Skip to contentSkip to conversation

Drivers: hv: vmbus: additional fixes for the setup/teardown path

PATCH Superseded Vitaly Kuznetsov 3 participants 8 messages Apr 21, 2015 Hyper-V/Azure core and drivers

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover Drivers: hv: vmbus: additional fixes for the setup/teardown path

K. Y.,

here are 3 additional patches for your "[PATCH 0/5] Drivers: hv: vmbus: Cleanup the vmbus unload path" series (with the fix I suggested). I tested the whole set on Gen2 Win2012R2 guest, load/unload path seems being stable. Can you please take a look?

Thanks,

Vitaly Kuznetsov (3):
Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths
Drivers: hv: vmbus: kill tasklets on module unload
Drivers: hv: vmbus: setup irq after synic is initialized

 drivers/hv/channel.c   |  7 ++++---
 drivers/hv/vmbus_drv.c | 10 +++++++---
 2 files changed, 11 insertions(+), 6 deletions(-)
Signature · 1 line
1.9.3
#1 lore raw

PATCH 1/3 Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

In case there was an error reported in the response to the CHANNELMSG_OPENCHANNEL call we need to do the cleanup as a vmbus_open() user won't be doing it after receiving an error. The cleanup should be done on all failure paths.

---
 drivers/hv/channel.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/hv/channel.c +4 -3
index 54da66d..836386f 100644
@@ -186,11 +186,12 @@ int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
 	list_del(&open_info->msglistentry);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 
-	if (err == 0)
-		newchannel->state = CHANNEL_OPENED_STATE;
+	if (err != 0)
+		goto error_gpadl;
 
+	newchannel->state = CHANNEL_OPENED_STATE;
 	kfree(open_info);
-	return err;
+	return 0;
 
 error1:
 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
Signature · 1 line
1.9.3
maintainer
#2 lore raw

RE: [PATCH 1/3] Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

Quoted 40 lines

-----Original Message-----
From: Vitaly Kuznetsov [mailto:[email protected]]
Sent: Tuesday, April 21, 2015 16:18
To: KY Srinivasan
Cc: Haiyang Zhang; [email protected]; linux- [email protected]; Dexuan Cui
Subject: [PATCH 1/3] Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

In case there was an error reported in the response to the CHANNELMSG_OPENCHANNEL
call we need to do the cleanup as a vmbus_open() user won't be doing it after
receiving an error. The cleanup should be done on all failure paths.

---
 drivers/hv/channel.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/hv/channel.c +4 -3
index 54da66d..836386f 100644
@@ -186,11 +186,12 @@ int vmbus_open(struct vmbus_channel
 *newchannel, u32 send_ringbuffer_size,
 	list_del(&open_info->msglistentry);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock,
 flags);
 
-	if (err == 0)
-		newchannel->state = CHANNEL_OPENED_STATE;
+	if (err != 0)
+		goto error_gpadl;
 
+	newchannel->state = CHANNEL_OPENED_STATE;
 	kfree(open_info);
-	return err;
+	return 0;
 
Quoted 3 lines

error1:
spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
--

Thanks for the patch, Vitaly!

The patch looks good to me except for 1 small issue:
I suppose open_info->response.open_result.status is different from Linux-style -EXXX error codes.

Maybe here we can return -EAGAIN if err != 0 ?

-- Dexuan

#3 lore raw

Re: [PATCH 1/3] Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

Quoted 46 lines · Dexuan Cui
Quoted 40 lines

-----Original Message-----
From: Vitaly Kuznetsov [mailto:[email protected]]
Sent: Tuesday, April 21, 2015 16:18
To: KY Srinivasan
Cc: Haiyang Zhang; [email protected]; linux- [email protected]; Dexuan Cui
Subject: [PATCH 1/3] Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

In case there was an error reported in the response to the CHANNELMSG_OPENCHANNEL
call we need to do the cleanup as a vmbus_open() user won't be doing it after
receiving an error. The cleanup should be done on all failure paths.

---
 drivers/hv/channel.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/hv/channel.c +4 -3
index 54da66d..836386f 100644
@@ -186,11 +186,12 @@ int vmbus_open(struct vmbus_channel
 *newchannel, u32 send_ringbuffer_size,
 	list_del(&open_info->msglistentry);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock,
 flags);
 
-	if (err == 0)
-		newchannel->state = CHANNEL_OPENED_STATE;
+	if (err != 0)
+		goto error_gpadl;
 
+	newchannel->state = CHANNEL_OPENED_STATE;
 	kfree(open_info);
-	return err;
+	return 0;
 
Quoted 3 lines

error1:
spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
--

Thanks for the patch, Vitaly!

Quoted 5 lines

The patch looks good to me except for 1 small issue:
I suppose open_info->response.open_result.status is different from Linux-style -EXXX error codes.

Maybe here we can return -EAGAIN if err != 0 ?

I think I inspected all vmbus_open() callers and they all just check ret != 0 but I agree we'd better return some -EXXXX code here. I'm not exactly sure if open_result.status != 0 usually means a permanent or a temporary failure but if it's temporary -EAGAIN here seems reasonable.

I'll send v2, thanks for the review!

Quoted 2 lines

-- Dexuan

Signature · 1 line
  Vitaly
#4 lore raw

Re: [PATCH 1/3] Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

In reply to Vitaly Kuznetsov

drivers/hv/channel.c int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size, · line 189

Quoted 11 lines · Vitaly Kuznetsov
drivers/hv/channel.c +1 -2
index 54da66d..836386f 100644
@@ -186,11 +186,12 @@ int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
 	list_del(&open_info->msglistentry);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 
-	if (err == 0)
-		newchannel->state = CHANNEL_OPENED_STATE;
+	if (err != 0)

Doesn't the double negative not make it slightly more confusing?

regards,
dan carpenter

#5 lore raw

Re: [PATCH 1/3] Drivers: hv: vmbus: do cleanup on all vmbus_open() failure paths

Quoted 9 lines · Dan Carpenter
Quoted 8 lines · Vitaly Kuznetsov
drivers/hv/channel.c +0 -0
index 54da66d..836386f 100644
@@ -186,11 +186,12 @@ int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
 	list_del(&open_info->msglistentry);
 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 
Quoted 6 lines

drivers/hv/channel.c int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size, · line 189

Quoted 3 lines
drivers/hv/channel.c +1 -2
-	if (err == 0)
-		newchannel->state = CHANNEL_OPENED_STATE;
+	if (err != 0)

Doesn't the double negative not make it slightly more confusing?

Point taken, will fix in v2 :-)

Quoted 2 lines

regards,
dan carpenter

Signature · 1 line
  Vitaly
#6 lore raw

PATCH 2/3 Drivers: hv: vmbus: kill tasklets on module unload

In reply to Vitaly Kuznetsov

Explicitly kill tasklets we create on module unload.

---
 drivers/hv/vmbus_drv.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/hv/vmbus_drv.c +4 -1
index 2b56260..cf20400 100644
@@ -1108,6 +1108,7 @@ static void __exit vmbus_exit(void)
 	hv_synic_clockevents_cleanup();
 	vmbus_disconnect();
 	hv_remove_vmbus_irq();
+	tasklet_kill(&msg_dpc);
 	vmbus_free_channels();
 	if (ms_hyperv.features & HV_FEATURE_GUEST_CRASH_MSR_AVAILABLE) {
 		atomic_notifier_chain_unregister(&panic_notifier_list,
@@ -1115,8 +1116,10 @@ static void __exit vmbus_exit(void)
 	}
 	bus_unregister(&hv_bus);
 	hv_cleanup();
-	for_each_online_cpu(cpu)
+	for_each_online_cpu(cpu) {
+		tasklet_kill(hv_context.event_dpc[cpu]);
 		smp_call_function_single(cpu, hv_synic_cleanup, NULL, 1);
+	}
 	acpi_bus_unregister_driver(&vmbus_acpi_driver);
 	hv_cpu_hotplug_quirk(false);
 }
Signature · 1 line
1.9.3
#7 lore raw

PATCH 3/3 Drivers: hv: vmbus: setup irq after synic is initialized

In reply to Vitaly Kuznetsov

vmbus_isr() uses synic pages which are being setup in hv_synic_alloc(), we need to register this irq handler only after we allocate all the required pages.

---
 drivers/hv/vmbus_drv.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/hv/vmbus_drv.c +3 -2
index cf20400..e922804 100644
@@ -822,11 +822,12 @@ static int vmbus_bus_init(int irq)
 	if (ret)
 		goto err_cleanup;
 
-	hv_setup_vmbus_irq(vmbus_isr);
-
 	ret = hv_synic_alloc();
 	if (ret)
 		goto err_alloc;
+
+	hv_setup_vmbus_irq(vmbus_isr);
+
 	/*
 	 * Initialize the per-cpu interrupt state and
 	 * connect to the host.
Signature · 1 line
1.9.3