| Message ID | 1353059674-17349-1-git-send-email-aaro.koskinen@iki.fi (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fre, 2012-11-16 at 11:54 +0200, Aaro Koskinen wrote: > Check that the AGP aperture can be mapped. This follows a similar change > done for Radeon (commit 365048ff, drm/radeon: AGP memory is only I/O if > the aperture can be mapped by the CPU.). > > The patch fixes the following error seen on G5 iMac: > > nouveau E[ DRM] failed to create kernel channel, -12 > > Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi> > --- > drivers/gpu/drm/nouveau/nouveau_bo.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c > index 35ac57f..5f0e7ef 100644 > --- a/drivers/gpu/drm/nouveau/nouveau_bo.c > +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c > @@ -1279,7 +1279,7 @@ nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) > if (drm->agp.stat == ENABLED) { > mem->bus.offset = mem->start << PAGE_SHIFT; > mem->bus.base = drm->agp.base; > - mem->bus.is_iomem = true; > + mem->bus.is_iomem = !dev->agp->cant_use_aperture; Looks good, though radeon also doesn't set TTM_MEMTYPE_FLAG_MAPPABLE for the TT memory type if the aperture can't be mapped directly. Maybe nouveau should do the same, or maybe it doesn't really matter. Not sure.
Hi, On Fri, Nov 16, 2012 at 03:16:09PM +0100, Michel Dänzer wrote: > On Fre, 2012-11-16 at 11:54 +0200, Aaro Koskinen wrote: > > Check that the AGP aperture can be mapped. This follows a similar change > > done for Radeon (commit 365048ff, drm/radeon: AGP memory is only I/O if > > the aperture can be mapped by the CPU.). > > > > The patch fixes the following error seen on G5 iMac: > > > > nouveau E[ DRM] failed to create kernel channel, -12 > > > > Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi> > > --- > > drivers/gpu/drm/nouveau/nouveau_bo.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c > > index 35ac57f..5f0e7ef 100644 > > --- a/drivers/gpu/drm/nouveau/nouveau_bo.c > > +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c > > @@ -1279,7 +1279,7 @@ nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) > > if (drm->agp.stat == ENABLED) { > > mem->bus.offset = mem->start << PAGE_SHIFT; > > mem->bus.base = drm->agp.base; > > - mem->bus.is_iomem = true; > > + mem->bus.is_iomem = !dev->agp->cant_use_aperture; > > Looks good, though radeon also doesn't set TTM_MEMTYPE_FLAG_MAPPABLE for > the TT memory type if the aperture can't be mapped directly. Maybe > nouveau should do the same, or maybe it doesn't really matter. Not sure. This would fail with -EINVAL if the MAPPABLE flag is not set. So I think it's needed? Looking at radeon_ttm.c, the MAPPABLE flag seems to be set in both cases: 141 man->flags = TTM_MEMTYPE_FLAG_MAPPABLE | TTM_MEMTYPE_FLAG_CMA; 142 #if __OS_HAS_AGP 143 if (rdev->flags & RADEON_IS_AGP) { 144 if (!(drm_core_has_AGP(rdev->ddev) && rdev->ddev->agp)) { 145 DRM_ERROR("AGP is not enabled for memory type %u\n", 146 (unsigned)type); 147 return -EINVAL; 148 } 149 if (!rdev->ddev->agp->cant_use_aperture) 150 man->flags = TTM_MEMTYPE_FLAG_MAPPABLE; A.
On Fre, 2012-11-16 at 20:11 +0200, Aaro Koskinen wrote: > On Fri, Nov 16, 2012 at 03:16:09PM +0100, Michel Dänzer wrote: > > On Fre, 2012-11-16 at 11:54 +0200, Aaro Koskinen wrote: > > > Check that the AGP aperture can be mapped. This follows a similar change > > > done for Radeon (commit 365048ff, drm/radeon: AGP memory is only I/O if > > > the aperture can be mapped by the CPU.). > > > > > > The patch fixes the following error seen on G5 iMac: > > > > > > nouveau E[ DRM] failed to create kernel channel, -12 > > > > > > Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi> > > > --- > > > drivers/gpu/drm/nouveau/nouveau_bo.c | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c > > > index 35ac57f..5f0e7ef 100644 > > > --- a/drivers/gpu/drm/nouveau/nouveau_bo.c > > > +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c > > > @@ -1279,7 +1279,7 @@ nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) > > > if (drm->agp.stat == ENABLED) { > > > mem->bus.offset = mem->start << PAGE_SHIFT; > > > mem->bus.base = drm->agp.base; > > > - mem->bus.is_iomem = true; > > > + mem->bus.is_iomem = !dev->agp->cant_use_aperture; > > > > Looks good, though radeon also doesn't set TTM_MEMTYPE_FLAG_MAPPABLE for > > the TT memory type if the aperture can't be mapped directly. Maybe > > nouveau should do the same, or maybe it doesn't really matter. Not sure. > > This would fail with -EINVAL if the MAPPABLE flag is not set. So I think > it's needed? > > Looking at radeon_ttm.c, the MAPPABLE flag seems to be set in both cases: Ah, indeed, I misread that. Your patch is Reviewed-by: Michel Dänzer <michel@daenzer.net>
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 35ac57f..5f0e7ef 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -1279,7 +1279,7 @@ nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) if (drm->agp.stat == ENABLED) { mem->bus.offset = mem->start << PAGE_SHIFT; mem->bus.base = drm->agp.base; - mem->bus.is_iomem = true; + mem->bus.is_iomem = !dev->agp->cant_use_aperture; } #endif break;
Check that the AGP aperture can be mapped. This follows a similar change done for Radeon (commit 365048ff, drm/radeon: AGP memory is only I/O if the aperture can be mapped by the CPU.). The patch fixes the following error seen on G5 iMac: nouveau E[ DRM] failed to create kernel channel, -12 Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi> --- drivers/gpu/drm/nouveau/nouveau_bo.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)