From d8f60cfc93452d0554f6a701aa8e3236cbee4636 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 1 Dec 2009 13:43:46 -0500 Subject: [PATCH] drm/radeon/kms: Add support for interrupts on r6xx/r7xx chips (v3) This enables the use of interrupts on r6xx/r7xx hardware. Interrupts are implemented via a ring buffer. The GPU adds interrupts vectors to the ring and the host reads them off in the interrupt handler. The interrupt controller requires firmware like the CP. This firmware must be installed and accessble to the firmware loader for interrupts to function. MSIs don't seem to work on my RS780. They work fine on all my discrete cards. I'm not sure about other RS780s or RS880s. I've disabled MSIs on RS780 and RS880, but it would probably be worth checking on some other systems. v2 - fix some checkpatch.pl problems; re-read the disp int status reg if we restart the ih; v3 - remove the irq handler if r600_irq_init() fails; remove spinlock in r600_ih_ring_fini(); move ih rb overflow check to r600_get_ih_wptr(); move irq ack to separate function; Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie --- drivers/gpu/drm/radeon/r600.c | 561 ++++++++++++++++++++++++++++++-- drivers/gpu/drm/radeon/r600_blit_kms.c | 4 +- drivers/gpu/drm/radeon/r600d.h | 159 ++++++++- drivers/gpu/drm/radeon/radeon.h | 26 +- drivers/gpu/drm/radeon/radeon_asic.h | 2 + drivers/gpu/drm/radeon/radeon_device.c | 2 + drivers/gpu/drm/radeon/radeon_drv.h | 1 - drivers/gpu/drm/radeon/radeon_fence.c | 38 --- drivers/gpu/drm/radeon/radeon_irq_kms.c | 12 +- drivers/gpu/drm/radeon/rv770.c | 24 +- 10 files changed, 754 insertions(+), 75 deletions(-) diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 3dbd93e..5067ab7 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -38,8 +38,10 @@ #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 /* Firmware Names */ MODULE_FIRMWARE("radeon/R600_pfp.bin"); @@ -62,6 +64,8 @@ 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"); int r600_debugfs_mc_info_init(struct radeon_device *rdev); @@ -1114,11 +1118,12 @@ 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; @@ -1132,30 +1137,62 @@ 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; default: BUG(); } 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); @@ -1179,6 +1216,18 @@ 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); @@ -1191,6 +1240,8 @@ 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; } @@ -1437,10 +1488,14 @@ 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 */ /* 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); + /* 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_dma(struct radeon_device *rdev, @@ -1463,18 +1518,6 @@ int r600_copy_blit(struct radeon_device *rdev, return 0; } -int r600_irq_process(struct radeon_device *rdev) -{ - /* FIXME: implement */ - return 0; -} - -int r600_irq_set(struct radeon_device *rdev) -{ - /* FIXME: implement */ - return 0; -} - int r600_set_surface_reg(struct radeon_device *rdev, int reg, uint32_t tiling_flags, uint32_t pitch, uint32_t offset, uint32_t obj_size) @@ -1527,6 +1570,16 @@ int r600_startup(struct radeon_device *rdev) return r; } + /* Enable IRQ */ + rdev->irq.sw_int = true; + r = r600_irq_init(rdev); + if (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) return r; @@ -1661,11 +1714,19 @@ int r600_init(struct radeon_device *rdev) r = radeon_object_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); + rdev->ih.ring_obj = NULL; + r600_ih_ring_init(rdev, 64 * 1024); + + 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; @@ -1712,6 +1773,8 @@ void r600_fini(struct radeon_device *rdev) r600_suspend(rdev); r600_blit_fini(rdev); + r600_irq_fini(rdev); + radeon_irq_kms_fini(rdev); radeon_ring_fini(rdev); r600_wb_fini(rdev); r600_pcie_gart_fini(rdev); @@ -1806,8 +1869,452 @@ 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.align_mask = 4 - 1; +} + +static int r600_ih_ring_alloc(struct radeon_device *rdev, unsigned ring_size) +{ + int r; + + rdev->ih.ring_size = ring_size; + /* Allocate ring buffer */ + if (rdev->ih.ring_obj == NULL) { + r = radeon_object_create(rdev, NULL, rdev->ih.ring_size, + true, + RADEON_GEM_DOMAIN_GTT, + false, + &rdev->ih.ring_obj); + if (r) { + DRM_ERROR("radeon: failed to create ih ring buffer (%d).\n", r); + return r; + } + r = radeon_object_pin(rdev->ih.ring_obj, + RADEON_GEM_DOMAIN_GTT, + &rdev->ih.gpu_addr); + if (r) { + DRM_ERROR("radeon: failed to pin ih ring buffer (%d).\n", r); + return r; + } + r = radeon_object_kmap(rdev->ih.ring_obj, + (void **)&rdev->ih.ring); + if (r) { + DRM_ERROR("radeon: failed to map ih ring buffer (%d).\n", r); + return r; + } + } + rdev->ih.ptr_mask = (rdev->cp.ring_size / 4) - 1; + rdev->ih.rptr = 0; + + return 0; +} + +static void r600_ih_ring_fini(struct radeon_device *rdev) +{ + if (rdev->ih.ring_obj) { + radeon_object_kunmap(rdev->ih.ring_obj); + radeon_object_unpin(rdev->ih.ring_obj); + radeon_object_unref(&rdev->ih.ring_obj); + rdev->ih.ring = NULL; + rdev->ih.ring_obj = NULL; + } +} + +static void r600_rlc_stop(struct radeon_device *rdev) +{ + + if (rdev->family >= CHIP_RV770) { + /* 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_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; +} + +static 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; +} + +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, rdev->ih.ring_size); + 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 */ + WREG32(CP_INT_CNTL, 0); + WREG32(GRBM_INT_CNTL, 0); + WREG32(DxMODE_INT_MASK, 0); + + /* enable irqs */ + r600_enable_interrupts(rdev); + + return ret; +} + +void r600_irq_fini(struct radeon_device *rdev) +{ + r600_disable_interrupts(rdev); + r600_rlc_stop(rdev); + r600_ih_ring_fini(rdev); +} + +int r600_irq_set(struct radeon_device *rdev) +{ + uint32_t cp_int_cntl = CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE; + uint32_t mode_int = 0; + + /* don't enable anything if the ih is disabled */ + if (!rdev->ih.enabled) + return 0; + + 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; + } + + WREG32(CP_INT_CNTL, cp_int_cntl); + WREG32(DxMODE_INT_MASK, mode_int); + + return 0; +} + +static inline void r600_irq_ack(struct radeon_device *rdev, u32 disp_int) +{ + + 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); + +} + +void r600_irq_disable(struct radeon_device *rdev) +{ + u32 disp_int; + + r600_disable_interrupts(rdev); + /* Wait and acknowledge irq */ + mdelay(1); + if (ASIC_IS_DCE3(rdev)) + disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS); + else + disp_int = RREG32(DISP_INTERRUPT_STATUS); + r600_irq_ack(rdev, disp_int); +} + +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) { + WARN_ON(1); + /* XXX deal with overflow */ + DRM_ERROR("IH RB overflow\n"); + tmp = RREG32(IH_RB_CNTL); + tmp |= IH_WPTR_OVERFLOW_CLEAR; + WREG32(IH_RB_CNTL, tmp); + } + wptr = wptr & WPTR_OFFSET_MASK; + return wptr; +} +/* 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 + * 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 last_entry = rdev->ih.ring_size - 16; + u32 ring_index, disp_int; + unsigned long flags; + + DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr); + + 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 */ + if (ASIC_IS_DCE3(rdev)) + disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS); + else + disp_int = RREG32(DISP_INTERRUPT_STATUS); + r600_irq_ack(rdev, disp_int); + + 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); + 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_ERROR("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); + 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_ERROR("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + 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; + default: + DRM_ERROR("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + + /* wptr/rptr are in bytes! */ + if (rptr == last_entry) + rptr = 0; + else + rptr += 16; + } + /* make sure wptr hasn't changed while processing */ + wptr = r600_get_ih_wptr(rdev); + if (wptr != rdev->ih.wptr) + goto restart_ih; + rdev->ih.rptr = rptr; + WREG32(IH_RB_RPTR, rdev->ih.rptr); + spin_unlock_irqrestore(&rdev->ih.lock, flags); + return IRQ_HANDLED; +} /* * Debugfs info diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/radeon/r600_blit_kms.c index dbf716e..c20909c 100644 --- a/drivers/gpu/drm/radeon/r600_blit_kms.c +++ b/drivers/gpu/drm/radeon/r600_blit_kms.c @@ -569,9 +569,9 @@ int r600_blit_prepare_copy(struct radeon_device *rdev, int size_bytes) ring_size = num_loops * dwords_per_loop; /* set default + shaders */ ring_size += 40; /* shaders + def state */ - ring_size += 3; /* fence emit for VB IB */ + ring_size += 5; /* fence emit for VB IB */ ring_size += 5; /* done copy */ - ring_size += 3; /* fence emit for done copy */ + ring_size += 5; /* fence emit for done copy */ r = radeon_ring_lock(rdev, ring_size); WARN_ON(r); diff --git a/drivers/gpu/drm/radeon/r600d.h b/drivers/gpu/drm/radeon/r600d.h index b7f4ce2..61ccde5 100644 --- a/drivers/gpu/drm/radeon/r600d.h +++ b/drivers/gpu/drm/radeon/r600d.h @@ -456,7 +456,163 @@ #define WAIT_2D_IDLECLEAN_bit (1 << 16) #define WAIT_3D_IDLECLEAN_bit (1 << 17) - +#define IH_RB_CNTL 0x3e00 +# define IH_RB_ENABLE (1 << 0) +# define IH_IB_SIZE(x) ((x) << 1) /* log2 */ +# define IH_RB_FULL_DRAIN_ENABLE (1 << 6) +# define IH_WPTR_WRITEBACK_ENABLE (1 << 8) +# define IH_WPTR_WRITEBACK_TIMER(x) ((x) << 9) /* log2 */ +# define IH_WPTR_OVERFLOW_ENABLE (1 << 16) +# define IH_WPTR_OVERFLOW_CLEAR (1 << 31) +#define IH_RB_BASE 0x3e04 +#define IH_RB_RPTR 0x3e08 +#define IH_RB_WPTR 0x3e0c +# define RB_OVERFLOW (1 << 0) +# define WPTR_OFFSET_MASK 0x3fffc +#define IH_RB_WPTR_ADDR_HI 0x3e10 +#define IH_RB_WPTR_ADDR_LO 0x3e14 +#define IH_CNTL 0x3e18 +# define ENABLE_INTR (1 << 0) +# define IH_MC_SWAP(x) ((x) << 2) +# define IH_MC_SWAP_NONE 0 +# define IH_MC_SWAP_16BIT 1 +# define IH_MC_SWAP_32BIT 2 +# define IH_MC_SWAP_64BIT 3 +# define RPTR_REARM (1 << 4) +# define MC_WRREQ_CREDIT(x) ((x) << 15) +# define MC_WR_CLEAN_CNT(x) ((x) << 20) + +#define RLC_CNTL 0x3f00 +# define RLC_ENABLE (1 << 0) +#define RLC_HB_BASE 0x3f10 +#define RLC_HB_CNTL 0x3f0c +#define RLC_HB_RPTR 0x3f20 +#define RLC_HB_WPTR 0x3f1c +#define RLC_HB_WPTR_LSB_ADDR 0x3f14 +#define RLC_HB_WPTR_MSB_ADDR 0x3f18 +#define RLC_MC_CNTL 0x3f44 +#define RLC_UCODE_CNTL 0x3f48 +#define RLC_UCODE_ADDR 0x3f2c +#define RLC_UCODE_DATA 0x3f30 + +#define SRBM_SOFT_RESET 0xe60 +# define SOFT_RESET_RLC (1 << 13) + +#define CP_INT_CNTL 0xc124 +# define CNTX_BUSY_INT_ENABLE (1 << 19) +# define CNTX_EMPTY_INT_ENABLE (1 << 20) +# define SCRATCH_INT_ENABLE (1 << 25) +# define TIME_STAMP_INT_ENABLE (1 << 26) +# define IB2_INT_ENABLE (1 << 29) +# define IB1_INT_ENABLE (1 << 30) +# define RB_INT_ENABLE (1 << 31) +#define CP_INT_STATUS 0xc128 +# define SCRATCH_INT_STAT (1 << 25) +# define TIME_STAMP_INT_STAT (1 << 26) +# define IB2_INT_STAT (1 << 29) +# define IB1_INT_STAT (1 << 30) +# define RB_INT_STAT (1 << 31) + +#define GRBM_INT_CNTL 0x8060 +# define RDERR_INT_ENABLE (1 << 0) +# define WAIT_COUNT_TIMEOUT_INT_ENABLE (1 << 1) +# define GUI_IDLE_INT_ENABLE (1 << 19) + +#define INTERRUPT_CNTL 0x5468 +# define IH_DUMMY_RD_OVERRIDE (1 << 0) +# define IH_DUMMY_RD_EN (1 << 1) +# define IH_REQ_NONSNOOP_EN (1 << 3) +# define GEN_IH_INT_EN (1 << 8) +#define INTERRUPT_CNTL2 0x546c + +#define D1MODE_VBLANK_STATUS 0x6534 +#define D2MODE_VBLANK_STATUS 0x6d34 +# define DxMODE_VBLANK_OCCURRED (1 << 0) +# define DxMODE_VBLANK_ACK (1 << 4) +# define DxMODE_VBLANK_STAT (1 << 12) +# define DxMODE_VBLANK_INTERRUPT (1 << 16) +# define DxMODE_VBLANK_INTERRUPT_TYPE (1 << 17) +#define D1MODE_VLINE_STATUS 0x653c +#define D2MODE_VLINE_STATUS 0x6d3c +# define DxMODE_VLINE_OCCURRED (1 << 0) +# define DxMODE_VLINE_ACK (1 << 4) +# define DxMODE_VLINE_STAT (1 << 12) +# define DxMODE_VLINE_INTERRUPT (1 << 16) +# define DxMODE_VLINE_INTERRUPT_TYPE (1 << 17) +#define DxMODE_INT_MASK 0x6540 +# define D1MODE_VBLANK_INT_MASK (1 << 0) +# define D1MODE_VLINE_INT_MASK (1 << 4) +# define D2MODE_VBLANK_INT_MASK (1 << 8) +# define D2MODE_VLINE_INT_MASK (1 << 12) +#define DCE3_DISP_INTERRUPT_STATUS 0x7ddc +# define DC_HPD1_INTERRUPT (1 << 18) +# define DC_HPD2_INTERRUPT (1 << 19) +#define DISP_INTERRUPT_STATUS 0x7edc +# define LB_D1_VLINE_INTERRUPT (1 << 2) +# define LB_D2_VLINE_INTERRUPT (1 << 3) +# define LB_D1_VBLANK_INTERRUPT (1 << 4) +# define LB_D2_VBLANK_INTERRUPT (1 << 5) +# define DACA_AUTODETECT_INTERRUPT (1 << 16) +# define DACB_AUTODETECT_INTERRUPT (1 << 17) +# define DC_HOT_PLUG_DETECT1_INTERRUPT (1 << 18) +# define DC_HOT_PLUG_DETECT2_INTERRUPT (1 << 19) +# define DC_I2C_SW_DONE_INTERRUPT (1 << 20) +# define DC_I2C_HW_DONE_INTERRUPT (1 << 21) +#define DCE3_DISP_INTERRUPT_STATUS_CONTINUE 0x7de8 +# define DC_HPD4_INTERRUPT (1 << 14) +# define DC_HPD4_RX_INTERRUPT (1 << 15) +# define DC_HPD3_INTERRUPT (1 << 28) +# define DC_HPD1_RX_INTERRUPT (1 << 29) +# define DC_HPD2_RX_INTERRUPT (1 << 30) +#define DCE3_DISP_INTERRUPT_STATUS_CONTINUE2 0x7dec +# define DC_HPD3_RX_INTERRUPT (1 << 0) +# define DIGA_DP_VID_STREAM_DISABLE_INTERRUPT (1 << 1) +# define DIGA_DP_STEER_FIFO_OVERFLOW_INTERRUPT (1 << 2) +# define DIGB_DP_VID_STREAM_DISABLE_INTERRUPT (1 << 3) +# define DIGB_DP_STEER_FIFO_OVERFLOW_INTERRUPT (1 << 4) +# define AUX1_SW_DONE_INTERRUPT (1 << 5) +# define AUX1_LS_DONE_INTERRUPT (1 << 6) +# define AUX2_SW_DONE_INTERRUPT (1 << 7) +# define AUX2_LS_DONE_INTERRUPT (1 << 8) +# define AUX3_SW_DONE_INTERRUPT (1 << 9) +# define AUX3_LS_DONE_INTERRUPT (1 << 10) +# define AUX4_SW_DONE_INTERRUPT (1 << 11) +# define AUX4_LS_DONE_INTERRUPT (1 << 12) +# define DIGA_DP_FAST_TRAINING_COMPLETE_INTERRUPT (1 << 13) +# define DIGB_DP_FAST_TRAINING_COMPLETE_INTERRUPT (1 << 14) +/* DCE 3.2 */ +# define AUX5_SW_DONE_INTERRUPT (1 << 15) +# define AUX5_LS_DONE_INTERRUPT (1 << 16) +# define AUX6_SW_DONE_INTERRUPT (1 << 17) +# define AUX6_LS_DONE_INTERRUPT (1 << 18) +# define DC_HPD5_INTERRUPT (1 << 19) +# define DC_HPD5_RX_INTERRUPT (1 << 20) +# define DC_HPD6_INTERRUPT (1 << 21) +# define DC_HPD6_RX_INTERRUPT (1 << 22) + +#define DCE3_DACA_AUTODETECT_INT_CONTROL 0x7038 +#define DCE3_DACB_AUTODETECT_INT_CONTROL 0x7138 +#define DACA_AUTODETECT_INT_CONTROL 0x7838 +#define DACB_AUTODETECT_INT_CONTROL 0x7a38 +# define DACx_AUTODETECT_ACK (1 << 0) +# define DACx_AUTODETECT_INT_ENABLE (1 << 16) + +#define DC_HOT_PLUG_DETECT1_INT_CONTROL 0x7d08 +#define DC_HOT_PLUG_DETECT2_INT_CONTROL 0x7d18 +#define DC_HOT_PLUG_DETECT3_INT_CONTROL 0x7d2c +# define DC_HOT_PLUG_DETECTx_INT_ACK (1 << 0) +# define DC_HOT_PLUG_DETECTx_INT_POLARITY (1 << 8) +# define DC_HOT_PLUG_DETECTx_INT_EN (1 << 16) +/* DCE 3.2 */ +#define DC_HPD1_INT_CONTROL 0x7d04 +#define DC_HPD2_INT_CONTROL 0x7d10 +#define DC_HPD3_INT_CONTROL 0x7d1c +#define DC_HPD4_INT_CONTROL 0x7d28 +# define DC_HPDx_INT_ACK (1 << 0) +# define DC_HPDx_INT_POLARITY (1 << 8) +# define DC_HPDx_INT_EN (1 << 16) +# define DC_HPDx_RX_INT_ACK (1 << 20) +# define DC_HPDx_RX_INT_EN (1 << 24) /* * PM4 @@ -500,7 +656,6 @@ #define PACKET3_WAIT_REG_MEM 0x3C #define PACKET3_MEM_WRITE 0x3D #define PACKET3_INDIRECT_BUFFER 0x32 -#define PACKET3_CP_INTERRUPT 0x40 #define PACKET3_SURFACE_SYNC 0x43 # define PACKET3_CB0_DEST_BASE_ENA (1 << 6) # define PACKET3_TC_ACTION_ENA (1 << 23) diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 15b9e03..0b8dad6 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -399,6 +399,23 @@ struct radeon_cp { bool ready; }; +/* + * R6xx+ IH ring + */ +struct r600_ih { + struct radeon_object *ring_obj; + volatile uint32_t *ring; + unsigned rptr; + unsigned wptr; + unsigned wptr_old; + unsigned ring_size; + uint64_t gpu_addr; + uint32_t align_mask; + uint32_t ptr_mask; + spinlock_t lock; + bool enabled; +}; + struct r600_blit { struct radeon_object *shader_obj; u64 shader_gpu_addr; @@ -792,8 +809,10 @@ struct radeon_device { struct radeon_surface_reg surface_regs[RADEON_GEM_MAX_SURFACES]; const struct firmware *me_fw; /* all family ME firmware */ const struct firmware *pfp_fw; /* r6/700 PFP firmware */ + const struct firmware *rlc_fw; /* r6/700 RLC firmware */ struct r600_blit r600_blit; int msi_enabled; /* msi enabled */ + struct r600_ih ih; /* r6/700 interrupt ring */ }; int radeon_device_init(struct radeon_device *rdev, @@ -1108,7 +1127,12 @@ extern void r600_wb_disable(struct radeon_device *rdev); extern void r600_scratch_init(struct radeon_device *rdev); extern int r600_blit_init(struct radeon_device *rdev); extern void r600_blit_fini(struct radeon_device *rdev); -extern int r600_cp_init_microcode(struct radeon_device *rdev); +extern int r600_init_microcode(struct radeon_device *rdev); extern int r600_gpu_reset(struct radeon_device *rdev); +/* r600 irq */ +extern int r600_irq_init(struct radeon_device *rdev); +extern void r600_irq_fini(struct radeon_device *rdev); +extern void r600_ih_ring_init(struct radeon_device *rdev, unsigned ring_size); +extern int r600_irq_set(struct radeon_device *rdev); #endif diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index c7a7f84..755f505 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -480,6 +480,7 @@ static struct radeon_asic r600_asic = { .ring_ib_execute = &r600_ring_ib_execute, .irq_set = &r600_irq_set, .irq_process = &r600_irq_process, + .get_vblank_counter = &rs600_get_vblank_counter, .fence_ring_emit = &r600_fence_ring_emit, .cs_parse = &r600_cs_parse, .copy_blit = &r600_copy_blit, @@ -520,6 +521,7 @@ static struct radeon_asic rv770_asic = { .ring_ib_execute = &r600_ring_ib_execute, .irq_set = &r600_irq_set, .irq_process = &r600_irq_process, + .get_vblank_counter = &rs600_get_vblank_counter, .fence_ring_emit = &r600_fence_ring_emit, .cs_parse = &r600_cs_parse, .copy_blit = &r600_copy_blit, diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index c43a690..c962f34 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -562,6 +562,8 @@ int radeon_device_init(struct radeon_device *rdev, mutex_init(&rdev->cs_mutex); mutex_init(&rdev->ib_pool.mutex); mutex_init(&rdev->cp.mutex); + if (rdev->family >= CHIP_R600) + spin_lock_init(&rdev->ih.lock); rwlock_init(&rdev->fence_drv.lock); INIT_LIST_HEAD(&rdev->gem.objects); diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h index 350962e..e137852 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.h +++ b/drivers/gpu/drm/radeon/radeon_drv.h @@ -1104,7 +1104,6 @@ extern u32 radeon_get_scratch(drm_radeon_private_t *dev_priv, int index); # define R600_IT_WAIT_REG_MEM 0x00003C00 # define R600_IT_MEM_WRITE 0x00003D00 # define R600_IT_INDIRECT_BUFFER 0x00003200 -# define R600_IT_CP_INTERRUPT 0x00004000 # define R600_IT_SURFACE_SYNC 0x00004300 # define R600_CB0_DEST_BASE_ENA (1 << 6) # define R600_TC_ACTION_ENA (1 << 23) diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 3beb26d..ab2a8b1 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -168,37 +168,6 @@ bool radeon_fence_signaled(struct radeon_fence *fence) return signaled; } -int r600_fence_wait(struct radeon_fence *fence, bool intr, bool lazy) -{ - struct radeon_device *rdev; - int ret = 0; - - rdev = fence->rdev; - - __set_current_state(intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE); - - while (1) { - if (radeon_fence_signaled(fence)) - break; - - if (time_after_eq(jiffies, fence->timeout)) { - ret = -EBUSY; - break; - } - - if (lazy) - schedule_timeout(1); - - if (intr && signal_pending(current)) { - ret = -ERESTARTSYS; - break; - } - } - __set_current_state(TASK_RUNNING); - return ret; -} - - int radeon_fence_wait(struct radeon_fence *fence, bool intr) { struct radeon_device *rdev; @@ -216,13 +185,6 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr) return 0; } - if (rdev->family >= CHIP_R600) { - r = r600_fence_wait(fence, intr, 0); - if (r == -ERESTARTSYS) - return -EBUSY; - return r; - } - retry: cur_jiffies = jiffies; timeout = HZ / 100; diff --git a/drivers/gpu/drm/radeon/radeon_irq_kms.c b/drivers/gpu/drm/radeon/radeon_irq_kms.c index a0fe623..84f8a6f 100644 --- a/drivers/gpu/drm/radeon/radeon_irq_kms.c +++ b/drivers/gpu/drm/radeon/radeon_irq_kms.c @@ -94,10 +94,18 @@ int radeon_irq_kms_init(struct radeon_device *rdev) } /* enable msi */ rdev->msi_enabled = 0; - if (rdev->family >= CHIP_RV380) { + /* MSIs don't seem to work on my rs780; + * not sure about rs880 or other rs780s. + * Needs more investigation. + */ + if ((rdev->family >= CHIP_RV380) && + (rdev->family != CHIP_RS780) && + (rdev->family != CHIP_RS880)) { int ret = pci_enable_msi(rdev->pdev); - if (!ret) + if (!ret) { rdev->msi_enabled = 1; + DRM_INFO("radeon: using MSI.\n"); + } } drm_irq_install(rdev->ddev); rdev->irq.installed = true; diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index f546284..479684b 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -887,6 +887,16 @@ static int rv770_startup(struct radeon_device *rdev) return r; } + /* Enable IRQ */ + rdev->irq.sw_int = true; + r = r600_irq_init(rdev); + if (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) return r; @@ -1005,11 +1015,19 @@ int rv770_init(struct radeon_device *rdev) r = radeon_object_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); + rdev->ih.ring_obj = NULL; + r600_ih_ring_init(rdev, 64 * 1024); + + 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; @@ -1055,6 +1073,8 @@ void rv770_fini(struct radeon_device *rdev) rv770_suspend(rdev); r600_blit_fini(rdev); + r600_irq_fini(rdev); + radeon_irq_kms_fini(rdev); radeon_ring_fini(rdev); r600_wb_fini(rdev); rv770_pcie_gart_fini(rdev); -- 1.8.2.3