction(struct dm_pool_metadata *pmd)
{
int r;
@@ -790,6 +804,9 @@ static int __commit_transaction(struct dm_pool_metadata *pmd)
*/
BUILD_BUG_ON(sizeof(struct thin_disk_superblock) > 512);
+ if (!pmd->changed)
+ return 0;
+
r = __write_changed_details(pmd);
if (r < 0)
return r;
@@ -819,6 +836,8 @@ static int __commit_transaction(struct dm_pool_metadata *pmd)
copy_sm_roots(pmd, disk_super);
+ pmd->changed = false;
+
return dm_tm_commit(pmd->tm, sblock);
}
@@ -853,6 +872,7 @@ struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
pmd->time = 0;
INIT_LIST_HEAD(&pmd->thin_devices);
pmd->fail_io = false;
+ pmd->changed = false;
pmd->bdev = bdev;
pmd->data_block_size = data_block_size;
@@ -967,7 +987,8 @@ static int __open_device(struct dm_pool_metadata *pmd,
(*td)->pmd = pmd;
(*td)->id = dev;
(*td)->open_count = 1;
- (*td)->changed = changed;
+ if (changed)
+ __set_thin_changed(*td);
(*td)->aborted_with_changes = false;
(*td)->mapped_blocks = le64_to_cpu(details_le.mapped_blocks);
(*td)->transaction_id = le64_to_cpu(details_le.transaction_id);
@@ -1051,7 +1072,7 @@ static int __set_snapshot_details(struct dm_pool_metadata *pmd,
if (r)
return r;
- td->changed = 1;
+ __set_thin_changed(td);
td->snapshotted_time = time;
snap->mapped_blocks = td->mapped_blocks;
@@ -1131,7 +1152,7 @@ int dm_pool_create_snap(struct dm_pool_metadata *pmd,
return r;
}
-static int __delete_device(struct dm_pool_metadata *pmd, dm_thin_id dev)
+static int __delete_thin(struct dm_pool_metadata *pmd, dm_thin_id dev)
{
int r;
uint64_t key = dev;
@@ -1158,6 +1179,8 @@ static int __delete_device(struct dm_pool_metadata *pmd, dm_thin_id dev)
if (r)
return r;
+ pmd->changed = true;
+
return 0;
}
@@ -1168,7 +1191,7 @@ int dm_pool_delete_thin_device(struct dm_pool_metadata *pmd,
down_write(&pmd->root_lock);
if (!pmd->fail_io)
- r = __delete_device(pmd, dev);
+ r = __delete_thin(pmd, dev);
up_write(&pmd->root_lock);
return r;
@@ -1276,6 +1299,9 @@ static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
disk_super = dm_block_data(sblock);
disk_super->held_root = cpu_to_le64(held_root);
dm_bm_unlock(sblock);
+
+ pmd->changed = true;
+
return 0;
}
@@ -1324,6 +1350,8 @@ static int __release_metadata_snap(struct dm_pool_metadata *pmd)
dm_tm_unlock(pmd->tm, copy);
+ pmd->changed = true;
+
return 0;
}
@@ -1558,7 +1586,7 @@ static int __insert(struct dm_thin_device *td, dm_block_t block,
if (r)
return r;
- td->changed = 1;
+ __set_thin_changed(td);
if (inserted)
td->mapped_blocks++;
@@ -1589,7 +1617,7 @@ static int __remove(struct dm_thin_device *td, dm_block_t block)
return r;
td->mapped_blocks--;
- td->changed = 1;
+ __set_thin_changed(td);
return 0;
}
@@ -1643,7 +1671,7 @@ static int __remove_range(struct dm_thin_device *td, dm_block_t begin, dm_block_
}
td->mapped_blocks -= total_count;
- td->changed = 1;
+ __set_thin_changed(td);
/*
* Reinsert the mapping tree.
@@ -1735,16 +1763,10 @@ bool dm_thin_changed_this_transaction(struct dm_thin_device *td)
bool dm_pool_changed_this_transaction(struct dm_pool_metadata *pmd)
{
- bool r = false;
- struct dm_thin_device *td, *tmp;
+ bool r;
down_read(&pmd->root_lock);
- list_for_each_entry_safe(td, tmp, &pmd->thin_devices, list) {
- if (td->changed) {
- r = td->changed;
- break;
- }
- }
+ r = pmd->changed;
up_read(&pmd->root_lock);
return r;
@@ -1782,7 +1804,7 @@ int dm_pool_commit_metadata(struct dm_pool_metadata *pmd)
goto out;
r = __commit_transaction(pmd);
- if (r <= 0)
+ if (r < 0)
goto out;
/*
--
2.18.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 3/3] dm thin metadata: check __commit_transaction()'s return
2019-04-15 21:17 [PATCH 0/3] dm thin metadata: handful of improvements Mike Snitzer
2019-04-15 21:17 ` [PATCH 1/3] dm space map common: zero entire ll_disk Mike Snitzer
2019-04-15 21:17 ` [PATCH 2/3] dm thin metadata: do not write metadata if no changes occurred Mike Snitzer
@ 2019-04-15 21:17 ` Mike Snitzer
2 siblings, 0 replies; 6+ messages in thread
From: Mike Snitzer @ 2019-04-15 21:17 UTC (permalink / raw)
To: dm-devel; +Cc: ejt
Fix __reserve_metadata_snap() to return early if __commit_transaction()
fails.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
---
drivers/md/dm-thin-metadata.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c
index f4e8327abc75..3ba27f5f7da9 100644
--- a/drivers/md/dm-thin-metadata.c
+++ b/drivers/md/dm-thin-metadata.c
@@ -1248,7 +1248,12 @@ static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
* We commit to ensure the btree roots which we increment in a
* moment are up to date.
*/
- __commit_transaction(pmd);
+ r = __commit_transaction(pmd);
+ if (r < 0) {
+ DMWARN("%s: __commit_transaction() failed, error = %d",
+ __func__, r);
+ return r;
+ }
/*
* Copy the superblock.
--
2.18.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH 2/3] dm thin metadata: do not write metadata if no changes occurred
2019-04-15 21:17 ` [PATCH 2/3] dm thin metadata: do not write metadata if no changes occurred Mike Snitzer
@ 2019-04-16 2:39 ` Mike Snitzer
2019-04-17 1:30 ` [PATCH v2 " Mike Snitzer
0 siblings, 1 reply; 6+ messages in thread
From: Mike Snitzer @ 2019-04-16 2:39 UTC (permalink / raw)
To: dm-devel; +Cc: ejt
On Mon, Apr 15 2019 at 5:17pm -0400,
Mike Snitzer <snitzer@redhat.com> wrote:
> Otherwise, just loading a thin-pool and then removing it will cause its
> metadata to be changed (e.g. superblock written) -- even without any
> thin devices being made active or metadata snapshot being manipulated.
>
> Add 'changed' flag to struct dm_pool_metadata. Set both td->changed and
> pmd->changed using __set_thin_changed() -- this assures pmd->changed is
> always set when td->changed is. For methods that don't involve changing
> a td, set pmd->changed directly.
>
> Rename __delete_device() to __delete_thin() -- improves symmetry with
> __create_thin().
>
> Update dm_pool_changed_this_transaction() to simply return pmd->changed.
>
> Fix dm_pool_commit_metadata() to open the next transaction if the return
> from __commit_transaction() is 0. Not seeing why the early return ever
> made since for a return of 0 given that dm-io's async_io(), as used by
> bufio, always returns 0.
This v1 patch was incomplete. However this approach's need to establish
pmd->changed everywhere that alters metadata, without doing so in terms
of a thin device, is like whack-a-mole.
But I think this could be simplified to only set a 'pmd->accessed'
(never clear it) and only__commit_transaction() if pmd->accessed --
so only pool messages that change metadata would need to set
pmd->accessed.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v2 2/3] dm thin metadata: do not write metadata if no changes occurred
2019-04-16 2:39 ` Mike Snitzer
@ 2019-04-17 1:30 ` Mike Snitzer
0 siblings, 0 replies; 6+ messages in thread
From: Mike Snitzer @ 2019-04-17 1:30 UTC (permalink / raw)
To: dm-devel; +Cc: ejt
Otherwise, just loading a thin-pool and then removing it will cause its
metadata to be changed (e.g. superblock written) -- even without any
thin devices being made active or metadata snapshot being manipulated.
Add 'in_service' flag to struct dm_pool_metadata and use a new
dm_pool_set_in_service() interface to set it if either a thin device is
activated or a message is sent to the thin-pool. Once 'in_service' is
set it is never cleared. __commit_transaction() will return 0 if
'in_service' is not set.
Also fix dm_pool_commit_metadata() to open the next transaction if the
return from __commit_transaction() is 0. Not seeing why the early
return ever made since for a return of 0 given that dm-io's async_io(),
as used by bufio, always returns 0.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
---
drivers/md/dm-thin-metadata.c | 23 ++++++++++++++++++++++-
drivers/md/dm-thin-metadata.h | 2 ++
drivers/md/dm-thin.c | 9 +++++++++
3 files changed, 33 insertions(+), 1 deletion(-)
diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c
index 7381e477a945..fcff30fc07cd 100644
--- a/drivers/md/dm-thin-metadata.c
+++ b/drivers/md/dm-thin-metadata.c
@@ -201,6 +201,13 @@ struct dm_pool_metadata {
*/
bool fail_io:1;
+ /*
+ * Set once a thin-pool has been accessed through one of the interfaces
+ * that imply the pool is in-service (e.g. thin devices created/deleted,
+ * thin-pool message, metadata snapshots, etc).
+ */
+ bool in_service:1;
+
/*
* Reading the space map roots can fail, so we read it into these
* buffers before the superblock is locked and updated.
@@ -790,6 +797,9 @@ static int __commit_transaction(struct dm_pool_metadata *pmd)
*/
BUILD_BUG_ON(sizeof(struct thin_disk_superblock) > 512);
+ if (unlikely(!pmd->in_service))
+ return 0;
+
r = __write_changed_details(pmd);
if (r < 0)
return r;
@@ -853,6 +863,7 @@ struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
pmd->time = 0;
INIT_LIST_HEAD(&pmd->thin_devices);
pmd->fail_io = false;
+ pmd->in_service = false;
pmd->bdev = bdev;
pmd->data_block_size = data_block_size;
@@ -1727,6 +1738,16 @@ int dm_pool_dec_data_range(struct dm_pool_metadata *pmd, dm_block_t b, dm_block_
return r;
}
+void dm_pool_set_in_service(struct dm_pool_metadata *pmd)
+{
+ if (!pmd->in_service) {
+ down_write(&pmd->root_lock);
+ if (!pmd->in_service)
+ pmd->in_service = true;
+ up_write(&pmd->root_lock);
+ }
+}
+
bool dm_thin_changed_this_transaction(struct dm_thin_device *td)
{
int r;
@@ -1787,7 +1808,7 @@ int dm_pool_commit_metadata(struct dm_pool_metadata *pmd)
goto out;
r = __commit_transaction(pmd);
- if (r <= 0)
+ if (r < 0)
goto out;
/*
diff --git a/drivers/md/dm-thin-metadata.h b/drivers/md/dm-thin-metadata.h
index f6be0d733c20..8192530984be 100644
--- a/drivers/md/dm-thin-metadata.h
+++ b/drivers/md/dm-thin-metadata.h
@@ -47,6 +47,8 @@ struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
int dm_pool_metadata_close(struct dm_pool_metadata *pmd);
+void dm_pool_set_in_service(struct dm_pool_metadata *pmd);
+
/*
* Compat feature flags. Any incompat flags beyond the ones
* specified below will prevent use of the thin metadata.
diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
index fcd887703f95..b6dbbdfc9118 100644
--- a/drivers/md/dm-thin.c
+++ b/drivers/md/dm-thin.c
@@ -3822,6 +3822,13 @@ static int pool_message(struct dm_target *ti, unsigned argc, char **argv,
return -EOPNOTSUPP;
}
+ /*
+ * Any message sent to a thin-pool warrants putting it in-service
+ * (even if the message is malformed, it speaks to a user
+ * attempting to interface with the thin-pool)
+ */
+ dm_pool_set_in_service(pool->pmd);
+
if (!strcasecmp(argv[0], "create_thin"))
r = process_create_thin_mesg(argc, argv, pool);
@@ -4222,6 +4229,8 @@ static int thin_ctr(struct dm_target *ti, unsigned argc, char **argv)
goto bad_pool;
}
+ dm_pool_set_in_service(tc->pool->pmd);
+
r = dm_pool_open_thin_device(tc->pool->pmd, tc->dev_id, &tc->td);
if (r) {
ti->error = "Couldn't open thin internal device";
--
2.18.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
end of thread, other threads:[~2019-04-17 1:30 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-04-15 21:17 [PATCH 0/3] dm thin metadata: handful of improvements Mike Snitzer
2019-04-15 21:17 ` [PATCH 1/3] dm space map common: zero entire ll_disk Mike Snitzer
2019-04-15 21:17 ` [PATCH 2/3] dm thin metadata: do not write metadata if no changes occurred Mike Snitzer
2019-04-16 2:39 ` Mike Snitzer
2019-04-17 1:30 ` [PATCH v2 " Mike Snitzer
2019-04-15 21:17 ` [PATCH 3/3] dm thin metadata: check __commit_transaction()'s return Mike Snitzer
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.