sanitizers: Avoid building genassym.c and genoffset.c with sanitizers
ClosedPublic
Actions

Authored by markj on Oct 11 2023, 1:46 PM.

Details

Reviewers

emaste
andrew
imp

Commits

rG8e1a7e29b6cc: sanitizers: Avoid building genassym.c and genoffset.c with sanitizers

Summary

Some, particularly KASAN, may insert redzones around global symbols,
resulting in incorrect offset definitions because genassym.sh defines
some semantics using symbol sizes.

Ideally I would be able to define this pattern in one place, but I
haven't found a way to define a GENSYM_CFLAGS that actually works for
all of the consumers (kern.post.mk, kmod.mk, sys/conf/files*).

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 53925
Build 50815: arc lint + arc unit

Event Timeline

markj requested review of this revision.Oct 11 2023, 1:46 PM

markj created this revision.

Harbormaster completed remote builds in B53924: Diff 128566.Oct 11 2023, 1:46 PM

I wonder why amd64 is the only one with a <32>_genassym.o

Funny, I stumbled on the genassym/genoffset mechanisms yesterday. FWIW, LGTM.

In D42157#961811, @emaste wrote:

I wonder why amd64 is the only one with a <32>_genassym.o

You mean, why doesn't arm64 have one as well? It looks like ia32_genassym defines symbols used by the 32-bit signal trampoline. On arm64, aarch32_sigcode is stuck at the end of locore and uses hex constants instead of actual instructions, so it doesn't need to define SIGF_UC etc.

It would be nice if there was a common place we could define this pattern, but I'm not aware of a clean way to do it. I had the same issue in 0e8b3ab34883efedeeb5e0d3baa90ea98de1e692. Note there are a couple of modules in that change that have their own genassym invocations.

This revision is now accepted and ready to land.Oct 11 2023, 2:06 PM

Fix genassym invocations in sys/modules

This revision now requires review to proceed.Oct 11 2023, 2:10 PM

Herald added a subscriber: rgrimes. · View Herald TranscriptOct 11 2023, 2:10 PM

Harbormaster completed remote builds in B53925: Diff 128567.Oct 11 2023, 2:10 PM

emaste accepted this revision.Oct 11 2023, 2:12 PM

This revision is now accepted and ready to land.Oct 11 2023, 2:12 PM

I guess this is a good indication that we should find a common way to do this. But I think it's fine to commit this to fix the immediate issue and see if @imp or @sjg has a suggestion.