ZFS fix to make xattr=sa logging to ZIL on create/remove/update. #9078
Conversation
Codecov Report
@@ Coverage Diff @@
## master #9078 +/- ##
==========================================
+ Coverage 75.17% 78.88% +3.70%
==========================================
Files 402 400 -2
Lines 128071 121758 -6313
==========================================
- Hits 96283 96050 -233
+ Misses 31788 25708 -6080
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
ahrens
added
the
Status: Design Review Needed
|
Can you clarify whether we are talking about when synchronous semantics are requested vs not? How do we request those semantics? I think that we only log these operations do disk if sync=always or when we later do another (unrelated) sync operation, is that right? It looks like with the default of xattr=on, these would be logged via zpl_xattr_set_dir()'s calls to zfs_create(), zpl_write_common(), etc. These would do zfs_log_create() but not zil_commit(). Are those semantics correct? I think that with xattr=sa, we are not creating the itx at all (i.e. not calling zfs_log_create()). So there is no way for these operations to be logged. That does indeed seem like a bug. |
Good question. Unlike illumos and FreeBSD, the
Adding support to log these wasn't included in the original implementation, and we never managed to revisit this and add the missing support. Which is definitely a bug. A new ZIL log record type does seem like the correct way to handle this, but we'll need to register a new feature flag for this. Otherwise, replaying a log with these entries on another OpenZFS platform will result in a crash. |
Thanks Matthew and Brain for the review comments. yes, exact semantics (sync or async) required for the operation are not very clear. For xattr=on, it does log the xattr write/update operation via zpl_write_common() path as Metthew mentioned and with sync=always, log would be committed to disk. But for xattr=sa, it doesn't log the operation. So, for xattr=sa, as well, similarly, we can log the record and commit (zil_commit) to disk as part of operation if sync=always. Thanks Brian, I would check the feature flag part. I am new to zfs and still learning it. |
Do you think that we should also change the xattr=on behavior to be always sync? In either case, I think that like all other ZFS operations (AFAIK), xattr operations (whether xattr=on or xattr=sa) should be async by default (i.e. not persisted to disk immediately via zil_commit()). |
jsai20
force-pushed
the
issue-8768
branch
4 times, most recently
from
a7f8e9a
to
12fac85
Compare
For xattr=on, xattr's are synced to zil via zpl_xattr_set_dir()->zpl_write_common()->zpl_write_common_iovec()->zfs_write() { In my diff, which enables xattr=sa logging to zil, for now I have kept xattr=sa behavior same as xattr=on, means do zil_commit() only if dataset property "sync" is set to "always". So, xattr=sa attributes doesn't lost on node crash similar to xattr=on, but it's guaranteed only when sync=always is set. I think, decision about "xatt=on" and "xattr=sa" should always do zil_commit() needs more involved discussion and not much related to diff here. Appropriate diff/pull request for the same can be opened separately, as discussed and decided. I have updated diff with feature flag change. Please have a look and provide feedback on same. |
Why should xattr=on always sync to disk immediately?
I agree that it isn't related to this PR, so we can proceed with |
I meant, I am not very sure, either it should be always sync or not. I can't find any specific reference to it.
Thanks. |
|
When talking new feature flags - there's an ultimate trade-off between these two options with the current design:
It is explained a bit better in the comment in the OverlayFS PR #9414 I don't understand the need for those functions to be called in the syncing context fully, but I suppose it is to be sure that the incr/decr actually lands on disk by the end of the txg. So, might I suggest adding new txtypes for incr/decr of ZFS pool features? That way, we could incr/decr the features no matter the context; the only trade off here is that on import, to check for the features used by the pool, we'd have to walk the IL. I'm not sure whether this can be done in a flat way for the whole pool, or whether we have to walk the dataset structure and then walk the associated IL records... @behlendorf @ahrens perhaps you would know more and could advise on the matter? I wouldn't mind doing the work, but I'd be glad for some guidance :) We would probably want to use a new feature flag for feature incr/decr's in the IL, but that is something all platforms could potentially benefit from right away (unlike the two txtypes needed for renameat2, for example). |
|
@snajpa What you are suggesting would most likely be implemented something like this: When we do a read-write import of the pool, when we walk all the ZIL's to claim their blocks, we also check if there are any unrecognized record types. If so, we fail the import. One downside with this is that the record types are just an enum, so all implementations of ZFS must agree about what each value means. Alternatively, we could add a single feature flag now which is always activated (refcount=1), and add a single new ZIL record type. This new record type would specify which other feature flag must be supported in order to import the pool read-write. The other feature flag would be specified by fully-qualified name, so it wouldn't have the downside mentioned above. But it doesn't solve the backwards-compatibility for the first thing that needs it. Personally, I think we should err on the side of simplicity. We already have the feature flag infrastructure. We shouldn't shy away from using it to introduce new feature flags and accept the fact that older/different implementations won't be able to access the pool (at least not writeable). But I could be convinced otherwise depending on the specifics. IMO, the problem this PR is solving is not that acute, so folks desiring compatibility can leave the feature disabled with little impact. |
|
It looks like maybe we decided on adding a new on-disk feature flag for this, but the code review process stalled out. @jsai20 if you were to rebase this, would it be ready for final code review and integration, as far as you know? |
Thanks Matt. ya, code review was stalled after around conversation on new on-disk feature flag. |
jsai20
force-pushed
the
issue-8768
branch
2 times, most recently
from
2e3d681
to
00688c9
Compare
|
This will need to be applied on the FreeBSD side as well, and the test updated to use helpers from libtest for working with tunables and xattrs. |
Done. |
jsai20
force-pushed
the
issue-8768
branch
3 times, most recently
from
2d79d97
to
fbc2d93
Compare
jsai20
force-pushed
the
issue-8768
branch
4 times, most recently
from
71b0bd0
to
c5085ab
Compare
|
@jsai20 thanks for your patience, if can you rebase this one last time it should be good to go. I'd like to try and get this merged by the end of the week. @ahrens @nabijaczleweli @problame @freqlabs last call, if you have any remaining concerns please post a comment. |
behlendorf
added
Status: Accepted
cmd/zdb/zdb_il.c
Outdated
| char *val; | ||
| int i; | ||
|
|
||
| name = (char *)(lr + 1); |
There was a problem hiding this comment.
| name = (char *)(lr + 1); | |
| char *name = (char *)(lr + 1); |
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
|
@jsai20 thanks for iterating with us on this, merged! |
Thanks @behlendorf @ahrens @freqlabs @problame @nabijaczleweli @scineram and othre reviewers. Thanks much for refining it! |
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. So, xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change adds xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This makes xattr=sa behavior similar to xattr=on.
Implementation notes:
The actual logging is fairly straight-forward and does not warrant
additional explanation.
However, it has been 14 years since we last added new TX types
to the ZIL [1], hence this is the first time we do it after the
introduction of zpool features. Therefore, here is an overview of the
feature activation and deactivation workflow:
1. The feature must be enabled. Otherwise, we don't log the new
record type. This ensures compatibility with older software.
2. The feature is activated per-dataset, since the ZIL is per-dataset.
3. If the feature is enabled and dataset is not for zvol, any append to
the ZIL chain will activate the feature for the dataset. Likewise
for starting a new ZIL chain.
4. A dataset that doesn't have a ZIL chain has the feature deactivated.
We ensure (3) by activating on the first zil_commit() after the feature
was enabled. Since activating the features requires waiting for txg
sync, the first zil_commit() after enabling the feature will be slower
than usual. The downside is that this is really a conservative
approximation: even if we never append a 'TX_SETSAXATTR' to the ZIL
chain, we pay the penalty for feature activation. The upside is that the
user is in control of when we pay the penalty, i.e., upon enabling the
feature.
We ensure (4) by hooking into zil_sync(), where ZIL destroy actually
happens.
One more piece on feature activation, since it's spread across
multiple functions:
zil_commit()
zil_process_commit_list()
if lwb == NULL // first zil_commit since zil_open
zil_create()
if no log block pointer in ZIL header:
if feature enabled and not active:
// CASE 1
enable, COALESCE txg wait with dmu_tx that allocated the
log block
else // log block was allocated earlier than this zil_open
if feature enabled and not active:
// CASE 2
enable, EXPLICIT txg wait
else // already have an in-DRAM LWB
if feature enabled and not active:
// this happens when we enable the feature after zil_create
// CASE 3
enable, EXPLICIT txg wait
[1] illumos/illumos-gate@da6c28a
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Christian Schwarz <christian.schwarz@nutanix.com>
Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com>
Closes openzfs#8768
Closes openzfs#9078
As such, there are no specific synchronous semantics defined for
the xattrs. But for xattr=on, it does log to ZIL and zil_commit() is
done, if sync=always is set on dataset. This provides sync semantics
for xattr=on with sync=always set on dataset.
For the xattr=sa implementation, it doesn't log to ZIL, so, even with
sync=always, xattrs are not guaranteed to be synced before xattr call
returns to caller. so xattr can be lost if system crash happens, before
txg carrying xattr transaction is synced.
This change makes xattr=sa logging to ZIL on xattr create/remove/update
and xattrs are synced to ZIL (zil_commit() done) for sync=always.
This make xattr=sa behavior similar to xattr=on.
This could also provide basic framework to support implementing sync
semantics at file level for xattr=sa.
Signed-off-by: Jitendra Patidar jitendra.patidar@nutanix.com
Closes #8768