Use of WARN family of macros resulting in changed section false positives about kpatch HOT 4 CLOSED

Amazing analyse, so the root problem is the code generation of the macro __FILE__ and __LINE__.

from kpatch.

For now I think it's ok if we just document this behavior. It is kind of annoying, but technically, it's working as expected. If the WARN call's line number changes, then it makes sense that the value of __LINE__ will be different and the function will be replaced to use the new value of __LINE__ in the WARN call.

The worst case is that we end up patching some functions that haven't really changed. If that turns out to be a problem for the user, we might be able to fix this in a different way. For example, they could modify the patch to redefine the WARN macro for the affected files, such that it hard codes the old line number instead of using __LINE__, so that the affected function is recompiled to be identical.

from kpatch.

@jpoimboe sounds like a reasonable (and the only) solution for now. Detecting line number changes when they are encoded in mov instructions... not sure how we could detect that and therefore determine equivalence at the object level.

from kpatch.

Since we have no plans to fix this, I'll just document it in the README as expected behavior.

A future fix might involve changing the WARN macro upstream so that it works kind of like the BUG macro, with its own warn_table.

from kpatch.