Re: [PATCH] Fix the use of sigaltstack to return to the saved main stack.

[James Y Knight <jyknight@google.com>]

[-- Attachment #1.1: Type: text/plain, Size: 848 bytes --]

Updated patches attached.

W.r.t.:
> Here, "set to" is
> probably something the resolution of Austin Group issue 1187 failed to
> fix; it should probably be "includes" rather than "is set to". But I'm
> not sure it makes sense to have any flags set alongside SS_DISABLE
> anyway.

While the SS_AUTODISARM flag has no effect if specified alongside
SS_DISABLE, the kernel still accepts and stores it. So A subsequent call to
sigaltstack can return SS_DISABLE|SS_AUTODISARM in the "old" flags value.
To avoid the case where the old value returned from sigaltstack is not
accepted back as the input, I used the "includes" semantics here.


On Wed, Jul 10, 2019 at 2:39 PM Rich Felker <dalias@libc.org> wrote:

> On Wed, Jul 10, 2019 at 02:04:18PM -0400, James Y Knight wrote:
> > On Tue, Jul 9, 2019 at 3:30 PM Rich Felker <dalias@libc.org> wrote:
>
>

[-- Attachment #1.2: Type: text/html, Size: 1371 bytes --]

[-- Attachment #2: 0001-Fix-the-use-of-sigaltstack-to-return-to-the-saved-ma.patch --]
[-- Type: text/x-patch, Size: 1347 bytes --]

From f6a83b477b5b1b7bd1706327f9d6de19b6efe386 Mon Sep 17 00:00:00 2001
From: James Y Knight <jyknight@google.com>
Date: Wed, 10 Jul 2019 15:22:58 -0400
Subject: [PATCH] Fix the use of sigaltstack to return to the saved main stack.

Previously, musl would reject the call with -ENOMEM, because the main
stack typically has ss_size == 0 and ss_flags == SS_DISABLE.
---
 src/signal/sigaltstack.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/signal/sigaltstack.c b/src/signal/sigaltstack.c
index cfa3f5c1..b0d799d2 100644
--- a/src/signal/sigaltstack.c
+++ b/src/signal/sigaltstack.c
@@ -4,11 +4,19 @@
 
 int sigaltstack(const stack_t *restrict ss, stack_t *restrict old)
 {
+	// We must check requirements which Linux fails to verify in the syscall
+	// itself.
 	if (ss) {
-		if (ss->ss_size < MINSIGSTKSZ) {
+		// The syscall does already check against MINSIGSTKSZ, however,
+		// the kernel's value is smaller than musl's value on some
+		// architectures. Thus, although this check may appear
+		// redundant, it is not.
+		if (!(ss->ss_flags & SS_DISABLE) && ss->ss_size < MINSIGSTKSZ) {
 			errno = ENOMEM;
 			return -1;
 		}
+		// Linux ignores SS_ONSTACK on input, but POSIX requires an
+		// error.
 		if (ss->ss_flags & SS_ONSTACK) {
 			errno = EINVAL;
 			return -1;
-- 
2.22.0.410.gd8fdbe21b5-goog


[-- Attachment #3: 0001-Verify-that-returning-to-the-original-stack-doesn-t-.patch --]
[-- Type: text/x-patch, Size: 1166 bytes --]

From 313618c8e00df106528e24cabc38379b4783f51e Mon Sep 17 00:00:00 2001
From: James Y Knight <jyknight@google.com>
Date: Tue, 9 Jul 2019 14:31:24 -0400
Subject: [PATCH] Verify that returning to the original stack doesn't return an
 error (e.g. ENOMEM).

---
 src/regression/sigaltstack.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/regression/sigaltstack.c b/src/regression/sigaltstack.c
index bfdc44a..2bfe329 100644
--- a/src/regression/sigaltstack.c
+++ b/src/regression/sigaltstack.c
@@ -30,7 +30,7 @@ static void handler(int sig)
 
 int main(void)
 {
-	stack_t ss;
+	stack_t ss, oldss;
 	struct sigaction sa;
 
 	ss.ss_sp = stack;
@@ -39,7 +39,7 @@ int main(void)
 	sa.sa_handler = handler;
 	sa.sa_flags = SA_ONSTACK;
 
-	T(sigaltstack(&ss, 0));
+	T(sigaltstack(&ss, &oldss));
 	T(sigfillset(&sa.sa_mask));
 	T(sigaction(SIGUSR1, &sa, 0));
 	T(raise(SIGUSR1));
@@ -56,7 +56,7 @@ int main(void)
 		t_error("sigaltstack with bad ss_flags should have failed with EINVAL, "
 			"got %s\n", strerror(errno));
 	errno = 0;
-	T(sigaltstack(0, 0));
+	T(sigaltstack(&oldss, 0));
 
 	return t_status;
 }
-- 
2.22.0.410.gd8fdbe21b5-goog