Re: [PATCH 4/8] drm/ttm: factor out ttm_bo_mmap_vma_setup

From: Thomas Zimmermann <tzimmermann@suse.de>
To: Gerd Hoffmann <kraxel@redhat.com>
Cc: David Airlie <airlied@linux.ie>,
	Daniel Vetter <daniel.vetter@ffwll.ch>,
	open list <linux-kernel@vger.kernel.org>,
	dri-devel@lists.freedesktop.org, Huang Rui <ray.huang@amd.com>,
	Christian Koenig <christian.koenig@amd.com>
Subject: Re: [PATCH 4/8] drm/ttm: factor out ttm_bo_mmap_vma_setup
Date: Tue, 17 Sep 2019 11:38:25 +0200	[thread overview]
Message-ID: <8b8ce387-1b3c-0a9c-3aaf-6294b3b81018@suse.de> (raw)
In-Reply-To: <20190917083425.kwwqyn463gn3mghf@sirius.home.kraxel.org>

[-- Attachment #1.1: Type: text/plain, Size: 2730 bytes --]

Hi Gerd

Am 17.09.19 um 10:34 schrieb Gerd Hoffmann:
> On Fri, Sep 13, 2019 at 03:05:34PM +0200, Thomas Zimmermann wrote:
> 
>>> +void ttm_bo_mmap_vma_setup(struct ttm_buffer_object *bo, struct vm_area_struct *vma)
>>> +{
>>> +	vma->vm_ops = &ttm_bo_vm_ops;
>>> +
>>> +	/*
>>> +	 * Note: We're transferring the bo reference to
>>> +	 * vma->vm_private_data here.
>>> +	 */
>>> +
>>> +	vma->vm_private_data = bo;
>>> +
>>> +	/*
>>> +	 * We'd like to use VM_PFNMAP on shared mappings, where
>>> +	 * (vma->vm_flags & VM_SHARED) != 0, for performance reasons,
>>> +	 * but for some reason VM_PFNMAP + x86 PAT + write-combine is very
>>> +	 * bad for performance. Until that has been sorted out, use
>>> +	 * VM_MIXEDMAP on all mappings. See freedesktop.org bug #75719
>>> +	 */
>>> +	vma->vm_flags |= VM_MIXEDMAP;
>>> +	vma->vm_flags |= VM_IO | VM_DONTEXPAND | VM_DONTDUMP;
>>> +}
>>> +EXPORT_SYMBOL(ttm_bo_mmap_vma_setup);
>>
>> To me, this function looks like an internal helper that should rather
>> remain internal.
> 
> Well, I'm moving that to a helper exactly to avoid drm gem ttm helpers
> messing with ttm internals.  To not them initialize vm_flags for
> example, and to avoid exporting ttm_bo_vm_ops.  Also to make sure ttm bo
> vma's are initialized the same way no matter which code path was taken
> to mmap the object.

It may not be worth blocking on this, so

  Acked-by: Thomas Zimmermann <tzimmermann@suse.de>

But I still think it's not a good interface because it exposes internal
details.

Please consider another idea: how about splitting off the ttm_bo_get()
and vma-flags setup of ttm_fbdev_mmap() into a separate function, like this:

void ttm_bo_mmap_refed(vma, bo)
{
	ttm_bo_get(bo)
	ttm_bo_mmap_vma_setup(vma);
}
EXPORT_SYMBOL(ttm_bo_mmap_refed)

int ttm_fbdev_mmap(vma, bo)
{
        if (vma->vm_pgoff != 0)
                return -EACCES;
	ttm_bo_mmap_refed(vma, bo);
	return 0;
}

That would allow to keep _vma_setup() an internal function.

ttm_fbdev_mmap() sounds like it is only for fbdev and the only user is
amdgpu. Can it be moved out of ttm entirely?

Best regards
Thomas

>> As mentioned in my reply to patch 5, maybe re-using
>> ttm_fbdev_mmap() could help.
> 
> No, the check in that function prevents that from working.
> 
> cheers,
>   Gerd
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
> 

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Linux GmbH, Maxfeldstrasse 5, 90409 Nuernberg, Germany
GF: Felix Imendörffer, Mary Higgins, Sri Rasiah
HRB 21284 (AG Nürnberg)

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]