4.20.16-stable review

index bfe1639df02d..97fc498dc767 100644
@@ -47,6 +47,24 @@ static inline bool drm_arch_can_wc_memory(void)
 	return false;
 #elif defined(CONFIG_MIPS) && defined(CONFIG_CPU_LOONGSON3)
 	return false;
+#elif defined(CONFIG_ARM) || defined(CONFIG_ARM64)
+	/*
+	 * The DRM driver stack is designed to work with cache coherent devices
+	 * only, but permits an optimization to be enabled in some cases, where
+	 * for some buffers, both the CPU and the GPU use uncached mappings,
+	 * removing the need for DMA snooping and allocation in the CPU caches.
+	 *
+	 * The use of uncached GPU mappings relies on the correct implementation
+	 * of the PCIe NoSnoop TLP attribute by the platform, otherwise the GPU
+	 * will use cached mappings nonetheless. On x86 platforms, this does not
+	 * seem to matter, as uncached CPU mappings will snoop the caches in any
+	 * case. However, on ARM and arm64, enabling this optimization on a
+	 * platform where NoSnoop is ignored results in loss of coherency, which
+	 * breaks correct operation of the device. Since we have no way of
+	 * detecting whether NoSnoop works or not, just disable this
+	 * optimization entirely for ARM and arm64.
+	 */
+	return false;
 #else
 	return true;
 #endif

@@ -274,6 +274,7 @@ static unsigned int ir_update_mapping(st
 				      unsigned int new_keycode)
 {
 	int old_keycode = rc_map->scan[index].keycode;
+	int i;
 
 	/* Did the user wish to remove the mapping? */
 	if (new_keycode == KEY_RESERVED || new_keycode == KEY_UNKNOWN) {

@@ -288,9 +289,20 @@ static unsigned int ir_update_mapping(st
 			old_keycode == KEY_RESERVED ? "New" : "Replacing",
 			rc_map->scan[index].scancode, new_keycode);
 		rc_map->scan[index].keycode = new_keycode;
+		__set_bit(new_keycode, dev->input_dev->keybit);
 	}
 
 	if (old_keycode != KEY_RESERVED) {
+		/* A previous mapping was updated... */
+		__clear_bit(old_keycode, dev->input_dev->keybit);
+		/* ... but another scancode might use the same keycode */
+		for (i = 0; i < rc_map->len; i++) {
+			if (rc_map->scan[i].keycode == old_keycode) {
+				__set_bit(old_keycode, dev->input_dev->keybit);
+				break;
+			}
+		}
+
 		/* Possibly shrink the keytable, failure is not a problem */
 		ir_resize_table(dev, rc_map, GFP_ATOMIC);
 	}

@@ -1750,7 +1762,6 @@ static int rc_prepare_rx_device(struct r
 	set_bit(EV_REP, dev->input_dev->evbit);
 	set_bit(EV_MSC, dev->input_dev->evbit);
 	set_bit(MSC_SCAN, dev->input_dev->mscbit);
-	bitmap_fill(dev->input_dev->keybit, KEY_CNT);
 
 	/* Pointer/mouse events */
 	set_bit(EV_REL, dev->input_dev->evbit);

@@ -432,31 +432,6 @@ static void pcie_pme_remove(struct pcie_
 	kfree(get_service_data(srv));
 }
 
-static int pcie_pme_runtime_suspend(struct pcie_device *srv)
-{
-	struct pcie_pme_service_data *data = get_service_data(srv);
-
-	spin_lock_irq(&data->lock);
-	pcie_pme_interrupt_enable(srv->port, false);
-	pcie_clear_root_pme_status(srv->port);
-	data->noirq = true;
-	spin_unlock_irq(&data->lock);
-
-	return 0;
-}
-
-static int pcie_pme_runtime_resume(struct pcie_device *srv)
-{
-	struct pcie_pme_service_data *data = get_service_data(srv);
-
-	spin_lock_irq(&data->lock);
-	pcie_pme_interrupt_enable(srv->port, true);
-	data->noirq = false;
-	spin_unlock_irq(&data->lock);
-
-	return 0;
-}
-
 static struct pcie_port_service_driver pcie_pme_driver = {
 	.name		= "pcie_pme",
 	.port_type	= PCI_EXP_TYPE_ROOT_PORT,

@@ -464,8 +439,6 @@ static struct pcie_port_service_driver p
 
 	.probe		= pcie_pme_probe,
 	.suspend	= pcie_pme_suspend,
-	.runtime_suspend = pcie_pme_runtime_suspend,
-	.runtime_resume	= pcie_pme_runtime_resume,
 	.resume		= pcie_pme_resume,
 	.remove		= pcie_pme_remove,
 };

@@ -545,8 +545,7 @@ static void sk_psock_destroy_deferred(st
 	struct sk_psock *psock = container_of(gc, struct sk_psock, gc);
 
 	/* No sk_callback_lock since already detached. */
-	if (psock->parser.enabled)
-		strp_done(&psock->parser.strp);
+	strp_done(&psock->parser.strp);
 
 	cancel_work_sync(&psock->work);
 

@@ -545,6 +545,7 @@ static void sk_psock_destroy_deferred(st
 	struct sk_psock *psock = container_of(gc, struct sk_psock, gc);
 
 	/* No sk_callback_lock since already detached. */
+	strp_stop(&psock->parser.strp);
 	strp_done(&psock->parser.strp);
 
 	cancel_work_sync(&psock->work);

@@ -107,7 +107,7 @@ static int glock_wake_function(wait_queu
 
 static wait_queue_head_t *glock_waitqueue(struct lm_lockname *name)
 {
-	u32 hash = jhash2((u32 *)name, sizeof(*name) / 4, 0);
+	u32 hash = jhash2((u32 *)name, ht_parms.key_len / 4, 0);
 
 	return glock_wait_table + hash_32(hash, GLOCK_WAIT_TABLE_BITS);
 }

@@ -15,74 +15,77 @@
 
 #include <trace/events/erofs.h>
 
-/* based on the value of qn->len is accurate */
-static inline int dirnamecmp(struct qstr *qn,
-	struct qstr *qd, unsigned int *matched)
+struct erofs_qstr {
+	const unsigned char *name;
+	const unsigned char *end;
+};
+
+/* based on the end of qn is accurate and it must have the trailing '\0' */
+static inline int dirnamecmp(const struct erofs_qstr *qn,
+			     const struct erofs_qstr *qd,
+			     unsigned int *matched)
 {
-	unsigned int i = *matched, len = min(qn->len, qd->len);
-loop:
-	if (unlikely(i >= len)) {
-		*matched = i;
-		if (qn->len < qd->len) {
-			/*
-			 * actually (qn->len == qd->len)
-			 * when qd->name[i] == '\0'
-			 */
-			return qd->name[i] == '\0' ? 0 : -1;
-		}
-		return (qn->len > qd->len);
-	}
+	unsigned int i = *matched;
 
-	if (qn->name[i] != qd->name[i]) {
-		*matched = i;
-		return qn->name[i] > qd->name[i] ? 1 : -1;
+	/*
+	 * on-disk error, let's only BUG_ON in the debugging mode.
+	 * otherwise, it will return 1 to just skip the invalid name
+	 * and go on (in consideration of the lookup performance).
+	 */
+	DBG_BUGON(qd->name > qd->end);
+
+	/* qd could not have trailing '\0' */
+	/* However it is absolutely safe if < qd->end */
+	while (qd->name + i < qd->end && qd->name[i] != '\0') {
+		if (qn->name[i] != qd->name[i]) {
+			*matched = i;
+			return qn->name[i] > qd->name[i] ? 1 : -1;
+		}
+		++i;
 	}
-
-	++i;
-	goto loop;
+	*matched = i;
+	/* See comments in __d_alloc on the terminating NUL character */
+	return qn->name[i] == '\0' ? 0 : 1;
 }
 
-static struct erofs_dirent *find_target_dirent(
-	struct qstr *name,
-	u8 *data, int maxsize)
+#define nameoff_from_disk(off, sz)	(le16_to_cpu(off) & ((sz) - 1))
+
+static struct erofs_dirent *find_target_dirent(struct erofs_qstr *name,
+					       u8 *data,
+					       unsigned int dirblksize,
+					       const int ndirents)
 {
-	unsigned int ndirents, head, back;
+	int head, back;
 	unsigned int startprfx, endprfx;
 	struct erofs_dirent *const de = (struct erofs_dirent *)data;
 
-	/* make sure that maxsize is valid */
-	BUG_ON(maxsize < sizeof(struct erofs_dirent));
-
-	ndirents = le16_to_cpu(de->nameoff) / sizeof(*de);
-
-	/* corrupted dir (may be unnecessary...) */
-	BUG_ON(!ndirents);
-
-	head = 0;
+	/* since the 1st dirent has been evaluated previously */
+	head = 1;
 	back = ndirents - 1;
 	startprfx = endprfx = 0;
 
 	while (head <= back) {
-		unsigned int mid = head + (back - head) / 2;
-		unsigned int nameoff = le16_to_cpu(de[mid].nameoff);
+		const int mid = head + (back - head) / 2;
+		const int nameoff = nameoff_from_disk(de[mid].nameoff,
+						      dirblksize);
 		unsigned int matched = min(startprfx, endprfx);
-
-		struct qstr dname = QSTR_INIT(data + nameoff,
-			unlikely(mid >= ndirents - 1) ?
-				maxsize - nameoff :
-				le16_to_cpu(de[mid + 1].nameoff) - nameoff);
+		struct erofs_qstr dname = {
+			.name = data + nameoff,
+			.end = unlikely(mid >= ndirents - 1) ?
+				data + dirblksize :
+				data + nameoff_from_disk(de[mid + 1].nameoff,
+							 dirblksize)
+		};
 
 		/* string comparison without already matched prefix */
 		int ret = dirnamecmp(name, &dname, &matched);
 
-		if (unlikely(!ret))
+		if (unlikely(!ret)) {
 			return de + mid;
-		else if (ret > 0) {
+		} else if (ret > 0) {
 			head = mid + 1;
 			startprfx = matched;
-		} else if (unlikely(mid < 1))	/* fix "mid" overflow */
-			break;
-		else {
+		} else {
 			back = mid - 1;
 			endprfx = matched;
 		}

@@ -91,12 +94,12 @@ static struct erofs_dirent *find_target_
 	return ERR_PTR(-ENOENT);
 }
 
-static struct page *find_target_block_classic(
-	struct inode *dir,
-	struct qstr *name, int *_diff)
+static struct page *find_target_block_classic(struct inode *dir,
+					      struct erofs_qstr *name,
+					      int *_ndirents)
 {
 	unsigned int startprfx, endprfx;
-	unsigned int head, back;
+	int head, back;
 	struct address_space *const mapping = dir->i_mapping;
 	struct page *candidate = ERR_PTR(-ENOENT);
 

@@ -105,41 +108,43 @@ static struct page *find_target_block_cl
 	back = inode_datablocks(dir) - 1;
 
 	while (head <= back) {
-		unsigned int mid = head + (back - head) / 2;
+		const int mid = head + (back - head) / 2;
 		struct page *page = read_mapping_page(mapping, mid, NULL);
 
-		if (IS_ERR(page)) {
-exact_out:
-			if (!IS_ERR(candidate)) /* valid candidate */
-				put_page(candidate);
-			return page;
-		} else {
-			int diff;
-			unsigned int ndirents, matched;
-			struct qstr dname;
+		if (!IS_ERR(page)) {
 			struct erofs_dirent *de = kmap_atomic(page);
-			unsigned int nameoff = le16_to_cpu(de->nameoff);
-
-			ndirents = nameoff / sizeof(*de);
+			const int nameoff = nameoff_from_disk(de->nameoff,
+							      EROFS_BLKSIZ);
+			const int ndirents = nameoff / sizeof(*de);
+			int diff;
+			unsigned int matched;
+			struct erofs_qstr dname;
 
-			/* corrupted dir (should have one entry at least) */
-			BUG_ON(!ndirents || nameoff > PAGE_SIZE);
+			if (unlikely(!ndirents)) {
+				DBG_BUGON(1);
+				kunmap_atomic(de);
+				put_page(page);
+				page = ERR_PTR(-EIO);
+				goto out;
+			}
 
 			matched = min(startprfx, endprfx);
 
 			dname.name = (u8 *)de + nameoff;
-			dname.len = ndirents == 1 ?
-				/* since the rest of the last page is 0 */
-				EROFS_BLKSIZ - nameoff
-				: le16_to_cpu(de[1].nameoff) - nameoff;
+			if (ndirents == 1)
+				dname.end = (u8 *)de + EROFS_BLKSIZ;
+			else
+				dname.end = (u8 *)de +
+					nameoff_from_disk(de[1].nameoff,
+							  EROFS_BLKSIZ);
 
 			/* string comparison without already matched prefix */
 			diff = dirnamecmp(name, &dname, &matched);
 			kunmap_atomic(de);
 
 			if (unlikely(!diff)) {
-				*_diff = 0;
-				goto exact_out;
+				*_ndirents = 0;
+				goto out;
 			} else if (diff > 0) {
 				head = mid + 1;
 				startprfx = matched;

@@ -147,45 +152,51 @@ exact_out:
 				if (likely(!IS_ERR(candidate)))
 					put_page(candidate);
 				candidate = page;
+				*_ndirents = ndirents;
 			} else {
 				put_page(page);
 
-				if (unlikely(mid < 1))	/* fix "mid" overflow */
-					break;
-
 				back = mid - 1;
 				endprfx = matched;
 			}
+			continue;
 		}
+out:		/* free if the candidate is valid */
+		if (!IS_ERR(candidate))
+			put_page(candidate);
+		return page;
 	}
-	*_diff = 1;
 	return candidate;
 }
 
 int erofs_namei(struct inode *dir,
-	struct qstr *name,
-	erofs_nid_t *nid, unsigned int *d_type)
+		struct qstr *name,
+		erofs_nid_t *nid, unsigned int *d_type)
 {
-	int diff;
+	int ndirents;
 	struct page *page;
-	u8 *data;
+	void *data;
 	struct erofs_dirent *de;
+	struct erofs_qstr qn;
 
 	if (unlikely(!dir->i_size))
 		return -ENOENT;
 
-	diff = 1;
-	page = find_target_block_classic(dir, name, &diff);
+	qn.name = name->name;
+	qn.end = name->name + name->len;
+
+	ndirents = 0;
+	page = find_target_block_classic(dir, &qn, &ndirents);
 
 	if (unlikely(IS_ERR(page)))
 		return PTR_ERR(page);
 
 	data = kmap_atomic(page);
 	/* the target page has been mapped */
-	de = likely(diff) ?
-		/* since the rest of the last page is 0 */
-		find_target_dirent(name, data, EROFS_BLKSIZ) :
-		(struct erofs_dirent *)data;
+	if (ndirents)
+		de = find_target_dirent(&qn, data, EROFS_BLKSIZ, ndirents);
+	else
+		de = (struct erofs_dirent *)data;
 
 	if (likely(!IS_ERR(de))) {
 		*nid = le64_to_cpu(de->nid);

@@ -636,15 +636,15 @@ static int ath9k_of_init(struct ath_soft
 		ret = ath9k_eeprom_request(sc, eeprom_name);
 		if (ret)
 			return ret;
+
+		ah->ah_flags &= ~AH_USE_EEPROM;
+		ah->ah_flags |= AH_NO_EEP_SWAP;
 	}
 
 	mac = of_get_mac_address(np);
 	if (mac)
 		ether_addr_copy(common->macaddr, mac);
 
-	ah->ah_flags &= ~AH_USE_EEPROM;
-	ah->ah_flags |= AH_NO_EEP_SWAP;
-
 	return 0;
 }
 

@@ -1995,7 +1995,7 @@ static int x86_pmu_commit_txn(struct pmu
  */
 static void free_fake_cpuc(struct cpu_hw_events *cpuc)
 {
-	kfree(cpuc->shared_regs);
+	intel_cpuc_finish(cpuc);
 	kfree(cpuc);
 }
 

@@ -2007,14 +2007,11 @@ static struct cpu_hw_events *allocate_fa
 	cpuc = kzalloc(sizeof(*cpuc), GFP_KERNEL);
 	if (!cpuc)
 		return ERR_PTR(-ENOMEM);
-
-	/* only needed, if we have extra_regs */
-	if (x86_pmu.extra_regs) {
-		cpuc->shared_regs = allocate_shared_regs(cpu);
-		if (!cpuc->shared_regs)
-			goto error;
-	}
 	cpuc->is_fake = 1;
+
+	if (intel_cpuc_prepare(cpuc, cpu))
+		goto error;
+
 	return cpuc;
 error:
 	free_fake_cpuc(cpuc);

@@ -3398,7 +3398,7 @@ ssize_t intel_event_sysfs_show(char *pag
 	return x86_event_sysfs_show(page, config, event);
 }
 
-struct intel_shared_regs *allocate_shared_regs(int cpu)
+static struct intel_shared_regs *allocate_shared_regs(int cpu)
 {
 	struct intel_shared_regs *regs;
 	int i;

@@ -3430,10 +3430,9 @@ static struct intel_excl_cntrs *allocate
 	return c;
 }
 
-static int intel_pmu_cpu_prepare(int cpu)
-{
-	struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
 
+int intel_cpuc_prepare(struct cpu_hw_events *cpuc, int cpu)
+{
 	if (x86_pmu.extra_regs || x86_pmu.lbr_sel_map) {
 		cpuc->shared_regs = allocate_shared_regs(cpu);
 		if (!cpuc->shared_regs)

@@ -3443,7 +3442,7 @@ static int intel_pmu_cpu_prepare(int cpu
 	if (x86_pmu.flags & PMU_FL_EXCL_CNTRS) {
 		size_t sz = X86_PMC_IDX_MAX * sizeof(struct event_constraint);
 
-		cpuc->constraint_list = kzalloc(sz, GFP_KERNEL);
+		cpuc->constraint_list = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
 		if (!cpuc->constraint_list)
 			goto err_shared_regs;
 

@@ -3468,6 +3467,11 @@ err:
 	return -ENOMEM;
 }
 
+static int intel_pmu_cpu_prepare(int cpu)
+{
+	return intel_cpuc_prepare(&per_cpu(cpu_hw_events, cpu), cpu);
+}
+
 static void flip_smm_bit(void *data)
 {
 	unsigned long set = *(unsigned long *)data;

@@ -3542,9 +3546,8 @@ static void intel_pmu_cpu_starting(int c
 	}
 }
 
-static void free_excl_cntrs(int cpu)
+static void free_excl_cntrs(struct cpu_hw_events *cpuc)
 {
-	struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
 	struct intel_excl_cntrs *c;
 
 	c = cpuc->excl_cntrs;

@@ -3565,9 +3568,8 @@ static void intel_pmu_cpu_dying(int cpu)
 		disable_counter_freeze();
 }
 
-static void intel_pmu_cpu_dead(int cpu)
+void intel_cpuc_finish(struct cpu_hw_events *cpuc)
 {
-	struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
 	struct intel_shared_regs *pc;
 
 	pc = cpuc->shared_regs;

@@ -3577,7 +3579,12 @@ static void intel_pmu_cpu_dead(int cpu)
 		cpuc->shared_regs = NULL;
 	}
 
-	free_excl_cntrs(cpu);
+	free_excl_cntrs(cpuc);
+}
+
+static void intel_pmu_cpu_dead(int cpu)
+{
+	intel_cpuc_finish(&per_cpu(cpu_hw_events, cpu));
 }
 
 static void intel_pmu_sched_task(struct perf_event_context *ctx,

@@ -4715,7 +4722,7 @@ static __init int fixup_ht_bug(void)
 	hardlockup_detector_perf_restart();
 
 	for_each_online_cpu(c)
-		free_excl_cntrs(c);
+		free_excl_cntrs(&per_cpu(cpu_hw_events, c));
 
 	cpus_read_unlock();
 	pr_info("PMU erratum BJ122, BV98, HSD29 workaround disabled, HT off\n");

@@ -889,7 +889,8 @@ struct event_constraint *
 x86_get_event_constraints(struct cpu_hw_events *cpuc, int idx,
 			  struct perf_event *event);
 
-struct intel_shared_regs *allocate_shared_regs(int cpu);
+extern int intel_cpuc_prepare(struct cpu_hw_events *cpuc, int cpu);
+extern void intel_cpuc_finish(struct cpu_hw_events *cpuc);
 
 int intel_pmu_init(void);
 

@@ -1025,9 +1026,13 @@ static inline int intel_pmu_init(void)
 	return 0;
 }
 
-static inline struct intel_shared_regs *allocate_shared_regs(int cpu)
+static inline int intel_cpuc_prepare(struct cpu_hw_event *cpuc, int cpu)
+{
+	return 0;
+}
+
+static inline void intel_cpuc_finish(struct cpu_hw_event *cpuc)
 {
-	return NULL;
 }
 
 static inline int is_ht_workaround_enabled(void)

@@ -2769,6 +2769,35 @@ intel_stop_scheduling(struct cpu_hw_even
 }
 
 static struct event_constraint *
+dyn_constraint(struct cpu_hw_events *cpuc, struct event_constraint *c, int idx)
+{
+	WARN_ON_ONCE(!cpuc->constraint_list);
+
+	if (!(c->flags & PERF_X86_EVENT_DYNAMIC)) {
+		struct event_constraint *cx;
+
+		/*
+		 * grab pre-allocated constraint entry
+		 */
+		cx = &cpuc->constraint_list[idx];
+
+		/*
+		 * initialize dynamic constraint
+		 * with static constraint
+		 */
+		*cx = *c;
+
+		/*
+		 * mark constraint as dynamic
+		 */
+		cx->flags |= PERF_X86_EVENT_DYNAMIC;
+		c = cx;
+	}
+
+	return c;
+}
+
+static struct event_constraint *
 intel_get_excl_constraints(struct cpu_hw_events *cpuc, struct perf_event *event,
 			   int idx, struct event_constraint *c)
 {

@@ -2798,27 +2827,7 @@ intel_get_excl_constraints(struct cpu_hw
 	 * only needed when constraint has not yet
 	 * been cloned (marked dynamic)
 	 */
-	if (!(c->flags & PERF_X86_EVENT_DYNAMIC)) {
-		struct event_constraint *cx;
-
-		/*
-		 * grab pre-allocated constraint entry
-		 */
-		cx = &cpuc->constraint_list[idx];
-
-		/*
-		 * initialize dynamic constraint
-		 * with static constraint
-		 */
-		*cx = *c;
-
-		/*
-		 * mark constraint as dynamic, so we
-		 * can free it later on
-		 */
-		cx->flags |= PERF_X86_EVENT_DYNAMIC;
-		c = cx;
-	}
+	c = dyn_constraint(cpuc, c, idx);
 
 	/*
 	 * From here on, the constraint is dynamic.

@@ -342,6 +342,7 @@
 /* Intel-defined CPU features, CPUID level 0x00000007:0 (EDX), word 18 */
 #define X86_FEATURE_AVX512_4VNNIW	(18*32+ 2) /* AVX-512 Neural Network Instructions */
 #define X86_FEATURE_AVX512_4FMAPS	(18*32+ 3) /* AVX-512 Multiply Accumulation Single precision */
+#define X86_FEATURE_TSX_FORCE_ABORT	(18*32+13) /* "" TSX_FORCE_ABORT */
 #define X86_FEATURE_PCONFIG		(18*32+18) /* Intel PCONFIG */
 #define X86_FEATURE_SPEC_CTRL		(18*32+26) /* "" Speculation Control (IBRS + IBPB) */
 #define X86_FEATURE_INTEL_STIBP		(18*32+27) /* "" Single Thread Indirect Branch Predictors */

@@ -630,6 +630,12 @@
 
 #define MSR_IA32_TSC_DEADLINE		0x000006E0
 
+
+#define MSR_TSX_FORCE_ABORT		0x0000010F
+
+#define MSR_TFA_RTM_FORCE_ABORT_BIT	0
+#define MSR_TFA_RTM_FORCE_ABORT		BIT_ULL(MSR_TFA_RTM_FORCE_ABORT_BIT)
+
 /* P4/Xeon+ specific */
 #define MSR_IA32_MCG_EAX		0x00000180
 #define MSR_IA32_MCG_EBX		0x00000181

@@ -1999,6 +1999,39 @@ static void intel_pmu_nhm_enable_all(int
 	intel_pmu_enable_all(added);
 }
 
+static void intel_set_tfa(struct cpu_hw_events *cpuc, bool on)
+{
+	u64 val = on ? MSR_TFA_RTM_FORCE_ABORT : 0;
+
+	if (cpuc->tfa_shadow != val) {
+		cpuc->tfa_shadow = val;
+		wrmsrl(MSR_TSX_FORCE_ABORT, val);
+	}
+}
+
+static void intel_tfa_commit_scheduling(struct cpu_hw_events *cpuc, int idx, int cntr)
+{
+	/*
+	 * We're going to use PMC3, make sure TFA is set before we touch it.
+	 */
+	if (cntr == 3 && !cpuc->is_fake)
+		intel_set_tfa(cpuc, true);
+}
+
+static void intel_tfa_pmu_enable_all(int added)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	/*
+	 * If we find PMC3 is no longer used when we enable the PMU, we can
+	 * clear TFA.
+	 */
+	if (!test_bit(3, cpuc->active_mask))
+		intel_set_tfa(cpuc, false);
+
+	intel_pmu_enable_all(added);
+}
+
 static void enable_counter_freeze(void)
 {
 	update_debugctlmsr(get_debugctlmsr() |

@@ -3354,6 +3387,26 @@ glp_get_event_constraints(struct cpu_hw_
 	return c;
 }
 
+static bool allow_tsx_force_abort = true;
+
+static struct event_constraint *
+tfa_get_event_constraints(struct cpu_hw_events *cpuc, int idx,
+			  struct perf_event *event)
+{
+	struct event_constraint *c = hsw_get_event_constraints(cpuc, idx, event);
+
+	/*
+	 * Without TFA we must not use PMC3.
+	 */
+	if (!allow_tsx_force_abort && test_bit(3, c->idxmsk)) {
+		c = dyn_constraint(cpuc, c, idx);
+		c->idxmsk64 &= ~(1ULL << 3);
+		c->weight--;
+	}
+
+	return c;
+}
+
 /*
  * Broadwell:
  *

@@ -3448,13 +3501,15 @@ int intel_cpuc_prepare(struct cpu_hw_eve
 			goto err;
 	}
 
-	if (x86_pmu.flags & PMU_FL_EXCL_CNTRS) {
+	if (x86_pmu.flags & (PMU_FL_EXCL_CNTRS | PMU_FL_TFA)) {
 		size_t sz = X86_PMC_IDX_MAX * sizeof(struct event_constraint);
 
 		cpuc->constraint_list = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
 		if (!cpuc->constraint_list)
 			goto err_shared_regs;
+	}
 
+	if (x86_pmu.flags & PMU_FL_EXCL_CNTRS) {
 		cpuc->excl_cntrs = allocate_excl_cntrs(cpu);
 		if (!cpuc->excl_cntrs)
 			goto err_constraint_list;

@@ -3564,9 +3619,10 @@ static void free_excl_cntrs(struct cpu_h
 		if (c->core_id == -1 || --c->refcnt == 0)
 			kfree(c);
 		cpuc->excl_cntrs = NULL;
-		kfree(cpuc->constraint_list);
-		cpuc->constraint_list = NULL;
 	}
+
+	kfree(cpuc->constraint_list);
+	cpuc->constraint_list = NULL;
 }
 
 static void intel_pmu_cpu_dying(int cpu)

@@ -4086,8 +4142,11 @@ static struct attribute *intel_pmu_caps_
        NULL
 };
 
+DEVICE_BOOL_ATTR(allow_tsx_force_abort, 0644, allow_tsx_force_abort);
+
 static struct attribute *intel_pmu_attrs[] = {
 	&dev_attr_freeze_on_smi.attr,
+	NULL, /* &dev_attr_allow_tsx_force_abort.attr.attr */
 	NULL,
 };
 

@@ -4580,6 +4639,15 @@ __init int intel_pmu_init(void)
 		tsx_attr = hsw_tsx_events_attrs;
 		intel_pmu_pebs_data_source_skl(
 			boot_cpu_data.x86_model == INTEL_FAM6_SKYLAKE_X);
+
+		if (boot_cpu_has(X86_FEATURE_TSX_FORCE_ABORT)) {
+			x86_pmu.flags |= PMU_FL_TFA;
+			x86_pmu.get_event_constraints = tfa_get_event_constraints;
+			x86_pmu.enable_all = intel_tfa_pmu_enable_all;
+			x86_pmu.commit_scheduling = intel_tfa_commit_scheduling;
+			intel_pmu_attrs[1] = &dev_attr_allow_tsx_force_abort.attr.attr;
+		}
+
 		pr_cont("Skylake events, ");
 		name = "skylake";
 		break;

@@ -243,6 +243,11 @@ struct cpu_hw_events {
 	int excl_thread_id; /* 0 or 1 */
 
 	/*
+	 * SKL TSX_FORCE_ABORT shadow
+	 */
+	u64				tfa_shadow;
+
+	/*
 	 * AMD specific bits
 	 */
 	struct amd_nb			*amd_nb;

@@ -681,6 +686,7 @@ do {									\
 #define PMU_FL_EXCL_CNTRS	0x4 /* has exclusive counter requirements  */
 #define PMU_FL_EXCL_ENABLED	0x8 /* exclusive counter active */
 #define PMU_FL_PEBS_ALL		0x10 /* all events are valid PEBS events */
+#define PMU_FL_TFA		0x20 /* deal with TSX force abort */
 
 #define EVENT_VAR(_id)  event_attr_##_id
 #define EVENT_PTR(_id) &event_attr_##_id.attr.attr