[PATCH 1/2] drm/radeon: allways add the VM clear duplicate

[Alex Deucher]

On Wed, Jul 8, 2015 at 12:31 PM, Christian König
<deathsimple at vodafone.de> wrote:
> On 08.07.2015 18:14, Alex Deucher wrote:
>>
>> On Wed, Jul 8, 2015 at 12:08 PM, Christian König
>> <deathsimple at vodafone.de> wrote:
>>>
>>> On 08.07.2015 17:57, Alex Deucher wrote:
>>>>
>>>> On Wed, Jul 8, 2015 at 3:56 AM, Christian König
>>>> <deathsimple at vodafone.de>
>>>> wrote:
>>>>>
>>>>> From: Christian König <christian.koenig at amd.com>
>>>>>
>>>>> We need to allways add the VM clear duplicate of the BO_VA,
>>>>> no matter what the old status was.
>>>>>
>>>>> Signed-off-by: Christian König <christian.koenig at amd.com>
>>>>> Test-by: hadack at gmx.de
>>>>
>>>> This patch does not apply cleanly.  It also adds an undefined label
>>>> error_unreserve.  Is there an additional patch required?
>>>
>>>
>>> That's possible, but I thought the depending patch would have already
>>> been
>>> upstream.
>>>
>>> Can you create a drm-fixes-4.2 branch I can rebase on?
>>
>> http://cgit.freedesktop.org/~agd5f/linux/log/?h=drm-fixes-4.2-wip
>
>
> Ah, here is the problem. This branch don't yet contains the fixes which went
> upstream with 4.1.
>
> What should we do? I don't want to change those patches and create a merge
> conflict when you merge in 4.1.

I'll rebase by tree on Linus' tree.  Dave is on vacation so his tree is lagging.

Alex

>
> Christian.
>
>
>>
>> Thanks,
>>
>> Alex
>>
>>> Thanks,
>>> Christian.
>>>
>>>
>>>> Alex
>>>>
>>>>> ---
>>>>>    drivers/gpu/drm/radeon/radeon_vm.c | 34
>>>>> ++++++++++++++++------------------
>>>>>    1 file changed, 16 insertions(+), 18 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/radeon/radeon_vm.c
>>>>> b/drivers/gpu/drm/radeon/radeon_vm.c
>>>>> index ec10533..0310965 100644
>>>>> --- a/drivers/gpu/drm/radeon/radeon_vm.c
>>>>> +++ b/drivers/gpu/drm/radeon/radeon_vm.c
>>>>> @@ -493,29 +493,27 @@ int radeon_vm_bo_set_addr(struct radeon_device
>>>>> *rdev,
>>>>>           }
>>>>>
>>>>>           if (bo_va->it.start || bo_va->it.last) {
>>>>> -               spin_lock(&vm->status_lock);
>>>>> -               if (list_empty(&bo_va->vm_status)) {
>>>>> -                       /* add a clone of the bo_va to clear the old
>>>>> address */
>>>>> -                       struct radeon_bo_va *tmp;
>>>>> -                       spin_unlock(&vm->status_lock);
>>>>> -                       tmp = kzalloc(sizeof(struct radeon_bo_va),
>>>>> GFP_KERNEL);
>>>>> -                       if (!tmp) {
>>>>> -                               mutex_unlock(&vm->mutex);
>>>>> -                               r = -ENOMEM;
>>>>> -                               goto error_unreserve;
>>>>> -                       }
>>>>> -                       tmp->it.start = bo_va->it.start;
>>>>> -                       tmp->it.last = bo_va->it.last;
>>>>> -                       tmp->vm = vm;
>>>>> -                       tmp->bo = radeon_bo_ref(bo_va->bo);
>>>>> -                       spin_lock(&vm->status_lock);
>>>>> -                       list_add(&tmp->vm_status, &vm->freed);
>>>>> +               /* add a clone of the bo_va to clear the old address */
>>>>> +               struct radeon_bo_va *tmp;
>>>>> +               tmp = kzalloc(sizeof(struct radeon_bo_va), GFP_KERNEL);
>>>>> +               if (!tmp) {
>>>>> +                       mutex_unlock(&vm->mutex);
>>>>> +                       r = -ENOMEM;
>>>>> +                       goto error_unreserve;
>>>>>                   }
>>>>> -               spin_unlock(&vm->status_lock);
>>>>> +               tmp->it.start = bo_va->it.start;
>>>>> +               tmp->it.last = bo_va->it.last;
>>>>> +               tmp->vm = vm;
>>>>> +               tmp->bo = radeon_bo_ref(bo_va->bo);
>>>>>
>>>>>                   interval_tree_remove(&bo_va->it, &vm->va);
>>>>>                   bo_va->it.start = 0;
>>>>>                   bo_va->it.last = 0;
>>>>> +
>>>>> +               spin_lock(&vm->status_lock);
>>>>> +               list_del_init(&bo_va->vm_status);
>>>>> +               list_add(&tmp->vm_status, &vm->freed);
>>>>> +               spin_unlock(&vm->status_lock);
>>>>>           }
>>>>>
>>>>>           if (soffset || eoffset) {
>>>>> --
>>>>> 1.9.1
>>>>>
>