[V2] PCI: print pcie max_payload_size default setting info to help diagnosis

PCIe device mps maybe not equal to its bridge mps after doing hotplug.
This may result unexpected problem like
http://marc.info/?l=linux-scsi&m=134788365823217&w=2. So add mps log for current
mps setting to help diagnosis an issue like this.

Signed-off-by: Yijing Wang <wangyijing@huawei.com>
---
 drivers/pci/probe.c |   29 +++++++++++++++++++++++++++++
 1 files changed, 29 insertions(+), 0 deletions(-)

On Thu, Oct 11, 2012 at 5:03 AM, Yijing Wang <wangyijing@huawei.com> wrote:
> PCIe device mps maybe not equal to its bridge mps after doing hotplug.
> This may result unexpected problem like
> http://marc.info/?l=linux-scsi&m=134788365823217&w=2. So add mps log for current
> mps setting to help diagnosis an issue like this.
>
> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
> ---
>  drivers/pci/probe.c |   29 +++++++++++++++++++++++++++++
>  1 files changed, 29 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
> index ec909af..a31e615 100644
> --- a/drivers/pci/probe.c
> +++ b/drivers/pci/probe.c
> @@ -1533,6 +1533,33 @@ static void pcie_write_mrrs(struct pci_dev *dev)
>                         "with pci=pcie_bus_safe.\n");
>  }
>
> +static int pcie_mps_info(struct pci_dev *dev, void *data)
> +{
> +       int mps, mpss;
> +
> +       if (!pci_is_pcie(dev))
> +               return -EINVAL;
> +
> +       mps = pcie_get_mps(dev);
> +       mpss = 128 << dev->pcie_mpss;
> +       dev_printk(KERN_DEBUG, &dev->dev, "PCI-E Max_Payload_Size Supported %d, "
> +                       "Max_Payload_Size %d\n", mpss, mps);
> +
> +       return 0;
> +}
> +
> +static void pcie_dump_mps_info(struct pci_bus *bus)
> +{
> +       int mps, mpss;
> +
> +       mps = pcie_get_mps(bus->self);
> +       mpss = 128 << bus->self->pcie_mpss;
> +       dev_printk(KERN_DEBUG, &bus->self->dev, "PCI-E Max_Payload_Size Supported %d, "
> +               "Max_Payload_Size %d\n", mpss, mps);
> +
> +       pci_walk_bus(bus, pcie_mps_info, NULL);
> +}
> +
>  static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
>  {
>         int mps, orig_mps;
> @@ -1564,6 +1591,8 @@ void pcie_bus_configure_settings(struct pci_bus *bus, u8 mpss)
>         if (!pci_is_pcie(bus->self))
>                 return;
>
> +       pcie_dump_mps_info(bus);
> +

The patch looks fine, but a better way of doing things would be to
create a new default pcie_bus_config state (PCIE_BUS_WARN) that:
* prints the default MPS of the device
* prints the suggested MPS
* notifies the user that they would be better off running
PCIE_BUS_SAFE (if that is the case).

I have a rough version of this in my tree already.  I'll try and push
it out tonight.

Thanks,
Jon


>         if (pcie_bus_config == PCIE_BUS_TUNE_OFF)
>                 return;
>
> --
> 1.7.1
>
>
--
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

On 2012/10/12 8:15, Jon Mason wrote:
> On Thu, Oct 11, 2012 at 5:03 AM, Yijing Wang <wangyijing@huawei.com> wrote:
>> PCIe device mps maybe not equal to its bridge mps after doing hotplug.
>> This may result unexpected problem like
>> http://marc.info/?l=linux-scsi&m=134788365823217&w=2. So add mps log for current
>> mps setting to help diagnosis an issue like this.
>>
>> Signed-off-by: Yijing Wang <wangyijing@huawei.com>
>> ---
>>  drivers/pci/probe.c |   29 +++++++++++++++++++++++++++++
>>  1 files changed, 29 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
>> index ec909af..a31e615 100644
>> --- a/drivers/pci/probe.c
>> +++ b/drivers/pci/probe.c
>> @@ -1533,6 +1533,33 @@ static void pcie_write_mrrs(struct pci_dev *dev)
>>                         "with pci=pcie_bus_safe.\n");
>>  }
>>
>> +static int pcie_mps_info(struct pci_dev *dev, void *data)
>> +{
>> +       int mps, mpss;
>> +
>> +       if (!pci_is_pcie(dev))
>> +               return -EINVAL;
>> +
>> +       mps = pcie_get_mps(dev);
>> +       mpss = 128 << dev->pcie_mpss;
>> +       dev_printk(KERN_DEBUG, &dev->dev, "PCI-E Max_Payload_Size Supported %d, "
>> +                       "Max_Payload_Size %d\n", mpss, mps);
>> +
>> +       return 0;
>> +}
>> +
>> +static void pcie_dump_mps_info(struct pci_bus *bus)
>> +{
>> +       int mps, mpss;
>> +
>> +       mps = pcie_get_mps(bus->self);
>> +       mpss = 128 << bus->self->pcie_mpss;
>> +       dev_printk(KERN_DEBUG, &bus->self->dev, "PCI-E Max_Payload_Size Supported %d, "
>> +               "Max_Payload_Size %d\n", mpss, mps);
>> +
>> +       pci_walk_bus(bus, pcie_mps_info, NULL);
>> +}
>> +
>>  static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
>>  {
>>         int mps, orig_mps;
>> @@ -1564,6 +1591,8 @@ void pcie_bus_configure_settings(struct pci_bus *bus, u8 mpss)
>>         if (!pci_is_pcie(bus->self))
>>                 return;
>>
>> +       pcie_dump_mps_info(bus);
>> +
> 
> The patch looks fine, but a better way of doing things would be to
> create a new default pcie_bus_config state (PCIE_BUS_WARN) that:
> * prints the default MPS of the device
> * prints the suggested MPS
> * notifies the user that they would be better off running
> PCIE_BUS_SAFE (if that is the case).
> 

I agree, your solution is better, I will drop this patch.

> I have a rough version of this in my tree already.  I'll try and push
> it out tonight.
> 
> Thanks,
> Jon
> 
> 
>>         if (pcie_bus_config == PCIE_BUS_TUNE_OFF)
>>                 return;
>>
>> --
>> 1.7.1
>>
>>
> 
> .
>