c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-23 02:44:41 +03:00
Code Issues Packages Projects Releases Wiki Activity

Refcounted DSL Crypto Key Mappings

Browse Source 
Since native ZFS encryption was merged, we have been fighting
against a series of bugs that come down to the same problem: Key
mappings (which must be present during all I/O operations) are
created and destroyed based on dataset ownership, but I/Os can
have traditionally been allowed to "leak" into the next txg after
the dataset is disowned.

In the past we have attempted to solve this problem by trying to
ensure that datasets are disowned ater all I/O is finished by
calling txg_wait_synced(), but we have repeatedly found edge cases
that need to be squashed and code paths that might incur a high
number of txg syncs. This patch attempts to resolve this issue
differently, by adding a reference to the key mapping for each txg
it is dirtied in. By doing so, we can remove many of the
unnecessary calls to txg_wait_synced() we have added in the past
and ensure we don't need to deal with this problem in the future.

Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #7949
This commit is contained in:
Tom Caputi
2018-10-03 12:47:11 -04:00
committed by Brian Behlendorf
parent f65fbee1e7
commit 52ce99dd61
9 changed files with 195 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zfs/zil.c
+8 -18
View File
@@ -3236,8 +3236,8 @@ zil_suspend(const char *osname, void **cookiep)
* grabbing a reference to it. If the key isn't loaded we have no
* choice but to return an error until the wrapping key is loaded.
*/
if (os->os_encrypted && spa_keystore_create_mapping(os->os_spa,
dmu_objset_ds(os), FTAG) != 0) {
if (os->os_encrypted &&
dsl_dataset_create_key_mapping(dmu_objset_ds(os)) != 0) {
zilog->zl_suspend--;
mutex_exit(&zilog->zl_lock);
dsl_dataset_long_rele(dmu_objset_ds(os), suspend_tag);
@@ -3259,9 +3259,10 @@ zil_suspend(const char *osname, void **cookiep)
zil_commit_impl(zilog, 0);
/*
* Now that we've ensured all lwb's are LWB_STATE_DONE, we use
* txg_wait_synced() to ensure the data from the zilog has
* migrated to the main pool before calling zil_destroy().
* Now that we've ensured all lwb's are LWB_STATE_DONE,
* txg_wait_synced() will be called from within zil_destroy(),
* which will ensure the data from the zilog has migrated to the
* main pool before it returns.
*/
txg_wait_synced(zilog->zl_dmu_pool, 0);
@@ -3272,19 +3273,8 @@ zil_suspend(const char *osname, void **cookiep)
cv_broadcast(&zilog->zl_cv_suspend);
mutex_exit(&zilog->zl_lock);
if (os->os_encrypted) {
/*
* Encrypted datasets need to wait for all data to be
* synced out before removing the mapping.
*
* XXX: Depending on the number of datasets with
* outstanding ZIL data on a given log device, this
* might cause spa_offline_log() to take a long time.
*/
txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
VERIFY0(spa_keystore_remove_mapping(os->os_spa,
dmu_objset_id(os), FTAG));
}
if (os->os_encrypted)
dsl_dataset_remove_key_mapping(dmu_objset_ds(os));
if (cookiep == NULL)
zil_resume(os);