mirror of
git://nv-tegra.nvidia.com/linux-nvgpu.git
synced 2025-12-24 10:34:43 +03:00
gpu: nvgpu: ltc: move chip specific files to hal
Move ltc chip speciifc files to hal from common JIRA NVGPU-2044 Change-Id: If3f5e77fce1dfa94336e1be616833cef5b91839b Signed-off-by: Seshendra Gadagottu <sgadagottu@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/2070186 Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
committed by
mobile promotions
parent
6be73f561a
commit
a15d1fa72c
352
drivers/gpu/nvgpu/hal/ltc/ltc_gm20b.c
Normal file
352
drivers/gpu/nvgpu/hal/ltc/ltc_gm20b.c
Normal file
@@ -0,0 +1,352 @@
|
||||
/*
|
||||
* GM20B L2
|
||||
*
|
||||
* Copyright (c) 2014-2019 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 <trace/events/gk20a.h>
|
||||
|
||||
#include <nvgpu/timers.h>
|
||||
#include <nvgpu/enabled.h>
|
||||
#include <nvgpu/bug.h>
|
||||
#include <nvgpu/ltc.h>
|
||||
#include <nvgpu/io.h>
|
||||
#include <nvgpu/utils.h>
|
||||
#include <nvgpu/gk20a.h>
|
||||
|
||||
#include <nvgpu/hw/gm20b/hw_ltc_gm20b.h>
|
||||
#include <nvgpu/hw/gm20b/hw_top_gm20b.h>
|
||||
#include <nvgpu/hw/gm20b/hw_pri_ringmaster_gm20b.h>
|
||||
|
||||
#include "ltc_gm20b.h"
|
||||
|
||||
void gm20b_ltc_init_fs_state(struct gk20a *g)
|
||||
{
|
||||
u32 reg;
|
||||
|
||||
nvgpu_log_info(g, "initialize gm20b l2");
|
||||
|
||||
g->max_ltc_count = gk20a_readl(g, top_num_ltcs_r());
|
||||
g->ltc_count = gk20a_readl(g, pri_ringmaster_enum_ltc_r());
|
||||
nvgpu_log_info(g, "%d ltcs out of %d", g->ltc_count, g->max_ltc_count);
|
||||
|
||||
reg = gk20a_readl(g, ltc_ltcs_ltss_cbc_param_r());
|
||||
g->slices_per_ltc = ltc_ltcs_ltss_cbc_param_slices_per_ltc_v(reg);;
|
||||
g->cacheline_size =
|
||||
U32(512) << ltc_ltcs_ltss_cbc_param_cache_line_size_v(reg);
|
||||
|
||||
gk20a_writel(g, ltc_ltcs_ltss_cbc_num_active_ltcs_r(),
|
||||
g->ltc_count);
|
||||
gk20a_writel(g, ltc_ltcs_misc_ltc_num_active_ltcs_r(),
|
||||
g->ltc_count);
|
||||
|
||||
gk20a_writel(g, ltc_ltcs_ltss_dstg_cfg0_r(),
|
||||
gk20a_readl(g, ltc_ltc0_lts0_dstg_cfg0_r()) |
|
||||
ltc_ltcs_ltss_dstg_cfg0_vdc_4to2_disable_m());
|
||||
|
||||
/* Disable LTC interrupts */
|
||||
reg = gk20a_readl(g, ltc_ltcs_ltss_intr_r());
|
||||
reg &= ~ltc_ltcs_ltss_intr_en_evicted_cb_m();
|
||||
reg &= ~ltc_ltcs_ltss_intr_en_illegal_compstat_access_m();
|
||||
reg &= ~ltc_ltcs_ltss_intr_en_illegal_compstat_m();
|
||||
gk20a_writel(g, ltc_ltcs_ltss_intr_r(), reg);
|
||||
}
|
||||
|
||||
void gm20b_ltc_lts_isr(struct gk20a *g, unsigned int ltc, unsigned int slice)
|
||||
{
|
||||
u32 ltc_intr;
|
||||
u32 ltc_stride = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE);
|
||||
u32 lts_stride = nvgpu_get_litter_value(g, GPU_LIT_LTS_STRIDE);
|
||||
|
||||
ltc_intr = gk20a_readl(g, ltc_ltc0_lts0_intr_r() +
|
||||
ltc_stride * ltc +
|
||||
lts_stride * slice);
|
||||
nvgpu_err(g, "ltc%d, slice %d: %08x",
|
||||
ltc, slice, ltc_intr);
|
||||
gk20a_writel(g, ltc_ltc0_lts0_intr_r() +
|
||||
ltc_stride * ltc +
|
||||
lts_stride * slice,
|
||||
ltc_intr);
|
||||
}
|
||||
|
||||
void gm20b_ltc_isr(struct gk20a *g, unsigned int ltc)
|
||||
{
|
||||
unsigned int slice;
|
||||
|
||||
for (slice = 0U; slice < g->slices_per_ltc; slice++) {
|
||||
gm20b_ltc_lts_isr(g, ltc, slice);
|
||||
}
|
||||
}
|
||||
|
||||
/*
|
||||
* Performs a full flush of the L2 cache.
|
||||
*/
|
||||
void gm20b_flush_ltc(struct gk20a *g)
|
||||
{
|
||||
struct nvgpu_timeout timeout;
|
||||
unsigned int ltc;
|
||||
u32 ltc_stride = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE);
|
||||
bool is_clean_pending_set = false;
|
||||
bool is_invalidate_pending_set = false;
|
||||
|
||||
/* Clean... */
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_tstg_cmgmt1_r(),
|
||||
ltc_ltcs_ltss_tstg_cmgmt1_clean_pending_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt1_max_cycles_between_cleans_3_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt1_clean_wait_for_fb_to_pull_true_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt1_clean_evict_last_class_true_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt1_clean_evict_normal_class_true_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt1_clean_evict_first_class_true_f());
|
||||
|
||||
/* Wait on each LTC individually. */
|
||||
for (ltc = 0; ltc < g->ltc_count; ltc++) {
|
||||
u32 op_pending;
|
||||
|
||||
/*
|
||||
* Use 5ms - this should be sufficient time to flush the cache.
|
||||
* On tegra, rough EMC BW available for old tegra chips (newer
|
||||
* chips are strictly faster) can be estimated as follows:
|
||||
*
|
||||
* Lowest reasonable EMC clock speed will be around 102MHz on
|
||||
* t124 for display enabled boards and generally fixed to max
|
||||
* for non-display boards (since they are generally plugged in).
|
||||
*
|
||||
* Thus, the available BW is 64b * 2 * 102MHz = 1.3GB/s. Of that
|
||||
* BW the GPU will likely get about half (display and overhead/
|
||||
* utilization inefficiency eating the rest) so 650MB/s at
|
||||
* worst. Assuming at most 1MB of GPU L2 cache (less for most
|
||||
* chips) worst case is we take 1MB/650MB/s = 1.5ms.
|
||||
*
|
||||
* So 5ms timeout here should be more than sufficient.
|
||||
*/
|
||||
nvgpu_timeout_init(g, &timeout, 5, NVGPU_TIMER_CPU_TIMER);
|
||||
|
||||
do {
|
||||
int cmgmt1 = ltc_ltc0_ltss_tstg_cmgmt1_r() +
|
||||
ltc * ltc_stride;
|
||||
op_pending = gk20a_readl(g, cmgmt1);
|
||||
is_clean_pending_set = (op_pending &
|
||||
ltc_ltc0_ltss_tstg_cmgmt1_clean_pending_f()) != 0U;
|
||||
} while (is_clean_pending_set &&
|
||||
(nvgpu_timeout_expired_msg(&timeout,
|
||||
"L2 flush timeout!") == 0));
|
||||
}
|
||||
|
||||
/* And invalidate. */
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_tstg_cmgmt0_r(),
|
||||
ltc_ltcs_ltss_tstg_cmgmt0_invalidate_pending_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt0_max_cycles_between_invalidates_3_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt0_invalidate_evict_last_class_true_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt0_invalidate_evict_normal_class_true_f() |
|
||||
ltc_ltcs_ltss_tstg_cmgmt0_invalidate_evict_first_class_true_f());
|
||||
|
||||
/* Wait on each LTC individually. */
|
||||
for (ltc = 0; ltc < g->ltc_count; ltc++) {
|
||||
u32 op_pending;
|
||||
|
||||
/* Again, 5ms. */
|
||||
nvgpu_timeout_init(g, &timeout, 5, NVGPU_TIMER_CPU_TIMER);
|
||||
|
||||
do {
|
||||
int cmgmt0 = ltc_ltc0_ltss_tstg_cmgmt0_r() +
|
||||
ltc * ltc_stride;
|
||||
op_pending = gk20a_readl(g, cmgmt0);
|
||||
is_invalidate_pending_set = (op_pending &
|
||||
ltc_ltc0_ltss_tstg_cmgmt0_invalidate_pending_f()) != 0U;
|
||||
} while (is_invalidate_pending_set &&
|
||||
(nvgpu_timeout_expired_msg(&timeout,
|
||||
"L2 flush timeout!") == 0));
|
||||
}
|
||||
}
|
||||
|
||||
int gm20b_determine_L2_size_bytes(struct gk20a *g)
|
||||
{
|
||||
u32 lts_per_ltc;
|
||||
u32 ways;
|
||||
u32 sets;
|
||||
u32 bytes_per_line;
|
||||
u32 active_ltcs;
|
||||
u32 cache_size;
|
||||
|
||||
u32 tmp;
|
||||
u32 active_sets_value;
|
||||
|
||||
tmp = gk20a_readl(g, ltc_ltc0_lts0_tstg_cfg1_r());
|
||||
ways = hweight32(ltc_ltc0_lts0_tstg_cfg1_active_ways_v(tmp));
|
||||
|
||||
active_sets_value = ltc_ltc0_lts0_tstg_cfg1_active_sets_v(tmp);
|
||||
if (active_sets_value == ltc_ltc0_lts0_tstg_cfg1_active_sets_all_v()) {
|
||||
sets = 64U;
|
||||
} else if (active_sets_value ==
|
||||
ltc_ltc0_lts0_tstg_cfg1_active_sets_half_v()) {
|
||||
sets = 32U;
|
||||
} else if (active_sets_value ==
|
||||
ltc_ltc0_lts0_tstg_cfg1_active_sets_quarter_v()) {
|
||||
sets = 16U;
|
||||
} else {
|
||||
nvgpu_err(g, "Unknown constant %u for active sets",
|
||||
(unsigned)active_sets_value);
|
||||
sets = 0U;
|
||||
}
|
||||
|
||||
active_ltcs = g->gr.num_fbps;
|
||||
|
||||
/* chip-specific values */
|
||||
lts_per_ltc = 2U;
|
||||
bytes_per_line = 128U;
|
||||
cache_size = active_ltcs * lts_per_ltc * ways * sets * bytes_per_line;
|
||||
|
||||
return cache_size;
|
||||
}
|
||||
|
||||
/*
|
||||
* Sets the ZBC color for the passed index.
|
||||
*/
|
||||
void gm20b_ltc_set_zbc_color_entry(struct gk20a *g,
|
||||
u32 *color_l2,
|
||||
u32 index)
|
||||
{
|
||||
u32 i;
|
||||
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_dstg_zbc_index_r(),
|
||||
ltc_ltcs_ltss_dstg_zbc_index_address_f(index));
|
||||
|
||||
for (i = 0;
|
||||
i < ltc_ltcs_ltss_dstg_zbc_color_clear_value__size_1_v(); i++) {
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltcs_ltss_dstg_zbc_color_clear_value_r(i),
|
||||
color_l2[i]);
|
||||
}
|
||||
}
|
||||
|
||||
/*
|
||||
* Sets the ZBC depth for the passed index.
|
||||
*/
|
||||
void gm20b_ltc_set_zbc_depth_entry(struct gk20a *g,
|
||||
u32 depth_val,
|
||||
u32 index)
|
||||
{
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_dstg_zbc_index_r(),
|
||||
ltc_ltcs_ltss_dstg_zbc_index_address_f(index));
|
||||
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltcs_ltss_dstg_zbc_depth_clear_value_r(),
|
||||
depth_val);
|
||||
}
|
||||
|
||||
void gm20b_ltc_set_enabled(struct gk20a *g, bool enabled)
|
||||
{
|
||||
u32 reg_f = ltc_ltcs_ltss_tstg_set_mgmt_2_l2_bypass_mode_enabled_f();
|
||||
u32 reg = gk20a_readl(g, ltc_ltcs_ltss_tstg_set_mgmt_2_r());
|
||||
|
||||
if (enabled) {
|
||||
/* bypass disabled (normal caching ops) */
|
||||
reg &= ~reg_f;
|
||||
} else {
|
||||
/* bypass enabled (no caching) */
|
||||
reg |= reg_f;
|
||||
}
|
||||
|
||||
gk20a_writel(g, ltc_ltcs_ltss_tstg_set_mgmt_2_r(), reg);
|
||||
}
|
||||
|
||||
/*
|
||||
* LTC pri addressing
|
||||
*/
|
||||
bool gm20b_ltc_pri_is_ltc_addr(struct gk20a *g, u32 addr)
|
||||
{
|
||||
return ((addr >= ltc_pltcg_base_v()) && (addr < ltc_pltcg_extent_v()));
|
||||
}
|
||||
|
||||
bool gm20b_ltc_is_ltcs_ltss_addr(struct gk20a *g, u32 addr)
|
||||
{
|
||||
u32 ltc_shared_base = ltc_ltcs_ltss_v();
|
||||
u32 lts_stride = nvgpu_get_litter_value(g, GPU_LIT_LTS_STRIDE);
|
||||
|
||||
return (addr >= ltc_shared_base) &&
|
||||
(addr < (ltc_shared_base + lts_stride));
|
||||
}
|
||||
|
||||
bool gm20b_ltc_is_ltcn_ltss_addr(struct gk20a *g, u32 addr)
|
||||
{
|
||||
u32 lts_shared_base = ltc_ltc0_ltss_v();
|
||||
u32 lts_stride = nvgpu_get_litter_value(g, GPU_LIT_LTS_STRIDE);
|
||||
u32 addr_mask = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE) - 1U;
|
||||
u32 base_offset = lts_shared_base & addr_mask;
|
||||
u32 end_offset = base_offset + lts_stride;
|
||||
|
||||
return (!gm20b_ltc_is_ltcs_ltss_addr(g, addr)) &&
|
||||
((addr & addr_mask) >= base_offset) &&
|
||||
((addr & addr_mask) < end_offset);
|
||||
}
|
||||
|
||||
static void gm20b_ltc_update_ltc_lts_addr(struct gk20a *g, u32 addr, u32 ltc_num,
|
||||
u32 *priv_addr_table,
|
||||
u32 *priv_addr_table_index)
|
||||
{
|
||||
u32 num_ltc_slices = g->ops.top.get_max_lts_per_ltc(g);
|
||||
u32 index = *priv_addr_table_index;
|
||||
u32 lts_num;
|
||||
u32 ltc_stride = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE);
|
||||
u32 lts_stride = nvgpu_get_litter_value(g, GPU_LIT_LTS_STRIDE);
|
||||
|
||||
for (lts_num = 0; lts_num < num_ltc_slices; lts_num++) {
|
||||
priv_addr_table[index++] = ltc_ltc0_lts0_v() +
|
||||
ltc_num * ltc_stride +
|
||||
lts_num * lts_stride +
|
||||
(addr & (lts_stride - 1U));
|
||||
}
|
||||
|
||||
*priv_addr_table_index = index;
|
||||
}
|
||||
|
||||
void gm20b_ltc_split_lts_broadcast_addr(struct gk20a *g, u32 addr,
|
||||
u32 *priv_addr_table,
|
||||
u32 *priv_addr_table_index)
|
||||
{
|
||||
u32 num_ltc = g->ltc_count;
|
||||
u32 i, start, ltc_num = 0;
|
||||
u32 pltcg_base = ltc_pltcg_base_v();
|
||||
u32 ltc_stride = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE);
|
||||
|
||||
for (i = 0; i < num_ltc; i++) {
|
||||
start = pltcg_base + i * ltc_stride;
|
||||
if ((addr >= start) && (addr < (start + ltc_stride))) {
|
||||
ltc_num = i;
|
||||
break;
|
||||
}
|
||||
}
|
||||
gm20b_ltc_update_ltc_lts_addr(g, addr, ltc_num, priv_addr_table,
|
||||
priv_addr_table_index);
|
||||
}
|
||||
|
||||
void gm20b_ltc_split_ltc_broadcast_addr(struct gk20a *g, u32 addr,
|
||||
u32 *priv_addr_table,
|
||||
u32 *priv_addr_table_index)
|
||||
{
|
||||
u32 num_ltc = g->ltc_count;
|
||||
u32 ltc_num;
|
||||
|
||||
for (ltc_num = 0; ltc_num < num_ltc; ltc_num++) {
|
||||
gm20b_ltc_update_ltc_lts_addr(g, addr, ltc_num,
|
||||
priv_addr_table, priv_addr_table_index);
|
||||
}
|
||||
}
|
||||
55
drivers/gpu/nvgpu/hal/ltc/ltc_gm20b.h
Normal file
55
drivers/gpu/nvgpu/hal/ltc/ltc_gm20b.h
Normal file
@@ -0,0 +1,55 @@
|
||||
/*
|
||||
* GM20B L2
|
||||
*
|
||||
* Copyright (c) 2014-2019, 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.
|
||||
*/
|
||||
|
||||
#ifndef NVGPU_LTC_GM20B
|
||||
#define NVGPU_LTC_GM20B
|
||||
|
||||
#include <nvgpu/types.h>
|
||||
|
||||
struct gk20a;
|
||||
struct gr_gk20a;
|
||||
|
||||
int gm20b_determine_L2_size_bytes(struct gk20a *g);
|
||||
void gm20b_ltc_set_zbc_color_entry(struct gk20a *g,
|
||||
u32 *color_l2,
|
||||
u32 index);
|
||||
void gm20b_ltc_set_zbc_depth_entry(struct gk20a *g,
|
||||
u32 depth_val,
|
||||
u32 index);
|
||||
void gm20b_ltc_set_enabled(struct gk20a *g, bool enabled);
|
||||
void gm20b_ltc_init_fs_state(struct gk20a *g);
|
||||
void gm20b_ltc_isr(struct gk20a *g, unsigned int ltc);
|
||||
void gm20b_ltc_lts_isr(struct gk20a *g, unsigned int ltc, unsigned int slice);
|
||||
void gm20b_flush_ltc(struct gk20a *g);
|
||||
bool gm20b_ltc_pri_is_ltc_addr(struct gk20a *g, u32 addr);
|
||||
bool gm20b_ltc_is_ltcs_ltss_addr(struct gk20a *g, u32 addr);
|
||||
bool gm20b_ltc_is_ltcn_ltss_addr(struct gk20a *g, u32 addr);
|
||||
void gm20b_ltc_split_lts_broadcast_addr(struct gk20a *g, u32 addr,
|
||||
u32 *priv_addr_table,
|
||||
u32 *priv_addr_table_index);
|
||||
void gm20b_ltc_split_ltc_broadcast_addr(struct gk20a *g, u32 addr,
|
||||
u32 *priv_addr_table,
|
||||
u32 *priv_addr_table_index);
|
||||
|
||||
#endif
|
||||
157
drivers/gpu/nvgpu/hal/ltc/ltc_gp10b.c
Normal file
157
drivers/gpu/nvgpu/hal/ltc/ltc_gp10b.c
Normal file
@@ -0,0 +1,157 @@
|
||||
/*
|
||||
* GP10B L2
|
||||
*
|
||||
* Copyright (c) 2014-2019, 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 <trace/events/gk20a.h>
|
||||
|
||||
#include <nvgpu/ltc.h>
|
||||
#include <nvgpu/log.h>
|
||||
#include <nvgpu/enabled.h>
|
||||
#include <nvgpu/io.h>
|
||||
#include <nvgpu/timers.h>
|
||||
#include <nvgpu/gk20a.h>
|
||||
|
||||
#include <nvgpu/hw/gp10b/hw_ltc_gp10b.h>
|
||||
|
||||
#include "ltc_gm20b.h"
|
||||
#include "ltc_gp10b.h"
|
||||
|
||||
int gp10b_determine_L2_size_bytes(struct gk20a *g)
|
||||
{
|
||||
u32 tmp;
|
||||
int ret;
|
||||
|
||||
nvgpu_log_fn(g, " ");
|
||||
|
||||
tmp = gk20a_readl(g, ltc_ltc0_lts0_tstg_info_1_r());
|
||||
|
||||
ret = g->ltc_count *
|
||||
ltc_ltc0_lts0_tstg_info_1_slice_size_in_kb_v(tmp) * 1024U *
|
||||
ltc_ltc0_lts0_tstg_info_1_slices_per_l2_v(tmp);
|
||||
|
||||
nvgpu_log(g, gpu_dbg_info, "L2 size: %d\n", ret);
|
||||
|
||||
nvgpu_log_fn(g, "done");
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
void gp10b_ltc_lts_isr(struct gk20a *g, unsigned int ltc, unsigned int slice)
|
||||
{
|
||||
u32 offset;
|
||||
u32 ltc_intr;
|
||||
u32 ltc_stride = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE);
|
||||
u32 lts_stride = nvgpu_get_litter_value(g, GPU_LIT_LTS_STRIDE);
|
||||
|
||||
offset = ltc_stride * ltc + lts_stride * slice;
|
||||
ltc_intr = gk20a_readl(g, ltc_ltc0_lts0_intr_r() + offset);
|
||||
|
||||
/* Detect and handle ECC errors */
|
||||
if ((ltc_intr &
|
||||
ltc_ltcs_ltss_intr_ecc_sec_error_pending_f()) != 0U) {
|
||||
u32 ecc_stats_reg_val;
|
||||
|
||||
nvgpu_err(g,
|
||||
"Single bit error detected in GPU L2!");
|
||||
|
||||
ecc_stats_reg_val =
|
||||
gk20a_readl(g,
|
||||
ltc_ltc0_lts0_dstg_ecc_report_r() + offset);
|
||||
g->ecc.ltc.ecc_sec_count[ltc][slice].counter +=
|
||||
ltc_ltc0_lts0_dstg_ecc_report_sec_count_v(ecc_stats_reg_val);
|
||||
ecc_stats_reg_val &=
|
||||
~(ltc_ltc0_lts0_dstg_ecc_report_sec_count_m());
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltc0_lts0_dstg_ecc_report_r() + offset,
|
||||
ecc_stats_reg_val);
|
||||
if (g->ops.mm.l2_flush(g, true) != 0) {
|
||||
nvgpu_err(g, "l2_flush failed");
|
||||
}
|
||||
}
|
||||
if ((ltc_intr &
|
||||
ltc_ltcs_ltss_intr_ecc_ded_error_pending_f()) != 0U) {
|
||||
u32 ecc_stats_reg_val;
|
||||
|
||||
nvgpu_err(g,
|
||||
"Double bit error detected in GPU L2!");
|
||||
|
||||
ecc_stats_reg_val =
|
||||
gk20a_readl(g,
|
||||
ltc_ltc0_lts0_dstg_ecc_report_r() + offset);
|
||||
g->ecc.ltc.ecc_ded_count[ltc][slice].counter +=
|
||||
ltc_ltc0_lts0_dstg_ecc_report_ded_count_v(ecc_stats_reg_val);
|
||||
ecc_stats_reg_val &=
|
||||
~(ltc_ltc0_lts0_dstg_ecc_report_ded_count_m());
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltc0_lts0_dstg_ecc_report_r() + offset,
|
||||
ecc_stats_reg_val);
|
||||
}
|
||||
|
||||
nvgpu_err(g, "ltc%d, slice %d: %08x",
|
||||
ltc, slice, ltc_intr);
|
||||
nvgpu_writel_check(g, ltc_ltc0_lts0_intr_r() +
|
||||
ltc_stride * ltc + lts_stride * slice,
|
||||
ltc_intr);
|
||||
}
|
||||
|
||||
void gp10b_ltc_isr(struct gk20a *g, unsigned int ltc)
|
||||
{
|
||||
unsigned int slice;
|
||||
|
||||
for (slice = 0U; slice < g->slices_per_ltc; slice++) {
|
||||
gp10b_ltc_lts_isr(g, ltc, slice);
|
||||
}
|
||||
}
|
||||
|
||||
void gp10b_ltc_init_fs_state(struct gk20a *g)
|
||||
{
|
||||
u32 ltc_intr;
|
||||
|
||||
gm20b_ltc_init_fs_state(g);
|
||||
|
||||
gk20a_writel(g, ltc_ltca_g_axi_pctrl_r(),
|
||||
ltc_ltca_g_axi_pctrl_user_sid_f(g->ltc_streamid));
|
||||
|
||||
/* Enable ECC interrupts */
|
||||
ltc_intr = gk20a_readl(g, ltc_ltcs_ltss_intr_r());
|
||||
ltc_intr |= ltc_ltcs_ltss_intr_en_ecc_sec_error_enabled_f() |
|
||||
ltc_ltcs_ltss_intr_en_ecc_ded_error_enabled_f();
|
||||
gk20a_writel(g, ltc_ltcs_ltss_intr_r(),
|
||||
ltc_intr);
|
||||
}
|
||||
|
||||
void gp10b_ltc_set_enabled(struct gk20a *g, bool enabled)
|
||||
{
|
||||
u32 reg_f = ltc_ltcs_ltss_tstg_set_mgmt_2_l2_bypass_mode_enabled_f();
|
||||
u32 reg = gk20a_readl(g, ltc_ltcs_ltss_tstg_set_mgmt_2_r());
|
||||
|
||||
if (enabled) {
|
||||
/* bypass disabled (normal caching ops) */
|
||||
reg &= ~reg_f;
|
||||
} else {
|
||||
/* bypass enabled (no caching) */
|
||||
reg |= reg_f;
|
||||
}
|
||||
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_tstg_set_mgmt_2_r(), reg);
|
||||
}
|
||||
33
drivers/gpu/nvgpu/hal/ltc/ltc_gp10b.h
Normal file
33
drivers/gpu/nvgpu/hal/ltc/ltc_gp10b.h
Normal file
@@ -0,0 +1,33 @@
|
||||
/*
|
||||
* Copyright (c) 2014-2019, 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.
|
||||
*/
|
||||
|
||||
#ifndef LTC_GP10B_H
|
||||
#define LTC_GP10B_H
|
||||
struct gk20a;
|
||||
struct gpu_ops;
|
||||
|
||||
int gp10b_determine_L2_size_bytes(struct gk20a *g);
|
||||
void gp10b_ltc_init_fs_state(struct gk20a *g);
|
||||
void gp10b_ltc_set_enabled(struct gk20a *g, bool enabled);
|
||||
void gp10b_ltc_isr(struct gk20a *g, unsigned int ltc);
|
||||
void gp10b_ltc_lts_isr(struct gk20a *g, unsigned int ltc, unsigned int slice);
|
||||
#endif
|
||||
244
drivers/gpu/nvgpu/hal/ltc/ltc_gv11b.c
Normal file
244
drivers/gpu/nvgpu/hal/ltc/ltc_gv11b.c
Normal file
@@ -0,0 +1,244 @@
|
||||
/*
|
||||
* GV11B LTC
|
||||
*
|
||||
* Copyright (c) 2016-2019, 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/io.h>
|
||||
#include <nvgpu/gk20a.h>
|
||||
#include <nvgpu/nvgpu_err.h>
|
||||
|
||||
#include "ltc_gp10b.h"
|
||||
#include "ltc_gv11b.h"
|
||||
|
||||
#include <nvgpu/hw/gv11b/hw_ltc_gv11b.h>
|
||||
#include <nvgpu/hw/gv11b/hw_mc_gv11b.h>
|
||||
#include <nvgpu/hw/gv11b/hw_top_gv11b.h>
|
||||
#include <nvgpu/hw/gv11b/hw_mc_gv11b.h>
|
||||
|
||||
#include <nvgpu/utils.h>
|
||||
|
||||
/*
|
||||
* Sets the ZBC stencil for the passed index.
|
||||
*/
|
||||
void gv11b_ltc_set_zbc_stencil_entry(struct gk20a *g,
|
||||
u32 stencil_depth,
|
||||
u32 index)
|
||||
{
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_dstg_zbc_index_r(),
|
||||
ltc_ltcs_ltss_dstg_zbc_index_address_f(index));
|
||||
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltcs_ltss_dstg_zbc_stencil_clear_value_r(),
|
||||
stencil_depth);
|
||||
}
|
||||
|
||||
void gv11b_ltc_init_fs_state(struct gk20a *g)
|
||||
{
|
||||
u32 ltc_intr;
|
||||
u32 reg;
|
||||
|
||||
nvgpu_log_info(g, "initialize gv11b l2");
|
||||
|
||||
g->max_ltc_count = gk20a_readl(g, top_num_ltcs_r());
|
||||
g->ltc_count = g->ops.priv_ring.enum_ltc(g);
|
||||
nvgpu_log_info(g, "%u ltcs out of %u", g->ltc_count, g->max_ltc_count);
|
||||
|
||||
reg = gk20a_readl(g, ltc_ltcs_ltss_cbc_param_r());
|
||||
g->slices_per_ltc = ltc_ltcs_ltss_cbc_param_slices_per_ltc_v(reg);;
|
||||
g->cacheline_size =
|
||||
U32(512) << ltc_ltcs_ltss_cbc_param_cache_line_size_v(reg);
|
||||
|
||||
/* Disable LTC interrupts */
|
||||
reg = gk20a_readl(g, ltc_ltcs_ltss_intr_r());
|
||||
reg &= ~ltc_ltcs_ltss_intr_en_evicted_cb_m();
|
||||
reg &= ~ltc_ltcs_ltss_intr_en_illegal_compstat_access_m();
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_intr_r(), reg);
|
||||
|
||||
if (g->ops.ltc.intr_en_illegal_compstat != NULL) {
|
||||
g->ops.ltc.intr_en_illegal_compstat(g,
|
||||
g->ltc_intr_en_illegal_compstat);
|
||||
}
|
||||
|
||||
/* Enable ECC interrupts */
|
||||
ltc_intr = gk20a_readl(g, ltc_ltcs_ltss_intr_r());
|
||||
ltc_intr |= ltc_ltcs_ltss_intr_en_ecc_sec_error_enabled_f() |
|
||||
ltc_ltcs_ltss_intr_en_ecc_ded_error_enabled_f();
|
||||
nvgpu_writel_check(g, ltc_ltcs_ltss_intr_r(),
|
||||
ltc_intr);
|
||||
}
|
||||
|
||||
void gv11b_ltc_intr_en_illegal_compstat(struct gk20a *g, bool enable)
|
||||
{
|
||||
u32 val;
|
||||
|
||||
/* disble/enble illegal_compstat interrupt */
|
||||
val = gk20a_readl(g, ltc_ltcs_ltss_intr_r());
|
||||
if (enable) {
|
||||
val = set_field(val,
|
||||
ltc_ltcs_ltss_intr_en_illegal_compstat_m(),
|
||||
ltc_ltcs_ltss_intr_en_illegal_compstat_enabled_f());
|
||||
} else {
|
||||
val = set_field(val,
|
||||
ltc_ltcs_ltss_intr_en_illegal_compstat_m(),
|
||||
ltc_ltcs_ltss_intr_en_illegal_compstat_disabled_f());
|
||||
}
|
||||
gk20a_writel(g, ltc_ltcs_ltss_intr_r(), val);
|
||||
}
|
||||
|
||||
void gv11b_ltc_lts_isr(struct gk20a *g, unsigned int ltc, unsigned int slice)
|
||||
{
|
||||
u32 offset;
|
||||
u32 ltc_intr3;
|
||||
u32 ecc_status, ecc_addr, dstg_ecc_addr, corrected_cnt, uncorrected_cnt;
|
||||
u32 corrected_delta, uncorrected_delta;
|
||||
u32 corrected_overflow, uncorrected_overflow;
|
||||
u32 ltc_stride = nvgpu_get_litter_value(g, GPU_LIT_LTC_STRIDE);
|
||||
u32 lts_stride = nvgpu_get_litter_value(g, GPU_LIT_LTS_STRIDE);
|
||||
|
||||
offset = ltc_stride * ltc + lts_stride * slice;
|
||||
ltc_intr3 = gk20a_readl(g, ltc_ltc0_lts0_intr3_r() +
|
||||
offset);
|
||||
|
||||
/* Detect and handle ECC PARITY errors */
|
||||
if ((ltc_intr3 &
|
||||
(ltc_ltcs_ltss_intr3_ecc_uncorrected_m() |
|
||||
ltc_ltcs_ltss_intr3_ecc_corrected_m())) != 0U) {
|
||||
|
||||
ecc_status = gk20a_readl(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_status_r() +
|
||||
offset);
|
||||
ecc_addr = gk20a_readl(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_address_r() +
|
||||
offset);
|
||||
dstg_ecc_addr = gk20a_readl(g,
|
||||
ltc_ltc0_lts0_dstg_ecc_address_r() +
|
||||
offset);
|
||||
corrected_cnt = gk20a_readl(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_corrected_err_count_r() + offset);
|
||||
uncorrected_cnt = gk20a_readl(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_uncorrected_err_count_r() + offset);
|
||||
|
||||
corrected_delta =
|
||||
ltc_ltc0_lts0_l2_cache_ecc_corrected_err_count_total_v(corrected_cnt);
|
||||
uncorrected_delta =
|
||||
ltc_ltc0_lts0_l2_cache_ecc_uncorrected_err_count_total_v(uncorrected_cnt);
|
||||
corrected_overflow = ecc_status &
|
||||
ltc_ltc0_lts0_l2_cache_ecc_status_corrected_err_total_counter_overflow_m();
|
||||
|
||||
uncorrected_overflow = ecc_status &
|
||||
ltc_ltc0_lts0_l2_cache_ecc_status_uncorrected_err_total_counter_overflow_m();
|
||||
|
||||
/* clear the interrupt */
|
||||
if ((corrected_delta > 0U) || (corrected_overflow != 0U)) {
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_corrected_err_count_r() + offset, 0);
|
||||
}
|
||||
if ((uncorrected_delta > 0U) || (uncorrected_overflow !=0U)) {
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_uncorrected_err_count_r() + offset, 0);
|
||||
}
|
||||
|
||||
nvgpu_writel_check(g,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_status_r() + offset,
|
||||
ltc_ltc0_lts0_l2_cache_ecc_status_reset_task_f());
|
||||
|
||||
/* update counters per slice */
|
||||
if (corrected_overflow != 0U) {
|
||||
corrected_delta += BIT32(ltc_ltc0_lts0_l2_cache_ecc_corrected_err_count_total_s());
|
||||
}
|
||||
if (uncorrected_overflow != 0U) {
|
||||
uncorrected_delta += BIT32(ltc_ltc0_lts0_l2_cache_ecc_uncorrected_err_count_total_s());
|
||||
}
|
||||
|
||||
g->ecc.ltc.ecc_sec_count[ltc][slice].counter += corrected_delta;
|
||||
g->ecc.ltc.ecc_ded_count[ltc][slice].counter += uncorrected_delta;
|
||||
nvgpu_log(g, gpu_dbg_intr,
|
||||
"ltc:%d lts: %d cache ecc interrupt intr: 0x%x", ltc, slice, ltc_intr3);
|
||||
|
||||
if ((ecc_status & ltc_ltc0_lts0_l2_cache_ecc_status_corrected_err_rstg_m()) != 0U) {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_RSTG_ECC_CORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_sec_count[ltc][slice].counter);
|
||||
nvgpu_log(g, gpu_dbg_intr, "rstg ecc error corrected");
|
||||
}
|
||||
if ((ecc_status & ltc_ltc0_lts0_l2_cache_ecc_status_uncorrected_err_rstg_m()) != 0U) {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_RSTG_ECC_UNCORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_ded_count[ltc][slice].counter);
|
||||
nvgpu_log(g, gpu_dbg_intr, "rstg ecc error uncorrected");
|
||||
}
|
||||
if ((ecc_status & ltc_ltc0_lts0_l2_cache_ecc_status_corrected_err_tstg_m()) != 0U) {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_TSTG_ECC_CORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_sec_count[ltc][slice].counter);
|
||||
nvgpu_log(g, gpu_dbg_intr, "tstg ecc error corrected");
|
||||
}
|
||||
if ((ecc_status & ltc_ltc0_lts0_l2_cache_ecc_status_uncorrected_err_tstg_m()) != 0U) {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_TSTG_ECC_UNCORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_ded_count[ltc][slice].counter);
|
||||
nvgpu_log(g, gpu_dbg_intr, "tstg ecc error uncorrected");
|
||||
}
|
||||
if ((ecc_status & ltc_ltc0_lts0_l2_cache_ecc_status_corrected_err_dstg_m()) != 0U) {
|
||||
if ((dstg_ecc_addr & ltc_ltc0_lts0_dstg_ecc_address_info_ram_m()) == 0U) {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_DSTG_ECC_CORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_sec_count[ltc][slice].counter);
|
||||
} else {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_DSTG_BE_ECC_CORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_sec_count[ltc][slice].counter);
|
||||
}
|
||||
nvgpu_log(g, gpu_dbg_intr, "dstg ecc error corrected");
|
||||
}
|
||||
if ((ecc_status & ltc_ltc0_lts0_l2_cache_ecc_status_uncorrected_err_dstg_m()) != 0U) {
|
||||
if ((dstg_ecc_addr & ltc_ltc0_lts0_dstg_ecc_address_info_ram_m()) == 0U) {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_DSTG_ECC_UNCORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_ded_count[ltc][slice].counter);
|
||||
} else {
|
||||
nvgpu_ltc_report_ecc_error(g, ltc, slice,
|
||||
GPU_LTC_CACHE_DSTG_BE_ECC_UNCORRECTED, ecc_addr,
|
||||
g->ecc.ltc.ecc_ded_count[ltc][slice].counter);
|
||||
}
|
||||
nvgpu_log(g, gpu_dbg_intr, "dstg ecc error uncorrected");
|
||||
}
|
||||
|
||||
if ((corrected_overflow != 0U) || (uncorrected_overflow != 0U)) {
|
||||
nvgpu_info(g, "ecc counter overflow!");
|
||||
}
|
||||
|
||||
nvgpu_log(g, gpu_dbg_intr,
|
||||
"ecc error address: 0x%x", ecc_addr);
|
||||
}
|
||||
|
||||
gp10b_ltc_lts_isr(g, ltc, slice);
|
||||
}
|
||||
|
||||
void gv11b_ltc_isr(struct gk20a *g, unsigned int ltc)
|
||||
{
|
||||
unsigned int slice;
|
||||
|
||||
for (slice = 0U; slice < g->slices_per_ltc; slice++) {
|
||||
gv11b_ltc_lts_isr(g, ltc, slice);
|
||||
}
|
||||
}
|
||||
35
drivers/gpu/nvgpu/hal/ltc/ltc_gv11b.h
Normal file
35
drivers/gpu/nvgpu/hal/ltc/ltc_gv11b.h
Normal file
@@ -0,0 +1,35 @@
|
||||
/*
|
||||
* Copyright (c) 2016-2019, 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.
|
||||
*/
|
||||
|
||||
#ifndef LTC_GV11B_H
|
||||
#define LTC_GV11B_H
|
||||
struct gk20a;
|
||||
|
||||
void gv11b_ltc_set_zbc_stencil_entry(struct gk20a *g,
|
||||
u32 stencil_depth,
|
||||
u32 index);
|
||||
void gv11b_ltc_init_fs_state(struct gk20a *g);
|
||||
void gv11b_ltc_intr_en_illegal_compstat(struct gk20a *g, bool enable);
|
||||
void gv11b_ltc_isr(struct gk20a *g, unsigned int ltc);
|
||||
void gv11b_ltc_lts_isr(struct gk20a *g, unsigned int ltc, unsigned int slice);
|
||||
|
||||
#endif
|
||||
59
drivers/gpu/nvgpu/hal/ltc/ltc_tu104.c
Normal file
59
drivers/gpu/nvgpu/hal/ltc/ltc_tu104.c
Normal file
@@ -0,0 +1,59 @@
|
||||
/*
|
||||
* Copyright (c) 2018-2019, 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/types.h>
|
||||
#include <nvgpu/ltc.h>
|
||||
#include <nvgpu/comptags.h>
|
||||
#include <nvgpu/io.h>
|
||||
#include <nvgpu/timers.h>
|
||||
#include <nvgpu/gk20a.h>
|
||||
|
||||
#include <trace/events/gk20a.h>
|
||||
|
||||
#include "ltc_tu104.h"
|
||||
|
||||
#include "ltc_gv11b.h"
|
||||
|
||||
#include <nvgpu/hw/tu104/hw_ltc_tu104.h>
|
||||
|
||||
void ltc_tu104_init_fs_state(struct gk20a *g)
|
||||
{
|
||||
u32 reg;
|
||||
|
||||
gv11b_ltc_init_fs_state(g);
|
||||
|
||||
reg = nvgpu_readl(g, ltc_ltcs_ltss_cbc_param2_r());
|
||||
g->slices_per_ltc =
|
||||
ltc_ltcs_ltss_cbc_param2_slices_per_ltc_v(reg);
|
||||
g->cacheline_size =
|
||||
U32(512) << ltc_ltcs_ltss_cbc_param2_cache_line_size_v(reg);
|
||||
|
||||
/* disable PLC compression */
|
||||
reg = nvgpu_readl(g, ltc_ltcs_ltss_tstg_set_mgmt_1_r());
|
||||
reg = set_field(reg,
|
||||
ltc_ltcs_ltss_tstg_set_mgmt_1_plc_recompress_plc_m(),
|
||||
ltc_ltcs_ltss_tstg_set_mgmt_1_plc_recompress_plc_disabled_f());
|
||||
reg = set_field(reg,
|
||||
ltc_ltcs_ltss_tstg_set_mgmt_1_plc_recompress_rmw_m(),
|
||||
ltc_ltcs_ltss_tstg_set_mgmt_1_plc_recompress_rmw_disabled_f());
|
||||
nvgpu_writel(g, ltc_ltcs_ltss_tstg_set_mgmt_1_r(), reg);
|
||||
}
|
||||
32
drivers/gpu/nvgpu/hal/ltc/ltc_tu104.h
Normal file
32
drivers/gpu/nvgpu/hal/ltc/ltc_tu104.h
Normal file
@@ -0,0 +1,32 @@
|
||||
/*
|
||||
* Copyright (c) 2018-2019, 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.
|
||||
*/
|
||||
|
||||
#ifndef LTC_TU104_H
|
||||
#define LTC_TU104_H
|
||||
|
||||
#include <nvgpu/types.h>
|
||||
|
||||
struct gk20a;
|
||||
|
||||
void ltc_tu104_init_fs_state(struct gk20a *g);
|
||||
|
||||
#endif
|
||||
Reference in New Issue
Block a user