signal: Use a smaller struct siginfo in the kernel

We reserve 128 bytes for struct siginfo but only use about 48 bytes on
64bit and 32 bytes on 32bit.  Someday we might use more but it is unlikely
to be anytime soon.

Userspace seems content with just enough bytes of siginfo to implement
sigqueue.  Or in the case of checkpoint/restart reinjecting signals
the kernel has sent.

Reducing the stack footprint and the work to copy siginfo around from
2 cachelines to 1 cachelines seems worth doing even if I don't have
benchmarks to show a performance difference.

Suggested-by: Linus Torvalds <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>

Author: ebiederm

include/linux/signal.h

@@ -22,6 +22,8 @@ static inline void clear_siginfo(kernel_siginfo_t *info)
 	memset(info, 0, sizeof(*info));
 }
 
+#define SI_EXPANSION_SIZE (sizeof(struct siginfo) - sizeof(struct kernel_siginfo))
+
 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from);
 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from);
 

include/linux/signal_types.h

@@ -10,10 +10,7 @@
 #include <uapi/linux/signal.h>
 
 typedef struct kernel_siginfo {
-	union {
-		__SIGINFO;
-		int _si_pad[SI_MAX_SIZE/sizeof(int)];
-	};
+	__SIGINFO;
 } kernel_siginfo_t;
 
 /*

kernel/signal.c

@@ -2844,27 +2844,48 @@ COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
 }
 #endif
 
+static const struct {
+	unsigned char limit, layout;
+} sig_sicodes[] = {
+	[SIGILL]  = { NSIGILL,  SIL_FAULT },
+	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
+	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
+	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
+	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
+#if defined(SIGEMT)
+	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
+#endif
+	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
+	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
+	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
+};
+
+static bool known_siginfo_layout(int sig, int si_code)
+{
+	if (si_code == SI_KERNEL)
+		return true;
+	else if ((si_code > SI_USER)) {
+		if (sig_specific_sicodes(sig)) {
+			if (si_code <= sig_sicodes[sig].limit)
+				return true;
+		}
+		else if (si_code <= NSIGPOLL)
+			return true;
+	}
+	else if (si_code >= SI_DETHREAD)
+		return true;
+	else if (si_code == SI_ASYNCNL)
+		return true;
+	return false;
+}
+
 enum siginfo_layout siginfo_layout(int sig, int si_code)
 {
 	enum siginfo_layout layout = SIL_KILL;
 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
-		static const struct {
-			unsigned char limit, layout;
-		} filter[] = {
-			[SIGILL]  = { NSIGILL,  SIL_FAULT },
-			[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
-			[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
-			[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
-			[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
-#if defined(SIGEMT)
-			[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
-#endif
-			[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
-			[SIGPOLL] = { NSIGPOLL, SIL_POLL },
-			[SIGSYS]  = { NSIGSYS,  SIL_SYS },
-		};
-		if ((sig < ARRAY_SIZE(filter)) && (si_code <= filter[sig].limit)) {
-			layout = filter[sig].layout;
+		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
+		    (si_code <= sig_sicodes[sig].limit)) {
+			layout = sig_sicodes[sig].layout;
 			/* Handle the exceptions */
 			if ((sig == SIGBUS) &&
 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
@@ -2889,17 +2910,42 @@ enum siginfo_layout siginfo_layout(int sig, int si_code)
 	return layout;
 }
 
+static inline char __user *si_expansion(const siginfo_t __user *info)
+{
+	return ((char __user *)info) + sizeof(struct kernel_siginfo);
+}
+
 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
 {
+	char __user *expansion = si_expansion(to);
 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
 		return -EFAULT;
+	if (clear_user(expansion, SI_EXPANSION_SIZE))
+		return -EFAULT;
 	return 0;
 }
 
 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
 {
-	if (copy_from_user(to, from, sizeof(struct siginfo)))
+	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
 		return -EFAULT;
+	if (unlikely(!known_siginfo_layout(to->si_signo, to->si_code))) {
+		char __user *expansion = si_expansion(from);
+		char buf[SI_EXPANSION_SIZE];
+		int i;
+		/*
+		 * An unknown si_code might need more than
+		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
+		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
+		 * will return this data to userspace exactly.
+		 */
+		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
+			return -EFAULT;
+		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
+			if (buf[i] != 0)
+				return -E2BIG;
+		}
+	}
 	return 0;
 }