CRASH SIGFPE with glibc 2.39

[MikeWalrus]

Describe the bug
glibc changes field offsets again in 2.39, which causes SIGFPE in __libc_early_init.
See #5437 #5695.

To Reproduce
Run drrun with any client and program on a linux system with glibc 2.39 (e.g. ArchLinux).

Expected behavior
No crashes with SIGFPE.

Screenshots or Pasted Text
See the previous issue #5437.

Versions

• What version of DynamoRIO are you using? 9591e25 and 10.0.0
• Does the latest build from https://github.com/DynamoRIO/dynamorio/releases solve the problem? No
• What operating system version are you running on? ArchLinux (glibc 2.39)
• Is your application 32-bit or 64-bit? 64-bit

[derekbruening]

Xref glibc causing pain in other ways: #5860. One possibility is to stop supporting glibc in clients and DR's own "normal" binaries: require clients to use musl or somesuch, and switch to musl in DR's binaries, since the glibc owners are making it quite difficult to keep using it in our project not just for the private loader but also for "normal" uses such as in the drrun launcher where we want backward compatibility without having to build on an old toolchain.

[MikeWalrus]

In my opinion, DynamoRIO's capability of being able to work with system glibc reasonably well is a great advantage over pin. I think we should just keep an eye on new versions of glibc and fix broken fields, unless this becomes too unmanageable of course. :)

[derekbruening]

In my opinion, DynamoRIO's capability of being able to work with system glibc reasonably well is a great advantage over pin. I think we should just keep an eye on new versions of glibc and fix broken fields, unless this becomes too unmanageable of course. :)

This is useful feedback. Could you elaborate: are there features missing from musl you want to use in your clients, or you want to avoid having to build/link your clients in non-standard ways?

[MikeWalrus]

are there features missing from musl you want to use in your clients, or you want to avoid having to build/link your clients in non-standard ways?

I'd say it's more of the latter, although I can imagine someone would prefer glibc because of its performance or some third-party libraries require glibc.

[jdw1023]

Hi, I am running into the SIGFPE crash on Arch Linux using glibc 2.39-4.
After some investigation, it looks like the offset changed in the glibc 2.39(-4?) on Arch Linux. I am not sure why the offset is different (the offset in d21d6e3 looks like the right offset with using glibc 2.39-0ubuntu8.1_amd64 on ubuntu).

Left Arch Linux glibc 2.39-4, Right Ubuntu glibc 2.39-0ubuntu8.1_amd64

I got drrun to work after patching offset to the following. Just want to leave this comment incase anyone ran into the same issue.

+0x02d8 _dl_tls_static_size  : size_t
+0x02e0 _dl_tls_static_align : size_t

glibc version info:

% /usr/lib/libc.so.6
GNU C Library (GNU libc) stable release version 2.39.
Copyright (C) 2024 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.
Compiled by GNU CC version 13.2.1 20240417.
libc ABIs: UNIQUE IFUNC ABSOLUTE
Minimum supported kernel: 4.4.0
For bug reporting instructions, please see:
<https://gitlab.archlinux.org/archlinux/packaging/packages/glibc/-/issues>.
% pacman -Q glibc
glibc 2.39-4