Re: [PATCH 1/2, pci] pci: Addition of PCI_DEV_FLAGS_QUIET_PCI_REALIGN attribute to the PCI subsystem

["Desnes Augusto Nunes do Rosário" <desnesn@linux.vnet.ibm.com>]

Hello Bjorn,

On 03/14/2018 03:06 PM, Bjorn Helgaas wrote:
> On Wed, Mar 14, 2018 at 01:34:54PM -0300, Desnes A. Nunes do Rosario wrote:
>> Add PCI_DEV_FLAGS_QUIET_PCI_REALIGN to pci_dev_flags and use it to
>> silent PCI realignment messages if the flag is turned on by a driver.
>>
>> Signed-off-by: Desnes A. Nunes do Rosario <desnesn@linux.vnet.ibm.com>
>> ---
>>   drivers/pci/pci.c       | 3 ++-
>>   drivers/pci/setup-res.c | 3 ++-
>>   include/linux/pci.h     | 2 ++
>>   3 files changed, 6 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>> index 8c71d1a66cdd..be197c944e5f 100644
>> --- a/drivers/pci/pci.c
>> +++ b/drivers/pci/pci.c
>> @@ -5505,7 +5505,8 @@ void pci_reassigndev_resource_alignment(struct pci_dev *dev)
>>   		return;
>>   	}
>>   
>> -	pci_info(dev, "Disabling memory decoding and releasing memory resources\n");
>> +	if (!(dev->dev_flags & PCI_DEV_FLAGS_QUIET_PCI_REALIGN))
>> +		pci_info(dev, "Disabling memory decoding and releasing memory resources\n");
>>   	pci_read_config_word(dev, PCI_COMMAND, &command);
>>   	command &= ~PCI_COMMAND_MEMORY;
>>   	pci_write_config_word(dev, PCI_COMMAND, command);
>> diff --git a/drivers/pci/setup-res.c b/drivers/pci/setup-res.c
>> index 369d48d6c6f1..00a538def763 100644
>> --- a/drivers/pci/setup-res.c
>> +++ b/drivers/pci/setup-res.c
>> @@ -172,7 +172,8 @@ EXPORT_SYMBOL(pci_claim_resource);
>>   
>>   void pci_disable_bridge_window(struct pci_dev *dev)
>>   {
>> -	pci_info(dev, "disabling bridge mem windows\n");
>> +	if (!(dev->dev_flags & PCI_DEV_FLAGS_QUIET_PCI_REALIGN))
>> +		pci_info(dev, "disabling bridge mem windows\n");
> 
> As far as I'm concerned, we can just remove these messages completely.
> I don't think there's any real value there.

After I found out that this was happening to all PCI devices on powerpc 
due to the __weak
pcibios_default_alignment() interface (necessary for VFIO passthrough 
and performance), I confess that this was my first approach to this 
matter; however I couldn't vouch the need of these messages on other 
architectures.

If there are no further concerns, I definitely prefer sending a second 
version of this patch only eliminating these messages and attesting the 
reason why.

Thank you very much for your review Bjorn,

-- 
Desnes A. Nunes do Rosário