X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fgpu%2Fdrm%2Fradeon%2Fr600.c;h=acec26b70d0de17e0dbec645170b3ba6d34432c9;hb=4d60173fc1b12b0c308f861620fe8e2a84f6e5da;hp=609719490ec28c26b064df9d43eb642954fbb442;hpb=c1176d6f03e1085797ce83648a2c76ae15a2b515;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 6097194..acec26b 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -25,12 +25,14 @@ * Alex Deucher * Jerome Glisse */ +#include #include #include #include #include "drmP.h" #include "radeon_drm.h" #include "radeon.h" +#include "radeon_asic.h" #include "radeon_mode.h" #include "r600d.h" #include "atom.h" @@ -38,8 +40,13 @@ #define PFP_UCODE_SIZE 576 #define PM4_UCODE_SIZE 1792 +#define RLC_UCODE_SIZE 768 #define R700_PFP_UCODE_SIZE 848 #define R700_PM4_UCODE_SIZE 1360 +#define R700_RLC_UCODE_SIZE 1024 +#define EVERGREEN_PFP_UCODE_SIZE 1120 +#define EVERGREEN_PM4_UCODE_SIZE 1376 +#define EVERGREEN_RLC_UCODE_SIZE 768 /* Firmware Names */ MODULE_FIRMWARE("radeon/R600_pfp.bin"); @@ -62,6 +69,20 @@ MODULE_FIRMWARE("radeon/RV730_pfp.bin"); MODULE_FIRMWARE("radeon/RV730_me.bin"); MODULE_FIRMWARE("radeon/RV710_pfp.bin"); MODULE_FIRMWARE("radeon/RV710_me.bin"); +MODULE_FIRMWARE("radeon/R600_rlc.bin"); +MODULE_FIRMWARE("radeon/R700_rlc.bin"); +MODULE_FIRMWARE("radeon/CEDAR_pfp.bin"); +MODULE_FIRMWARE("radeon/CEDAR_me.bin"); +MODULE_FIRMWARE("radeon/CEDAR_rlc.bin"); +MODULE_FIRMWARE("radeon/REDWOOD_pfp.bin"); +MODULE_FIRMWARE("radeon/REDWOOD_me.bin"); +MODULE_FIRMWARE("radeon/REDWOOD_rlc.bin"); +MODULE_FIRMWARE("radeon/JUNIPER_pfp.bin"); +MODULE_FIRMWARE("radeon/JUNIPER_me.bin"); +MODULE_FIRMWARE("radeon/JUNIPER_rlc.bin"); +MODULE_FIRMWARE("radeon/CYPRESS_pfp.bin"); +MODULE_FIRMWARE("radeon/CYPRESS_me.bin"); +MODULE_FIRMWARE("radeon/CYPRESS_rlc.bin"); int r600_debugfs_mc_info_init(struct radeon_device *rdev); @@ -69,27 +90,780 @@ int r600_debugfs_mc_info_init(struct radeon_device *rdev); int r600_mc_wait_for_idle(struct radeon_device *rdev); void r600_gpu_init(struct radeon_device *rdev); void r600_fini(struct radeon_device *rdev); +void r600_irq_disable(struct radeon_device *rdev); -/* - * R600 PCIE GART - */ -int r600_gart_clear_page(struct radeon_device *rdev, int i) +void r600_pm_get_dynpm_state(struct radeon_device *rdev) { - void __iomem *ptr = (void *)rdev->gart.table.vram.ptr; - u64 pte; + int i; - if (i < 0 || i > rdev->gart.num_gpu_pages) - return -EINVAL; - pte = 0; - writeq(pte, ((void __iomem *)ptr) + (i * 8)); - return 0; + rdev->pm.dynpm_can_upclock = true; + rdev->pm.dynpm_can_downclock = true; + + /* power state array is low to high, default is first */ + if ((rdev->flags & RADEON_IS_IGP) || (rdev->family == CHIP_R600)) { + int min_power_state_index = 0; + + if (rdev->pm.num_power_states > 2) + min_power_state_index = 1; + + switch (rdev->pm.dynpm_planned_action) { + case DYNPM_ACTION_MINIMUM: + rdev->pm.requested_power_state_index = min_power_state_index; + rdev->pm.requested_clock_mode_index = 0; + rdev->pm.dynpm_can_downclock = false; + break; + case DYNPM_ACTION_DOWNCLOCK: + if (rdev->pm.current_power_state_index == min_power_state_index) { + rdev->pm.requested_power_state_index = rdev->pm.current_power_state_index; + rdev->pm.dynpm_can_downclock = false; + } else { + if (rdev->pm.active_crtc_count > 1) { + for (i = 0; i < rdev->pm.num_power_states; i++) { + if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY) + continue; + else if (i >= rdev->pm.current_power_state_index) { + rdev->pm.requested_power_state_index = + rdev->pm.current_power_state_index; + break; + } else { + rdev->pm.requested_power_state_index = i; + break; + } + } + } else + rdev->pm.requested_power_state_index = + rdev->pm.current_power_state_index - 1; + } + rdev->pm.requested_clock_mode_index = 0; + /* don't use the power state if crtcs are active and no display flag is set */ + if ((rdev->pm.active_crtc_count > 0) && + (rdev->pm.power_state[rdev->pm.requested_power_state_index]. + clock_info[rdev->pm.requested_clock_mode_index].flags & + RADEON_PM_MODE_NO_DISPLAY)) { + rdev->pm.requested_power_state_index++; + } + break; + case DYNPM_ACTION_UPCLOCK: + if (rdev->pm.current_power_state_index == (rdev->pm.num_power_states - 1)) { + rdev->pm.requested_power_state_index = rdev->pm.current_power_state_index; + rdev->pm.dynpm_can_upclock = false; + } else { + if (rdev->pm.active_crtc_count > 1) { + for (i = (rdev->pm.num_power_states - 1); i >= 0; i--) { + if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY) + continue; + else if (i <= rdev->pm.current_power_state_index) { + rdev->pm.requested_power_state_index = + rdev->pm.current_power_state_index; + break; + } else { + rdev->pm.requested_power_state_index = i; + break; + } + } + } else + rdev->pm.requested_power_state_index = + rdev->pm.current_power_state_index + 1; + } + rdev->pm.requested_clock_mode_index = 0; + break; + case DYNPM_ACTION_DEFAULT: + rdev->pm.requested_power_state_index = rdev->pm.default_power_state_index; + rdev->pm.requested_clock_mode_index = 0; + rdev->pm.dynpm_can_upclock = false; + break; + case DYNPM_ACTION_NONE: + default: + DRM_ERROR("Requested mode for not defined action\n"); + return; + } + } else { + /* XXX select a power state based on AC/DC, single/dualhead, etc. */ + /* for now just select the first power state and switch between clock modes */ + /* power state array is low to high, default is first (0) */ + if (rdev->pm.active_crtc_count > 1) { + rdev->pm.requested_power_state_index = -1; + /* start at 1 as we don't want the default mode */ + for (i = 1; i < rdev->pm.num_power_states; i++) { + if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY) + continue; + else if ((rdev->pm.power_state[i].type == POWER_STATE_TYPE_PERFORMANCE) || + (rdev->pm.power_state[i].type == POWER_STATE_TYPE_BATTERY)) { + rdev->pm.requested_power_state_index = i; + break; + } + } + /* if nothing selected, grab the default state. */ + if (rdev->pm.requested_power_state_index == -1) + rdev->pm.requested_power_state_index = 0; + } else + rdev->pm.requested_power_state_index = 1; + + switch (rdev->pm.dynpm_planned_action) { + case DYNPM_ACTION_MINIMUM: + rdev->pm.requested_clock_mode_index = 0; + rdev->pm.dynpm_can_downclock = false; + break; + case DYNPM_ACTION_DOWNCLOCK: + if (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index) { + if (rdev->pm.current_clock_mode_index == 0) { + rdev->pm.requested_clock_mode_index = 0; + rdev->pm.dynpm_can_downclock = false; + } else + rdev->pm.requested_clock_mode_index = + rdev->pm.current_clock_mode_index - 1; + } else { + rdev->pm.requested_clock_mode_index = 0; + rdev->pm.dynpm_can_downclock = false; + } + /* don't use the power state if crtcs are active and no display flag is set */ + if ((rdev->pm.active_crtc_count > 0) && + (rdev->pm.power_state[rdev->pm.requested_power_state_index]. + clock_info[rdev->pm.requested_clock_mode_index].flags & + RADEON_PM_MODE_NO_DISPLAY)) { + rdev->pm.requested_clock_mode_index++; + } + break; + case DYNPM_ACTION_UPCLOCK: + if (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index) { + if (rdev->pm.current_clock_mode_index == + (rdev->pm.power_state[rdev->pm.requested_power_state_index].num_clock_modes - 1)) { + rdev->pm.requested_clock_mode_index = rdev->pm.current_clock_mode_index; + rdev->pm.dynpm_can_upclock = false; + } else + rdev->pm.requested_clock_mode_index = + rdev->pm.current_clock_mode_index + 1; + } else { + rdev->pm.requested_clock_mode_index = + rdev->pm.power_state[rdev->pm.requested_power_state_index].num_clock_modes - 1; + rdev->pm.dynpm_can_upclock = false; + } + break; + case DYNPM_ACTION_DEFAULT: + rdev->pm.requested_power_state_index = rdev->pm.default_power_state_index; + rdev->pm.requested_clock_mode_index = 0; + rdev->pm.dynpm_can_upclock = false; + break; + case DYNPM_ACTION_NONE: + default: + DRM_ERROR("Requested mode for not defined action\n"); + return; + } + } + + DRM_DEBUG("Requested: e: %d m: %d p: %d\n", + rdev->pm.power_state[rdev->pm.requested_power_state_index]. + clock_info[rdev->pm.requested_clock_mode_index].sclk, + rdev->pm.power_state[rdev->pm.requested_power_state_index]. + clock_info[rdev->pm.requested_clock_mode_index].mclk, + rdev->pm.power_state[rdev->pm.requested_power_state_index]. + pcie_lanes); +} + +static int r600_pm_get_type_index(struct radeon_device *rdev, + enum radeon_pm_state_type ps_type, + int instance) +{ + int i; + int found_instance = -1; + + for (i = 0; i < rdev->pm.num_power_states; i++) { + if (rdev->pm.power_state[i].type == ps_type) { + found_instance++; + if (found_instance == instance) + return i; + } + } + /* return default if no match */ + return rdev->pm.default_power_state_index; +} + +void rs780_pm_init_profile(struct radeon_device *rdev) +{ + if (rdev->pm.num_power_states == 2) { + /* default */ + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0; + /* low sh */ + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + /* mid sh */ + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0; + /* high sh */ + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0; + /* low mh */ + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + /* mid mh */ + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0; + /* high mh */ + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0; + } else if (rdev->pm.num_power_states == 3) { + /* default */ + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0; + /* low sh */ + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + /* mid sh */ + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0; + /* high sh */ + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0; + /* low mh */ + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + /* mid mh */ + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0; + /* high mh */ + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0; + } else { + /* default */ + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0; + /* low sh */ + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + /* mid sh */ + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0; + /* high sh */ + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 3; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0; + /* low mh */ + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + /* mid mh */ + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0; + /* high mh */ + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 3; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0; + } +} + +void r600_pm_init_profile(struct radeon_device *rdev) +{ + if (rdev->family == CHIP_R600) { + /* XXX */ + /* default */ + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0; + /* low sh */ + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + /* mid sh */ + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0; + /* high sh */ + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0; + /* low mh */ + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + /* mid mh */ + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0; + /* high mh */ + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0; + } else { + if (rdev->pm.num_power_states < 4) { + /* default */ + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 2; + /* low sh */ + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + /* mid sh */ + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1; + /* high sh */ + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 1; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 2; + /* low mh */ + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + /* low mh */ + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1; + /* high mh */ + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 2; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 2; + } else { + /* default */ + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 2; + /* low sh */ + if (rdev->flags & RADEON_IS_MOBILITY) { + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0); + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0); + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + } else { + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0; + } + /* mid sh */ + if (rdev->flags & RADEON_IS_MOBILITY) { + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0); + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0); + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1; + } else { + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1; + } + /* high sh */ + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 2; + /* low mh */ + if (rdev->flags & RADEON_IS_MOBILITY) { + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1); + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1); + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + } else { + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0; + } + /* mid mh */ + if (rdev->flags & RADEON_IS_MOBILITY) { + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1); + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1); + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1; + } else { + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1; + } + /* high mh */ + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = + r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0; + rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 2; + } + } +} + +void r600_pm_misc(struct radeon_device *rdev) +{ + int requested_index = rdev->pm.requested_power_state_index; + struct radeon_power_state *ps = &rdev->pm.power_state[requested_index]; + struct radeon_voltage *voltage = &ps->clock_info[0].voltage; + + if ((voltage->type == VOLTAGE_SW) && voltage->voltage) { + if (voltage->voltage != rdev->pm.current_vddc) { + radeon_atom_set_voltage(rdev, voltage->voltage); + rdev->pm.current_vddc = voltage->voltage; + } + } +} + +bool r600_gui_idle(struct radeon_device *rdev) +{ + if (RREG32(GRBM_STATUS) & GUI_ACTIVE) + return false; + else + return true; +} + +/* hpd for digital panel detect/disconnect */ +bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd) +{ + bool connected = false; + + if (ASIC_IS_DCE3(rdev)) { + switch (hpd) { + case RADEON_HPD_1: + if (RREG32(DC_HPD1_INT_STATUS) & DC_HPDx_SENSE) + connected = true; + break; + case RADEON_HPD_2: + if (RREG32(DC_HPD2_INT_STATUS) & DC_HPDx_SENSE) + connected = true; + break; + case RADEON_HPD_3: + if (RREG32(DC_HPD3_INT_STATUS) & DC_HPDx_SENSE) + connected = true; + break; + case RADEON_HPD_4: + if (RREG32(DC_HPD4_INT_STATUS) & DC_HPDx_SENSE) + connected = true; + break; + /* DCE 3.2 */ + case RADEON_HPD_5: + if (RREG32(DC_HPD5_INT_STATUS) & DC_HPDx_SENSE) + connected = true; + break; + case RADEON_HPD_6: + if (RREG32(DC_HPD6_INT_STATUS) & DC_HPDx_SENSE) + connected = true; + break; + default: + break; + } + } else { + switch (hpd) { + case RADEON_HPD_1: + if (RREG32(DC_HOT_PLUG_DETECT1_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE) + connected = true; + break; + case RADEON_HPD_2: + if (RREG32(DC_HOT_PLUG_DETECT2_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE) + connected = true; + break; + case RADEON_HPD_3: + if (RREG32(DC_HOT_PLUG_DETECT3_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE) + connected = true; + break; + default: + break; + } + } + return connected; } +void r600_hpd_set_polarity(struct radeon_device *rdev, + enum radeon_hpd_id hpd) +{ + u32 tmp; + bool connected = r600_hpd_sense(rdev, hpd); + + if (ASIC_IS_DCE3(rdev)) { + switch (hpd) { + case RADEON_HPD_1: + tmp = RREG32(DC_HPD1_INT_CONTROL); + if (connected) + tmp &= ~DC_HPDx_INT_POLARITY; + else + tmp |= DC_HPDx_INT_POLARITY; + WREG32(DC_HPD1_INT_CONTROL, tmp); + break; + case RADEON_HPD_2: + tmp = RREG32(DC_HPD2_INT_CONTROL); + if (connected) + tmp &= ~DC_HPDx_INT_POLARITY; + else + tmp |= DC_HPDx_INT_POLARITY; + WREG32(DC_HPD2_INT_CONTROL, tmp); + break; + case RADEON_HPD_3: + tmp = RREG32(DC_HPD3_INT_CONTROL); + if (connected) + tmp &= ~DC_HPDx_INT_POLARITY; + else + tmp |= DC_HPDx_INT_POLARITY; + WREG32(DC_HPD3_INT_CONTROL, tmp); + break; + case RADEON_HPD_4: + tmp = RREG32(DC_HPD4_INT_CONTROL); + if (connected) + tmp &= ~DC_HPDx_INT_POLARITY; + else + tmp |= DC_HPDx_INT_POLARITY; + WREG32(DC_HPD4_INT_CONTROL, tmp); + break; + case RADEON_HPD_5: + tmp = RREG32(DC_HPD5_INT_CONTROL); + if (connected) + tmp &= ~DC_HPDx_INT_POLARITY; + else + tmp |= DC_HPDx_INT_POLARITY; + WREG32(DC_HPD5_INT_CONTROL, tmp); + break; + /* DCE 3.2 */ + case RADEON_HPD_6: + tmp = RREG32(DC_HPD6_INT_CONTROL); + if (connected) + tmp &= ~DC_HPDx_INT_POLARITY; + else + tmp |= DC_HPDx_INT_POLARITY; + WREG32(DC_HPD6_INT_CONTROL, tmp); + break; + default: + break; + } + } else { + switch (hpd) { + case RADEON_HPD_1: + tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL); + if (connected) + tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY; + else + tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY; + WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp); + break; + case RADEON_HPD_2: + tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL); + if (connected) + tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY; + else + tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY; + WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp); + break; + case RADEON_HPD_3: + tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL); + if (connected) + tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY; + else + tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY; + WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp); + break; + default: + break; + } + } +} + +void r600_hpd_init(struct radeon_device *rdev) +{ + struct drm_device *dev = rdev->ddev; + struct drm_connector *connector; + + if (ASIC_IS_DCE3(rdev)) { + u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4) | DC_HPDx_RX_INT_TIMER(0xfa); + if (ASIC_IS_DCE32(rdev)) + tmp |= DC_HPDx_EN; + + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + switch (radeon_connector->hpd.hpd) { + case RADEON_HPD_1: + WREG32(DC_HPD1_CONTROL, tmp); + rdev->irq.hpd[0] = true; + break; + case RADEON_HPD_2: + WREG32(DC_HPD2_CONTROL, tmp); + rdev->irq.hpd[1] = true; + break; + case RADEON_HPD_3: + WREG32(DC_HPD3_CONTROL, tmp); + rdev->irq.hpd[2] = true; + break; + case RADEON_HPD_4: + WREG32(DC_HPD4_CONTROL, tmp); + rdev->irq.hpd[3] = true; + break; + /* DCE 3.2 */ + case RADEON_HPD_5: + WREG32(DC_HPD5_CONTROL, tmp); + rdev->irq.hpd[4] = true; + break; + case RADEON_HPD_6: + WREG32(DC_HPD6_CONTROL, tmp); + rdev->irq.hpd[5] = true; + break; + default: + break; + } + } + } else { + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + switch (radeon_connector->hpd.hpd) { + case RADEON_HPD_1: + WREG32(DC_HOT_PLUG_DETECT1_CONTROL, DC_HOT_PLUG_DETECTx_EN); + rdev->irq.hpd[0] = true; + break; + case RADEON_HPD_2: + WREG32(DC_HOT_PLUG_DETECT2_CONTROL, DC_HOT_PLUG_DETECTx_EN); + rdev->irq.hpd[1] = true; + break; + case RADEON_HPD_3: + WREG32(DC_HOT_PLUG_DETECT3_CONTROL, DC_HOT_PLUG_DETECTx_EN); + rdev->irq.hpd[2] = true; + break; + default: + break; + } + } + } + if (rdev->irq.installed) + r600_irq_set(rdev); +} + +void r600_hpd_fini(struct radeon_device *rdev) +{ + struct drm_device *dev = rdev->ddev; + struct drm_connector *connector; + + if (ASIC_IS_DCE3(rdev)) { + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + switch (radeon_connector->hpd.hpd) { + case RADEON_HPD_1: + WREG32(DC_HPD1_CONTROL, 0); + rdev->irq.hpd[0] = false; + break; + case RADEON_HPD_2: + WREG32(DC_HPD2_CONTROL, 0); + rdev->irq.hpd[1] = false; + break; + case RADEON_HPD_3: + WREG32(DC_HPD3_CONTROL, 0); + rdev->irq.hpd[2] = false; + break; + case RADEON_HPD_4: + WREG32(DC_HPD4_CONTROL, 0); + rdev->irq.hpd[3] = false; + break; + /* DCE 3.2 */ + case RADEON_HPD_5: + WREG32(DC_HPD5_CONTROL, 0); + rdev->irq.hpd[4] = false; + break; + case RADEON_HPD_6: + WREG32(DC_HPD6_CONTROL, 0); + rdev->irq.hpd[5] = false; + break; + default: + break; + } + } + } else { + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + switch (radeon_connector->hpd.hpd) { + case RADEON_HPD_1: + WREG32(DC_HOT_PLUG_DETECT1_CONTROL, 0); + rdev->irq.hpd[0] = false; + break; + case RADEON_HPD_2: + WREG32(DC_HOT_PLUG_DETECT2_CONTROL, 0); + rdev->irq.hpd[1] = false; + break; + case RADEON_HPD_3: + WREG32(DC_HOT_PLUG_DETECT3_CONTROL, 0); + rdev->irq.hpd[2] = false; + break; + default: + break; + } + } + } +} + +/* + * R600 PCIE GART + */ void r600_pcie_gart_tlb_flush(struct radeon_device *rdev) { unsigned i; u32 tmp; + /* flush hdp cache so updates hit vram */ + WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1); + WREG32(VM_CONTEXT0_INVALIDATION_LOW_ADDR, rdev->mc.gtt_start >> 12); WREG32(VM_CONTEXT0_INVALIDATION_HIGH_ADDR, (rdev->mc.gtt_end - 1) >> 12); WREG32(VM_CONTEXT0_REQUEST_RESPONSE, REQUEST_TYPE(1)); @@ -136,6 +910,7 @@ int r600_pcie_gart_enable(struct radeon_device *rdev) r = radeon_gart_table_vram_pin(rdev); if (r) return r; + radeon_gart_restore(rdev); /* Setup L2 cache */ WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING | @@ -180,7 +955,7 @@ int r600_pcie_gart_enable(struct radeon_device *rdev) void r600_pcie_gart_disable(struct radeon_device *rdev) { u32 tmp; - int i; + int i, r; /* Disable all tables */ for (i = 0; i < 7; i++) @@ -208,16 +983,20 @@ void r600_pcie_gart_disable(struct radeon_device *rdev) WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp); WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp); if (rdev->gart.table.vram.robj) { - radeon_object_kunmap(rdev->gart.table.vram.robj); - radeon_object_unpin(rdev->gart.table.vram.robj); + r = radeon_bo_reserve(rdev->gart.table.vram.robj, false); + if (likely(r == 0)) { + radeon_bo_kunmap(rdev->gart.table.vram.robj); + radeon_bo_unpin(rdev->gart.table.vram.robj); + radeon_bo_unreserve(rdev->gart.table.vram.robj); + } } } void r600_pcie_gart_fini(struct radeon_device *rdev) { + radeon_gart_fini(rdev); r600_pcie_gart_disable(rdev); radeon_gart_table_vram_free(rdev); - radeon_gart_fini(rdev); } void r600_agp_enable(struct radeon_device *rdev) @@ -335,15 +1114,74 @@ static void r600_mc_program(struct radeon_device *rdev) rv515_vga_render_disable(rdev); } +/** + * r600_vram_gtt_location - try to find VRAM & GTT location + * @rdev: radeon device structure holding all necessary informations + * @mc: memory controller structure holding memory informations + * + * Function will place try to place VRAM at same place as in CPU (PCI) + * address space as some GPU seems to have issue when we reprogram at + * different address space. + * + * If there is not enough space to fit the unvisible VRAM after the + * aperture then we limit the VRAM size to the aperture. + * + * If we are using AGP then place VRAM adjacent to AGP aperture are we need + * them to be in one from GPU point of view so that we can program GPU to + * catch access outside them (weird GPU policy see ??). + * + * This function will never fails, worst case are limiting VRAM or GTT. + * + * Note: GTT start, end, size should be initialized before calling this + * function on AGP platform. + */ +void r600_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc) +{ + u64 size_bf, size_af; + + if (mc->mc_vram_size > 0xE0000000) { + /* leave room for at least 512M GTT */ + dev_warn(rdev->dev, "limiting VRAM\n"); + mc->real_vram_size = 0xE0000000; + mc->mc_vram_size = 0xE0000000; + } + if (rdev->flags & RADEON_IS_AGP) { + size_bf = mc->gtt_start; + size_af = 0xFFFFFFFF - mc->gtt_end + 1; + if (size_bf > size_af) { + if (mc->mc_vram_size > size_bf) { + dev_warn(rdev->dev, "limiting VRAM\n"); + mc->real_vram_size = size_bf; + mc->mc_vram_size = size_bf; + } + mc->vram_start = mc->gtt_start - mc->mc_vram_size; + } else { + if (mc->mc_vram_size > size_af) { + dev_warn(rdev->dev, "limiting VRAM\n"); + mc->real_vram_size = size_af; + mc->mc_vram_size = size_af; + } + mc->vram_start = mc->gtt_end; + } + mc->vram_end = mc->vram_start + mc->mc_vram_size - 1; + dev_info(rdev->dev, "VRAM: %lluM 0x%08llX - 0x%08llX (%lluM used)\n", + mc->mc_vram_size >> 20, mc->vram_start, + mc->vram_end, mc->real_vram_size >> 20); + } else { + u64 base = 0; + if (rdev->flags & RADEON_IS_IGP) + base = (RREG32(MC_VM_FB_LOCATION) & 0xFFFF) << 24; + radeon_vram_location(rdev, &rdev->mc, base); + radeon_gtt_location(rdev, mc); + } +} + int r600_mc_init(struct radeon_device *rdev) { - fixed20_12 a; u32 tmp; - int chansize; - int r; + int chansize, numchan; /* Get VRAM informations */ - rdev->mc.vram_width = 128; rdev->mc.vram_is_ddr = true; tmp = RREG32(RAMCFG); if (tmp & CHANSIZE_OVERRIDE) { @@ -353,97 +1191,35 @@ int r600_mc_init(struct radeon_device *rdev) } else { chansize = 32; } - if (rdev->family == CHIP_R600) { - rdev->mc.vram_width = 8 * chansize; - } else if (rdev->family == CHIP_RV670) { - rdev->mc.vram_width = 4 * chansize; - } else if ((rdev->family == CHIP_RV610) || - (rdev->family == CHIP_RV620)) { - rdev->mc.vram_width = chansize; - } else if ((rdev->family == CHIP_RV630) || - (rdev->family == CHIP_RV635)) { - rdev->mc.vram_width = 2 * chansize; + tmp = RREG32(CHMAP); + switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) { + case 0: + default: + numchan = 1; + break; + case 1: + numchan = 2; + break; + case 2: + numchan = 4; + break; + case 3: + numchan = 8; + break; } + rdev->mc.vram_width = numchan * chansize; /* Could aper size report 0 ? */ rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); /* Setup GPU memory space */ rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE); rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE); + rdev->mc.visible_vram_size = rdev->mc.aper_size; + r600_vram_gtt_location(rdev, &rdev->mc); - if (rdev->mc.mc_vram_size > rdev->mc.aper_size) - rdev->mc.mc_vram_size = rdev->mc.aper_size; - - if (rdev->mc.real_vram_size > rdev->mc.aper_size) - rdev->mc.real_vram_size = rdev->mc.aper_size; - - if (rdev->flags & RADEON_IS_AGP) { - r = radeon_agp_init(rdev); - if (r) - return r; - /* gtt_size is setup by radeon_agp_init */ - rdev->mc.gtt_location = rdev->mc.agp_base; - tmp = 0xFFFFFFFFUL - rdev->mc.agp_base - rdev->mc.gtt_size; - /* Try to put vram before or after AGP because we - * we want SYSTEM_APERTURE to cover both VRAM and - * AGP so that GPU can catch out of VRAM/AGP access - */ - if (rdev->mc.gtt_location > rdev->mc.mc_vram_size) { - /* Enought place before */ - rdev->mc.vram_location = rdev->mc.gtt_location - - rdev->mc.mc_vram_size; - } else if (tmp > rdev->mc.mc_vram_size) { - /* Enought place after */ - rdev->mc.vram_location = rdev->mc.gtt_location + - rdev->mc.gtt_size; - } else { - /* Try to setup VRAM then AGP might not - * not work on some card - */ - rdev->mc.vram_location = 0x00000000UL; - rdev->mc.gtt_location = rdev->mc.mc_vram_size; - } - } else { - if (rdev->family == CHIP_RS780 || rdev->family == CHIP_RS880) { - rdev->mc.vram_location = (RREG32(MC_VM_FB_LOCATION) & - 0xFFFF) << 24; - rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024; - tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size; - if ((0xFFFFFFFFUL - tmp) >= rdev->mc.gtt_size) { - /* Enough place after vram */ - rdev->mc.gtt_location = tmp; - } else if (rdev->mc.vram_location >= rdev->mc.gtt_size) { - /* Enough place before vram */ - rdev->mc.gtt_location = 0; - } else { - /* Not enough place after or before shrink - * gart size - */ - if (rdev->mc.vram_location > (0xFFFFFFFFUL - tmp)) { - rdev->mc.gtt_location = 0; - rdev->mc.gtt_size = rdev->mc.vram_location; - } else { - rdev->mc.gtt_location = tmp; - rdev->mc.gtt_size = 0xFFFFFFFFUL - tmp; - } - } - rdev->mc.gtt_location = rdev->mc.mc_vram_size; - } else { - rdev->mc.vram_location = 0x00000000UL; - rdev->mc.gtt_location = rdev->mc.mc_vram_size; - rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024; - } - } - rdev->mc.vram_start = rdev->mc.vram_location; - rdev->mc.vram_end = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; - rdev->mc.gtt_start = rdev->mc.gtt_location; - rdev->mc.gtt_end = rdev->mc.gtt_location + rdev->mc.gtt_size - 1; - /* FIXME: we should enforce default clock in case GPU is not in - * default setup - */ - a.full = rfixed_const(100); - rdev->pm.sclk.full = rfixed_const(rdev->clock.default_sclk); - rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a); + if (rdev->flags & RADEON_IS_IGP) + rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev); + radeon_update_bandwidth_info(rdev); return 0; } @@ -470,7 +1246,6 @@ int r600_gpu_soft_reset(struct radeon_device *rdev) S_008014_DB2_BUSY(1) | S_008014_DB3_BUSY(1) | S_008014_CB0_BUSY(1) | S_008014_CB1_BUSY(1) | S_008014_CB2_BUSY(1) | S_008014_CB3_BUSY(1); - u32 srbm_reset = 0; u32 tmp; dev_info(rdev->dev, "GPU softreset \n"); @@ -485,7 +1260,7 @@ int r600_gpu_soft_reset(struct radeon_device *rdev) dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); } /* Disable CP parsing/prefetching */ - WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(0xff)); + WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1)); /* Check if any of the rendering block is busy and reset it */ if ((RREG32(R_008010_GRBM_STATUS) & grbm_busy_mask) || (RREG32(R_008014_GRBM_STATUS2) & grbm2_busy_mask)) { @@ -504,72 +1279,56 @@ int r600_gpu_soft_reset(struct radeon_device *rdev) S_008020_SOFT_RESET_VGT(1); dev_info(rdev->dev, " R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp); WREG32(R_008020_GRBM_SOFT_RESET, tmp); - (void)RREG32(R_008020_GRBM_SOFT_RESET); - udelay(50); + RREG32(R_008020_GRBM_SOFT_RESET); + mdelay(15); WREG32(R_008020_GRBM_SOFT_RESET, 0); - (void)RREG32(R_008020_GRBM_SOFT_RESET); } /* Reset CP (we always reset CP) */ tmp = S_008020_SOFT_RESET_CP(1); dev_info(rdev->dev, "R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp); WREG32(R_008020_GRBM_SOFT_RESET, tmp); - (void)RREG32(R_008020_GRBM_SOFT_RESET); - udelay(50); + RREG32(R_008020_GRBM_SOFT_RESET); + mdelay(15); WREG32(R_008020_GRBM_SOFT_RESET, 0); - (void)RREG32(R_008020_GRBM_SOFT_RESET); - /* Reset others GPU block if necessary */ - if (G_000E50_RLC_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_RLC(1); - if (G_000E50_GRBM_RQ_PENDING(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_GRBM(1); - if (G_000E50_HI_RQ_PENDING(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_IH(1); - if (G_000E50_VMC_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_VMC(1); - if (G_000E50_MCB_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_MC(1); - if (G_000E50_MCDZ_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_MC(1); - if (G_000E50_MCDY_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_MC(1); - if (G_000E50_MCDX_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_MC(1); - if (G_000E50_MCDW_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_MC(1); - if (G_000E50_RLC_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_RLC(1); - if (G_000E50_SEM_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_SEM(1); - if (G_000E50_BIF_BUSY(RREG32(R_000E50_SRBM_STATUS))) - srbm_reset |= S_000E60_SOFT_RESET_BIF(1); - dev_info(rdev->dev, " R_000E60_SRBM_SOFT_RESET=0x%08X\n", srbm_reset); - WREG32(R_000E60_SRBM_SOFT_RESET, srbm_reset); - (void)RREG32(R_000E60_SRBM_SOFT_RESET); - udelay(50); - WREG32(R_000E60_SRBM_SOFT_RESET, 0); - (void)RREG32(R_000E60_SRBM_SOFT_RESET); - WREG32(R_000E60_SRBM_SOFT_RESET, srbm_reset); - (void)RREG32(R_000E60_SRBM_SOFT_RESET); - udelay(50); - WREG32(R_000E60_SRBM_SOFT_RESET, 0); - (void)RREG32(R_000E60_SRBM_SOFT_RESET); /* Wait a little for things to settle down */ - udelay(50); + mdelay(1); dev_info(rdev->dev, " R_008010_GRBM_STATUS=0x%08X\n", RREG32(R_008010_GRBM_STATUS)); dev_info(rdev->dev, " R_008014_GRBM_STATUS2=0x%08X\n", RREG32(R_008014_GRBM_STATUS2)); dev_info(rdev->dev, " R_000E50_SRBM_STATUS=0x%08X\n", RREG32(R_000E50_SRBM_STATUS)); - /* After reset we need to reinit the asic as GPU often endup in an - * incoherent state. - */ - atom_asic_init(rdev->mode_info.atom_context); rv515_mc_resume(rdev, &save); return 0; } -int r600_gpu_reset(struct radeon_device *rdev) +bool r600_gpu_is_lockup(struct radeon_device *rdev) +{ + u32 srbm_status; + u32 grbm_status; + u32 grbm_status2; + int r; + + srbm_status = RREG32(R_000E50_SRBM_STATUS); + grbm_status = RREG32(R_008010_GRBM_STATUS); + grbm_status2 = RREG32(R_008014_GRBM_STATUS2); + if (!G_008010_GUI_ACTIVE(grbm_status)) { + r100_gpu_lockup_update(&rdev->config.r300.lockup, &rdev->cp); + return false; + } + /* force CP activities */ + r = radeon_ring_lock(rdev, 2); + if (!r) { + /* PACKET2 NOP */ + radeon_ring_write(rdev, 0x80000000); + radeon_ring_write(rdev, 0x80000000); + radeon_ring_unlock_commit(rdev); + } + rdev->cp.rptr = RREG32(R600_CP_RB_RPTR); + return r100_gpu_cp_is_lockup(rdev, &rdev->config.r300.lockup, &rdev->cp); +} + +int r600_asic_reset(struct radeon_device *rdev) { return r600_gpu_soft_reset(rdev); } @@ -698,6 +1457,9 @@ void r600_gpu_init(struct radeon_device *rdev) { u32 tiling_config; u32 ramcfg; + u32 backend_map; + u32 cc_rb_backend_disable; + u32 cc_gc_shader_pipe_config; u32 tmp; int i, j; u32 sq_config; @@ -807,8 +1569,11 @@ void r600_gpu_init(struct radeon_device *rdev) default: break; } + rdev->config.r600.tiling_npipes = rdev->config.r600.max_tile_pipes; + rdev->config.r600.tiling_nbanks = 4 << ((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT); tiling_config |= BANK_TILING((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT); tiling_config |= GROUP_SIZE(0); + rdev->config.r600.tiling_group_size = 256; tmp = (ramcfg & NOOFROWS_MASK) >> NOOFROWS_SHIFT; if (tmp > 3) { tiling_config |= ROW_TILING(3); @@ -818,24 +1583,34 @@ void r600_gpu_init(struct radeon_device *rdev) tiling_config |= SAMPLE_SPLIT(tmp); } tiling_config |= BANK_SWAPS(1); - tmp = r600_get_tile_pipe_to_backend_map(rdev->config.r600.max_tile_pipes, - rdev->config.r600.max_backends, - (0xff << rdev->config.r600.max_backends) & 0xff); - tiling_config |= BACKEND_MAP(tmp); + + cc_rb_backend_disable = RREG32(CC_RB_BACKEND_DISABLE) & 0x00ff0000; + cc_rb_backend_disable |= + BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << rdev->config.r600.max_backends) & R6XX_MAX_BACKENDS_MASK); + + cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG) & 0xffffff00; + cc_gc_shader_pipe_config |= + INACTIVE_QD_PIPES((R6XX_MAX_PIPES_MASK << rdev->config.r600.max_pipes) & R6XX_MAX_PIPES_MASK); + cc_gc_shader_pipe_config |= + INACTIVE_SIMDS((R6XX_MAX_SIMDS_MASK << rdev->config.r600.max_simds) & R6XX_MAX_SIMDS_MASK); + + backend_map = r600_get_tile_pipe_to_backend_map(rdev->config.r600.max_tile_pipes, + (R6XX_MAX_BACKENDS - + r600_count_pipe_bits((cc_rb_backend_disable & + R6XX_MAX_BACKENDS_MASK) >> 16)), + (cc_rb_backend_disable >> 16)); + + tiling_config |= BACKEND_MAP(backend_map); WREG32(GB_TILING_CONFIG, tiling_config); WREG32(DCP_TILING_CONFIG, tiling_config & 0xffff); WREG32(HDP_TILING_CONFIG, tiling_config & 0xffff); - tmp = BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << rdev->config.r600.max_backends) & R6XX_MAX_BACKENDS_MASK); - WREG32(CC_RB_BACKEND_DISABLE, tmp); - /* Setup pipes */ - tmp = INACTIVE_QD_PIPES((R6XX_MAX_PIPES_MASK << rdev->config.r600.max_pipes) & R6XX_MAX_PIPES_MASK); - tmp |= INACTIVE_SIMDS((R6XX_MAX_SIMDS_MASK << rdev->config.r600.max_simds) & R6XX_MAX_SIMDS_MASK); - WREG32(CC_GC_SHADER_PIPE_CONFIG, tmp); - WREG32(GC_USER_SHADER_PIPE_CONFIG, tmp); + WREG32(CC_RB_BACKEND_DISABLE, cc_rb_backend_disable); + WREG32(CC_GC_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config); + WREG32(GC_USER_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config); - tmp = R6XX_MAX_BACKENDS - r600_count_pipe_bits(tmp & INACTIVE_QD_PIPES_MASK); + tmp = R6XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK) >> 8); WREG32(VGT_OUT_DEALLOC_CNTL, (tmp * 4) & DEALLOC_DIST_MASK); WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, ((tmp * 4) - 2) & VTX_REUSE_DEPTH_MASK); @@ -859,7 +1634,8 @@ void r600_gpu_init(struct radeon_device *rdev) ((rdev->family) == CHIP_RV630) || ((rdev->family) == CHIP_RV610) || ((rdev->family) == CHIP_RV620) || - ((rdev->family) == CHIP_RS780)) { + ((rdev->family) == CHIP_RS780) || + ((rdev->family) == CHIP_RS880)) { WREG32(DB_DEBUG, PREZ_MUST_WAIT_FOR_POSTZ_DONE); } else { WREG32(DB_DEBUG, 0); @@ -876,7 +1652,8 @@ void r600_gpu_init(struct radeon_device *rdev) tmp = RREG32(SQ_MS_FIFO_SIZES); if (((rdev->family) == CHIP_RV610) || ((rdev->family) == CHIP_RV620) || - ((rdev->family) == CHIP_RS780)) { + ((rdev->family) == CHIP_RS780) || + ((rdev->family) == CHIP_RS880)) { tmp = (CACHE_FIFO_SIZE(0xa) | FETCH_FIFO_HIWATER(0xa) | DONE_FIFO_HIWATER(0xe0) | @@ -919,7 +1696,8 @@ void r600_gpu_init(struct radeon_device *rdev) NUM_ES_STACK_ENTRIES(0)); } else if (((rdev->family) == CHIP_RV610) || ((rdev->family) == CHIP_RV620) || - ((rdev->family) == CHIP_RS780)) { + ((rdev->family) == CHIP_RS780) || + ((rdev->family) == CHIP_RS880)) { /* no vertex cache */ sq_config &= ~VC_ENABLE; @@ -976,7 +1754,8 @@ void r600_gpu_init(struct radeon_device *rdev) if (((rdev->family) == CHIP_RV610) || ((rdev->family) == CHIP_RV620) || - ((rdev->family) == CHIP_RS780)) { + ((rdev->family) == CHIP_RS780) || + ((rdev->family) == CHIP_RS880)) { WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(TC_ONLY)); } else { WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC)); @@ -1002,8 +1781,9 @@ void r600_gpu_init(struct radeon_device *rdev) tmp = rdev->config.r600.max_pipes * 16; switch (rdev->family) { case CHIP_RV610: - case CHIP_RS780: case CHIP_RV620: + case CHIP_RS780: + case CHIP_RS880: tmp += 32; break; case CHIP_RV670: @@ -1044,8 +1824,9 @@ void r600_gpu_init(struct radeon_device *rdev) switch (rdev->family) { case CHIP_RV610: - case CHIP_RS780: case CHIP_RV620: + case CHIP_RS780: + case CHIP_RS880: tmp = TC_L2_SIZE(8); break; case CHIP_RV630: @@ -1096,7 +1877,6 @@ void r600_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v) (void)RREG32(PCIE_PORT_DATA); } - /* * CP & Ring */ @@ -1105,11 +1885,12 @@ void r600_cp_stop(struct radeon_device *rdev) WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1)); } -int r600_cp_init_microcode(struct radeon_device *rdev) +int r600_init_microcode(struct radeon_device *rdev) { struct platform_device *pdev; const char *chip_name; - size_t pfp_req_size, me_req_size; + const char *rlc_chip_name; + size_t pfp_req_size, me_req_size, rlc_req_size; char fw_name[30]; int err; @@ -1123,30 +1904,83 @@ int r600_cp_init_microcode(struct radeon_device *rdev) } switch (rdev->family) { - case CHIP_R600: chip_name = "R600"; break; - case CHIP_RV610: chip_name = "RV610"; break; - case CHIP_RV630: chip_name = "RV630"; break; - case CHIP_RV620: chip_name = "RV620"; break; - case CHIP_RV635: chip_name = "RV635"; break; - case CHIP_RV670: chip_name = "RV670"; break; + case CHIP_R600: + chip_name = "R600"; + rlc_chip_name = "R600"; + break; + case CHIP_RV610: + chip_name = "RV610"; + rlc_chip_name = "R600"; + break; + case CHIP_RV630: + chip_name = "RV630"; + rlc_chip_name = "R600"; + break; + case CHIP_RV620: + chip_name = "RV620"; + rlc_chip_name = "R600"; + break; + case CHIP_RV635: + chip_name = "RV635"; + rlc_chip_name = "R600"; + break; + case CHIP_RV670: + chip_name = "RV670"; + rlc_chip_name = "R600"; + break; case CHIP_RS780: - case CHIP_RS880: chip_name = "RS780"; break; - case CHIP_RV770: chip_name = "RV770"; break; + case CHIP_RS880: + chip_name = "RS780"; + rlc_chip_name = "R600"; + break; + case CHIP_RV770: + chip_name = "RV770"; + rlc_chip_name = "R700"; + break; case CHIP_RV730: - case CHIP_RV740: chip_name = "RV730"; break; - case CHIP_RV710: chip_name = "RV710"; break; + case CHIP_RV740: + chip_name = "RV730"; + rlc_chip_name = "R700"; + break; + case CHIP_RV710: + chip_name = "RV710"; + rlc_chip_name = "R700"; + break; + case CHIP_CEDAR: + chip_name = "CEDAR"; + rlc_chip_name = "CEDAR"; + break; + case CHIP_REDWOOD: + chip_name = "REDWOOD"; + rlc_chip_name = "REDWOOD"; + break; + case CHIP_JUNIPER: + chip_name = "JUNIPER"; + rlc_chip_name = "JUNIPER"; + break; + case CHIP_CYPRESS: + case CHIP_HEMLOCK: + chip_name = "CYPRESS"; + rlc_chip_name = "CYPRESS"; + break; default: BUG(); } - if (rdev->family >= CHIP_RV770) { + if (rdev->family >= CHIP_CEDAR) { + pfp_req_size = EVERGREEN_PFP_UCODE_SIZE * 4; + me_req_size = EVERGREEN_PM4_UCODE_SIZE * 4; + rlc_req_size = EVERGREEN_RLC_UCODE_SIZE * 4; + } else if (rdev->family >= CHIP_RV770) { pfp_req_size = R700_PFP_UCODE_SIZE * 4; me_req_size = R700_PM4_UCODE_SIZE * 4; + rlc_req_size = R700_RLC_UCODE_SIZE * 4; } else { pfp_req_size = PFP_UCODE_SIZE * 4; me_req_size = PM4_UCODE_SIZE * 12; + rlc_req_size = RLC_UCODE_SIZE * 4; } - DRM_INFO("Loading %s CP Microcode\n", chip_name); + DRM_INFO("Loading %s Microcode\n", chip_name); snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name); err = request_firmware(&rdev->pfp_fw, fw_name, &pdev->dev); @@ -1170,6 +2004,18 @@ int r600_cp_init_microcode(struct radeon_device *rdev) rdev->me_fw->size, fw_name); err = -EINVAL; } + + snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", rlc_chip_name); + err = request_firmware(&rdev->rlc_fw, fw_name, &pdev->dev); + if (err) + goto out; + if (rdev->rlc_fw->size != rlc_req_size) { + printk(KERN_ERR + "r600_rlc: Bogus length %zu in firmware \"%s\"\n", + rdev->rlc_fw->size, fw_name); + err = -EINVAL; + } + out: platform_device_unregister(pdev); @@ -1182,6 +2028,8 @@ out: rdev->pfp_fw = NULL; release_firmware(rdev->me_fw); rdev->me_fw = NULL; + release_firmware(rdev->rlc_fw); + rdev->rlc_fw = NULL; } return err; } @@ -1236,12 +2084,15 @@ int r600_cp_start(struct radeon_device *rdev) } radeon_ring_write(rdev, PACKET3(PACKET3_ME_INITIALIZE, 5)); radeon_ring_write(rdev, 0x1); - if (rdev->family < CHIP_RV770) { - radeon_ring_write(rdev, 0x3); - radeon_ring_write(rdev, rdev->config.r600.max_hw_contexts - 1); - } else { + if (rdev->family >= CHIP_CEDAR) { + radeon_ring_write(rdev, 0x0); + radeon_ring_write(rdev, rdev->config.evergreen.max_hw_contexts - 1); + } else if (rdev->family >= CHIP_RV770) { radeon_ring_write(rdev, 0x0); radeon_ring_write(rdev, rdev->config.rv770.max_hw_contexts - 1); + } else { + radeon_ring_write(rdev, 0x3); + radeon_ring_write(rdev, rdev->config.r600.max_hw_contexts - 1); } radeon_ring_write(rdev, PACKET3_ME_INITIALIZE_DEVICE_ID(1)); radeon_ring_write(rdev, 0); @@ -1267,19 +2118,17 @@ int r600_cp_resume(struct radeon_device *rdev) /* Set ring buffer size */ rb_bufsz = drm_order(rdev->cp.ring_size / 8); + tmp = RB_NO_UPDATE | (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz; #ifdef __BIG_ENDIAN - WREG32(CP_RB_CNTL, BUF_SWAP_32BIT | RB_NO_UPDATE | - (drm_order(4096/8) << 8) | rb_bufsz); -#else - WREG32(CP_RB_CNTL, RB_NO_UPDATE | (drm_order(4096/8) << 8) | rb_bufsz); + tmp |= BUF_SWAP_32BIT; #endif + WREG32(CP_RB_CNTL, tmp); WREG32(CP_SEM_WAIT_TIMER, 0x4); /* Set the write pointer delay */ WREG32(CP_RB_WPTR_DELAY, 0); /* Initialize the ring buffer's read and write pointers */ - tmp = RREG32(CP_RB_CNTL); WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA); WREG32(CP_RB_RPTR_WR, 0); WREG32(CP_RB_WPTR, 0); @@ -1321,6 +2170,12 @@ void r600_ring_init(struct radeon_device *rdev, unsigned ring_size) rdev->cp.align_mask = 16 - 1; } +void r600_cp_fini(struct radeon_device *rdev) +{ + r600_cp_stop(rdev); + radeon_ring_fini(rdev); +} + /* * GPU scratch registers helpers function. @@ -1378,10 +2233,16 @@ int r600_ring_test(struct radeon_device *rdev) void r600_wb_disable(struct radeon_device *rdev) { + int r; + WREG32(SCRATCH_UMSK, 0); if (rdev->wb.wb_obj) { - radeon_object_kunmap(rdev->wb.wb_obj); - radeon_object_unpin(rdev->wb.wb_obj); + r = radeon_bo_reserve(rdev->wb.wb_obj, false); + if (unlikely(r != 0)) + return; + radeon_bo_kunmap(rdev->wb.wb_obj); + radeon_bo_unpin(rdev->wb.wb_obj); + radeon_bo_unreserve(rdev->wb.wb_obj); } } @@ -1389,7 +2250,7 @@ void r600_wb_fini(struct radeon_device *rdev) { r600_wb_disable(rdev); if (rdev->wb.wb_obj) { - radeon_object_unref(&rdev->wb.wb_obj); + radeon_bo_unref(&rdev->wb.wb_obj); rdev->wb.wb = NULL; rdev->wb.wb_obj = NULL; } @@ -1400,22 +2261,29 @@ int r600_wb_enable(struct radeon_device *rdev) int r; if (rdev->wb.wb_obj == NULL) { - r = radeon_object_create(rdev, NULL, 4096, true, - RADEON_GEM_DOMAIN_GTT, false, &rdev->wb.wb_obj); + r = radeon_bo_create(rdev, NULL, RADEON_GPU_PAGE_SIZE, true, + RADEON_GEM_DOMAIN_GTT, &rdev->wb.wb_obj); if (r) { - dev_warn(rdev->dev, "failed to create WB buffer (%d).\n", r); + dev_warn(rdev->dev, "(%d) create WB bo failed\n", r); + return r; + } + r = radeon_bo_reserve(rdev->wb.wb_obj, false); + if (unlikely(r != 0)) { + r600_wb_fini(rdev); return r; } - r = radeon_object_pin(rdev->wb.wb_obj, RADEON_GEM_DOMAIN_GTT, + r = radeon_bo_pin(rdev->wb.wb_obj, RADEON_GEM_DOMAIN_GTT, &rdev->wb.gpu_addr); if (r) { - dev_warn(rdev->dev, "failed to pin WB buffer (%d).\n", r); + radeon_bo_unreserve(rdev->wb.wb_obj); + dev_warn(rdev->dev, "(%d) pin WB bo failed\n", r); r600_wb_fini(rdev); return r; } - r = radeon_object_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb); + r = radeon_bo_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb); + radeon_bo_unreserve(rdev->wb.wb_obj); if (r) { - dev_warn(rdev->dev, "failed to map WB buffer (%d).\n", r); + dev_warn(rdev->dev, "(%d) map WB bo failed\n", r); r600_wb_fini(rdev); return r; } @@ -1430,41 +2298,41 @@ int r600_wb_enable(struct radeon_device *rdev) void r600_fence_ring_emit(struct radeon_device *rdev, struct radeon_fence *fence) { + /* Also consider EVENT_WRITE_EOP. it handles the interrupts + timestamps + events */ + + radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0)); + radeon_ring_write(rdev, CACHE_FLUSH_AND_INV_EVENT); + /* wait for 3D idle clean */ + radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1)); + radeon_ring_write(rdev, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); + radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit); /* Emit fence sequence & fire IRQ */ radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(rdev, ((rdev->fence_drv.scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2)); radeon_ring_write(rdev, fence->seq); -} - -int r600_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, - uint64_t dst_offset, - unsigned num_pages, - struct radeon_fence *fence) -{ - /* FIXME: implement */ - return 0; + /* CP_INTERRUPT packet 3 no longer exists, use packet 0 */ + radeon_ring_write(rdev, PACKET0(CP_INT_STATUS, 0)); + radeon_ring_write(rdev, RB_INT_STAT); } int r600_copy_blit(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_pages, struct radeon_fence *fence) { - r600_blit_prepare_copy(rdev, num_pages * 4096); - r600_kms_blit_copy(rdev, src_offset, dst_offset, num_pages * 4096); - r600_blit_done_copy(rdev, fence); - return 0; -} - -int r600_irq_process(struct radeon_device *rdev) -{ - /* FIXME: implement */ - return 0; -} + int r; -int r600_irq_set(struct radeon_device *rdev) -{ - /* FIXME: implement */ + mutex_lock(&rdev->r600_blit.mutex); + rdev->r600_blit.vb_ib = NULL; + r = r600_blit_prepare_copy(rdev, num_pages * RADEON_GPU_PAGE_SIZE); + if (r) { + if (rdev->r600_blit.vb_ib) + radeon_ib_free(rdev, &rdev->r600_blit.vb_ib); + mutex_unlock(&rdev->r600_blit.mutex); + return r; + } + r600_kms_blit_copy(rdev, src_offset, dst_offset, num_pages * RADEON_GPU_PAGE_SIZE); + r600_blit_done_copy(rdev, fence); + mutex_unlock(&rdev->r600_blit.mutex); return 0; } @@ -1503,6 +2371,14 @@ int r600_startup(struct radeon_device *rdev) { int r; + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { + r = r600_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + r600_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { r600_agp_enable(rdev); @@ -1512,13 +2388,33 @@ int r600_startup(struct radeon_device *rdev) return r; } r600_gpu_init(rdev); - - r = radeon_object_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->r600_blit.shader_gpu_addr); + r = r600_blit_init(rdev); + if (r) { + r600_blit_fini(rdev); + rdev->asic->copy = NULL; + dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); + } + /* pin copy shader into vram */ + if (rdev->r600_blit.shader_obj) { + r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); + if (unlikely(r != 0)) + return r; + r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM, + &rdev->r600_blit.shader_gpu_addr); + radeon_bo_unreserve(rdev->r600_blit.shader_obj); + if (r) { + dev_err(rdev->dev, "(%d) pin blit object failed\n", r); + return r; + } + } + /* Enable IRQ */ + r = r600_irq_init(rdev); if (r) { - DRM_ERROR("failed to pin blit object %d\n", r); + DRM_ERROR("radeon: IH init failed (%d).\n", r); + radeon_irq_kms_fini(rdev); return r; } + r600_irq_set(rdev); r = radeon_ring_init(rdev, rdev->cp.ring_size); if (r) @@ -1575,18 +2471,35 @@ int r600_resume(struct radeon_device *rdev) DRM_ERROR("radeon: failled testing IB (%d).\n", r); return r; } + + r = r600_audio_init(rdev); + if (r) { + DRM_ERROR("radeon: audio resume failed\n"); + return r; + } + return r; } int r600_suspend(struct radeon_device *rdev) { + int r; + + r600_audio_fini(rdev); /* FIXME: we should wait for ring to be empty */ r600_cp_stop(rdev); rdev->cp.ready = false; + r600_irq_suspend(rdev); r600_wb_disable(rdev); r600_pcie_gart_disable(rdev); /* unpin shaders bo */ - radeon_object_unpin(rdev->r600_blit.shader_obj); + if (rdev->r600_blit.shader_obj) { + r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); + if (!r) { + radeon_bo_unpin(rdev->r600_blit.shader_obj); + radeon_bo_unreserve(rdev->r600_blit.shader_obj); + } + } return 0; } @@ -1624,7 +2537,11 @@ int r600_init(struct radeon_device *rdev) if (r) return r; /* Post card if necessary */ - if (!r600_card_posted(rdev) && rdev->bios) { + if (!r600_card_posted(rdev)) { + if (!rdev->bios) { + dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n"); + return -EINVAL; + } DRM_INFO("GPU not posted. posting now...\n"); atom_asic_init(rdev->mode_info.atom_context); } @@ -1632,6 +2549,7 @@ int r600_init(struct radeon_device *rdev) r600_scratch_init(rdev); /* Initialize surface registers */ radeon_surface_init(rdev); + /* Initialize clocks */ radeon_get_clock_info(rdev->ddev); r = radeon_clocks_init(rdev); if (r) @@ -1640,73 +2558,78 @@ int r600_init(struct radeon_device *rdev) r = radeon_fence_driver_init(rdev); if (r) return r; + if (rdev->flags & RADEON_IS_AGP) { + r = radeon_agp_init(rdev); + if (r) + radeon_agp_disable(rdev); + } r = r600_mc_init(rdev); if (r) return r; /* Memory manager */ - r = radeon_object_init(rdev); + r = radeon_bo_init(rdev); if (r) return r; + + r = radeon_irq_kms_init(rdev); + if (r) + return r; + rdev->cp.ring_obj = NULL; r600_ring_init(rdev, 1024 * 1024); - if (!rdev->me_fw || !rdev->pfp_fw) { - r = r600_cp_init_microcode(rdev); - if (r) { - DRM_ERROR("Failed to load firmware!\n"); - return r; - } - } + rdev->ih.ring_obj = NULL; + r600_ih_ring_init(rdev, 64 * 1024); r = r600_pcie_gart_init(rdev); if (r) return r; rdev->accel_working = true; - r = r600_blit_init(rdev); - if (r) { - DRM_ERROR("radeon: failled blitter (%d).\n", r); - return r; - } - r = r600_startup(rdev); if (r) { - r600_suspend(rdev); + dev_err(rdev->dev, "disabling GPU acceleration\n"); + r600_cp_fini(rdev); r600_wb_fini(rdev); - radeon_ring_fini(rdev); + r600_irq_fini(rdev); + radeon_irq_kms_fini(rdev); r600_pcie_gart_fini(rdev); rdev->accel_working = false; } if (rdev->accel_working) { r = radeon_ib_pool_init(rdev); if (r) { - DRM_ERROR("radeon: failled initializing IB pool (%d).\n", r); - rdev->accel_working = false; - } - r = r600_ib_test(rdev); - if (r) { - DRM_ERROR("radeon: failled testing IB (%d).\n", r); + dev_err(rdev->dev, "IB initialization failed (%d).\n", r); rdev->accel_working = false; + } else { + r = r600_ib_test(rdev); + if (r) { + dev_err(rdev->dev, "IB test failed (%d).\n", r); + rdev->accel_working = false; + } } } + + r = r600_audio_init(rdev); + if (r) + return r; /* TODO error handling */ return 0; } void r600_fini(struct radeon_device *rdev) { - /* Suspend operations */ - r600_suspend(rdev); - + r600_audio_fini(rdev); r600_blit_fini(rdev); - radeon_ring_fini(rdev); + r600_cp_fini(rdev); r600_wb_fini(rdev); + r600_irq_fini(rdev); + radeon_irq_kms_fini(rdev); r600_pcie_gart_fini(rdev); + radeon_agp_fini(rdev); radeon_gem_fini(rdev); radeon_fence_driver_fini(rdev); radeon_clocks_fini(rdev); - if (rdev->flags & RADEON_IS_AGP) - radeon_agp_fini(rdev); - radeon_object_fini(rdev); + radeon_bo_fini(rdev); radeon_atombios_fini(rdev); kfree(rdev->bios); rdev->bios = NULL; @@ -1792,8 +2715,725 @@ int r600_ib_test(struct radeon_device *rdev) return r; } +/* + * Interrupts + * + * Interrupts use a ring buffer on r6xx/r7xx hardware. It works pretty + * the same as the CP ring buffer, but in reverse. Rather than the CPU + * writing to the ring and the GPU consuming, the GPU writes to the ring + * and host consumes. As the host irq handler processes interrupts, it + * increments the rptr. When the rptr catches up with the wptr, all the + * current interrupts have been processed. + */ + +void r600_ih_ring_init(struct radeon_device *rdev, unsigned ring_size) +{ + u32 rb_bufsz; + /* Align ring size */ + rb_bufsz = drm_order(ring_size / 4); + ring_size = (1 << rb_bufsz) * 4; + rdev->ih.ring_size = ring_size; + rdev->ih.ptr_mask = rdev->ih.ring_size - 1; + rdev->ih.rptr = 0; +} +static int r600_ih_ring_alloc(struct radeon_device *rdev) +{ + int r; + + /* Allocate ring buffer */ + if (rdev->ih.ring_obj == NULL) { + r = radeon_bo_create(rdev, NULL, rdev->ih.ring_size, + true, + RADEON_GEM_DOMAIN_GTT, + &rdev->ih.ring_obj); + if (r) { + DRM_ERROR("radeon: failed to create ih ring buffer (%d).\n", r); + return r; + } + r = radeon_bo_reserve(rdev->ih.ring_obj, false); + if (unlikely(r != 0)) + return r; + r = radeon_bo_pin(rdev->ih.ring_obj, + RADEON_GEM_DOMAIN_GTT, + &rdev->ih.gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->ih.ring_obj); + DRM_ERROR("radeon: failed to pin ih ring buffer (%d).\n", r); + return r; + } + r = radeon_bo_kmap(rdev->ih.ring_obj, + (void **)&rdev->ih.ring); + radeon_bo_unreserve(rdev->ih.ring_obj); + if (r) { + DRM_ERROR("radeon: failed to map ih ring buffer (%d).\n", r); + return r; + } + } + return 0; +} + +static void r600_ih_ring_fini(struct radeon_device *rdev) +{ + int r; + if (rdev->ih.ring_obj) { + r = radeon_bo_reserve(rdev->ih.ring_obj, false); + if (likely(r == 0)) { + radeon_bo_kunmap(rdev->ih.ring_obj); + radeon_bo_unpin(rdev->ih.ring_obj); + radeon_bo_unreserve(rdev->ih.ring_obj); + } + radeon_bo_unref(&rdev->ih.ring_obj); + rdev->ih.ring = NULL; + rdev->ih.ring_obj = NULL; + } +} + +void r600_rlc_stop(struct radeon_device *rdev) +{ + + if ((rdev->family >= CHIP_RV770) && + (rdev->family <= CHIP_RV740)) { + /* r7xx asics need to soft reset RLC before halting */ + WREG32(SRBM_SOFT_RESET, SOFT_RESET_RLC); + RREG32(SRBM_SOFT_RESET); + udelay(15000); + WREG32(SRBM_SOFT_RESET, 0); + RREG32(SRBM_SOFT_RESET); + } + + WREG32(RLC_CNTL, 0); +} + +static void r600_rlc_start(struct radeon_device *rdev) +{ + WREG32(RLC_CNTL, RLC_ENABLE); +} + +static int r600_rlc_init(struct radeon_device *rdev) +{ + u32 i; + const __be32 *fw_data; + + if (!rdev->rlc_fw) + return -EINVAL; + + r600_rlc_stop(rdev); + + WREG32(RLC_HB_BASE, 0); + WREG32(RLC_HB_CNTL, 0); + WREG32(RLC_HB_RPTR, 0); + WREG32(RLC_HB_WPTR, 0); + WREG32(RLC_HB_WPTR_LSB_ADDR, 0); + WREG32(RLC_HB_WPTR_MSB_ADDR, 0); + WREG32(RLC_MC_CNTL, 0); + WREG32(RLC_UCODE_CNTL, 0); + + fw_data = (const __be32 *)rdev->rlc_fw->data; + if (rdev->family >= CHIP_CEDAR) { + for (i = 0; i < EVERGREEN_RLC_UCODE_SIZE; i++) { + WREG32(RLC_UCODE_ADDR, i); + WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); + } + } else if (rdev->family >= CHIP_RV770) { + for (i = 0; i < R700_RLC_UCODE_SIZE; i++) { + WREG32(RLC_UCODE_ADDR, i); + WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); + } + } else { + for (i = 0; i < RLC_UCODE_SIZE; i++) { + WREG32(RLC_UCODE_ADDR, i); + WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); + } + } + WREG32(RLC_UCODE_ADDR, 0); + + r600_rlc_start(rdev); + + return 0; +} + +static void r600_enable_interrupts(struct radeon_device *rdev) +{ + u32 ih_cntl = RREG32(IH_CNTL); + u32 ih_rb_cntl = RREG32(IH_RB_CNTL); + + ih_cntl |= ENABLE_INTR; + ih_rb_cntl |= IH_RB_ENABLE; + WREG32(IH_CNTL, ih_cntl); + WREG32(IH_RB_CNTL, ih_rb_cntl); + rdev->ih.enabled = true; +} + +void r600_disable_interrupts(struct radeon_device *rdev) +{ + u32 ih_rb_cntl = RREG32(IH_RB_CNTL); + u32 ih_cntl = RREG32(IH_CNTL); + + ih_rb_cntl &= ~IH_RB_ENABLE; + ih_cntl &= ~ENABLE_INTR; + WREG32(IH_RB_CNTL, ih_rb_cntl); + WREG32(IH_CNTL, ih_cntl); + /* set rptr, wptr to 0 */ + WREG32(IH_RB_RPTR, 0); + WREG32(IH_RB_WPTR, 0); + rdev->ih.enabled = false; + rdev->ih.wptr = 0; + rdev->ih.rptr = 0; +} + +static void r600_disable_interrupt_state(struct radeon_device *rdev) +{ + u32 tmp; + + WREG32(CP_INT_CNTL, 0); + WREG32(GRBM_INT_CNTL, 0); + WREG32(DxMODE_INT_MASK, 0); + if (ASIC_IS_DCE3(rdev)) { + WREG32(DCE3_DACA_AUTODETECT_INT_CONTROL, 0); + WREG32(DCE3_DACB_AUTODETECT_INT_CONTROL, 0); + tmp = RREG32(DC_HPD1_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD1_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD2_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD2_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD3_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD3_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD4_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD4_INT_CONTROL, tmp); + if (ASIC_IS_DCE32(rdev)) { + tmp = RREG32(DC_HPD5_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD5_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD6_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD6_INT_CONTROL, tmp); + } + } else { + WREG32(DACA_AUTODETECT_INT_CONTROL, 0); + WREG32(DACB_AUTODETECT_INT_CONTROL, 0); + tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY; + WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp); + tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY; + WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp); + tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY; + WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp); + } +} + +int r600_irq_init(struct radeon_device *rdev) +{ + int ret = 0; + int rb_bufsz; + u32 interrupt_cntl, ih_cntl, ih_rb_cntl; + + /* allocate ring */ + ret = r600_ih_ring_alloc(rdev); + if (ret) + return ret; + + /* disable irqs */ + r600_disable_interrupts(rdev); + + /* init rlc */ + ret = r600_rlc_init(rdev); + if (ret) { + r600_ih_ring_fini(rdev); + return ret; + } + + /* setup interrupt control */ + /* set dummy read address to ring address */ + WREG32(INTERRUPT_CNTL2, rdev->ih.gpu_addr >> 8); + interrupt_cntl = RREG32(INTERRUPT_CNTL); + /* IH_DUMMY_RD_OVERRIDE=0 - dummy read disabled with msi, enabled without msi + * IH_DUMMY_RD_OVERRIDE=1 - dummy read controlled by IH_DUMMY_RD_EN + */ + interrupt_cntl &= ~IH_DUMMY_RD_OVERRIDE; + /* IH_REQ_NONSNOOP_EN=1 if ring is in non-cacheable memory, e.g., vram */ + interrupt_cntl &= ~IH_REQ_NONSNOOP_EN; + WREG32(INTERRUPT_CNTL, interrupt_cntl); + + WREG32(IH_RB_BASE, rdev->ih.gpu_addr >> 8); + rb_bufsz = drm_order(rdev->ih.ring_size / 4); + + ih_rb_cntl = (IH_WPTR_OVERFLOW_ENABLE | + IH_WPTR_OVERFLOW_CLEAR | + (rb_bufsz << 1)); + /* WPTR writeback, not yet */ + /*ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE;*/ + WREG32(IH_RB_WPTR_ADDR_LO, 0); + WREG32(IH_RB_WPTR_ADDR_HI, 0); + + WREG32(IH_RB_CNTL, ih_rb_cntl); + + /* set rptr, wptr to 0 */ + WREG32(IH_RB_RPTR, 0); + WREG32(IH_RB_WPTR, 0); + + /* Default settings for IH_CNTL (disabled at first) */ + ih_cntl = MC_WRREQ_CREDIT(0x10) | MC_WR_CLEAN_CNT(0x10); + /* RPTR_REARM only works if msi's are enabled */ + if (rdev->msi_enabled) + ih_cntl |= RPTR_REARM; + +#ifdef __BIG_ENDIAN + ih_cntl |= IH_MC_SWAP(IH_MC_SWAP_32BIT); +#endif + WREG32(IH_CNTL, ih_cntl); + + /* force the active interrupt state to all disabled */ + if (rdev->family >= CHIP_CEDAR) + evergreen_disable_interrupt_state(rdev); + else + r600_disable_interrupt_state(rdev); + + /* enable irqs */ + r600_enable_interrupts(rdev); + + return ret; +} + +void r600_irq_suspend(struct radeon_device *rdev) +{ + r600_irq_disable(rdev); + r600_rlc_stop(rdev); +} + +void r600_irq_fini(struct radeon_device *rdev) +{ + r600_irq_suspend(rdev); + r600_ih_ring_fini(rdev); +} + +int r600_irq_set(struct radeon_device *rdev) +{ + u32 cp_int_cntl = CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE; + u32 mode_int = 0; + u32 hpd1, hpd2, hpd3, hpd4 = 0, hpd5 = 0, hpd6 = 0; + u32 grbm_int_cntl = 0; + u32 hdmi1, hdmi2; + + if (!rdev->irq.installed) { + WARN(1, "Can't enable IRQ/MSI because no handler is installed.\n"); + return -EINVAL; + } + /* don't enable anything if the ih is disabled */ + if (!rdev->ih.enabled) { + r600_disable_interrupts(rdev); + /* force the active interrupt state to all disabled */ + r600_disable_interrupt_state(rdev); + return 0; + } + + hdmi1 = RREG32(R600_HDMI_BLOCK1 + R600_HDMI_CNTL) & ~R600_HDMI_INT_EN; + if (ASIC_IS_DCE3(rdev)) { + hdmi2 = RREG32(R600_HDMI_BLOCK3 + R600_HDMI_CNTL) & ~R600_HDMI_INT_EN; + hpd1 = RREG32(DC_HPD1_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd2 = RREG32(DC_HPD2_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd3 = RREG32(DC_HPD3_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd4 = RREG32(DC_HPD4_INT_CONTROL) & ~DC_HPDx_INT_EN; + if (ASIC_IS_DCE32(rdev)) { + hpd5 = RREG32(DC_HPD5_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~DC_HPDx_INT_EN; + } + } else { + hdmi2 = RREG32(R600_HDMI_BLOCK2 + R600_HDMI_CNTL) & ~R600_HDMI_INT_EN; + hpd1 = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd2 = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd3 = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL) & ~DC_HPDx_INT_EN; + } + + if (rdev->irq.sw_int) { + DRM_DEBUG("r600_irq_set: sw int\n"); + cp_int_cntl |= RB_INT_ENABLE; + } + if (rdev->irq.crtc_vblank_int[0]) { + DRM_DEBUG("r600_irq_set: vblank 0\n"); + mode_int |= D1MODE_VBLANK_INT_MASK; + } + if (rdev->irq.crtc_vblank_int[1]) { + DRM_DEBUG("r600_irq_set: vblank 1\n"); + mode_int |= D2MODE_VBLANK_INT_MASK; + } + if (rdev->irq.hpd[0]) { + DRM_DEBUG("r600_irq_set: hpd 1\n"); + hpd1 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[1]) { + DRM_DEBUG("r600_irq_set: hpd 2\n"); + hpd2 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[2]) { + DRM_DEBUG("r600_irq_set: hpd 3\n"); + hpd3 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[3]) { + DRM_DEBUG("r600_irq_set: hpd 4\n"); + hpd4 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[4]) { + DRM_DEBUG("r600_irq_set: hpd 5\n"); + hpd5 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[5]) { + DRM_DEBUG("r600_irq_set: hpd 6\n"); + hpd6 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hdmi[0]) { + DRM_DEBUG("r600_irq_set: hdmi 1\n"); + hdmi1 |= R600_HDMI_INT_EN; + } + if (rdev->irq.hdmi[1]) { + DRM_DEBUG("r600_irq_set: hdmi 2\n"); + hdmi2 |= R600_HDMI_INT_EN; + } + if (rdev->irq.gui_idle) { + DRM_DEBUG("gui idle\n"); + grbm_int_cntl |= GUI_IDLE_INT_ENABLE; + } + + WREG32(CP_INT_CNTL, cp_int_cntl); + WREG32(DxMODE_INT_MASK, mode_int); + WREG32(GRBM_INT_CNTL, grbm_int_cntl); + WREG32(R600_HDMI_BLOCK1 + R600_HDMI_CNTL, hdmi1); + if (ASIC_IS_DCE3(rdev)) { + WREG32(R600_HDMI_BLOCK3 + R600_HDMI_CNTL, hdmi2); + WREG32(DC_HPD1_INT_CONTROL, hpd1); + WREG32(DC_HPD2_INT_CONTROL, hpd2); + WREG32(DC_HPD3_INT_CONTROL, hpd3); + WREG32(DC_HPD4_INT_CONTROL, hpd4); + if (ASIC_IS_DCE32(rdev)) { + WREG32(DC_HPD5_INT_CONTROL, hpd5); + WREG32(DC_HPD6_INT_CONTROL, hpd6); + } + } else { + WREG32(R600_HDMI_BLOCK2 + R600_HDMI_CNTL, hdmi2); + WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, hpd1); + WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, hpd2); + WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, hpd3); + } + + return 0; +} + +static inline void r600_irq_ack(struct radeon_device *rdev, + u32 *disp_int, + u32 *disp_int_cont, + u32 *disp_int_cont2) +{ + u32 tmp; + + if (ASIC_IS_DCE3(rdev)) { + *disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS); + *disp_int_cont = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE); + *disp_int_cont2 = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE2); + } else { + *disp_int = RREG32(DISP_INTERRUPT_STATUS); + *disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE); + *disp_int_cont2 = 0; + } + + if (*disp_int & LB_D1_VBLANK_INTERRUPT) + WREG32(D1MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK); + if (*disp_int & LB_D1_VLINE_INTERRUPT) + WREG32(D1MODE_VLINE_STATUS, DxMODE_VLINE_ACK); + if (*disp_int & LB_D2_VBLANK_INTERRUPT) + WREG32(D2MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK); + if (*disp_int & LB_D2_VLINE_INTERRUPT) + WREG32(D2MODE_VLINE_STATUS, DxMODE_VLINE_ACK); + if (*disp_int & DC_HPD1_INTERRUPT) { + if (ASIC_IS_DCE3(rdev)) { + tmp = RREG32(DC_HPD1_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD1_INT_CONTROL, tmp); + } else { + tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp); + } + } + if (*disp_int & DC_HPD2_INTERRUPT) { + if (ASIC_IS_DCE3(rdev)) { + tmp = RREG32(DC_HPD2_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD2_INT_CONTROL, tmp); + } else { + tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp); + } + } + if (*disp_int_cont & DC_HPD3_INTERRUPT) { + if (ASIC_IS_DCE3(rdev)) { + tmp = RREG32(DC_HPD3_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD3_INT_CONTROL, tmp); + } else { + tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp); + } + } + if (*disp_int_cont & DC_HPD4_INTERRUPT) { + tmp = RREG32(DC_HPD4_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD4_INT_CONTROL, tmp); + } + if (ASIC_IS_DCE32(rdev)) { + if (*disp_int_cont2 & DC_HPD5_INTERRUPT) { + tmp = RREG32(DC_HPD5_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD5_INT_CONTROL, tmp); + } + if (*disp_int_cont2 & DC_HPD6_INTERRUPT) { + tmp = RREG32(DC_HPD5_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD6_INT_CONTROL, tmp); + } + } + if (RREG32(R600_HDMI_BLOCK1 + R600_HDMI_STATUS) & R600_HDMI_INT_PENDING) { + WREG32_P(R600_HDMI_BLOCK1 + R600_HDMI_CNTL, R600_HDMI_INT_ACK, ~R600_HDMI_INT_ACK); + } + if (ASIC_IS_DCE3(rdev)) { + if (RREG32(R600_HDMI_BLOCK3 + R600_HDMI_STATUS) & R600_HDMI_INT_PENDING) { + WREG32_P(R600_HDMI_BLOCK3 + R600_HDMI_CNTL, R600_HDMI_INT_ACK, ~R600_HDMI_INT_ACK); + } + } else { + if (RREG32(R600_HDMI_BLOCK2 + R600_HDMI_STATUS) & R600_HDMI_INT_PENDING) { + WREG32_P(R600_HDMI_BLOCK2 + R600_HDMI_CNTL, R600_HDMI_INT_ACK, ~R600_HDMI_INT_ACK); + } + } +} + +void r600_irq_disable(struct radeon_device *rdev) +{ + u32 disp_int, disp_int_cont, disp_int_cont2; + + r600_disable_interrupts(rdev); + /* Wait and acknowledge irq */ + mdelay(1); + r600_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2); + r600_disable_interrupt_state(rdev); +} + +static inline u32 r600_get_ih_wptr(struct radeon_device *rdev) +{ + u32 wptr, tmp; + + /* XXX use writeback */ + wptr = RREG32(IH_RB_WPTR); + + if (wptr & RB_OVERFLOW) { + /* When a ring buffer overflow happen start parsing interrupt + * from the last not overwritten vector (wptr + 16). Hopefully + * this should allow us to catchup. + */ + dev_warn(rdev->dev, "IH ring buffer overflow (0x%08X, %d, %d)\n", + wptr, rdev->ih.rptr, (wptr + 16) + rdev->ih.ptr_mask); + rdev->ih.rptr = (wptr + 16) & rdev->ih.ptr_mask; + tmp = RREG32(IH_RB_CNTL); + tmp |= IH_WPTR_OVERFLOW_CLEAR; + WREG32(IH_RB_CNTL, tmp); + } + return (wptr & rdev->ih.ptr_mask); +} + +/* r600 IV Ring + * Each IV ring entry is 128 bits: + * [7:0] - interrupt source id + * [31:8] - reserved + * [59:32] - interrupt source data + * [127:60] - reserved + * + * The basic interrupt vector entries + * are decoded as follows: + * src_id src_data description + * 1 0 D1 Vblank + * 1 1 D1 Vline + * 5 0 D2 Vblank + * 5 1 D2 Vline + * 19 0 FP Hot plug detection A + * 19 1 FP Hot plug detection B + * 19 2 DAC A auto-detection + * 19 3 DAC B auto-detection + * 21 4 HDMI block A + * 21 5 HDMI block B + * 176 - CP_INT RB + * 177 - CP_INT IB1 + * 178 - CP_INT IB2 + * 181 - EOP Interrupt + * 233 - GUI Idle + * + * Note, these are based on r600 and may need to be + * adjusted or added to on newer asics + */ + +int r600_irq_process(struct radeon_device *rdev) +{ + u32 wptr = r600_get_ih_wptr(rdev); + u32 rptr = rdev->ih.rptr; + u32 src_id, src_data; + u32 ring_index, disp_int, disp_int_cont, disp_int_cont2; + unsigned long flags; + bool queue_hotplug = false; + + DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr); + if (!rdev->ih.enabled) + return IRQ_NONE; + + spin_lock_irqsave(&rdev->ih.lock, flags); + + if (rptr == wptr) { + spin_unlock_irqrestore(&rdev->ih.lock, flags); + return IRQ_NONE; + } + if (rdev->shutdown) { + spin_unlock_irqrestore(&rdev->ih.lock, flags); + return IRQ_NONE; + } + +restart_ih: + /* display interrupts */ + r600_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2); + + rdev->ih.wptr = wptr; + while (rptr != wptr) { + /* wptr/rptr are in bytes! */ + ring_index = rptr / 4; + src_id = rdev->ih.ring[ring_index] & 0xff; + src_data = rdev->ih.ring[ring_index + 1] & 0xfffffff; + + switch (src_id) { + case 1: /* D1 vblank/vline */ + switch (src_data) { + case 0: /* D1 vblank */ + if (disp_int & LB_D1_VBLANK_INTERRUPT) { + drm_handle_vblank(rdev->ddev, 0); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + disp_int &= ~LB_D1_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D1 vblank\n"); + } + break; + case 1: /* D1 vline */ + if (disp_int & LB_D1_VLINE_INTERRUPT) { + disp_int &= ~LB_D1_VLINE_INTERRUPT; + DRM_DEBUG("IH: D1 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 5: /* D2 vblank/vline */ + switch (src_data) { + case 0: /* D2 vblank */ + if (disp_int & LB_D2_VBLANK_INTERRUPT) { + drm_handle_vblank(rdev->ddev, 1); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + disp_int &= ~LB_D2_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D2 vblank\n"); + } + break; + case 1: /* D1 vline */ + if (disp_int & LB_D2_VLINE_INTERRUPT) { + disp_int &= ~LB_D2_VLINE_INTERRUPT; + DRM_DEBUG("IH: D2 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 19: /* HPD/DAC hotplug */ + switch (src_data) { + case 0: + if (disp_int & DC_HPD1_INTERRUPT) { + disp_int &= ~DC_HPD1_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD1\n"); + } + break; + case 1: + if (disp_int & DC_HPD2_INTERRUPT) { + disp_int &= ~DC_HPD2_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD2\n"); + } + break; + case 4: + if (disp_int_cont & DC_HPD3_INTERRUPT) { + disp_int_cont &= ~DC_HPD3_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD3\n"); + } + break; + case 5: + if (disp_int_cont & DC_HPD4_INTERRUPT) { + disp_int_cont &= ~DC_HPD4_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD4\n"); + } + break; + case 10: + if (disp_int_cont2 & DC_HPD5_INTERRUPT) { + disp_int_cont2 &= ~DC_HPD5_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD5\n"); + } + break; + case 12: + if (disp_int_cont2 & DC_HPD6_INTERRUPT) { + disp_int_cont2 &= ~DC_HPD6_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD6\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 21: /* HDMI */ + DRM_DEBUG("IH: HDMI: 0x%x\n", src_data); + r600_audio_schedule_polling(rdev); + break; + case 176: /* CP_INT in ring buffer */ + case 177: /* CP_INT in IB1 */ + case 178: /* CP_INT in IB2 */ + DRM_DEBUG("IH: CP int: 0x%08x\n", src_data); + radeon_fence_process(rdev); + break; + case 181: /* CP EOP event */ + DRM_DEBUG("IH: CP EOP\n"); + break; + case 233: /* GUI IDLE */ + DRM_DEBUG("IH: CP EOP\n"); + rdev->pm.gui_idle = true; + wake_up(&rdev->irq.idle_queue); + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + + /* wptr/rptr are in bytes! */ + rptr += 16; + rptr &= rdev->ih.ptr_mask; + } + /* make sure wptr hasn't changed while processing */ + wptr = r600_get_ih_wptr(rdev); + if (wptr != rdev->ih.wptr) + goto restart_ih; + if (queue_hotplug) + queue_work(rdev->wq, &rdev->hotplug_work); + rdev->ih.rptr = rptr; + WREG32(IH_RB_RPTR, rdev->ih.rptr); + spin_unlock_irqrestore(&rdev->ih.lock, flags); + return IRQ_HANDLED; +} /* * Debugfs info @@ -1805,21 +3445,21 @@ static int r600_debugfs_cp_ring_info(struct seq_file *m, void *data) struct drm_info_node *node = (struct drm_info_node *) m->private; struct drm_device *dev = node->minor->dev; struct radeon_device *rdev = dev->dev_private; - uint32_t rdp, wdp; unsigned count, i, j; radeon_ring_free_size(rdev); - rdp = RREG32(CP_RB_RPTR); - wdp = RREG32(CP_RB_WPTR); - count = (rdp + rdev->cp.ring_size - wdp) & rdev->cp.ptr_mask; + count = (rdev->cp.ring_size / 4) - rdev->cp.ring_free_dw; seq_printf(m, "CP_STAT 0x%08x\n", RREG32(CP_STAT)); - seq_printf(m, "CP_RB_WPTR 0x%08x\n", wdp); - seq_printf(m, "CP_RB_RPTR 0x%08x\n", rdp); + seq_printf(m, "CP_RB_WPTR 0x%08x\n", RREG32(CP_RB_WPTR)); + seq_printf(m, "CP_RB_RPTR 0x%08x\n", RREG32(CP_RB_RPTR)); + seq_printf(m, "driver's copy of the CP_RB_WPTR 0x%08x\n", rdev->cp.wptr); + seq_printf(m, "driver's copy of the CP_RB_RPTR 0x%08x\n", rdev->cp.rptr); seq_printf(m, "%u free dwords in ring\n", rdev->cp.ring_free_dw); seq_printf(m, "%u dwords in ring\n", count); + i = rdev->cp.rptr; for (j = 0; j <= count; j++) { - i = (rdp + j) & rdev->cp.ptr_mask; seq_printf(m, "r[%04d]=0x%08x\n", i, rdev->cp.ring[i]); + i = (i + 1) & rdev->cp.ptr_mask; } return 0; } @@ -1849,3 +3489,18 @@ int r600_debugfs_mc_info_init(struct radeon_device *rdev) return 0; #endif } + +/** + * r600_ioctl_wait_idle - flush host path cache on wait idle ioctl + * rdev: radeon device structure + * bo: buffer object struct which userspace is waiting for idle + * + * Some R6XX/R7XX doesn't seems to take into account HDP flush performed + * through ring buffer, this leads to corruption in rendering, see + * http://bugzilla.kernel.org/show_bug.cgi?id=15186 to avoid this we + * directly perform HDP flush by writing register through MMIO. + */ +void r600_ioctl_wait_idle(struct radeon_device *rdev, struct radeon_bo *bo) +{ + WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1); +}