Re: [PATCH 0/4] PCI: Enable LTR/OBFF before device is used by driver

[Don Dutile <ddutile@redhat.com>]

On 06/08/2012 02:02 PM, Myron Stowe wrote:
> On Fri, Jun 8, 2012 at 11:31 AM, Bjorn Helgaas<bhelgaas@google.com>  wrote:
>> On Fri, Jun 8, 2012 at 1:01 AM, Xudong Hao<xudong.hao@intel.com>  wrote:
>>> The series of patches enable LTR and OBFF before device is used by driver, and
>>> introduce a couple of functions to save/restore LTR latency value.
>>>
>>> Patch 1/4 introduce new function pci_obff_supported() as pci_ltr_support().
>>>
>>> Patch 2/4 enable LTR(Latency tolerance reporting) before device is used by
>>> driver.
>>>
>>> Patch 3/4 enable OBFF(optimized buffer flush/fill) before device is used by
>>> driver.
>>>
>>> Patch 4/4 introduce a couple of functions pci_save_ltr_value() and
>>> pci_restore_ltr_value() to save and restore LTR latency value, while device is
>>> reset.
>>
>> We need some justification for these patches.  Why do we want them?
>> Do they improve performance?  Reduce power consumption?  How have they
>> been tested?  How can we be confident that these features work
>> correctly on hardware in the field?  Should or could the BIOS enable
>> them itself, based on OEM testing and desire to support these
>> features?
>
> I too am a little nervous about these changes due to Jesse's earlier response
> (see http://marc.info/?l=linux-pci&m=133372610102933&w=2) where he indicated:
>    "Given how device specific these extensions are, I'd expect you'd need
>     to know about each specific device anyway, which is why I think the
>     control belongs in the driver."
>
> Having these features enabled by default may be too aggressive.  Not saying it
> is not correct - something you may be able to inform us about, especially since
> you are with Intel - just make me nervous without further information.
>
> Myron
>
+1; like AER, I prefer the enablement be in the driver; when/if the
feature has proven itself reliable, then the kernel can enable it by default
In the case of the kernel & driver doing an enable, it won't hurt.
If want hook to disable by boot parameter, the kernel would have to clear
on scan, and put the disable *after* driver probe.

>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
> --
> 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