fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-15 20:20:25 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

amd/vpelib: Fix memory leak from segment ctx

Browse source 
[WHY]
There is potential memory leak from vpe_alloc_segment_ctx.
This memory leak occurs only in multi-frame VPE tests where
between vpe_create and vpe_destroy, multiple calls are made
to vpe_check_support that allocates new segment context without
releasing the old one.

[HOW]
Allocate segment_ctx only when it is not already allocated. If it is
already allocated, check whether re-allocation is needed. If not, skip
the allocation.

Signed-off-by: Phoebe Chen <phoebe.chen@amd.com>
Reviewed-by: Roy Chan <Roy.Chan@amd.com>
Reviewed-by: Jesse Agate <Jesse.Agate@amd.com>
Acked-by: ChuanYu Tseng <ChuanYu.Tseng@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35012>
This commit is contained in:
Chen, Phoebe 2025-02-12 16:48:17 -05:00 • committed by ChuanYu Tseng (Max)
parent c1a3d021f3
commit f2e65c02e6
4 changed files with 39 additions and 24 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

17
src/amd/vpelib/src/chip/vpe10/vpe10_resource.c
View file

@ -249,9 +249,10 @@ enum vpe_status vpe10_set_num_segments(struct vpe_priv *vpe_priv, struct stream_
struct scaler_data *scl_data, struct vpe_rect *src_rect, struct vpe_rect *dst_rect,
uint32_t *max_seg_width, uint32_t recout_width_alignment)
{
uint16_t num_segs;
struct dpp *dpp = vpe_priv->resource.dpp[0];
const uint32_t max_lb_size = dpp->funcs->get_line_buffer_size();
uint16_t num_segs;
struct dpp *dpp = vpe_priv->resource.dpp[0];
const uint32_t max_lb_size = dpp->funcs->get_line_buffer_size();
enum vpe_status res = VPE_STATUS_OK;
(void)recout_width_alignment;
@ -259,13 +260,13 @@ enum vpe_status vpe10_set_num_segments(struct vpe_priv *vpe_priv, struct stream_
num_segs = vpe_get_num_segments(vpe_priv, src_rect, dst_rect, *max_seg_width);
stream_ctx->segment_ctx = vpe_alloc_segment_ctx(vpe_priv, num_segs);
if (!stream_ctx->segment_ctx)
return VPE_STATUS_NO_MEMORY;
res = vpe_alloc_segment_ctx(vpe_priv, stream_ctx, num_segs);
stream_ctx->num_segments = num_segs;
if (res == VPE_STATUS_OK) {
stream_ctx->num_segments = num_segs;
}
return VPE_STATUS_OK;
return res;
}
bool vpe10_get_dcc_compression_output_cap(const struct vpe *vpe, const struct vpe_dcc_surface_param *params, struct vpe_surface_dcc_cap *cap)

17
src/amd/vpelib/src/chip/vpe11/vpe11_resource.c
View file

@ -237,9 +237,10 @@ enum vpe_status vpe11_set_num_segments(struct vpe_priv *vpe_priv, struct stream_
struct scaler_data *scl_data, struct vpe_rect *src_rect, struct vpe_rect *dst_rect,
uint32_t *max_seg_width, uint32_t recout_width_alignment)
{
uint16_t num_segs;
struct dpp *dpp = vpe_priv->resource.dpp[0];
const uint32_t max_lb_size = dpp->funcs->get_line_buffer_size();
uint16_t num_segs;
struct dpp *dpp = vpe_priv->resource.dpp[0];
const uint32_t max_lb_size = dpp->funcs->get_line_buffer_size();
enum vpe_status res = VPE_STATUS_OK;
(void)recout_width_alignment;
*max_seg_width = min(*max_seg_width, max_lb_size / scl_data->taps.v_taps);
@ -250,13 +251,13 @@ enum vpe_status vpe11_set_num_segments(struct vpe_priv *vpe_priv, struct stream_
num_segs += (vpe_priv->vpe_num_instance - (num_segs % vpe_priv->vpe_num_instance));
}
stream_ctx->segment_ctx = vpe_alloc_segment_ctx(vpe_priv, num_segs);
if (!stream_ctx->segment_ctx)
return VPE_STATUS_NO_MEMORY;
res = vpe_alloc_segment_ctx(vpe_priv, stream_ctx, num_segs);
stream_ctx->num_segments = num_segs;
if (res == VPE_STATUS_OK) {
stream_ctx->num_segments = num_segs;
}
return VPE_STATUS_OK;
return res;
}
bool vpe11_validate_cached_param(struct vpe_priv *vpe_priv, const struct vpe_build_param *param)

3
src/amd/vpelib/src/core/inc/resource.h
View file

@ -133,7 +133,8 @@ enum vpe_status vpe_construct_resource(
void vpe_destroy_resource(struct vpe_priv *vpe_priv, struct resource *res);
/** alloc segment ctx*/
struct segment_ctx *vpe_alloc_segment_ctx(struct vpe_priv *vpe_priv, uint16_t num_segments);
enum vpe_status vpe_alloc_segment_ctx(
struct vpe_priv *vpe_priv, struct stream_ctx *stream_ctx, uint16_t num_segments);
/** stream ctx */
struct stream_ctx *vpe_alloc_stream_ctx(struct vpe_priv *vpe_priv, uint32_t num_streams);

26
src/amd/vpelib/src/core/resource.c
View file

@ -136,16 +136,28 @@ void vpe_destroy_resource(struct vpe_priv *vpe_priv, struct resource *res)
}
}
struct segment_ctx *vpe_alloc_segment_ctx(struct vpe_priv *vpe_priv, uint16_t num_segments)
enum vpe_status vpe_alloc_segment_ctx(
struct vpe_priv *vpe_priv, struct stream_ctx *stream_ctx, uint16_t num_segments)
{
struct segment_ctx *segment_ctx_base;
// If segment_ctx is already allocated, check if re-allocation needed
if (stream_ctx->segment_ctx) {
if (num_segments != stream_ctx->num_segments) {
// Need to re-allocate segment_ctx. Free it first
vpe_free(stream_ctx->segment_ctx);
stream_ctx->segment_ctx = NULL;
} else {
// No need for re-allocation. Return
return VPE_STATUS_OK;
}
}
segment_ctx_base = (struct segment_ctx *)vpe_zalloc(sizeof(struct segment_ctx) * num_segments);
stream_ctx->segment_ctx =
(struct segment_ctx *)vpe_zalloc(sizeof(struct segment_ctx) * num_segments);
if (!stream_ctx->segment_ctx) {
return VPE_STATUS_NO_MEMORY;
}
if (!segment_ctx_base)
return NULL;
return segment_ctx_base;
return VPE_STATUS_OK;
}
static enum vpe_status create_input_config_vector(struct stream_ctx *stream_ctx)