signal part 4 (sigsuspend etc)

arch/x86/kernel/cpu.c

@@ -368,7 +368,7 @@ void setup_x86_ap(void (*next_func)(void))
 }
 
 void arch_show_interrupt_context(const void *reg);
-void set_signal(int, void *, int);
+void set_signal(int, void *);
 void check_signal(long, void *);
 
 void handle_interrupt(int vector, struct x86_regs *regs)
@@ -409,7 +409,7 @@ void gpe_handler(struct x86_regs *regs)
 	kprintf("General protection fault (err: %lx, %lx:%lx)\n",
 	        regs->error, regs->cs, regs->rip);
 	arch_show_interrupt_context(regs);
-	set_signal(SIGILL, regs, 1);
+	set_signal(SIGILL, regs);
 	check_signal(0, regs);
 	// panic("GPF");
 }

arch/x86/kernel/include/signal.h

@@ -33,6 +33,16 @@ struct sigaction {
 	sigset_t sa_mask;
 };
 
+//// work in progress
+struct sigbackup {
+	unsigned long rc; // return code of rt_sigreturn (rax)
+	void	*uctx;
+	sigset_t mask;
+	double	fpregs[16];
+	
+};
+////
+
 struct k_sigaction {
         struct sigaction sa;
 };

arch/x86/kernel/include/syscall_list.h

@@ -41,6 +41,7 @@ SYSCALL_DELEGATED(20, writev)
 SYSCALL_DELEGATED(21, access)
 SYSCALL_HANDLED(24, sched_yield)
 SYSCALL_HANDLED(28, madvise)
+SYSCALL_HANDLED(34, pause)
 SYSCALL_HANDLED(39, getpid)
 SYSCALL_HANDLED(56, clone)
 SYSCALL_HANDLED(59, execve)

arch/x86/kernel/syscall.c

@@ -95,6 +95,8 @@ SYSCALL_DECLARE(rt_sigreturn)
 	regs -= 16;
         memcpy(regs, proc->sigstack, 128);
 
+	proc->sigmask.__val[0] = proc->supmask.__val[0];
+
         return proc->sigrc;
 }
 
@@ -104,58 +106,18 @@ extern void interrupt_syscall(int all);
 extern int num_processors;
 
 void
-check_signal(unsigned long rc, unsigned long *regs)
+do_signal(unsigned long rc, unsigned long *regs, struct process *proc, struct sig_pending *pending)
 {
-	struct process *proc;
 	struct k_sigaction *k;
 	int	sig;
-	struct sig_pending *pending;
-	struct sig_pending *next;
-	struct list_head *head;
-	ihk_spinlock_t *lock;
 	__sigset_t w;
 	int	irqstate;
 
-	if(clv == NULL)
-		return;
-	proc = cpu_local_var(current);
-	if(proc == NULL || proc->pid == 0)
-		return;
-
-	w = proc->sigmask.__val[0];
-	lock = &proc->sigshared->lock;
-	head = &proc->sigshared->sigpending;
-	pending = NULL;
-	irqstate = ihk_mc_spinlock_lock(lock);
-	list_for_each_entry_safe(pending, next, head, list){
-		if(!(pending->sigmask.__val[0] & w)){
-			list_del(&pending->list);
-			break;
-		}
-	}
-	if(&pending->list == head)
-		pending = NULL;
-	ihk_mc_spinlock_unlock(lock, irqstate);
-
-	if(!pending){
-		lock = &proc->sigpendinglock;
-		head = &proc->sigpending;
-		irqstate = ihk_mc_spinlock_lock(lock);
-		list_for_each_entry_safe(pending, next, head, list){
-			if(!(pending->sigmask.__val[0] & w)){
-				list_del(&pending->list);
-				break;
-			}
-		}
-		if(&pending->list == head)
-			pending = NULL;
-		ihk_mc_spinlock_unlock(lock, irqstate);
-	}
-	if(!pending)
-		return;
-
 	for(w = pending->sigmask.__val[0], sig = 0; w; sig++, w >>= 1);
 
+	if(sig == SIGKILL || sig == SIGTERM)
+		terminate(0, sig, (ihk_mc_user_context_t *)regs[14]);
+
 	irqstate = ihk_mc_spinlock_lock(&proc->sighandler->lock);
 	if(regs == NULL){ /* call from syscall */
 		asm volatile ("movq %%gs:132,%0" : "=r" (regs));
@@ -174,14 +136,6 @@ check_signal(unsigned long rc, unsigned long *regs)
 	else if(k->sa.sa_handler){
 		unsigned long *usp; /* user stack */
 
-		if(regs[14] & 0x8000000000000000){ // kernel addr
-			int flag = ihk_mc_spinlock_lock(lock);
-			list_add(&pending->list, head);
-			ihk_mc_spinlock_unlock(lock, flag);
-			ihk_mc_spinlock_unlock(&proc->sighandler->lock, irqstate);
-			return;
-		}
-
 		usp = (void *)regs[14];
 		memcpy(proc->sigstack, regs, 128);
 		proc->sigrc = rc;
@@ -192,6 +146,7 @@ check_signal(unsigned long rc, unsigned long *regs)
 		regs[11] = (unsigned long)k->sa.sa_handler;
 		regs[14] = (unsigned long)usp;
 		kfree(pending);
+		proc->sigmask.__val[0] |= pending->sigmask.__val[0];
 		ihk_mc_spinlock_unlock(&proc->sighandler->lock, irqstate);
 	}
 	else{
@@ -203,6 +158,63 @@ check_signal(unsigned long rc, unsigned long *regs)
 	}
 }
 
+void
+check_signal(unsigned long rc, unsigned long *regs)
+{
+	struct process *proc;
+	struct sig_pending *pending;
+	struct sig_pending *next;
+	struct list_head *head;
+	ihk_spinlock_t *lock;
+	__sigset_t w;
+	int	irqstate;
+
+	if(clv == NULL)
+		return;
+	proc = cpu_local_var(current);
+	if(proc == NULL || proc->pid == 0)
+		return;
+
+	if(regs != NULL && (regs[14] & 0x8000000000000000))
+		return;
+
+	for(;;){
+		w = proc->sigmask.__val[0];
+		lock = &proc->sigshared->lock;
+		head = &proc->sigshared->sigpending;
+		pending = NULL;
+		irqstate = ihk_mc_spinlock_lock(lock);
+		list_for_each_entry_safe(pending, next, head, list){
+			if(!(pending->sigmask.__val[0] & w)){
+				list_del(&pending->list);
+				break;
+			}
+		}
+		if(&pending->list == head)
+			pending = NULL;
+		ihk_mc_spinlock_unlock(lock, irqstate);
+
+		if(!pending){
+			lock = &proc->sigpendinglock;
+			head = &proc->sigpending;
+			irqstate = ihk_mc_spinlock_lock(lock);
+			list_for_each_entry_safe(pending, next, head, list){
+				if(!(pending->sigmask.__val[0] & w)){
+					list_del(&pending->list);
+					break;
+				}
+			}
+			if(&pending->list == head)
+				pending = NULL;
+			ihk_mc_spinlock_unlock(lock, irqstate);
+		}
+		if(!pending)
+			return;
+
+		do_signal(rc, regs, proc, pending);
+	}
+}
+
 unsigned long
 do_kill(int pid, int tid, int sig)
 {
@@ -284,6 +296,7 @@ do_kill(int pid, int tid, int sig)
 		}
 		else{
 			list_add_tail(&pending->list, head);
+			proc->sigevent = 1;
 		}
 	}
 	if(tid == -1){
@@ -297,17 +310,16 @@ do_kill(int pid, int tid, int sig)
 }
 
 void
-set_signal(int sig, unsigned long *regs, int nonmaskable)
+set_signal(int sig, unsigned long *regs)
 {
 	struct process *proc = cpu_local_var(current);
 
 	if(proc == NULL || proc->pid == 0)
 		return;
 
-	if(__sigmask(sig) & proc->sigmask.__val[0]){
+	if((__sigmask(sig) & proc->sigmask.__val[0]) ||
-		if(nonmaskable){
+	   (regs[14] & 0x8000000000000000))
-			terminate(0, sig, (ihk_mc_user_context_t *)regs[14]);
+		terminate(0, sig, (ihk_mc_user_context_t *)regs[14]);
-		}
+	else
-	}
+		do_kill(proc->pid, proc->tid, sig);
-	do_kill(proc->pid, proc->tid, sig);
 }

kernel/mem.c

@@ -164,7 +164,7 @@ static struct ihk_mc_interrupt_handler query_free_mem_handler = {
 	.priv = NULL,
 };
 
-void set_signal(int, unsigned long *, int);
+void set_signal(int, unsigned long *);
 void check_signal(long, unsigned long *);
 
 static void unhandled_page_fault(struct process *proc, void *fault_addr, void *regs)
@@ -234,10 +234,10 @@ static void page_fault_handler(void *fault_addr, uint64_t reason, void *regs)
 				reason, regs, error);
 		unhandled_page_fault(proc, fault_addr, regs);
 		if (error == -ERANGE) {
-			set_signal(SIGBUS, regs, 1);
+			set_signal(SIGBUS, regs);
 		}
 		else {
-			set_signal(SIGSEGV, regs, 1);
+			set_signal(SIGSEGV, regs);
 		}
 		check_signal(0, regs);
 		goto out;

kernel/syscall.c

@@ -93,6 +93,7 @@ static char *syscall_name[] MCKERNEL_UNUSED = {
 };
 
 void check_signal(long rc, unsigned long *regs);
+void do_signal(long rc, unsigned long *regs, struct process *proc, struct sig_pending *pending);
 int copy_from_user(struct process *, void *, const void *, size_t);
 int copy_to_user(struct process *, void *, const void *, size_t);
 
@@ -1154,37 +1155,132 @@ SYSCALL_DECLARE(rt_sigpending)
 
 SYSCALL_DECLARE(rt_sigtimedwait)
 {
-/*
+	struct process *proc = cpu_local_var(current);
-sigset_t *
+	const sigset_t *set = (const sigset_t *)ihk_mc_syscall_arg0(ctx);
-siginfo_t *
+	siginfo_t *info = (siginfo_t *)ihk_mc_syscall_arg1(ctx);
-struct timespec *
+	void *timeout = (void *)ihk_mc_syscall_arg2(ctx);
-size_t
+	size_t sigsetsize = (size_t)ihk_mc_syscall_arg3(ctx);
-*/
+	siginfo_t winfo;
-	return 0;
+	__sigset_t wset;
+	long wtimeout[2];
+
+	if (sigsetsize > sizeof(sigset_t))
+		return -EINVAL;
+
+	memset(&winfo, '\0', sizeof winfo);
+	if(copy_from_user(proc, &wset, set, sizeof wset))
+		return -EFAULT;
+	if(copy_from_user(proc, wtimeout, timeout, sizeof wtimeout))
+		return -EFAULT;
+	if(copy_to_user(proc, info, &winfo, sizeof winfo))
+		return -EFAULT;
+
+	return -EOPNOTSUPP;
 }
 
 SYSCALL_DECLARE(rt_sigqueueinfo)
 {
-/*
+	struct process *proc = cpu_local_var(current);
-pid_t
+	int pid = (int)ihk_mc_syscall_arg0(ctx);
-int
+	int sig = (int)ihk_mc_syscall_arg1(ctx);
-siginfo_t *
+	siginfo_t *info = (siginfo_t *)ihk_mc_syscall_arg2(ctx);
-*/
+	siginfo_t winfo;
-	return 0;
+
+	if(copy_from_user(proc, &winfo, info, sizeof winfo))
+		return -EFAULT;
+
+	return -EOPNOTSUPP;
+}
+
+static int
+do_sigsuspend(struct process *proc, const sigset_t *set)
+{
+	__sigset_t wset;
+	int flag;
+	struct sig_pending *pending;
+	struct list_head *head;
+	ihk_spinlock_t *lock;
+
+	wset = set->__val[0];
+	wset &= ~__sigmask(SIGKILL);
+	wset &= ~__sigmask(SIGTERM);
+	proc->sigmask.__val[0] = wset;
+
+	for(;;){
+		while(proc->sigevent == 0);
+		proc->sigevent = 0;
+
+		lock = &proc->sigshared->lock;
+		head = &proc->sigshared->sigpending;
+		flag = ihk_mc_spinlock_lock(lock);
+		list_for_each_entry(pending, head, list){
+			if(!(pending->sigmask.__val[0] & wset))
+				break;
+		}
+
+		if(&pending->list == head){
+			ihk_mc_spinlock_unlock(lock, flag);
+
+			lock = &proc->sigpendinglock;
+			head = &proc->sigpending;
+			flag = ihk_mc_spinlock_lock(lock);
+			list_for_each_entry(pending, head, list){
+				if(!(pending->sigmask.__val[0] & wset))
+					break;
+			}
+		}
+		if(&pending->list == head){
+			ihk_mc_spinlock_unlock(lock, flag);
+			continue;
+		}
+
+		list_del(&pending->list);
+		ihk_mc_spinlock_unlock(lock, flag);
+		do_signal(-EINTR, NULL, proc, pending);
+		break;
+	}
+	return -EINTR;
+}
+
+
+SYSCALL_DECLARE(pause)
+{
+	struct process *proc = cpu_local_var(current);
+
+	return do_sigsuspend(proc, &proc->sigmask);
 }
 
 SYSCALL_DECLARE(rt_sigsuspend)
 {
-/*
+	struct process *proc = cpu_local_var(current);
-sigset_t *
+	const sigset_t *set = (const sigset_t *)ihk_mc_syscall_arg0(ctx);
-size_t
+	size_t sigsetsize = (size_t)ihk_mc_syscall_arg1(ctx);
-*/
+	sigset_t wset;
-	return 0;
+
+	if (sigsetsize > sizeof(sigset_t))
+		return -EINVAL;
+	if(copy_from_user(proc, &wset, set, sizeof wset))
+		return -EFAULT;
+
+	return do_sigsuspend(proc, &wset);
 }
 
 SYSCALL_DECLARE(sigaltstack)
 {
-	return 0;
+	struct process *proc = cpu_local_var(current);
+	const stack_t *ss = (const stack_t *)ihk_mc_syscall_arg0(ctx);
+	stack_t *oss = (stack_t *)ihk_mc_syscall_arg1(ctx);
+	stack_t	wss;
+
+	memset(&wss, '\0', sizeof wss);
+	if(oss)
+		if(copy_to_user(proc, oss, &wss, sizeof wss))
+			return -EFAULT;
+	if(ss)
+		if(copy_from_user(proc, &wss, ss, sizeof wss))
+			return -EFAULT;
+
+	return -EOPNOTSUPP;
 }
 
 SYSCALL_DECLARE(madvise)