perf/x86: Only allow rdpmc if a perf_event is mapped

We currently allow any process to use rdpmc.  This significantly
weakens the protection offered by PR_TSC_DISABLED, and it could be
helpful to users attempting to exploit timing attacks.

Since we can't enable access to individual counters, use a very
coarse heuristic to limit access to rdpmc: allow access only when
a perf_event is mmapped.  This protects seccomp sandboxes.

There is plenty of room to further tighen these restrictions.  For
example, this allows rdpmc for any x86_pmu event, but it's only
useful for self-monitoring tasks.

As a side effect, cap_user_rdpmc will now be false for AMD uncore
events.  This isn't a real regression, since .event_idx is disabled
for these events anyway for the time being.  Whenever that gets
re-added, the cap_user_rdpmc code can be adjusted or refactored
accordingly.
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/a2bdb3cf3a1d70c26980d7c6dddfbaa69f3182bf.1414190806.git.luto@amacapital.netSigned-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/include/asm/mmu.h

@@ -19,6 +19,8 @@ typedef struct {
 
 	struct mutex lock;
 	void __user *vdso;
+
+	atomic_t perf_rdpmc_allowed;	/* nonzero if rdpmc is allowed */
 } mm_context_t;
 
 #ifdef CONFIG_SMP

arch/x86/include/asm/mmu_context.h

@@ -18,6 +18,18 @@ static inline void paravirt_activate_mm(struct mm_struct *prev,
 }
 #endif	/* !CONFIG_PARAVIRT */
 
+#ifdef CONFIG_PERF_EVENTS
+static inline void load_mm_cr4(struct mm_struct *mm)
+{
+	if (atomic_read(&mm->context.perf_rdpmc_allowed))
+		cr4_set_bits(X86_CR4_PCE);
+	else
+		cr4_clear_bits(X86_CR4_PCE);
+}
+#else
+static inline void load_mm_cr4(struct mm_struct *mm) {}
+#endif
+
 /*
  * Used for LDT copy/destruction.
  */
@@ -52,6 +64,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/* Stop flush ipis for the previous mm */
 		cpumask_clear_cpu(cpu, mm_cpumask(prev));
 
+		/* Load per-mm CR4 state */
+		load_mm_cr4(next);
+
 		/*
 		 * Load the LDT, if the LDT is different.
 		 *
@@ -87,6 +102,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			 */
 			load_cr3(next->pgd);
 			trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL);
+			load_mm_cr4(next);
 			load_LDT_nolock(&next->context);
 		}
 	}

arch/x86/kernel/cpu/perf_event.c

@@ -31,6 +31,7 @@
 #include <asm/nmi.h>
 #include <asm/smp.h>
 #include <asm/alternative.h>
+#include <asm/mmu_context.h>
 #include <asm/tlbflush.h>
 #include <asm/timer.h>
 #include <asm/desc.h>
@@ -1328,8 +1329,6 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 		break;
 
 	case CPU_STARTING:
-		if (x86_pmu.attr_rdpmc)
-			cr4_set_bits(X86_CR4_PCE);
 		if (x86_pmu.cpu_starting)
 			x86_pmu.cpu_starting(cpu);
 		break;
@@ -1805,14 +1804,44 @@ static int x86_pmu_event_init(struct perf_event *event)
 			event->destroy(event);
 	}
 
+	if (ACCESS_ONCE(x86_pmu.attr_rdpmc))
+		event->hw.flags |= PERF_X86_EVENT_RDPMC_ALLOWED;
+
 	return err;
 }
 
+static void refresh_pce(void *ignored)
+{
+	if (current->mm)
+		load_mm_cr4(current->mm);
+}
+
+static void x86_pmu_event_mapped(struct perf_event *event)
+{
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
+		return;
+
+	if (atomic_inc_return(&current->mm->context.perf_rdpmc_allowed) == 1)
+		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+}
+
+static void x86_pmu_event_unmapped(struct perf_event *event)
+{
+	if (!current->mm)
+		return;
+
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
+		return;
+
+	if (atomic_dec_and_test(&current->mm->context.perf_rdpmc_allowed))
+		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+}
+
 static int x86_pmu_event_idx(struct perf_event *event)
 {
 	int idx = event->hw.idx;
 
-	if (!x86_pmu.attr_rdpmc)
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
 		return 0;
 
 	if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) {
@@ -1830,16 +1859,6 @@ static ssize_t get_attr_rdpmc(struct device *cdev,
 	return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
 }
 
-static void change_rdpmc(void *info)
-{
-	bool enable = !!(unsigned long)info;
-
-	if (enable)
-		cr4_set_bits(X86_CR4_PCE);
-	else
-		cr4_clear_bits(X86_CR4_PCE);
-}
-
 static ssize_t set_attr_rdpmc(struct device *cdev,
 			      struct device_attribute *attr,
 			      const char *buf, size_t count)
@@ -1854,11 +1873,7 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
 	if (x86_pmu.attr_rdpmc_broken)
 		return -ENOTSUPP;
 
-	if (!!val != !!x86_pmu.attr_rdpmc) {
 	x86_pmu.attr_rdpmc = !!val;
-		on_each_cpu(change_rdpmc, (void *)val, 1);
-	}
-
 	return count;
 }
 
@@ -1901,6 +1916,9 @@ static struct pmu pmu = {
 
 	.event_init		= x86_pmu_event_init,
 
+	.event_mapped		= x86_pmu_event_mapped,
+	.event_unmapped		= x86_pmu_event_unmapped,
+
 	.add			= x86_pmu_add,
 	.del			= x86_pmu_del,
 	.start			= x86_pmu_start,
@@ -1922,7 +1940,8 @@ void arch_perf_update_userpage(struct perf_event *event,
 
 	userpg->cap_user_time = 0;
 	userpg->cap_user_time_zero = 0;
-	userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc;
+	userpg->cap_user_rdpmc =
+		!!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED);
 	userpg->pmc_width = x86_pmu.cntval_bits;
 
 	if (!sched_clock_stable())

arch/x86/kernel/cpu/perf_event.h

@@ -71,6 +71,8 @@ struct event_constraint {
 #define PERF_X86_EVENT_COMMITTED	0x8 /* event passed commit_txn */
 #define PERF_X86_EVENT_PEBS_LD_HSW	0x10 /* haswell style datala, load */
 #define PERF_X86_EVENT_PEBS_NA_HSW	0x20 /* haswell style datala, unknown */
+#define PERF_X86_EVENT_RDPMC_ALLOWED	0x40 /* grant rdpmc permission */
+
 
 struct amd_nb {
 	int nb_id;  /* NorthBridge id */