Patchwork [v4,3/5] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject

login
register
mail settings
Submitter Yijing Wang
Date Sept. 21, 2012, 3:44 a.m.
Message ID <1348199056-7696-4-git-send-email-wangyijing@huawei.com>
Download mbox | patch
Permalink /patch/185567/
State Superseded
Headers show

Comments

Yijing Wang - Sept. 21, 2012, 3:44 a.m.
When we do hot plug for pci devices that were injected aer errors, some newly created child buses'
pci_ops will be assigned to pci_ops_aer. Aer_inject module will not track these pci_ops_aer(not
list in pci_bus_ops_list),so we should clean all of these when rmmod aer_inject module.

Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Huang Ying <ying.huang@intel.com>
---
 drivers/pci/pcie/aer/aer_inject.c |   24 ++++++++++++++++++++++++
 1 files changed, 24 insertions(+), 0 deletions(-)
Huang Ying - Sept. 21, 2012, 5:13 a.m.
On Fri, 2012-09-21 at 11:44 +0800, Yijing Wang wrote:
> When we do hot plug for pci devices that were injected aer errors, some newly created child buses'
> pci_ops will be assigned to pci_ops_aer. Aer_inject module will not track these pci_ops_aer(not
> list in pci_bus_ops_list),so we should clean all of these when rmmod aer_inject module.
> 
> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
> Signed-off-by: Huang Ying <ying.huang@intel.com>
> ---
>  drivers/pci/pcie/aer/aer_inject.c |   24 ++++++++++++++++++++++++
>  1 files changed, 24 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
> index 0123120..e04cf24 100644
> --- a/drivers/pci/pcie/aer/aer_inject.c
> +++ b/drivers/pci/pcie/aer/aer_inject.c
> @@ -299,6 +299,29 @@ static void pci_bus_ops_init(struct pci_bus_ops *bus_ops,
>  	bus_ops->ops = ops;
>  }
>  
> +static void pci_clean_child_aer_ops(struct pci_bus *bus)
> +{
> +	struct pci_bus *child;
> +
> +	list_for_each_entry(child, &bus->children, node) {
> +		if (child->ops == &pci_ops_aer)
> +			pci_bus_set_ops(child, bus->ops);
> +		pci_clean_child_aer_ops(child);
> +	}
> +}
> +
> +/* find pci_ops_aer from root bus, and replace it by parent bus's pci_ops.
> + * pci_ops of root bus won't be pci_ops_aer here*/
> +static void clean_untracked_pci_ops_aer(void)
> +{
> +	struct pci_bus_ops *bus_ops;
> +
> +	list_for_each_entry(bus_ops, &pci_bus_ops_list, list) {
> +		if (pci_is_root_bus(bus_ops->bus))

Why do cleanup only for root bus?

Best Regards,
Huang Ying

> +			pci_clean_child_aer_ops(bus_ops->bus);
> +	}
> +}
> +
>  static int pci_bus_set_aer_ops(struct pci_bus *bus)
>  {
>  	struct pci_ops *ops;
> @@ -560,6 +583,7 @@ static void __exit aer_inject_exit(void)
>  	while ((bus_ops = pci_bus_ops_get(bus_ops)))
>  		pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
>  
> +	clean_untracked_pci_ops_aer();
>  	while ((bus_ops = pci_bus_ops_pop()))
>  		kfree(bus_ops);
>  


--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Yijing Wang - Sept. 21, 2012, 5:47 a.m.
On 2012/9/21 13:13, Huang Ying wrote:
> On Fri, 2012-09-21 at 11:44 +0800, Yijing Wang wrote:
>> When we do hot plug for pci devices that were injected aer errors, some newly created child buses'
>> pci_ops will be assigned to pci_ops_aer. Aer_inject module will not track these pci_ops_aer(not
>> list in pci_bus_ops_list),so we should clean all of these when rmmod aer_inject module.
>>
>> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
>> Signed-off-by: Huang Ying <ying.huang@intel.com>
>> ---
>>  drivers/pci/pcie/aer/aer_inject.c |   24 ++++++++++++++++++++++++
>>  1 files changed, 24 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
>> index 0123120..e04cf24 100644
>> --- a/drivers/pci/pcie/aer/aer_inject.c
>> +++ b/drivers/pci/pcie/aer/aer_inject.c
>> @@ -299,6 +299,29 @@ static void pci_bus_ops_init(struct pci_bus_ops *bus_ops,
>>  	bus_ops->ops = ops;
>>  }
>>  
>> +static void pci_clean_child_aer_ops(struct pci_bus *bus)
>> +{
>> +	struct pci_bus *child;
>> +
>> +	list_for_each_entry(child, &bus->children, node) {
>> +		if (child->ops == &pci_ops_aer)
>> +			pci_bus_set_ops(child, bus->ops);
>> +		pci_clean_child_aer_ops(child);
>> +	}
>> +}
>> +
>> +/* find pci_ops_aer from root bus, and replace it by parent bus's pci_ops.
>> + * pci_ops of root bus won't be pci_ops_aer here*/
>> +static void clean_untracked_pci_ops_aer(void)
>> +{
>> +	struct pci_bus_ops *bus_ops;
>> +
>> +	list_for_each_entry(bus_ops, &pci_bus_ops_list, list) {
>> +		if (pci_is_root_bus(bus_ops->bus))
> 
> Why do cleanup only for root bus?
> 

Because the bus with untracked pci_ops_aer always is the child bus of the root bus
whose pci_ops_aer is tracked in pci_bus_ops_list. So here only do cleanup for root bus to
avoid unnecessary pci_clean_child_aer_ops() called.
eg.
If inject aer errors into 0000:46:00.0, so the pci_ops of bus 0000:40 and bus 0000:46 will be assign to pci_ops_aer
hot remove and hot add 0000:46:00.0, so newly created bus 0000:47 0000:48 and 0000:49 have untracked pci_ops_aer;
In this case 0000:40 and 0000:46 are tracked in pci_bus_ops_list, but 0000:47 0000:48 and 0000:49 are not.
So when we clean untracked pci_ops_aer, I think only cleanup root bus's child bus is enough. cleanup bus 0000:46's child bus
is unnecessary.

 |           +-07.0-[0000:46-49]----00.0-[0000:47-49]--+-02.0-[0000:48]--+-00.0  Intel Corporation 82576 Gigabit Network Connection
 |           |                                         |                 \-00.1  Intel Corporation 82576 Gigabit Network Connection
 |           |                                         \-04.0-[0000:49]--+-00.0  Intel Corporation 82576 Gigabit Network Connection
 |           |                                                           \-00.1  Intel Corporation 82576 Gigabit Network Connection

Thanks
Yijing


> Best Regards,
> Huang Ying
> 
>> +			pci_clean_child_aer_ops(bus_ops->bus);
>> +	}
>> +}
>> +
>>  static int pci_bus_set_aer_ops(struct pci_bus *bus)
>>  {
>>  	struct pci_ops *ops;
>> @@ -560,6 +583,7 @@ static void __exit aer_inject_exit(void)
>>  	while ((bus_ops = pci_bus_ops_get(bus_ops)))
>>  		pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
>>  
>> +	clean_untracked_pci_ops_aer();
>>  	while ((bus_ops = pci_bus_ops_pop()))
>>  		kfree(bus_ops);
>>  
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-pci" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> .
>
Huang Ying - Sept. 21, 2012, 6:15 a.m.
On Fri, 2012-09-21 at 13:47 +0800, Yijing Wang wrote:
> On 2012/9/21 13:13, Huang Ying wrote:
> > On Fri, 2012-09-21 at 11:44 +0800, Yijing Wang wrote:
> >> When we do hot plug for pci devices that were injected aer errors, some newly created child buses'
> >> pci_ops will be assigned to pci_ops_aer. Aer_inject module will not track these pci_ops_aer(not
> >> list in pci_bus_ops_list),so we should clean all of these when rmmod aer_inject module.
> >>
> >> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
> >> Signed-off-by: Huang Ying <ying.huang@intel.com>

Please remove my signed-off.

> >> ---
> >>  drivers/pci/pcie/aer/aer_inject.c |   24 ++++++++++++++++++++++++
> >>  1 files changed, 24 insertions(+), 0 deletions(-)
> >>
> >> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
> >> index 0123120..e04cf24 100644
> >> --- a/drivers/pci/pcie/aer/aer_inject.c
> >> +++ b/drivers/pci/pcie/aer/aer_inject.c
> >> @@ -299,6 +299,29 @@ static void pci_bus_ops_init(struct pci_bus_ops *bus_ops,
> >>  	bus_ops->ops = ops;
> >>  }
> >>  
> >> +static void pci_clean_child_aer_ops(struct pci_bus *bus)
> >> +{
> >> +	struct pci_bus *child;
> >> +
> >> +	list_for_each_entry(child, &bus->children, node) {
> >> +		if (child->ops == &pci_ops_aer)
> >> +			pci_bus_set_ops(child, bus->ops);
> >> +		pci_clean_child_aer_ops(child);
> >> +	}
> >> +}
> >> +
> >> +/* find pci_ops_aer from root bus, and replace it by parent bus's pci_ops.
> >> + * pci_ops of root bus won't be pci_ops_aer here*/
> >> +static void clean_untracked_pci_ops_aer(void)
> >> +{
> >> +	struct pci_bus_ops *bus_ops;
> >> +
> >> +	list_for_each_entry(bus_ops, &pci_bus_ops_list, list) {
> >> +		if (pci_is_root_bus(bus_ops->bus))
> > 
> > Why do cleanup only for root bus?
> > 
> 
> Because the bus with untracked pci_ops_aer always is the child bus of the root bus
> whose pci_ops_aer is tracked in pci_bus_ops_list. So here only do cleanup for root bus to
> avoid unnecessary pci_clean_child_aer_ops() called.
> eg.
> If inject aer errors into 0000:46:00.0, so the pci_ops of bus 0000:40 and bus 0000:46 will be assign to pci_ops_aer
> hot remove and hot add 0000:46:00.0, so newly created bus 0000:47 0000:48 and 0000:49 have untracked pci_ops_aer;
> In this case 0000:40 and 0000:46 are tracked in pci_bus_ops_list, but 0000:47 0000:48 and 0000:49 are not.
> So when we clean untracked pci_ops_aer, I think only cleanup root bus's child bus is enough. cleanup bus 0000:46's child bus
> is unnecessary.
> 
>  |           +-07.0-[0000:46-49]----00.0-[0000:47-49]--+-02.0-[0000:48]--+-00.0  Intel Corporation 82576 Gigabit Network Connection
>  |           |                                         |                 \-00.1  Intel Corporation 82576 Gigabit Network Connection
>  |           |                                         \-04.0-[0000:49]--+-00.0  Intel Corporation 82576 Gigabit Network Connection
>  |           |                                                           \-00.1  Intel Corporation 82576 Gigabit Network Connection

I see.

Reviewed-by: Huang Ying <ying.huang@intel.com>

Best Regards,
Huang Ying

> Thanks
> Yijing
> 
> 
> > Best Regards,
> > Huang Ying
> > 
> >> +			pci_clean_child_aer_ops(bus_ops->bus);
> >> +	}
> >> +}
> >> +
> >>  static int pci_bus_set_aer_ops(struct pci_bus *bus)
> >>  {
> >>  	struct pci_ops *ops;
> >> @@ -560,6 +583,7 @@ static void __exit aer_inject_exit(void)
> >>  	while ((bus_ops = pci_bus_ops_get(bus_ops)))
> >>  		pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
> >>  
> >> +	clean_untracked_pci_ops_aer();
> >>  	while ((bus_ops = pci_bus_ops_pop()))
> >>  		kfree(bus_ops);
> >>  
> > 
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-pci" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> > .
> > 
> 
> 


--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
index 0123120..e04cf24 100644
--- a/drivers/pci/pcie/aer/aer_inject.c
+++ b/drivers/pci/pcie/aer/aer_inject.c
@@ -299,6 +299,29 @@  static void pci_bus_ops_init(struct pci_bus_ops *bus_ops,
 	bus_ops->ops = ops;
 }
 
+static void pci_clean_child_aer_ops(struct pci_bus *bus)
+{
+	struct pci_bus *child;
+
+	list_for_each_entry(child, &bus->children, node) {
+		if (child->ops == &pci_ops_aer)
+			pci_bus_set_ops(child, bus->ops);
+		pci_clean_child_aer_ops(child);
+	}
+}
+
+/* find pci_ops_aer from root bus, and replace it by parent bus's pci_ops.
+ * pci_ops of root bus won't be pci_ops_aer here*/
+static void clean_untracked_pci_ops_aer(void)
+{
+	struct pci_bus_ops *bus_ops;
+
+	list_for_each_entry(bus_ops, &pci_bus_ops_list, list) {
+		if (pci_is_root_bus(bus_ops->bus))
+			pci_clean_child_aer_ops(bus_ops->bus);
+	}
+}
+
 static int pci_bus_set_aer_ops(struct pci_bus *bus)
 {
 	struct pci_ops *ops;
@@ -560,6 +583,7 @@  static void __exit aer_inject_exit(void)
 	while ((bus_ops = pci_bus_ops_get(bus_ops)))
 		pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
 
+	clean_untracked_pci_ops_aer();
 	while ((bus_ops = pci_bus_ops_pop()))
 		kfree(bus_ops);