integration tests: CONFIG_MODVERSIONS and linux-5.18.0/symvers-disagreement-FAIL.patch about kpatch HOT 8 CLOSED

I got hit by missing CONFIG_MODVERSIONS this week. Shall we add a check in kpatch-build? Or better, shall we select MODVERSIONS with CONFIG_LIVEPATCH? (Maybe I should ask the second question in the mail list?)

from kpatch.

I got hit by missing CONFIG_MODVERSIONS this week. Shall we add a check in kpatch-build? Or better, shall we select MODVERSIONS with CONFIG_LIVEPATCH? (Maybe I should ask the second question in the mail list?)

Hi @liu-song-6 ,

• When you say you "got hit by missing CONFIG_MODVERSIONS", is that in reference to kpatch integration tests, or did you have to debug a kpatch that tweaked exported symbol abi?
• The kpatch integration tests assume the same result for all arches and configs. Perhaps they could be made more aware of their environment.
• MODVERSIONS is a nice safety feature when building kpatches, but maybe there is other tooling that generates livepatches that can make do without it? And/or some distros that would want to livepatch even without MODVERSIONS?

from kpatch.

When you say you "got hit by missing CONFIG_MODVERSIONS", is that in reference to kpatch integration tests, or did you have to debug a kpatch that tweaked exported symbol abi?

I was trying to make kpatch-build work for a kernel without CONFIG_MODVERSIONS. kpatch-build works, but the patch cannot load. The fix was to reenable CONFIG_MODVERSIONS.

I think we can have kpatch-build check whether CONFIG_MODVERSIONS is enabled. If it is not, kpatch-build fails. Is this reasonable?

from kpatch.

I was trying to make kpatch-build work for a kernel without CONFIG_MODVERSIONS. kpatch-build works, but the patch cannot load. The fix was to reenable CONFIG_MODVERSIONS.ing to make kpatch-build work for a kernel without CONFIG_MODVERSIONS. kpatch-build works, but the patch cannot load. The fix was to reenable CONFIG_MODVERSIONS.

Interesting, so setting CONFIG_MODVERSIONS alone fixed the kpatch.ko load, or did it help point out inadvertent symvers change?

I think we can have kpatch-build check whether CONFIG_MODVERSIONS is enabled. If it is not, kpatch-build fails. Is this reasonable?

That would affect distros like Fedora, which probably aren't the best use case for kpatch, but certainly handy for development and testing. How about a warning?

from kpatch.

Interesting, so setting CONFIG_MODVERSIONS alone fixed the kpatch.ko load, or did it help point out inadvertent symvers change?

In this case, setting CONFIG_MODVERSIONS fixes the issue.

That would affect distros like Fedora, which probably aren't the best use case for kpatch, but certainly handy for development and testing. How about a warning?

A warning should also work.

Thanks,
Song

from kpatch.

What error message did you see when the patch failed to load? I'm trying to understand why enabling CONFIG_MODVERSIONS fixed it.

from kpatch.

This issue has been open for 30 days with no activity and no assignee. It will be closed in 7 days unless a comment is added.

from kpatch.

This issue was closed because it was inactive for 7 days after being marked stale.

from kpatch.