ptrace_stop: fix the race with ptrace detach+attach
PATCH 1/2 ptrace_stop: fix the race with ptrace detach+attach
If the tracer went away (may_ptrace_stop() failed), ptrace_stop() drops tasklist and then changes the ->state from TASK_TRACED to TASK_RUNNING.
This can fool another tracer which attaches to us in between. Change the ->state under tasklist_lock to ensure that ptrace_check_attach() can't wrongly succeed. Also, remove the unnecessary mb().
kernel/signal.c renamed from kernel/signal.c~1_ptrace_stop +3 -3
@@ -1628,11 +1628,11 @@ static void ptrace_stop(int exit_code, i }else {/* * By the time we got the lock, our tracer went away. -* Don't stop here. +* Don't drop the lock yet, another tracer may come. */ -read_unlock (&tasklist_lock); -set_current_state (TASK_RUNNING ); +__set_current_state (TASK_RUNNING ); current->exit_code = nostop_code; +read_unlock (&tasklist_lock); }/*
Quoted 5 lines · Oleg Nesterov
If the tracer went away (
may_ptrace_stop()failed),ptrace_stop()drops tasklist and then changes the ->state fromTASK_TRACEDtoTASK_RUNNING.This can fool another tracer which attaches to us in between. Change the ->state under
tasklist_lockto ensure thatptrace_check_attach()can't wrongly succeed.
ptrace_check_attach? Both do read_lock -- can run in parallel, so how can it help?
Quoted 17 lines
kernel/signal.c renamed from kernel/signal.c~1_ptrace_stop +3
-3@@ -1628,11 +1628,11 @@ static void ptrace_stop(int exit_code, i }else {/* * By the time we got the lock, our tracer went away. -* Don't stop here. +* Don't drop the lock yet, another tracer may come. */ -read_unlock (&tasklist_lock); -set_current_state (TASK_RUNNING ); +__set_current_state (TASK_RUNNING ); current->exit_code = nostop_code; +read_unlock (&tasklist_lock); }/*
Quoted 9 lines · Alexey Dobriyan
Quoted 5 lines · Oleg Nesterov
If the tracer went away (
may_ptrace_stop()failed),ptrace_stop()drops tasklist and then changes the ->state fromTASK_TRACEDtoTASK_RUNNING.This can fool another tracer which attaches to us in between. Change the ->state under
tasklist_lockto ensure thatptrace_check_attach()can't wrongly succeed.
ptrace_check_attach? Both doread_lock-- can run in parallel,
Yep.
Quoted 1 line
so how can it help?
read_lock prevents ptrace_attach(), so the new tracer can't attach and then do ptrace_check_attach().
Oleg.
Your change looks correct to me.
Thanks,
Roland