linux-kernel - [PATCH v2 1/3] entry: Check that syscall entries and syscall exits match

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <fb6f42d81eaaec2c4b3bcb657f0dd7d7bdc136c9.1614104065.git.luto@kernel.org>
Date:   Tue, 23 Feb 2021 10:15:08 -0800
From:   Andy Lutomirski <luto@...nel.org>
To:     x86@...nel.org
Cc:     LKML <linux-kernel@...r.kernel.org>,
        Andy Lutomirski <luto@...nel.org>
Subject: [PATCH v2 1/3] entry: Check that syscall entries and syscall exits match

If arch code calls the wrong kernel entry helpers, syscall entries and
exits can get out of sync.  Add a new field to task_struct to track the
syscall state and validate that it transitions correctly.

Signed-off-by: Andy Lutomirski <luto@...nel.org>
---
 include/linux/entry-common.h | 11 +++++++++++
 include/linux/sched.h        |  1 +
 init/init_task.c             |  9 +++++++++
 kernel/entry/common.c        | 25 ++++++++++++++++++++++++-
 4 files changed, 45 insertions(+), 1 deletion(-)

diff --git a/include/linux/entry-common.h b/include/linux/entry-common.h
index ca86a00abe86..c2463b6b71fe 100644
--- a/include/linux/entry-common.h
+++ b/include/linux/entry-common.h
@@ -60,6 +60,17 @@
 	 _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | _TIF_NOTIFY_SIGNAL |	\
 	 ARCH_EXIT_TO_USER_MODE_WORK)
 
+/*
+ * task_struct::kentry_syscall_state
+ *
+ * kentry_syscall_state is reset to zero on syscall exit.  For efficiency
+ * reasons, if CONFIG_PROVE_LOCKING=n, kentry_syscall_state is permitted
+ * to remain 0 even inside a syscall.
+ */
+#ifdef CONFIG_PROVE_LOCKING
+# define KENTRY_SYSCALL_STATE_IN_SYSCALL		1
+#endif
+
 /**
  * arch_check_user_regs - Architecture specific sanity check for user mode regs
  * @regs:	Pointer to currents pt_regs
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 6e3a5eeec509..691057a3b48d 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1002,6 +1002,7 @@ struct task_struct {
 #endif
 	struct seccomp			seccomp;
 	struct syscall_user_dispatch	syscall_dispatch;
+	u32				kentry_syscall_state;
 
 	/* Thread group tracking: */
 	u64				parent_exec_id;
diff --git a/init/init_task.c b/init/init_task.c
index 8a992d73e6fb..91050c4f0bb3 100644
--- a/init/init_task.c
+++ b/init/init_task.c
@@ -12,6 +12,7 @@
 #include <linux/audit.h>
 #include <linux/numa.h>
 #include <linux/scs.h>
+#include <linux/entry-common.h>
 
 #include <linux/uaccess.h>
 
@@ -212,6 +213,14 @@ struct task_struct init_task
 #ifdef CONFIG_SECCOMP
 	.seccomp	= { .filter_count = ATOMIC_INIT(0) },
 #endif
+#ifdef CONFIG_PROVE_LOCKING
+	/*
+	 * The init task, and kernel threads in general, are considered
+	 * to be "in a syscall".  This way they can execve() and then exit
+	 * the supposed syscall that they were in to go to user mode.
+	 */
+	.kentry_syscall_state = KENTRY_SYSCALL_STATE_IN_SYSCALL,
+#endif
 };
 EXPORT_SYMBOL(init_task);
 
diff --git a/kernel/entry/common.c b/kernel/entry/common.c
index 378341642f94..7e971b866ad2 100644
--- a/kernel/entry/common.c
+++ b/kernel/entry/common.c
@@ -83,7 +83,16 @@ static long syscall_trace_enter(struct pt_regs *regs, long syscall,
 static __always_inline long
 __syscall_enter_from_user_work(struct pt_regs *regs, long syscall)
 {
-	unsigned long work = READ_ONCE(current_thread_info()->syscall_work);
+	unsigned long work;
+
+	if (IS_ENABLED(CONFIG_PROVE_LOCKING)) {
+		WARN_ONCE(current->kentry_syscall_state,
+			  "entering syscall %ld while already in a syscall",
+			  syscall);
+		current->kentry_syscall_state = KENTRY_SYSCALL_STATE_IN_SYSCALL;
+	}
+
+	work = READ_ONCE(current_thread_info()->syscall_work);
 
 	if (work & SYSCALL_WORK_ENTER)
 		syscall = syscall_trace_enter(regs, syscall, work);
@@ -195,6 +204,12 @@ static void exit_to_user_mode_prepare(struct pt_regs *regs)
 {
 	unsigned long ti_work = READ_ONCE(current_thread_info()->flags);
 
+	if (IS_ENABLED(CONFIG_PROVE_LOCKING)) {
+		WARN_ONCE(current->kentry_syscall_state &
+			  KENTRY_SYSCALL_STATE_IN_SYSCALL,
+			  "exiting to user mode while in syscall context");
+	}
+
 	lockdep_assert_irqs_disabled();
 
 	if (unlikely(ti_work & EXIT_TO_USER_MODE_WORK))
@@ -282,6 +297,14 @@ static void syscall_exit_to_user_mode_prepare(struct pt_regs *regs)
 	 */
 	if (unlikely(work & SYSCALL_WORK_EXIT))
 		syscall_exit_work(regs, work);
+
+	if (IS_ENABLED(CONFIG_PROVE_LOCKING)) {
+		WARN_ONCE(!(current->kentry_syscall_state &
+			    KENTRY_SYSCALL_STATE_IN_SYSCALL),
+			  "exiting syscall %lu without entering first", nr);
+	}
+
+	current->kentry_syscall_state = 0;
 }
 
 static __always_inline void __syscall_exit_to_user_mode_work(struct pt_regs *regs)
-- 
2.29.2