-
Couldn't load subscription status.
- Fork 21
DLPX-87572 sdb: want live kernel tests to find kernel regressions early #337
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: develop
Are you sure you want to change the base?
Conversation
37c2944 to
3a92736
Compare
1f1cb88 to
abf539e
Compare
2ac14fc to
52feab9
Compare
= Problem With our switch to the new v5.15 kernel a subset of SDB commands broke without us realizing until we actually needed them. Our regression dumps helps us to ensure we don't introduce regressions for older kernels when developing new features but they can't help us in detecting changes in the upstream kernel or ZFS that break our commands. = This Patch This patch attempts to provide a rudimentary mechanism for catching regression introduced by the upstream Ubuntu kernels by running a few basic SDB commands in a Github action that's run nightly and for every PR. Specifically this patch makes it so we have such a test for each Ubuntu LTS kernel starting from 20.04 (currently the `ubuntu-latest` Github runner tag points to 22.04 so we'd test that twice but in the future that tag will point to 24.04, etc...). We also change for all the available Python versions for each Ubuntu version to further ensure SDB's compatibility with future Python versions. = Misc Notes In order to use SDB in the Github runner I had to introduce an extra script that downloads the kernel's debug info. See the `install-live-kernel-dbg.sh` script for more info. I also made sure to decouple the apt-install of the python-dev files to its own shell script too as different Ubuntu versions ship with different Python versions. See `install-python-dev.sh` for more info. = Potential Future Items In the future we may want to detect whenever our ZFS commands are not getting out of date. `test_live_kernel.sh` has a way of detecting whether the ZFS module is installed and running a few ZFS commands on the live kernel. The idea is that we can either introduce Github Actions like the upstream openzfs that install our kernel module to the runner and run the commands there OR we can create a BlackBox test that clones the repo and runs this script.
52feab9 to
fbc215c
Compare
Codecov Report
❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more. @@ Coverage Diff @@
## develop #337 +/- ##
===========================================
+ Coverage 85.18% 85.21% +0.03%
===========================================
Files 67 67
Lines 3077 3070 -7
===========================================
- Hits 2621 2616 -5
+ Misses 456 454 -2
|
bae4146 to
fbc215c
Compare
= Problem Currently having our own custom function for figuring out a task's state has two drawbacks: 1] As we saw in a2bdd57 things can get out of date and it is up to us to fix them. 2] There are some weird edge cases that we don't handle as well like the following crash that I have never been able to reproduce locally but it occasionally reproduces in the Github runners of PR delphix#337: ``` Traceback (most recent call last): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/internal/repl.py", line 107, in eval_cmd for obj in invoke([], input_): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 152, in invoke yield from execute_pipeline(first_input, pipeline) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 84, in execute_pipeline yield from massage_input_and_call(pipeline[-1], this_input) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 67, in massage_input_and_call yield from cmd.call(objs) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 413, in call yield from self.__invalid_memory_objects_check( File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 358, in __invalid_memory_objects_check for obj in objs: File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 625, in _call self.pretty_print(self.caller(objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 407, in pretty_print self.print_stacks(filter(self.match_stack, objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 382, in print_stacks for stack_key, tasks in KernelStacks.aggregate_stacks(objs): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 375, in aggregate_stacks stack_key = (KernelStacks.task_struct_get_state(task), File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 221, in task_struct_get_state return KernelStacks.TASK_STATES[(state | exit_state) & 0x7f] KeyError: 101 ``` = This Patch Uses the drgn helper whose implementation handles more edge cases and is more probable to stay up to date with the latest kernels while keeping backwards compatibility. = Testing The above stack trace that I was able to reproduce consistently in that PR no longer shows up with this patch.
= Problem Currently having our own custom function for figuring out a task's state has two drawbacks: 1] As we saw in a2bdd57 things can get out of date and it is up to us to fix them. 2] There are some weird edge cases that we don't handle as well like the following crash that I have never been able to reproduce locally but it occasionally reproduces in the Github runners of PR delphix#337: ``` Traceback (most recent call last): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/internal/repl.py", line 107, in eval_cmd for obj in invoke([], input_): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 152, in invoke yield from execute_pipeline(first_input, pipeline) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 84, in execute_pipeline yield from massage_input_and_call(pipeline[-1], this_input) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 67, in massage_input_and_call yield from cmd.call(objs) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 413, in call yield from self.__invalid_memory_objects_check( File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 358, in __invalid_memory_objects_check for obj in objs: File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 625, in _call self.pretty_print(self.caller(objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 407, in pretty_print self.print_stacks(filter(self.match_stack, objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 382, in print_stacks for stack_key, tasks in KernelStacks.aggregate_stacks(objs): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 375, in aggregate_stacks stack_key = (KernelStacks.task_struct_get_state(task), File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 221, in task_struct_get_state return KernelStacks.TASK_STATES[(state | exit_state) & 0x7f] KeyError: 101 ``` = This Patch Uses the drgn helper whose implementation handles more edge cases and is more probable to stay up to date with the latest kernels while keeping backwards compatibility. = Testing The above stack trace that I was able to reproduce consistently in that PR no longer shows up with this patch.
= Problem Currently having our own custom function for figuring out a task's state has two drawbacks: 1] As we saw in a2bdd57 things can get out of date and it is up to us to fix them. 2] There are some weird edge cases that we don't handle as well like the following crash that I have never been able to reproduce locally but it occasionally reproduces in the Github runners of PR delphix#337: ``` Traceback (most recent call last): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/internal/repl.py", line 107, in eval_cmd for obj in invoke([], input_): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 152, in invoke yield from execute_pipeline(first_input, pipeline) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 84, in execute_pipeline yield from massage_input_and_call(pipeline[-1], this_input) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 67, in massage_input_and_call yield from cmd.call(objs) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 413, in call yield from self.__invalid_memory_objects_check( File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 358, in __invalid_memory_objects_check for obj in objs: File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 625, in _call self.pretty_print(self.caller(objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 407, in pretty_print self.print_stacks(filter(self.match_stack, objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 382, in print_stacks for stack_key, tasks in KernelStacks.aggregate_stacks(objs): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 375, in aggregate_stacks stack_key = (KernelStacks.task_struct_get_state(task), File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 221, in task_struct_get_state return KernelStacks.TASK_STATES[(state | exit_state) & 0x7f] KeyError: 101 ``` = This Patch Uses the drgn helper whose implementation handles more edge cases and is more probable to stay up to date with the latest kernels while keeping backwards compatibility. = Testing The above stack trace that I was able to reproduce consistently in that PR no longer shows up with this patch.
= Problem Currently having our own custom function for figuring out a task's state has two drawbacks: 1] As we saw in a2bdd57 things can get out of date and it is up to us to fix them. 2] There are some weird edge cases that we don't handle as well like the following crash that I have never been able to reproduce locally but it occasionally reproduces in the Github runners of PR #337: ``` Traceback (most recent call last): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/internal/repl.py", line 107, in eval_cmd for obj in invoke([], input_): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 152, in invoke yield from execute_pipeline(first_input, pipeline) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 84, in execute_pipeline yield from massage_input_and_call(pipeline[-1], this_input) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/pipeline.py", line 67, in massage_input_and_call yield from cmd.call(objs) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 413, in call yield from self.__invalid_memory_objects_check( File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 358, in __invalid_memory_objects_check for obj in objs: File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/command.py", line 625, in _call self.pretty_print(self.caller(objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 407, in pretty_print self.print_stacks(filter(self.match_stack, objs)) File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 382, in print_stacks for stack_key, tasks in KernelStacks.aggregate_stacks(objs): File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 375, in aggregate_stacks stack_key = (KernelStacks.task_struct_get_state(task), File "/usr/local/lib/python3.8/dist-packages/sdb-0.1.0-py3.8.egg/sdb/commands/linux/stacks.py", line 221, in task_struct_get_state return KernelStacks.TASK_STATES[(state | exit_state) & 0x7f] KeyError: 101 ``` = This Patch Uses the drgn helper whose implementation handles more edge cases and is more probable to stay up to date with the latest kernels while keeping backwards compatibility. = Testing The above stack trace that I was able to reproduce consistently in that PR no longer shows up with this patch.
= Problem
With our switch to the new v5.15 kernel a subset of SDB commands broke without us realizing until we actually needed them. Our regression dumps helps us to ensure we don't introduce regressions for older kernels when developing new features but they can't help us in detecting changes in the upstream kernel or ZFS that break our commands.
= This Patch
This patch attempts to provide a rudimentary mechanism for catching regression introduced by the upstream Ubuntu kernels by running a few basic SDB commands in a Github action that's run nightly and for every PR.
Specifically this patch makes it so we have such a test for each Ubuntu LTS kernel starting from 20.04 (currently the
ubuntu-latestGithub runner tag points to 22.04 so we'd test that twice but in the future that tag will point to 24.04, etc...).We also change for all the available Python versions for each Ubuntu version to further ensure SDB's compatibility with future Python versions.
= Misc Notes
In order to use SDB in the Github runner I had to introduce an extra script that downloads the kernel's debug info. See the
install-live-kernel-dbg.shscript for more info.I also made sure to decouple the apt-install of the python-dev files to its own shell script too as different Ubuntu versions ship with different Python versions. See
install-python-dev.shfor more info.= Potential Future Items
In the future we may want to detect whenever our ZFS commands are not getting out of date.
test_live_kernel.shhas a way of detecting whether the ZFS module is installed and running a few ZFS commands on the live kernel. The idea is that we can either introduce Github Actions like the upstream openzfs that install our kernel module to the runner and run the commands there OR we can create a BlackBox test that clones the repo and runs this script.