[PATCH] iov: initialize NumVFs register to 0 at the end of sriov_init()

[PATCH] iov: initialize NumVFs register to 0 at the end of sriov_init()

[Ethan Zhao]

After commit 4449f079722c ("PCI: Calculate maximum number of buses
required for VFs"),the initial value of NumVFs register was set to
non-zero after sriov_init() and no VFs was enabled in device driver.
this changed the behaviour of kernel exported by lspci and sysfs etc.
so this patch initialize the NumVFs register to zero after the
calculation of max_VF_buses was done.

Tested on stable 4.1 and passed building on stable 4.2-rc7

Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com>
Tested-by: Sriharsha Yadagudde <sriharsha.devdas@oracle.com>
---
 drivers/pci/iov.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
index ee0ebff..6969084 100644
--- a/drivers/pci/iov.c
+++ b/drivers/pci/iov.c
@@ -476,6 +476,7 @@ found:
 	dev->is_physfn = 1;
 	iov->max_VF_buses = virtfn_max_buses(dev);
 
+	pci_iov_set_numvfs(dev, 0);
 	return 0;
 
 failed:
-- 
1.8.3.1

Re: [PATCH] iov: initialize NumVFs register to 0 at the end of sriov_init()

[Bjorn Helgaas]

On Fri, Aug 21, 2015 at 06:51:25PM +0900, Ethan Zhao wrote:
> After commit 4449f079722c ("PCI: Calculate maximum number of buses
> required for VFs"),the initial value of NumVFs register was set to
> non-zero after sriov_init() and no VFs was enabled in device driver.
> this changed the behaviour of kernel exported by lspci and sysfs etc.
> so this patch initialize the NumVFs register to zero after the
> calculation of max_VF_buses was done.
> 
> Tested on stable 4.1 and passed building on stable 4.2-rc7
> 
> Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com>
> Tested-by: Sriharsha Yadagudde <sriharsha.devdas@oracle.com>
> ---
>  drivers/pci/iov.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
> index ee0ebff..6969084 100644
> --- a/drivers/pci/iov.c
> +++ b/drivers/pci/iov.c
> @@ -476,6 +476,7 @@ found:
>  	dev->is_physfn = 1;
>  	iov->max_VF_buses = virtfn_max_buses(dev);
>  
> +	pci_iov_set_numvfs(dev, 0);

I think it would be better to put this in virtfn_max_buses(), where we
clobbered numVFs in the first place.  I'd also read the original value and
restore it, e.g.,

    pci_read_config_word(dev, iov->pos + PCI_SRIOV_NUM_VF, &numvfs);
    for (nr_virtfn = 1; nr_virtfn <= iov->total_VFs; nr_virtfn++) {
      ...
    }
    pci_iov_set_numvfs(dev, numvfs);
    return max;

I know sriov_init() sets numVFs to zero before it calls virtfn_max_buses(),
but why rely on that extra knowledge?

>  	return 0;
>  
>  failed:
> -- 
> 1.8.3.1
> 

Re: [PATCH] iov: initialize NumVFs register to 0 at the end of sriov_init()

[ethan zhao]

On 2015/9/16 0:10, Bjorn Helgaas wrote:
> On Fri, Aug 21, 2015 at 06:51:25PM +0900, Ethan Zhao wrote:
>> After commit 4449f079722c ("PCI: Calculate maximum number of buses
>> required for VFs"),the initial value of NumVFs register was set to
>> non-zero after sriov_init() and no VFs was enabled in device driver.
>> this changed the behaviour of kernel exported by lspci and sysfs etc.
>> so this patch initialize the NumVFs register to zero after the
>> calculation of max_VF_buses was done.
>>
>> Tested on stable 4.1 and passed building on stable 4.2-rc7
>>
>> Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com>
>> Tested-by: Sriharsha Yadagudde <sriharsha.devdas@oracle.com>
>> ---
>>   drivers/pci/iov.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
>> index ee0ebff..6969084 100644
>> --- a/drivers/pci/iov.c
>> +++ b/drivers/pci/iov.c
>> @@ -476,6 +476,7 @@ found:
>>   	dev->is_physfn = 1;
>>   	iov->max_VF_buses = virtfn_max_buses(dev);
>>   
>> +	pci_iov_set_numvfs(dev, 0);
> I think it would be better to put this in virtfn_max_buses(), where we
> clobbered numVFs in the first place.  I'd also read the original value and
> restore it, e.g.,
>
>      pci_read_config_word(dev, iov->pos + PCI_SRIOV_NUM_VF, &numvfs);
>      for (nr_virtfn = 1; nr_virtfn <= iov->total_VFs; nr_virtfn++) {
>        ...
>      }
>      pci_iov_set_numvfs(dev, numvfs);
>      return max;
>
> I know sriov_init() sets numVFs to zero before it calls virtfn_max_buses(),
> but why rely on that extra knowledge?
  Yes, I think also virtfn_max_buses() should restore the register 
before it returns.
  Will revise it and resend.

  Thanks,
  Ethan
>>   	return 0;
>>   
>>   failed:
>> -- 
>> 1.8.3.1
>>