Bug #1371
closedlibc_r not functional with new malloc
0%
Description
I think the new malloc broke libc_r. Running firefox, I get this:
% firefox3
Fatal error 'Cannot allocate red zone for initial thread' at line ? in
file /usr/src/lib/libc_r/uthread/uthread_init.c (errno = ?)
zsh: abort firefox3
cheers
simon
Updated by corecode about 18 years ago
Matthew Dillon wrote:
BTW we don't support libc_r any more. We haven't for a while.
It is libthread_xu or nothing.
why do we build and install it still then? I'm trying to figure out the
cause of MASSIVE slowdown with thunderbird and firefox, and I thought an
alternative thread lib could help tracking it down.
cheers
simon
Updated by corecode about 18 years ago
fixed in 5f2211a11e7c56b07e5d8ce5c278b86d06210131
Updated by dillon about 18 years ago
:I think the new malloc broke libc_r. Running firefox, I get this:
:
:% firefox3
:Fatal error 'Cannot allocate red zone for initial thread' at line ? in=20
:file /usr/src/lib/libc_r/uthread/uthread_init.c (errno =3D ?)
:zsh: abort firefox3
:
:cheers
: simon
No, this is a bug in libc_r not using the proper MAP_* flags when
it mmap()'s the thread stack or the thread stack's guard page. The
kernel previously did not catch the problem but now it does, and I
fixed libthread_xu to use the proper MAP_ flags.
Instead of properly generating ENOMEM or some other related failure
when the user VM map fills up in older kernels would allow MAP_STACK
mappings to improperly overlap MAP_STACK mapped areas, creating a
situation where the mapped stack can dive into the non-MAP_STACK
mapping without generating a VM fault or a guard fault, resulting in
massive corruption instead of a clean memory allocation error.
BTW we don't support libc_r any more. We haven't for a while.
It is libthread_xu or nothing.
-Matt