Page MenuHomeFreeBSD

Reserve Arm MTE SIGSEGV codes
AbandonedPublic

Authored by emaste on Dec 27 2020, 6:30 PM.
Tags
None
Referenced Files
Unknown Object (File)
Mon, Dec 9, 4:48 AM
Unknown Object (File)
Oct 21 2024, 5:16 PM
Unknown Object (File)
Oct 9 2024, 11:14 PM
Unknown Object (File)
Oct 7 2024, 8:11 PM
Unknown Object (File)
Sep 8 2024, 1:49 AM
Unknown Object (File)
Sep 7 2024, 2:05 PM
Unknown Object (File)
Sep 7 2024, 10:49 AM
Unknown Object (File)
Sep 4 2024, 10:57 AM
Subscribers
None

Details

Reviewers
kib
mhorne
markj
Summary

LLDB is adding support for sync MTE exception and async MTE reporting. Claim the same values as used on Linux for future compatibility.

Diff Detail

Lint
Lint Skipped
Unit
Tests Skipped

Event Timeline

emaste created this revision.
sys/sys/signal.h
323–325

This line is part of the SEGV_ACCERR description.

I am curious what is the value of having FreeBSD si_code numerically equal to the Linux' one ?

When I added PKU si_code, I effectively started MD values at 100.

I am curious what is the value of having FreeBSD si_code numerically equal to the Linux' one ?

Application developers might have a tendency to do things like

#ifndef SEGV_MTEAERR
#define SEGV_MTEAERR 8
#endif

(from https://reviews.llvm.org/D93495)

Even though this is not really valid it seems sensible to me to just use the Linux values.

I do not think this is a valid practice, and we should not encourage it.

IMO it buys us nothing, while making some havoc in the SEGV_ value space. I would prefer to keep new values around SEGV_PKUERR, as any other new MD si_codes.