UEK-NEXT support #127
Merged
UEK-NEXT support #127
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
oracle-contributor-agreement
bot
added
the
OCA Verified
brenns10
force-pushed
the
fix-ueknext-tests
branch
3 times, most recently
from
7169419
to
648a271
Compare
Kernel 6.4 adopted a new module memory layout structure which is more flexible than before. Rather than complicate the "ModuleLayout" structure which contained lots of details from the previous implementation, I've gone ahead and removed as much detail as possible. Now the module helpers simply return a list of module address ranges, which may be text, data, rodata, or anything really. In addition, to more efficiently look up module addresses, add support for the kernel module address tree. This avoids iterating over the (possibly long) module list each time you want to find out which module an address belongs to. These helpers ended up being really nice, and I upstreamed them to drgn. Each helper function which was upstreamed has a TODO entry so that we can find and remove them when we raise our minimum required version to drgn 0.0.28. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
This also somewhat simplifies the legacy code thanks to a new helper for iterating over all tasks in a task group. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
I'm not sure the root cause, but some workers now have NULL task fields. I'm confident this has always been legal, but just didn't happen in tests prior to v6.9 due to some kernel internals. Now that it happens, let's fix this case by searching for a worker with a non-NULL task. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Also, "NR_UNSTABLE_NFS" was never being set non-zero because we were looking it up in mm_stats, not node_zone_stats. Fix that too. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
In 6.3 and 6.4 there was a push to make struct bus_type and struct class const. This means that the private pointers were removed, and replaced by accessor functions. Implement a drgn version of each accessor. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Previous commits have resolved all of the compatibility issues as of UEK-next 6.9.0-2. Enable the tests so we can run them in CI. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
The minimum required drgn version is 0.0.25 for drgn-tools. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
These errors are not fully resolved in any drgn version, and they're not fatal either. Leave the documentation links, but take away the TODO since there's nothing here to fix. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
An upstreamed version is available within the standard d_path() function, starting from 0.0.29. Once this is the minimum required version of drgn, we can delete this function. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
In v6.10 there was a cleanup of struct block_device, removing the bd_inode field and combining bd_read_only, bd_partno, and others into a single __bd_flags atomic field. Fix the helpers for these changes. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
Computing size and read-only is duplicated here. Instead, use the helpers from drgn_tools.block, which are already fixed for v6.10 and later. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
The information for partition info can be easily gleaned from /sys/class/block on live systems. Add unit tests to verify the information is correct, so that our test can detect block-related changes like the ones corrected recently. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
This matches the output of /proc/mounts, which is important because the test cases compare the output from this helper against /proc/mounts. With this, tests pass on my laptop which happens to have some FUSE mounts that populate s_subtype. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
The max_active field got moved into the workqueue_struct. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
This should have been done a while ago. The "smp" corelens module already disallows running on live systems. IPIs fly too fast for us to keep up in userspace, so running the full module tends to cause issues. In this case, the specific issue had to do with unwinding the stack on the running task, which raises an error in drgn. Orabug: 37296325 Signed-off-by: Stephen Brennan <[email protected]>
brenns10
force-pushed
the
fix-ueknext-tests
branch
from
648a271
to
dd29e99
Compare
gtmoth
approved these changes
Nov 19, 2024
gtmoth
approved these changes
Nov 19, 2024
gtmoth
approved these changes
Nov 19, 2024
imran-kn
reviewed
Nov 19, 2024
imran-kn
reviewed
Nov 19, 2024
jianfenw
approved these changes
Nov 20, 2024
biger410
approved these changes
Nov 20, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This pull request begins running the Github CI tests against UEK-next, which is currently at 6.11. It includes a whole crop of fixes to helpers that allow the tests to pass. This is the first version of UEK-next that I'm actually doing this for, so it includes a lot of changes from UEK7 era up till the present.
These fixes are are necessary, but almost certainly incomplete. The Github CI tests are very small QEMU VMs, and I did manually verify that the tests pass against my laptop as well. There are plenty of subsystems not exercised by my laptop or the limited VM in the CI tests.
I'm breaking these changes down into a few categories for which I believe subject matter experts should review. If you're tagged here, please take a look at the subset of changes mentioned (note the Git SHAs may not be correct, as I may need to amend things to add bug references).