Merge tag 'perf-urgent-2020-04-12' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf fixes from Thomas Gleixner:
 "Three fixes/updates for perf:

   - Fix the perf event cgroup tracking which tries to track the cgroup
     even for disabled events.

   - Add Ice Lake server support for uncore events

   - Disable pagefaults when retrieving the physical address in the
     sampling code"

* tag 'perf-urgent-2020-04-12' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/core: Disable page faults when getting phys address
  perf/x86/intel/uncore: Add Ice Lake server uncore support
  perf/cgroup: Correct indirection in perf_less_group_idx()
  perf/core: Fix event cgroup tracking

arch/x86/events/intel/uncore.c

@@ -1476,6 +1476,12 @@ static const struct intel_uncore_init_fun tgl_l_uncore_init __initconst = {
 	.mmio_init = tgl_l_uncore_mmio_init,
 };
 
+static const struct intel_uncore_init_fun icx_uncore_init __initconst = {
+	.cpu_init = icx_uncore_cpu_init,
+	.pci_init = icx_uncore_pci_init,
+	.mmio_init = icx_uncore_mmio_init,
+};
+
 static const struct intel_uncore_init_fun snr_uncore_init __initconst = {
 	.cpu_init = snr_uncore_cpu_init,
 	.pci_init = snr_uncore_pci_init,
@@ -1511,6 +1517,8 @@ static const struct x86_cpu_id intel_uncore_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_L,		&icl_uncore_init),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_NNPI,	&icl_uncore_init),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE,		&icl_uncore_init),
+	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_D,		&icx_uncore_init),
+	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_X,		&icx_uncore_init),
 	X86_MATCH_INTEL_FAM6_MODEL(TIGERLAKE_L,		&tgl_l_uncore_init),
 	X86_MATCH_INTEL_FAM6_MODEL(TIGERLAKE,		&tgl_uncore_init),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT_D,	&snr_uncore_init),

arch/x86/events/intel/uncore.h

@@ -550,6 +550,9 @@ void skx_uncore_cpu_init(void);
 int snr_uncore_pci_init(void);
 void snr_uncore_cpu_init(void);
 void snr_uncore_mmio_init(void);
+int icx_uncore_pci_init(void);
+void icx_uncore_cpu_init(void);
+void icx_uncore_mmio_init(void);
 
 /* uncore_nhmex.c */
 void nhmex_uncore_cpu_init(void);

kernel/events/core.c

@@ -983,16 +983,10 @@ perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
 	event->shadow_ctx_time = now - t->timestamp;
 }
 
-/*
- * Update cpuctx->cgrp so that it is set when first cgroup event is added and
- * cleared when last cgroup event is removed.
- */
 static inline void
-list_update_cgroup_event(struct perf_event *event,
-			 struct perf_event_context *ctx, bool add)
+perf_cgroup_event_enable(struct perf_event *event, struct perf_event_context *ctx)
 {
 	struct perf_cpu_context *cpuctx;
-	struct list_head *cpuctx_entry;
 
 	if (!is_cgroup_event(event))
 		return;
@@ -1009,28 +1003,41 @@ list_update_cgroup_event(struct perf_event *event,
 	 * because if the first would mismatch, the second would not try again
 	 * and we would leave cpuctx->cgrp unset.
 	 */
-	if (add && !cpuctx->cgrp) {
+	if (ctx->is_active && !cpuctx->cgrp) {
 		struct perf_cgroup *cgrp = perf_cgroup_from_task(current, ctx);
 
 		if (cgroup_is_descendant(cgrp->css.cgroup, event->cgrp->css.cgroup))
 			cpuctx->cgrp = cgrp;
 	}
 
-	if (add && ctx->nr_cgroups++)
+	if (ctx->nr_cgroups++)
 		return;
-	else if (!add && --ctx->nr_cgroups)
+
+	list_add(&cpuctx->cgrp_cpuctx_entry,
+			per_cpu_ptr(&cgrp_cpuctx_list, event->cpu));
+}
+
+static inline void
+perf_cgroup_event_disable(struct perf_event *event, struct perf_event_context *ctx)
+{
+	struct perf_cpu_context *cpuctx;
+
+	if (!is_cgroup_event(event))
 		return;
 
-	/* no cgroup running */
-	if (!add)
+	/*
+	 * Because cgroup events are always per-cpu events,
+	 * @ctx == &cpuctx->ctx.
+	 */
+	cpuctx = container_of(ctx, struct perf_cpu_context, ctx);
+
+	if (--ctx->nr_cgroups)
+		return;
+
+	if (ctx->is_active && cpuctx->cgrp)
 		cpuctx->cgrp = NULL;
 
-	cpuctx_entry = &cpuctx->cgrp_cpuctx_entry;
-	if (add)
-		list_add(cpuctx_entry,
-			 per_cpu_ptr(&cgrp_cpuctx_list, event->cpu));
-	else
-		list_del(cpuctx_entry);
+	list_del(&cpuctx->cgrp_cpuctx_entry);
 }
 
 #else /* !CONFIG_CGROUP_PERF */
@@ -1096,11 +1103,14 @@ static inline u64 perf_cgroup_event_time(struct perf_event *event)
 }
 
 static inline void
-list_update_cgroup_event(struct perf_event *event,
-			 struct perf_event_context *ctx, bool add)
+perf_cgroup_event_enable(struct perf_event *event, struct perf_event_context *ctx)
 {
 }
 
+static inline void
+perf_cgroup_event_disable(struct perf_event *event, struct perf_event_context *ctx)
+{
+}
 #endif
 
 /*
@@ -1791,13 +1801,14 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx)
 		add_event_to_groups(event, ctx);
 	}
 
-	list_update_cgroup_event(event, ctx, true);
-
 	list_add_rcu(&event->event_entry, &ctx->event_list);
 	ctx->nr_events++;
 	if (event->attr.inherit_stat)
 		ctx->nr_stat++;
 
+	if (event->state > PERF_EVENT_STATE_OFF)
+		perf_cgroup_event_enable(event, ctx);
+
 	ctx->generation++;
 }
 
@@ -1976,8 +1987,6 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 
 	event->attach_state &= ~PERF_ATTACH_CONTEXT;
 
-	list_update_cgroup_event(event, ctx, false);
-
 	ctx->nr_events--;
 	if (event->attr.inherit_stat)
 		ctx->nr_stat--;
@@ -1994,8 +2003,10 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 	 * of error state is by explicit re-enabling
 	 * of the event
 	 */
-	if (event->state > PERF_EVENT_STATE_OFF)
+	if (event->state > PERF_EVENT_STATE_OFF) {
+		perf_cgroup_event_disable(event, ctx);
 		perf_event_set_state(event, PERF_EVENT_STATE_OFF);
+	}
 
 	ctx->generation++;
 }
@@ -2226,6 +2237,7 @@ event_sched_out(struct perf_event *event,
 
 	if (READ_ONCE(event->pending_disable) >= 0) {
 		WRITE_ONCE(event->pending_disable, -1);
+		perf_cgroup_event_disable(event, ctx);
 		state = PERF_EVENT_STATE_OFF;
 	}
 	perf_event_set_state(event, state);
@@ -2363,6 +2375,7 @@ static void __perf_event_disable(struct perf_event *event,
 		event_sched_out(event, cpuctx, ctx);
 
 	perf_event_set_state(event, PERF_EVENT_STATE_OFF);
+	perf_cgroup_event_disable(event, ctx);
 }
 
 /*
@@ -2746,7 +2759,7 @@ static int  __perf_install_in_context(void *info)
 	}
 
 #ifdef CONFIG_CGROUP_PERF
-	if (is_cgroup_event(event)) {
+	if (event->state > PERF_EVENT_STATE_OFF && is_cgroup_event(event)) {
 		/*
 		 * If the current cgroup doesn't match the event's
 		 * cgroup, we should not try to schedule it.
@@ -2906,6 +2919,7 @@ static void __perf_event_enable(struct perf_event *event,
 		ctx_sched_out(ctx, cpuctx, EVENT_TIME);
 
 	perf_event_set_state(event, PERF_EVENT_STATE_INACTIVE);
+	perf_cgroup_event_enable(event, ctx);
 
 	if (!ctx->is_active)
 		return;
@@ -3508,7 +3522,8 @@ static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
 
 static bool perf_less_group_idx(const void *l, const void *r)
 {
-	const struct perf_event *le = l, *re = r;
+	const struct perf_event *le = *(const struct perf_event **)l;
+	const struct perf_event *re = *(const struct perf_event **)r;
 
 	return le->group_index < re->group_index;
 }
@@ -3616,8 +3631,10 @@ static int merge_sched_in(struct perf_event *event, void *data)
 	}
 
 	if (event->state == PERF_EVENT_STATE_INACTIVE) {
-		if (event->attr.pinned)
+		if (event->attr.pinned) {
+			perf_cgroup_event_disable(event, ctx);
 			perf_event_set_state(event, PERF_EVENT_STATE_ERROR);
+		}
 
 		*can_add_hw = 0;
 		ctx->rotate_necessary = 1;
@@ -6917,9 +6934,12 @@ static u64 perf_virt_to_phys(u64 virt)
 		 * Try IRQ-safe __get_user_pages_fast first.
 		 * If failed, leave phys_addr as 0.
 		 */
-		if ((current->mm != NULL) &&
-		    (__get_user_pages_fast(virt, 1, 0, &p) == 1))
+		if (current->mm != NULL) {
+			pagefault_disable();
+			if (__get_user_pages_fast(virt, 1, 0, &p) == 1)
 				phys_addr = page_to_phys(p) + virt % PAGE_SIZE;
+			pagefault_enable();
+		}
 
 		if (p)
 			put_page(p);