[tip:x86/asm] x86/vdso: Disallow vvar access to vclock IO for never-used vclocks

[tip-bot for Andy Lutomirski]

Commit-ID:  bd902c536298830e4d126dcf6491b46d3f1bf96e
Gitweb:     http://git.kernel.org/tip/bd902c536298830e4d126dcf6491b46d3f1bf96e
Author:     Andy Lutomirski <luto@xxxxxxxxxx>
AuthorDate: Tue, 29 Dec 2015 20:12:24 -0800
Committer:  Ingo Molnar <mingo@xxxxxxxxxx>
CommitDate: Tue, 12 Jan 2016 11:59:35 +0100

x86/vdso: Disallow vvar access to vclock IO for never-used vclocks

It makes me uncomfortable that even modern systems grant every
process direct read access to the HPET.

While fixing this for real without regressing anything is a mess
(unmapping the HPET is tricky because we don't adequately track
all the mappings), we can do almost as well by tracking which
vclocks have ever been used and only allowing pages associated
with used vclocks to be faulted in.

This will cause rogue programs that try to peek at the HPET to
get SIGBUS instead on most systems.

We can't restrict faults to vclock pages that are associated
with the currently selected vclock due to a race: a process
could start to access the HPET for the first time and race
against a switch away from the HPET as the current clocksource.
We can't segfault the process trying to peek at the HPET in this
case, even though the process isn't going to do anything useful
with the data.

Signed-off-by: Andy Lutomirski <luto@xxxxxxxxxx>
Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>
Cc: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxxxx>
Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
Cc: Fenghua Yu <fenghua.yu@xxxxxxxxx>
Cc: H. Peter Anvin <hpa@xxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Oleg Nesterov <oleg@xxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Quentin Casasnovas <quentin.casasnovas@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Link: http://lkml.kernel.org/r/e79d06295625c02512277737ab55085a498ac5d8.1451446564.git.luto@xxxxxxxxxx
Signed-off-by: Ingo Molnar <mingo@xxxxxxxxxx>
---
 arch/x86/entry/vdso/vma.c               | 4 ++--
 arch/x86/entry/vsyscall/vsyscall_gtod.c | 9 ++++++++-
 arch/x86/include/asm/clocksource.h      | 9 +++++----
 arch/x86/include/asm/vgtod.h            | 6 ++++++
 4 files changed, 21 insertions(+), 7 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 4b5461b..7c912fe 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -130,7 +130,7 @@ static int vvar_fault(const struct vm_special_mapping *sm,
 				    __pa_symbol(&__vvar_page) >> PAGE_SHIFT);
 	} else if (sym_offset == image->sym_hpet_page) {
 #ifdef CONFIG_HPET_TIMER
-		if (hpet_address) {
+		if (hpet_address && vclock_was_used(VCLOCK_HPET)) {
 			ret = vm_insert_pfn_prot(
 				vma,
 				(unsigned long)vmf->virtual_address,
@@ -141,7 +141,7 @@ static int vvar_fault(const struct vm_special_mapping *sm,
 	} else if (sym_offset == image->sym_pvclock_page) {
 		struct pvclock_vsyscall_time_info *pvti =
 			pvclock_pvti_cpu0_va();
-		if (pvti) {
+		if (pvti && vclock_was_used(VCLOCK_PVCLOCK)) {
 			ret = vm_insert_pfn(
 				vma,
 				(unsigned long)vmf->virtual_address,
diff --git a/arch/x86/entry/vsyscall/vsyscall_gtod.c b/arch/x86/entry/vsyscall/vsyscall_gtod.c
index 51e3304..0fb3a10 100644
--- a/arch/x86/entry/vsyscall/vsyscall_gtod.c
+++ b/arch/x86/entry/vsyscall/vsyscall_gtod.c
@@ -16,6 +16,8 @@
 #include <asm/vgtod.h>
 #include <asm/vvar.h>
 
+int vclocks_used __read_mostly;
+
 DEFINE_VVAR(struct vsyscall_gtod_data, vsyscall_gtod_data);
 
 void update_vsyscall_tz(void)
@@ -26,12 +28,17 @@ void update_vsyscall_tz(void)
 
 void update_vsyscall(struct timekeeper *tk)
 {
+	int vclock_mode = tk->tkr_mono.clock->archdata.vclock_mode;
 	struct vsyscall_gtod_data *vdata = &vsyscall_gtod_data;
 
+	/* Mark the new vclock used. */
+	BUILD_BUG_ON(VCLOCK_MAX >= 32);
+	WRITE_ONCE(vclocks_used, READ_ONCE(vclocks_used) | (1 << vclock_mode));
+
 	gtod_write_begin(vdata);
 
 	/* copy vsyscall data */
-	vdata->vclock_mode	= tk->tkr_mono.clock->archdata.vclock_mode;
+	vdata->vclock_mode	= vclock_mode;
 	vdata->cycle_last	= tk->tkr_mono.cycle_last;
 	vdata->mask		= tk->tkr_mono.mask;
 	vdata->mult		= tk->tkr_mono.mult;
diff --git a/arch/x86/include/asm/clocksource.h b/arch/x86/include/asm/clocksource.h
index eda81dc..d194266 100644
--- a/arch/x86/include/asm/clocksource.h
+++ b/arch/x86/include/asm/clocksource.h
@@ -3,10 +3,11 @@
 #ifndef _ASM_X86_CLOCKSOURCE_H
 #define _ASM_X86_CLOCKSOURCE_H
 
-#define VCLOCK_NONE 0  /* No vDSO clock available.	*/
-#define VCLOCK_TSC  1  /* vDSO should use vread_tsc.	*/
-#define VCLOCK_HPET 2  /* vDSO should use vread_hpet.	*/
-#define VCLOCK_PVCLOCK 3 /* vDSO should use vread_pvclock. */
+#define VCLOCK_NONE	0  /* No vDSO clock available.	*/
+#define VCLOCK_TSC	1  /* vDSO should use vread_tsc.	*/
+#define VCLOCK_HPET	2  /* vDSO should use vread_hpet.	*/
+#define VCLOCK_PVCLOCK	3 /* vDSO should use vread_pvclock. */
+#define VCLOCK_MAX	3
 
 struct arch_clocksource_data {
 	int vclock_mode;
diff --git a/arch/x86/include/asm/vgtod.h b/arch/x86/include/asm/vgtod.h
index f556c48..e728699 100644
--- a/arch/x86/include/asm/vgtod.h
+++ b/arch/x86/include/asm/vgtod.h
@@ -37,6 +37,12 @@ struct vsyscall_gtod_data {
 };
 extern struct vsyscall_gtod_data vsyscall_gtod_data;
 
+extern int vclocks_used;
+static inline bool vclock_was_used(int vclock)
+{
+	return READ_ONCE(vclocks_used) & (1 << vclock);
+}
+
 static inline unsigned gtod_read_begin(const struct vsyscall_gtod_data *s)
 {
 	unsigned ret;