remove rlimit hacks from multi-threaded set*id() code

the code being removed was introduced to work around "partial failure"
of multi-threaded set*id() operations, where some threads would
succeed in changing their ids but an RLIMIT_NPROC setting would
prevent the rest from succeeding, leaving the process in an
inconsistent and dangerous state. however, the workaround code did not
handle important usage cases like swapping real and effective uids
then restoring their original values, and the wrongful kernel
enforcement of RLIMIT_NPROC at setuid time was removed in Linux 3.1,
making the workaround obsolete.

since the partial failure still is dangerous on old kernels, and could
in principle happen on post-fix kernels as well if set*id() syscalls
fail for another spurious reason such as resource-related failures,
new code is added to detect and forcibly kill the process if/when such
a situation arises. future documentation releases should be updated to
reflect that setting RLIMIT_NPROC to RLIM_INFINITY is necessary to
avoid this forced-kill on old kernels. ideally, at some point the
kernel will get proper multi-threaded set*id() syscalls capable of
performing their actions atomically, and all of the userspace code to
emulate them can be treated as a fallback for outdated kernels.

1 files changed, 15 insertions, 23 deletions

diff --git a/src/unistd/setxid.c b/src/unistd/setxid.c
index 2f651a11..9e37ddc4 100644
--- a/src/unistd/setxid.c
+++ b/src/unistd/setxid.c
@@ -1,43 +1,35 @@
 #include <unistd.h>
 #include <errno.h>
-#include <sys/resource.h>
 #include "syscall.h"
 #include "libc.h"
+#include "pthread_impl.h"
 
 struct ctx {
 	int id, eid, sid;
-	int nr, rlim, err;
+	int nr, err;
 };
 
-/* We jump through hoops to eliminate the possibility of partial failures. */
-
-int __setrlimit(int, const struct rlimit *);
-
 static void do_setxid(void *p)
 {
 	struct ctx *c = p;
-	if (c->err) return;
-	if (c->rlim && c->id >= 0 && c->id != getuid()) {
-		struct rlimit inf = { RLIM_INFINITY, RLIM_INFINITY }, old;
-		getrlimit(RLIMIT_NPROC, &old);
-		if ((c->err = -__setrlimit(RLIMIT_NPROC, &inf)) && libc.threads_minus_1)
-			return;
-		c->err = -__syscall(c->nr, c->id, c->eid, c->sid);
-		__setrlimit(RLIMIT_NPROC, &old);
-		return;
+	if (c->err>0) return;
+	int ret = -__syscall(c->nr, c->id, c->eid, c->sid);
+	if (ret && !c->err) {
+		/* If one thread fails to set ids after another has already
+		 * succeeded, forcibly killing the process is the only safe
+		 * thing to do. State is inconsistent and dangerous. Use
+		 * SIGKILL because it is uncatchable. */
+		__block_all_sigs(0);
+		__syscall(SYS_kill, __syscall(SYS_getpid), SIGKILL);
 	}
-	c->err = -__syscall(c->nr, c->id, c->eid, c->sid);
+	c->err = ret;
 }
 
 int __setxid(int nr, int id, int eid, int sid)
 {
-	struct ctx c = { .nr = nr, .id = id, .eid = eid, .sid = sid };
-	switch (nr) {
-	case SYS_setuid:
-	case SYS_setreuid:
-	case SYS_setresuid:
-		c.rlim = 1;
-	}
+	/* err is initially nonzero so that failure of the first thread does not
+	 * trigger the safety kill above. */
+	struct ctx c = { .nr = nr, .id = id, .eid = eid, .sid = sid, .err = -1 };
 	__synccall(do_setxid, &c);
 	if (c.err) {
 		errno = c.err;