[PATCH 3/3] staging: vme_user: alloc buffer when setting slave

Aaron Sierra asierra at xes-inc.com
Fri Sep 9 20:44:55 UTC 2016


----- Original Message -----
> From: "Martyn Welch" <martyn at welchs.me.uk>
> Sent: Friday, September 9, 2016 3:03:57 PM

> On Fri, Sep 09, 2016 at 02:41:43PM -0500, Aaron Sierra wrote:
>> ----- Original Message -----
>> > From: "Martyn Welch" <martyn at welchs.me.uk>
>> > To: "Aaron Sierra" <asierra at xes-inc.com>
>> > Cc: "Manohar Vanga" <manohar.vanga at gmail.com>, "Greg Kroah-Hartman"
>> > <gregkh at linuxfoundation.org>, "devel"
>> > <devel at driverdev.osuosl.org>, "Egor Uleyskiy" <egor.ulieiskii at gmail.com>,
>> > "Dmitry Kalinkin" <dmitry.kalinkin at gmail.com>
>> > Sent: Thursday, September 8, 2016 2:34:32 PM
>> > Subject: Re: [PATCH 3/3] staging: vme_user: alloc buffer when setting slave
>> 
>> > On Fri, Sep 02, 2016 at 04:17:01PM -0500, Aaron Sierra wrote:
>> >> Some bridges support a minimum slave window larger than this driver's
>> >> 128 KB default buffer size. Also bridges typically require the PCI and
>> >> VME bases to be aligned to the size of the window.
>> >> 
>> > 
>> > Which bridge has a minimum slave window larger than 128KB?
>> 
>> Martyn,
>> 
>> The ALTHEA 7910 PCIe-VME bridge from IOxOS Technologies can require a
>> PCI-side window be larger than 128 KiB.
>> 
>> I've been meaning to submit our driver for it.
>> 
> 
> Ah, OK. In that case please can you submit the driver first. At this point
> in time there are no in tree use cases for the changes your proposing.
> 

Martyn,

I could submit my driver in the state that it's in, but I don't think it's
quite ready for review. I decided to submit these patches now since I
noticed that code I rely on had been removed from the kernel because of
disuse. This patch provides the use case AND helps support my bridge.

Nothing in this patch is really ALTHEA-specific. I could easily adjust the
language in my commit log to omit the implicit reference to the ALTHEA
bridge. Would something like this be more appealing?

    staging: vme_user: alloc buffer when setting slave

    Don't allocate more PCI space than is absolutely needed to support the
    VME bridge window. We don't know how much space is needed, until the
    user calls the VME_SET_SLAVE ioctl.

-Aaron S.


More information about the devel mailing list