[PATCH] drm/xe: Fix build error in xe_ggtt.c

Lucas De Marchi lucas.demarchi at intel.com
Fri Mar 1 04:23:55 UTC 2024


On Thu, Feb 29, 2024 at 12:52:26AM +0000, Matthew Brost wrote:
>On Wed, Feb 28, 2024 at 12:49:05PM -0600, Lucas De Marchi wrote:
>> On Wed, Feb 28, 2024 at 03:47:13PM +0100, Thomas Hellström wrote:
>> > On Mon, 2024-02-26 at 15:40 +0000, Matthew Brost wrote:
>> > > On Mon, Feb 26, 2024 at 10:05:58AM +0100, Thomas Hellström wrote:
>> > > > On Sat, 2024-02-24 at 16:14 -0800, Matthew Brost wrote:
>> > > > > Need to include io-64-nonatomic-lo-hi.h for writeq function.
>> > > >
>> > > > As I understand it, the choice of header here determines the dword
>> > > > write order on 32-bit systems that don't have an atomic writeq(),
>> > > >
>> > > > So is writing the low dword first the correct order in this case?
>> > > > Perhaps add a motivation in the commit message?
>> > > >
>> > >
>> > > "Cleanup some layering in GGTT" removed xe_mmio.h from xe_gt.c and
>> > > that
>> > > file includes linux/io-64-nonatomic-lo-hi.h. Perhaps it is better
>> > > just
>> > > to include xe_mmio.h again in xe_gt.c?
>> >
>> > I think it then makes sense to use your original patch, so that it's
>> > easier to follow for each subsystem what ordering is used.
>>
>> what original patch, I only find this one in the mailing list.
>> and it seems correct to me, following the "include what you use"
>> approach.
>>
>
>I'm also confused by Thomas comment.
>
>Let me ask directly - do you prefer xe_mmio.h or
>linux/io-64-nonatomic-lo-hi.h to be included in xe_ggtt.c?

sync'ed with Thomas and he's onboard with what's in this patch.

Applied, thanks!

[1/1] drm/xe: Fix build error in xe_ggtt.c
       commit: d0a5fb2e0a5abeaad983c5c5c52b88ccef3aaae2

Lucas De Marchi


More information about the Intel-xe mailing list