nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-22 17:36:20 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: enable CONFIG_GK20A_DEVFREQ for k5.9

Browse Source 
Enable CONFIG_GK20A_DEVFREQ and apply corresponding
changes for kernel-5.9
- Remove frequency clipping of devfreq min/max frequency
  constraint due to devfreq already takes care of it.
- Register available GPU frequencies to OPP framework
  due to devfreq access available device frequencies
  through OPP frameworks during device frequency
  transition.

Bug 200639056

Change-Id: I72e4a7825ae9ca814791dc283138d17a5cfbe8e2
Signed-off-by: Aaron Tian <atian@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2400107
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
Aaron Tian
2020-08-18 17:48:35 +08:00
committed by Alex Waterman
parent c8add76c8d
commit 5d740f98b0
2 changed files with 58 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
drivers/gpu/nvgpu/Makefile.linux.configs
View File

@@ -148,7 +148,6 @@ ifneq ($(findstring stable,$(NV_BUILD_KERNEL_OPTIONS)),)
CONFIG_GK20A_DEVFREQ := n CONFIG_GK20A_DEVFREQ := n
CONFIG_GK20A_PM_QOS := n CONFIG_GK20A_PM_QOS := n
else ifneq ($(findstring 5.9,$(NV_BUILD_KERNEL_OPTIONS)),) else ifneq ($(findstring 5.9,$(NV_BUILD_KERNEL_OPTIONS)),)
CONFIG_GK20A_DEVFREQ := n
CONFIG_GK20A_PM_QOS := n CONFIG_GK20A_PM_QOS := n
endif endif

59
drivers/gpu/nvgpu/os/linux/scale.c
View File

@@ -19,6 +19,7 @@
#include <linux/devfreq.h> #include <linux/devfreq.h>
#include <linux/export.h> #include <linux/export.h>
#include <linux/pm_qos.h> #include <linux/pm_qos.h>
#include <linux/version.h>
#include <governor.h> #include <governor.h>
@@ -149,9 +150,11 @@ static int gk20a_scale_target(struct device *dev, unsigned long *freq,
{ {
struct gk20a_platform *platform = dev_get_drvdata(dev); struct gk20a_platform *platform = dev_get_drvdata(dev);
struct gk20a *g = platform->g; struct gk20a *g = platform->g;
struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
struct gk20a_scale_profile *profile = g->scale_profile; struct gk20a_scale_profile *profile = g->scale_profile;
#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 8, 0)
struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
struct devfreq *devfreq = l->devfreq; struct devfreq *devfreq = l->devfreq;
#endif
unsigned long local_freq = *freq; unsigned long local_freq = *freq;
unsigned long rounded_rate; unsigned long rounded_rate;
unsigned long min_freq = 0, max_freq = 0; unsigned long min_freq = 0, max_freq = 0;
@@ -173,8 +176,18 @@ static int gk20a_scale_target(struct device *dev, unsigned long *freq,
* In case we have conflict (min_freq > max_freq) after above * In case we have conflict (min_freq > max_freq) after above
* steps, we ensure that max_freq wins over min_freq * steps, we ensure that max_freq wins over min_freq
*/ */
#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 8, 0)
min_freq = max_t(u32, devfreq->min_freq, profile->qos_min_freq); min_freq = max_t(u32, devfreq->min_freq, profile->qos_min_freq);
max_freq = min_t(u32, devfreq->max_freq, profile->qos_max_freq); max_freq = min_t(u32, devfreq->max_freq, profile->qos_max_freq);
#else
/*
* devfreq takes care of min/max freq clipping in update_devfreq() then
* invoked devfreq->profile->target(), thus we only need to do freq
* clipping based on pm_qos constraint
*/
min_freq = profile->qos_min_freq;
max_freq = profile->qos_max_freq;
#endif
if (min_freq > max_freq) if (min_freq > max_freq)
min_freq = max_freq; min_freq = max_freq;
@@ -320,6 +333,46 @@ static int get_cur_freq(struct device *dev, unsigned long *freq)
return 0; return 0;
} }
#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 8, 0)
static int register_gpu_opp(struct device *dev)
{
return 0;
}
static void unregister_gpu_opp(struct device *dev)
{
}
#else
static void unregister_gpu_opp(struct device *dev)
{
dev_pm_opp_remove_all_dynamic(dev);
}
static int register_gpu_opp(struct device *dev)
{
struct gk20a_platform *platform = dev_get_drvdata(dev);
struct gk20a *g = platform->g;
struct gk20a_scale_profile *profile = g->scale_profile;
unsigned long *freq_table = profile->devfreq_profile.freq_table;
int max_states = profile->devfreq_profile.max_state;
int i;
int err = 0;
for (i = 0; i < max_states; ++i) {
err = dev_pm_opp_add(dev, freq_table[i], 0);
if (err) {
nvgpu_err(g,
"Failed to add OPP %lu: %d\n",
freq_table[i],
err);
unregister_gpu_opp(dev);
break;
}
}
return err;
}
#endif
/* /*
* gk20a_scale_init(dev) * gk20a_scale_init(dev)
@@ -362,6 +415,8 @@ void gk20a_scale_init(struct device *dev)
struct devfreq *devfreq; struct devfreq *devfreq;
int error = 0; int error = 0;
register_gpu_opp(dev);
profile->devfreq_profile.initial_freq = profile->devfreq_profile.initial_freq =
profile->devfreq_profile.freq_table[0]; profile->devfreq_profile.freq_table[0];
profile->devfreq_profile.target = gk20a_scale_target; profile->devfreq_profile.target = gk20a_scale_target;
@@ -432,6 +487,8 @@ void gk20a_scale_exit(struct device *dev)
err = devfreq_remove_device(l->devfreq); err = devfreq_remove_device(l->devfreq);
l->devfreq = NULL; l->devfreq = NULL;
unregister_gpu_opp(dev);
} }
nvgpu_kfree(g, g->scale_profile); nvgpu_kfree(g, g->scale_profile);