Re: Linux 5.1.9 build failure with CONFIG_NOUVEAU_LEGACY_CTX_SUPPORT=n

[Daniel Vetter]

On Tue, Jun 11, 2019 at 10:08:10PM +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>

Reviewed-by: Daniel Vetter <daniel.vetter@xxxxxxxx>

Not-entirely-upstream-sha1-but-equivalent: bed2dd8421 ("drm/ttm:
Quick-test mmap offset in ttm_bo_mmap()")

Cheers, Daniel

> 
> ---
>  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);
>  }
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch