Re: Linux 5.1.9 build failure with CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT=n

From: Sven Joachim
Date: Tue Jun 11 2019 - 15:48:14 EST


On 2019-06-11 22:08 +0300, Thomas Backlund wrote:

> Den 11-06-2019 kl. 20:40, skrev Greg Kroah-Hartman:
>> On Tue, Jun 11, 2019 at 07:33:16PM +0200, Daniel Vetter wrote:
>>> On Tue, Jun 11, 2019 at 5:37 PM Greg Kroah-Hartman
>>> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>>> On Tue, Jun 11, 2019 at 03:56:35PM +0200, Sven Joachim wrote:
>>>>> Commit 1e07d63749 ("drm/nouveau: add kconfig option to turn off nouveau
>>>>> legacy contexts. (v3)") has caused a build failure for me when I
>>>>> actually tried that option (CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT=n):
>>>>>
>>>>> ,----
>>>>> | Kernel: arch/x86/boot/bzImage is ready (#1)
>>>>> | Building modules, stage 2.
>>>>> | MODPOST 290 modules
>>>>> | ERROR: "drm_legacy_mmap" [drivers/gpu/drm/nouveau/nouveau.ko] undefined!
>>>>> | scripts/Makefile.modpost:91: recipe for target '__modpost' failed
>>>>> `----
>>>
>>> Calling drm_legacy_mmap is definitely not a great idea. I think either
>>> we need a custom patch to remove that out on older kernels, or maybe
>>> even #ifdef if you want to be super paranoid about breaking stuff ...
>>>
>>>>> Upstream does not have that problem, as commit bed2dd8421 ("drm/ttm:
>>>>> Quick-test mmap offset in ttm_bo_mmap()") has removed the use of
>>>>> drm_legacy_mmap from nouveau_ttm.c. Unfortunately that commit does not
>>>>> apply in 5.1.9.
>>>>>
>>>>> Most likely 4.19.50 and 4.14.125 are also affected, I haven't tested
>>>>> them yet.
>>>>
>>>> They probably are.
>>>>
>>>> Should I just revert this patch in the stable tree, or add some other
>>>> patch (like the one pointed out here, which seems an odd patch for
>>>> stable...)
>>>
>>> ... or backport the above patch, that should be save to do too. Not
>>> sure what stable folks prefer?
>>
>> The above patch does not apply to all of the stable branches, so how
>> about I just revert this? People can live with this option not able to
>> turn off for now, and if they really want it, they can use a newer
>> kernel, right?
>>
>
> Or add the simple fix suggested by Daniel (if I understand correctly):
>
>
> From: Thomas Backlund <tmb@xxxxxxxxxx>
>
> Setting CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT=n (added by commit:
> b30a43ac7132) causes the build to fail with:
>
> ERROR: "drm_legacy_mmap" [drivers/gpu/drm/nouveau/nouveau.ko] undefined!
>
> Fix that by adding check for CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT around
> the code using drm_legacy_mmap()
>
> Fixes: b30a43ac7132 drm/nouveau: add kconfig option to turn off
> nouveau legacy contexts. (v3)
> Signed-off-by: Thomas Backlund <tmb@xxxxxxxxxx>
>
> ---
> drivers/gpu/drm/nouveau/nouveau_ttm.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> --- a/drivers/gpu/drm/nouveau/nouveau_ttm.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_ttm.c
> @@ -168,8 +168,10 @@ nouveau_ttm_mmap(struct file *filp, stru
> struct drm_file *file_priv = filp->private_data;
> struct nouveau_drm *drm = nouveau_drm(file_priv->minor->dev);
>
> +#if defined(CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT)
> if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET))
> return drm_legacy_mmap(filp, vma);
> +#endif
>
> return ttm_bo_mmap(filp, vma, &drm->ttm.bdev);
> }

That's not quite correct, I am afraid. If
CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT is not defined, you still need to do
the test, but return -EINVAL. Something along these lines:

diff --git a/drivers/gpu/drm/nouveau/nouveau_ttm.c b/drivers/gpu/drm/nouveau/nouveau_ttm.c
index 1543c2f8d3d3..05d513d54555 100644
--- a/drivers/gpu/drm/nouveau/nouveau_ttm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_ttm.c
@@ -169,7 +169,11 @@ nouveau_ttm_mmap(struct file *filp, struct vm_area_struct *vma)
struct nouveau_drm *drm = nouveau_drm(file_priv->minor->dev);

if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET))
+#if defined(CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT)
return drm_legacy_mmap(filp, vma);
+#else
+ return -EINVAL;
+#endif

return ttm_bo_mmap(filp, vma, &drm->ttm.bdev);
}


At least that builds for me, need to reboot to check whether it works.

Cheers,
Sven