Fix sigaltstack corruption among cloned threads
Subsystems (3)
PATCH 1/1 Fix sigaltstack corruption among cloned threads
This patch fixes alternate signal stack corruption among cloned threads with CLONE_SIGHAND (and CLONE_VM) for linux-2.6.16-rc6.
The value of alternate signal stack is currently inherited after a call of clone(... CLONE_SIGHAND | CLONE_VM). But if sigaltstack is set by a parent thread, and then if multiple cloned child threads (+ parent threads) call signal handler at the same time, some threads may be conflicted - because they share to use the same alternative signal stack region. Finally they get sigsegv. It's an undesirable race condition. Note that child threads created from NPTL pthread_create() also hit this conflict when the parent thread uses sigaltstack, without my patch.
To fix this problem, this patch clears the child threads' sigaltstack information like exec(). This behavior follows the SUSv3 specification. In SUSv3, pthread_create() says "The alternate stack shall not be inherited (when new threads are initialized)". It means that sigaltstack should be cleared when sigaltstack memory space is shared by cloned threads with CLONE_SIGHAND.
Note that I chose "if (clone_flags & CLONE_SIGHAND)" line because:
- If clone_flags line is not existed, fork() does not inherit sigaltstack.
- CLONE_VM is another choice, but vfork() does not inherit sigaltstack.
- CLONE_SIGHAND implies CLONE_VM, and it looks suitable.
- CLONE_THREAD is another candidate, and includes CLONE_SIGHAND + CLONE_VM, but this flag has a bit different semantics.
I decided to use CLONE_SIGHAND.
-- gotom
kernel/fork.c +7 -0
@@ -1062,6 +1062,13 @@ static task_t *copy_process(unsigned lon p->clear_child_tid = (clone_flags &CLONE_CHILD_CLEARTID ) ? child_tidptr:NULL ;/* +* sigaltstack should be cleared when CLONE_SIGHAND (and CLONE_VM) is +* specified. +*/ +if (clone_flags &CLONE_SIGHAND ) + p->sas_ss_sp = p->sas_ss_size =0 ; + +/* * Syscall tracing should be turned off in the child regardless * of CLONE_PTRACE. */
kernel/fork.c static task_t *copy_process(unsigned lon · line 1069
Quoted 5 lines · GOTO Masanori
kernel/fork.c +5
-0+* sigaltstack should be cleared when CLONE_SIGHAND (and CLONE_VM) is +* specified. +*/ +if (clone_flags &CLONE_SIGHAND ) + p->sas_ss_sp = p->sas_ss_size =0 ;
I agree in general, but why base it on CLONE_SIGHAND? The problem results from using the same address space. So it should be
if (clone_flags & CLONE_VM)
The fact that both these flags are used at the same time in all cases today shouldn't hide the real reason for this requirement which is sharing the address space.
Quoted 7 lines · Ulrich Drepper
kernel/fork.c static task_t *copy_process(unsigned lon · line 1069
Quoted 5 lines · GOTO Masanori
kernel/fork.c +5
-0+* sigaltstack should be cleared when CLONE_SIGHAND (and CLONE_VM) is +* specified. +*/ +if (clone_flags &CLONE_SIGHAND ) + p->sas_ss_sp = p->sas_ss_size =0 ;
Quoted 8 lines
I agree in general, but why base it on
CLONE_SIGHAND? The problem results from using the same address space. So it should beif (
clone_flags&CLONE_VM)The fact that both these flags are used at the same time in all cases today shouldn't hide the real reason for this requirement which is sharing the address space.
Because vfork also sets CLONE_VM and vfork isn't supposed to reset alternate stack setting. For vfork that's not a problem, as the parent task will not continue until the vfork child execve's. So, if you want to use CLONE_VM bit, you'd need to use
if ((clone_flags & (CLONE_VM | CLONE_VFORK)) == CLONE_VM) p->sas_ss_sp = p->sas_ss_size = 0; Jakub
Quoted 2 lines · Jakub Jelinek
Because vfork also sets
CLONE_VMand vfork isn't supposed to reset alternate stack setting.
I'd say this is rather an oversight than a real requirement (we fixed quite a few such problems in the POSIX spec of vfork) but it should be as safe as any other solution. So, I'm OK with the extended CLONE_VM test.