From: Zack Rusin <[email protected]>
[ Upstream commit f54c4442893b8dfbd3aff8e903c54dfff1aef990 ]
ttm_mem_type_manager_func.get_node was changed to return -ENOSPC
instead of setting the node pointer to NULL. Unfortunately
vmwgfx still had two places where it was explicitly converting
-ENOSPC to 0 causing regressions. This fixes those spots by
allowing -ENOSPC to be returned. That seems to fix recent
regressions with vmwgfx.
Signed-off-by: Zack Rusin <[email protected]>
Reviewed-by: Roland Scheidegger <[email protected]>
Reviewed-by: Martin Krastev <[email protected]>
Sigend-off-by: Roland Scheidegger <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>
---
drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c | 2 +-
drivers/gpu/drm/vmwgfx/vmwgfx_thp.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c b/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c
index 7da752ca1c34b..b93c558dd86e0 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c
@@ -57,7 +57,7 @@ static int vmw_gmrid_man_get_node(struct ttm_mem_type_manager *man,
id = ida_alloc_max(&gman->gmr_ida, gman->max_gmr_ids - 1, GFP_KERNEL);
if (id < 0)
- return (id != -ENOMEM ? 0 : id);
+ return id;
spin_lock(&gman->lock);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c b/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c
index b7c816ba71663..c8b9335bccd8d 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c
@@ -95,7 +95,7 @@ found_unlock:
mem->start = node->start;
}
- return 0;
+ return ret;
}
--
2.25.1
Hi,
this commit should NOT be applied to 5.8.
It fixes a regression introduced by
58e4d686d456c3e356439ae160ff4a0728940b8e (drm/ttm: cleanup
ttm_mem_type_manager_func.get_node interface v3) which is part of 5.9
but not 5.8.
Applying this to 5.8 will very likely break things. I don't know why it
ended up as a candidate for 5.8.
Roland
Am 12.10.20 um 15:30 schrieb Greg Kroah-Hartman:
> From: Zack Rusin <[email protected]>
>
> [ Upstream commit f54c4442893b8dfbd3aff8e903c54dfff1aef990 ]
>
> ttm_mem_type_manager_func.get_node was changed to return -ENOSPC
> instead of setting the node pointer to NULL. Unfortunately
> vmwgfx still had two places where it was explicitly converting
> -ENOSPC to 0 causing regressions. This fixes those spots by
> allowing -ENOSPC to be returned. That seems to fix recent
> regressions with vmwgfx.
>
> Signed-off-by: Zack Rusin <[email protected]>
> Reviewed-by: Roland Scheidegger <[email protected]>
> Reviewed-by: Martin Krastev <[email protected]>
> Sigend-off-by: Roland Scheidegger <[email protected]>
> Signed-off-by: Sasha Levin <[email protected]>
> ---
> drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c | 2 +-
> drivers/gpu/drm/vmwgfx/vmwgfx_thp.c | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c b/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c
> index 7da752ca1c34b..b93c558dd86e0 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_gmrid_manager.c
> @@ -57,7 +57,7 @@ static int vmw_gmrid_man_get_node(struct ttm_mem_type_manager *man,
>
> id = ida_alloc_max(&gman->gmr_ida, gman->max_gmr_ids - 1, GFP_KERNEL);
> if (id < 0)
> - return (id != -ENOMEM ? 0 : id);
> + return id;
>
> spin_lock(&gman->lock);
>
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c b/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c
> index b7c816ba71663..c8b9335bccd8d 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_thp.c
> @@ -95,7 +95,7 @@ found_unlock:
> mem->start = node->start;
> }
>
> - return 0;
> + return ret;
> }
>
>
>
On Tue, Oct 13, 2020 at 05:55:31PM +0200, Roland Scheidegger wrote:
> Hi,
>
> this commit should NOT be applied to 5.8.
> It fixes a regression introduced by
> 58e4d686d456c3e356439ae160ff4a0728940b8e (drm/ttm: cleanup
> ttm_mem_type_manager_func.get_node interface v3) which is part of 5.9
> but not 5.8.
> Applying this to 5.8 will very likely break things. I don't know why it
> ended up as a candidate for 5.8.
Now dropped, thanks. And it was probably picked up due to the wording
in the changelog text, along with a lack of a "Fixes:" tag that pointed
at the exact change it fixed up, which would have shown that this is a
5.9-only thing.
thanks,
greg k-h