4.13.9-stable review

@@ -599,9 +599,13 @@ static const struct x86_cpu_id deadline_
 
 static void apic_check_deadline_errata(void)
 {
-	const struct x86_cpu_id *m = x86_match_cpu(deadline_match);
+	const struct x86_cpu_id *m;
 	u32 rev;
 
+	if (!boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
+		return;
+
+	m = x86_match_cpu(deadline_match);
 	if (!m)
 		return;
 

@@ -602,7 +602,8 @@ static void apic_check_deadline_errata(v
 	const struct x86_cpu_id *m;
 	u32 rev;
 
-	if (!boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
+	if (!boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER) ||
+	    boot_cpu_has(X86_FEATURE_HYPERVISOR))
 		return;
 
 	m = x86_match_cpu(deadline_match);

@@ -309,10 +309,11 @@ static char *get_config_name(struct list
 static struct perf_evsel *
 __add_event(struct list_head *list, int *idx,
 	    struct perf_event_attr *attr,
-	    char *name, struct cpu_map *cpus,
+	    char *name, struct perf_pmu *pmu,
 	    struct list_head *config_terms)
 {
 	struct perf_evsel *evsel;
+	struct cpu_map *cpus = pmu ? pmu->cpus : NULL;
 
 	event_attr_init(attr);
 

@@ -323,7 +324,7 @@ __add_event(struct list_head *list, int
 	(*idx)++;
 	evsel->cpus        = cpu_map__get(cpus);
 	evsel->own_cpus    = cpu_map__get(cpus);
-	evsel->system_wide = !!cpus;
+	evsel->system_wide = pmu ? pmu->is_uncore : false;
 
 	if (name)
 		evsel->name = strdup(name);

@@ -1232,7 +1233,7 @@ int parse_events_add_pmu(struct parse_ev
 
 	if (!head_config) {
 		attr.type = pmu->type;
-		evsel = __add_event(list, &data->idx, &attr, NULL, pmu->cpus, NULL);
+		evsel = __add_event(list, &data->idx, &attr, NULL, pmu, NULL);
 		return evsel ? 0 : -ENOMEM;
 	}
 

@@ -1253,7 +1254,7 @@ int parse_events_add_pmu(struct parse_ev
 		return -EINVAL;
 
 	evsel = __add_event(list, &data->idx, &attr,
-			    get_config_name(head_config), pmu->cpus,
+			    get_config_name(head_config), pmu,
 			    &config_terms);
 	if (evsel) {
 		evsel->unit = info.unit;

84 lines not on this page · raw message

@@ -22,6 +22,7 @@ struct perf_pmu {
 	char *name;
 	__u32 type;
 	bool selectable;
+	bool is_uncore;
 	struct perf_event_attr *default_config;
 	struct cpu_map *cpus;
 	struct list_head format;  /* HEAD struct perf_pmu_format -> list */

@@ -6,17 +6,6 @@
 
 #include "internal.h"
 
-static inline bool check_pmd(struct page_vma_mapped_walk *pvmw)
-{
-	pmd_t pmde;
-	/*
-	 * Make sure we don't re-load pmd between present and !trans_huge check.
-	 * We need a consistent view.
-	 */
-	pmde = READ_ONCE(*pvmw->pmd);
-	return pmd_present(pmde) && !pmd_trans_huge(pmde);
-}
-
 static inline bool not_found(struct page_vma_mapped_walk *pvmw)
 {
 	page_vma_mapped_walk_done(pvmw);

@@ -106,6 +95,7 @@ bool page_vma_mapped_walk(struct page_vm
 	pgd_t *pgd;
 	p4d_t *p4d;
 	pud_t *pud;
+	pmd_t pmde;
 
 	/* The only possible pmd mapping has been handled on last iteration */
 	if (pvmw->pmd && !pvmw->pte)

@@ -138,7 +128,13 @@ restart:
 	if (!pud_present(*pud))
 		return false;
 	pvmw->pmd = pmd_offset(pud, pvmw->address);
-	if (pmd_trans_huge(*pvmw->pmd)) {
+	/*
+	 * Make sure the pmd value isn't cached in a register by the
+	 * compiler and used as a stale value after we've observed a
+	 * subsequent update.
+	 */
+	pmde = READ_ONCE(*pvmw->pmd);
+	if (pmd_trans_huge(pmde)) {
 		pvmw->ptl = pmd_lock(mm, pvmw->pmd);
 		if (!pmd_present(*pvmw->pmd))
 			return not_found(pvmw);

@@ -153,9 +149,8 @@ restart:
 			spin_unlock(pvmw->ptl);
 			pvmw->ptl = NULL;
 		}
-	} else {
-		if (!check_pmd(pvmw))
-			return false;
+	} else if (!pmd_present(pmde)) {
+		return false;
 	}
 	if (!map_pte(pvmw))
 		goto next_pte;

@@ -2031,6 +2031,8 @@ static const struct hid_device_id hid_ha
 	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_BM084) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_DEFT_WIRED) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_DEFT_WIRELESS) },
+	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_HUGE_WIRED) },
+	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_HUGE_WIRELESS) },
 #endif
 #if IS_ENABLED(CONFIG_HID_ELO)
 	{ HID_USB_DEVICE(USB_VENDOR_ID_ELO, 0x0009) },

@@ -3,6 +3,7 @@
  *  Copyright (c) 2010 Richard Nauber <[email protected]>
  *  Copyright (c) 2016 Yuxuan Shui <[email protected]>
  *  Copyright (c) 2017 Diego Elio Pettenò <[email protected]>
+ *  Copyright (c) 2017 Alex Manoussakis <[email protected]>
  */
 
 /*

@@ -32,9 +33,11 @@ static __u8 *elecom_report_fixup(struct
 		break;
 	case USB_DEVICE_ID_ELECOM_DEFT_WIRED:
 	case USB_DEVICE_ID_ELECOM_DEFT_WIRELESS:
-		/* The DEFT trackball has eight buttons, but its descriptor only
-		 * reports five, disabling the three Fn buttons on the top of
-		 * the mouse.
+	case USB_DEVICE_ID_ELECOM_HUGE_WIRED:
+	case USB_DEVICE_ID_ELECOM_HUGE_WIRELESS:
+		/* The DEFT/HUGE trackball has eight buttons, but its descriptor
+		 * only reports five, disabling the three Fn buttons on the top
+		 * of the mouse.
 		 *
 		 * Apply the following diff to the descriptor:
 		 *

@@ -62,7 +65,7 @@ static __u8 *elecom_report_fixup(struct
 		 * End Collection,                     End Collection,
 		 */
 		if (*rsize == 213 && rdesc[13] == 5 && rdesc[21] == 5) {
-			hid_info(hdev, "Fixing up Elecom DEFT Fn buttons\n");
+			hid_info(hdev, "Fixing up Elecom DEFT/HUGE Fn buttons\n");
 			rdesc[13] = 8; /* Button/Variable Report Count */
 			rdesc[21] = 8; /* Button/Variable Usage Maximum */
 			rdesc[29] = 0; /* Button/Constant Report Count */

@@ -76,6 +79,8 @@ static const struct hid_device_id elecom
 	{ HID_BLUETOOTH_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_BM084) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_DEFT_WIRED) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_DEFT_WIRELESS) },
+	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_HUGE_WIRED) },
+	{ HID_USB_DEVICE(USB_VENDOR_ID_ELECOM, USB_DEVICE_ID_ELECOM_HUGE_WIRELESS) },
 	{ }
 };
 MODULE_DEVICE_TABLE(hid, elecom_devices);

@@ -366,6 +366,8 @@
 #define USB_DEVICE_ID_ELECOM_BM084	0x0061
 #define USB_DEVICE_ID_ELECOM_DEFT_WIRED	0x00fe
 #define USB_DEVICE_ID_ELECOM_DEFT_WIRELESS	0x00ff
+#define USB_DEVICE_ID_ELECOM_HUGE_WIRED	0x010c
+#define USB_DEVICE_ID_ELECOM_HUGE_WIRELESS	0x010d
 
 #define USB_VENDOR_ID_DREAM_CHEEKY	0x1d34
 #define USB_DEVICE_ID_DREAM_CHEEKY_WN	0x0004

@@ -177,6 +177,11 @@ int vmbus_open(struct vmbus_channel *new
 		      &vmbus_connection.chn_msg_list);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 
+	if (newchannel->rescind) {
+		err = -ENODEV;
+		goto error_free_gpadl;
+	}
+
 	ret = vmbus_post_msg(open_msg,
 			     sizeof(struct vmbus_channel_open_channel), true);
 

@@ -421,6 +426,11 @@ int vmbus_establish_gpadl(struct vmbus_c
 
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 
+	if (channel->rescind) {
+		ret = -ENODEV;
+		goto cleanup;
+	}
+
 	ret = vmbus_post_msg(gpadlmsg, msginfo->msgsize -
 			     sizeof(*msginfo), true);
 	if (ret != 0)

@@ -494,6 +504,10 @@ int vmbus_teardown_gpadl(struct vmbus_ch
 	list_add_tail(&info->msglistentry,
 		      &vmbus_connection.chn_msg_list);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
+
+	if (channel->rescind)
+		goto post_msg_err;
+
 	ret = vmbus_post_msg(msg, sizeof(struct vmbus_channel_gpadl_teardown),
 			     true);
 

@@ -451,6 +451,12 @@ static void vmbus_process_offer(struct v
 	/* Make sure this is a new offer */
 	mutex_lock(&vmbus_connection.channel_mutex);
 
+	/*
+	 * Now that we have acquired the channel_mutex,
+	 * we can release the potentially racing rescind thread.
+	 */
+	atomic_dec(&vmbus_connection.offer_in_progress);
+
 	list_for_each_entry(channel, &vmbus_connection.chn_list, listentry) {
 		if (!uuid_le_cmp(channel->offermsg.offer.if_type,
 			newchannel->offermsg.offer.if_type) &&

@@ -481,7 +487,6 @@ static void vmbus_process_offer(struct v
 			channel->num_sc++;
 			spin_unlock_irqrestore(&channel->lock, flags);
 		} else {
-			atomic_dec(&vmbus_connection.offer_in_progress);
 			goto err_free_chan;
 		}
 	}

@@ -510,7 +515,6 @@ static void vmbus_process_offer(struct v
 	if (!fnew) {
 		if (channel->sc_creation_callback != NULL)
 			channel->sc_creation_callback(newchannel);
-		atomic_dec(&vmbus_connection.offer_in_progress);
 		return;
 	}
 

@@ -541,7 +545,7 @@ static void vmbus_process_offer(struct v
 		goto err_deq_chan;
 	}
 
-	atomic_dec(&vmbus_connection.offer_in_progress);
+	newchannel->probe_done = true;
 	return;
 
 err_deq_chan:

@@ -882,8 +886,27 @@ static void vmbus_onoffer_rescind(struct
 	channel->rescind = true;
 	spin_unlock_irqrestore(&channel->lock, flags);
 
+	/*
+	 * Now that we have posted the rescind state, perform
+	 * rescind related cleanup.
+	 */
 	vmbus_rescind_cleanup(channel);
 
+	/*
+	 * Now wait for offer handling to complete.
+	 */
+	while (READ_ONCE(channel->probe_done) == false) {
+		/*
+		 * We wait here until any channel offer is currently
+		 * being processed.
+		 */
+		msleep(1);
+	}
+
+	/*
+	 * At this point, the rescind handling can proceed safely.
+	 */
+
 	if (channel->device_obj) {
 		if (channel->chn_rescind_callback) {
 			channel->chn_rescind_callback(channel);

@@ -940,6 +940,9 @@ static void vmbus_chan_sched(struct hv_p
 			if (channel->offermsg.child_relid != relid)
 				continue;
 
+			if (channel->rescind)
+				continue;
+
 			switch (channel->callback_mode) {
 			case HV_CALL_ISR:
 				vmbus_channel_isr(channel);

@@ -895,6 +895,8 @@ struct vmbus_channel {
 	 */
 	enum hv_numa_policy affinity_policy;
 
+	bool probe_done;
+
 };
 
 static inline bool is_hvsock_channel(const struct vmbus_channel *c)

@@ -640,6 +640,7 @@ void vmbus_close(struct vmbus_channel *c
 		 */
 		return;
 	}
+	mutex_lock(&vmbus_connection.channel_mutex);
 	/*
 	 * Close all the sub-channels first and then close the
 	 * primary channel.

@@ -648,16 +649,15 @@ void vmbus_close(struct vmbus_channel *c
 		cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
 		vmbus_close_internal(cur_channel);
 		if (cur_channel->rescind) {
-			mutex_lock(&vmbus_connection.channel_mutex);
-			hv_process_channel_removal(cur_channel,
+			hv_process_channel_removal(
 					   cur_channel->offermsg.child_relid);
-			mutex_unlock(&vmbus_connection.channel_mutex);
 		}
 	}
 	/*
 	 * Now close the primary.
 	 */
 	vmbus_close_internal(channel);
+	mutex_unlock(&vmbus_connection.channel_mutex);
 }
 EXPORT_SYMBOL_GPL(vmbus_close);
 

@@ -159,7 +159,7 @@ static void vmbus_rescind_cleanup(struct
 
 
 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
-
+	channel->rescind = true;
 	list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
 				msglistentry) {
 

@@ -381,14 +381,21 @@ static void vmbus_release_relid(u32 reli
 		       true);
 }
 
-void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid)
+void hv_process_channel_removal(u32 relid)
 {
 	unsigned long flags;
-	struct vmbus_channel *primary_channel;
+	struct vmbus_channel *primary_channel, *channel;
 
-	BUG_ON(!channel->rescind);
 	BUG_ON(!mutex_is_locked(&vmbus_connection.channel_mutex));
 
+	/*
+	 * Make sure channel is valid as we may have raced.
+	 */
+	channel = relid2channel(relid);
+	if (!channel)
+		return;
+
+	BUG_ON(!channel->rescind);
 	if (channel->target_cpu != get_cpu()) {
 		put_cpu();
 		smp_call_function_single(channel->target_cpu,

@@ -515,6 +522,7 @@ static void vmbus_process_offer(struct v
 	if (!fnew) {
 		if (channel->sc_creation_callback != NULL)
 			channel->sc_creation_callback(newchannel);
+		newchannel->probe_done = true;
 		return;
 	}
 

@@ -843,7 +851,6 @@ static void vmbus_onoffer_rescind(struct
 {
 	struct vmbus_channel_rescind_offer *rescind;
 	struct vmbus_channel *channel;
-	unsigned long flags;
 	struct device *dev;
 
 	rescind = (struct vmbus_channel_rescind_offer *)hdr;

@@ -882,16 +889,6 @@ static void vmbus_onoffer_rescind(struct
 		return;
 	}
 
-	spin_lock_irqsave(&channel->lock, flags);
-	channel->rescind = true;
-	spin_unlock_irqrestore(&channel->lock, flags);
-
-	/*
-	 * Now that we have posted the rescind state, perform
-	 * rescind related cleanup.
-	 */
-	vmbus_rescind_cleanup(channel);
-
 	/*
 	 * Now wait for offer handling to complete.
 	 */

@@ -910,6 +907,7 @@ static void vmbus_onoffer_rescind(struct
 	if (channel->device_obj) {
 		if (channel->chn_rescind_callback) {
 			channel->chn_rescind_callback(channel);
+			vmbus_rescind_cleanup(channel);
 			return;
 		}
 		/*

@@ -918,6 +916,7 @@ static void vmbus_onoffer_rescind(struct
 		 */
 		dev = get_device(&channel->device_obj->device);
 		if (dev) {
+			vmbus_rescind_cleanup(channel);
 			vmbus_device_unregister(channel->device_obj);
 			put_device(dev);
 		}

@@ -930,16 +929,16 @@ static void vmbus_onoffer_rescind(struct
 		 * 1. Close all sub-channels first
 		 * 2. Then close the primary channel.
 		 */
+		mutex_lock(&vmbus_connection.channel_mutex);
+		vmbus_rescind_cleanup(channel);
 		if (channel->state == CHANNEL_OPEN_STATE) {
 			/*
 			 * The channel is currently not open;
 			 * it is safe for us to cleanup the channel.
 			 */
-			mutex_lock(&vmbus_connection.channel_mutex);
-			hv_process_channel_removal(channel,
-						channel->offermsg.child_relid);
-			mutex_unlock(&vmbus_connection.channel_mutex);
+			hv_process_channel_removal(rescind->child_relid);
 		}
+		mutex_unlock(&vmbus_connection.channel_mutex);
 	}
 }
 

@@ -768,8 +768,7 @@ static void vmbus_device_release(struct
 	struct vmbus_channel *channel = hv_dev->channel;
 
 	mutex_lock(&vmbus_connection.channel_mutex);
-	hv_process_channel_removal(channel,
-				   channel->offermsg.child_relid);
+	hv_process_channel_removal(channel->offermsg.child_relid);
 	mutex_unlock(&vmbus_connection.channel_mutex);
 	kfree(hv_dev);
 

@@ -1455,7 +1455,7 @@ extern bool vmbus_prep_negotiate_resp(st
 				const int *srv_version, int srv_vercnt,
 				int *nego_fw_version, int *nego_srv_version);
 
-void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid);
+void hv_process_channel_removal(u32 relid);
 
 void vmbus_setevent(struct vmbus_channel *channel);
 /*

@@ -94,30 +94,6 @@ hv_set_next_write_location(struct hv_rin
 	ring_info->ring_buffer->write_index = next_write_location;
 }
 
-/* Get the next read location for the specified ring buffer. */
-static inline u32
-hv_get_next_read_location(const struct hv_ring_buffer_info *ring_info)
-{
-	return ring_info->ring_buffer->read_index;
-}
-
-/*
- * Get the next read location + offset for the specified ring buffer.
- * This allows the caller to skip.
- */
-static inline u32
-hv_get_next_readlocation_withoffset(const struct hv_ring_buffer_info *ring_info,
-				    u32 offset)
-{
-	u32 next = ring_info->ring_buffer->read_index;
-
-	next += offset;
-	if (next >= ring_info->ring_datasize)
-		next -= ring_info->ring_datasize;
-
-	return next;
-}
-
 /* Set the next read location for the specified ring buffer. */
 static inline void
 hv_set_next_read_location(struct hv_ring_buffer_info *ring_info,

@@ -142,29 +118,6 @@ hv_get_ring_bufferindices(struct hv_ring
 }
 
 /*
- * Helper routine to copy to source from ring buffer.
- * Assume there is enough room. Handles wrap-around in src case only!!
- */
-static u32 hv_copyfrom_ringbuffer(
-	const struct hv_ring_buffer_info *ring_info,
-	void				*dest,
-	u32				destlen,
-	u32				start_read_offset)
-{
-	void *ring_buffer = hv_get_ring_buffer(ring_info);
-	u32 ring_buffer_size = hv_get_ring_buffersize(ring_info);
-
-	memcpy(dest, ring_buffer + start_read_offset, destlen);
-
-	start_read_offset += destlen;
-	if (start_read_offset >= ring_buffer_size)
-		start_read_offset -= ring_buffer_size;
-
-	return start_read_offset;
-}
-
-
-/*
  * Helper routine to copy from source to ring buffer.
  * Assume there is enough room. Handles wrap-around in dest case only!!
  */

@@ -334,33 +287,22 @@ int hv_ringbuffer_write(struct vmbus_cha
 	return 0;
 }
 
-static inline void
-init_cached_read_index(struct hv_ring_buffer_info *rbi)
-{
-	rbi->cached_read_index = rbi->ring_buffer->read_index;
-}
-
 int hv_ringbuffer_read(struct vmbus_channel *channel,
 		       void *buffer, u32 buflen, u32 *buffer_actual_len,
 		       u64 *requestid, bool raw)
 {
-	u32 bytes_avail_toread;
-	u32 next_read_location;
-	u64 prev_indices = 0;
-	struct vmpacket_descriptor desc;
-	u32 offset;
-	u32 packetlen;
-	struct hv_ring_buffer_info *inring_info = &channel->inbound;
+	struct vmpacket_descriptor *desc;
+	u32 packetlen, offset;
 
-	if (buflen <= 0)
+	if (unlikely(buflen == 0))
 		return -EINVAL;
 
 	*buffer_actual_len = 0;
 	*requestid = 0;
 
-	bytes_avail_toread = hv_get_bytes_to_read(inring_info);
 	/* Make sure there is something to read */
-	if (bytes_avail_toread < sizeof(desc)) {
+	desc = hv_pkt_iter_first(channel);
+	if (desc == NULL) {
 		/*
 		 * No error is set when there is even no header, drivers are
 		 * supposed to analyze buffer_actual_len.

@@ -368,48 +310,22 @@ int hv_ringbuffer_read(struct vmbus_chan
 		return 0;
 	}
 
-	init_cached_read_index(inring_info);
-
-	next_read_location = hv_get_next_read_location(inring_info);
-	next_read_location = hv_copyfrom_ringbuffer(inring_info, &desc,
-						    sizeof(desc),
-						    next_read_location);
-
-	offset = raw ? 0 : (desc.offset8 << 3);
-	packetlen = (desc.len8 << 3) - offset;
+	offset = raw ? 0 : (desc->offset8 << 3);
+	packetlen = (desc->len8 << 3) - offset;
 	*buffer_actual_len = packetlen;
-	*requestid = desc.trans_id;
-
-	if (bytes_avail_toread < packetlen + offset)
-		return -EAGAIN;
+	*requestid = desc->trans_id;
 
-	if (packetlen > buflen)
+	if (unlikely(packetlen > buflen))
 		return -ENOBUFS;
 
-	next_read_location =
-		hv_get_next_readlocation_withoffset(inring_info, offset);
+	/* since ring is double mapped, only one copy is necessary */
+	memcpy(buffer, (const char *)desc + offset, packetlen);
 
-	next_read_location = hv_copyfrom_ringbuffer(inring_info,
-						buffer,
-						packetlen,
-						next_read_location);
-
-	next_read_location = hv_copyfrom_ringbuffer(inring_info,
-						&prev_indices,
-						sizeof(u64),
-						next_read_location);
+	/* Advance ring index to next packet descriptor */
+	__hv_pkt_iter_next(channel, desc);
 
-	/*
-	 * Make sure all reads are done before we update the read index since
-	 * the writer may start writing to the read area once the read index
-	 * is updated.
-	 */
-	virt_mb();
-
-	/* Update the read index */
-	hv_set_next_read_location(inring_info, next_read_location);
-
-	hv_signal_on_read(channel);
+	/* Notify host of update */
+	hv_pkt_iter_close(channel);
 
 	return 0;
 }

@@ -442,7 +358,7 @@ struct vmpacket_descriptor *hv_pkt_iter_
 	struct hv_ring_buffer_info *rbi = &channel->inbound;
 
 	/* set state for later hv_signal_on_read() */
-	init_cached_read_index(rbi);
+	rbi->cached_read_index = rbi->ring_buffer->read_index;
 
 	if (hv_pkt_iter_avail(rbi) < sizeof(struct vmpacket_descriptor))
 		return NULL;

@@ -29,6 +29,7 @@
 #include <linux/uio.h>
 #include <linux/vmalloc.h>
 #include <linux/slab.h>
+#include <linux/prefetch.h>
 
 #include "hyperv_vmbus.h"
 

@@ -357,7 +358,7 @@ struct vmpacket_descriptor *hv_pkt_iter_
 {
 	struct hv_ring_buffer_info *rbi = &channel->inbound;
 
-	/* set state for later hv_signal_on_read() */
+	/* set state for later hv_pkt_iter_close */
 	rbi->cached_read_index = rbi->ring_buffer->read_index;
 
 	if (hv_pkt_iter_avail(rbi) < sizeof(struct vmpacket_descriptor))

@@ -400,6 +401,8 @@ EXPORT_SYMBOL_GPL(__hv_pkt_iter_next);
 void hv_pkt_iter_close(struct vmbus_channel *channel)
 {
 	struct hv_ring_buffer_info *rbi = &channel->inbound;
+	u32 cur_write_sz, cached_write_sz;
+	u32 pending_sz;
 
 	/*
 	 * Make sure all reads are done before we update the read index since

@@ -409,6 +412,31 @@ void hv_pkt_iter_close(struct vmbus_chan
 	virt_rmb();
 	rbi->ring_buffer->read_index = rbi->priv_read_index;
 
-	hv_signal_on_read(channel);
+	/*
+	 * Issue a full memory barrier before making the signaling decision.
+	 * Here is the reason for having this barrier:
+	 * If the reading of the pend_sz (in this function)
+	 * were to be reordered and read before we commit the new read
+	 * index (in the calling function)  we could
+	 * have a problem. If the host were to set the pending_sz after we
+	 * have sampled pending_sz and go to sleep before we commit the
+	 * read index, we could miss sending the interrupt. Issue a full
+	 * memory barrier to address this.
+	 */
+	virt_mb();
+
+	pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
+	/* If the other end is not blocked on write don't bother. */
+	if (pending_sz == 0)
+		return;
+
+	cur_write_sz = hv_get_bytes_to_write(rbi);
+
+	if (cur_write_sz < pending_sz)
+		return;
+
+	cached_write_sz = hv_get_cached_bytes_to_write(rbi);
+	if (cached_write_sz < pending_sz)
+		vmbus_setevent(channel);
 }
 EXPORT_SYMBOL_GPL(hv_pkt_iter_close);

55 lines not on this page · raw message

@@ -358,9 +358,6 @@ struct vmpacket_descriptor *hv_pkt_iter_
 {
 	struct hv_ring_buffer_info *rbi = &channel->inbound;
 
-	/* set state for later hv_pkt_iter_close */
-	rbi->cached_read_index = rbi->ring_buffer->read_index;
-
 	if (hv_pkt_iter_avail(rbi) < sizeof(struct vmpacket_descriptor))
 		return NULL;
 

@@ -388,10 +385,7 @@ __hv_pkt_iter_next(struct vmbus_channel
 		rbi->priv_read_index -= dsize;
 
 	/* more data? */
-	if (hv_pkt_iter_avail(rbi) < sizeof(struct vmpacket_descriptor))
-		return NULL;
-	else
-		return hv_get_ring_buffer(rbi) + rbi->priv_read_index;
+	return hv_pkt_iter_first(channel);
 }
 EXPORT_SYMBOL_GPL(__hv_pkt_iter_next);
 

@@ -401,7 +395,7 @@ EXPORT_SYMBOL_GPL(__hv_pkt_iter_next);
 void hv_pkt_iter_close(struct vmbus_channel *channel)
 {
 	struct hv_ring_buffer_info *rbi = &channel->inbound;
-	u32 cur_write_sz, cached_write_sz;
+	u32 orig_write_sz = hv_get_bytes_to_write(rbi);
 	u32 pending_sz;
 
 	/*

@@ -430,13 +424,10 @@ void hv_pkt_iter_close(struct vmbus_chan
 	if (pending_sz == 0)
 		return;
 
-	cur_write_sz = hv_get_bytes_to_write(rbi);
-
-	if (cur_write_sz < pending_sz)
+	if (hv_get_bytes_to_write(rbi) < pending_sz)
 		return;
 
-	cached_write_sz = hv_get_cached_bytes_to_write(rbi);
-	if (cached_write_sz < pending_sz)
+	if (orig_write_sz < pending_sz)
 		vmbus_setevent(channel);
 }
 EXPORT_SYMBOL_GPL(hv_pkt_iter_close);

@@ -127,7 +127,6 @@ struct hv_ring_buffer_info {
 	u32 ring_data_startoffset;
 	u32 priv_write_index;
 	u32 priv_read_index;
-	u32 cached_read_index;
 };
 
 /*

@@ -180,19 +179,6 @@ static inline u32 hv_get_bytes_to_write(
 	return write;
 }
 
-static inline u32 hv_get_cached_bytes_to_write(
-	const struct hv_ring_buffer_info *rbi)
-{
-	u32 read_loc, write_loc, dsize, write;
-
-	dsize = rbi->ring_datasize;
-	read_loc = rbi->cached_read_index;
-	write_loc = rbi->ring_buffer->write_index;
-
-	write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
-		read_loc - write_loc;
-	return write;
-}
 /*
  * VMBUS version is 32 bit entity broken up into
  * two 16 bit quantities: major_number. minor_number.

@@ -396,7 +396,6 @@ void hv_pkt_iter_close(struct vmbus_chan
 {
 	struct hv_ring_buffer_info *rbi = &channel->inbound;
 	u32 orig_write_sz = hv_get_bytes_to_write(rbi);
-	u32 pending_sz;
 
 	/*
 	 * Make sure all reads are done before we update the read index since

@@ -419,15 +418,27 @@ void hv_pkt_iter_close(struct vmbus_chan
 	 */
 	virt_mb();
 
-	pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
-	/* If the other end is not blocked on write don't bother. */
-	if (pending_sz == 0)
+	/* If host has disabled notifications then skip */
+	if (rbi->ring_buffer->interrupt_mask)
 		return;
 
-	if (hv_get_bytes_to_write(rbi) < pending_sz)
-		return;
+	if (rbi->ring_buffer->feature_bits.feat_pending_send_sz) {
+		u32 pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
+
+		/*
+		 * If there was space before we began iteration,
+		 * then host was not blocked. Also handles case where
+		 * pending_sz is zero then host has nothing pending
+		 * and does not need to be signaled.
+		 */
+		if (orig_write_sz > pending_sz)
+			return;
+
+		/* If pending write will not fit, don't give false hope. */
+		if (hv_get_bytes_to_write(rbi) < pending_sz)
+			return;
+	}
 
-	if (orig_write_sz < pending_sz)
-		vmbus_setevent(channel);
+	vmbus_setevent(channel);
 }
 EXPORT_SYMBOL_GPL(hv_pkt_iter_close);