Kernel: Fix signal delivery when no syscall is made

This fixes a regression introduced by the new software context
switching where the Kernel would not deliver a signal unless the
process is making system calls. This is because the TSS no longer
updates the CS value, so the scheduler never considered delivery
as the process always appeared to be in kernel mode. With software
context switching we can just set up the signal trampoline at
any time and when the processor returns back to user mode it'll
get executed. This should fix e.g. killing programs that are
stuck in some tight loop that doesn't make any system calls and
is only pre-empted by the timer interrupt.

Fixes #2958
This commit is contained in:
Tom 2020-08-02 12:08:22 -06:00 committed by Andreas Kling
parent 6e54d0c072
commit f011c420c1
Notes: sideshowbarker 2024-07-19 04:22:26 +09:00
4 changed files with 10 additions and 39 deletions

View file

@ -773,7 +773,7 @@ void Process::terminate_due_to_signal(u8 signal)
{
ASSERT_INTERRUPTS_DISABLED();
ASSERT(signal < 32);
dbg() << "Terminating due to signal " << signal;
dbg() << "Terminating " << *this << " due to signal " << signal;
m_termination_status = 0;
m_termination_signal = signal;
die();

View file

@ -405,17 +405,6 @@ bool Scheduler::pick_next()
Thread::for_each_living([&](Thread& thread) -> IterationDecision {
if (!thread.has_unmasked_pending_signals())
return IterationDecision::Continue;
// FIXME: It would be nice if the Scheduler didn't have to worry about who is "current"
// For now, avoid dispatching signals to "current" and do it in a scheduling pass
// while some other process is interrupted. Otherwise a mess will be made.
if (&thread == current_thread)
return IterationDecision::Continue;
// We know how to interrupt blocked processes, but if they are just executing
// at some random point in the kernel, let them continue.
// Before returning to userspace from a syscall, we will block a thread if it has any
// pending unmasked signals, allowing it to be dispatched then.
if (thread.in_kernel() && !thread.is_blocked() && !thread.is_stopped())
return IterationDecision::Continue;
// NOTE: dispatch_one_pending_signal() may unblock the process.
bool was_blocked = thread.is_blocked();
if (thread.dispatch_one_pending_signal() == ShouldUnblockThread::No)

View file

@ -144,12 +144,11 @@ void Thread::set_should_die()
m_should_die = true;
if (is_blocked()) {
ASSERT(in_kernel());
ASSERT(m_blocker != nullptr);
// We're blocked in the kernel.
m_blocker->set_interrupted_by_death();
unblock();
} else if (!in_kernel()) {
} else {
// We're executing in userspace (and we're clearly
// not the current thread). No need to unwind, so
// set the state to dying right away. This also
@ -446,7 +445,7 @@ ShouldUnblockThread Thread::dispatch_signal(u8 signal)
ASSERT(!process().is_ring0());
#ifdef SIGNAL_DEBUG
klog() << "dispatch_signal <- " << signal;
klog() << "signal: dispatch signal " << signal << " to " << *this;
#endif
auto& action = m_signal_action_data[signal];
@ -518,7 +517,7 @@ ShouldUnblockThread Thread::dispatch_signal(u8 signal)
if (handler_vaddr.as_ptr() == SIG_IGN) {
#ifdef SIGNAL_DEBUG
klog() << "ignored signal " << signal;
klog() << "signal: " << *this << " ignored signal " << signal;
#endif
return ShouldUnblockThread::Yes;
}
@ -572,31 +571,16 @@ ShouldUnblockThread Thread::dispatch_signal(u8 signal)
};
// We now place the thread state on the userspace stack.
// Note that when we are in the kernel (ie. blocking) we cannot use the
// tss, as that will contain kernel state; instead, we use a RegisterState.
// Note that we use a RegisterState.
// Conversely, when the thread isn't blocking the RegisterState may not be
// valid (fork, exec etc) but the tss will, so we use that instead.
if (!in_kernel()) {
u32* stack = &m_tss.esp;
setup_stack(m_tss, stack);
m_tss.cs = GDT_SELECTOR_CODE3 | 3;
m_tss.ds = GDT_SELECTOR_DATA3 | 3;
m_tss.es = GDT_SELECTOR_DATA3 | 3;
m_tss.fs = GDT_SELECTOR_DATA3 | 3;
m_tss.gs = GDT_SELECTOR_TLS | 3;
m_tss.eip = g_return_to_ring3_from_signal_trampoline.get();
// FIXME: This state is such a hack. It avoids trouble if 'current' is the process receiving a signal.
set_state(Skip1SchedulerPass);
} else {
auto& regs = get_register_dump_from_stack();
u32* stack = &regs.userspace_esp;
setup_stack(regs, stack);
regs.eip = g_return_to_ring3_from_signal_trampoline.get();
}
auto& regs = get_register_dump_from_stack();
u32* stack = &regs.userspace_esp;
setup_stack(regs, stack);
regs.eip = g_return_to_ring3_from_signal_trampoline.get();
#ifdef SIGNAL_DEBUG
klog() << "signal: Okay, {" << state_string() << "} has been primed with signal handler " << String::format("%w", m_tss.cs) << ":" << String::format("%x", m_tss.eip);
klog() << "signal: Okay, " << *this << " {" << state_string() << "} has been primed with signal handler " << String::format("%w", m_tss.cs) << ":" << String::format("%x", m_tss.eip) << " to deliver " << signal;
#endif
return ShouldUnblockThread::Yes;
}

View file

@ -274,8 +274,6 @@ public:
bool has_blocker() const { return m_blocker != nullptr; }
const Blocker& blocker() const;
bool in_kernel() const { return (m_tss.cs & 0x03) == 0; }
u32 cpu() const { return m_cpu.load(AK::MemoryOrder::memory_order_consume); }
void set_cpu(u32 cpu) { m_cpu.store(cpu, AK::MemoryOrder::memory_order_release); }
u32 affinity() const { return m_cpu_affinity; }