I've been experiencing some issues with the Iris driver recently and I think it coincides with the glibc 2.30 release
https://bugs.freedesktop.org/show_bug.cgi?id=111376
I've just rebuild my system with gcc 9.2 and glibc 2.30 uisng these flags:
CFLAGS="-O3 -march=native -pipe -flto=8"
CXXFLAGS="${CFLAGS}"
LDFLAGS="${CFLAGS} -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,--build-id=sha1"
(Just -O3 -march=native -pipe for glibc)
And now I'm seeing segfaults in libacl when systemd-tmpfiles starts preventing bootup
0x00007ffff7cc79ce in __acl_entry_pp_compare () from /lib64/libacl.so.1
(gdb) bt
#0 0x00007ffff7cc79ce in __acl_entry_pp_compare () from /lib64/libacl.so.1
#1 0x00007ffff7b2fec4 in ?? () from /lib64/libc.so.6
#2 0x00007ffff7b2fe52 in ?? () from /lib64/libc.so.6
#3 0x00007ffff7b2fe35 in ?? () from /lib64/libc.so.6
#4 0x00007ffff7b30212 in qsort_r () from /lib64/libc.so.6
#5 0x00007ffff7cc86e4 in __acl_reorder_obj_p () from /lib64/libacl.so.1
#6 0x00007ffff7cc89a8 in __acl_from_xattr () from /lib64/libacl.so.1
#7 0x00007ffff7cc9ce0 in acl_get_file () from /lib64/libacl.so.1
#8 0x00007ffff7e4921d in acls_for_file () from /lib/systemd/libsystemd-shared-243.so
#9 0x000055555555c9a5 in ?? ()
#10 0x000055555555cd55 in ?? ()
#11 0x000055555555fe19 in ?? ()
#12 0x000055555555895c in ?? ()
#13 0x00007ffff7b17e8b in __libc_start_main () from /lib64/libc.so.6
#14 0x0000555555558a8a in ?? ()
(gdb) Quit
(gdb) Quit
(gdb) q
with an unstripped sys-apps/acl
If I compile it with -O3 -march=native -pipe the issues go away
With the Iris driver dropping LTO, or enabling the debug flag, or using clang makes the issue go away
Has anyone else seen these issues?
I've been experiencing some issues with the Iris driver recently and I think it coincides with the glibc 2.30 release
https://bugs.freedesktop.org/show_bug.cgi?id=111376
I've just rebuild my system with gcc 9.2 and glibc 2.30 uisng these flags:
CFLAGS="-O3 -march=native -pipe -flto=8"
CXXFLAGS="${CFLAGS}"
LDFLAGS="${CFLAGS} -Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,--build-id=sha1"
(Just -O3 -march=native -pipe for glibc)
And now I'm seeing segfaults in libacl when systemd-tmpfiles starts preventing bootup
0x00007ffff7cc79ce in __acl_entry_pp_compare () from /lib64/libacl.so.1
(gdb) bt
#0 0x00007ffff7cc79ce in __acl_entry_pp_compare () from /lib64/libacl.so.1
#1 0x00007ffff7b2fec4 in ?? () from /lib64/libc.so.6
#2 0x00007ffff7b2fe52 in ?? () from /lib64/libc.so.6
#3 0x00007ffff7b2fe35 in ?? () from /lib64/libc.so.6
#4 0x00007ffff7b30212 in qsort_r () from /lib64/libc.so.6
#5 0x00007ffff7cc86e4 in __acl_reorder_obj_p () from /lib64/libacl.so.1
#6 0x00007ffff7cc89a8 in __acl_from_xattr () from /lib64/libacl.so.1
#7 0x00007ffff7cc9ce0 in acl_get_file () from /lib64/libacl.so.1
#8 0x00007ffff7e4921d in acls_for_file () from /lib/systemd/libsystemd-shared-243.so
#9 0x000055555555c9a5 in ?? ()
#10 0x000055555555cd55 in ?? ()
#11 0x000055555555fe19 in ?? ()
#12 0x000055555555895c in ?? ()
#13 0x00007ffff7b17e8b in __libc_start_main () from /lib64/libc.so.6
#14 0x0000555555558a8a in ?? ()
(gdb) Quit
(gdb) Quit
(gdb) q
with an unstripped sys-apps/acl
If I compile it with -O3 -march=native -pipe the issues go away
With the Iris driver dropping LTO, or enabling the debug flag, or using clang makes the issue go away
Has anyone else seen these issues?