Re: [Mesa-dev] [PATCH 07/16] radeonsi: add reference count to si_compute
On Mon, Aug 21, 2017 at 8:36 AM, Gert Wollnywrote: > Am Sonntag, den 20.08.2017, 15:48 +0200 schrieb Marek Olšák: >> The patch seems OK to me. > > Hmm, > > you're right in that I was wrong about "program", I still need learn to > think in patches. > > However, when I look at si_delete_compute_state where > si_compute_reference is called: > > >>+ si_compute_reference(, NULL); >>+} > ... >>+static inline void >>+si_compute_reference(struct si_compute **dst, struct si_compute *src) >>+{ >>+ if (pipe_reference(&(*dst)->reference, >reference)) >>+ si_destroy_compute(*dst); > > then src will be a NULL pointer and de-referencing it is most likely > already undefined behavior. > > Anyway, in the end the address of a member is taken (i.e. > src+offsetof(src, reference), and passed as a pointer, and this pointer > is later de-referenced in > > pipe_reference_described > > if it is not 0. > > For now this runs through because "reference" is the first element in > the si_compute struct, and hence its offset is 0 and >reference > will also be 0, but the moment someone decides to reorder si_compute > this may no longer the case and the code would break. > > At the least it should be commented that the element reference within > the structure si_compute must always be at offset 0. You're right. All pipe_reference declarations should always be first. And that's why they are always first. It's an unwritten rule in Gallium. Marek ___ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
Re: [Mesa-dev] [PATCH 07/16] radeonsi: add reference count to si_compute
Am Sonntag, den 20.08.2017, 15:48 +0200 schrieb Marek Olšák: > The patch seems OK to me. Hmm, you're right in that I was wrong about "program", I still need learn to think in patches. However, when I look at si_delete_compute_state where si_compute_reference is called: >+ si_compute_reference(, NULL); >+} ... >+static inline void >+si_compute_reference(struct si_compute **dst, struct si_compute *src) >+{ >+ if (pipe_reference(&(*dst)->reference, >reference)) >+ si_destroy_compute(*dst); then src will be a NULL pointer and de-referencing it is most likely already undefined behavior. Anyway, in the end the address of a member is taken (i.e. src+offsetof(src, reference), and passed as a pointer, and this pointer is later de-referenced in pipe_reference_described if it is not 0. For now this runs through because "reference" is the first element in the si_compute struct, and hence its offset is 0 and >reference will also be 0, but the moment someone decides to reorder si_compute this may no longer the case and the code would break. At the least it should be commented that the element reference within the structure si_compute must always be at offset 0. Best, Gert ___ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
Re: [Mesa-dev] [PATCH 07/16] radeonsi: add reference count to si_compute
The patch seems OK to me. Marek On Wed, Aug 16, 2017 at 1:50 PM, Gert Wollnywrote: > Hello Nicolai, > > I've spotted a little problem in you code (see below). > > Best, > Gert > > > Am Mittwoch, den 16.08.2017, 13:05 +0200 schrieb Nicolai Hähnle: >> From: Nicolai Hähnle >> >> To allow keep-alive for deferred logging. >> --- >> src/gallium/drivers/radeonsi/si_compute.c | 24 ++--- >> --- >> src/gallium/drivers/radeonsi/si_compute.h | 14 ++ >> 2 files changed, 28 insertions(+), 10 deletions(-) >> >> diff --git a/src/gallium/drivers/radeonsi/si_compute.c >> b/src/gallium/drivers/radeonsi/si_compute.c >> index 5efdd3919d2..8c016ba65ab 100644 >> --- a/src/gallium/drivers/radeonsi/si_compute.c >> +++ b/src/gallium/drivers/radeonsi/si_compute.c >> @@ -151,6 +151,7 @@ static void *si_create_compute_state( >> struct si_screen *sscreen = (struct si_screen *)ctx->screen; >> struct si_compute *program = CALLOC_STRUCT(si_compute); >> >> + pipe_reference_init(>reference, 1); >> program->screen = (struct si_screen *)ctx->screen; >> program->ir_type = cso->ir_type; >> program->local_size = cso->req_local_mem; >> @@ -855,20 +856,24 @@ static void si_launch_grid( >> sctx->b.flags |= SI_CONTEXT_CS_PARTIAL_FLUSH; >> } >> >> +void si_destroy_compute(struct si_compute *program) >> +{ >> + if (program->ir_type == PIPE_SHADER_IR_TGSI) { >> + util_queue_drop_job(>screen- >> >shader_compiler_queue, >> + >ready); >> + util_queue_fence_destroy(>ready); >> + } >> + >> + si_shader_destroy(>shader); >> + FREE(program); >> +} >> >> static void si_delete_compute_state(struct pipe_context *ctx, void* >> state){ >> struct si_compute *program = (struct si_compute *)state; >> struct si_context *sctx = (struct si_context*)ctx; >> >> - if (!state) { >> + if (!state) >> return; >> - } >> - >> - if (program->ir_type == PIPE_SHADER_IR_TGSI) { >> - util_queue_drop_job(>screen- >> >shader_compiler_queue, >> - >ready); >> - util_queue_fence_destroy(>ready); >> - } >> >> if (program == sctx->cs_shader_state.program) >> sctx->cs_shader_state.program = NULL; >> @@ -876,8 +881,7 @@ static void si_delete_compute_state(struct >> pipe_context *ctx, void* state){ >> if (program == sctx->cs_shader_state.emitted_program) >> sctx->cs_shader_state.emitted_program = NULL; >> >> - si_shader_destroy(>shader); >> - FREE(program); >> + si_compute_reference(, NULL); >> } > Here you call "si_compute_reference" with parameter src=NULL and de- > reference it in that function. > Likewise you call FREE on "program" and also dereference it there. > (see below). > >> >> static void si_set_compute_resources(struct pipe_context * ctx_, >> diff --git a/src/gallium/drivers/radeonsi/si_compute.h >> b/src/gallium/drivers/radeonsi/si_compute.h >> index 268817b23a6..c19b701fc71 100644 >> --- a/src/gallium/drivers/radeonsi/si_compute.h >> +++ b/src/gallium/drivers/radeonsi/si_compute.h >> @@ -24,11 +24,14 @@ >> #ifndef SI_COMPUTE_H >> #define SI_COMPUTE_H >> >> +#include "util/u_inlines.h" >> + >> #include "si_shader.h" >> >> #define MAX_GLOBAL_BUFFERS 22 >> >> struct si_compute { >> + struct pipe_reference reference; >> struct si_screen *screen; >> struct tgsi_token *tokens; >> struct util_queue_fence ready; >> @@ -53,4 +56,15 @@ struct si_compute { >> unsigned uses_bindless_images:1; >> }; >> >> +void si_destroy_compute(struct si_compute *program); >> + >> +static inline void >> +si_compute_reference(struct si_compute **dst, struct si_compute >> *src) >> +{ >> + if (pipe_reference(&(*dst)->reference, >reference)) > + si_destroy_compute(*dst); > > When called from si_delete_compute_state *dst has been FREEed and src > is NULL. > >> + >> + *dst = src; >> +} >> + >> #endif /* SI_COMPUTE_H */ > ___ > mesa-dev mailing list > mesa-dev@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/mesa-dev ___ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
[Mesa-dev] [PATCH 07/16] radeonsi: add reference count to si_compute
From: Nicolai HähnleTo allow keep-alive for deferred logging. --- src/gallium/drivers/radeonsi/si_compute.c | 24 ++-- src/gallium/drivers/radeonsi/si_compute.h | 14 ++ 2 files changed, 28 insertions(+), 10 deletions(-) diff --git a/src/gallium/drivers/radeonsi/si_compute.c b/src/gallium/drivers/radeonsi/si_compute.c index 5efdd3919d2..8c016ba65ab 100644 --- a/src/gallium/drivers/radeonsi/si_compute.c +++ b/src/gallium/drivers/radeonsi/si_compute.c @@ -151,6 +151,7 @@ static void *si_create_compute_state( struct si_screen *sscreen = (struct si_screen *)ctx->screen; struct si_compute *program = CALLOC_STRUCT(si_compute); + pipe_reference_init(>reference, 1); program->screen = (struct si_screen *)ctx->screen; program->ir_type = cso->ir_type; program->local_size = cso->req_local_mem; @@ -855,20 +856,24 @@ static void si_launch_grid( sctx->b.flags |= SI_CONTEXT_CS_PARTIAL_FLUSH; } +void si_destroy_compute(struct si_compute *program) +{ + if (program->ir_type == PIPE_SHADER_IR_TGSI) { + util_queue_drop_job(>screen->shader_compiler_queue, + >ready); + util_queue_fence_destroy(>ready); + } + + si_shader_destroy(>shader); + FREE(program); +} static void si_delete_compute_state(struct pipe_context *ctx, void* state){ struct si_compute *program = (struct si_compute *)state; struct si_context *sctx = (struct si_context*)ctx; - if (!state) { + if (!state) return; - } - - if (program->ir_type == PIPE_SHADER_IR_TGSI) { - util_queue_drop_job(>screen->shader_compiler_queue, - >ready); - util_queue_fence_destroy(>ready); - } if (program == sctx->cs_shader_state.program) sctx->cs_shader_state.program = NULL; @@ -876,8 +881,7 @@ static void si_delete_compute_state(struct pipe_context *ctx, void* state){ if (program == sctx->cs_shader_state.emitted_program) sctx->cs_shader_state.emitted_program = NULL; - si_shader_destroy(>shader); - FREE(program); + si_compute_reference(, NULL); } static void si_set_compute_resources(struct pipe_context * ctx_, diff --git a/src/gallium/drivers/radeonsi/si_compute.h b/src/gallium/drivers/radeonsi/si_compute.h index 268817b23a6..c19b701fc71 100644 --- a/src/gallium/drivers/radeonsi/si_compute.h +++ b/src/gallium/drivers/radeonsi/si_compute.h @@ -24,11 +24,14 @@ #ifndef SI_COMPUTE_H #define SI_COMPUTE_H +#include "util/u_inlines.h" + #include "si_shader.h" #define MAX_GLOBAL_BUFFERS 22 struct si_compute { + struct pipe_reference reference; struct si_screen *screen; struct tgsi_token *tokens; struct util_queue_fence ready; @@ -53,4 +56,15 @@ struct si_compute { unsigned uses_bindless_images:1; }; +void si_destroy_compute(struct si_compute *program); + +static inline void +si_compute_reference(struct si_compute **dst, struct si_compute *src) +{ + if (pipe_reference(&(*dst)->reference, >reference)) + si_destroy_compute(*dst); + + *dst = src; +} + #endif /* SI_COMPUTE_H */ -- 2.11.0 ___ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev