Skip to content
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

Prevent .eh_frame from being emitted for -C panic=abort #112403

Merged
merged 1 commit into from
Jun 16, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion compiler/rustc_mir_transform/src/check_alignment.rs
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ use rustc_middle::mir::{
};
use rustc_middle::ty::{Ty, TyCtxt, TypeAndMut};
use rustc_session::Session;
use rustc_target::spec::PanicStrategy;

pub struct CheckAlignment;

Expand Down Expand Up @@ -236,7 +237,11 @@ fn insert_alignment_check<'tcx>(
required: Operand::Copy(alignment),
found: Operand::Copy(addr),
}),
unwind: UnwindAction::Terminate,
unwind: if tcx.sess.panic_strategy() == PanicStrategy::Unwind {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This feels like something we're going to regress on with the next similar check in MIR -- can we introduce another run of AbortUnwindingCalls or otherwise make this more general?

(I'm not super familiar with mir opts so not sure if that pass is very expensive).

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Currently this pass is the only pass that can add a new call, and given that this pass is only enabled with debug assertions, re-running AbortUnwindingCalls feels a bit unnecessary?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we run AbortUnwindingCalls after it (or rather, run the alignment checks before it)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

alternatively, MIR validation could check that no normal unwind actions are present on panic=abort

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

UnwindAction::Terminate may be present in -Cpanic=abort when C-unwind is used so we couldn't just do it in validation.

Lifting the CheckAlignment pass might make sense.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

terminate existing with panic=unwind could happen, but other unwind actions with panic=abort shouldn't, right? that could be validated

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's entirely orthogonal to this PR though.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I just checked we couldn't lift CheckAlignment, because alignment checks are not doable for CTFE, so it must stay as an optimization pass. We also couldn't delay AbortUnwindingCalls, because it needs to be run before generator lowering.

Copy link
Member

@RalfJung RalfJung Jun 26, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I understood @saethlin correctly, this could be Unreachable always?

Either way there should be a comment explaining why this is unreachable. "Because we use the non-unwinding panic machinery" seems like a good answer to me; it is explicitly intended to not unwind so I have no issue with this non-local dependency.

EDIT: Ah, #112599 already did this. :)

UnwindAction::Terminate
} else {
UnwindAction::Unreachable
},
},
});
}
10 changes: 10 additions & 0 deletions tests/run-make/panic-abort-eh_frame/Makefile
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
# only-linux
#
# This test ensures that `panic=abort` code (without `C-unwind`, that is) should not have any
# unwinding related `.eh_frame` sections emitted.

include ../tools.mk

all:
$(RUSTC) foo.rs --crate-type=lib --emit=obj=$(TMPDIR)/foo.o -Cpanic=abort
objdump --dwarf=frames $(TMPDIR)/foo.o | $(CGREP) -v 'DW_CFA'
10 changes: 10 additions & 0 deletions tests/run-make/panic-abort-eh_frame/foo.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
#![no_std]

#[panic_handler]
fn handler(_: &core::panic::PanicInfo<'_>) -> ! {
loop {}
}

pub unsafe fn oops(x: *const u32) -> u32 {
*x
}