diff options
| author | Ryan Moeller <[email protected]> | 2020-03-03 13:28:09 -0500 |
|---|---|---|
| committer | GitHub <[email protected]> | 2020-03-03 10:28:09 -0800 |
| commit | 0a0f9a7dc6e4f215089cf0aabb230d5dd5dfbd1e (patch) | |
| tree | d994b7072bd1ccff7e2d96566b347edc87904185 /tests/zfs-tests | |
| parent | 9bb907bc3fdcaa132ecb0f8326c8815f27824ce4 (diff) | |
ZTS: Provide for nested cleanup routines
Shared test library functions lack a simple way to ensure proper
cleanup in the event of a failure. The `log_onexit` cleanup pattern
cannot be used in library functions because it uses one global
variable to store the cleanup command.
An example of where this is a serious issue is when a tunable that
artifically stalls kernel progress gets activated and then some check
fails. Unless the caller knows about the tunable and sets it back,
the system will be left in a bad state.
To solve this problem, turn the global cleanup variable into a stack.
Provide push and pop functions to add additional cleanup steps and
remove them after it is safe again.
The first use of this new functionality is in attempt_during_removal,
which sets REMOVAL_SUSPEND_PROGRESS.
Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: John Kennedy <[email protected]>
Signed-off-by: Ryan Moeller <[email protected]>
Closes #10080
Diffstat (limited to 'tests/zfs-tests')
| -rw-r--r-- | tests/zfs-tests/tests/functional/removal/removal.kshlib | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/tests/zfs-tests/tests/functional/removal/removal.kshlib b/tests/zfs-tests/tests/functional/removal/removal.kshlib index e1f43fbe7..231991e82 100644 --- a/tests/zfs-tests/tests/functional/removal/removal.kshlib +++ b/tests/zfs-tests/tests/functional/removal/removal.kshlib @@ -60,6 +60,7 @@ function attempt_during_removal # pool disk callback [args] typeset callback=$3 shift 3 + log_onexit_push set_tunable32 REMOVAL_SUSPEND_PROGRESS 0 set_tunable32 REMOVAL_SUSPEND_PROGRESS 1 log_must zpool remove $pool $disk @@ -80,6 +81,7 @@ function attempt_during_removal # pool disk callback [args] log_must is_pool_removing $pool set_tunable32 REMOVAL_SUSPEND_PROGRESS 0 + log_onexit_pop log_must wait_for_removal $pool log_mustnot vdevs_in_pool $pool $disk |