Kernel: Signals are not queued anymore

Posix doesn't require signal queing if you don't use sigqueue() which
we don't support. Process also has its own pending signal mask.

kernel/include/kernel/Process.h

@@ -160,6 +160,7 @@ namespace Kernel
 		BAN::UniqPtr<GeneralAllocator> m_general_allocator;
 
 		vaddr_t m_signal_handlers[_SIGMAX + 1] { };
+		uint64_t m_signal_pending_mask { 0 };
 
 		bool m_is_userspace { false };
 		userspace_info_t m_userspace_info;

kernel/include/kernel/Thread.h

@@ -1,6 +1,5 @@
 #pragma once
 
-#include <BAN/CircularQueue.h>
 #include <BAN/NoCopyMove.h>
 #include <BAN/RefPtr.h>
 #include <BAN/UniqPtr.h>
@@ -50,8 +49,8 @@ namespace Kernel
 
 		bool has_signal_to_execute() const;
 		void set_signal_done(int signal);
-		void handle_next_signal();
-		void queue_signal(int signal);
+		void handle_signal(int signal = 0);
+		bool add_signal(int signal);
 
 		void set_return_rsp(uintptr_t& rsp) { m_return_rsp = &rsp; }
 		void set_return_rip(uintptr_t& rip) { m_return_rip = &rip; }
@@ -109,8 +108,8 @@ namespace Kernel
 		uintptr_t*					m_return_rsp		{ nullptr };
 		uintptr_t*					m_return_rip		{ nullptr };
 
-		BAN::CircularQueue<int, 10>	m_signal_queue;
-		uint64_t					m_signal_mask		{ 0 };
+		uint64_t					m_signal_pending_mask	{ 0 };
+		uint64_t					m_signal_block_mask		{ 0 };
 		int							m_handling_signal		{ 0 };
 		static_assert(_SIGMAX < 64);
 

kernel/kernel/Process.cpp

@@ -845,14 +845,14 @@ namespace Kernel
 			return Process::current().sys_raise(signal);
 		
 		LockGuard process_guard(s_process_lock);
-		CriticalScope _;
-
 		for (auto* process : s_processes)
 		{
 			if (process->pid() == pid)
 			{
-				if (signal)
-					process->m_threads.front()->queue_signal(signal);
+				if (signal == 0)
+					return 0;
+				CriticalScope _;
+				process->m_signal_pending_mask |= 1ull << signal;
 				return 0;
 			}
 		}
@@ -864,14 +864,21 @@ namespace Kernel
 	{
 		if (signal < _SIGMIN || signal > _SIGMAX)
 			return BAN::Error::from_errno(EINVAL);
-		ASSERT(m_threads.size() == 1);
+		ASSERT(this == &Process::current());
+		
 		CriticalScope _;
+		
+		// FIXME: support raise with signal blocked
 		Thread& current = Thread::current();
-		current.queue_signal(signal);
-		current.handle_next_signal();
+		if (current.add_signal(signal))
+		{
+			current.handle_signal(signal);
 			return 0;
 		}
 
+		ASSERT_NOT_REACHED();
+	}
+
 	BAN::ErrorOr<long> Process::sys_tcsetpgrp(int fd, pid_t pgid)
 	{
 		LockGuard _(m_lock);

kernel/kernel/Scheduler.cpp

@@ -248,7 +248,7 @@ namespace Kernel
 				start_thread(current->rsp(), current->rip());
 			case Thread::State::Executing:
 				while (current->has_signal_to_execute() && current->state() == Thread::State::Executing)
-					current->handle_next_signal();
+					current->handle_signal();
 				// fall through
 			case Thread::State::Terminating:
 				continue_thread(current->rsp(), current->rip());

kernel/kernel/Thread.cpp

@@ -183,7 +183,7 @@ namespace Kernel
 		// Setup stack for returning
 		{
 			// FIXME: don't use PageTableScope
-			PageTableScope _(m_process->page_table());
+			PageTableScope _(process().page_table());
 			write_to_stack(m_rsp, this);
 			write_to_stack(m_rsp, &Thread::on_exit);
 			write_to_stack(m_rsp, nullptr);
@@ -204,12 +204,13 @@ namespace Kernel
 		m_rsp = stack_base() + stack_size();
 		m_rip = (uintptr_t)entry;
 
-		m_signal_mask = ~0ull;
+		m_signal_pending_mask = 0;
+		m_signal_block_mask = ~0ull;
 
 		// Setup stack for returning
 		{
 			// FIXME: don't use PageTableScope
-			PageTableScope _(m_process->page_table());
+			PageTableScope _(process().page_table());
 			write_to_stack(m_rsp, this);
 			write_to_stack(m_rsp, &Thread::on_exit);
 			write_to_stack(m_rsp, m_process);
@@ -218,7 +219,10 @@ namespace Kernel
 
 	bool Thread::has_signal_to_execute() const
 	{
-		return !m_signal_queue.empty() && !m_handling_signal;
+		if (!m_process || m_handling_signal)
+			return false;
+		uint64_t full_pending_mask = m_signal_pending_mask | m_process->m_signal_pending_mask;
+		return full_pending_mask & ~m_signal_block_mask;
 	}
 
 	void Thread::set_signal_done(int signal)
@@ -232,26 +236,40 @@ namespace Kernel
 			m_handling_signal = 0;
 	}
 
-	void Thread::handle_next_signal()
+	void Thread::handle_signal(int signal)
 	{
 		ASSERT(!interrupts_enabled());
-		ASSERT(!m_signal_queue.empty());
 		ASSERT(&Thread::current() == this);
 		ASSERT(is_userspace());
 
-		int signal = m_signal_queue.front();
-		ASSERT(signal >= _SIGMIN && signal <= _SIGMAX);
-		m_signal_queue.pop();
+		if (signal == 0)
+		{
+			uint64_t full_pending_mask = m_signal_pending_mask | process().m_signal_pending_mask;
+			for (signal = _SIGMIN; signal <= _SIGMAX; signal++)
+			{
+				uint64_t mask = 1ull << signal;
+				if ((full_pending_mask & mask) && !(m_signal_block_mask & mask))
+					break;
+			}
+			ASSERT(signal <= _SIGMAX);
+		}
+		else
+		{
+			uint64_t full_pending_mask = m_signal_pending_mask | process().m_signal_pending_mask;
+			uint64_t mask = 1ull << signal;
+			ASSERT(full_pending_mask & mask);
+			ASSERT(!(m_signal_block_mask & mask));
+		}
 
 		uintptr_t& return_rsp = this->return_rsp();
 		uintptr_t& return_rip = this->return_rip();
 
 		vaddr_t signal_handler = process().m_signal_handlers[signal];
 
-		// Skip masked and ignored signals
-		if (m_signal_mask & (1ull << signal))
-			;
-		else if (signal_handler == (vaddr_t)SIG_IGN)
+		m_signal_pending_mask &= ~(1ull << signal);
+		process().m_signal_pending_mask &= ~(1ull << signal);
+
+		if (signal_handler == (vaddr_t)SIG_IGN)
 			;
 		else if (signal_handler != (vaddr_t)SIG_DFL)
 		{
@@ -321,17 +339,18 @@ namespace Kernel
 		}
 	}
 
-	void Thread::queue_signal(int signal)
+	bool Thread::add_signal(int signal)
 	{
 		ASSERT(!interrupts_enabled());
-		if (m_signal_queue.full())
+		uint64_t mask = 1ull << signal;
+		if (!(m_signal_block_mask & mask))
 		{
-			dwarnln("Signal queue full");
-			return;
-		}
-		m_signal_queue.push(signal);
+			m_signal_pending_mask |= mask;
 			if (this != &Thread::current())
 				Scheduler::get().unblock_thread(tid());
+			return true;
+		}
+		return false;
 	}
 
 	void Thread::validate_stack() const