* [PATCH v2 1/4] PCI/AER: Fix pci_ops return NULL when hotplug a pci bus doing aer error inject
@ 2012-09-12 12:33 Yijing Wang
2012-09-12 12:33 ` [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject Yijing Wang
` (2 more replies)
0 siblings, 3 replies; 10+ messages in thread
From: Yijing Wang @ 2012-09-12 12:33 UTC (permalink / raw)
To: Bjorn Helgaas, Huang Ying, Chen Gong
Cc: jiang.liu, Hanjun Guo, linux-pci, Yijing Wang
When we inject aer errors to the target pcie device by aer_inject module, the pci_ops of pci
bus which the target device is on will be assigned to pci_ops_aer.So if the target pci device
is a bridge, once we hot-remove and hot-add the bridge, the newly created child bus's pci_ops
will be assigned to pci_ops_aer too.Now every access to the child bus's devices will result to
system panic, because it get a NULL pci_ops in pci_read_aer/pci_write_aer.
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Reviewed-by: Sven Dietrich <Sven.Dietrich@huawei.com>
---
drivers/pci/pcie/aer/aer_inject.c | 23 +++++++++++++++++++++++
1 files changed, 23 insertions(+), 0 deletions(-)
diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
index 4e24cb8..0816483 100644
--- a/drivers/pci/pcie/aer/aer_inject.c
+++ b/drivers/pci/pcie/aer/aer_inject.c
@@ -109,6 +109,19 @@ static struct aer_error *__find_aer_error_by_dev(struct pci_dev *dev)
return __find_aer_error((u16)domain, dev->bus->number, dev->devfn);
}
+static bool pci_is_upstream_bus(struct pci_bus *bus, struct pci_bus *up_bus)
+{
+ struct pci_bus *pbus = bus->parent;
+
+ while (pbus) {
+ if (pbus == up_bus)
+ return true;
+ pbus = pbus->parent;
+ }
+
+ return false;
+}
+
/* inject_lock must be held before calling */
static struct pci_ops *__find_pci_bus_ops(struct pci_bus *bus)
{
@@ -118,6 +131,13 @@ static struct pci_ops *__find_pci_bus_ops(struct pci_bus *bus)
if (bus_ops->bus == bus)
return bus_ops->ops;
}
+
+ /* can't find bus_ops, fall back to get bus_ops of upstream bus */
+ list_for_each_entry(bus_ops, &pci_bus_ops_list, list) {
+ if (pci_is_upstream_bus(bus, bus_ops->bus))
+ return bus_ops->ops;
+ }
+
return NULL;
}
@@ -208,6 +228,7 @@ static int pci_read_aer(struct pci_bus *bus, unsigned int devfn, int where,
}
out:
ops = __find_pci_bus_ops(bus);
+ BUG_ON(!ops);
spin_unlock_irqrestore(&inject_lock, flags);
return ops->read(bus, devfn, where, size, val);
}
@@ -243,6 +264,7 @@ int pci_write_aer(struct pci_bus *bus, unsigned int devfn, int where, int size,
}
out:
ops = __find_pci_bus_ops(bus);
+ BUG_ON(!ops);
spin_unlock_irqrestore(&inject_lock, flags);
return ops->write(bus, devfn, where, size, val);
}
@@ -506,6 +528,7 @@ static struct miscdevice aer_inject_device = {
.fops = &aer_inject_fops,
};
+
static int __init aer_inject_init(void)
{
return misc_register(&aer_inject_device);
--
1.7.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject
2012-09-12 12:33 [PATCH v2 1/4] PCI/AER: Fix pci_ops return NULL when hotplug a pci bus doing aer error inject Yijing Wang
@ 2012-09-12 12:33 ` Yijing Wang
2012-09-13 1:43 ` Huang Ying
2012-09-12 12:33 ` [PATCH v2 3/4] PCI/AER: Clean pci_bus_ops when related pci bus was removed Yijing Wang
2012-09-12 12:33 ` [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject Yijing Wang
2 siblings, 1 reply; 10+ messages in thread
From: Yijing Wang @ 2012-09-12 12:33 UTC (permalink / raw)
To: Bjorn Helgaas, Huang Ying, Chen Gong
Cc: jiang.liu, Hanjun Guo, linux-pci, Yijing Wang
Since hot plug for pci devices while doing aer inject, some newly created child buses'
pci_ops will be assigned to pci_ops_aer. Aer_inject does not track these pci_ops_aer(not
list in pci_bus_ops_list),we should clean all of these when rmmod aer_inject module.
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
---
drivers/pci/pcie/aer/aer_inject.c | 32 ++++++++++++++++++++++++++++++++
1 files changed, 32 insertions(+), 0 deletions(-)
diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
index 0816483..9d195ae 100644
--- a/drivers/pci/pcie/aer/aer_inject.c
+++ b/drivers/pci/pcie/aer/aer_inject.c
@@ -283,6 +283,28 @@ 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 *bus = NULL;
+
+ while ((bus = pci_find_next_bus(bus)) != NULL)
+ pci_clean_child_aer_ops(bus);
+}
+
+
static int pci_bus_set_aer_ops(struct pci_bus *bus)
{
struct pci_ops *ops;
@@ -546,6 +568,8 @@ static void __exit aer_inject_exit(void)
pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
kfree(bus_ops);
}
+
+ clean_untracked_pci_aer_ops();
spin_lock_irqsave(&inject_lock, flags);
list_for_each_entry_safe(err, err_next, &einjected, list) {
@@ -553,6 +577,14 @@ static void __exit aer_inject_exit(void)
kfree(err);
}
spin_unlock_irqrestore(&inject_lock, flags);
+
+ /* Inject aer errors and hotplug the same pcie device
+ * maybe assign some newly created buses' pci_ops pci_ops_aer.
+ * Since these pci_ops_aer are not tracked in pci_bus_ops_list,
+ * we need to find and clean untracked pci_ops_aer before aer_inject
+ * module exit
+ */
+ clean_untracked_pci_ops_aer();
}
module_init(aer_inject_init);
--
1.7.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject
2012-09-12 12:33 ` [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject Yijing Wang
@ 2012-09-13 1:43 ` Huang Ying
2012-09-13 10:54 ` Yijing Wang
0 siblings, 1 reply; 10+ messages in thread
From: Huang Ying @ 2012-09-13 1:43 UTC (permalink / raw)
To: Yijing Wang; +Cc: Bjorn Helgaas, Chen Gong, jiang.liu, Hanjun Guo, linux-pci
On Wed, 2012-09-12 at 20:33 +0800, Yijing Wang wrote:
> Since hot plug for pci devices while doing aer inject, some newly created child buses'
> pci_ops will be assigned to pci_ops_aer. Aer_inject does not track these pci_ops_aer(not
> list in pci_bus_ops_list),we should clean all of these when rmmod aer_inject module.
>
> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
> ---
> drivers/pci/pcie/aer/aer_inject.c | 32 ++++++++++++++++++++++++++++++++
> 1 files changed, 32 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
> index 0816483..9d195ae 100644
> --- a/drivers/pci/pcie/aer/aer_inject.c
> +++ b/drivers/pci/pcie/aer/aer_inject.c
> @@ -283,6 +283,28 @@ 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*/
Why scan all pci buses? Is it possible just to check buses in
pci_bus_ops_list.
Best Regards,
Huang Ying
> +static void clean_untracked_pci_ops_aer(void)
> +{
> + struct pci_bus *bus = NULL;
> +
> + while ((bus = pci_find_next_bus(bus)) != NULL)
> + pci_clean_child_aer_ops(bus);
> +}
> +
> +
> static int pci_bus_set_aer_ops(struct pci_bus *bus)
> {
> struct pci_ops *ops;
> @@ -546,6 +568,8 @@ static void __exit aer_inject_exit(void)
> pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
> kfree(bus_ops);
> }
> +
> + clean_untracked_pci_aer_ops();
>
> spin_lock_irqsave(&inject_lock, flags);
> list_for_each_entry_safe(err, err_next, &einjected, list) {
> @@ -553,6 +577,14 @@ static void __exit aer_inject_exit(void)
> kfree(err);
> }
> spin_unlock_irqrestore(&inject_lock, flags);
> +
> + /* Inject aer errors and hotplug the same pcie device
> + * maybe assign some newly created buses' pci_ops pci_ops_aer.
> + * Since these pci_ops_aer are not tracked in pci_bus_ops_list,
> + * we need to find and clean untracked pci_ops_aer before aer_inject
> + * module exit
> + */
> + clean_untracked_pci_ops_aer();
> }
>
> module_init(aer_inject_init);
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject
2012-09-13 1:43 ` Huang Ying
@ 2012-09-13 10:54 ` Yijing Wang
2012-09-14 0:56 ` Huang Ying
0 siblings, 1 reply; 10+ messages in thread
From: Yijing Wang @ 2012-09-13 10:54 UTC (permalink / raw)
To: Huang Ying; +Cc: Bjorn Helgaas, Chen Gong, jiang.liu, Hanjun Guo, linux-pci
On 2012/9/13 9:43, Huang Ying wrote:
> On Wed, 2012-09-12 at 20:33 +0800, Yijing Wang wrote:
>> Since hot plug for pci devices while doing aer inject, some newly created child buses'
>> pci_ops will be assigned to pci_ops_aer. Aer_inject does not track these pci_ops_aer(not
>> list in pci_bus_ops_list),we should clean all of these when rmmod aer_inject module.
>>
>> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
>> ---
>> drivers/pci/pcie/aer/aer_inject.c | 32 ++++++++++++++++++++++++++++++++
>> 1 files changed, 32 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
>> index 0816483..9d195ae 100644
>> --- a/drivers/pci/pcie/aer/aer_inject.c
>> +++ b/drivers/pci/pcie/aer/aer_inject.c
>> @@ -283,6 +283,28 @@ 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*/
>
> Why scan all pci buses? Is it possible just to check buses in
> pci_bus_ops_list.
>
Hi Huang Ying,
Some buses' pci_ops_aer may be not assigned by pci_bus_set_aer_ops() function,
As the first patch [PATCH v2 1/4] said. Some newly created buses' pci_ops assigned to
pci_ops_aer because their parent just doing aer inject, parent bus's pci_ops was pci_ops_aer.
So when we rmmod aer_inject module, these buses' pci_ops_aer will be invalid.
1、Tracked pci_ops_aer will be cleaned and pci_ops will reassign to their original pci_ops saved in pci_bus_ops_list.
2、Un-tracked pci_ops_aer won't be cleaned.so we need to fix it.
> Best Regards,
> Huang Ying
>
>> +static void clean_untracked_pci_ops_aer(void)
>> +{
>> + struct pci_bus *bus = NULL;
>> +
>> + while ((bus = pci_find_next_bus(bus)) != NULL)
>> + pci_clean_child_aer_ops(bus);
>> +}
>> +
>> +
>> static int pci_bus_set_aer_ops(struct pci_bus *bus)
>> {
>> struct pci_ops *ops;
>> @@ -546,6 +568,8 @@ static void __exit aer_inject_exit(void)
>> pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
>> kfree(bus_ops);
>> }
>> +
>> + clean_untracked_pci_aer_ops();
>>
>> spin_lock_irqsave(&inject_lock, flags);
>> list_for_each_entry_safe(err, err_next, &einjected, list) {
>> @@ -553,6 +577,14 @@ static void __exit aer_inject_exit(void)
>> kfree(err);
>> }
>> spin_unlock_irqrestore(&inject_lock, flags);
>> +
>> + /* Inject aer errors and hotplug the same pcie device
>> + * maybe assign some newly created buses' pci_ops pci_ops_aer.
>> + * Since these pci_ops_aer are not tracked in pci_bus_ops_list,
>> + * we need to find and clean untracked pci_ops_aer before aer_inject
>> + * module exit
>> + */
>> + clean_untracked_pci_ops_aer();
>> }
>>
>> module_init(aer_inject_init);
>
>
>
> .
>
--
Thanks!
Yijing
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject
2012-09-13 10:54 ` Yijing Wang
@ 2012-09-14 0:56 ` Huang Ying
2012-09-14 1:29 ` Yijing Wang
0 siblings, 1 reply; 10+ messages in thread
From: Huang Ying @ 2012-09-14 0:56 UTC (permalink / raw)
To: Yijing Wang; +Cc: Bjorn Helgaas, Chen Gong, jiang.liu, Hanjun Guo, linux-pci
On Thu, 2012-09-13 at 18:54 +0800, Yijing Wang wrote:
> On 2012/9/13 9:43, Huang Ying wrote:
> > On Wed, 2012-09-12 at 20:33 +0800, Yijing Wang wrote:
> >> Since hot plug for pci devices while doing aer inject, some newly created child buses'
> >> pci_ops will be assigned to pci_ops_aer. Aer_inject does not track these pci_ops_aer(not
> >> list in pci_bus_ops_list),we should clean all of these when rmmod aer_inject module.
> >>
> >> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
> >> ---
> >> drivers/pci/pcie/aer/aer_inject.c | 32 ++++++++++++++++++++++++++++++++
> >> 1 files changed, 32 insertions(+), 0 deletions(-)
> >>
> >> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
> >> index 0816483..9d195ae 100644
> >> --- a/drivers/pci/pcie/aer/aer_inject.c
> >> +++ b/drivers/pci/pcie/aer/aer_inject.c
> >> @@ -283,6 +283,28 @@ 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*/
> >
> > Why scan all pci buses? Is it possible just to check buses in
> > pci_bus_ops_list.
> >
>
> Hi Huang Ying,
> Some buses' pci_ops_aer may be not assigned by pci_bus_set_aer_ops() function,
> As the first patch [PATCH v2 1/4] said. Some newly created buses' pci_ops assigned to
> pci_ops_aer because their parent just doing aer inject, parent bus's pci_ops was pci_ops_aer.
> So when we rmmod aer_inject module, these buses' pci_ops_aer will be invalid.
>
> 1、Tracked pci_ops_aer will be cleaned and pci_ops will reassign to their original pci_ops saved in pci_bus_ops_list.
> 2、Un-tracked pci_ops_aer won't be cleaned.so we need to fix it.
Can we just check all bus and its children in pci_bus_ops_list? If my
understanding were correct, bus itself or some of its parents should be
saved in pci_bus_ops_list?
Best Regards,
Huang Ying
> >
> >> +static void clean_untracked_pci_ops_aer(void)
> >> +{
> >> + struct pci_bus *bus = NULL;
> >> +
> >> + while ((bus = pci_find_next_bus(bus)) != NULL)
> >> + pci_clean_child_aer_ops(bus);
> >> +}
> >> +
> >> +
> >> static int pci_bus_set_aer_ops(struct pci_bus *bus)
> >> {
> >> struct pci_ops *ops;
> >> @@ -546,6 +568,8 @@ static void __exit aer_inject_exit(void)
> >> pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
> >> kfree(bus_ops);
> >> }
> >> +
> >> + clean_untracked_pci_aer_ops();
> >>
> >> spin_lock_irqsave(&inject_lock, flags);
> >> list_for_each_entry_safe(err, err_next, &einjected, list) {
> >> @@ -553,6 +577,14 @@ static void __exit aer_inject_exit(void)
> >> kfree(err);
> >> }
> >> spin_unlock_irqrestore(&inject_lock, flags);
> >> +
> >> + /* Inject aer errors and hotplug the same pcie device
> >> + * maybe assign some newly created buses' pci_ops pci_ops_aer.
> >> + * Since these pci_ops_aer are not tracked in pci_bus_ops_list,
> >> + * we need to find and clean untracked pci_ops_aer before aer_inject
> >> + * module exit
> >> + */
> >> + clean_untracked_pci_ops_aer();
> >> }
> >>
> >> module_init(aer_inject_init);
> >
> >
> >
> > .
> >
>
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject
2012-09-14 0:56 ` Huang Ying
@ 2012-09-14 1:29 ` Yijing Wang
0 siblings, 0 replies; 10+ messages in thread
From: Yijing Wang @ 2012-09-14 1:29 UTC (permalink / raw)
To: Huang Ying; +Cc: Bjorn Helgaas, Chen Gong, jiang.liu, Hanjun Guo, linux-pci
On 2012/9/14 8:56, Huang Ying wrote:
> On Thu, 2012-09-13 at 18:54 +0800, Yijing Wang wrote:
>> On 2012/9/13 9:43, Huang Ying wrote:
>>> On Wed, 2012-09-12 at 20:33 +0800, Yijing Wang wrote:
>>>> Since hot plug for pci devices while doing aer inject, some newly created child buses'
>>>> pci_ops will be assigned to pci_ops_aer. Aer_inject does not track these pci_ops_aer(not
>>>> list in pci_bus_ops_list),we should clean all of these when rmmod aer_inject module.
>>>>
>>>> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
>>>> ---
>>>> drivers/pci/pcie/aer/aer_inject.c | 32 ++++++++++++++++++++++++++++++++
>>>> 1 files changed, 32 insertions(+), 0 deletions(-)
>>>>
>>>> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
>>>> index 0816483..9d195ae 100644
>>>> --- a/drivers/pci/pcie/aer/aer_inject.c
>>>> +++ b/drivers/pci/pcie/aer/aer_inject.c
>>>> @@ -283,6 +283,28 @@ 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*/
>>>
>>> Why scan all pci buses? Is it possible just to check buses in
>>> pci_bus_ops_list.
>>>
>>
>> Hi Huang Ying,
>> Some buses' pci_ops_aer may be not assigned by pci_bus_set_aer_ops() function,
>> As the first patch [PATCH v2 1/4] said. Some newly created buses' pci_ops assigned to
>> pci_ops_aer because their parent just doing aer inject, parent bus's pci_ops was pci_ops_aer.
>> So when we rmmod aer_inject module, these buses' pci_ops_aer will be invalid.
>>
>> 1、Tracked pci_ops_aer will be cleaned and pci_ops will reassign to their original pci_ops saved in pci_bus_ops_list.
>> 2、Un-tracked pci_ops_aer won't be cleaned.so we need to fix it.
>
> Can we just check all bus and its children in pci_bus_ops_list? If my
> understanding were correct, bus itself or some of its parents should be
> saved in pci_bus_ops_list?
>
Yes, good idea! I think the buses that owned untracked pci_ops_aer always are child buses of the root buses
tracked in the pci_bus_ops_list. I will improve this patch in this way.
Thanks
Yijing
> Best Regards,
> Huang Ying
>
>>>
>>>> +static void clean_untracked_pci_ops_aer(void)
>>>> +{
>>>> + struct pci_bus *bus = NULL;
>>>> +
>>>> + while ((bus = pci_find_next_bus(bus)) != NULL)
>>>> + pci_clean_child_aer_ops(bus);
>>>> +}
>>>> +
>>>> +
>>>> static int pci_bus_set_aer_ops(struct pci_bus *bus)
>>>> {
>>>> struct pci_ops *ops;
>>>> @@ -546,6 +568,8 @@ static void __exit aer_inject_exit(void)
>>>> pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
>>>> kfree(bus_ops);
>>>> }
>>>> +
>>>> + clean_untracked_pci_aer_ops();
>>>>
>>>> spin_lock_irqsave(&inject_lock, flags);
>>>> list_for_each_entry_safe(err, err_next, &einjected, list) {
>>>> @@ -553,6 +577,14 @@ static void __exit aer_inject_exit(void)
>>>> kfree(err);
>>>> }
>>>> spin_unlock_irqrestore(&inject_lock, flags);
>>>> +
>>>> + /* Inject aer errors and hotplug the same pcie device
>>>> + * maybe assign some newly created buses' pci_ops pci_ops_aer.
>>>> + * Since these pci_ops_aer are not tracked in pci_bus_ops_list,
>>>> + * we need to find and clean untracked pci_ops_aer before aer_inject
>>>> + * module exit
>>>> + */
>>>> + clean_untracked_pci_ops_aer();
>>>> }
>>>>
>>>> module_init(aer_inject_init);
>>>
>>>
>>>
>>> .
>>>
>>
>>
>
>
>
> .
>
--
Thanks!
Yijing
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v2 3/4] PCI/AER: Clean pci_bus_ops when related pci bus was removed
2012-09-12 12:33 [PATCH v2 1/4] PCI/AER: Fix pci_ops return NULL when hotplug a pci bus doing aer error inject Yijing Wang
2012-09-12 12:33 ` [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject Yijing Wang
@ 2012-09-12 12:33 ` Yijing Wang
2012-09-12 12:33 ` [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject Yijing Wang
2 siblings, 0 replies; 10+ messages in thread
From: Yijing Wang @ 2012-09-12 12:33 UTC (permalink / raw)
To: Bjorn Helgaas, Huang Ying, Chen Gong
Cc: jiang.liu, Hanjun Guo, linux-pci, Yijing Wang
When Inject aer errors to the target pci device, a pci_bus_ops will
be allocated for the pci device's pci bus.When the pci bus was removed,
we should also release the pci_bus_ops.
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Reviewed-by: Sven Dietrich <Sven.Dietrich@huawei.com>
---
drivers/pci/pcie/aer/aer_inject.c | 48 ++++++++++++++++++++++++++++++++++++-
1 files changed, 47 insertions(+), 1 deletions(-)
diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
index 9d195ae..bd043db 100644
--- a/drivers/pci/pcie/aer/aer_inject.c
+++ b/drivers/pci/pcie/aer/aer_inject.c
@@ -550,10 +550,55 @@ static struct miscdevice aer_inject_device = {
.fops = &aer_inject_fops,
};
+static void aer_clean_pci_bus_ops(struct pci_dev *dev)
+{
+ unsigned long flags;
+ struct pci_bus_ops *bus_ops, *tmp_ops;
+ struct pci_bus *bus;
+ bus = dev->subordinate;
+ if (!bus)
+ return;
+
+ spin_lock_irqsave(&inject_lock, flags);
+ list_for_each_entry_safe(bus_ops, tmp_ops, &pci_bus_ops_list, list)
+ if (bus_ops->bus == bus) {
+ list_del(&bus_ops->list);
+ kfree(bus_ops);
+ break;
+ }
+ spin_unlock_irqrestore(&inject_lock, flags);
+}
+
+static int aer_hp_notify_fn(struct notifier_block *nb,
+ unsigned long event, void *data)
+{
+ switch (event) {
+ case BUS_NOTIFY_DEL_DEVICE:
+ aer_clean_pci_bus_ops(to_pci_dev(data));
+ break;
+ default:
+ return NOTIFY_DONE;
+ }
+
+ return NOTIFY_OK;
+}
+
+static struct notifier_block aerinj_hp_notifier = {
+ .notifier_call = &aer_hp_notify_fn,
+};
static int __init aer_inject_init(void)
{
- return misc_register(&aer_inject_device);
+ int ret;
+ ret = misc_register(&aer_inject_device);
+ if (ret)
+ goto out;
+
+ ret = bus_register_notifier(&pci_bus_type, &aerinj_hp_notifier);
+ if (ret)
+ misc_deregister(&aer_inject_device);
+out:
+ return ret;
}
static void __exit aer_inject_exit(void)
@@ -562,6 +607,7 @@ static void __exit aer_inject_exit(void)
unsigned long flags;
struct pci_bus_ops *bus_ops;
+ bus_unregister_notifier(&pci_bus_type, &aerinj_hp_notifier);
misc_deregister(&aer_inject_device);
while ((bus_ops = pci_bus_ops_pop())) {
--
1.7.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject
2012-09-12 12:33 [PATCH v2 1/4] PCI/AER: Fix pci_ops return NULL when hotplug a pci bus doing aer error inject Yijing Wang
2012-09-12 12:33 ` [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject Yijing Wang
2012-09-12 12:33 ` [PATCH v2 3/4] PCI/AER: Clean pci_bus_ops when related pci bus was removed Yijing Wang
@ 2012-09-12 12:33 ` Yijing Wang
2012-09-13 1:29 ` Huang Ying
2 siblings, 1 reply; 10+ messages in thread
From: Yijing Wang @ 2012-09-12 12:33 UTC (permalink / raw)
To: Bjorn Helgaas, Huang Ying, Chen Gong
Cc: jiang.liu, Hanjun Guo, linux-pci, Yijing Wang
Fix a small race condition window between pci_bus_ops_pop() and
pci_bus_set_ops() functions.
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
---
drivers/pci/pcie/aer/aer_inject.c | 8 ++------
1 files changed, 2 insertions(+), 6 deletions(-)
diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
index bd043db..69c2d77 100644
--- a/drivers/pci/pcie/aer/aer_inject.c
+++ b/drivers/pci/pcie/aer/aer_inject.c
@@ -141,12 +141,11 @@ static struct pci_ops *__find_pci_bus_ops(struct pci_bus *bus)
return NULL;
}
+/* inject_lock must be held before calling */
static struct pci_bus_ops *pci_bus_ops_pop(void)
{
- unsigned long flags;
struct pci_bus_ops *bus_ops = NULL;
- spin_lock_irqsave(&inject_lock, flags);
if (list_empty(&pci_bus_ops_list))
bus_ops = NULL;
else {
@@ -154,7 +153,6 @@ static struct pci_bus_ops *pci_bus_ops_pop(void)
list_del(lh);
bus_ops = list_entry(lh, struct pci_bus_ops, list);
}
- spin_unlock_irqrestore(&inject_lock, flags);
return bus_ops;
}
@@ -610,14 +608,12 @@ static void __exit aer_inject_exit(void)
bus_unregister_notifier(&pci_bus_type, &aerinj_hp_notifier);
misc_deregister(&aer_inject_device);
+ spin_lock_irqsave(&inject_lock, flags);
while ((bus_ops = pci_bus_ops_pop())) {
pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
kfree(bus_ops);
}
- clean_untracked_pci_aer_ops();
-
- spin_lock_irqsave(&inject_lock, flags);
list_for_each_entry_safe(err, err_next, &einjected, list) {
list_del(&err->list);
kfree(err);
--
1.7.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject
2012-09-12 12:33 ` [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject Yijing Wang
@ 2012-09-13 1:29 ` Huang Ying
2012-09-13 10:59 ` Yijing Wang
0 siblings, 1 reply; 10+ messages in thread
From: Huang Ying @ 2012-09-13 1:29 UTC (permalink / raw)
To: Yijing Wang; +Cc: Bjorn Helgaas, Chen Gong, jiang.liu, Hanjun Guo, linux-pci
On Wed, 2012-09-12 at 20:33 +0800, Yijing Wang wrote:
> Fix a small race condition window between pci_bus_ops_pop() and
> pci_bus_set_ops() functions.
>
> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
> Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
> ---
> drivers/pci/pcie/aer/aer_inject.c | 8 ++------
> 1 files changed, 2 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
> index bd043db..69c2d77 100644
> --- a/drivers/pci/pcie/aer/aer_inject.c
> +++ b/drivers/pci/pcie/aer/aer_inject.c
> @@ -141,12 +141,11 @@ static struct pci_ops *__find_pci_bus_ops(struct pci_bus *bus)
> return NULL;
> }
>
> +/* inject_lock must be held before calling */
> static struct pci_bus_ops *pci_bus_ops_pop(void)
> {
> - unsigned long flags;
> struct pci_bus_ops *bus_ops = NULL;
>
> - spin_lock_irqsave(&inject_lock, flags);
> if (list_empty(&pci_bus_ops_list))
> bus_ops = NULL;
> else {
> @@ -154,7 +153,6 @@ static struct pci_bus_ops *pci_bus_ops_pop(void)
> list_del(lh);
> bus_ops = list_entry(lh, struct pci_bus_ops, list);
> }
> - spin_unlock_irqrestore(&inject_lock, flags);
> return bus_ops;
> }
>
> @@ -610,14 +608,12 @@ static void __exit aer_inject_exit(void)
> bus_unregister_notifier(&pci_bus_type, &aerinj_hp_notifier);
> misc_deregister(&aer_inject_device);
>
> + spin_lock_irqsave(&inject_lock, flags);
> while ((bus_ops = pci_bus_ops_pop())) {
> pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
> kfree(bus_ops);
> }
>
> - clean_untracked_pci_aer_ops();
> -
> - spin_lock_irqsave(&inject_lock, flags);
> list_for_each_entry_safe(err, err_next, &einjected, list) {
> list_del(&err->list);
> kfree(err);
This may trigger a AB BA dead lock. In aer_inject_exit, the lock
sequence is:
inject_lock -> pci_lock
In pci config read/write path, the lock sequence is:
pci_lock -> inject_lock
With lockdep enabled, you may found warning on that.
Best Regards,
Huang YIng
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject
2012-09-13 1:29 ` Huang Ying
@ 2012-09-13 10:59 ` Yijing Wang
0 siblings, 0 replies; 10+ messages in thread
From: Yijing Wang @ 2012-09-13 10:59 UTC (permalink / raw)
To: Huang Ying; +Cc: Bjorn Helgaas, Chen Gong, jiang.liu, Hanjun Guo, linux-pci
On 2012/9/13 9:29, Huang Ying wrote:
> On Wed, 2012-09-12 at 20:33 +0800, Yijing Wang wrote:
>> Fix a small race condition window between pci_bus_ops_pop() and
>> pci_bus_set_ops() functions.
>>
>> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
>> Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
>> ---
>> drivers/pci/pcie/aer/aer_inject.c | 8 ++------
>> 1 files changed, 2 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/pci/pcie/aer/aer_inject.c b/drivers/pci/pcie/aer/aer_inject.c
>> index bd043db..69c2d77 100644
>> --- a/drivers/pci/pcie/aer/aer_inject.c
>> +++ b/drivers/pci/pcie/aer/aer_inject.c
>> @@ -141,12 +141,11 @@ static struct pci_ops *__find_pci_bus_ops(struct pci_bus *bus)
>> return NULL;
>> }
>>
>> +/* inject_lock must be held before calling */
>> static struct pci_bus_ops *pci_bus_ops_pop(void)
>> {
>> - unsigned long flags;
>> struct pci_bus_ops *bus_ops = NULL;
>>
>> - spin_lock_irqsave(&inject_lock, flags);
>> if (list_empty(&pci_bus_ops_list))
>> bus_ops = NULL;
>> else {
>> @@ -154,7 +153,6 @@ static struct pci_bus_ops *pci_bus_ops_pop(void)
>> list_del(lh);
>> bus_ops = list_entry(lh, struct pci_bus_ops, list);
>> }
>> - spin_unlock_irqrestore(&inject_lock, flags);
>> return bus_ops;
>> }
>>
>> @@ -610,14 +608,12 @@ static void __exit aer_inject_exit(void)
>> bus_unregister_notifier(&pci_bus_type, &aerinj_hp_notifier);
>> misc_deregister(&aer_inject_device);
>>
>> + spin_lock_irqsave(&inject_lock, flags);
>> while ((bus_ops = pci_bus_ops_pop())) {
>> pci_bus_set_ops(bus_ops->bus, bus_ops->ops);
>> kfree(bus_ops);
>> }
>>
>> - clean_untracked_pci_aer_ops();
>> -
>> - spin_lock_irqsave(&inject_lock, flags);
>> list_for_each_entry_safe(err, err_next, &einjected, list) {
>> list_del(&err->list);
>> kfree(err);
>
> This may trigger a AB BA dead lock. In aer_inject_exit, the lock
> sequence is:
>
> inject_lock -> pci_lock
>
> In pci config read/write path, the lock sequence is:
>
> pci_lock -> inject_lock
>
> With lockdep enabled, you may found warning on that.
OK, I will carry out a detailed test with lockdep enabled, and try to improve this patch.
Thanks for review and comments!
------
Thanks
Yijing
>
> Best Regards,
> Huang YIng
>
>
>
> .
>
--
Thanks!
Yijing
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2012-09-14 1:31 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-09-12 12:33 [PATCH v2 1/4] PCI/AER: Fix pci_ops return NULL when hotplug a pci bus doing aer error inject Yijing Wang
2012-09-12 12:33 ` [PATCH v2 2/4] PCI/AER: clean all untracked pci_ops_aer when rmmod aer_inject Yijing Wang
2012-09-13 1:43 ` Huang Ying
2012-09-13 10:54 ` Yijing Wang
2012-09-14 0:56 ` Huang Ying
2012-09-14 1:29 ` Yijing Wang
2012-09-12 12:33 ` [PATCH v2 3/4] PCI/AER: Clean pci_bus_ops when related pci bus was removed Yijing Wang
2012-09-12 12:33 ` [PATCH v2 4/4] PCI/AER: fix a small race condition window when rmmod aer_inject Yijing Wang
2012-09-13 1:29 ` Huang Ying
2012-09-13 10:59 ` Yijing Wang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).