[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-26 Thread Ben Skeggs
On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: > On 01/25/2012 10:41 AM, Ben Skeggs wrote: > > > > My main concern is that we blindly and unnecessarily set up GPU bindings and > > end up with unnecessary code in TTM, and furthermore that we communicate > > that bad practice to future dr

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 09:39 +0100, Thomas Hellstrom wrote: > On 01/25/2012 09:05 AM, Ben Skeggs wrote: > > On Wed, 2012-01-25 at 08:43 +0100, Thomas Hellstrom wrote: > >> On 01/25/2012 06:34 AM, Ben Skeggs wrote: > >>> From: Ben Skeggs > >>> > >>> Both changes in dc97b3409a790d2a21aac6e5cdb99558b59

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 07:12 PM, Dave Airlie wrote: > On Wed, Jan 25, 2012 at 5:19 PM, Thomas Hellstrom > wrote: >> On 01/25/2012 04:37 PM, Jerome Glisse wrote: >>> On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggswrote: On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: > On 01/25/2012 10

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 08:24 +, Dave Airlie wrote: > On Wed, Jan 25, 2012 at 5:34 AM, Ben Skeggs wrote: > > From: Ben Skeggs > > > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > > regressions in the nouveau driver. > > > > move_notify() was originally able to presum

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 06:34 AM, Ben Skeggs wrote: > From: Ben Skeggs > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > regressions in the nouveau driver. > > move_notify() was originally able to presume that bo->mem is the old node, > and new_mem is the new node. The above commi

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 04:37 PM, Jerome Glisse wrote: > On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs wrote: >> On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: >>> On 01/25/2012 10:41 AM, Ben Skeggs wrote: My main concern is that we blindly and unnecessarily set up GPU bindings and >>>

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 04:37 PM, Jerome Glisse wrote: > On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs wrote: >> On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: >>> On 01/25/2012 10:41 AM, Ben Skeggs wrote: My main concern is that we blindly and unnecessarily set up GPU bindings and >>>

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Dave Airlie
On Wed, Jan 25, 2012 at 5:19 PM, Thomas Hellstrom wrote: > On 01/25/2012 04:37 PM, Jerome Glisse wrote: >> >> On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs ?wrote: >>> >>> On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: On 01/25/2012 10:41 AM, Ben Skeggs wrote: > > My

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 08:43 +0100, Thomas Hellstrom wrote: > On 01/25/2012 06:34 AM, Ben Skeggs wrote: > > From: Ben Skeggs > > > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > > regressions in the nouveau driver. > > > > move_notify() was originally able to presume that

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
From: Ben Skeggs Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious regressions in the nouveau driver. move_notify() was originally able to presume that bo->mem is the old node, and new_mem is the new node. The above commit moves the call to move_notify() to after move() has

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 10:41 AM, Ben Skeggs wrote: > > My main concern is that we blindly and unnecessarily set up GPU bindings and > end up with unnecessary code in TTM, and furthermore that we communicate > that bad practice to future driver writers. > This "unnecessary code" is like 5 lines of cleanup if

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Jerome Glisse
On Wed, Jan 25, 2012 at 1:21 PM, Thomas Hellstrom wrote: > On 01/25/2012 07:12 PM, Dave Airlie wrote: >> >> On Wed, Jan 25, 2012 at 5:19 PM, Thomas Hellstrom >> ?wrote: >>> >>> On 01/25/2012 04:37 PM, Jerome Glisse wrote: On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs ?wrote: > >

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Jerome Glisse
On Wed, Jan 25, 2012 at 1:21 PM, Thomas Hellstrom wrote: > On 01/25/2012 07:12 PM, Dave Airlie wrote: >> >> On Wed, Jan 25, 2012 at 5:19 PM, Thomas Hellstrom >>  wrote: >>> >>> On 01/25/2012 04:37 PM, Jerome Glisse wrote: On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs  wrote: > >>

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Jerome Glisse
On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs wrote: > On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: >> On 01/25/2012 10:41 AM, Ben Skeggs wrote: >> > >> > My main concern is that we blindly and unnecessarily set up GPU bindings >> > and >> > end up with unnecessary code in TTM, and f

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 07:12 PM, Dave Airlie wrote: On Wed, Jan 25, 2012 at 5:19 PM, Thomas Hellstrom wrote: On 01/25/2012 04:37 PM, Jerome Glisse wrote: On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggswrote: On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: On 01/25/2012 10:41 AM, Ben Skegg

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Dave Airlie
On Wed, Jan 25, 2012 at 5:19 PM, Thomas Hellstrom wrote: > On 01/25/2012 04:37 PM, Jerome Glisse wrote: >> >> On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs  wrote: >>> >>> On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: On 01/25/2012 10:41 AM, Ben Skeggs wrote: > > My m

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 09:05 AM, Ben Skeggs wrote: > On Wed, 2012-01-25 at 08:43 +0100, Thomas Hellstrom wrote: >> On 01/25/2012 06:34 AM, Ben Skeggs wrote: >>> From: Ben Skeggs >>> >>> Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious >>> regressions in the nouveau driver. >>> >>> move

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 06:34 AM, Ben Skeggs wrote: From: Ben Skeggs Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious regressions in the nouveau driver. move_notify() was originally able to presume that bo->mem is the old node, and new_mem is the new node. The above commit moves th

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 04:37 PM, Jerome Glisse wrote: On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs wrote: On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: On 01/25/2012 10:41 AM, Ben Skeggs wrote: My main concern is that we blindly and unnecessarily set up GPU bindings and end up with unnece

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 04:37 PM, Jerome Glisse wrote: On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs wrote: On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: On 01/25/2012 10:41 AM, Ben Skeggs wrote: My main concern is that we blindly and unnecessarily set up GPU bindings and end up with unnece

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 06:34 AM, Ben Skeggs wrote: > From: Ben Skeggs > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > regressions in the nouveau driver. > > move_notify() was originally able to presume that bo->mem is the old node, > and new_mem is the new node. The above commi

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Dave Airlie
On Wed, Jan 25, 2012 at 5:34 AM, Ben Skeggs wrote: > From: Ben Skeggs > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > regressions in the nouveau driver. > > move_notify() was originally able to presume that bo->mem is the old node, > and new_mem is the new node. ?The

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Jerome Glisse
On Wed, Jan 25, 2012 at 10:21 AM, Ben Skeggs wrote: > On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: >> On 01/25/2012 10:41 AM, Ben Skeggs wrote: >> > >> > My main concern is that we blindly and unnecessarily set up GPU bindings >> > and >> > end up with unnecessary code in TTM, and f

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 15:33 +0100, Thomas Hellstrom wrote: > On 01/25/2012 10:41 AM, Ben Skeggs wrote: > > > > My main concern is that we blindly and unnecessarily set up GPU bindings and > > end up with unnecessary code in TTM, and furthermore that we communicate > > that bad practice to future dr

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 10:41 AM, Ben Skeggs wrote: My main concern is that we blindly and unnecessarily set up GPU bindings and end up with unnecessary code in TTM, and furthermore that we communicate that bad practice to future driver writers. This "unnecessary code" is like 5 lines of cleanup if someth

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 09:39 +0100, Thomas Hellstrom wrote: > On 01/25/2012 09:05 AM, Ben Skeggs wrote: > > On Wed, 2012-01-25 at 08:43 +0100, Thomas Hellstrom wrote: > >> On 01/25/2012 06:34 AM, Ben Skeggs wrote: > >>> From: Ben Skeggs > >>> > >>> Both changes in dc97b3409a790d2a21aac6e5cdb99558b59

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 09:05 AM, Ben Skeggs wrote: On Wed, 2012-01-25 at 08:43 +0100, Thomas Hellstrom wrote: On 01/25/2012 06:34 AM, Ben Skeggs wrote: From: Ben Skeggs Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious regressions in the nouveau driver. move_notify() was originall

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 08:24 +, Dave Airlie wrote: > On Wed, Jan 25, 2012 at 5:34 AM, Ben Skeggs wrote: > > From: Ben Skeggs > > > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > > regressions in the nouveau driver. > > > > move_notify() was originally able to presum

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Dave Airlie
On Wed, Jan 25, 2012 at 5:34 AM, Ben Skeggs wrote: > From: Ben Skeggs > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > regressions in the nouveau driver. > > move_notify() was originally able to presume that bo->mem is the old node, > and new_mem is the new node.  The

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Ben Skeggs
On Wed, 2012-01-25 at 08:43 +0100, Thomas Hellstrom wrote: > On 01/25/2012 06:34 AM, Ben Skeggs wrote: > > From: Ben Skeggs > > > > Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious > > regressions in the nouveau driver. > > > > move_notify() was originally able to presume that

Re: [PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-25 Thread Thomas Hellstrom
On 01/25/2012 06:34 AM, Ben Skeggs wrote: From: Ben Skeggs Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious regressions in the nouveau driver. move_notify() was originally able to presume that bo->mem is the old node, and new_mem is the new node. The above commit moves th

[PATCH] drm/ttm: fix two regressions since move_notify changes

2012-01-24 Thread Ben Skeggs
From: Ben Skeggs Both changes in dc97b3409a790d2a21aac6e5cdb99558b5944119 cause serious regressions in the nouveau driver. move_notify() was originally able to presume that bo->mem is the old node, and new_mem is the new node. The above commit moves the call to move_notify() to after move() has