Files
linux-nvgpu/drivers/gpu/nvgpu/common/fifo/tsg_subctx.c
Sagar Kamble ef99d9f010 gpu: nvgpu: implement scg, pbdma and cilp rules
Only certain combination of channels of GFX/Compute object classes can
be assigned to particular pbdma and/or VEID. CILP can be enabled only
in certain configs. Implement checks for the configurations verified
during alloc_obj_ctx and/or setting preemption mode.

Bug 3677982

Change-Id: Ie7026cbb240819c1727b3736ed34044d7138d3cd
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2719995
Reviewed-by: Ankur Kishore <ankkishore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
2022-09-08 21:00:30 -07:00

414 lines
10 KiB
C

/*
* Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
#include <nvgpu/tsg.h>
#include <nvgpu/tsg_subctx.h>
#include <nvgpu/channel.h>
#include <nvgpu/enabled.h>
#include <nvgpu/kmem.h>
#include <nvgpu/list.h>
#include <nvgpu/gk20a.h>
#include <nvgpu/log.h>
#include <nvgpu/gr/subctx.h>
#include <nvgpu/gr/ctx_mappings.h>
#include "tsg_subctx_priv.h"
static inline struct nvgpu_tsg_subctx *
nvgpu_tsg_subctx_from_tsg_entry(struct nvgpu_list_node *node)
{
return (struct nvgpu_tsg_subctx *)
((uintptr_t)node - offsetof(struct nvgpu_tsg_subctx, tsg_entry));
};
static struct nvgpu_tsg_subctx *nvgpu_tsg_subctx_from_id(struct nvgpu_tsg *tsg,
u32 subctx_id)
{
struct nvgpu_tsg_subctx *subctx = NULL;
nvgpu_list_for_each_entry(subctx, &tsg->subctx_list,
nvgpu_tsg_subctx, tsg_entry) {
if (subctx->subctx_id == subctx_id) {
return subctx;
}
}
return NULL;
}
int nvgpu_tsg_subctx_bind_channel(struct nvgpu_tsg *tsg,
struct nvgpu_channel *ch)
{
struct nvgpu_tsg_subctx *subctx = NULL;
struct gk20a *g = tsg->g;
nvgpu_log(g, gpu_dbg_gr, " ");
if (!nvgpu_is_enabled(g, NVGPU_SUPPORT_TSG_SUBCONTEXTS)) {
return 0;
}
subctx = nvgpu_tsg_subctx_from_id(tsg, ch->subctx_id);
if (subctx != NULL) {
if (subctx->vm != ch->vm) {
nvgpu_err(g, "subctx vm mismatch");
return -EINVAL;
}
goto add_ch_subctx;
}
nvgpu_log(g, gpu_dbg_gr, "Allocating subctx %u", ch->subctx_id);
subctx = nvgpu_kzalloc(g, sizeof(struct nvgpu_tsg_subctx));
if (subctx == NULL) {
nvgpu_err(g, "Failed to allocate subctx");
return -ENOMEM;
}
subctx->subctx_id = ch->subctx_id;
subctx->tsg = tsg;
subctx->vm = ch->vm;
nvgpu_init_list_node(&subctx->ch_list);
nvgpu_init_list_node(&subctx->tsg_entry);
nvgpu_list_add_tail(&subctx->tsg_entry, &tsg->subctx_list);
add_ch_subctx:
ch->subctx = subctx;
nvgpu_list_add_tail(&ch->subctx_entry, &subctx->ch_list);
nvgpu_log(g, gpu_dbg_gr, "done");
return 0;
}
void nvgpu_tsg_subctx_unbind_channel(struct nvgpu_tsg *tsg,
struct nvgpu_channel *ch)
{
struct nvgpu_tsg_subctx *subctx;
struct gk20a *g = tsg->g;
nvgpu_log(g, gpu_dbg_gr, " ");
if (!nvgpu_is_enabled(g, NVGPU_SUPPORT_TSG_SUBCONTEXTS)) {
return;
}
subctx = ch->subctx;
nvgpu_assert(subctx != NULL);
nvgpu_list_del(&ch->subctx_entry);
if (nvgpu_list_empty(&subctx->ch_list)) {
if (g->ops.tsg.remove_subctx_channel_hw != NULL) {
g->ops.tsg.remove_subctx_channel_hw(ch);
}
if (g->ops.gr.setup.free_subctx != NULL) {
g->ops.gr.setup.free_subctx(ch);
subctx->gr_subctx = NULL;
}
nvgpu_list_del(&subctx->tsg_entry);
nvgpu_kfree(tsg->g, subctx);
}
ch->subctx = NULL;
nvgpu_log(g, gpu_dbg_gr, "done");
}
int nvgpu_tsg_subctx_alloc_gr_subctx(struct gk20a *g, struct nvgpu_channel *ch)
{
struct nvgpu_tsg_subctx *subctx;
nvgpu_log(g, gpu_dbg_gr, " ");
if (!nvgpu_is_enabled(g, NVGPU_SUPPORT_TSG_SUBCONTEXTS)) {
return 0;
}
subctx = ch->subctx;
if (subctx == NULL) {
nvgpu_err(g, "channel not bound to TSG subctx");
return -EINVAL;
}
if (subctx->gr_subctx == NULL) {
subctx->gr_subctx = nvgpu_gr_subctx_alloc(g);
if (subctx->gr_subctx == NULL) {
nvgpu_err(g, "gr_subctx alloc failed");
return -ENOMEM;
}
}
nvgpu_log(g, gpu_dbg_gr, "done");
return 0;
}
int nvgpu_tsg_subctx_setup_subctx_header(struct gk20a *g,
struct nvgpu_channel *ch)
{
struct nvgpu_tsg_subctx *subctx;
int err;
nvgpu_log(g, gpu_dbg_gr, " ");
if (!nvgpu_is_enabled(g, NVGPU_SUPPORT_TSG_SUBCONTEXTS)) {
return 0;
}
subctx = ch->subctx;
if ((subctx == NULL) || (subctx->gr_subctx == NULL)) {
nvgpu_err(g, "channel not bound to TSG/GR subctx");
return -EINVAL;
}
err = nvgpu_gr_subctx_setup_header(g, subctx->gr_subctx, subctx->vm);
if (err != 0) {
nvgpu_err(g, "gr_subctx header setup failed %d", err);
return err;
}
nvgpu_log(g, gpu_dbg_gr, "done");
return 0;
}
struct nvgpu_gr_subctx *nvgpu_tsg_subctx_get_gr_subctx(
struct nvgpu_tsg_subctx *subctx)
{
return subctx->gr_subctx;
}
u32 nvgpu_tsg_subctx_get_id(struct nvgpu_tsg_subctx *subctx)
{
return subctx->subctx_id;
}
void nvgpu_tsg_subctx_set_replayable(struct nvgpu_tsg_subctx *subctx,
bool replayable)
{
subctx->replayable = replayable;
}
bool nvgpu_tsg_subctx_get_replayable(struct nvgpu_tsg_subctx *subctx)
{
return subctx->replayable;
}
struct vm_gk20a *nvgpu_tsg_subctx_get_vm(struct nvgpu_tsg_subctx *subctx)
{
return subctx->vm;
}
struct nvgpu_gr_ctx_mappings *nvgpu_tsg_subctx_alloc_or_get_mappings(
struct gk20a *g,
struct nvgpu_tsg *tsg,
struct nvgpu_channel *ch)
{
struct nvgpu_gr_ctx_mappings *mappings = NULL;
struct nvgpu_gr_subctx *gr_subctx = NULL;
struct vm_gk20a *vm = ch->vm;
nvgpu_log(g, gpu_dbg_gr, " ");
nvgpu_assert(ch->subctx != NULL);
nvgpu_assert(ch->subctx->vm == vm);
mappings = nvgpu_gr_ctx_mappings_get_subctx_mappings(g, tsg, vm);
if (mappings != NULL) {
goto add_gr_subctx;
}
mappings = nvgpu_gr_ctx_mappings_create_subctx_mappings(g, tsg, vm);
if (mappings == NULL) {
nvgpu_err(g, "failed to allocate gr_ctx mappings");
return NULL;
}
add_gr_subctx:
gr_subctx = nvgpu_tsg_subctx_get_gr_subctx(ch->subctx);
nvgpu_assert(gr_subctx != NULL);
nvgpu_gr_ctx_mappings_add_gr_subctx(mappings, gr_subctx);
nvgpu_log(g, gpu_dbg_gr, "done");
return mappings;
}
#ifdef CONFIG_NVGPU_GFXP
static struct nvgpu_gr_ctx_mappings *nvgpu_tsg_subctx_get_veid0_mappings(
struct gk20a *g,
struct nvgpu_tsg *tsg)
{
struct nvgpu_gr_ctx_mappings *mappings = NULL;
struct nvgpu_tsg_subctx *subctx = NULL;
subctx = nvgpu_tsg_subctx_from_id(tsg, CHANNEL_INFO_VEID0);
if (subctx == NULL) {
nvgpu_log(g, gpu_dbg_gr, "VEID0 subctx not available");
return NULL;
}
mappings = nvgpu_gr_subctx_get_mappings(subctx->gr_subctx);
if (mappings == NULL) {
nvgpu_log(g, gpu_dbg_gr, "VEID0 mappings not available");
return NULL;
}
return mappings;
}
void nvgpu_tsg_subctxs_set_preemption_buffer_va(
struct nvgpu_tsg_subctx *tsg_subctx)
{
struct nvgpu_gr_ctx_mappings *veid0_mappings;
struct nvgpu_tsg_subctx *subctx = NULL;
struct nvgpu_tsg *tsg = tsg_subctx->tsg;
struct gk20a *g = tsg->g;
nvgpu_log(g, gpu_dbg_gr, " ");
nvgpu_rwsem_down_read(&tsg->ch_list_lock);
veid0_mappings = nvgpu_tsg_subctx_get_veid0_mappings(g, tsg);
if (veid0_mappings == NULL) {
nvgpu_rwsem_up_read(&tsg->ch_list_lock);
return;
}
nvgpu_list_for_each_entry(subctx, &tsg->subctx_list,
nvgpu_tsg_subctx, tsg_entry) {
if (subctx->gr_subctx != NULL) {
nvgpu_gr_subctx_set_preemption_buffer_va(g,
subctx->gr_subctx, veid0_mappings);
}
}
nvgpu_rwsem_up_read(&tsg->ch_list_lock);
nvgpu_log(g, gpu_dbg_gr, "done");
}
void nvgpu_tsg_subctxs_clear_preemption_buffer_va(
struct nvgpu_tsg_subctx *tsg_subctx)
{
struct nvgpu_tsg_subctx *subctx = NULL;
struct nvgpu_tsg *tsg = tsg_subctx->tsg;
struct gk20a *g = tsg->g;
nvgpu_log(g, gpu_dbg_gr, " ");
nvgpu_list_for_each_entry(subctx, &tsg->subctx_list,
nvgpu_tsg_subctx, tsg_entry) {
if (subctx->gr_subctx != NULL) {
nvgpu_gr_subctx_clear_preemption_buffer_va(g,
subctx->gr_subctx);
}
}
nvgpu_log(g, gpu_dbg_gr, "done");
}
#endif /* CONFIG_NVGPU_GFXP */
#ifdef CONFIG_NVGPU_DEBUGGER
void nvgpu_tsg_subctxs_set_pm_buffer_va(struct nvgpu_tsg *tsg,
bool set_pm_ctx_gpu_va)
{
struct nvgpu_tsg_subctx *subctx = NULL;
struct gk20a *g = tsg->g;
nvgpu_log(g, gpu_dbg_gr, " ");
nvgpu_rwsem_down_read(&tsg->ch_list_lock);
nvgpu_list_for_each_entry(subctx, &tsg->subctx_list,
nvgpu_tsg_subctx, tsg_entry) {
if (subctx->gr_subctx != NULL) {
nvgpu_gr_subctx_set_hwpm_ptr(g, subctx->gr_subctx,
set_pm_ctx_gpu_va);
}
}
nvgpu_rwsem_up_read(&tsg->ch_list_lock);
nvgpu_log(g, gpu_dbg_gr, "done");
}
#endif /* CONFIG_NVGPU_DEBUGGER */
static inline struct nvgpu_channel *
nvgpu_channel_from_subctx_entry(struct nvgpu_list_node *node)
{
return (struct nvgpu_channel *)
((uintptr_t)node - offsetof(struct nvgpu_channel, subctx_entry));
};
bool nvgpu_tsg_channel_type_active(struct nvgpu_tsg *tsg,
bool match_subctx, u32 subctx_id,
bool match_pbdma, u32 pbdma_id,
bool (*is_valid_class)(u32 class_num))
{
struct nvgpu_tsg_subctx *subctx = NULL;
bool channel_active = false;
struct gk20a *g = tsg->g;
struct nvgpu_channel *ch;
nvgpu_log(g, gpu_dbg_gr, " ");
if (is_valid_class == NULL) {
return false;
}
nvgpu_rwsem_down_write(&tsg->ch_list_lock);
nvgpu_list_for_each_entry(subctx, &tsg->subctx_list,
nvgpu_tsg_subctx, tsg_entry) {
if ((match_subctx && (subctx->subctx_id == subctx_id)) ||
(!match_subctx && (subctx->subctx_id != subctx_id))) {
nvgpu_list_for_each_entry(ch, &subctx->ch_list,
nvgpu_channel, subctx_entry) {
if ((*is_valid_class)(ch->obj_class)) {
if ((match_pbdma && (ch->runqueue_sel == pbdma_id)) ||
(!match_pbdma && (ch->runqueue_sel != pbdma_id))) {
channel_active = true;
break;
}
}
}
if (channel_active == true) {
break;
}
}
}
nvgpu_rwsem_up_write(&tsg->ch_list_lock);
nvgpu_log(g, gpu_dbg_gr, "done");
return channel_active;
}