Commit 6759a0a7 authored by Marek Olšák's avatar Marek Olšák Committed by Alex Deucher

drm/radeon/kms: implement timestamp userspace query (v2)

Returns a snapshot of the GPU clock counter.  Needed
for certain OpenGL extensions.

v2: agd5f
- address Jerome's comments
- add function documentation
Signed-off-by: default avatarMarek Olšák <maraeo@gmail.com>
Reviewed-by: default avatarJerome Glisse <jglisse@redhat.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent b51ad12a
......@@ -3789,3 +3789,23 @@ static void r600_pcie_gen2_enable(struct radeon_device *rdev)
WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);
}
}
/**
* r600_get_gpu_clock - return GPU clock counter snapshot
*
* @rdev: radeon_device pointer
*
* Fetches a GPU clock counter snapshot (R6xx-cayman).
* Returns the 64 bit clock counter snapshot.
*/
uint64_t r600_get_gpu_clock(struct radeon_device *rdev)
{
uint64_t clock;
mutex_lock(&rdev->gpu_clock_mutex);
WREG32(RLC_CAPTURE_GPU_CLOCK_COUNT, 1);
clock = (uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_LSB) |
((uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
mutex_unlock(&rdev->gpu_clock_mutex);
return clock;
}
......@@ -602,6 +602,9 @@
#define RLC_HB_WPTR 0x3f1c
#define RLC_HB_WPTR_LSB_ADDR 0x3f14
#define RLC_HB_WPTR_MSB_ADDR 0x3f18
#define RLC_GPU_CLOCK_COUNT_LSB 0x3f38
#define RLC_GPU_CLOCK_COUNT_MSB 0x3f3c
#define RLC_CAPTURE_GPU_CLOCK_COUNT 0x3f40
#define RLC_MC_CNTL 0x3f44
#define RLC_UCODE_CNTL 0x3f48
#define RLC_UCODE_ADDR 0x3f2c
......
......@@ -1534,6 +1534,7 @@ struct radeon_device {
unsigned debugfs_count;
/* virtual memory */
struct radeon_vm_manager vm_manager;
struct mutex gpu_clock_mutex;
};
int radeon_device_init(struct radeon_device *rdev,
......
......@@ -368,6 +368,7 @@ void r600_kms_blit_copy(struct radeon_device *rdev,
unsigned num_gpu_pages,
struct radeon_sa_bo *vb);
int r600_mc_wait_for_idle(struct radeon_device *rdev);
uint64_t r600_get_gpu_clock(struct radeon_device *rdev);
/*
* rv770,rv730,rv710,rv740
......@@ -468,5 +469,6 @@ int si_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm, int id);
void si_vm_unbind(struct radeon_device *rdev, struct radeon_vm *vm);
void si_vm_tlb_flush(struct radeon_device *rdev, struct radeon_vm *vm);
int si_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib);
uint64_t si_get_gpu_clock(struct radeon_device *rdev);
#endif
......@@ -1009,6 +1009,7 @@ int radeon_device_init(struct radeon_device *rdev,
atomic_set(&rdev->ih.lock, 0);
mutex_init(&rdev->gem.mutex);
mutex_init(&rdev->pm.mutex);
mutex_init(&rdev->gpu_clock_mutex);
init_rwsem(&rdev->pm.mclk_lock);
init_rwsem(&rdev->exclusive_lock);
init_waitqueue_head(&rdev->irq.vblank_queue);
......
......@@ -61,9 +61,10 @@
* 2.17.0 - add STRMOUT_BASE_UPDATE for r7xx
* 2.18.0 - r600-eg: allow "invalid" DB formats
* 2.19.0 - r600-eg: MSAA textures
* 2.20.0 - r600-si: RADEON_INFO_TIMESTAMP query
*/
#define KMS_DRIVER_MAJOR 2
#define KMS_DRIVER_MINOR 19
#define KMS_DRIVER_MINOR 20
#define KMS_DRIVER_PATCHLEVEL 0
int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags);
int radeon_driver_unload_kms(struct drm_device *dev);
......
......@@ -29,6 +29,7 @@
#include "drm_sarea.h"
#include "radeon.h"
#include "radeon_drm.h"
#include "radeon_asic.h"
#include <linux/vga_switcheroo.h>
#include <linux/slab.h>
......@@ -167,17 +168,39 @@ static void radeon_set_filp_rights(struct drm_device *dev,
int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
{
struct radeon_device *rdev = dev->dev_private;
struct drm_radeon_info *info;
struct drm_radeon_info *info = data;
struct radeon_mode_info *minfo = &rdev->mode_info;
uint32_t *value_ptr;
uint32_t value;
uint32_t value, *value_ptr;
uint64_t value64, *value_ptr64;
struct drm_crtc *crtc;
int i, found;
info = data;
/* TIMESTAMP is a 64-bit value, needs special handling. */
if (info->request == RADEON_INFO_TIMESTAMP) {
if (rdev->family >= CHIP_R600) {
value_ptr64 = (uint64_t*)((unsigned long)info->value);
if (rdev->family >= CHIP_TAHITI) {
value64 = si_get_gpu_clock(rdev);
} else {
value64 = r600_get_gpu_clock(rdev);
}
if (DRM_COPY_TO_USER(value_ptr64, &value64, sizeof(value64))) {
DRM_ERROR("copy_to_user %s:%u\n", __func__, __LINE__);
return -EFAULT;
}
return 0;
} else {
DRM_DEBUG_KMS("timestamp is r6xx+ only!\n");
return -EINVAL;
}
}
value_ptr = (uint32_t *)((unsigned long)info->value);
if (DRM_COPY_FROM_USER(&value, value_ptr, sizeof(value)))
if (DRM_COPY_FROM_USER(&value, value_ptr, sizeof(value))) {
DRM_ERROR("copy_from_user %s:%u\n", __func__, __LINE__);
return -EFAULT;
}
switch (info->request) {
case RADEON_INFO_DEVICE_ID:
......@@ -337,7 +360,7 @@ int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
return -EINVAL;
}
if (DRM_COPY_TO_USER(value_ptr, &value, sizeof(uint32_t))) {
DRM_ERROR("copy_to_user\n");
DRM_ERROR("copy_to_user %s:%u\n", __func__, __LINE__);
return -EFAULT;
}
return 0;
......
......@@ -3968,3 +3968,22 @@ void si_fini(struct radeon_device *rdev)
rdev->bios = NULL;
}
/**
* si_get_gpu_clock - return GPU clock counter snapshot
*
* @rdev: radeon_device pointer
*
* Fetches a GPU clock counter snapshot (SI).
* Returns the 64 bit clock counter snapshot.
*/
uint64_t si_get_gpu_clock(struct radeon_device *rdev)
{
uint64_t clock;
mutex_lock(&rdev->gpu_clock_mutex);
WREG32(RLC_CAPTURE_GPU_CLOCK_COUNT, 1);
clock = (uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_LSB) |
((uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
mutex_unlock(&rdev->gpu_clock_mutex);
return clock;
}
......@@ -698,6 +698,9 @@
#define RLC_UCODE_ADDR 0xC32C
#define RLC_UCODE_DATA 0xC330
#define RLC_GPU_CLOCK_COUNT_LSB 0xC338
#define RLC_GPU_CLOCK_COUNT_MSB 0xC33C
#define RLC_CAPTURE_GPU_CLOCK_COUNT 0xC340
#define RLC_MC_CNTL 0xC344
#define RLC_UCODE_CNTL 0xC348
......
......@@ -964,6 +964,8 @@ struct drm_radeon_cs {
#define RADEON_INFO_IB_VM_MAX_SIZE 0x0f
/* max pipes - needed for compute shaders */
#define RADEON_INFO_MAX_PIPES 0x10
/* timestamp for GL_ARB_timer_query (OpenGL), returns the current GPU clock */
#define RADEON_INFO_TIMESTAMP 0x11
struct drm_radeon_info {
uint32_t request;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment