[PATCH 18/27] glamor: Allow nested mapping of pixmaps.
Markus Wick
markus at selfnet.de
Wed Mar 12 12:34:15 PDT 2014
Am 2014-03-11 22:30, schrieb Eric Anholt:
> The common pattern is to do nested if statements making calls to
> prepare_access() and then pop those mappings back off in each set of
> braces. Some cases checked for src == dst to avoid leaking mappings,
> but others didn't. Others didn't even do the nested mappings, so a
> failure in the outer map would result in trying to umap the inner and
> failing.
>
> By allowing nested mappings, we can fix both problems by not requiring
> the care from the caller, plus we can allow a simpler nesting of all
> the prepares in one if statement.
>
> Signed-off-by: Eric Anholt <eric at anholt.net>
> ---
> glamor/glamor_core.c | 19 ++++++++++++++++++-
> glamor/glamor_priv.h | 6 ++++++
> 2 files changed, 24 insertions(+), 1 deletion(-)
>
> diff --git a/glamor/glamor_core.c b/glamor/glamor_core.c
> index 5883809..7a7ca08 100644
> --- a/glamor/glamor_core.c
> +++ b/glamor/glamor_core.c
> @@ -105,6 +105,19 @@ Bool
> glamor_prepare_access(DrawablePtr drawable, glamor_access_t access)
> {
> PixmapPtr pixmap = glamor_get_drawable_pixmap(drawable);
> + glamor_pixmap_private *pixmap_priv =
> glamor_get_pixmap_private(pixmap);
> +
> + if (pixmap->devPrivate.ptr) {
> + /* Already mapped, nothing needs to be done. Note that we
> + * aren't allowing promotion from RO to RW, because it would
> + * require re-mapping the PBO.
> + */
> + assert(!GLAMOR_PIXMAP_PRIV_HAS_FBO(pixmap_priv) ||
> + access == GLAMOR_ACCESS_RO ||
> + pixmap_priv->base.mapped_for_write);
> + return TRUE;
> + }
> + pixmap_priv->base.mapped_for_write = (access == GLAMOR_ACCESS_RW);
>
> return glamor_download_pixmap_to_cpu(pixmap, access);
> }
> @@ -300,7 +313,11 @@ glamor_finish_access(DrawablePtr drawable,
> glamor_access_t access_mode)
> if (!GLAMOR_PIXMAP_PRIV_HAS_FBO_DOWNLOADED(pixmap_priv))
> return;
>
> - if (access_mode != GLAMOR_ACCESS_RO) {
> + /* If we are doing a series of unmaps from a nested map, we're
> done. */
> + if (!pixmap->devPrivate.ptr)
> + return;
In my opinion, there should be a note that this will unmap on the
innermost call to finish_access, but the outer functions maybe still
want to access this pixmap.
> +
> + if (pixmap_priv->base.mapped_for_write) {
> glamor_restore_pixmap_to_texture(pixmap);
> }
>
> diff --git a/glamor/glamor_priv.h b/glamor/glamor_priv.h
> index 7f41025..24a3575 100644
> --- a/glamor/glamor_priv.h
> +++ b/glamor/glamor_priv.h
> @@ -410,6 +410,12 @@ typedef struct glamor_pixmap_clipped_regions {
> typedef struct glamor_pixmap_private_base {
> glamor_pixmap_type_t type;
> enum glamor_fbo_state gl_fbo;
> + /**
> + * If devPrivate.ptr is non-NULL (meaning we're within
> + * glamor_prepare_access), determies whether we should re-upload
> + * that data on glamor_finish_access().
> + */
> + bool mapped_for_write;
Why haven't you used the enum glamor_access? Memory footprint shouldn't
matter that much.
> unsigned char is_picture:1;
> unsigned char gl_tex:1;
> glamor_pixmap_fbo *fbo;
More information about the xorg-devel
mailing list