Drivers: hv: vmbus: Resume after fixing up old primary channels

When the host re-offers the primary channels upon resume, the host only guarantees the Instance GUID doesn't change, so vmbus_bus_suspend() should invalidate channel->offermsg.child_relid and figure out the number of primary channels that need to be fixed up upon resume. Upon resume, vmbus_onoffer() finds the old channel structs, and maps the new offers to the old channels, and fixes up the old structs, and finally the resume callbacks of the VSC drivers will re-open the channels.

Dexuan Cui committed 7 years ago

Commit d8bd2d41 parent b307b385 files · +101 −20Browse files

Unified · Split

5 files

drivers/hv/channel_mgmt.c +65 −20
drivers/hv/connection.c +2 −0
drivers/hv/hyperv_vmbus.h +14 −0
drivers/hv/vmbus_drv.c +17 −0
include/linux/hyperv.h +3 −0

drivers/hv/channel_mgmt.c+65−20

@@ -407,7 +407,15 @@ void hv_process_channel_removal(struct vmbus_channel *channel) · 406 unmodified lines
407	cpumask_clear_cpu(channel->target_cpu,	407	cpumask_clear_cpu(channel->target_cpu,
408	&primary_channel->alloced_cpus_in_node);	408	&primary_channel->alloced_cpus_in_node);
409		409
410	vmbus_release_relid(channel->offermsg.child_relid);	410	/*
		411	* Upon suspend, an in-use hv_sock channel is marked as "rescinded" and
		412	* the relid is invalidated; after hibernation, when the user-space app
		413	* destroys the channel, the relid is INVALID_RELID, and in this case
		414	* it's unnecessary and unsafe to release the old relid, since the same
		415	* relid can refer to a completely different channel now.
		416	*/
		417	if (channel->offermsg.child_relid != INVALID_RELID)
		418	vmbus_release_relid(channel->offermsg.child_relid);
411		419
412	free_channel(channel);	420	free_channel(channel);
413	}	421	}
@@ -851,6 +859,36 @@ void vmbus_initiate_unload(bool crash) · 437 unmodified lines
851	vmbus_wait_for_unload();	859	vmbus_wait_for_unload();
852	}	860	}
853		861
		862	static void check_ready_for_resume_event(void)
		863	{
		864	/*
		865	* If all the old primary channels have been fixed up, then it's safe
		866	* to resume.
		867	*/
		868	if (atomic_dec_and_test(&vmbus_connection.nr_chan_fixup_on_resume))
		869	complete(&vmbus_connection.ready_for_resume_event);
		870	}
		871
		872	static void vmbus_setup_channel_state(struct vmbus_channel *channel,
		873	struct vmbus_channel_offer_channel *offer)
		874	{
		875	/*
		876	* Setup state for signalling the host.
		877	*/
		878	channel->sig_event = VMBUS_EVENT_CONNECTION_ID;
		879
		880	if (vmbus_proto_version != VERSION_WS2008) {
		881	channel->is_dedicated_interrupt =
		882	(offer->is_dedicated_interrupt != 0);
		883	channel->sig_event = offer->connection_id;
		884	}
		885
		886	memcpy(&channel->offermsg, offer,
		887	sizeof(struct vmbus_channel_offer_channel));
		888	channel->monitor_grp = (u8)offer->monitorid / 32;
		889	channel->monitor_bit = (u8)offer->monitorid % 32;
		890	}
		891
854	/*	892	/*
855	* find_primary_channel_by_offer - Get the channel object given the new offer.	893	* find_primary_channel_by_offer - Get the channel object given the new offer.
856	* This is only used in the resume path of hibernation.	894	* This is only used in the resume path of hibernation.
@@ -902,14 +940,29 @@ static void vmbus_onoffer(struct vmbus_channel_message_header *hdr) · 45 unmodified lines
902	atomic_dec(&vmbus_connection.offer_in_progress);	940	atomic_dec(&vmbus_connection.offer_in_progress);
903		941
904	/*	942	/*
905	* We're resuming from hibernation: we expect the host to send	943	* We're resuming from hibernation: all the sub-channel and
906	* exactly the same offers that we had before the hibernation.	944	* hv_sock channels we had before the hibernation should have
		945	* been cleaned up, and now we must be seeing a re-offered
		946	* primary channel that we had before the hibernation.
907	*/	947	*/
		948
		949	WARN_ON(oldchannel->offermsg.child_relid != INVALID_RELID);
		950	/* Fix up the relid. */
		951	oldchannel->offermsg.child_relid = offer->child_relid;
		952
908	offer_sz = sizeof(*offer);	953	offer_sz = sizeof(*offer);
909	if (memcmp(offer, &oldchannel->offermsg, offer_sz) == 0)	954	if (memcmp(offer, &oldchannel->offermsg, offer_sz) == 0) {
		955	check_ready_for_resume_event();
910	return;	956	return;
		957	}
911		958
912	pr_debug("Mismatched offer from the host (relid=%d)\n",	959	/*
		960	* This is not an error, since the host can also change the
		961	* other field(s) of the offer, e.g. on WS RS5 (Build 17763),
		962	* the offer->connection_id of the Mellanox VF vmbus device
		963	* can change when the host reoffers the device upon resume.
		964	*/
		965	pr_debug("vmbus offer changed: relid=%d\n",
913	offer->child_relid);	966	offer->child_relid);
914		967
915	print_hex_dump_debug("Old vmbus offer: ", DUMP_PREFIX_OFFSET,	968	print_hex_dump_debug("Old vmbus offer: ", DUMP_PREFIX_OFFSET,
@@ -917,6 +970,12 @@ static void vmbus_onoffer(struct vmbus_channel_message_header *hdr) · 1 unmodified line
917	false);	970	false);
918	print_hex_dump_debug("New vmbus offer: ", DUMP_PREFIX_OFFSET,	971	print_hex_dump_debug("New vmbus offer: ", DUMP_PREFIX_OFFSET,
919	16, 4, offer, offer_sz, false);	972	16, 4, offer, offer_sz, false);
		973
		974	/* Fix up the old channel. */
		975	vmbus_setup_channel_state(oldchannel, offer);
		976
		977	check_ready_for_resume_event();
		978
920	return;	979	return;
921	}	980	}
922		981
@@ -929,21 +988,7 @@ static void vmbus_onoffer(struct vmbus_channel_message_header *hdr) · 6 unmodified lines
929	return;	988	return;
930	}	989	}
931		990
932	/*	991	vmbus_setup_channel_state(newchannel, offer);
933	* Setup state for signalling the host.
934	*/
935	newchannel->sig_event = VMBUS_EVENT_CONNECTION_ID;
936
937	if (vmbus_proto_version != VERSION_WS2008) {
938	newchannel->is_dedicated_interrupt =
939	(offer->is_dedicated_interrupt != 0);
940	newchannel->sig_event = offer->connection_id;
941	}
942
943	memcpy(&newchannel->offermsg, offer,
944	sizeof(struct vmbus_channel_offer_channel));
945	newchannel->monitor_grp = (u8)offer->monitorid / 32;
946	newchannel->monitor_bit = (u8)offer->monitorid % 32;
947		992
948	vmbus_process_offer(newchannel);	993	vmbus_process_offer(newchannel);
949	}	994	}

drivers/hv/connection.c+2−0

@@ -29,6 +29,8 @@ struct vmbus_connection vmbus_connection = { · 28 unmodified lines
29		29
30	.ready_for_suspend_event= COMPLETION_INITIALIZER(	30	.ready_for_suspend_event= COMPLETION_INITIALIZER(
31	vmbus_connection.ready_for_suspend_event),	31	vmbus_connection.ready_for_suspend_event),
		32	.ready_for_resume_event = COMPLETION_INITIALIZER(
		33	vmbus_connection.ready_for_resume_event),
32	};	34	};
33	EXPORT_SYMBOL_GPL(vmbus_connection);	35	EXPORT_SYMBOL_GPL(vmbus_connection);
34		36

drivers/hv/hyperv_vmbus.h+14−0

@@ -272,6 +272,20 @@ struct vmbus_connection { · 271 unmodified lines
272	* drop to zero.	272	* drop to zero.
273	*/	273	*/
274	struct completion ready_for_suspend_event;	274	struct completion ready_for_suspend_event;
		275
		276	/*
		277	* The number of primary channels that should be "fixed up"
		278	* upon resume: these channels are re-offered upon resume, and some
		279	* fields of the channel offers (i.e. child_relid and connection_id)
		280	* can change, so the old offermsg must be fixed up, before the resume
		281	* callbacks of the VSC drivers start to further touch the channels.
		282	*/
		283	atomic_t nr_chan_fixup_on_resume;
		284	/*
		285	* vmbus_bus_resume() waits for "nr_chan_fixup_on_resume" to
		286	* drop to zero.
		287	*/
		288	struct completion ready_for_resume_event;
275	};	289	};
276		290
277		291

drivers/hv/vmbus_drv.c+17−0

 	if (atomic_read(&vmbus_connection.nr_chan_close_on_suspend) > 0)
 		wait_for_completion(&vmbus_connection.ready_for_suspend_event);
+	WARN_ON(atomic_read(&vmbus_connection.nr_chan_fixup_on_resume) != 0);
 	mutex_lock(&vmbus_connection.channel_mutex);
 	list_for_each_entry(channel, &vmbus_connection.chn_list, listentry) {
+		/*
+		 * Invalidate the field. Upon resume, vmbus_onoffer() will fix
+		 * up the field, and the other fields (if necessary).
+		 */
+		channel->offermsg.child_relid = INVALID_RELID;
 		if (is_hvsock_channel(channel)) {
 			if (!channel->rescind) {
 				pr_err("hv_sock channel not rescinded!\n");
 			WARN_ON_ONCE(1);
 		}
 		spin_unlock_irqrestore(&channel->lock, flags);
+		atomic_inc(&vmbus_connection.nr_chan_fixup_on_resume);
 	}
 	mutex_unlock(&vmbus_connection.channel_mutex);
 	vmbus_connection.conn_state = DISCONNECTED;
+	/* Reset the event for the next resume. */
+	reinit_completion(&vmbus_connection.ready_for_resume_event);
 	return 0;
 }
 	if (ret != 0)
 		return ret;
+	WARN_ON(atomic_read(&vmbus_connection.nr_chan_fixup_on_resume) == 0);
 	vmbus_request_offers();
+	wait_for_completion(&vmbus_connection.ready_for_resume_event);
 	/* Reset the event for the next suspend. */
 	reinit_completion(&vmbus_connection.ready_for_suspend_event);

include/linux/hyperv.h+3−0

@@ -426,6 +426,9 @@ enum vmbus_channel_message_type { · 425 unmodified lines
426	CHANNELMSG_COUNT	426	CHANNELMSG_COUNT
427	};	427	};
428		428
		429	/* Hyper-V supports about 2048 channels, and the RELIDs start with 1. */
		430	#define INVALID_RELID U32_MAX
		431
429	struct vmbus_channel_message_header {	432	struct vmbus_channel_message_header {
430	enum vmbus_channel_message_type msgtype;	433	enum vmbus_channel_message_type msgtype;
431	u32 padding;	434	u32 padding;

Drivers: hv: vmbus: Resume after fixing up old primary channels

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code