From 42a0cc3478584d4d63f68f2f5af021ddbea771fa Mon Sep 17 00:00:00 2001
From: Jann Horn <jannh@google.com>
Date: Mon, 25 Jun 2018 18:34:10 +0200
Subject: [PATCH] sys: don't hold uts_sem while accessing userspace memory
Git-commit: 42a0cc3478584d4d63f68f2f5af021ddbea771fa
Patch-mainline: v4.19-rc1
References: bnc#1106995

Mhocko@suse.com: 
dropped alpha and sparc parts as we do not build them and the patch doesn't
apply cleanly on them.

Holding uts_sem as a writer while accessing userspace memory allows a
namespace admin to stall all processes that attempt to take uts_sem.
Instead, move data through stack buffers and don't access userspace memory
while uts_sem is held.

Cc: stable@vger.kernel.org
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Michal Hocko <mhocko@suse.com>

---
 kernel/sys.c            |   95 ++++++++++++++++++++++--------------------------
 kernel/utsname_sysctl.c |   41 ++++++++++++--------
 2 files changed, 70 insertions(+), 66 deletions(-)

--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -1142,18 +1142,19 @@ static int override_release(char __user
 
 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
 {
-	int errno = 0;
+	struct new_utsname tmp;
 
 	down_read(&uts_sem);
-	if (copy_to_user(name, utsname(), sizeof *name))
-		errno = -EFAULT;
+	memcpy(&tmp, utsname(), sizeof(tmp));
 	up_read(&uts_sem);
+	if (copy_to_user(name, &tmp, sizeof(tmp)))
+		return -EFAULT;
 
-	if (!errno && override_release(name->release, sizeof(name->release)))
-		errno = -EFAULT;
-	if (!errno && override_architecture(name))
-		errno = -EFAULT;
-	return errno;
+	if (override_release(name->release, sizeof(name->release)))
+		return -EFAULT;
+	if (override_architecture(name))
+		return -EFAULT;
+	return 0;
 }
 
 #ifdef __ARCH_WANT_SYS_OLD_UNAME
@@ -1162,55 +1163,46 @@ SYSCALL_DEFINE1(newuname, struct new_uts
  */
 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
 {
-	int error = 0;
+	struct old_utsname tmp;
 
 	if (!name)
 		return -EFAULT;
 
 	down_read(&uts_sem);
-	if (copy_to_user(name, utsname(), sizeof(*name)))
-		error = -EFAULT;
+	memcpy(&tmp, utsname(), sizeof(tmp));
 	up_read(&uts_sem);
+	if (copy_to_user(name, &tmp, sizeof(tmp)))
+		return -EFAULT;
 
-	if (!error && override_release(name->release, sizeof(name->release)))
-		error = -EFAULT;
-	if (!error && override_architecture(name))
-		error = -EFAULT;
-	return error;
+	if (override_release(name->release, sizeof(name->release)))
+		return -EFAULT;
+	if (override_architecture(name))
+		return -EFAULT;
+	return 0;
 }
 
 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
 {
-	int error;
+	struct oldold_utsname tmp = {};
 
 	if (!name)
 		return -EFAULT;
-	if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
-		return -EFAULT;
 
 	down_read(&uts_sem);
-	error = __copy_to_user(&name->sysname, &utsname()->sysname,
-			       __OLD_UTS_LEN);
-	error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
-	error |= __copy_to_user(&name->nodename, &utsname()->nodename,
-				__OLD_UTS_LEN);
-	error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
-	error |= __copy_to_user(&name->release, &utsname()->release,
-				__OLD_UTS_LEN);
-	error |= __put_user(0, name->release + __OLD_UTS_LEN);
-	error |= __copy_to_user(&name->version, &utsname()->version,
-				__OLD_UTS_LEN);
-	error |= __put_user(0, name->version + __OLD_UTS_LEN);
-	error |= __copy_to_user(&name->machine, &utsname()->machine,
-				__OLD_UTS_LEN);
-	error |= __put_user(0, name->machine + __OLD_UTS_LEN);
+	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
+	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
+	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
+	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
+	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
 	up_read(&uts_sem);
+	if (copy_to_user(name, &tmp, sizeof(tmp)))
+		return -EFAULT;
 
-	if (!error && override_architecture(name))
-		error = -EFAULT;
-	if (!error && override_release(name->release, sizeof(name->release)))
-		error = -EFAULT;
-	return error ? -EFAULT : 0;
+	if (override_architecture(name))
+		return -EFAULT;
+	if (override_release(name->release, sizeof(name->release)))
+		return -EFAULT;
+	return 0;
 }
 #endif
 
@@ -1224,17 +1216,18 @@ SYSCALL_DEFINE2(sethostname, char __user
 
 	if (len < 0 || len > __NEW_UTS_LEN)
 		return -EINVAL;
-	down_write(&uts_sem);
 	errno = -EFAULT;
 	if (!copy_from_user(tmp, name, len)) {
-		struct new_utsname *u = utsname();
+		struct new_utsname *u;
 
+		down_write(&uts_sem);
+		u = utsname();
 		memcpy(u->nodename, tmp, len);
 		memset(u->nodename + len, 0, sizeof(u->nodename) - len);
 		errno = 0;
 		uts_proc_notify(UTS_PROC_HOSTNAME);
+		up_write(&uts_sem);
 	}
-	up_write(&uts_sem);
 	return errno;
 }
 
@@ -1242,8 +1235,9 @@ SYSCALL_DEFINE2(sethostname, char __user
 
 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
 {
-	int i, errno;
+	int i;
 	struct new_utsname *u;
+	char tmp[__NEW_UTS_LEN + 1];
 
 	if (len < 0)
 		return -EINVAL;
@@ -1252,11 +1246,11 @@ SYSCALL_DEFINE2(gethostname, char __user
 	i = 1 + strlen(u->nodename);
 	if (i > len)
 		i = len;
-	errno = 0;
-	if (copy_to_user(name, u->nodename, i))
-		errno = -EFAULT;
+	memcpy(tmp, u->nodename, i);
 	up_read(&uts_sem);
-	return errno;
+	if (copy_to_user(name, tmp, i))
+		return -EFAULT;
+	return 0;
 }
 
 #endif
@@ -1275,17 +1269,18 @@ SYSCALL_DEFINE2(setdomainname, char __us
 	if (len < 0 || len > __NEW_UTS_LEN)
 		return -EINVAL;
 
-	down_write(&uts_sem);
 	errno = -EFAULT;
 	if (!copy_from_user(tmp, name, len)) {
-		struct new_utsname *u = utsname();
+		struct new_utsname *u;
 
+		down_write(&uts_sem);
+		u = utsname();
 		memcpy(u->domainname, tmp, len);
 		memset(u->domainname + len, 0, sizeof(u->domainname) - len);
 		errno = 0;
 		uts_proc_notify(UTS_PROC_DOMAINNAME);
+		up_write(&uts_sem);
 	}
-	up_write(&uts_sem);
 	return errno;
 }
 
--- a/kernel/utsname_sysctl.c
+++ b/kernel/utsname_sysctl.c
@@ -17,7 +17,7 @@
 
 #ifdef CONFIG_PROC_SYSCTL
 
-static void *get_uts(struct ctl_table *table, int write)
+static void *get_uts(struct ctl_table *table)
 {
 	char *which = table->data;
 	struct uts_namespace *uts_ns;
@@ -25,21 +25,9 @@ static void *get_uts(struct ctl_table *t
 	uts_ns = current->nsproxy->uts_ns;
 	which = (which - (char *)&init_uts_ns) + (char *)uts_ns;
 
-	if (!write)
-		down_read(&uts_sem);
-	else
-		down_write(&uts_sem);
 	return which;
 }
 
-static void put_uts(struct ctl_table *table, int write, void *which)
-{
-	if (!write)
-		up_read(&uts_sem);
-	else
-		up_write(&uts_sem);
-}
-
 /*
  *	Special case of dostring for the UTS structure. This has locks
  *	to observe. Should this be in kernel/sys.c ????
@@ -49,13 +37,34 @@ static int proc_do_uts_string(struct ctl
 {
 	struct ctl_table uts_table;
 	int r;
+	char tmp_data[__NEW_UTS_LEN + 1];
+
 	memcpy(&uts_table, table, sizeof(uts_table));
-	uts_table.data = get_uts(table, write);
+	uts_table.data = tmp_data;
+
+	/*
+	 * Buffer the value in tmp_data so that proc_dostring() can be called
+	 * without holding any locks.
+	 * We also need to read the original value in the write==1 case to
+	 * support partial writes.
+	 */
+	down_read(&uts_sem);
+	memcpy(tmp_data, get_uts(table), sizeof(tmp_data));
+	up_read(&uts_sem);
 	r = proc_dostring(&uts_table, write, buffer, lenp, ppos);
-	put_uts(table, write, uts_table.data);
 
-	if (write)
+	if (write) {
+		/*
+		 * Write back the new value.
+		 * Note that, since we dropped uts_sem, the result can
+		 * theoretically be incorrect if there are two parallel writes
+		 * at non-zero offsets to the same sysctl.
+		 */
+		down_write(&uts_sem);
+		memcpy(get_uts(table), tmp_data, sizeof(tmp_data));
+		up_write(&uts_sem);
 		proc_sys_poll_notify(table->poll);
+	}
 
 	return r;
 }