Merge remote branch 'intel/drm-intel-next' of ../drm-next into drm-core-next

* 'intel/drm-intel-next' of ../drm-next: (755 commits)
  drm/i915: Only wait on a pending flip if we intend to write to the buffer
  drm/i915/dp: Sanity check eDP existence
  drm/i915: Rebind the buffer if its alignment constraints changes with tiling
  drm/i915: Disable GPU semaphores by default
  drm/i915: Do not overflow the MMADDR write FIFO
  Revert "drm/i915: fix corruptions on i8xx due to relaxed fencing"
  drm/i915: Don't save/restore hardware status page address register
  drm/i915: don't store the reg value for HWS_PGA
  drm/i915: fix memory corruption with GM965 and >4GB RAM
  Linux 2.6.38-rc7
  Revert "TPM: Long default timeout fix"
  drm/i915: Re-enable GPU semaphores for SandyBridge mobile
  drm/i915: Replace vblank PM QoS with "Interrupt-Based AGPBUSY#"
  Revert "drm/i915: Use PM QoS to prevent C-State starvation of gen3 GPU"
  drm/i915: Allow relocation deltas outside of target bo
  drm/i915: Silence an innocuous compiler warning for an unused variable
  fs/block_dev.c: fix new kernel-doc warning
  ACPI: Fix build for CONFIG_NET unset
  mm: <asm-generic/pgtable.h> must include <linux/mm_types.h>
  x86: Use u32 instead of long to set reset vector back to 0
  ...

Conflicts:
	drivers/gpu/drm/i915/i915_gem.c

34 files changed, 451 insertions, 200 deletions

diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c
index d56f08d3cbd..a2199fe9fa9 100644
--- a/drivers/gpu/drm/radeon/atombios_crtc.c
+++ b/drivers/gpu/drm/radeon/atombios_crtc.c
@@ -48,29 +48,29 @@ static void atombios_overscan_setup(struct drm_crtc *crtc,
 
 	switch (radeon_crtc->rmx_type) {
 	case RMX_CENTER:
-		args.usOverscanTop = (adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2;
-		args.usOverscanBottom = (adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2;
-		args.usOverscanLeft = (adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2;
-		args.usOverscanRight = (adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2;
+		args.usOverscanTop = cpu_to_le16((adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2);
+		args.usOverscanBottom = cpu_to_le16((adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2);
+		args.usOverscanLeft = cpu_to_le16((adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2);
+		args.usOverscanRight = cpu_to_le16((adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2);
 		break;
 	case RMX_ASPECT:
 		a1 = mode->crtc_vdisplay * adjusted_mode->crtc_hdisplay;
 		a2 = adjusted_mode->crtc_vdisplay * mode->crtc_hdisplay;
 
 		if (a1 > a2) {
-			args.usOverscanLeft = (adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2;
-			args.usOverscanRight = (adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2;
+			args.usOverscanLeft = cpu_to_le16((adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2);
+			args.usOverscanRight = cpu_to_le16((adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2);
 		} else if (a2 > a1) {
-			args.usOverscanLeft = (adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2;
-			args.usOverscanRight = (adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2;
+			args.usOverscanLeft = cpu_to_le16((adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2);
+			args.usOverscanRight = cpu_to_le16((adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2);
 		}
 		break;
 	case RMX_FULL:
 	default:
-		args.usOverscanRight = radeon_crtc->h_border;
-		args.usOverscanLeft = radeon_crtc->h_border;
-		args.usOverscanBottom = radeon_crtc->v_border;
-		args.usOverscanTop = radeon_crtc->v_border;
+		args.usOverscanRight = cpu_to_le16(radeon_crtc->h_border);
+		args.usOverscanLeft = cpu_to_le16(radeon_crtc->h_border);
+		args.usOverscanBottom = cpu_to_le16(radeon_crtc->v_border);
+		args.usOverscanTop = cpu_to_le16(radeon_crtc->v_border);
 		break;
 	}
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
@@ -419,23 +419,23 @@ static void atombios_crtc_program_ss(struct drm_crtc *crtc,
 	memset(&args, 0, sizeof(args));
 
 	if (ASIC_IS_DCE5(rdev)) {
-		args.v3.usSpreadSpectrumAmountFrac = 0;
+		args.v3.usSpreadSpectrumAmountFrac = cpu_to_le16(0);
 		args.v3.ucSpreadSpectrumType = ss->type;
 		switch (pll_id) {
 		case ATOM_PPLL1:
 			args.v3.ucSpreadSpectrumType |= ATOM_PPLL_SS_TYPE_V3_P1PLL;
-			args.v3.usSpreadSpectrumAmount = ss->amount;
-			args.v3.usSpreadSpectrumStep = ss->step;
+			args.v3.usSpreadSpectrumAmount = cpu_to_le16(ss->amount);
+			args.v3.usSpreadSpectrumStep = cpu_to_le16(ss->step);
 			break;
 		case ATOM_PPLL2:
 			args.v3.ucSpreadSpectrumType |= ATOM_PPLL_SS_TYPE_V3_P2PLL;
-			args.v3.usSpreadSpectrumAmount = ss->amount;
-			args.v3.usSpreadSpectrumStep = ss->step;
+			args.v3.usSpreadSpectrumAmount = cpu_to_le16(ss->amount);
+			args.v3.usSpreadSpectrumStep = cpu_to_le16(ss->step);
 			break;
 		case ATOM_DCPLL:
 			args.v3.ucSpreadSpectrumType |= ATOM_PPLL_SS_TYPE_V3_DCPLL;
-			args.v3.usSpreadSpectrumAmount = 0;
-			args.v3.usSpreadSpectrumStep = 0;
+			args.v3.usSpreadSpectrumAmount = cpu_to_le16(0);
+			args.v3.usSpreadSpectrumStep = cpu_to_le16(0);
 			break;
 		case ATOM_PPLL_INVALID:
 			return;
@@ -447,18 +447,18 @@ static void atombios_crtc_program_ss(struct drm_crtc *crtc,
 		switch (pll_id) {
 		case ATOM_PPLL1:
 			args.v2.ucSpreadSpectrumType |= ATOM_PPLL_SS_TYPE_V2_P1PLL;
-			args.v2.usSpreadSpectrumAmount = ss->amount;
-			args.v2.usSpreadSpectrumStep = ss->step;
+			args.v2.usSpreadSpectrumAmount = cpu_to_le16(ss->amount);
+			args.v2.usSpreadSpectrumStep = cpu_to_le16(ss->step);
 			break;
 		case ATOM_PPLL2:
 			args.v2.ucSpreadSpectrumType |= ATOM_PPLL_SS_TYPE_V2_P2PLL;
-			args.v2.usSpreadSpectrumAmount = ss->amount;
-			args.v2.usSpreadSpectrumStep = ss->step;
+			args.v2.usSpreadSpectrumAmount = cpu_to_le16(ss->amount);
+			args.v2.usSpreadSpectrumStep = cpu_to_le16(ss->step);
 			break;
 		case ATOM_DCPLL:
 			args.v2.ucSpreadSpectrumType |= ATOM_PPLL_SS_TYPE_V2_DCPLL;
-			args.v2.usSpreadSpectrumAmount = 0;
-			args.v2.usSpreadSpectrumStep = 0;
+			args.v2.usSpreadSpectrumAmount = cpu_to_le16(0);
+			args.v2.usSpreadSpectrumStep = cpu_to_le16(0);
 			break;
 		case ATOM_PPLL_INVALID:
 			return;
@@ -538,7 +538,6 @@ static u32 atombios_adjust_pll(struct drm_crtc *crtc,
 			pll->flags |= RADEON_PLL_PREFER_HIGH_FB_DIV;
 		else
 			pll->flags |= RADEON_PLL_PREFER_LOW_REF_DIV;
-
 	}
 
 	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
@@ -555,29 +554,28 @@ static u32 atombios_adjust_pll(struct drm_crtc *crtc,
 					dp_clock = dig_connector->dp_clock;
 				}
 			}
-/* this might work properly with the new pll algo */
-#if 0 /* doesn't work properly on some laptops */
+
 			/* use recommended ref_div for ss */
 			if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
 				if (ss_enabled) {
 					if (ss->refdiv) {
+						pll->flags |= RADEON_PLL_PREFER_MINM_OVER_MAXP;
 						pll->flags |= RADEON_PLL_USE_REF_DIV;
 						pll->reference_div = ss->refdiv;
+						if (ASIC_IS_AVIVO(rdev))
+							pll->flags |= RADEON_PLL_USE_FRAC_FB_DIV;
 					}
 				}
 			}
-#endif
+
 			if (ASIC_IS_AVIVO(rdev)) {
 				/* DVO wants 2x pixel clock if the DVO chip is in 12 bit mode */
 				if (radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1)
 					adjusted_clock = mode->clock * 2;
 				if (radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT))
 					pll->flags |= RADEON_PLL_PREFER_CLOSEST_LOWER;
-				/* rv515 needs more testing with this option */
-				if (rdev->family != CHIP_RV515) {
-					if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
-						pll->flags |= RADEON_PLL_IS_LCD;
-				}
+				if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
+					pll->flags |= RADEON_PLL_IS_LCD;
 			} else {
 				if (encoder->encoder_type != DRM_MODE_ENCODER_DAC)
 					pll->flags |= RADEON_PLL_NO_ODD_POST_DIV;
@@ -664,10 +662,12 @@ static u32 atombios_adjust_pll(struct drm_crtc *crtc,
 						   index, (uint32_t *)&args);
 				adjusted_clock = le32_to_cpu(args.v3.sOutput.ulDispPllFreq) * 10;
 				if (args.v3.sOutput.ucRefDiv) {
+					pll->flags |= RADEON_PLL_USE_FRAC_FB_DIV;
 					pll->flags |= RADEON_PLL_USE_REF_DIV;
 					pll->reference_div = args.v3.sOutput.ucRefDiv;
 				}
 				if (args.v3.sOutput.ucPostDiv) {
+					pll->flags |= RADEON_PLL_USE_FRAC_FB_DIV;
 					pll->flags |= RADEON_PLL_USE_POST_DIV;
 					pll->post_div = args.v3.sOutput.ucPostDiv;
 				}
@@ -721,14 +721,14 @@ static void atombios_crtc_set_dcpll(struct drm_crtc *crtc,
 			 * SetPixelClock provides the dividers
 			 */
 			args.v5.ucCRTC = ATOM_CRTC_INVALID;
-			args.v5.usPixelClock = dispclk;
+			args.v5.usPixelClock = cpu_to_le16(dispclk);
 			args.v5.ucPpll = ATOM_DCPLL;
 			break;
 		case 6:
 			/* if the default dcpll clock is specified,
 			 * SetPixelClock provides the dividers
 			 */
-			args.v6.ulDispEngClkFreq = dispclk;
+			args.v6.ulDispEngClkFreq = cpu_to_le32(dispclk);
 			args.v6.ucPpll = ATOM_DCPLL;
 			break;
 		default:
@@ -957,11 +957,7 @@ static void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode
 	/* adjust pixel clock as needed */
 	adjusted_clock = atombios_adjust_pll(crtc, mode, pll, ss_enabled, &ss);
 
-	/* rv515 seems happier with the old algo */
-	if (rdev->family == CHIP_RV515)
-		radeon_compute_pll_legacy(pll, adjusted_clock, &pll_clock, &fb_div, &frac_fb_div,
-					  &ref_div, &post_div);
-	else if (ASIC_IS_AVIVO(rdev))
+	if (ASIC_IS_AVIVO(rdev))
 		radeon_compute_pll_avivo(pll, adjusted_clock, &pll_clock, &fb_div, &frac_fb_div,
 					 &ref_div, &post_div);
 	else
@@ -995,9 +991,9 @@ static void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode
 	}
 }
 
-static int evergreen_crtc_do_set_base(struct drm_crtc *crtc,
-				      struct drm_framebuffer *fb,
-				      int x, int y, int atomic)
+static int dce4_crtc_do_set_base(struct drm_crtc *crtc,
+				 struct drm_framebuffer *fb,
+				 int x, int y, int atomic)
 {
 	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
@@ -1137,12 +1133,6 @@ static int evergreen_crtc_do_set_base(struct drm_crtc *crtc,
 	WREG32(EVERGREEN_VIEWPORT_SIZE + radeon_crtc->crtc_offset,
 	       (crtc->mode.hdisplay << 16) | crtc->mode.vdisplay);
 
-	if (crtc->mode.flags & DRM_MODE_FLAG_INTERLACE)
-		WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset,
-		       EVERGREEN_INTERLEAVE_EN);
-	else
-		WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset, 0);
-
 	if (!atomic && fb && fb != crtc->fb) {
 		radeon_fb = to_radeon_framebuffer(fb);
 		rbo = gem_to_radeon_bo(radeon_fb->obj);
@@ -1300,12 +1290,6 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 	WREG32(AVIVO_D1MODE_VIEWPORT_SIZE + radeon_crtc->crtc_offset,
 	       (crtc->mode.hdisplay << 16) | crtc->mode.vdisplay);
 
-	if (crtc->mode.flags & DRM_MODE_FLAG_INTERLACE)
-		WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset,
-		       AVIVO_D1MODE_INTERLEAVE_EN);
-	else
-		WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset, 0);
-
 	if (!atomic && fb && fb != crtc->fb) {
 		radeon_fb = to_radeon_framebuffer(fb);
 		rbo = gem_to_radeon_bo(radeon_fb->obj);
@@ -1329,7 +1313,7 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y,
 	struct radeon_device *rdev = dev->dev_private;
 
 	if (ASIC_IS_DCE4(rdev))
-		return evergreen_crtc_do_set_base(crtc, old_fb, x, y, 0);
+		return dce4_crtc_do_set_base(crtc, old_fb, x, y, 0);
 	else if (ASIC_IS_AVIVO(rdev))
 		return avivo_crtc_do_set_base(crtc, old_fb, x, y, 0);
 	else
@@ -1344,7 +1328,7 @@ int atombios_crtc_set_base_atomic(struct drm_crtc *crtc,
        struct radeon_device *rdev = dev->dev_private;
 
 	if (ASIC_IS_DCE4(rdev))
-		return evergreen_crtc_do_set_base(crtc, fb, x, y, 1);
+		return dce4_crtc_do_set_base(crtc, fb, x, y, 1);
 	else if (ASIC_IS_AVIVO(rdev))
 		return avivo_crtc_do_set_base(crtc, fb, x, y, 1);
 	else
diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c
index d4045223d0f..789441ed983 100644
--- a/drivers/gpu/drm/radeon/evergreen.c
+++ b/drivers/gpu/drm/radeon/evergreen.c
@@ -1192,7 +1192,11 @@ void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
 	radeon_ring_write(rdev, 1);
 	/* FIXME: implement */
 	radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
-	radeon_ring_write(rdev, ib->gpu_addr & 0xFFFFFFFC);
+	radeon_ring_write(rdev,
+#ifdef __BIG_ENDIAN
+			  (2 << 0) |
+#endif
+			  (ib->gpu_addr & 0xFFFFFFFC));
 	radeon_ring_write(rdev, upper_32_bits(ib->gpu_addr) & 0xFF);
 	radeon_ring_write(rdev, ib->length_dw);
 }
@@ -1207,7 +1211,11 @@ static int evergreen_cp_load_microcode(struct radeon_device *rdev)
 		return -EINVAL;
 
 	r700_cp_stop(rdev);
-	WREG32(CP_RB_CNTL, RB_NO_UPDATE | (15 << 8) | (3 << 0));
+	WREG32(CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+	       BUF_SWAP_32BIT |
+#endif
+	       RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3));
 
 	fw_data = (const __be32 *)rdev->pfp_fw->data;
 	WREG32(CP_PFP_UCODE_ADDR, 0);
@@ -1326,7 +1334,11 @@ int evergreen_cp_resume(struct radeon_device *rdev)
 	WREG32(CP_RB_WPTR, 0);
 
 	/* set the wb address wether it's enabled or not */
-	WREG32(CP_RB_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC);
+	WREG32(CP_RB_RPTR_ADDR,
+#ifdef __BIG_ENDIAN
+	       RB_RPTR_SWAP(2) |
+#endif
+	       ((rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC));
 	WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF);
 	WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF);
 
@@ -2627,8 +2639,8 @@ restart_ih:
 	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;
+		src_id =  le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff;
+		src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff;
 
 		switch (src_id) {
 		case 1: /* D1 vblank/vline */
diff --git a/drivers/gpu/drm/radeon/evergreen_blit_kms.c b/drivers/gpu/drm/radeon/evergreen_blit_kms.c
index 2ed930e02f3..3218287f4c5 100644
--- a/drivers/gpu/drm/radeon/evergreen_blit_kms.c
+++ b/drivers/gpu/drm/radeon/evergreen_blit_kms.c
@@ -55,7 +55,7 @@ set_render_target(struct radeon_device *rdev, int format,
 	if (h < 8)
 		h = 8;
 
-	cb_color_info = ((format << 2) | (1 << 24));
+	cb_color_info = ((format << 2) | (1 << 24) | (1 << 8));
 	pitch = (w / 8) - 1;
 	slice = ((w * h) / 64) - 1;
 
@@ -133,6 +133,9 @@ set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
 
 	/* high addr, stride */
 	sq_vtx_constant_word2 = ((upper_32_bits(gpu_addr) & 0xff) | (16 << 8));
+#ifdef __BIG_ENDIAN
+	sq_vtx_constant_word2 |= (2 << 30);
+#endif
 	/* xyzw swizzles */
 	sq_vtx_constant_word3 = (0 << 3) | (1 << 6) | (2 << 9) | (3 << 12);
 
@@ -173,7 +176,7 @@ set_tex_resource(struct radeon_device *rdev,
 	sq_tex_resource_word0 = (1 << 0); /* 2D */
 	sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 6) |
 				  ((w - 1) << 18));
-	sq_tex_resource_word1 = ((h - 1) << 0);
+	sq_tex_resource_word1 = ((h - 1) << 0) | (1 << 28);
 	/* xyzw swizzles */
 	sq_tex_resource_word4 = (0 << 16) | (1 << 19) | (2 << 22) | (3 << 25);
 
@@ -221,7 +224,11 @@ draw_auto(struct radeon_device *rdev)
 	radeon_ring_write(rdev, DI_PT_RECTLIST);
 
 	radeon_ring_write(rdev, PACKET3(PACKET3_INDEX_TYPE, 0));
-	radeon_ring_write(rdev, DI_INDEX_SIZE_16_BIT);
+	radeon_ring_write(rdev,
+#ifdef __BIG_ENDIAN
+			  (2 << 2) |
+#endif
+			  DI_INDEX_SIZE_16_BIT);
 
 	radeon_ring_write(rdev, PACKET3(PACKET3_NUM_INSTANCES, 0));
 	radeon_ring_write(rdev, 1);
@@ -541,7 +548,7 @@ static inline uint32_t i2f(uint32_t input)
 int evergreen_blit_init(struct radeon_device *rdev)
 {
 	u32 obj_size;
-	int r, dwords;
+	int i, r, dwords;
 	void *ptr;
 	u32 packet2s[16];
 	int num_packet2s = 0;
@@ -557,7 +564,7 @@ int evergreen_blit_init(struct radeon_device *rdev)
 
 	dwords = rdev->r600_blit.state_len;
 	while (dwords & 0xf) {
-		packet2s[num_packet2s++] = PACKET2(0);
+		packet2s[num_packet2s++] = cpu_to_le32(PACKET2(0));
 		dwords++;
 	}
 
@@ -598,8 +605,10 @@ int evergreen_blit_init(struct radeon_device *rdev)
 	if (num_packet2s)
 		memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
 			    packet2s, num_packet2s * 4);
-	memcpy(ptr + rdev->r600_blit.vs_offset, evergreen_vs, evergreen_vs_size * 4);
-	memcpy(ptr + rdev->r600_blit.ps_offset, evergreen_ps, evergreen_ps_size * 4);
+	for (i = 0; i < evergreen_vs_size; i++)
+		*(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(evergreen_vs[i]);
+	for (i = 0; i < evergreen_ps_size; i++)
+		*(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(evergreen_ps[i]);
 	radeon_bo_kunmap(rdev->r600_blit.shader_obj);
 	radeon_bo_unreserve(rdev->r600_blit.shader_obj);
 
diff --git a/drivers/gpu/drm/radeon/evergreen_blit_shaders.c b/drivers/gpu/drm/radeon/evergreen_blit_shaders.c
index ef1d28c07fb..3a10399e006 100644
--- a/drivers/gpu/drm/radeon/evergreen_blit_shaders.c
+++ b/drivers/gpu/drm/radeon/evergreen_blit_shaders.c
@@ -311,11 +311,19 @@ const u32 evergreen_vs[] =
 	0x00000000,
 	0x3c000000,
 	0x67961001,
+#ifdef __BIG_ENDIAN
+	0x000a0000,
+#else
 	0x00080000,
+#endif
 	0x00000000,
 	0x1c000000,
 	0x67961000,
+#ifdef __BIG_ENDIAN
+	0x00020008,
+#else
 	0x00000008,
+#endif
 	0x00000000,
 };
 
diff --git a/drivers/gpu/drm/radeon/evergreend.h b/drivers/gpu/drm/radeon/evergreend.h
index 21e839bd20e..9aaa3f0c937 100644
--- a/drivers/gpu/drm/radeon/evergreend.h
+++ b/drivers/gpu/drm/radeon/evergreend.h
@@ -98,6 +98,7 @@
 #define		BUF_SWAP_32BIT					(2 << 16)
 #define	CP_RB_RPTR					0x8700
 #define	CP_RB_RPTR_ADDR					0xC10C
+#define		RB_RPTR_SWAP(x)					((x) << 0)
 #define	CP_RB_RPTR_ADDR_HI				0xC110
 #define	CP_RB_RPTR_WR					0xC108
 #define	CP_RB_WPTR					0xC114
diff --git a/drivers/gpu/drm/radeon/mkregtable.c b/drivers/gpu/drm/radeon/mkregtable.c
index 607241c6a8a..5a82b6b7584 100644
--- a/drivers/gpu/drm/radeon/mkregtable.c
+++ b/drivers/gpu/drm/radeon/mkregtable.c
@@ -673,8 +673,10 @@ static int parser_auth(struct table *t, const char *filename)
 	last_reg = strtol(last_reg_s, NULL, 16);
 
 	do {
-		if (fgets(buf, 1024, file) == NULL)
+		if (fgets(buf, 1024, file) == NULL) {
+			fclose(file);
 			return -1;
+		}
 		len = strlen(buf);
 		if (ftell(file) == end)
 			done = 1;
@@ -685,6 +687,7 @@ static int parser_auth(struct table *t, const char *filename)
 				fprintf(stderr,
 					"Error matching regular expression %d in %s\n",
 					r, filename);
+				fclose(file);
 				return -1;
 			} else {
 				buf[match[0].rm_eo] = 0;
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index 5f15820efe1..93fa735c8c1 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -1427,6 +1427,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->zb.robj = reloc->robj;
 		track->zb.offset = idx_value;
+		track->zb_dirty = true;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_RB3D_COLOROFFSET:
@@ -1439,6 +1440,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->cb[0].robj = reloc->robj;
 		track->cb[0].offset = idx_value;
+		track->cb_dirty = true;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_PP_TXOFFSET_0:
@@ -1454,6 +1456,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		}
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[i].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_CUBIC_OFFSET_T0_0:
 	case RADEON_PP_CUBIC_OFFSET_T0_1:
@@ -1471,6 +1474,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		track->textures[0].cube_info[i].offset = idx_value;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[0].cube_info[i].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_CUBIC_OFFSET_T1_0:
 	case RADEON_PP_CUBIC_OFFSET_T1_1:
@@ -1488,6 +1492,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		track->textures[1].cube_info[i].offset = idx_value;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[1].cube_info[i].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_CUBIC_OFFSET_T2_0:
 	case RADEON_PP_CUBIC_OFFSET_T2_1:
@@ -1505,9 +1510,12 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		track->textures[2].cube_info[i].offset = idx_value;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[2].cube_info[i].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	case RADEON_RE_WIDTH_HEIGHT:
 		track->maxy = ((idx_value >> 16) & 0x7FF);
+		track->cb_dirty = true;
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_COLORPITCH:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1528,9 +1536,11 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		ib[idx] = tmp;
 
 		track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
+		track->cb_dirty = true;
 		break;
 	case RADEON_RB3D_DEPTHPITCH:
 		track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_CNTL:
 		switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
@@ -1555,6 +1565,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			return -EINVAL;
 		}
 		track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
+		track->cb_dirty = true;
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_ZSTENCILCNTL:
 		switch (idx_value & 0xf) {
@@ -1572,6 +1584,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		default:
 			break;
 		}
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_ZPASS_ADDR:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1588,6 +1601,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			uint32_t temp = idx_value >> 4;
 			for (i = 0; i < track->num_texture; i++)
 				track->textures[i].enabled = !!(temp & (1 << i));
+			track->tex_dirty = true;
 		}
 		break;
 	case RADEON_SE_VF_CNTL:
@@ -1602,12 +1616,14 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		i = (reg - RADEON_PP_TEX_SIZE_0) / 8;
 		track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
 		track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_TEX_PITCH_0:
 	case RADEON_PP_TEX_PITCH_1:
 	case RADEON_PP_TEX_PITCH_2:
 		i = (reg - RADEON_PP_TEX_PITCH_0) / 8;
 		track->textures[i].pitch = idx_value + 32;
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_TXFILTER_0:
 	case RADEON_PP_TXFILTER_1:
@@ -1621,6 +1637,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		tmp = (idx_value >> 27) & 0x7;
 		if (tmp == 2 || tmp == 6)
 			track->textures[i].roundup_h = false;
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_TXFORMAT_0:
 	case RADEON_PP_TXFORMAT_1:
@@ -1673,6 +1690,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
 		track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
+		track->tex_dirty = true;
 		break;
 	case RADEON_PP_CUBIC_FACES_0:
 	case RADEON_PP_CUBIC_FACES_1:
@@ -1683,6 +1701,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
 			track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
 			track->textures[i].cube_info[face].height = 1 << ((tmp >> ((face * 8) + 4)) & 0xf);
 		}
+		track->tex_dirty = true;
 		break;
 	default:
 		printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
@@ -3318,9 +3337,9 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 	unsigned long size;
 	unsigned prim_walk;
 	unsigned nverts;
-	unsigned num_cb = track->num_cb;
+	unsigned num_cb = track->cb_dirty ? track->num_cb : 0;
 
-	if (!track->zb_cb_clear && !track->color_channel_mask &&
+	if (num_cb && !track->zb_cb_clear && !track->color_channel_mask &&
 	    !track->blend_read_enable)
 		num_cb = 0;
 
@@ -3341,7 +3360,9 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 			return -EINVAL;
 		}
 	}
-	if (track->z_enabled) {
+	track->cb_dirty = false;
+
+	if (track->zb_dirty && track->z_enabled) {
 		if (track->zb.robj == NULL) {
 			DRM_ERROR("[drm] No buffer for z buffer !\n");
 			return -EINVAL;
@@ -3358,6 +3379,28 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 			return -EINVAL;
 		}
 	}
+	track->zb_dirty = false;
+
+	if (track->aa_dirty && track->aaresolve) {
+		if (track->aa.robj == NULL) {
+			DRM_ERROR("[drm] No buffer for AA resolve buffer %d !\n", i);
+			return -EINVAL;
+		}
+		/* I believe the format comes from colorbuffer0. */
+		size = track->aa.pitch * track->cb[0].cpp * track->maxy;
+		size += track->aa.offset;
+		if (size > radeon_bo_size(track->aa.robj)) {
+			DRM_ERROR("[drm] Buffer too small for AA resolve buffer %d "
+				  "(need %lu have %lu) !\n", i, size,
+				  radeon_bo_size(track->aa.robj));
+			DRM_ERROR("[drm] AA resolve buffer %d (%u %u %u %u)\n",
+				  i, track->aa.pitch, track->cb[0].cpp,
+				  track->aa.offset, track->maxy);
+			return -EINVAL;
+		}
+	}
+	track->aa_dirty = false;
+
 	prim_walk = (track->vap_vf_cntl >> 4) & 0x3;
 	if (track->vap_vf_cntl & (1 << 14)) {
 		nverts = track->vap_alt_nverts;
@@ -3417,13 +3460,23 @@ int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track)
 			  prim_walk);
 		return -EINVAL;
 	}
-	return r100_cs_track_texture_check(rdev, track);
+
+	if (track->tex_dirty) {
+		track->tex_dirty = false;
+		return r100_cs_track_texture_check(rdev, track);
+	}
+	return 0;
 }
 
 void r100_cs_track_clear(struct radeon_device *rdev, struct r100_cs_track *track)
 {
 	unsigned i, face;
 
+	track->cb_dirty = true;
+	track->zb_dirty = true;
+	track->tex_dirty = true;
+	track->aa_dirty = true;
+
 	if (rdev->family < CHIP_R300) {
 		track->num_cb = 1;
 		if (rdev->family <= CHIP_RS200)
@@ -3437,6 +3490,8 @@ void r100_cs_track_clear(struct radeon_device *rdev, struct r100_cs_track *track
 		track->num_texture = 16;
 		track->maxy = 4096;
 		track->separate_cube = 0;
+		track->aaresolve = false;
+		track->aa.robj = NULL;
 	}
 
 	for (i = 0; i < track->num_cb; i++) {
@@ -3746,8 +3801,6 @@ static int r100_startup(struct radeon_device *rdev)
 	r100_mc_program(rdev);
 	/* Resume clock */
 	r100_clock_startup(rdev);
-	/* Initialize GPU configuration (# pipes, ...) */
-//	r100_gpu_init(rdev);
 	/* Initialize GART (initialize after TTM so we can allocate
 	 * memory through TTM but finalize after TTM) */
 	r100_enable_bm(rdev);
diff --git a/drivers/gpu/drm/radeon/r100_track.h b/drivers/gpu/drm/radeon/r100_track.h
index af65600e656..2fef9de7f36 100644
--- a/drivers/gpu/drm/radeon/r100_track.h
+++ b/drivers/gpu/drm/radeon/r100_track.h
@@ -52,14 +52,7 @@ struct r100_cs_track_texture {
 	unsigned                compress_format;
 };
 
-struct r100_cs_track_limits {
-	unsigned num_cb;
-	unsigned num_texture;
-	unsigned max_levels;
-};
-
 struct r100_cs_track {
-	struct radeon_device *rdev;
 	unsigned			num_cb;
 	unsigned                        num_texture;
 	unsigned			maxy;
@@ -73,11 +66,17 @@ struct r100_cs_track {
 	struct r100_cs_track_array	arrays[11];
 	struct r100_cs_track_cb 	cb[R300_MAX_CB];
 	struct r100_cs_track_cb 	zb;
+	struct r100_cs_track_cb 	aa;
 	struct r100_cs_track_texture	textures[R300_TRACK_MAX_TEXTURE];
 	bool				z_enabled;
 	bool                            separate_cube;
 	bool				zb_cb_clear;
 	bool				blend_read_enable;
+	bool				cb_dirty;
+	bool				zb_dirty;
+	bool				tex_dirty;
+	bool				aa_dirty;
+	bool				aaresolve;
 };
 
 int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track);
diff --git a/drivers/gpu/drm/radeon/r200.c b/drivers/gpu/drm/radeon/r200.c
index d2408c39561..f2405830041 100644
--- a/drivers/gpu/drm/radeon/r200.c
+++ b/drivers/gpu/drm/radeon/r200.c
@@ -184,6 +184,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->zb.robj = reloc->robj;
 		track->zb.offset = idx_value;
+		track->zb_dirty = true;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case RADEON_RB3D_COLOROFFSET:
@@ -196,6 +197,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->cb[0].robj = reloc->robj;
 		track->cb[0].offset = idx_value;
+		track->cb_dirty = true;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case R200_PP_TXOFFSET_0:
@@ -214,6 +216,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		}
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[i].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	case R200_PP_CUBIC_OFFSET_F1_0:
 	case R200_PP_CUBIC_OFFSET_F2_0:
@@ -257,9 +260,12 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		track->textures[i].cube_info[face - 1].offset = idx_value;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		track->textures[i].cube_info[face - 1].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	case RADEON_RE_WIDTH_HEIGHT:
 		track->maxy = ((idx_value >> 16) & 0x7FF);
+		track->cb_dirty = true;
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_COLORPITCH:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -280,9 +286,11 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		ib[idx] = tmp;
 
 		track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
+		track->cb_dirty = true;
 		break;
 	case RADEON_RB3D_DEPTHPITCH:
 		track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_CNTL:
 		switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
@@ -312,6 +320,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		}
 
 		track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
+		track->cb_dirty = true;
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_ZSTENCILCNTL:
 		switch (idx_value & 0xf) {
@@ -329,6 +339,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		default:
 			break;
 		}
+		track->zb_dirty = true;
 		break;
 	case RADEON_RB3D_ZPASS_ADDR:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -345,6 +356,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			uint32_t temp = idx_value >> 4;
 			for (i = 0; i < track->num_texture; i++)
 				track->textures[i].enabled = !!(temp & (1 << i));
+			track->tex_dirty = true;
 		}
 		break;
 	case RADEON_SE_VF_CNTL:
@@ -369,6 +381,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		i = (reg - R200_PP_TXSIZE_0) / 32;
 		track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
 		track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
+		track->tex_dirty = true;
 		break;
 	case R200_PP_TXPITCH_0:
 	case R200_PP_TXPITCH_1:
@@ -378,6 +391,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 	case R200_PP_TXPITCH_5:
 		i = (reg - R200_PP_TXPITCH_0) / 32;
 		track->textures[i].pitch = idx_value + 32;
+		track->tex_dirty = true;
 		break;
 	case R200_PP_TXFILTER_0:
 	case R200_PP_TXFILTER_1:
@@ -394,6 +408,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		tmp = (idx_value >> 27) & 0x7;
 		if (tmp == 2 || tmp == 6)
 			track->textures[i].roundup_h = false;
+		track->tex_dirty = true;
 		break;
 	case R200_PP_TXMULTI_CTL_0:
 	case R200_PP_TXMULTI_CTL_1:
@@ -432,6 +447,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			track->textures[i].tex_coord_type = 1;
 			break;
 		}
+		track->tex_dirty = true;
 		break;
 	case R200_PP_TXFORMAT_0:
 	case R200_PP_TXFORMAT_1:
@@ -488,6 +504,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
 		track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
+		track->tex_dirty = true;
 		break;
 	case R200_PP_CUBIC_FACES_0:
 	case R200_PP_CUBIC_FACES_1:
@@ -501,6 +518,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
 			track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
 			track->textures[i].cube_info[face].height = 1 << ((tmp >> ((face * 8) + 4)) & 0xf);
 		}
+		track->tex_dirty = true;
 		break;
 	default:
 		printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index 55fe5ba7def..069efa8c8ec 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -667,6 +667,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->cb[i].robj = reloc->robj;
 		track->cb[i].offset = idx_value;
+		track->cb_dirty = true;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case R300_ZB_DEPTHOFFSET:
@@ -679,6 +680,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		}
 		track->zb.robj = reloc->robj;
 		track->zb.offset = idx_value;
+		track->zb_dirty = true;
 		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
 		break;
 	case R300_TX_OFFSET_0:
@@ -717,6 +719,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		tmp |= tile_flags;
 		ib[idx] = tmp;
 		track->textures[i].robj = reloc->robj;
+		track->tex_dirty = true;
 		break;
 	/* Tracked registers */
 	case 0x2084:
@@ -743,6 +746,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		if (p->rdev->family < CHIP_RV515) {
 			track->maxy -= 1440;
 		}
+		track->cb_dirty = true;
+		track->zb_dirty = true;
 		break;
 	case 0x4E00:
 		/* RB3D_CCTL */
@@ -752,6 +757,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			return -EINVAL;
 		}
 		track->num_cb = ((idx_value >> 5) & 0x3) + 1;
+		track->cb_dirty = true;
 		break;
 	case 0x4E38:
 	case 0x4E3C:
@@ -814,6 +820,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 				  ((idx_value >> 21) & 0xF));
 			return -EINVAL;
 		}
+		track->cb_dirty = true;
 		break;
 	case 0x4F00:
 		/* ZB_CNTL */
@@ -822,6 +829,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		} else {
 			track->z_enabled = false;
 		}
+		track->zb_dirty = true;
 		break;
 	case 0x4F10:
 		/* ZB_FORMAT */
@@ -838,6 +846,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 				  (idx_value & 0xF));
 			return -EINVAL;
 		}
+		track->zb_dirty = true;
 		break;
 	case 0x4F24:
 		/* ZB_DEPTHPITCH */
@@ -861,14 +870,17 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		ib[idx] = tmp;
 
 		track->zb.pitch = idx_value & 0x3FFC;
+		track->zb_dirty = true;
 		break;
 	case 0x4104:
+		/* TX_ENABLE */
 		for (i = 0; i < 16; i++) {
 			bool enabled;
 
 			enabled = !!(idx_value & (1 << i));
 			track->textures[i].enabled = enabled;
 		}
+		track->tex_dirty = true;
 		break;
 	case 0x44C0:
 	case 0x44C4:
@@ -898,6 +910,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			track->textures[i].compress_format = R100_TRACK_COMP_NONE;
 			break;
 		case R300_TX_FORMAT_X16:
+		case R300_TX_FORMAT_FL_I16:
 		case R300_TX_FORMAT_Y8X8:
 		case R300_TX_FORMAT_Z5Y6X5:
 		case R300_TX_FORMAT_Z6Y5X5:
@@ -910,6 +923,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			track->textures[i].compress_format = R100_TRACK_COMP_NONE;
 			break;
 		case R300_TX_FORMAT_Y16X16:
+		case R300_TX_FORMAT_FL_I16A16:
 		case R300_TX_FORMAT_Z11Y11X10:
 		case R300_TX_FORMAT_Z10Y11X11:
 		case R300_TX_FORMAT_W8Z8Y8X8:
@@ -951,8 +965,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			DRM_ERROR("Invalid texture format %u\n",
 				  (idx_value & 0x1F));
 			return -EINVAL;
-			break;
 		}
+		track->tex_dirty = true;
 		break;
 	case 0x4400:
 	case 0x4404:
@@ -980,6 +994,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		if (tmp == 2 || tmp == 4 || tmp == 6) {
 			track->textures[i].roundup_h = false;
 		}
+		track->tex_dirty = true;
 		break;
 	case 0x4500:
 	case 0x4504:
@@ -1017,6 +1032,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 			DRM_ERROR("Forbidden bit TXFORMAT_MSB\n");
 			return -EINVAL;
 		}
+		track->tex_dirty = true;
 		break;
 	case 0x4480:
 	case 0x4484:
@@ -1046,6 +1062,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 		track->textures[i].use_pitch = !!tmp;
 		tmp = (idx_value >> 22) & 0xF;
 		track->textures[i].txdepth = tmp;
+		track->tex_dirty = true;
 		break;
 	case R300_ZB_ZPASS_ADDR:
 		r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1060,6 +1077,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x4e0c:
 		/* RB3D_COLOR_CHANNEL_MASK */
 		track->color_channel_mask = idx_value;
+		track->cb_dirty = true;
 		break;
 	case 0x43a4:
 		/* SC_HYPERZ_EN */
@@ -1073,6 +1091,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x4f1c:
 		/* ZB_BW_CNTL */
 		track->zb_cb_clear = !!(idx_value & (1 << 5));
+		track->cb_dirty = true;
+		track->zb_dirty = true;
 		if (p->rdev->hyperz_filp != p->filp) {
 			if (idx_value & (R300_HIZ_ENABLE |
 					 R300_RD_COMP_ENABLE |
@@ -1084,8 +1104,28 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
 	case 0x4e04:
 		/* RB3D_BLENDCNTL */
 		track->blend_read_enable = !!(idx_value & (1 << 2));
+		track->cb_dirty = true;
+		break;
+	case R300_RB3D_AARESOLVE_OFFSET:
+		r = r100_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
+				  idx, reg);
+			r100_cs_dump_packet(p, pkt);
+			return r;
+		}
+		track->aa.robj = reloc->robj;
+		track->aa.offset = idx_value;
+		track->aa_dirty = true;
+		ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
+		break;
+	case R300_RB3D_AARESOLVE_PITCH:
+		track->aa.pitch = idx_value & 0x3FFE;
+		track->aa_dirty = true;
 		break;
-	case 0x4f28: /* ZB_DEPTHCLEARVALUE */
+	case R300_RB3D_AARESOLVE_CTL:
+		track->aaresolve = idx_value & 0x1;
+		track->aa_dirty = true;
 		break;
 	case 0x4f30: /* ZB_MASK_OFFSET */
 	case 0x4f34: /* ZB_ZMASK_PITCH */
diff --git a/drivers/gpu/drm/radeon/r300_reg.h b/drivers/gpu/drm/radeon/r300_reg.h
index 1a0d5362cd7..f0bce399c9f 100644
--- a/drivers/gpu/drm/radeon/r300_reg.h
+++ b/drivers/gpu/drm/radeon/r300_reg.h
@@ -1371,6 +1371,8 @@
 #define R300_RB3D_COLORPITCH2               0x4E40 /* GUESS */
 #define R300_RB3D_COLORPITCH3               0x4E44 /* GUESS */
 
+#define R300_RB3D_AARESOLVE_OFFSET          0x4E80
+#define R300_RB3D_AARESOLVE_PITCH           0x4E84
 #define R300_RB3D_AARESOLVE_CTL             0x4E88
 /* gap */
 
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index 1cd56dc8c8a..b409b24207a 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -2106,7 +2106,11 @@ static int r600_cp_load_microcode(struct radeon_device *rdev)
 
 	r600_cp_stop(rdev);
 
-	WREG32(CP_RB_CNTL, RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3));
+	WREG32(CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+	       BUF_SWAP_32BIT |
+#endif
+	       RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3));
 
 	/* Reset cp */
 	WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP);
@@ -2193,7 +2197,11 @@ int r600_cp_resume(struct radeon_device *rdev)
 	WREG32(CP_RB_WPTR, 0);
 
 	/* set the wb address whether it's enabled or not */
-	WREG32(CP_RB_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC);
+	WREG32(CP_RB_RPTR_ADDR,
+#ifdef __BIG_ENDIAN
+	       RB_RPTR_SWAP(2) |
+#endif
+	       ((rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC));
 	WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF);
 	WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF);
 
@@ -2629,7 +2637,11 @@ void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
 {
 	/* FIXME: implement */
 	radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
-	radeon_ring_write(rdev, ib->gpu_addr & 0xFFFFFFFC);
+	radeon_ring_write(rdev,
+#ifdef __BIG_ENDIAN
+			  (2 << 0) |
+#endif
+			  (ib->gpu_addr & 0xFFFFFFFC));
 	radeon_ring_write(rdev, upper_32_bits(ib->gpu_addr) & 0xFF);
 	radeon_ring_write(rdev, ib->length_dw);
 }
@@ -3305,8 +3317,8 @@ restart_ih:
 	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;
+		src_id = le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff;
+		src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff;
 
 		switch (src_id) {
 		case 1: /* D1 vblank/vline */
diff --git a/drivers/gpu/drm/radeon/r600_blit.c b/drivers/gpu/drm/radeon/r600_blit.c
index ca5c29f7077..7f1043448d2 100644
--- a/drivers/gpu/drm/radeon/r600_blit.c
+++ b/drivers/gpu/drm/radeon/r600_blit.c
@@ -137,9 +137,9 @@ set_shaders(struct drm_device *dev)
 	ps = (u32 *) ((char *)dev->agp_buffer_map->handle + dev_priv->blit_vb->offset + 256);
 
 	for (i = 0; i < r6xx_vs_size; i++)
-		vs[i] = r6xx_vs[i];
+		vs[i] = cpu_to_le32(r6xx_vs[i]);
 	for (i = 0; i < r6xx_ps_size; i++)
-		ps[i] = r6xx_ps[i];
+		ps[i] = cpu_to_le32(r6xx_ps[i]);
 
 	dev_priv->blit_vb->used = 512;
 
@@ -192,6 +192,9 @@ set_vtx_resource(drm_radeon_private_t *dev_priv, u64 gpu_addr)
 	DRM_DEBUG("\n");
 
 	sq_vtx_constant_word2 = (((gpu_addr >> 32) & 0xff) | (16 << 8));
+#ifdef __BIG_ENDIAN
+	sq_vtx_constant_word2 |= (2 << 30);
+#endif
 
 	BEGIN_RING(9);
 	OUT_RING(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
@@ -291,7 +294,11 @@ draw_auto(drm_radeon_private_t *dev_priv)
 	OUT_RING(DI_PT_RECTLIST);
 
 	OUT_RING(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
+#ifdef __BIG_ENDIAN
+	OUT_RING((2 << 2) | DI_INDEX_SIZE_16_BIT);
+#else
 	OUT_RING(DI_INDEX_SIZE_16_BIT);
+#endif
 
 	OUT_RING(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
 	OUT_RING(1);
diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/radeon/r600_blit_kms.c
index 16e211a614d..2fed9175012 100644
--- a/drivers/gpu/drm/radeon/r600_blit_kms.c
+++ b/drivers/gpu/drm/radeon/r600_blit_kms.c
@@ -54,7 +54,7 @@ set_render_target(struct radeon_device *rdev, int format,
 	if (h < 8)
 		h = 8;
 
-	cb_color_info = ((format << 2) | (1 << 27));
+	cb_color_info = ((format << 2) | (1 << 27) | (1 << 8));
 	pitch = (w / 8) - 1;
 	slice = ((w * h) / 64) - 1;
 
@@ -165,6 +165,9 @@ set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
 	u32 sq_vtx_constant_word2;
 
 	sq_vtx_constant_word2 = ((upper_32_bits(gpu_addr) & 0xff) | (16 << 8));
+#ifdef __BIG_ENDIAN
+	sq_vtx_constant_word2 |= (2 << 30);
+#endif
 
 	radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 7));
 	radeon_ring_write(rdev, 0x460);
@@ -199,7 +202,7 @@ set_tex_resource(struct radeon_device *rdev,
 	if (h < 1)
 		h = 1;
 
-	sq_tex_resource_word0 = (1 << 0);
+	sq_tex_resource_word0 = (1 << 0) | (1 << 3);
 	sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 8) |
 				  ((w - 1) << 19));
 
@@ -253,7 +256,11 @@ draw_auto(struct radeon_device *rdev)
 	radeon_ring_write(rdev, DI_PT_RECTLIST);
 
 	radeon_ring_write(rdev, PACKET3(PACKET3_INDEX_TYPE, 0));
-	radeon_ring_write(rdev, DI_INDEX_SIZE_16_BIT);
+	radeon_ring_write(rdev,
+#ifdef __BIG_ENDIAN
+			  (2 << 2) |
+#endif
+			  DI_INDEX_SIZE_16_BIT);
 
 	radeon_ring_write(rdev, PACKET3(PACKET3_NUM_INSTANCES, 0));
 	radeon_ring_write(rdev, 1);
@@ -424,7 +431,11 @@ set_default_state(struct radeon_device *rdev)
 	dwords = ALIGN(rdev->r600_blit.state_len, 0x10);
 	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
 	radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
-	radeon_ring_write(rdev, gpu_addr & 0xFFFFFFFC);
+	radeon_ring_write(rdev,
+#ifdef __BIG_ENDIAN
+			  (2 << 0) |
+#endif
+			  (gpu_addr & 0xFFFFFFFC));
 	radeon_ring_write(rdev, upper_32_bits(gpu_addr) & 0xFF);
 	radeon_ring_write(rdev, dwords);
 
@@ -467,7 +478,7 @@ static inline uint32_t i2f(uint32_t input)
 int r600_blit_init(struct radeon_device *rdev)
 {
 	u32 obj_size;
-	int r, dwords;
+	int i, r, dwords;
 	void *ptr;
 	u32 packet2s[16];
 	int num_packet2s = 0;
@@ -486,7 +497,7 @@ int r600_blit_init(struct radeon_device *rdev)
 
 	dwords = rdev->r600_blit.state_len;
 	while (dwords & 0xf) {
-		packet2s[num_packet2s++] = PACKET2(0);
+		packet2s[num_packet2s++] = cpu_to_le32(PACKET2(0));
 		dwords++;
 	}
 
@@ -529,8 +540,10 @@ int r600_blit_init(struct radeon_device *rdev)
 	if (num_packet2s)
 		memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
 			    packet2s, num_packet2s * 4);
-	memcpy(ptr + rdev->r600_blit.vs_offset, r6xx_vs, r6xx_vs_size * 4);
-	memcpy(ptr + rdev->r600_blit.ps_offset, r6xx_ps, r6xx_ps_size * 4);
+	for (i = 0; i < r6xx_vs_size; i++)
+		*(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(r6xx_vs[i]);
+	for (i = 0; i < r6xx_ps_size; i++)
+		*(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(r6xx_ps[i]);
 	radeon_bo_kunmap(rdev->r600_blit.shader_obj);
 	radeon_bo_unreserve(rdev->r600_blit.shader_obj);
 
diff --git a/drivers/gpu/drm/radeon/r600_blit_shaders.c b/drivers/gpu/drm/radeon/r600_blit_shaders.c
index e8151c1d55b..2d1f6c5ee2a 100644
--- a/drivers/gpu/drm/radeon/r600_blit_shaders.c
+++ b/drivers/gpu/drm/radeon/r600_blit_shaders.c
@@ -684,7 +684,11 @@ const u32 r6xx_vs[] =
 	0x00000000,
 	0x3c000000,
 	0x68cd1000,
+#ifdef __BIG_ENDIAN
+	0x000a0000,
+#else
 	0x00080000,
+#endif
 	0x00000000,
 };
 
diff --git a/drivers/gpu/drm/radeon/r600_cp.c b/drivers/gpu/drm/radeon/r600_cp.c
index 4f4cd8b286d..c3ab959bdc7 100644
--- a/drivers/gpu/drm/radeon/r600_cp.c
+++ b/drivers/gpu/drm/radeon/r600_cp.c
@@ -396,6 +396,9 @@ static void r600_cp_load_microcode(drm_radeon_private_t *dev_priv)
 	r600_do_cp_stop(dev_priv);
 
 	RADEON_WRITE(R600_CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+		     R600_BUF_SWAP_32BIT |
+#endif
 		     R600_RB_NO_UPDATE |
 		     R600_RB_BLKSZ(15) |
 		     R600_RB_BUFSZ(3));
@@ -486,9 +489,12 @@ static void r700_cp_load_microcode(drm_radeon_private_t *dev_priv)
 	r600_do_cp_stop(dev_priv);
 
 	RADEON_WRITE(R600_CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+		     R600_BUF_SWAP_32BIT |
+#endif
 		     R600_RB_NO_UPDATE |
-		     (15 << 8) |
-		     (3 << 0));
+		     R600_RB_BLKSZ(15) |
+		     R600_RB_BUFSZ(3));
 
 	RADEON_WRITE(R600_GRBM_SOFT_RESET, R600_SOFT_RESET_CP);
 	RADEON_READ(R600_GRBM_SOFT_RESET);
@@ -550,8 +556,12 @@ static void r600_test_writeback(drm_radeon_private_t *dev_priv)
 
 	if (!dev_priv->writeback_works) {
 		/* Disable writeback to avoid unnecessary bus master transfer */
-		RADEON_WRITE(R600_CP_RB_CNTL, RADEON_READ(R600_CP_RB_CNTL) |
-			     RADEON_RB_NO_UPDATE);
+		RADEON_WRITE(R600_CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+			     R600_BUF_SWAP_32BIT |
+#endif
+			     RADEON_READ(R600_CP_RB_CNTL) |
+			     R600_RB_NO_UPDATE);
 		RADEON_WRITE(R600_SCRATCH_UMSK, 0);
 	}
 }
@@ -575,7 +585,11 @@ int r600_do_engine_reset(struct drm_device *dev)
 
 	RADEON_WRITE(R600_CP_RB_WPTR_DELAY, 0);
 	cp_rb_cntl = RADEON_READ(R600_CP_RB_CNTL);
-	RADEON_WRITE(R600_CP_RB_CNTL, R600_RB_RPTR_WR_ENA);
+	RADEON_WRITE(R600_CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+		     R600_BUF_SWAP_32BIT |
+#endif
+		     R600_RB_RPTR_WR_ENA);
 
 	RADEON_WRITE(R600_CP_RB_RPTR_WR, cp_ptr);
 	RADEON_WRITE(R600_CP_RB_WPTR, cp_ptr);
@@ -1838,7 +1852,10 @@ static void r600_cp_init_ring_buffer(struct drm_device *dev,
 			+ dev_priv->gart_vm_start;
 	}
 	RADEON_WRITE(R600_CP_RB_RPTR_ADDR,
-		     rptr_addr & 0xffffffff);
+#ifdef __BIG_ENDIAN
+		     (2 << 0) |
+#endif
+		     (rptr_addr & 0xfffffffc));
 	RADEON_WRITE(R600_CP_RB_RPTR_ADDR_HI,
 		     upper_32_bits(rptr_addr));
 
@@ -1889,7 +1906,7 @@ static void r600_cp_init_ring_buffer(struct drm_device *dev,
 	{
 		u64 scratch_addr;
 
-		scratch_addr = RADEON_READ(R600_CP_RB_RPTR_ADDR);
+		scratch_addr = RADEON_READ(R600_CP_RB_RPTR_ADDR) & 0xFFFFFFFC;
 		scratch_addr |= ((u64)RADEON_READ(R600_CP_RB_RPTR_ADDR_HI)) << 32;
 		scratch_addr += R600_SCRATCH_REG_OFFSET;
 		scratch_addr >>= 8;
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index fe0c8eb7601..0a0848f0346 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -388,17 +388,18 @@ static inline int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i)
 	}
 
 	if (!IS_ALIGNED(pitch, pitch_align)) {
-		dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n",
-			 __func__, __LINE__, pitch);
+		dev_warn(p->dev, "%s:%d cb pitch (%d, 0x%x, %d) invalid\n",
+			 __func__, __LINE__, pitch, pitch_align, array_mode);
 		return -EINVAL;
 	}
 	if (!IS_ALIGNED(height, height_align)) {
-		dev_warn(p->dev, "%s:%d cb height (%d) invalid\n",
-			 __func__, __LINE__, height);
+		dev_warn(p->dev, "%s:%d cb height (%d, 0x%x, %d) invalid\n",
+			 __func__, __LINE__, height, height_align, array_mode);
 		return -EINVAL;
 	}
 	if (!IS_ALIGNED(base_offset, base_align)) {
-		dev_warn(p->dev, "%s offset[%d] 0x%llx not aligned\n", __func__, i, base_offset);
+		dev_warn(p->dev, "%s offset[%d] 0x%llx 0x%llx, %d not aligned\n", __func__, i,
+			 base_offset, base_align, array_mode);
 		return -EINVAL;
 	}
 
@@ -413,7 +414,10 @@ static inline int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i)
 			 * broken userspace.
 			 */
 		} else {
-			dev_warn(p->dev, "%s offset[%d] %d %d %lu too big\n", __func__, i, track->cb_color_bo_offset[i], tmp, radeon_bo_size(track->cb_color_bo[i]));
+			dev_warn(p->dev, "%s offset[%d] %d %d %d %lu too big\n", __func__, i,
+				 array_mode,
+				 track->cb_color_bo_offset[i], tmp,
+				 radeon_bo_size(track->cb_color_bo[i]));
 			return -EINVAL;
 		}
 	}
@@ -548,17 +552,18 @@ static int r600_cs_track_check(struct radeon_cs_parser *p)
 			}
 
 			if (!IS_ALIGNED(pitch, pitch_align)) {
-				dev_warn(p->dev, "%s:%d db pitch (%d) invalid\n",
-					 __func__, __LINE__, pitch);
+				dev_warn(p->dev, "%s:%d db pitch (%d, 0x%x, %d) invalid\n",
+					 __func__, __LINE__, pitch, pitch_align, array_mode);
 				return -EINVAL;
 			}
 			if (!IS_ALIGNED(height, height_align)) {
-				dev_warn(p->dev, "%s:%d db height (%d) invalid\n",
-					 __func__, __LINE__, height);
+				dev_warn(p->dev, "%s:%d db height (%d, 0x%x, %d) invalid\n",
+					 __func__, __LINE__, height, height_align, array_mode);
 				return -EINVAL;
 			}
 			if (!IS_ALIGNED(base_offset, base_align)) {
-				dev_warn(p->dev, "%s offset[%d] 0x%llx not aligned\n", __func__, i, base_offset);
+				dev_warn(p->dev, "%s offset[%d] 0x%llx, 0x%llx, %d not aligned\n", __func__, i,
+					 base_offset, base_align, array_mode);
 				return -EINVAL;
 			}
 
@@ -566,9 +571,10 @@ static int r600_cs_track_check(struct radeon_cs_parser *p)
 			nviews = G_028004_SLICE_MAX(track->db_depth_view) + 1;
 			tmp = ntiles * bpe * 64 * nviews;
 			if ((tmp + track->db_offset) > radeon_bo_size(track->db_bo)) {
-				dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %d -> %u have %lu)\n",
-						track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset,
-						radeon_bo_size(track->db_bo));
+				dev_warn(p->dev, "z/stencil buffer (%d) too small (0x%08X %d %d %d -> %u have %lu)\n",
+					 array_mode,
+					 track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset,
+					 radeon_bo_size(track->db_bo));
 				return -EINVAL;
 			}
 		}
@@ -1350,18 +1356,18 @@ static inline int r600_check_texture_resource(struct radeon_cs_parser *p,  u32 i
 	/* XXX check height as well... */
 
 	if (!IS_ALIGNED(pitch, pitch_align)) {
-		dev_warn(p->dev, "%s:%d tex pitch (%d) invalid\n",
-			 __func__, __LINE__, pitch);
+		dev_warn(p->dev, "%s:%d tex pitch (%d, 0x%x, %d) invalid\n",
+			 __func__, __LINE__, pitch, pitch_align, G_038000_TILE_MODE(word0));
 		return -EINVAL;
 	}
 	if (!IS_ALIGNED(base_offset, base_align)) {
-		dev_warn(p->dev, "%s:%d tex base offset (0x%llx) invalid\n",
-			 __func__, __LINE__, base_offset);
+		dev_warn(p->dev, "%s:%d tex base offset (0x%llx, 0x%llx, %d) invalid\n",
+			 __func__, __LINE__, base_offset, base_align, G_038000_TILE_MODE(word0));
 		return -EINVAL;
 	}
 	if (!IS_ALIGNED(mip_offset, base_align)) {
-		dev_warn(p->dev, "%s:%d tex mip offset (0x%llx) invalid\n",
-			 __func__, __LINE__, mip_offset);
+		dev_warn(p->dev, "%s:%d tex mip offset (0x%llx, 0x%llx, %d) invalid\n",
+			 __func__, __LINE__, mip_offset, base_align, G_038000_TILE_MODE(word0));
 		return -EINVAL;
 	}
 
diff --git a/drivers/gpu/drm/radeon/r600d.h b/drivers/gpu/drm/radeon/r600d.h
index d1f598663da..b2b944bcd05 100644
--- a/drivers/gpu/drm/radeon/r600d.h
+++ b/drivers/gpu/drm/radeon/r600d.h
@@ -154,13 +154,14 @@
 #define		ROQ_IB2_START(x)				((x) << 8)
 #define	CP_RB_BASE					0xC100
 #define	CP_RB_CNTL					0xC104
-#define		RB_BUFSZ(x)					((x)<<0)
-#define		RB_BLKSZ(x)					((x)<<8)
-#define		RB_NO_UPDATE					(1<<27)
-#define		RB_RPTR_WR_ENA					(1<<31)
+#define		RB_BUFSZ(x)					((x) << 0)
+#define		RB_BLKSZ(x)					((x) << 8)
+#define		RB_NO_UPDATE					(1 << 27)
+#define		RB_RPTR_WR_ENA					(1 << 31)
 #define		BUF_SWAP_32BIT					(2 << 16)
 #define	CP_RB_RPTR					0x8700
 #define	CP_RB_RPTR_ADDR					0xC10C
+#define		RB_RPTR_SWAP(x)					((x) << 0)
 #define	CP_RB_RPTR_ADDR_HI				0xC110
 #define	CP_RB_RPTR_WR					0xC108
 #define	CP_RB_WPTR					0xC114
diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c
index 5c1cc7ad9a1..02d5c415f49 100644
--- a/drivers/gpu/drm/radeon/radeon_atombios.c
+++ b/drivers/gpu/drm/radeon/radeon_atombios.c
@@ -88,7 +88,7 @@ static inline struct radeon_i2c_bus_rec radeon_lookup_i2c_gpio(struct radeon_dev
 			/* some evergreen boards have bad data for this entry */
 			if (ASIC_IS_DCE4(rdev)) {
 				if ((i == 7) &&
-				    (gpio->usClkMaskRegisterIndex == 0x1936) &&
+				    (le16_to_cpu(gpio->usClkMaskRegisterIndex) == 0x1936) &&
 				    (gpio->sucI2cId.ucAccess == 0)) {
 					gpio->sucI2cId.ucAccess = 0x97;
 					gpio->ucDataMaskShift = 8;
@@ -101,7 +101,7 @@ static inline struct radeon_i2c_bus_rec radeon_lookup_i2c_gpio(struct radeon_dev
 			/* some DCE3 boards have bad data for this entry */
 			if (ASIC_IS_DCE3(rdev)) {
 				if ((i == 4) &&
-				    (gpio->usClkMaskRegisterIndex == 0x1fda) &&
+				    (le16_to_cpu(gpio->usClkMaskRegisterIndex) == 0x1fda) &&
 				    (gpio->sucI2cId.ucAccess == 0x94))
 					gpio->sucI2cId.ucAccess = 0x14;
 			}
@@ -172,7 +172,7 @@ void radeon_atombios_i2c_init(struct radeon_device *rdev)
 			/* some evergreen boards have bad data for this entry */
 			if (ASIC_IS_DCE4(rdev)) {
 				if ((i == 7) &&
-				    (gpio->usClkMaskRegisterIndex == 0x1936) &&
+				    (le16_to_cpu(gpio->usClkMaskRegisterIndex) == 0x1936) &&
 				    (gpio->sucI2cId.ucAccess == 0)) {
 					gpio->sucI2cId.ucAccess = 0x97;
 					gpio->ucDataMaskShift = 8;
@@ -185,7 +185,7 @@ void radeon_atombios_i2c_init(struct radeon_device *rdev)
 			/* some DCE3 boards have bad data for this entry */
 			if (ASIC_IS_DCE3(rdev)) {
 				if ((i == 4) &&
-				    (gpio->usClkMaskRegisterIndex == 0x1fda) &&
+				    (le16_to_cpu(gpio->usClkMaskRegisterIndex) == 0x1fda) &&
 				    (gpio->sucI2cId.ucAccess == 0x94))
 					gpio->sucI2cId.ucAccess = 0x14;
 			}
@@ -252,7 +252,7 @@ static inline struct radeon_gpio_rec radeon_lookup_gpio(struct radeon_device *rd
 			pin = &gpio_info->asGPIO_Pin[i];
 			if (id == pin->ucGPIO_ID) {
 				gpio.id = pin->ucGPIO_ID;
-				gpio.reg = pin->usGpioPin_AIndex * 4;
+				gpio.reg = le16_to_cpu(pin->usGpioPin_AIndex) * 4;
 				gpio.mask = (1 << pin->ucGpioPinBitShift);
 				gpio.valid = true;
 				break;
@@ -1274,11 +1274,11 @@ bool radeon_atombios_sideport_present(struct radeon_device *rdev)
 				      data_offset);
 		switch (crev) {
 		case 1:
-			if (igp_info->info.ulBootUpMemoryClock)
+			if (le32_to_cpu(igp_info->info.ulBootUpMemoryClock))
 				return true;
 			break;
 		case 2:
-			if (igp_info->info_2.ulBootUpSidePortClock)
+			if (le32_to_cpu(igp_info->info_2.ulBootUpSidePortClock))
 				return true;
 			break;
 		default:
@@ -1442,7 +1442,7 @@ bool radeon_atombios_get_asic_ss_info(struct radeon_device *rdev,
 
 			for (i = 0; i < num_indices; i++) {
 				if ((ss_info->info.asSpreadSpectrum[i].ucClockIndication == id) &&
-				    (clock <= ss_info->info.asSpreadSpectrum[i].ulTargetClockRange)) {
+				    (clock <= le32_to_cpu(ss_info->info.asSpreadSpectrum[i].ulTargetClockRange))) {
 					ss->percentage =
 						le16_to_cpu(ss_info->info.asSpreadSpectrum[i].usSpreadSpectrumPercentage);
 					ss->type = ss_info->info.asSpreadSpectrum[i].ucSpreadSpectrumMode;
@@ -1456,7 +1456,7 @@ bool radeon_atombios_get_asic_ss_info(struct radeon_device *rdev,
 				sizeof(ATOM_ASIC_SS_ASSIGNMENT_V2);
 			for (i = 0; i < num_indices; i++) {
 				if ((ss_info->info_2.asSpreadSpectrum[i].ucClockIndication == id) &&
-				    (clock <= ss_info->info_2.asSpreadSpectrum[i].ulTargetClockRange)) {
+				    (clock <= le32_to_cpu(ss_info->info_2.asSpreadSpectrum[i].ulTargetClockRange))) {
 					ss->percentage =
 						le16_to_cpu(ss_info->info_2.asSpreadSpectrum[i].usSpreadSpectrumPercentage);
 					ss->type = ss_info->info_2.asSpreadSpectrum[i].ucSpreadSpectrumMode;
@@ -1470,7 +1470,7 @@ bool radeon_atombios_get_asic_ss_info(struct radeon_device *rdev,
 				sizeof(ATOM_ASIC_SS_ASSIGNMENT_V3);
 			for (i = 0; i < num_indices; i++) {
 				if ((ss_info->info_3.asSpreadSpectrum[i].ucClockIndication == id) &&
-				    (clock <= ss_info->info_3.asSpreadSpectrum[i].ulTargetClockRange)) {
+				    (clock <= le32_to_cpu(ss_info->info_3.asSpreadSpectrum[i].ulTargetClockRange))) {
 					ss->percentage =
 						le16_to_cpu(ss_info->info_3.asSpreadSpectrum[i].usSpreadSpectrumPercentage);
 					ss->type = ss_info->info_3.asSpreadSpectrum[i].ucSpreadSpectrumMode;
@@ -1553,8 +1553,8 @@ struct radeon_encoder_atom_dig *radeon_atombios_get_lvds_info(struct
 		if (misc & ATOM_DOUBLE_CLOCK_MODE)
 			lvds->native_mode.flags |= DRM_MODE_FLAG_DBLSCAN;
 
-		lvds->native_mode.width_mm = lvds_info->info.sLCDTiming.usImageHSize;
-		lvds->native_mode.height_mm = lvds_info->info.sLCDTiming.usImageVSize;
+		lvds->native_mode.width_mm = le16_to_cpu(lvds_info->info.sLCDTiming.usImageHSize);
+		lvds->native_mode.height_mm = le16_to_cpu(lvds_info->info.sLCDTiming.usImageVSize);
 
 		/* set crtc values */
 		drm_mode_set_crtcinfo(&lvds->native_mode, CRTC_INTERLACE_HALVE_V);
@@ -1569,13 +1569,13 @@ struct radeon_encoder_atom_dig *radeon_atombios_get_lvds_info(struct
 			lvds->linkb = false;
 
 		/* parse the lcd record table */
-		if (lvds_info->info.usModePatchTableOffset) {
+		if (le16_to_cpu(lvds_info->info.usModePatchTableOffset)) {
 			ATOM_FAKE_EDID_PATCH_RECORD *fake_edid_record;
 			ATOM_PANEL_RESOLUTION_PATCH_RECORD *panel_res_record;
 			bool bad_record = false;
 			u8 *record = (u8 *)(mode_info->atom_context->bios +
 					    data_offset +
-					    lvds_info->info.usModePatchTableOffset);
+					    le16_to_cpu(lvds_info->info.usModePatchTableOffset));
 			while (*record != ATOM_RECORD_END_TYPE) {
 				switch (*record) {
 				case LCD_MODE_PATCH_RECORD_MODE_TYPE:
@@ -2189,7 +2189,7 @@ static u16 radeon_atombios_get_default_vddc(struct radeon_device *rdev)
 		firmware_info =
 			(union firmware_info *)(mode_info->atom_context->bios +
 						data_offset);
-		vddc = firmware_info->info_14.usBootUpVDDCVoltage;
+		vddc = le16_to_cpu(firmware_info->info_14.usBootUpVDDCVoltage);
 	}
 
 	return vddc;
@@ -2284,7 +2284,7 @@ static bool radeon_atombios_parse_pplib_clock_info(struct radeon_device *rdev,
 		rdev->pm.power_state[state_index].clock_info[mode_index].voltage.type =
 			VOLTAGE_SW;
 		rdev->pm.power_state[state_index].clock_info[mode_index].voltage.voltage =
-			clock_info->evergreen.usVDDC;
+			le16_to_cpu(clock_info->evergreen.usVDDC);
 	} else {
 		sclk = le16_to_cpu(clock_info->r600.usEngineClockLow);
 		sclk |= clock_info->r600.ucEngineClockHigh << 16;
@@ -2295,7 +2295,7 @@ static bool radeon_atombios_parse_pplib_clock_info(struct radeon_device *rdev,
 		rdev->pm.power_state[state_index].clock_info[mode_index].voltage.type =
 			VOLTAGE_SW;
 		rdev->pm.power_state[state_index].clock_info[mode_index].voltage.voltage =
-			clock_info->r600.usVDDC;
+			le16_to_cpu(clock_info->r600.usVDDC);
 	}
 
 	if (rdev->flags & RADEON_IS_IGP) {
@@ -2408,13 +2408,13 @@ static int radeon_atombios_parse_power_table_6(struct radeon_device *rdev)
 	radeon_atombios_add_pplib_thermal_controller(rdev, &power_info->pplib.sThermalController);
 	state_array = (struct StateArray *)
 		(mode_info->atom_context->bios + data_offset +
-		 power_info->pplib.usStateArrayOffset);
+		 le16_to_cpu(power_info->pplib.usStateArrayOffset));
 	clock_info_array = (struct ClockInfoArray *)
 		(mode_info->atom_context->bios + data_offset +
-		 power_info->pplib.usClockInfoArrayOffset);
+		 le16_to_cpu(power_info->pplib.usClockInfoArrayOffset));
 	non_clock_info_array = (struct NonClockInfoArray *)
 		(mode_info->atom_context->bios + data_offset +
-		 power_info->pplib.usNonClockInfoArrayOffset);
+		 le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset));
 	rdev->pm.power_state = kzalloc(sizeof(struct radeon_power_state) *
 				       state_array->ucNumEntries, GFP_KERNEL);
 	if (!rdev->pm.power_state)
@@ -2533,7 +2533,7 @@ uint32_t radeon_atom_get_engine_clock(struct radeon_device *rdev)
 	int index = GetIndexIntoMasterTable(COMMAND, GetEngineClock);
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
-	return args.ulReturnEngineClock;
+	return le32_to_cpu(args.ulReturnEngineClock);
 }
 
 uint32_t radeon_atom_get_memory_clock(struct radeon_device *rdev)
@@ -2542,7 +2542,7 @@ uint32_t radeon_atom_get_memory_clock(struct radeon_device *rdev)
 	int index = GetIndexIntoMasterTable(COMMAND, GetMemoryClock);
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
-	return args.ulReturnMemoryClock;
+	return le32_to_cpu(args.ulReturnMemoryClock);
 }
 
 void radeon_atom_set_engine_clock(struct radeon_device *rdev,
@@ -2551,7 +2551,7 @@ void radeon_atom_set_engine_clock(struct radeon_device *rdev,
 	SET_ENGINE_CLOCK_PS_ALLOCATION args;
 	int index = GetIndexIntoMasterTable(COMMAND, SetEngineClock);
 
-	args.ulTargetEngineClock = eng_clock;	/* 10 khz */
+	args.ulTargetEngineClock = cpu_to_le32(eng_clock);	/* 10 khz */
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
@@ -2565,7 +2565,7 @@ void radeon_atom_set_memory_clock(struct radeon_device *rdev,
 	if (rdev->flags & RADEON_IS_IGP)
 		return;
 
-	args.ulTargetMemoryClock = mem_clock;	/* 10 khz */
+	args.ulTargetMemoryClock = cpu_to_le32(mem_clock);	/* 10 khz */
 
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
diff --git a/drivers/gpu/drm/radeon/radeon_combios.c b/drivers/gpu/drm/radeon/radeon_combios.c
index d27ef74590c..cf7c8d5b4ec 100644
--- a/drivers/gpu/drm/radeon/radeon_combios.c
+++ b/drivers/gpu/drm/radeon/radeon_combios.c
@@ -1504,6 +1504,11 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 			   (rdev->pdev->subsystem_device == 0x4a48)) {
 			/* Mac X800 */
 			rdev->mode_info.connector_table = CT_MAC_X800;
+		} else if ((rdev->pdev->device == 0x4150) &&
+			   (rdev->pdev->subsystem_vendor == 0x1002) &&
+			   (rdev->pdev->subsystem_device == 0x4150)) {
+			/* Mac G5 9600 */
+			rdev->mode_info.connector_table = CT_MAC_G5_9600;
 		} else
 #endif /* CONFIG_PPC_PMAC */
 #ifdef CONFIG_PPC64
@@ -2022,6 +2027,48 @@ bool radeon_get_legacy_connector_info_from_table(struct drm_device *dev)
 					    CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I,
 					    &hpd);
 		break;
+	case CT_MAC_G5_9600:
+		DRM_INFO("Connector Table: %d (mac g5 9600)\n",
+			 rdev->mode_info.connector_table);
+		/* DVI - tv dac, dvo */
+		ddc_i2c = combios_setup_i2c_bus(rdev, DDC_DVI, 0, 0);
+		hpd.hpd = RADEON_HPD_1; /* ??? */
+		radeon_add_legacy_encoder(dev,
+					  radeon_get_encoder_enum(dev,
+								  ATOM_DEVICE_DFP2_SUPPORT,
+								  0),
+					  ATOM_DEVICE_DFP2_SUPPORT);
+		radeon_add_legacy_encoder(dev,
+					  radeon_get_encoder_enum(dev,
+								  ATOM_DEVICE_CRT2_SUPPORT,
+								  2),
+					  ATOM_DEVICE_CRT2_SUPPORT);
+		radeon_add_legacy_connector(dev, 0,
+					    ATOM_DEVICE_DFP2_SUPPORT |
+					    ATOM_DEVICE_CRT2_SUPPORT,
+					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+					    &hpd);
+		/* ADC - primary dac, internal tmds */
+		ddc_i2c = combios_setup_i2c_bus(rdev, DDC_VGA, 0, 0);
+		hpd.hpd = RADEON_HPD_2; /* ??? */
+		radeon_add_legacy_encoder(dev,
+					  radeon_get_encoder_enum(dev,
+								  ATOM_DEVICE_DFP1_SUPPORT,
+								  0),
+					  ATOM_DEVICE_DFP1_SUPPORT);
+		radeon_add_legacy_encoder(dev,
+					  radeon_get_encoder_enum(dev,
+								  ATOM_DEVICE_CRT1_SUPPORT,
+								  1),
+					  ATOM_DEVICE_CRT1_SUPPORT);
+		radeon_add_legacy_connector(dev, 1,
+					    ATOM_DEVICE_DFP1_SUPPORT |
+					    ATOM_DEVICE_CRT1_SUPPORT,
+					    DRM_MODE_CONNECTOR_DVII, &ddc_i2c,
+					    CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I,
+					    &hpd);
+		break;
 	default:
 		DRM_INFO("Connector table: %d (invalid)\n",
 			 rdev->mode_info.connector_table);
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index 0ca5eb21792..f0209be7a34 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -937,8 +937,11 @@ int radeon_resume_kms(struct drm_device *dev)
 int radeon_gpu_reset(struct radeon_device *rdev)
 {
 	int r;
+	int resched;
 
 	radeon_save_bios_scratch_regs(rdev);
+	/* block TTM */
+	resched = ttm_bo_lock_delayed_workqueue(&rdev->mman.bdev);
 	radeon_suspend(rdev);
 
 	r = radeon_asic_reset(rdev);
@@ -947,6 +950,7 @@ int radeon_gpu_reset(struct radeon_device *rdev)
 		radeon_resume(rdev);
 		radeon_restore_bios_scratch_regs(rdev);
 		drm_helper_resume_force_mode(rdev->ddev);
+		ttm_bo_unlock_delayed_workqueue(&rdev->mman.bdev, resched);
 		return 0;
 	}
 	/* bad news, how to tell it to userspace ? */
diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index 4409975a363..4be58793dc1 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/drivers/gpu/drm/radeon/radeon_display.c
@@ -793,6 +793,11 @@ static void avivo_get_fb_div(struct radeon_pll *pll,
 	tmp *= target_clock;
 	*fb_div = tmp / pll->reference_freq;
 	*frac_fb_div = tmp % pll->reference_freq;
+
+        if (*fb_div > pll->max_feedback_div)
+		*fb_div = pll->max_feedback_div;
+        else if (*fb_div < pll->min_feedback_div)
+                *fb_div = pll->min_feedback_div;
 }
 
 static u32 avivo_get_post_div(struct radeon_pll *pll,
@@ -826,6 +831,11 @@ static u32 avivo_get_post_div(struct radeon_pll *pll,
 			post_div--;
 	}
 
+	if (post_div > pll->max_post_div)
+		post_div = pll->max_post_div;
+	else if (post_div < pll->min_post_div)
+		post_div = pll->min_post_div;
+
 	return post_div;
 }
 
@@ -961,7 +971,7 @@ void radeon_compute_pll_legacy(struct radeon_pll *pll,
 		max_fractional_feed_div = pll->max_frac_feedback_div;
 	}
 
-	for (post_div = min_post_div; post_div <= max_post_div; ++post_div) {
+	for (post_div = max_post_div; post_div >= min_post_div; --post_div) {
 		uint32_t ref_div;
 
 		if ((pll->flags & RADEON_PLL_NO_ODD_POST_DIV) && (post_div & 1))
diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h
index 448eba89d1e..5cba46b9779 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.h
+++ b/drivers/gpu/drm/radeon/radeon_drv.h
@@ -1524,6 +1524,7 @@ extern u32 radeon_get_scratch(drm_radeon_private_t *dev_priv, int index);
 #define R600_CP_RB_CNTL                                        0xc104
 #       define R600_RB_BUFSZ(x)                                ((x) << 0)
 #       define R600_RB_BLKSZ(x)                                ((x) << 8)
+#	define R600_BUF_SWAP_32BIT		               (2 << 16)
 #       define R600_RB_NO_UPDATE                               (1 << 27)
 #       define R600_RB_RPTR_WR_ENA                             (1 << 31)
 #define R600_CP_RB_RPTR_WR                                     0xc108
diff --git a/drivers/gpu/drm/radeon/radeon_encoders.c b/drivers/gpu/drm/radeon/radeon_encoders.c
index d4a54224761..b4274883227 100644
--- a/drivers/gpu/drm/radeon/radeon_encoders.c
+++ b/drivers/gpu/drm/radeon/radeon_encoders.c
@@ -910,7 +910,7 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action, uint8_t
 
 	args.v1.ucAction = action;
 	if (action == ATOM_TRANSMITTER_ACTION_INIT) {
-		args.v1.usInitInfo = connector_object_id;
+		args.v1.usInitInfo = cpu_to_le16(connector_object_id);
 	} else if (action == ATOM_TRANSMITTER_ACTION_SETUP_VSEMPH) {
 		args.v1.asMode.ucLaneSel = lane_num;
 		args.v1.asMode.ucLaneSet = lane_set;
@@ -1140,7 +1140,7 @@ atombios_external_encoder_setup(struct drm_encoder *encoder,
 		case 3:
 			args.v3.sExtEncoder.ucAction = action;
 			if (action == EXTERNAL_ENCODER_ACTION_V3_ENCODER_INIT)
-				args.v3.sExtEncoder.usConnectorId = connector_object_id;
+				args.v3.sExtEncoder.usConnectorId = cpu_to_le16(connector_object_id);
 			else
 				args.v3.sExtEncoder.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10);
 			args.v3.sExtEncoder.ucEncoderMode = atombios_get_encoder_mode(encoder);
@@ -1570,11 +1570,21 @@ atombios_apply_encoder_quirks(struct drm_encoder *encoder,
 	}
 
 	/* set scaler clears this on some chips */
-	/* XXX check DCE4 */
-	if (!(radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT))) {
-		if (ASIC_IS_AVIVO(rdev) && (mode->flags & DRM_MODE_FLAG_INTERLACE))
-			WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset,
-			       AVIVO_D1MODE_INTERLEAVE_EN);
+	if (ASIC_IS_AVIVO(rdev) &&
+	    (!(radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT)))) {
+		if (ASIC_IS_DCE4(rdev)) {
+			if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+				WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset,
+				       EVERGREEN_INTERLEAVE_EN);
+			else
+				WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset, 0);
+		} else {
+			if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+				WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset,
+				       AVIVO_D1MODE_INTERLEAVE_EN);
+			else
+				WREG32(AVIVO_D1MODE_DATA_FORMAT + radeon_crtc->crtc_offset, 0);
+		}
 	}
 }
 
diff --git a/drivers/gpu/drm/radeon/radeon_fb.c b/drivers/gpu/drm/radeon/radeon_fb.c
index 28431e78ab5..0b7b486c97e 100644
--- a/drivers/gpu/drm/radeon/radeon_fb.c
+++ b/drivers/gpu/drm/radeon/radeon_fb.c
@@ -113,11 +113,14 @@ static int radeonfb_create_pinned_object(struct radeon_fbdev *rfbdev,
 	u32 tiling_flags = 0;
 	int ret;
 	int aligned_size, size;
+	int height = mode_cmd->height;
 
 	/* need to align pitch with crtc limits */
 	mode_cmd->pitch = radeon_align_pitch(rdev, mode_cmd->width, mode_cmd->bpp, fb_tiled) * ((mode_cmd->bpp + 1) / 8);
 
-	size = mode_cmd->pitch * mode_cmd->height;
+	if (rdev->family >= CHIP_R600)
+		height = ALIGN(mode_cmd->height, 8);
+	size = mode_cmd->pitch * height;
 	aligned_size = ALIGN(size, PAGE_SIZE);
 	ret = radeon_gem_object_create(rdev, aligned_size, 0,
 				       RADEON_GEM_DOMAIN_VRAM,
diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h
index c3f23f6ff60..5067d18d000 100644
--- a/drivers/gpu/drm/radeon/radeon_mode.h
+++ b/drivers/gpu/drm/radeon/radeon_mode.h
@@ -209,6 +209,7 @@ enum radeon_connector_table {
 	CT_EMAC,
 	CT_RN50_POWER,
 	CT_MAC_X800,
+	CT_MAC_G5_9600,
 };
 
 enum radeon_dvo_chip {
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index df5734d0c4a..e446979e0e0 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -791,9 +791,9 @@ static int radeon_ttm_debugfs_init(struct radeon_device *rdev)
 		radeon_mem_types_list[i].show = &radeon_mm_dump_table;
 		radeon_mem_types_list[i].driver_features = 0;
 		if (i == 0)
-			radeon_mem_types_list[i].data = &rdev->mman.bdev.man[TTM_PL_VRAM].priv;
+			radeon_mem_types_list[i].data = rdev->mman.bdev.man[TTM_PL_VRAM].priv;
 		else
-			radeon_mem_types_list[i].data = &rdev->mman.bdev.man[TTM_PL_TT].priv;
+			radeon_mem_types_list[i].data = rdev->mman.bdev.man[TTM_PL_TT].priv;
 
 	}
 	/* Add ttm page pool to debugfs */
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r300 b/drivers/gpu/drm/radeon/reg_srcs/r300
index b506ec1cab4..e8a1786b642 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/r300
+++ b/drivers/gpu/drm/radeon/reg_srcs/r300
@@ -683,9 +683,7 @@ r300 0x4f60
 0x4DF4 US_ALU_CONST_G_31
 0x4DF8 US_ALU_CONST_B_31
 0x4DFC US_ALU_CONST_A_31
-0x4E04 RB3D_BLENDCNTL_R3
 0x4E08 RB3D_ABLENDCNTL_R3
-0x4E0C RB3D_COLOR_CHANNEL_MASK
 0x4E10 RB3D_CONSTANT_COLOR
 0x4E14 RB3D_COLOR_CLEAR_VALUE
 0x4E18 RB3D_ROPCNTL_R3
@@ -706,13 +704,11 @@ r300 0x4f60
 0x4E74 RB3D_CMASK_WRINDEX
 0x4E78 RB3D_CMASK_DWORD
 0x4E7C RB3D_CMASK_RDINDEX
-0x4E80 RB3D_AARESOLVE_OFFSET
-0x4E84 RB3D_AARESOLVE_PITCH
-0x4E88 RB3D_AARESOLVE_CTL
 0x4EA0 RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD
 0x4EA4 RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD
 0x4F04 ZB_ZSTENCILCNTL
 0x4F08 ZB_STENCILREFMASK
 0x4F14 ZB_ZTOP
 0x4F18 ZB_ZCACHE_CTLSTAT
+0x4F28 ZB_DEPTHCLEARVALUE
 0x4F58 ZB_ZPASS_DATA
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r420 b/drivers/gpu/drm/radeon/reg_srcs/r420
index 8c1214c2390..722074e21e2 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/r420
+++ b/drivers/gpu/drm/radeon/reg_srcs/r420
@@ -130,7 +130,6 @@ r420 0x4f60
 0x401C GB_SELECT
 0x4020 GB_AA_CONFIG
 0x4024 GB_FIFO_SIZE
-0x4028 GB_Z_PEQ_CONFIG
 0x4100 TX_INVALTAGS
 0x4200 GA_POINT_S0
 0x4204 GA_POINT_T0
@@ -750,9 +749,7 @@ r420 0x4f60
 0x4DF4 US_ALU_CONST_G_31
 0x4DF8 US_ALU_CONST_B_31
 0x4DFC US_ALU_CONST_A_31
-0x4E04 RB3D_BLENDCNTL_R3
 0x4E08 RB3D_ABLENDCNTL_R3
-0x4E0C RB3D_COLOR_CHANNEL_MASK
 0x4E10 RB3D_CONSTANT_COLOR
 0x4E14 RB3D_COLOR_CLEAR_VALUE
 0x4E18 RB3D_ROPCNTL_R3
@@ -773,13 +770,11 @@ r420 0x4f60
 0x4E74 RB3D_CMASK_WRINDEX
 0x4E78 RB3D_CMASK_DWORD
 0x4E7C RB3D_CMASK_RDINDEX
-0x4E80 RB3D_AARESOLVE_OFFSET
-0x4E84 RB3D_AARESOLVE_PITCH
-0x4E88 RB3D_AARESOLVE_CTL
 0x4EA0 RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD
 0x4EA4 RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD
 0x4F04 ZB_ZSTENCILCNTL
 0x4F08 ZB_STENCILREFMASK
 0x4F14 ZB_ZTOP
 0x4F18 ZB_ZCACHE_CTLSTAT
+0x4F28 ZB_DEPTHCLEARVALUE
 0x4F58 ZB_ZPASS_DATA
diff --git a/drivers/gpu/drm/radeon/reg_srcs/rs600 b/drivers/gpu/drm/radeon/reg_srcs/rs600
index 0828d80396f..d9f62866bbc 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/rs600
+++ b/drivers/gpu/drm/radeon/reg_srcs/rs600
@@ -749,9 +749,7 @@ rs600 0x6d40
 0x4DF4 US_ALU_CONST_G_31
 0x4DF8 US_ALU_CONST_B_31
 0x4DFC US_ALU_CONST_A_31
-0x4E04 RB3D_BLENDCNTL_R3
 0x4E08 RB3D_ABLENDCNTL_R3
-0x4E0C RB3D_COLOR_CHANNEL_MASK
 0x4E10 RB3D_CONSTANT_COLOR
 0x4E14 RB3D_COLOR_CLEAR_VALUE
 0x4E18 RB3D_ROPCNTL_R3
@@ -772,13 +770,11 @@ rs600 0x6d40
 0x4E74 RB3D_CMASK_WRINDEX
 0x4E78 RB3D_CMASK_DWORD
 0x4E7C RB3D_CMASK_RDINDEX
-0x4E80 RB3D_AARESOLVE_OFFSET
-0x4E84 RB3D_AARESOLVE_PITCH
-0x4E88 RB3D_AARESOLVE_CTL
 0x4EA0 RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD
 0x4EA4 RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD
 0x4F04 ZB_ZSTENCILCNTL
 0x4F08 ZB_STENCILREFMASK
 0x4F14 ZB_ZTOP
 0x4F18 ZB_ZCACHE_CTLSTAT
+0x4F28 ZB_DEPTHCLEARVALUE
 0x4F58 ZB_ZPASS_DATA
diff --git a/drivers/gpu/drm/radeon/reg_srcs/rv515 b/drivers/gpu/drm/radeon/reg_srcs/rv515
index ef422bbacfc..911a8fbd32b 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/rv515
+++ b/drivers/gpu/drm/radeon/reg_srcs/rv515
@@ -164,7 +164,6 @@ rv515 0x6d40
 0x401C GB_SELECT
 0x4020 GB_AA_CONFIG
 0x4024 GB_FIFO_SIZE
-0x4028 GB_Z_PEQ_CONFIG
 0x4100 TX_INVALTAGS
 0x4114 SU_TEX_WRAP_PS3
 0x4118 PS3_ENABLE
@@ -461,9 +460,7 @@ rv515 0x6d40
 0x4DF4 US_ALU_CONST_G_31
 0x4DF8 US_ALU_CONST_B_31
 0x4DFC US_ALU_CONST_A_31
-0x4E04 RB3D_BLENDCNTL_R3
 0x4E08 RB3D_ABLENDCNTL_R3
-0x4E0C RB3D_COLOR_CHANNEL_MASK
 0x4E10 RB3D_CONSTANT_COLOR
 0x4E14 RB3D_COLOR_CLEAR_VALUE
 0x4E18 RB3D_ROPCNTL_R3
@@ -484,9 +481,6 @@ rv515 0x6d40
 0x4E74 RB3D_CMASK_WRINDEX
 0x4E78 RB3D_CMASK_DWORD
 0x4E7C RB3D_CMASK_RDINDEX
-0x4E80 RB3D_AARESOLVE_OFFSET
-0x4E84 RB3D_AARESOLVE_PITCH
-0x4E88 RB3D_AARESOLVE_CTL
 0x4EA0 RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD
 0x4EA4 RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD
 0x4EF8 RB3D_CONSTANT_COLOR_AR
@@ -496,4 +490,5 @@ rv515 0x6d40
 0x4F14 ZB_ZTOP
 0x4F18 ZB_ZCACHE_CTLSTAT
 0x4F58 ZB_ZPASS_DATA
+0x4F28 ZB_DEPTHCLEARVALUE
 0x4FD4 ZB_STENCILREFMASK_BF
diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c
index 0137d3e3728..6638c8e4c81 100644
--- a/drivers/gpu/drm/radeon/rs690.c
+++ b/drivers/gpu/drm/radeon/rs690.c
@@ -77,9 +77,9 @@ void rs690_pm_info(struct radeon_device *rdev)
 		switch (crev) {
 		case 1:
 			tmp.full = dfixed_const(100);
-			rdev->pm.igp_sideport_mclk.full = dfixed_const(info->info.ulBootUpMemoryClock);
+			rdev->pm.igp_sideport_mclk.full = dfixed_const(le32_to_cpu(info->info.ulBootUpMemoryClock));
 			rdev->pm.igp_sideport_mclk.full = dfixed_div(rdev->pm.igp_sideport_mclk, tmp);
-			if (info->info.usK8MemoryClock)
+			if (le16_to_cpu(info->info.usK8MemoryClock))
 				rdev->pm.igp_system_mclk.full = dfixed_const(le16_to_cpu(info->info.usK8MemoryClock));
 			else if (rdev->clock.default_mclk) {
 				rdev->pm.igp_system_mclk.full = dfixed_const(rdev->clock.default_mclk);
@@ -91,16 +91,16 @@ void rs690_pm_info(struct radeon_device *rdev)
 			break;
 		case 2:
 			tmp.full = dfixed_const(100);
-			rdev->pm.igp_sideport_mclk.full = dfixed_const(info->info_v2.ulBootUpSidePortClock);
+			rdev->pm.igp_sideport_mclk.full = dfixed_const(le32_to_cpu(info->info_v2.ulBootUpSidePortClock));
 			rdev->pm.igp_sideport_mclk.full = dfixed_div(rdev->pm.igp_sideport_mclk, tmp);
-			if (info->info_v2.ulBootUpUMAClock)
-				rdev->pm.igp_system_mclk.full = dfixed_const(info->info_v2.ulBootUpUMAClock);
+			if (le32_to_cpu(info->info_v2.ulBootUpUMAClock))
+				rdev->pm.igp_system_mclk.full = dfixed_const(le32_to_cpu(info->info_v2.ulBootUpUMAClock));
 			else if (rdev->clock.default_mclk)
 				rdev->pm.igp_system_mclk.full = dfixed_const(rdev->clock.default_mclk);
 			else
 				rdev->pm.igp_system_mclk.full = dfixed_const(66700);
 			rdev->pm.igp_system_mclk.full = dfixed_div(rdev->pm.igp_system_mclk, tmp);
-			rdev->pm.igp_ht_link_clk.full = dfixed_const(info->info_v2.ulHTLinkFreq);
+			rdev->pm.igp_ht_link_clk.full = dfixed_const(le32_to_cpu(info->info_v2.ulHTLinkFreq));
 			rdev->pm.igp_ht_link_clk.full = dfixed_div(rdev->pm.igp_ht_link_clk, tmp);
 			rdev->pm.igp_ht_link_width.full = dfixed_const(le16_to_cpu(info->info_v2.usMinHTLinkWidth));
 			break;
diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c
index 3a95999d2fe..ee5541c6a62 100644
--- a/drivers/gpu/drm/radeon/rv770.c
+++ b/drivers/gpu/drm/radeon/rv770.c
@@ -321,7 +321,11 @@ static int rv770_cp_load_microcode(struct radeon_device *rdev)
 		return -EINVAL;
 
 	r700_cp_stop(rdev);
-	WREG32(CP_RB_CNTL, RB_NO_UPDATE | (15 << 8) | (3 << 0));
+	WREG32(CP_RB_CNTL,
+#ifdef __BIG_ENDIAN
+	       BUF_SWAP_32BIT |
+#endif
+	       RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3));
 
 	/* Reset cp */
 	WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP);
diff --git a/drivers/gpu/drm/radeon/rv770d.h b/drivers/gpu/drm/radeon/rv770d.h
index abc8cf5a367..79fa588e9ed 100644
--- a/drivers/gpu/drm/radeon/rv770d.h
+++ b/drivers/gpu/drm/radeon/rv770d.h
@@ -76,10 +76,10 @@
 #define		ROQ_IB1_START(x)				((x) << 0)
 #define		ROQ_IB2_START(x)				((x) << 8)
 #define	CP_RB_CNTL					0xC104
-#define		RB_BUFSZ(x)					((x)<<0)
-#define		RB_BLKSZ(x)					((x)<<8)
-#define		RB_NO_UPDATE					(1<<27)
-#define		RB_RPTR_WR_ENA					(1<<31)
+#define		RB_BUFSZ(x)					((x) << 0)
+#define		RB_BLKSZ(x)					((x) << 8)
+#define		RB_NO_UPDATE					(1 << 27)
+#define		RB_RPTR_WR_ENA					(1 << 31)
 #define		BUF_SWAP_32BIT					(2 << 16)
 #define	CP_RB_RPTR					0x8700
 #define	CP_RB_RPTR_ADDR					0xC10C