linux-kernel - [PATCH entry v2 2/6] x86/entry, selftests: Further improve user entry sanity checks

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 for Android: free password hash cracker in your pocket

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

Message-Id: <881de09e786ab93ce56ee4a2437ba2c308afe7a9.1593795633.git.luto@kernel.org>
Date:   Fri,  3 Jul 2020 10:02:54 -0700
From:   Andy Lutomirski <luto@...nel.org>
To:     x86@...nel.org
Cc:     Andrew Cooper <andrew.cooper3@...rix.com>,
        Juergen Gross <jgross@...e.com>,
        LKML <linux-kernel@...r.kernel.org>,
        Andy Lutomirski <luto@...nel.org>
Subject: [PATCH entry v2 2/6] x86/entry, selftests: Further improve user entry sanity checks

Chasing down a Xen bug caused me to realize that the new entry sanity
checks are still fairly weak.  Add some more checks.

Signed-off-by: Andy Lutomirski <luto@...nel.org>
---
 arch/x86/entry/common.c                  | 19 +++++++++++++++++++
 tools/testing/selftests/x86/syscall_nt.c | 11 +++++++++++
 2 files changed, 30 insertions(+)

diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index f392a8bcd1c3..e83b3f14897c 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -49,6 +49,23 @@
 static void check_user_regs(struct pt_regs *regs)
 {
 	if (IS_ENABLED(CONFIG_DEBUG_ENTRY)) {
+		/*
+		 * Make sure that the entry code gave us a sensible EFLAGS
+		 * register.  Native because we want to check the actual CPU
+		 * state, not the interrupt state as imagined by Xen.
+		 */
+		unsigned long flags = native_save_fl();
+		WARN_ON_ONCE(flags & (X86_EFLAGS_AC | X86_EFLAGS_DF |
+				      X86_EFLAGS_NT));
+
+		/* We think we came from user mode. Make sure pt_regs agrees. */
+		WARN_ON_ONCE(!user_mode(regs));
+
+		/*
+		 * All entries from user mode (except #DF) should be on the
+		 * normal thread stack and should have user pt_regs in the
+		 * correct location.
+		 */
 		WARN_ON_ONCE(!on_thread_stack());
 		WARN_ON_ONCE(regs != task_pt_regs(current));
 	}
@@ -577,6 +594,7 @@ SYSCALL_DEFINE0(ni_syscall)
 bool noinstr idtentry_enter_cond_rcu(struct pt_regs *regs)
 {
 	if (user_mode(regs)) {
+		check_user_regs(regs);
 		enter_from_user_mode();
 		return false;
 	}
@@ -710,6 +728,7 @@ void noinstr idtentry_exit_cond_rcu(struct pt_regs *regs, bool rcu_exit)
  */
 void noinstr idtentry_enter_user(struct pt_regs *regs)
 {
+	check_user_regs(regs);
 	enter_from_user_mode();
 }
 
diff --git a/tools/testing/selftests/x86/syscall_nt.c b/tools/testing/selftests/x86/syscall_nt.c
index 970e5e14d96d..a108b80dd082 100644
--- a/tools/testing/selftests/x86/syscall_nt.c
+++ b/tools/testing/selftests/x86/syscall_nt.c
@@ -81,5 +81,16 @@ int main(void)
 	printf("[RUN]\tSet NT|AC|TF and issue a syscall\n");
 	do_it(X86_EFLAGS_NT | X86_EFLAGS_AC | X86_EFLAGS_TF);
 
+	/*
+	 * Now try DF.  This is evil and it's plausible that we will crash
+	 * glibc, but glibc would have to do something rather surprising
+	 * for this to happen.
+	 */
+	printf("[RUN]\tSet DF and issue a syscall\n");
+	do_it(X86_EFLAGS_DF);
+
+	printf("[RUN]\tSet TF|DF and issue a syscall\n");
+	do_it(X86_EFLAGS_TF | X86_EFLAGS_DF);
+
 	return nerrs == 0 ? 0 : 1;
 }
-- 
2.25.4