[PATCH] 1/3 Add pselect/ppoll() system calls

[PATCH] 1/3 Add pselect/ppoll() system calls

[David Woodhouse]

This patch adds the pselect() and ppoll() system calls. Most of this
implementation is as it was in the -mm kernel for a period of time
already before the difficulties with signal delivery became apparent.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

diff --git a/fs/select.c b/fs/select.c
index f10a103..4e0da96 100644
--- a/fs/select.c
+++ b/fs/select.c
@@ -179,12 +179,11 @@ get_max:
 #define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR)
 #define POLLEX_SET (POLLPRI)
 
-int do_select(int n, fd_set_bits *fds, long *timeout)
+int do_select(int n, fd_set_bits *fds, int64_t *timeout)
 {
 	struct poll_wqueues table;
 	poll_table *wait;
 	int retval, i;
-	long __timeout = *timeout;
 
 	rcu_read_lock();
 	retval = max_select_fd(n, fds);
@@ -196,11 +195,12 @@ int do_select(int n, fd_set_bits *fds, l
 
 	poll_initwait(&table);
 	wait = &table.pt;
-	if (!__timeout)
+	if (!*timeout)
 		wait = NULL;
 	retval = 0;
 	for (;;) {
 		unsigned long *rinp, *routp, *rexp, *inp, *outp, *exp;
+		long __timeout;
 
 		set_current_state(TASK_INTERRUPTIBLE);
 
@@ -255,22 +255,32 @@ int do_select(int n, fd_set_bits *fds, l
 				*rexp = res_ex;
 		}
 		wait = NULL;
-		if (retval || !__timeout || signal_pending(current))
+		if (retval || !*timeout || signal_pending(current))
 			break;
 		if(table.error) {
 			retval = table.error;
 			break;
 		}
+
+		if (*timeout < 0) {
+			/* Wait indefinitely */
+			__timeout = MAX_SCHEDULE_TIMEOUT;
+		} else if (unlikely(*timeout >= (int64_t)MAX_SCHEDULE_TIMEOUT - 1)) {
+			/* Wait for longer than MAX_SCHEDULE_TIMEOUT. Do it in a loop */
+			__timeout = MAX_SCHEDULE_TIMEOUT - 1;
+			*timeout -= __timeout;
+		} else {
+			__timeout = *timeout;
+			*timeout = 0;
+		}
 		__timeout = schedule_timeout(__timeout);
+		if (*timeout >= 0)
+			*timeout += __timeout;
 	}
 	__set_current_state(TASK_RUNNING);
 
 	poll_freewait(&table);
 
-	/*
-	 * Up-to-date the caller timeout.
-	 */
-	*timeout = __timeout;
 	return retval;
 }
 
@@ -295,36 +305,14 @@ static void select_bits_free(void *bits,
 #define MAX_SELECT_SECONDS \
 	((unsigned long) (MAX_SCHEDULE_TIMEOUT / HZ)-1)
 
-asmlinkage long
-sys_select(int n, fd_set __user *inp, fd_set __user *outp, fd_set __user *exp, struct timeval __user *tvp)
+static int core_sys_select(int n, fd_set __user *inp, fd_set __user *outp,
+			   fd_set __user *exp, int64_t *timeout)
 {
 	fd_set_bits fds;
 	char *bits;
-	long timeout;
 	int ret, size, max_fdset;
 	struct fdtable *fdt;
 
-	timeout = MAX_SCHEDULE_TIMEOUT;
-	if (tvp) {
-		time_t sec, usec;
-
-		if (!access_ok(VERIFY_READ, tvp, sizeof(*tvp))
-		    || __get_user(sec, &tvp->tv_sec)
-		    || __get_user(usec, &tvp->tv_usec)) {
-			ret = -EFAULT;
-			goto out_nofds;
-		}
-
-		ret = -EINVAL;
-		if (sec < 0 || usec < 0)
-			goto out_nofds;
-
-		if ((unsigned long) sec < MAX_SELECT_SECONDS) {
-			timeout = ROUND_UP(usec, 1000000/HZ);
-			timeout += sec * (unsigned long) HZ;
-		}
-	}
-
 	ret = -EINVAL;
 	if (n < 0)
 		goto out_nofds;
@@ -362,18 +350,7 @@ sys_select(int n, fd_set __user *inp, fd
 	zero_fd_set(n, fds.res_out);
 	zero_fd_set(n, fds.res_ex);
 
-	ret = do_select(n, &fds, &timeout);
-
-	if (tvp && !(current->personality & STICKY_TIMEOUTS)) {
-		time_t sec = 0, usec = 0;
-		if (timeout) {
-			sec = timeout / HZ;
-			usec = timeout % HZ;
-			usec *= (1000000/HZ);
-		}
-		put_user(sec, &tvp->tv_sec);
-		put_user(usec, &tvp->tv_usec);
-	}
+	ret = do_select(n, &fds, timeout);
 
 	if (ret < 0)
 		goto out;
@@ -395,6 +372,150 @@ out_nofds:
 	return ret;
 }
 
+asmlinkage long sys_select(int n, fd_set __user *inp, fd_set __user *outp,
+			fd_set __user *exp, struct timeval __user *tvp)
+{
+	int64_t timeout = -1;
+	struct timeval tv;
+	int ret;
+
+	if (tvp) {
+		if (copy_from_user(&tv, tvp, sizeof(tv)))
+			return -EFAULT;
+
+		if (tv.tv_sec < 0 || tv.tv_usec < 0)
+			return -EINVAL;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)tv.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(tv.tv_sec, 1000000/HZ);
+			timeout += tv.tv_sec * HZ;
+		}
+	}
+
+	ret = core_sys_select(n, inp, outp, exp, &timeout);
+
+	if (tvp) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		tv.tv_usec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ));
+		tv.tv_sec = timeout;
+		if (copy_to_user(tvp, &tv, sizeof(tv))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long sys_pselect7(int n, fd_set __user *inp, fd_set __user *outp,
+		fd_set __user *exp, struct timespec __user *tsp,
+		const sigset_t __user *sigmask, size_t sigsetsize)
+{
+	int64_t timeout = MAX_SCHEDULE_TIMEOUT;
+	sigset_t ksigmask, sigsaved;
+	struct timespec ts;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		if (ts.tv_sec < 0 || ts.tv_nsec < 0)
+			return -EINVAL;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)ts.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(ts.tv_sec, 1000000000/HZ);
+			timeout += ts.tv_sec * HZ;
+		}
+	}
+
+	if (sigmask) {
+		/* XXX: Don't preclude handling different sized sigset_t's.  */
+		if (sigsetsize != sizeof(sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
+			return -EFAULT;
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	ret = core_sys_select(n, inp, outp, exp, &timeout);
+
+	if (tsp) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		ts.tv_nsec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ)) * 1000;
+		ts.tv_sec = timeout;
+		if (copy_to_user(tsp, &ts, sizeof(ts))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND)
+				ret = -EINTR;
+		}
+	}
+
+	if (ret == -ERESTARTNOHAND) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	return ret;
+}
+
+/*
+ * Most architectures can't handle 7-argument syscalls. So we provide a
+ * 6-argument version where the sixth argument is a pointer to a structure
+ * which has a pointer to the sigset_t itself followed by a size_t containing
+ * the sigset size.
+ */
+asmlinkage long sys_pselect6(int n, fd_set __user *inp, fd_set __user *outp,
+	fd_set __user *exp, struct timespec __user *tsp, void __user *sig)
+{
+	size_t sigsetsize = 0;
+	sigset_t __user *up = NULL;
+
+	if (sig) {
+		if (!access_ok(VERIFY_READ, sig, sizeof(void *)+sizeof(size_t))
+		    || __get_user(up, (sigset_t * __user *)sig)
+		    || __get_user(sigsetsize,
+				(size_t * __user)(sig+sizeof(void *))))
+			return -EFAULT;
+	}
+
+	return sys_pselect7(n, inp, outp, exp, tsp, up, sigsetsize);
+}
+#endif /* TIF_RESTORE_SIGMASK */
+
 struct poll_list {
 	struct poll_list *next;
 	int len;
@@ -436,16 +557,19 @@ static void do_pollfd(unsigned int num, 
 }
 
 static int do_poll(unsigned int nfds,  struct poll_list *list,
-			struct poll_wqueues *wait, long timeout)
+		   struct poll_wqueues *wait, int64_t *timeout)
 {
 	int count = 0;
 	poll_table* pt = &wait->pt;
 
-	if (!timeout)
+	/* Optimise the no-wait case */
+	if (!(*timeout))
 		pt = NULL;
  
 	for (;;) {
 		struct poll_list *walk;
+		long __timeout;
+
 		set_current_state(TASK_INTERRUPTIBLE);
 		walk = list;
 		while(walk != NULL) {
@@ -453,18 +577,33 @@ static int do_poll(unsigned int nfds,  s
 			walk = walk->next;
 		}
 		pt = NULL;
-		if (count || !timeout || signal_pending(current))
+		if (count || !*timeout || signal_pending(current))
 			break;
 		count = wait->error;
 		if (count)
 			break;
-		timeout = schedule_timeout(timeout);
+
+		if (*timeout < 0) {
+			/* Wait indefinitely */
+			__timeout = MAX_SCHEDULE_TIMEOUT;
+		} else if (unlikely(*timeout >= (int64_t)MAX_SCHEDULE_TIMEOUT - 1)) {
+			/* Wait for longer than MAX_SCHEDULE_TIMEOUT. Do it in a loop */
+			__timeout = MAX_SCHEDULE_TIMEOUT - 1;
+			*timeout -= __timeout;
+		} else {
+			__timeout = *timeout;
+			*timeout = 0;
+		}
+
+		__timeout = schedule_timeout(__timeout);
+		if (*timeout >= 0)
+			*timeout += __timeout;
 	}
 	__set_current_state(TASK_RUNNING);
 	return count;
 }
 
-asmlinkage long sys_poll(struct pollfd __user * ufds, unsigned int nfds, long timeout)
+int do_sys_poll(struct pollfd __user *ufds, unsigned int nfds, int64_t *timeout)
 {
 	struct poll_wqueues table;
  	int fdcount, err;
@@ -482,14 +621,6 @@ asmlinkage long sys_poll(struct pollfd _
 	if (nfds > max_fdset && nfds > OPEN_MAX)
 		return -EINVAL;
 
-	if (timeout) {
-		/* Careful about overflow in the intermediate values */
-		if ((unsigned long) timeout < MAX_SCHEDULE_TIMEOUT / HZ)
-			timeout = (unsigned long)(timeout*HZ+999)/1000+1;
-		else /* Negative or overflow */
-			timeout = MAX_SCHEDULE_TIMEOUT;
-	}
-
 	poll_initwait(&table);
 
 	head = NULL;
@@ -519,6 +650,7 @@ asmlinkage long sys_poll(struct pollfd _
 		}
 		i -= pp->len;
 	}
+
 	fdcount = do_poll(nfds, head, &table, timeout);
 
 	/* OK, now copy the revents fields back to user space. */
@@ -547,3 +679,94 @@ out_fds:
 	poll_freewait(&table);
 	return err;
 }
+
+asmlinkage long sys_poll(struct pollfd __user *ufds, unsigned int nfds,
+			long timeout_msecs)
+{
+	int64_t timeout_jiffies = 0;
+
+	if (timeout_msecs) {
+#if HZ > 1000
+		/* We can only overflow if HZ > 1000 */
+		if (timeout_msecs / 1000 > (int64_t)0x7fffffffffffffffULL / (int64_t)HZ)
+			timeout_jiffies = -1;
+		else
+#endif
+			timeout_jiffies = msecs_to_jiffies(timeout_msecs);
+	}
+
+	return do_sys_poll(ufds, nfds, &timeout_jiffies);
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long sys_ppoll(struct pollfd __user *ufds, unsigned int nfds,
+	struct timespec __user *tsp, const sigset_t __user *sigmask,
+	size_t sigsetsize)
+{
+	sigset_t ksigmask, sigsaved;
+	struct timespec ts;
+	int64_t timeout = -1;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)ts.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(ts.tv_sec, 1000000000/HZ);
+			timeout += ts.tv_sec * HZ;
+		}
+	}
+
+	if (sigmask) {
+		/* XXX: Don't preclude handling different sized sigset_t's.  */
+		if (sigsetsize != sizeof(sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
+			return -EFAULT;
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	ret = do_sys_poll(ufds, nfds, &timeout);
+
+	/* We can restart this syscall, usually */
+	if (ret == -EINTR) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+		ret = -ERESTARTNOHAND;
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	if (tsp && timeout >= 0) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		/* Yes, we know it's actually an int64_t, but it's also positive. */
+		ts.tv_nsec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ)) * 1000;
+		ts.tv_sec = timeout;
+		if (copy_to_user(tsp, &ts, sizeof(ts))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND && timeout >= 0)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+#endif /* TIF_RESTORE_SIGMASK */
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -53,6 +53,8 @@
 #include <asm/mmu_context.h>
 #include <asm/ioctls.h>
 
+extern void sigset_from_compat(sigset_t *set, compat_sigset_t *compat);
+
 /*
  * Not all architectures have sys_utime, so implement this in terms
  * of sys_utimes.
@@ -1607,36 +1609,14 @@ static void select_bits_free(void *bits,
 #define MAX_SELECT_SECONDS \
 	((unsigned long) (MAX_SCHEDULE_TIMEOUT / HZ)-1)
 
-asmlinkage long
-compat_sys_select(int n, compat_ulong_t __user *inp, compat_ulong_t __user *outp,
-		compat_ulong_t __user *exp, struct compat_timeval __user *tvp)
+int compat_core_sys_select(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp, int64_t *timeout)
 {
 	fd_set_bits fds;
 	char *bits;
-	long timeout;
 	int size, max_fdset, ret = -EINVAL;
 	struct fdtable *fdt;
 
-	timeout = MAX_SCHEDULE_TIMEOUT;
-	if (tvp) {
-		time_t sec, usec;
-
-		if (!access_ok(VERIFY_READ, tvp, sizeof(*tvp))
-		    || __get_user(sec, &tvp->tv_sec)
-		    || __get_user(usec, &tvp->tv_usec)) {
-			ret = -EFAULT;
-			goto out_nofds;
-		}
-
-		if (sec < 0 || usec < 0)
-			goto out_nofds;
-
-		if ((unsigned long) sec < MAX_SELECT_SECONDS) {
-			timeout = ROUND_UP(usec, 1000000/HZ);
-			timeout += sec * (unsigned long) HZ;
-		}
-	}
-
 	if (n < 0)
 		goto out_nofds;
 
@@ -1673,19 +1653,7 @@ compat_sys_select(int n, compat_ulong_t 
 	zero_fd_set(n, fds.res_out);
 	zero_fd_set(n, fds.res_ex);
 
-	ret = do_select(n, &fds, &timeout);
-
-	if (tvp && !(current->personality & STICKY_TIMEOUTS)) {
-		time_t sec = 0, usec = 0;
-		if (timeout) {
-			sec = timeout / HZ;
-			usec = timeout % HZ;
-			usec *= (1000000/HZ);
-		}
-		if (put_user(sec, &tvp->tv_sec) ||
-		    put_user(usec, &tvp->tv_usec))
-			ret = -EFAULT;
-	}
+	ret = do_select(n, &fds, timeout);
 
 	if (ret < 0)
 		goto out;
@@ -1706,6 +1674,216 @@ out_nofds:
 	return ret;
 }
 
+asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+	struct compat_timeval __user *tvp)
+{
+	int64_t timeout = -1;
+	struct compat_timeval tv;
+	int ret;
+
+	if (tvp) {
+		if (copy_from_user(&tv, tvp, sizeof(tv)))
+			return -EFAULT;
+
+		if (tv.tv_sec < 0 || tv.tv_usec < 0)
+			return -EINVAL;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)tv.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(tv.tv_sec, 1000000/HZ);
+			timeout += tv.tv_sec * HZ;
+		}
+	}
+
+	ret = compat_core_sys_select(n, inp, outp, exp, &timeout);
+
+	if (tvp) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		tv.tv_usec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ));
+		tv.tv_sec = timeout;
+		if (copy_to_user(tvp, &tv, sizeof(tv))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long compat_sys_pselect7(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+	struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
+	compat_size_t sigsetsize)
+{
+	compat_sigset_t s32;
+	sigset_t ksigmask, sigsaved;
+	long timeout = MAX_SCHEDULE_TIMEOUT;
+	struct compat_timespec ts;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		if (ts.tv_sec < 0 || ts.tv_nsec < 0)
+			return -EINVAL;
+	}
+
+	if (sigmask) {
+		if (sigsetsize != sizeof(compat_sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&s32, sigmask, sizeof(s32)))
+			return -EFAULT;
+		sigset_from_compat(&ksigmask, &s32);
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	do {
+		if (tsp) {
+			if ((unsigned long)ts.tv_sec < MAX_SELECT_SECONDS) {
+				timeout = ROUND_UP(ts.tv_nsec, 1000000000/HZ);
+				timeout += ts.tv_sec * (unsigned long)HZ;
+				ts.tv_sec = 0;
+				ts.tv_nsec = 0;
+			} else {
+				ts.tv_sec -= MAX_SELECT_SECONDS;
+				timeout = MAX_SELECT_SECONDS * HZ;
+			}
+		}
+
+		ret = compat_core_sys_select(n, inp, outp, exp, &timeout);
+
+	} while (!ret && !timeout && tsp && (ts.tv_sec || ts.tv_nsec));
+
+	if (tsp && !(current->personality & STICKY_TIMEOUTS)) {
+		ts.tv_sec += timeout / HZ;
+		ts.tv_nsec += (timeout % HZ) * (1000000000/HZ);
+		if (ts.tv_nsec >= 1000000000) {
+			ts.tv_sec++;
+			ts.tv_nsec -= 1000000000;
+		}
+		(void)copy_to_user(tsp, &ts, sizeof(ts));
+	}
+
+	if (ret == -ERESTARTNOHAND) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	return ret;
+}
+
+asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+	struct compat_timespec __user *tsp, void __user *sig)
+{
+	compat_size_t sigsetsize = 0;
+	compat_uptr_t up = 0;
+
+	if (sig) {
+		if (!access_ok(VERIFY_READ, sig,
+				sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
+		    	__get_user(up, (compat_uptr_t __user *)sig) ||
+		    	__get_user(sigsetsize,
+				(compat_size_t __user *)(sig+sizeof(up))))
+			return -EFAULT;
+	}
+	return compat_sys_pselect7(n, inp, outp, exp, tsp, compat_ptr(up),
+					sigsetsize);
+}
+
+asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
+	unsigned int nfds, struct compat_timespec __user *tsp,
+	const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
+{
+	compat_sigset_t s32;
+	sigset_t ksigmask, sigsaved;
+	struct compat_timespec ts;
+	int64_t timeout = -1;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		/* We assume that ts.tv_sec is always lower than
+		   the number of seconds that can be expressed in
+		   an int64_t. Otherwise the compiler bitches at us */
+		timeout = ROUND_UP(ts.tv_sec, 1000000000/HZ);
+		timeout += ts.tv_sec * HZ;
+	}
+
+	if (sigmask) {
+		if (sigsetsize |= sizeof(compat_sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&s32, sigmask, sizeof(s32)))
+			return -EFAULT;
+		sigset_from_compat(&ksigmask, &s32);
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	ret = do_sys_poll(ufds, nfds, &timeout);
+
+	/* We can restart this syscall, usually */
+	if (ret == -EINTR) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+		ret = -ERESTARTNOHAND;
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	if (tsp && timeout >= 0) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		/* Yes, we know it's actually an int64_t, but it's also positive. */
+		ts.tv_nsec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ)) * 1000;
+		ts.tv_sec = timeout;
+		if (copy_to_user(tsp, &ts, sizeof(ts))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND && timeout >= 0)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+#endif /* TIF_RESTORE_SIGMASK */
+
 #if defined(CONFIG_NFSD) || defined(CONFIG_NFSD_MODULE)
 /* Stuff for NFS server syscalls... */
 struct compat_nfsctl_svc {
diff --git a/include/linux/poll.h b/include/linux/poll.h
index f6da702..5b9c85c 100644
--- a/include/linux/poll.h
+++ b/include/linux/poll.h
@@ -92,7 +92,11 @@ void zero_fd_set(unsigned long nr, unsig
 	memset(fdset, 0, FDS_BYTES(nr));
 }
 
-extern int do_select(int n, fd_set_bits *fds, long *timeout);
+#define MAX_INT64_SECONDS (((int64_t)(~((uint64_t)0)>>1)/HZ)-1)
+
+extern int do_select(int n, fd_set_bits *fds, int64_t *timeout);
+extern int do_sys_poll(struct pollfd __user * ufds, unsigned int nfds,
+		       int64_t *timeout);
 
 #endif /* KERNEL */
 
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2038bd2..bf7db28 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -800,6 +800,7 @@ struct task_struct {
 	struct sighand_struct *sighand;
 
 	sigset_t blocked, real_blocked;
+	sigset_t saved_sigmask;		/* To be restored with TIF_RESTORE_SIGMASK */
 	struct sigpending pending;
 
 	unsigned long sas_ss_sp;



-- 
dwmw2

[PATCH] 1/3 Add pselect/ppoll() system calls

[David Woodhouse]

This patch adds the pselect() and ppoll() system calls. Most of this
implementation is as it was in the -mm kernel for a period of time
already before the difficulties with signal delivery became apparent.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

diff --git a/fs/select.c b/fs/select.c
index f10a103..4e0da96 100644
--- a/fs/select.c
+++ b/fs/select.c
@@ -179,12 +179,11 @@ get_max:
 #define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR)
 #define POLLEX_SET (POLLPRI)
 
-int do_select(int n, fd_set_bits *fds, long *timeout)
+int do_select(int n, fd_set_bits *fds, int64_t *timeout)
 {
 	struct poll_wqueues table;
 	poll_table *wait;
 	int retval, i;
-	long __timeout = *timeout;
 
 	rcu_read_lock();
 	retval = max_select_fd(n, fds);
@@ -196,11 +195,12 @@ int do_select(int n, fd_set_bits *fds, l
 
 	poll_initwait(&table);
 	wait = &table.pt;
-	if (!__timeout)
+	if (!*timeout)
 		wait = NULL;
 	retval = 0;
 	for (;;) {
 		unsigned long *rinp, *routp, *rexp, *inp, *outp, *exp;
+		long __timeout;
 
 		set_current_state(TASK_INTERRUPTIBLE);
 
@@ -255,22 +255,32 @@ int do_select(int n, fd_set_bits *fds, l
 				*rexp = res_ex;
 		}
 		wait = NULL;
-		if (retval || !__timeout || signal_pending(current))
+		if (retval || !*timeout || signal_pending(current))
 			break;
 		if(table.error) {
 			retval = table.error;
 			break;
 		}
+
+		if (*timeout < 0) {
+			/* Wait indefinitely */
+			__timeout = MAX_SCHEDULE_TIMEOUT;
+		} else if (unlikely(*timeout >= (int64_t)MAX_SCHEDULE_TIMEOUT - 1)) {
+			/* Wait for longer than MAX_SCHEDULE_TIMEOUT. Do it in a loop */
+			__timeout = MAX_SCHEDULE_TIMEOUT - 1;
+			*timeout -= __timeout;
+		} else {
+			__timeout = *timeout;
+			*timeout = 0;
+		}
 		__timeout = schedule_timeout(__timeout);
+		if (*timeout >= 0)
+			*timeout += __timeout;
 	}
 	__set_current_state(TASK_RUNNING);
 
 	poll_freewait(&table);
 
-	/*
-	 * Up-to-date the caller timeout.
-	 */
-	*timeout = __timeout;
 	return retval;
 }
 
@@ -295,36 +305,14 @@ static void select_bits_free(void *bits,
 #define MAX_SELECT_SECONDS \
 	((unsigned long) (MAX_SCHEDULE_TIMEOUT / HZ)-1)
 
-asmlinkage long
-sys_select(int n, fd_set __user *inp, fd_set __user *outp, fd_set __user *exp, struct timeval __user *tvp)
+static int core_sys_select(int n, fd_set __user *inp, fd_set __user *outp,
+			   fd_set __user *exp, int64_t *timeout)
 {
 	fd_set_bits fds;
 	char *bits;
-	long timeout;
 	int ret, size, max_fdset;
 	struct fdtable *fdt;
 
-	timeout = MAX_SCHEDULE_TIMEOUT;
-	if (tvp) {
-		time_t sec, usec;
-
-		if (!access_ok(VERIFY_READ, tvp, sizeof(*tvp))
-		    || __get_user(sec, &tvp->tv_sec)
-		    || __get_user(usec, &tvp->tv_usec)) {
-			ret = -EFAULT;
-			goto out_nofds;
-		}
-
-		ret = -EINVAL;
-		if (sec < 0 || usec < 0)
-			goto out_nofds;
-
-		if ((unsigned long) sec < MAX_SELECT_SECONDS) {
-			timeout = ROUND_UP(usec, 1000000/HZ);
-			timeout += sec * (unsigned long) HZ;
-		}
-	}
-
 	ret = -EINVAL;
 	if (n < 0)
 		goto out_nofds;
@@ -362,18 +350,7 @@ sys_select(int n, fd_set __user *inp, fd
 	zero_fd_set(n, fds.res_out);
 	zero_fd_set(n, fds.res_ex);
 
-	ret = do_select(n, &fds, &timeout);
-
-	if (tvp && !(current->personality & STICKY_TIMEOUTS)) {
-		time_t sec = 0, usec = 0;
-		if (timeout) {
-			sec = timeout / HZ;
-			usec = timeout % HZ;
-			usec *= (1000000/HZ);
-		}
-		put_user(sec, &tvp->tv_sec);
-		put_user(usec, &tvp->tv_usec);
-	}
+	ret = do_select(n, &fds, timeout);
 
 	if (ret < 0)
 		goto out;
@@ -395,6 +372,150 @@ out_nofds:
 	return ret;
 }
 
+asmlinkage long sys_select(int n, fd_set __user *inp, fd_set __user *outp,
+			fd_set __user *exp, struct timeval __user *tvp)
+{
+	int64_t timeout = -1;
+	struct timeval tv;
+	int ret;
+
+	if (tvp) {
+		if (copy_from_user(&tv, tvp, sizeof(tv)))
+			return -EFAULT;
+
+		if (tv.tv_sec < 0 || tv.tv_usec < 0)
+			return -EINVAL;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)tv.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(tv.tv_sec, 1000000/HZ);
+			timeout += tv.tv_sec * HZ;
+		}
+	}
+
+	ret = core_sys_select(n, inp, outp, exp, &timeout);
+
+	if (tvp) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		tv.tv_usec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ));
+		tv.tv_sec = timeout;
+		if (copy_to_user(tvp, &tv, sizeof(tv))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long sys_pselect7(int n, fd_set __user *inp, fd_set __user *outp,
+		fd_set __user *exp, struct timespec __user *tsp,
+		const sigset_t __user *sigmask, size_t sigsetsize)
+{
+	int64_t timeout = MAX_SCHEDULE_TIMEOUT;
+	sigset_t ksigmask, sigsaved;
+	struct timespec ts;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		if (ts.tv_sec < 0 || ts.tv_nsec < 0)
+			return -EINVAL;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)ts.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(ts.tv_sec, 1000000000/HZ);
+			timeout += ts.tv_sec * HZ;
+		}
+	}
+
+	if (sigmask) {
+		/* XXX: Don't preclude handling different sized sigset_t's.  */
+		if (sigsetsize != sizeof(sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
+			return -EFAULT;
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	ret = core_sys_select(n, inp, outp, exp, &timeout);
+
+	if (tsp) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		ts.tv_nsec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ)) * 1000;
+		ts.tv_sec = timeout;
+		if (copy_to_user(tsp, &ts, sizeof(ts))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND)
+				ret = -EINTR;
+		}
+	}
+
+	if (ret == -ERESTARTNOHAND) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	return ret;
+}
+
+/*
+ * Most architectures can't handle 7-argument syscalls. So we provide a
+ * 6-argument version where the sixth argument is a pointer to a structure
+ * which has a pointer to the sigset_t itself followed by a size_t containing
+ * the sigset size.
+ */
+asmlinkage long sys_pselect6(int n, fd_set __user *inp, fd_set __user *outp,
+	fd_set __user *exp, struct timespec __user *tsp, void __user *sig)
+{
+	size_t sigsetsize = 0;
+	sigset_t __user *up = NULL;
+
+	if (sig) {
+		if (!access_ok(VERIFY_READ, sig, sizeof(void *)+sizeof(size_t))
+		    || __get_user(up, (sigset_t * __user *)sig)
+		    || __get_user(sigsetsize,
+				(size_t * __user)(sig+sizeof(void *))))
+			return -EFAULT;
+	}
+
+	return sys_pselect7(n, inp, outp, exp, tsp, up, sigsetsize);
+}
+#endif /* TIF_RESTORE_SIGMASK */
+
 struct poll_list {
 	struct poll_list *next;
 	int len;
@@ -436,16 +557,19 @@ static void do_pollfd(unsigned int num, 
 }
 
 static int do_poll(unsigned int nfds,  struct poll_list *list,
-			struct poll_wqueues *wait, long timeout)
+		   struct poll_wqueues *wait, int64_t *timeout)
 {
 	int count = 0;
 	poll_table* pt = &wait->pt;
 
-	if (!timeout)
+	/* Optimise the no-wait case */
+	if (!(*timeout))
 		pt = NULL;
  
 	for (;;) {
 		struct poll_list *walk;
+		long __timeout;
+
 		set_current_state(TASK_INTERRUPTIBLE);
 		walk = list;
 		while(walk != NULL) {
@@ -453,18 +577,33 @@ static int do_poll(unsigned int nfds,  s
 			walk = walk->next;
 		}
 		pt = NULL;
-		if (count || !timeout || signal_pending(current))
+		if (count || !*timeout || signal_pending(current))
 			break;
 		count = wait->error;
 		if (count)
 			break;
-		timeout = schedule_timeout(timeout);
+
+		if (*timeout < 0) {
+			/* Wait indefinitely */
+			__timeout = MAX_SCHEDULE_TIMEOUT;
+		} else if (unlikely(*timeout >= (int64_t)MAX_SCHEDULE_TIMEOUT - 1)) {
+			/* Wait for longer than MAX_SCHEDULE_TIMEOUT. Do it in a loop */
+			__timeout = MAX_SCHEDULE_TIMEOUT - 1;
+			*timeout -= __timeout;
+		} else {
+			__timeout = *timeout;
+			*timeout = 0;
+		}
+
+		__timeout = schedule_timeout(__timeout);
+		if (*timeout >= 0)
+			*timeout += __timeout;
 	}
 	__set_current_state(TASK_RUNNING);
 	return count;
 }
 
-asmlinkage long sys_poll(struct pollfd __user * ufds, unsigned int nfds, long timeout)
+int do_sys_poll(struct pollfd __user *ufds, unsigned int nfds, int64_t *timeout)
 {
 	struct poll_wqueues table;
  	int fdcount, err;
@@ -482,14 +621,6 @@ asmlinkage long sys_poll(struct pollfd _
 	if (nfds > max_fdset && nfds > OPEN_MAX)
 		return -EINVAL;
 
-	if (timeout) {
-		/* Careful about overflow in the intermediate values */
-		if ((unsigned long) timeout < MAX_SCHEDULE_TIMEOUT / HZ)
-			timeout = (unsigned long)(timeout*HZ+999)/1000+1;
-		else /* Negative or overflow */
-			timeout = MAX_SCHEDULE_TIMEOUT;
-	}
-
 	poll_initwait(&table);
 
 	head = NULL;
@@ -519,6 +650,7 @@ asmlinkage long sys_poll(struct pollfd _
 		}
 		i -= pp->len;
 	}
+
 	fdcount = do_poll(nfds, head, &table, timeout);
 
 	/* OK, now copy the revents fields back to user space. */
@@ -547,3 +679,94 @@ out_fds:
 	poll_freewait(&table);
 	return err;
 }
+
+asmlinkage long sys_poll(struct pollfd __user *ufds, unsigned int nfds,
+			long timeout_msecs)
+{
+	int64_t timeout_jiffies = 0;
+
+	if (timeout_msecs) {
+#if HZ > 1000
+		/* We can only overflow if HZ > 1000 */
+		if (timeout_msecs / 1000 > (int64_t)0x7fffffffffffffffULL / (int64_t)HZ)
+			timeout_jiffies = -1;
+		else
+#endif
+			timeout_jiffies = msecs_to_jiffies(timeout_msecs);
+	}
+
+	return do_sys_poll(ufds, nfds, &timeout_jiffies);
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long sys_ppoll(struct pollfd __user *ufds, unsigned int nfds,
+	struct timespec __user *tsp, const sigset_t __user *sigmask,
+	size_t sigsetsize)
+{
+	sigset_t ksigmask, sigsaved;
+	struct timespec ts;
+	int64_t timeout = -1;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)ts.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(ts.tv_sec, 1000000000/HZ);
+			timeout += ts.tv_sec * HZ;
+		}
+	}
+
+	if (sigmask) {
+		/* XXX: Don't preclude handling different sized sigset_t's.  */
+		if (sigsetsize != sizeof(sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
+			return -EFAULT;
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	ret = do_sys_poll(ufds, nfds, &timeout);
+
+	/* We can restart this syscall, usually */
+	if (ret == -EINTR) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+		ret = -ERESTARTNOHAND;
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	if (tsp && timeout >= 0) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		/* Yes, we know it's actually an int64_t, but it's also positive. */
+		ts.tv_nsec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ)) * 1000;
+		ts.tv_sec = timeout;
+		if (copy_to_user(tsp, &ts, sizeof(ts))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND && timeout >= 0)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+#endif /* TIF_RESTORE_SIGMASK */
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -53,6 +53,8 @@
 #include <asm/mmu_context.h>
 #include <asm/ioctls.h>
 
+extern void sigset_from_compat(sigset_t *set, compat_sigset_t *compat);
+
 /*
  * Not all architectures have sys_utime, so implement this in terms
  * of sys_utimes.
@@ -1607,36 +1609,14 @@ static void select_bits_free(void *bits,
 #define MAX_SELECT_SECONDS \
 	((unsigned long) (MAX_SCHEDULE_TIMEOUT / HZ)-1)
 
-asmlinkage long
-compat_sys_select(int n, compat_ulong_t __user *inp, compat_ulong_t __user *outp,
-		compat_ulong_t __user *exp, struct compat_timeval __user *tvp)
+int compat_core_sys_select(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp, int64_t *timeout)
 {
 	fd_set_bits fds;
 	char *bits;
-	long timeout;
 	int size, max_fdset, ret = -EINVAL;
 	struct fdtable *fdt;
 
-	timeout = MAX_SCHEDULE_TIMEOUT;
-	if (tvp) {
-		time_t sec, usec;
-
-		if (!access_ok(VERIFY_READ, tvp, sizeof(*tvp))
-		    || __get_user(sec, &tvp->tv_sec)
-		    || __get_user(usec, &tvp->tv_usec)) {
-			ret = -EFAULT;
-			goto out_nofds;
-		}
-
-		if (sec < 0 || usec < 0)
-			goto out_nofds;
-
-		if ((unsigned long) sec < MAX_SELECT_SECONDS) {
-			timeout = ROUND_UP(usec, 1000000/HZ);
-			timeout += sec * (unsigned long) HZ;
-		}
-	}
-
 	if (n < 0)
 		goto out_nofds;
 
@@ -1673,19 +1653,7 @@ compat_sys_select(int n, compat_ulong_t 
 	zero_fd_set(n, fds.res_out);
 	zero_fd_set(n, fds.res_ex);
 
-	ret = do_select(n, &fds, &timeout);
-
-	if (tvp && !(current->personality & STICKY_TIMEOUTS)) {
-		time_t sec = 0, usec = 0;
-		if (timeout) {
-			sec = timeout / HZ;
-			usec = timeout % HZ;
-			usec *= (1000000/HZ);
-		}
-		if (put_user(sec, &tvp->tv_sec) ||
-		    put_user(usec, &tvp->tv_usec))
-			ret = -EFAULT;
-	}
+	ret = do_select(n, &fds, timeout);
 
 	if (ret < 0)
 		goto out;
@@ -1706,6 +1674,216 @@ out_nofds:
 	return ret;
 }
 
+asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+	struct compat_timeval __user *tvp)
+{
+	int64_t timeout = -1;
+	struct compat_timeval tv;
+	int ret;
+
+	if (tvp) {
+		if (copy_from_user(&tv, tvp, sizeof(tv)))
+			return -EFAULT;
+
+		if (tv.tv_sec < 0 || tv.tv_usec < 0)
+			return -EINVAL;
+
+		/* Cast to uint64_t to make GCC stop complaining */
+		if ((uint64_t)tv.tv_sec >= (uint64_t)MAX_INT64_SECONDS)
+			timeout = -1;	/* infinite */
+		else {
+			timeout = ROUND_UP(tv.tv_sec, 1000000/HZ);
+			timeout += tv.tv_sec * HZ;
+		}
+	}
+
+	ret = compat_core_sys_select(n, inp, outp, exp, &timeout);
+
+	if (tvp) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		tv.tv_usec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ));
+		tv.tv_sec = timeout;
+		if (copy_to_user(tvp, &tv, sizeof(tv))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long compat_sys_pselect7(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+	struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
+	compat_size_t sigsetsize)
+{
+	compat_sigset_t s32;
+	sigset_t ksigmask, sigsaved;
+	long timeout = MAX_SCHEDULE_TIMEOUT;
+	struct compat_timespec ts;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		if (ts.tv_sec < 0 || ts.tv_nsec < 0)
+			return -EINVAL;
+	}
+
+	if (sigmask) {
+		if (sigsetsize != sizeof(compat_sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&s32, sigmask, sizeof(s32)))
+			return -EFAULT;
+		sigset_from_compat(&ksigmask, &s32);
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	do {
+		if (tsp) {
+			if ((unsigned long)ts.tv_sec < MAX_SELECT_SECONDS) {
+				timeout = ROUND_UP(ts.tv_nsec, 1000000000/HZ);
+				timeout += ts.tv_sec * (unsigned long)HZ;
+				ts.tv_sec = 0;
+				ts.tv_nsec = 0;
+			} else {
+				ts.tv_sec -= MAX_SELECT_SECONDS;
+				timeout = MAX_SELECT_SECONDS * HZ;
+			}
+		}
+
+		ret = compat_core_sys_select(n, inp, outp, exp, &timeout);
+
+	} while (!ret && !timeout && tsp && (ts.tv_sec || ts.tv_nsec));
+
+	if (tsp && !(current->personality & STICKY_TIMEOUTS)) {
+		ts.tv_sec += timeout / HZ;
+		ts.tv_nsec += (timeout % HZ) * (1000000000/HZ);
+		if (ts.tv_nsec >= 1000000000) {
+			ts.tv_sec++;
+			ts.tv_nsec -= 1000000000;
+		}
+		(void)copy_to_user(tsp, &ts, sizeof(ts));
+	}
+
+	if (ret == -ERESTARTNOHAND) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	return ret;
+}
+
+asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
+	compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+	struct compat_timespec __user *tsp, void __user *sig)
+{
+	compat_size_t sigsetsize = 0;
+	compat_uptr_t up = 0;
+
+	if (sig) {
+		if (!access_ok(VERIFY_READ, sig,
+				sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
+		    	__get_user(up, (compat_uptr_t __user *)sig) ||
+		    	__get_user(sigsetsize,
+				(compat_size_t __user *)(sig+sizeof(up))))
+			return -EFAULT;
+	}
+	return compat_sys_pselect7(n, inp, outp, exp, tsp, compat_ptr(up),
+					sigsetsize);
+}
+
+asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
+	unsigned int nfds, struct compat_timespec __user *tsp,
+	const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
+{
+	compat_sigset_t s32;
+	sigset_t ksigmask, sigsaved;
+	struct compat_timespec ts;
+	int64_t timeout = -1;
+	int ret;
+
+	if (tsp) {
+		if (copy_from_user(&ts, tsp, sizeof(ts)))
+			return -EFAULT;
+
+		/* We assume that ts.tv_sec is always lower than
+		   the number of seconds that can be expressed in
+		   an int64_t. Otherwise the compiler bitches at us */
+		timeout = ROUND_UP(ts.tv_sec, 1000000000/HZ);
+		timeout += ts.tv_sec * HZ;
+	}
+
+	if (sigmask) {
+		if (sigsetsize |= sizeof(compat_sigset_t))
+			return -EINVAL;
+		if (copy_from_user(&s32, sigmask, sizeof(s32)))
+			return -EFAULT;
+		sigset_from_compat(&ksigmask, &s32);
+
+		sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+		sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+	}
+
+	ret = do_sys_poll(ufds, nfds, &timeout);
+
+	/* We can restart this syscall, usually */
+	if (ret == -EINTR) {
+		/* Don't restore the signal mask yet. Let do_signal() deliver the signal
+		   on the way back to userspace, before the signal mask is restored. */
+		if (sigmask) {
+			memcpy(&current->saved_sigmask, &sigsaved, sizeof(sigsaved));
+			set_thread_flag(TIF_RESTORE_SIGMASK);
+		}
+		ret = -ERESTARTNOHAND;
+	} else if (sigmask)
+		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+	if (tsp && timeout >= 0) {
+		if (current->personality & STICKY_TIMEOUTS)
+			goto sticky;
+		/* Yes, we know it's actually an int64_t, but it's also positive. */
+		ts.tv_nsec = jiffies_to_usecs(do_div((*(uint64_t*)&timeout), HZ)) * 1000;
+		ts.tv_sec = timeout;
+		if (copy_to_user(tsp, &ts, sizeof(ts))) {
+		sticky:
+			/*
+			 * If an application puts its timeval in read-only
+			 * memory, we don't want the Linux-specific update to
+			 * the timeval to cause a fault after the select has
+			 * completed successfully. However, because we're not
+			 * updating the timeval, we can't restart the system
+			 * call.
+			 */
+			if (ret == -ERESTARTNOHAND && timeout >= 0)
+				ret = -EINTR;
+		}
+	}
+
+	return ret;
+}
+#endif /* TIF_RESTORE_SIGMASK */
+
 #if defined(CONFIG_NFSD) || defined(CONFIG_NFSD_MODULE)
 /* Stuff for NFS server syscalls... */
 struct compat_nfsctl_svc {
diff --git a/include/linux/poll.h b/include/linux/poll.h
index f6da702..5b9c85c 100644
--- a/include/linux/poll.h
+++ b/include/linux/poll.h
@@ -92,7 +92,11 @@ void zero_fd_set(unsigned long nr, unsig
 	memset(fdset, 0, FDS_BYTES(nr));
 }
 
-extern int do_select(int n, fd_set_bits *fds, long *timeout);
+#define MAX_INT64_SECONDS (((int64_t)(~((uint64_t)0)>>1)/HZ)-1)
+
+extern int do_select(int n, fd_set_bits *fds, int64_t *timeout);
+extern int do_sys_poll(struct pollfd __user * ufds, unsigned int nfds,
+		       int64_t *timeout);
 
 #endif /* KERNEL */
 
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2038bd2..bf7db28 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -800,6 +800,7 @@ struct task_struct {
 	struct sighand_struct *sighand;
 
 	sigset_t blocked, real_blocked;
+	sigset_t saved_sigmask;		/* To be restored with TIF_RESTORE_SIGMASK */
 	struct sigpending pending;
 
 	unsigned long sas_ss_sp;



-- 
dwmw2

[PATCH] 2/3 Add TIF_RESTORE_SIGMASK support for arch/powerpc

[David Woodhouse]

This patch implements the TIF_RESTORE_SIGMASK flag in the new
arch/powerpc kernel, for both 32-bit and 64-bit system call paths.

Mostly tested on 32-bit so far.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

diff --git a/arch/powerpc/kernel/entry_32.S b/arch/powerpc/kernel/entry_32.S
index 036b71d..6aa7386 100644
--- a/arch/powerpc/kernel/entry_32.S
+++ b/arch/powerpc/kernel/entry_32.S
@@ -227,7 +227,7 @@ ret_from_syscall:
 	MTMSRD(r10)
 	lwz	r9,TI_FLAGS(r12)
 	li	r8,-_LAST_ERRNO
-	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL)
+	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_RESTORE_SIGMASK)
 	bne-	syscall_exit_work
 	cmplw	0,r3,r8
 	blt+	syscall_exit_cont
@@ -357,7 +357,7 @@ save_user_nvgprs_cont:
 	lwz	r5,_MSR(r1)
 	andi.	r5,r5,MSR_PR
 	beq	ret_from_except
-	andi.	r0,r9,_TIF_SIGPENDING
+	andi.	r0,r9,_TIF_SIGPENDING|_TIF_RESTORE_SIGMASK
 	beq	ret_from_except
 	b	do_user_signal
 8:
@@ -683,7 +683,7 @@ user_exc_return:		/* r10 contains MSR_KE
 	/* Check current_thread_info()->flags */
 	rlwinm	r9,r1,0,0,(31-THREAD_SHIFT)
 	lwz	r9,TI_FLAGS(r9)
-	andi.	r0,r9,(_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL)
+	andi.	r0,r9,(_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_RESTORE_SIGMASK)
 	bne	do_work
 
 restore_user:
@@ -917,7 +917,7 @@ recheck:
 	lwz	r9,TI_FLAGS(r9)
 	andi.	r0,r9,_TIF_NEED_RESCHED
 	bne-	do_resched
-	andi.	r0,r9,_TIF_SIGPENDING
+	andi.	r0,r9,_TIF_SIGPENDING|_TIF_RESTORE_SIGMASK
 	beq	restore_user
 do_user_signal:			/* r10 contains MSR_KERNEL here */
 	ori	r10,r10,MSR_EE
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index 2eb6f54..22e592d 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -160,7 +160,7 @@ syscall_exit:
 	mtmsrd	r10,1
 	ld	r9,TI_FLAGS(r12)
 	li	r11,-_LAST_ERRNO
-	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SINGLESTEP|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_SAVE_NVGPRS|_TIF_NOERROR)
+	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SINGLESTEP|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_SAVE_NVGPRS|_TIF_NOERROR|_TIF_RESTORE_SIGMASK)
 	bne-	syscall_exit_work
 	cmpld	r3,r11
 	ld	r5,_CCR(r1)
--- a/arch/powerpc/kernel/signal_32.c
+++ b/arch/powerpc/kernel/signal_32.c
@@ -1177,7 +1131,7 @@ int do_signal(sigset_t *oldset, struct p
 {
 	siginfo_t info;
 	struct k_sigaction ka;
-	unsigned int frame, newsp;
+	unsigned int newsp;
 	int signr, ret;
 
 #ifdef CONFIG_PPC32
@@ -1188,11 +1142,11 @@ int do_signal(sigset_t *oldset, struct p
 	}
 #endif
 
-	if (!oldset)
+	if (test_thread_flag(TIF_RESTORE_SIGMASK))
+		oldset = &current->saved_sigmask;
+	else if (!oldset)
 		oldset = &current->blocked;
 
-	newsp = frame = 0;
-
 	signr = get_signal_to_deliver(&info, &ka, regs, NULL);
 #ifdef CONFIG_PPC32
 no_signal:
@@ -1222,8 +1176,14 @@ no_signal:
 		}
 	}
 
-	if (signr == 0)
+	if (signr == 0) {
+		/* No signal to deliver -- put the saved sigmask back */
+		if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
+			clear_thread_flag(TIF_RESTORE_SIGMASK);
+			sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
+		}
 		return 0;		/* no signals delivered */
+	}
 
 	if ((ka.sa.sa_flags & SA_ONSTACK) && current->sas_ss_size
 	    && !on_sig_stack(regs->gpr[1]))
@@ -1256,6 +1216,10 @@ no_signal:
 			sigaddset(&current->blocked, signr);
 		recalc_sigpending();
 		spin_unlock_irq(&current->sighand->siglock);
+		/* A signal was successfully delivered; the saved sigmask is in
+		   its frame, and we can clear the TIF_RESTORE_SIGMASK flag */
+		if (test_thread_flag(TIF_RESTORE_SIGMASK))
+			clear_thread_flag(TIF_RESTORE_SIGMASK);
 	}
 
 	return ret;
diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c
index 5462bef..7a6f0f7 100644
--- a/arch/powerpc/kernel/signal_64.c
+++ b/arch/powerpc/kernel/signal_64.c
@@ -554,11 +518,15 @@ int do_signal(sigset_t *oldset, struct p
 	if (test_thread_flag(TIF_32BIT))
 		return do_signal32(oldset, regs);
 
-	if (!oldset)
+	if (test_thread_flag(TIF_RESTORE_SIGMASK))
+		oldset = &current->saved_sigmask;
+	else if (!oldset)
 		oldset = &current->blocked;
 
 	signr = get_signal_to_deliver(&info, &ka, regs, NULL);
 	if (signr > 0) {
+		int ret;
+
 		/* Whee!  Actually deliver the signal.  */
 		if (TRAP(regs) == 0x0C00)
 			syscall_restart(regs, &ka);
@@ -571,7 +539,14 @@ int do_signal(sigset_t *oldset, struct p
 		if (current->thread.dabr)
 			set_dabr(current->thread.dabr);
 
-		return handle_signal(signr, &ka, &info, oldset, regs);
+		ret = handle_signal(signr, &ka, &info, oldset, regs);
+
+		/* If a signal was successfully delivered, the saved sigmask is in
+		   its frame, and we can clear the TIF_RESTORE_SIGMASK flag */
+		if (ret && test_thread_flag(TIF_RESTORE_SIGMASK))
+			clear_thread_flag(TIF_RESTORE_SIGMASK);
+
+		return ret;
 	}
 
 	if (TRAP(regs) == 0x0C00) {	/* System Call! */
@@ -587,6 +562,11 @@ int do_signal(sigset_t *oldset, struct p
 			regs->result = 0;
 		}
 	}
+	/* No signal to deliver -- put the saved sigmask back */
+	if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
+		clear_thread_flag(TIF_RESTORE_SIGMASK);
+		sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
+	}
 
 	return 0;
 }
diff --git a/arch/powerpc/kernel/systbl.S b/arch/powerpc/kernel/systbl.S
index 989f628..70bfff2 100644
--- a/arch/powerpc/kernel/systbl.S
+++ b/arch/powerpc/kernel/systbl.S
@@ -321,3 +321,5 @@ SYSCALL(inotify_add_watch)
 SYSCALL(inotify_rm_watch)
 SYSCALL(spu_run)
 SYSCALL(spu_create)
+COMPAT_SYS(pselect6)
+COMPAT_SYS(ppoll)	
diff --git a/fs/compat.c b/fs/compat.c
index 8186341..51747b9 100644
--- a/include/asm-powerpc/thread_info.h
+++ b/include/asm-powerpc/thread_info.h
@@ -125,6 +125,7 @@ static inline struct thread_info *curren
 #define TIF_RESTOREALL		12	/* Restore all regs (implies NOERROR) */
 #define TIF_SAVE_NVGPRS		13	/* Save r14-r31 in signal frame */
 #define TIF_NOERROR		14	/* Force successful syscall return */
+#define TIF_RESTORE_SIGMASK	15	/* Restore signal mask in do_signal */
 
 /* as above, but as bit values */
 #define _TIF_SYSCALL_TRACE	(1<<TIF_SYSCALL_TRACE)
@@ -141,10 +142,12 @@ static inline struct thread_info *curren
 #define _TIF_RESTOREALL		(1<<TIF_RESTOREALL)
 #define _TIF_SAVE_NVGPRS	(1<<TIF_SAVE_NVGPRS)
 #define _TIF_NOERROR		(1<<TIF_NOERROR)
+#define _TIF_RESTORE_SIGMASK	(1<<TIF_RESTORE_SIGMASK)
 #define _TIF_SYSCALL_T_OR_A	(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP)
 
 #define _TIF_USER_WORK_MASK	(_TIF_NOTIFY_RESUME | _TIF_SIGPENDING | \
-				 _TIF_NEED_RESCHED | _TIF_RESTOREALL)
+				 _TIF_NEED_RESCHED | _TIF_RESTOREALL | \
+				 _TIF_RESTORE_SIGMASK)
 #define _TIF_PERSYSCALL_MASK	(_TIF_RESTOREALL|_TIF_NOERROR|_TIF_SAVE_NVGPRS)
 
 #endif /* __KERNEL__ */
diff --git a/include/asm-powerpc/unistd.h b/include/asm-powerpc/unistd.h
index 19eaac3..a40cdff 100644
--- a/include/asm-powerpc/unistd.h
+++ b/include/asm-powerpc/unistd.h
@@ -298,8 +298,10 @@
 #define __NR_inotify_rm_watch	277
 #define __NR_spu_run		278
 #define __NR_spu_create		279
+#define __NR_pselect6		280
+#define __NR_ppoll		281
 
-#define __NR_syscalls		280
+#define __NR_syscalls		282
 
 #ifdef __KERNEL__
 #define __NR__exit __NR_exit

-- 
dwmw2

[PATCH] 2/3 Add TIF_RESTORE_SIGMASK support for arch/powerpc

[David Woodhouse]

This patch implements the TIF_RESTORE_SIGMASK flag in the new
arch/powerpc kernel, for both 32-bit and 64-bit system call paths.

Mostly tested on 32-bit so far.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

diff --git a/arch/powerpc/kernel/entry_32.S b/arch/powerpc/kernel/entry_32.S
index 036b71d..6aa7386 100644
--- a/arch/powerpc/kernel/entry_32.S
+++ b/arch/powerpc/kernel/entry_32.S
@@ -227,7 +227,7 @@ ret_from_syscall:
 	MTMSRD(r10)
 	lwz	r9,TI_FLAGS(r12)
 	li	r8,-_LAST_ERRNO
-	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL)
+	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_RESTORE_SIGMASK)
 	bne-	syscall_exit_work
 	cmplw	0,r3,r8
 	blt+	syscall_exit_cont
@@ -357,7 +357,7 @@ save_user_nvgprs_cont:
 	lwz	r5,_MSR(r1)
 	andi.	r5,r5,MSR_PR
 	beq	ret_from_except
-	andi.	r0,r9,_TIF_SIGPENDING
+	andi.	r0,r9,_TIF_SIGPENDING|_TIF_RESTORE_SIGMASK
 	beq	ret_from_except
 	b	do_user_signal
 8:
@@ -683,7 +683,7 @@ user_exc_return:		/* r10 contains MSR_KE
 	/* Check current_thread_info()->flags */
 	rlwinm	r9,r1,0,0,(31-THREAD_SHIFT)
 	lwz	r9,TI_FLAGS(r9)
-	andi.	r0,r9,(_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL)
+	andi.	r0,r9,(_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_RESTORE_SIGMASK)
 	bne	do_work
 
 restore_user:
@@ -917,7 +917,7 @@ recheck:
 	lwz	r9,TI_FLAGS(r9)
 	andi.	r0,r9,_TIF_NEED_RESCHED
 	bne-	do_resched
-	andi.	r0,r9,_TIF_SIGPENDING
+	andi.	r0,r9,_TIF_SIGPENDING|_TIF_RESTORE_SIGMASK
 	beq	restore_user
 do_user_signal:			/* r10 contains MSR_KERNEL here */
 	ori	r10,r10,MSR_EE
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index 2eb6f54..22e592d 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -160,7 +160,7 @@ syscall_exit:
 	mtmsrd	r10,1
 	ld	r9,TI_FLAGS(r12)
 	li	r11,-_LAST_ERRNO
-	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SINGLESTEP|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_SAVE_NVGPRS|_TIF_NOERROR)
+	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SINGLESTEP|_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_RESTOREALL|_TIF_SAVE_NVGPRS|_TIF_NOERROR|_TIF_RESTORE_SIGMASK)
 	bne-	syscall_exit_work
 	cmpld	r3,r11
 	ld	r5,_CCR(r1)
--- a/arch/powerpc/kernel/signal_32.c
+++ b/arch/powerpc/kernel/signal_32.c
@@ -1177,7 +1131,7 @@ int do_signal(sigset_t *oldset, struct p
 {
 	siginfo_t info;
 	struct k_sigaction ka;
-	unsigned int frame, newsp;
+	unsigned int newsp;
 	int signr, ret;
 
 #ifdef CONFIG_PPC32
@@ -1188,11 +1142,11 @@ int do_signal(sigset_t *oldset, struct p
 	}
 #endif
 
-	if (!oldset)
+	if (test_thread_flag(TIF_RESTORE_SIGMASK))
+		oldset = &current->saved_sigmask;
+	else if (!oldset)
 		oldset = &current->blocked;
 
-	newsp = frame = 0;
-
 	signr = get_signal_to_deliver(&info, &ka, regs, NULL);
 #ifdef CONFIG_PPC32
 no_signal:
@@ -1222,8 +1176,14 @@ no_signal:
 		}
 	}
 
-	if (signr == 0)
+	if (signr == 0) {
+		/* No signal to deliver -- put the saved sigmask back */
+		if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
+			clear_thread_flag(TIF_RESTORE_SIGMASK);
+			sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
+		}
 		return 0;		/* no signals delivered */
+	}
 
 	if ((ka.sa.sa_flags & SA_ONSTACK) && current->sas_ss_size
 	    && !on_sig_stack(regs->gpr[1]))
@@ -1256,6 +1216,10 @@ no_signal:
 			sigaddset(&current->blocked, signr);
 		recalc_sigpending();
 		spin_unlock_irq(&current->sighand->siglock);
+		/* A signal was successfully delivered; the saved sigmask is in
+		   its frame, and we can clear the TIF_RESTORE_SIGMASK flag */
+		if (test_thread_flag(TIF_RESTORE_SIGMASK))
+			clear_thread_flag(TIF_RESTORE_SIGMASK);
 	}
 
 	return ret;
diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c
index 5462bef..7a6f0f7 100644
--- a/arch/powerpc/kernel/signal_64.c
+++ b/arch/powerpc/kernel/signal_64.c
@@ -554,11 +518,15 @@ int do_signal(sigset_t *oldset, struct p
 	if (test_thread_flag(TIF_32BIT))
 		return do_signal32(oldset, regs);
 
-	if (!oldset)
+	if (test_thread_flag(TIF_RESTORE_SIGMASK))
+		oldset = &current->saved_sigmask;
+	else if (!oldset)
 		oldset = &current->blocked;
 
 	signr = get_signal_to_deliver(&info, &ka, regs, NULL);
 	if (signr > 0) {
+		int ret;
+
 		/* Whee!  Actually deliver the signal.  */
 		if (TRAP(regs) == 0x0C00)
 			syscall_restart(regs, &ka);
@@ -571,7 +539,14 @@ int do_signal(sigset_t *oldset, struct p
 		if (current->thread.dabr)
 			set_dabr(current->thread.dabr);
 
-		return handle_signal(signr, &ka, &info, oldset, regs);
+		ret = handle_signal(signr, &ka, &info, oldset, regs);
+
+		/* If a signal was successfully delivered, the saved sigmask is in
+		   its frame, and we can clear the TIF_RESTORE_SIGMASK flag */
+		if (ret && test_thread_flag(TIF_RESTORE_SIGMASK))
+			clear_thread_flag(TIF_RESTORE_SIGMASK);
+
+		return ret;
 	}
 
 	if (TRAP(regs) == 0x0C00) {	/* System Call! */
@@ -587,6 +562,11 @@ int do_signal(sigset_t *oldset, struct p
 			regs->result = 0;
 		}
 	}
+	/* No signal to deliver -- put the saved sigmask back */
+	if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
+		clear_thread_flag(TIF_RESTORE_SIGMASK);
+		sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
+	}
 
 	return 0;
 }
diff --git a/arch/powerpc/kernel/systbl.S b/arch/powerpc/kernel/systbl.S
index 989f628..70bfff2 100644
--- a/arch/powerpc/kernel/systbl.S
+++ b/arch/powerpc/kernel/systbl.S
@@ -321,3 +321,5 @@ SYSCALL(inotify_add_watch)
 SYSCALL(inotify_rm_watch)
 SYSCALL(spu_run)
 SYSCALL(spu_create)
+COMPAT_SYS(pselect6)
+COMPAT_SYS(ppoll)	
diff --git a/fs/compat.c b/fs/compat.c
index 8186341..51747b9 100644
--- a/include/asm-powerpc/thread_info.h
+++ b/include/asm-powerpc/thread_info.h
@@ -125,6 +125,7 @@ static inline struct thread_info *curren
 #define TIF_RESTOREALL		12	/* Restore all regs (implies NOERROR) */
 #define TIF_SAVE_NVGPRS		13	/* Save r14-r31 in signal frame */
 #define TIF_NOERROR		14	/* Force successful syscall return */
+#define TIF_RESTORE_SIGMASK	15	/* Restore signal mask in do_signal */
 
 /* as above, but as bit values */
 #define _TIF_SYSCALL_TRACE	(1<<TIF_SYSCALL_TRACE)
@@ -141,10 +142,12 @@ static inline struct thread_info *curren
 #define _TIF_RESTOREALL		(1<<TIF_RESTOREALL)
 #define _TIF_SAVE_NVGPRS	(1<<TIF_SAVE_NVGPRS)
 #define _TIF_NOERROR		(1<<TIF_NOERROR)
+#define _TIF_RESTORE_SIGMASK	(1<<TIF_RESTORE_SIGMASK)
 #define _TIF_SYSCALL_T_OR_A	(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP)
 
 #define _TIF_USER_WORK_MASK	(_TIF_NOTIFY_RESUME | _TIF_SIGPENDING | \
-				 _TIF_NEED_RESCHED | _TIF_RESTOREALL)
+				 _TIF_NEED_RESCHED | _TIF_RESTOREALL | \
+				 _TIF_RESTORE_SIGMASK)
 #define _TIF_PERSYSCALL_MASK	(_TIF_RESTOREALL|_TIF_NOERROR|_TIF_SAVE_NVGPRS)
 
 #endif /* __KERNEL__ */
diff --git a/include/asm-powerpc/unistd.h b/include/asm-powerpc/unistd.h
index 19eaac3..a40cdff 100644
--- a/include/asm-powerpc/unistd.h
+++ b/include/asm-powerpc/unistd.h
@@ -298,8 +298,10 @@
 #define __NR_inotify_rm_watch	277
 #define __NR_spu_run		278
 #define __NR_spu_create		279
+#define __NR_pselect6		280
+#define __NR_ppoll		281
 
-#define __NR_syscalls		280
+#define __NR_syscalls		282
 
 #ifdef __KERNEL__
 #define __NR__exit __NR_exit

-- 
dwmw2

[PATCH] 3/3 Generic sys_rt_sigsuspend

[David Woodhouse]

The TIF_RESTORE_SIGMASK flag allows us to have a generic implementation
of sys_rt_sigsuspend() instead of duplicating it for each architecture.
This provides such an implementation and makes arch/powerpc use it.

It also tidies up the ppc32 sys_sigsuspend() to use TIF_RESTORE_SIGMASK.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

--- a/arch/powerpc/kernel/signal_32.c
+++ b/arch/powerpc/kernel/signal_32.c
@@ -252,8 +252,7 @@ int do_signal(sigset_t *oldset, struct p
 /*
  * Atomically swap in the new signal mask, and wait for a signal.
  */
-long sys_sigsuspend(old_sigset_t mask, int p2, int p3, int p4, int p6, int p7,
-	       struct pt_regs *regs)
+long sys_sigsuspend(old_sigset_t mask)
 {
 	sigset_t saveset;
 
@@ -264,55 +263,10 @@ long sys_sigsuspend(old_sigset_t mask, i
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);
 
-	regs->result = -EINTR;
-	regs->gpr[3] = EINTR;
-	regs->ccr |= 0x10000000;
-	while (1) {
-		current->state = TASK_INTERRUPTIBLE;
-		schedule();
-		if (do_signal(&saveset, regs)) {
-			set_thread_flag(TIF_RESTOREALL);
-			return 0;
-		}
-	}
-}
-
-long sys_rt_sigsuspend(
-#ifdef CONFIG_PPC64
-		compat_sigset_t __user *unewset,
-#else
-		sigset_t __user *unewset,
-#endif
-		size_t sigsetsize, int p3, int p4,
-		int p6, int p7, struct pt_regs *regs)
-{
-	sigset_t saveset, newset;
-
-	/* XXX: Don't preclude handling different sized sigset_t's.  */
-	if (sigsetsize != sizeof(sigset_t))
-		return -EINVAL;
-
-	if (get_sigset_t(&newset, unewset))
-		return -EFAULT;
-	sigdelsetmask(&newset, ~_BLOCKABLE);
-
-	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
-	current->blocked = newset;
-	recalc_sigpending();
-	spin_unlock_irq(&current->sighand->siglock);
-
-	regs->result = -EINTR;
-	regs->gpr[3] = EINTR;
-	regs->ccr |= 0x10000000;
-	while (1) {
-		current->state = TASK_INTERRUPTIBLE;
-		schedule();
-		if (do_signal(&saveset, regs)) {
-			set_thread_flag(TIF_RESTOREALL);
-			return 0;
-		}
-	}
+ 	current->state = TASK_INTERRUPTIBLE;
+ 	schedule();
+ 	set_thread_flag(TIF_RESTORE_SIGMASK);
+ 	return -ERESTARTNOHAND;
 }
 
 #ifdef CONFIG_PPC32
diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c
index 5462bef..7a6f0f7 100644
--- a/arch/powerpc/kernel/signal_64.c
+++ b/arch/powerpc/kernel/signal_64.c
@@ -67,42 +67,6 @@ struct rt_sigframe {
 	char abigap[288];
 } __attribute__ ((aligned (16)));
 
-
-/*
- * Atomically swap in the new signal mask, and wait for a signal.
- */
-long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize, int p3, int p4,
-		       int p6, int p7, struct pt_regs *regs)
-{
-	sigset_t saveset, newset;
-
-	/* XXX: Don't preclude handling different sized sigset_t's.  */
-	if (sigsetsize != sizeof(sigset_t))
-		return -EINVAL;
-
-	if (copy_from_user(&newset, unewset, sizeof(newset)))
-		return -EFAULT;
-	sigdelsetmask(&newset, ~_BLOCKABLE);
-
-	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
-	current->blocked = newset;
-	recalc_sigpending();
-	spin_unlock_irq(&current->sighand->siglock);
-
-	regs->result = -EINTR;
-	regs->gpr[3] = EINTR;
-	regs->ccr |= 0x10000000;
-	while (1) {
-		current->state = TASK_INTERRUPTIBLE;
-		schedule();
-		if (do_signal(&saveset, regs)) {
-			set_thread_flag(TIF_RESTOREALL);
-			return 0;
-		}
-	}
-}
-
 long sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss, unsigned long r5,
 		     unsigned long r6, unsigned long r7, unsigned long r8,
 		     struct pt_regs *regs)
--- a/include/asm-powerpc/unistd.h
+++ b/include/asm-powerpc/unistd.h
@@ -444,11 +446,13 @@ type name(type1 arg1, type2 arg2, type3 
 #define __ARCH_WANT_SYS_SIGPENDING
 #define __ARCH_WANT_SYS_SIGPROCMASK
 #define __ARCH_WANT_SYS_RT_SIGACTION
+#define __ARCH_WANT_SYS_RT_SIGSUSPEND
 #ifdef CONFIG_PPC32
 #define __ARCH_WANT_OLD_STAT
 #endif
 #ifdef CONFIG_PPC64
 #define __ARCH_WANT_COMPAT_SYS_TIME
+#define __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND
 #endif
 
 /*
--- a/kernel/compat.c
+++ b/kernel/compat.c
@@ -855,3 +855,31 @@ asmlinkage long compat_sys_stime(compat_
 }
 
 #endif /* __ARCH_WANT_COMPAT_SYS_TIME */
+
+#ifdef __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND
+long compat_sys_rt_sigsuspend(compat_sigset_t __user *unewset, compat_size_t sigsetsize)
+{
+	sigset_t saveset, newset;
+	compat_sigset_t newset32;
+
+	/* XXX: Don't preclude handling different sized sigset_t's.  */
+	if (sigsetsize != sizeof(sigset_t))
+		return -EINVAL;
+
+	if (copy_from_user(&newset32, unewset, sizeof(compat_sigset_t)))
+		return -EFAULT;
+	sigset_from_compat(&newset, &newset32);
+	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
+
+	spin_lock_irq(&current->sighand->siglock);
+	saveset = current->blocked;
+	current->blocked = newset;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	current->state = TASK_INTERRUPTIBLE;
+	schedule();
+	set_thread_flag(TIF_RESTORE_SIGMASK);
+	return -ERESTARTNOHAND;
+}
+#endif /* __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND */
diff --git a/kernel/signal.c b/kernel/signal.c
index d7611f1..5e4c240 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2623,6 +2623,32 @@ sys_pause(void)
 
 #endif
 
+#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
+long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize)
+{
+	sigset_t saveset, newset;
+
+	/* XXX: Don't preclude handling different sized sigset_t's.  */
+	if (sigsetsize != sizeof(sigset_t))
+		return -EINVAL;
+
+	if (copy_from_user(&newset, unewset, sizeof(newset)))
+		return -EFAULT;
+	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
+
+	spin_lock_irq(&current->sighand->siglock);
+	saveset = current->blocked;
+	current->blocked = newset;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	current->state = TASK_INTERRUPTIBLE;
+	schedule();
+	set_thread_flag(TIF_RESTORE_SIGMASK);
+	return -ERESTARTNOHAND;
+}
+#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
+
 void __init signals_init(void)
 {
 	sigqueue_cachep =

-- 
dwmw2

[PATCH] 3/3 Generic sys_rt_sigsuspend

[David Woodhouse]

The TIF_RESTORE_SIGMASK flag allows us to have a generic implementation
of sys_rt_sigsuspend() instead of duplicating it for each architecture.
This provides such an implementation and makes arch/powerpc use it.

It also tidies up the ppc32 sys_sigsuspend() to use TIF_RESTORE_SIGMASK.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

--- a/arch/powerpc/kernel/signal_32.c
+++ b/arch/powerpc/kernel/signal_32.c
@@ -252,8 +252,7 @@ int do_signal(sigset_t *oldset, struct p
 /*
  * Atomically swap in the new signal mask, and wait for a signal.
  */
-long sys_sigsuspend(old_sigset_t mask, int p2, int p3, int p4, int p6, int p7,
-	       struct pt_regs *regs)
+long sys_sigsuspend(old_sigset_t mask)
 {
 	sigset_t saveset;
 
@@ -264,55 +263,10 @@ long sys_sigsuspend(old_sigset_t mask, i
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);
 
-	regs->result = -EINTR;
-	regs->gpr[3] = EINTR;
-	regs->ccr |= 0x10000000;
-	while (1) {
-		current->state = TASK_INTERRUPTIBLE;
-		schedule();
-		if (do_signal(&saveset, regs)) {
-			set_thread_flag(TIF_RESTOREALL);
-			return 0;
-		}
-	}
-}
-
-long sys_rt_sigsuspend(
-#ifdef CONFIG_PPC64
-		compat_sigset_t __user *unewset,
-#else
-		sigset_t __user *unewset,
-#endif
-		size_t sigsetsize, int p3, int p4,
-		int p6, int p7, struct pt_regs *regs)
-{
-	sigset_t saveset, newset;
-
-	/* XXX: Don't preclude handling different sized sigset_t's.  */
-	if (sigsetsize != sizeof(sigset_t))
-		return -EINVAL;
-
-	if (get_sigset_t(&newset, unewset))
-		return -EFAULT;
-	sigdelsetmask(&newset, ~_BLOCKABLE);
-
-	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
-	current->blocked = newset;
-	recalc_sigpending();
-	spin_unlock_irq(&current->sighand->siglock);
-
-	regs->result = -EINTR;
-	regs->gpr[3] = EINTR;
-	regs->ccr |= 0x10000000;
-	while (1) {
-		current->state = TASK_INTERRUPTIBLE;
-		schedule();
-		if (do_signal(&saveset, regs)) {
-			set_thread_flag(TIF_RESTOREALL);
-			return 0;
-		}
-	}
+ 	current->state = TASK_INTERRUPTIBLE;
+ 	schedule();
+ 	set_thread_flag(TIF_RESTORE_SIGMASK);
+ 	return -ERESTARTNOHAND;
 }
 
 #ifdef CONFIG_PPC32
diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c
index 5462bef..7a6f0f7 100644
--- a/arch/powerpc/kernel/signal_64.c
+++ b/arch/powerpc/kernel/signal_64.c
@@ -67,42 +67,6 @@ struct rt_sigframe {
 	char abigap[288];
 } __attribute__ ((aligned (16)));
 
-
-/*
- * Atomically swap in the new signal mask, and wait for a signal.
- */
-long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize, int p3, int p4,
-		       int p6, int p7, struct pt_regs *regs)
-{
-	sigset_t saveset, newset;
-
-	/* XXX: Don't preclude handling different sized sigset_t's.  */
-	if (sigsetsize != sizeof(sigset_t))
-		return -EINVAL;
-
-	if (copy_from_user(&newset, unewset, sizeof(newset)))
-		return -EFAULT;
-	sigdelsetmask(&newset, ~_BLOCKABLE);
-
-	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
-	current->blocked = newset;
-	recalc_sigpending();
-	spin_unlock_irq(&current->sighand->siglock);
-
-	regs->result = -EINTR;
-	regs->gpr[3] = EINTR;
-	regs->ccr |= 0x10000000;
-	while (1) {
-		current->state = TASK_INTERRUPTIBLE;
-		schedule();
-		if (do_signal(&saveset, regs)) {
-			set_thread_flag(TIF_RESTOREALL);
-			return 0;
-		}
-	}
-}
-
 long sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss, unsigned long r5,
 		     unsigned long r6, unsigned long r7, unsigned long r8,
 		     struct pt_regs *regs)
--- a/include/asm-powerpc/unistd.h
+++ b/include/asm-powerpc/unistd.h
@@ -444,11 +446,13 @@ type name(type1 arg1, type2 arg2, type3 
 #define __ARCH_WANT_SYS_SIGPENDING
 #define __ARCH_WANT_SYS_SIGPROCMASK
 #define __ARCH_WANT_SYS_RT_SIGACTION
+#define __ARCH_WANT_SYS_RT_SIGSUSPEND
 #ifdef CONFIG_PPC32
 #define __ARCH_WANT_OLD_STAT
 #endif
 #ifdef CONFIG_PPC64
 #define __ARCH_WANT_COMPAT_SYS_TIME
+#define __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND
 #endif
 
 /*
--- a/kernel/compat.c
+++ b/kernel/compat.c
@@ -855,3 +855,31 @@ asmlinkage long compat_sys_stime(compat_
 }
 
 #endif /* __ARCH_WANT_COMPAT_SYS_TIME */
+
+#ifdef __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND
+long compat_sys_rt_sigsuspend(compat_sigset_t __user *unewset, compat_size_t sigsetsize)
+{
+	sigset_t saveset, newset;
+	compat_sigset_t newset32;
+
+	/* XXX: Don't preclude handling different sized sigset_t's.  */
+	if (sigsetsize != sizeof(sigset_t))
+		return -EINVAL;
+
+	if (copy_from_user(&newset32, unewset, sizeof(compat_sigset_t)))
+		return -EFAULT;
+	sigset_from_compat(&newset, &newset32);
+	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
+
+	spin_lock_irq(&current->sighand->siglock);
+	saveset = current->blocked;
+	current->blocked = newset;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	current->state = TASK_INTERRUPTIBLE;
+	schedule();
+	set_thread_flag(TIF_RESTORE_SIGMASK);
+	return -ERESTARTNOHAND;
+}
+#endif /* __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND */
diff --git a/kernel/signal.c b/kernel/signal.c
index d7611f1..5e4c240 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2623,6 +2623,32 @@ sys_pause(void)
 
 #endif
 
+#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
+long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize)
+{
+	sigset_t saveset, newset;
+
+	/* XXX: Don't preclude handling different sized sigset_t's.  */
+	if (sigsetsize != sizeof(sigset_t))
+		return -EINVAL;
+
+	if (copy_from_user(&newset, unewset, sizeof(newset)))
+		return -EFAULT;
+	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
+
+	spin_lock_irq(&current->sighand->siglock);
+	saveset = current->blocked;
+	current->blocked = newset;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	current->state = TASK_INTERRUPTIBLE;
+	schedule();
+	set_thread_flag(TIF_RESTORE_SIGMASK);
+	return -ERESTARTNOHAND;
+}
+#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
+
 void __init signals_init(void)
 {
 	sigqueue_cachep =

-- 
dwmw2

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[Mika Penttilä]

David Woodhouse wrote:

>The TIF_RESTORE_SIGMASK flag allows us to have a generic implementation
>of sys_rt_sigsuspend() instead of duplicating it for each architecture.
>This provides such an implementation and makes arch/powerpc use it.
>
>It also tidies up the ppc32 sys_sigsuspend() to use TIF_RESTORE_SIGMASK.
>
>Signed-off-by: David Woodhouse <dwmw2@infradead.org>
>  
>
> 
>+#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
>+long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize)
>+{
>+	sigset_t saveset, newset;
>+
>+	/* XXX: Don't preclude handling different sized sigset_t's.  */
>+	if (sigsetsize != sizeof(sigset_t))
>+		return -EINVAL;
>+
>+	if (copy_from_user(&newset, unewset, sizeof(newset)))
>+		return -EFAULT;
>+	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
>+
>+	spin_lock_irq(&current->sighand->siglock);
>+	saveset = current->blocked;
>+	current->blocked = newset;
>+	recalc_sigpending();
>+	spin_unlock_irq(&current->sighand->siglock);
>+
>+	current->state = TASK_INTERRUPTIBLE;
>+	schedule();
>+	set_thread_flag(TIF_RESTORE_SIGMASK);
>+	return -ERESTARTNOHAND;
>+}
>+#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
>+
>  
>
You are not setting saved_sigmask here. And shouldn't it return -EINTR?

Thanks,
Mika

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[Mika Penttilä]

David Woodhouse wrote:

>The TIF_RESTORE_SIGMASK flag allows us to have a generic implementation
>of sys_rt_sigsuspend() instead of duplicating it for each architecture.
>This provides such an implementation and makes arch/powerpc use it.
>
>It also tidies up the ppc32 sys_sigsuspend() to use TIF_RESTORE_SIGMASK.
>
>Signed-off-by: David Woodhouse <dwmw2@infradead.org>
>  
>
> 
>+#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
>+long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize)
>+{
>+	sigset_t saveset, newset;
>+
>+	/* XXX: Don't preclude handling different sized sigset_t's.  */
>+	if (sigsetsize != sizeof(sigset_t))
>+		return -EINVAL;
>+
>+	if (copy_from_user(&newset, unewset, sizeof(newset)))
>+		return -EFAULT;
>+	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
>+
>+	spin_lock_irq(&current->sighand->siglock);
>+	saveset = current->blocked;
>+	current->blocked = newset;
>+	recalc_sigpending();
>+	spin_unlock_irq(&current->sighand->siglock);
>+
>+	current->state = TASK_INTERRUPTIBLE;
>+	schedule();
>+	set_thread_flag(TIF_RESTORE_SIGMASK);
>+	return -ERESTARTNOHAND;
>+}
>+#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
>+
>  
>
You are not setting saved_sigmask here. And shouldn't it return -EINTR?

Thanks,
Mika

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[David Woodhouse]

On Tue, 2005-11-29 at 07:18 +0200, Mika Penttilä wrote:
> You are not setting saved_sigmask here. 

Oops; well spotted. Thanks.

> And shouldn't it return -EINTR?

I believe not. The previous versions would loop until do_signal()
returned non-zero; i.e. until a signal was actually delivered.
By returning -ERESTARTNOHAND we achieve the same effect. If there's a
signal delivered, that gets magically converted to -EINTR, but if
there's no signal delivered, the syscall gets restarted.

diff -u b/kernel/compat.c b/kernel/compat.c
--- b/kernel/compat.c
+++ b/kernel/compat.c
@@ -859,7 +859,7 @@
 #ifdef __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND
 long compat_sys_rt_sigsuspend(compat_sigset_t __user *unewset, compat_size_t sigsetsize)
 {
-	sigset_t saveset, newset;
+	sigset_t newset;
 	compat_sigset_t newset32;
 
 	/* XXX: Don't preclude handling different sized sigset_t's.  */
@@ -872,7 +872,7 @@
 	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
 
 	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
+	current->saved_sigmask = current->blocked;
 	current->blocked = newset;
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);
diff -u b/kernel/signal.c b/kernel/signal.c
--- b/kernel/signal.c
+++ b/kernel/signal.c
@@ -2626,7 +2626,7 @@
 #ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
 long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize)
 {
-	sigset_t saveset, newset;
+	sigset_t newset;
 
 	/* XXX: Don't preclude handling different sized sigset_t's.  */
 	if (sigsetsize != sizeof(sigset_t))
@@ -2637,7 +2637,7 @@
 	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
 
 	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
+	current->saved_sigmask = current->blocked;
 	current->blocked = newset;
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);


-- 
dwmw2

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[David Woodhouse]

On Tue, 2005-11-29 at 07:18 +0200, Mika Penttilä wrote:
> You are not setting saved_sigmask here. 

Oops; well spotted. Thanks.

> And shouldn't it return -EINTR?

I believe not. The previous versions would loop until do_signal()
returned non-zero; i.e. until a signal was actually delivered.
By returning -ERESTARTNOHAND we achieve the same effect. If there's a
signal delivered, that gets magically converted to -EINTR, but if
there's no signal delivered, the syscall gets restarted.

diff -u b/kernel/compat.c b/kernel/compat.c
--- b/kernel/compat.c
+++ b/kernel/compat.c
@@ -859,7 +859,7 @@
 #ifdef __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND
 long compat_sys_rt_sigsuspend(compat_sigset_t __user *unewset, compat_size_t sigsetsize)
 {
-	sigset_t saveset, newset;
+	sigset_t newset;
 	compat_sigset_t newset32;
 
 	/* XXX: Don't preclude handling different sized sigset_t's.  */
@@ -872,7 +872,7 @@
 	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
 
 	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
+	current->saved_sigmask = current->blocked;
 	current->blocked = newset;
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);
diff -u b/kernel/signal.c b/kernel/signal.c
--- b/kernel/signal.c
+++ b/kernel/signal.c
@@ -2626,7 +2626,7 @@
 #ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
 long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize)
 {
-	sigset_t saveset, newset;
+	sigset_t newset;
 
 	/* XXX: Don't preclude handling different sized sigset_t's.  */
 	if (sigsetsize != sizeof(sigset_t))
@@ -2637,7 +2637,7 @@
 	sigdelsetmask(&newset, sigmask(SIGKILL)|sigmask(SIGSTOP));
 
 	spin_lock_irq(&current->sighand->siglock);
-	saveset = current->blocked;
+	current->saved_sigmask = current->blocked;
 	current->blocked = newset;
 	recalc_sigpending();
 	spin_unlock_irq(&current->sighand->siglock);


-- 
dwmw2

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[Daniel Jacobowitz]

On Tue, Nov 29, 2005 at 10:42:03AM +0000, David Woodhouse wrote:
> I believe not. The previous versions would loop until do_signal()
> returned non-zero; i.e. until a signal was actually delivered.
> By returning -ERESTARTNOHAND we achieve the same effect. If there's a
> signal delivered, that gets magically converted to -EINTR, but if
> there's no signal delivered, the syscall gets restarted.

And, crazy coincidence, I think this will fix the recently reported
ptrace attach bug.  Right now if you ptrace a process stuck in
sigsuspend, you can't easily force it to return to userspace.
I'll test that if these patches are merged.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[Daniel Jacobowitz]

On Tue, Nov 29, 2005 at 10:42:03AM +0000, David Woodhouse wrote:
> I believe not. The previous versions would loop until do_signal()
> returned non-zero; i.e. until a signal was actually delivered.
> By returning -ERESTARTNOHAND we achieve the same effect. If there's a
> signal delivered, that gets magically converted to -EINTR, but if
> there's no signal delivered, the syscall gets restarted.

And, crazy coincidence, I think this will fix the recently reported
ptrace attach bug.  Right now if you ptrace a process stuck in
sigsuspend, you can't easily force it to return to userspace.
I'll test that if these patches are merged.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[David Woodhouse]

On Tue, 2005-11-29 at 10:53 -0500, Daniel Jacobowitz wrote:
> And, crazy coincidence, I think this will fix the recently reported
> ptrace attach bug.  Right now if you ptrace a process stuck in
> sigsuspend, you can't easily force it to return to userspace.
> I'll test that if these patches are merged.

That seems to be true. What I get with my patches is...

# strace -p `pidof sigsusptest`
Process 1954 attached - interrupt to quit
rt_sigsuspend([])                       = ? ERESTARTNOHAND (To be restarted)
--- SIGALRM (Alarm clock) @ 0 (0) ---
sigreturn()                             = ? (mask now [])
fstat64(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(4, 64), ...}) = 0
ioctl(1, TCGETS, {B115200 opost isig icanon echo ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x3001f000
write(1, "r is -1\n", 8)                = 8
munmap(0x3001f000, 4096)                = 0
exit_group(8)                           = ?
Process 1954 detached

... whereas without them I get not only a failure to attach, until
there's a signal, but an unexplained SIGSEGV too...

# strace -p `pidof sigsusptest`
Process 3105 attached - interrupt to quit
--- SIGALRM (Alarm clock) @ 0 (0) ---
rt_sigsuspend([])                       = 14
rt_sigsuspend([] <unfinished ...>
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
Process 3105 detached

-- 
dwmw2

Re: [PATCH] 3/3 Generic sys_rt_sigsuspend

[David Woodhouse]

On Tue, 2005-11-29 at 10:53 -0500, Daniel Jacobowitz wrote:
> And, crazy coincidence, I think this will fix the recently reported
> ptrace attach bug.  Right now if you ptrace a process stuck in
> sigsuspend, you can't easily force it to return to userspace.
> I'll test that if these patches are merged.

That seems to be true. What I get with my patches is...

# strace -p `pidof sigsusptest`
Process 1954 attached - interrupt to quit
rt_sigsuspend([])                       = ? ERESTARTNOHAND (To be restarted)
--- SIGALRM (Alarm clock) @ 0 (0) ---
sigreturn()                             = ? (mask now [])
fstat64(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(4, 64), ...}) = 0
ioctl(1, TCGETS, {B115200 opost isig icanon echo ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x3001f000
write(1, "r is -1\n", 8)                = 8
munmap(0x3001f000, 4096)                = 0
exit_group(8)                           = ?
Process 1954 detached

... whereas without them I get not only a failure to attach, until
there's a signal, but an unexplained SIGSEGV too...

# strace -p `pidof sigsusptest`
Process 3105 attached - interrupt to quit
--- SIGALRM (Alarm clock) @ 0 (0) ---
rt_sigsuspend([])                       = 14
rt_sigsuspend([] <unfinished ...>
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
Process 3105 detached

-- 
dwmw2