Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / gpu / drm / gma500 / gma_display.c
diff --git a/kernel/drivers/gpu/drm/gma500/gma_display.c b/kernel/drivers/gpu/drm/gma500/gma_display.c
new file mode 100644 (file)
index 0000000..001b450
--- /dev/null
@@ -0,0 +1,791 @@
+/*
+ * Copyright © 2006-2011 Intel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Authors:
+ *     Eric Anholt <eric@anholt.net>
+ *     Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
+ */
+
+#include <drm/drmP.h>
+#include "gma_display.h"
+#include "psb_intel_drv.h"
+#include "psb_intel_reg.h"
+#include "psb_drv.h"
+#include "framebuffer.h"
+
+/**
+ * Returns whether any output on the specified pipe is of the specified type
+ */
+bool gma_pipe_has_type(struct drm_crtc *crtc, int type)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_mode_config *mode_config = &dev->mode_config;
+       struct drm_connector *l_entry;
+
+       list_for_each_entry(l_entry, &mode_config->connector_list, head) {
+               if (l_entry->encoder && l_entry->encoder->crtc == crtc) {
+                       struct gma_encoder *gma_encoder =
+                                               gma_attached_encoder(l_entry);
+                       if (gma_encoder->type == type)
+                               return true;
+               }
+       }
+
+       return false;
+}
+
+void gma_wait_for_vblank(struct drm_device *dev)
+{
+       /* Wait for 20ms, i.e. one cycle at 50hz. */
+       mdelay(20);
+}
+
+int gma_pipe_set_base(struct drm_crtc *crtc, int x, int y,
+                     struct drm_framebuffer *old_fb)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       struct psb_framebuffer *psbfb = to_psb_fb(crtc->primary->fb);
+       int pipe = gma_crtc->pipe;
+       const struct psb_offset *map = &dev_priv->regmap[pipe];
+       unsigned long start, offset;
+       u32 dspcntr;
+       int ret = 0;
+
+       if (!gma_power_begin(dev, true))
+               return 0;
+
+       /* no fb bound */
+       if (!crtc->primary->fb) {
+               dev_err(dev->dev, "No FB bound\n");
+               goto gma_pipe_cleaner;
+       }
+
+       /* We are displaying this buffer, make sure it is actually loaded
+          into the GTT */
+       ret = psb_gtt_pin(psbfb->gtt);
+       if (ret < 0)
+               goto gma_pipe_set_base_exit;
+       start = psbfb->gtt->offset;
+       offset = y * crtc->primary->fb->pitches[0] + x * (crtc->primary->fb->bits_per_pixel / 8);
+
+       REG_WRITE(map->stride, crtc->primary->fb->pitches[0]);
+
+       dspcntr = REG_READ(map->cntr);
+       dspcntr &= ~DISPPLANE_PIXFORMAT_MASK;
+
+       switch (crtc->primary->fb->bits_per_pixel) {
+       case 8:
+               dspcntr |= DISPPLANE_8BPP;
+               break;
+       case 16:
+               if (crtc->primary->fb->depth == 15)
+                       dspcntr |= DISPPLANE_15_16BPP;
+               else
+                       dspcntr |= DISPPLANE_16BPP;
+               break;
+       case 24:
+       case 32:
+               dspcntr |= DISPPLANE_32BPP_NO_ALPHA;
+               break;
+       default:
+               dev_err(dev->dev, "Unknown color depth\n");
+               ret = -EINVAL;
+               goto gma_pipe_set_base_exit;
+       }
+       REG_WRITE(map->cntr, dspcntr);
+
+       dev_dbg(dev->dev,
+               "Writing base %08lX %08lX %d %d\n", start, offset, x, y);
+
+       /* FIXME: Investigate whether this really is the base for psb and why
+                 the linear offset is named base for the other chips. map->surf
+                 should be the base and map->linoff the offset for all chips */
+       if (IS_PSB(dev)) {
+               REG_WRITE(map->base, offset + start);
+               REG_READ(map->base);
+       } else {
+               REG_WRITE(map->base, offset);
+               REG_READ(map->base);
+               REG_WRITE(map->surf, start);
+               REG_READ(map->surf);
+       }
+
+gma_pipe_cleaner:
+       /* If there was a previous display we can now unpin it */
+       if (old_fb)
+               psb_gtt_unpin(to_psb_fb(old_fb)->gtt);
+
+gma_pipe_set_base_exit:
+       gma_power_end(dev);
+       return ret;
+}
+
+/* Loads the palette/gamma unit for the CRTC with the prepared values */
+void gma_crtc_load_lut(struct drm_crtc *crtc)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       const struct psb_offset *map = &dev_priv->regmap[gma_crtc->pipe];
+       int palreg = map->palette;
+       int i;
+
+       /* The clocks have to be on to load the palette. */
+       if (!crtc->enabled)
+               return;
+
+       if (gma_power_begin(dev, false)) {
+               for (i = 0; i < 256; i++) {
+                       REG_WRITE(palreg + 4 * i,
+                                 ((gma_crtc->lut_r[i] +
+                                 gma_crtc->lut_adj[i]) << 16) |
+                                 ((gma_crtc->lut_g[i] +
+                                 gma_crtc->lut_adj[i]) << 8) |
+                                 (gma_crtc->lut_b[i] +
+                                 gma_crtc->lut_adj[i]));
+               }
+               gma_power_end(dev);
+       } else {
+               for (i = 0; i < 256; i++) {
+                       /* FIXME: Why pipe[0] and not pipe[..._crtc->pipe]? */
+                       dev_priv->regs.pipe[0].palette[i] =
+                                 ((gma_crtc->lut_r[i] +
+                                 gma_crtc->lut_adj[i]) << 16) |
+                                 ((gma_crtc->lut_g[i] +
+                                 gma_crtc->lut_adj[i]) << 8) |
+                                 (gma_crtc->lut_b[i] +
+                                 gma_crtc->lut_adj[i]);
+               }
+
+       }
+}
+
+void gma_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, u16 *blue,
+                       u32 start, u32 size)
+{
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       int i;
+       int end = (start + size > 256) ? 256 : start + size;
+
+       for (i = start; i < end; i++) {
+               gma_crtc->lut_r[i] = red[i] >> 8;
+               gma_crtc->lut_g[i] = green[i] >> 8;
+               gma_crtc->lut_b[i] = blue[i] >> 8;
+       }
+
+       gma_crtc_load_lut(crtc);
+}
+
+/**
+ * Sets the power management mode of the pipe and plane.
+ *
+ * This code should probably grow support for turning the cursor off and back
+ * on appropriately at the same time as we're turning the pipe off/on.
+ */
+void gma_crtc_dpms(struct drm_crtc *crtc, int mode)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       int pipe = gma_crtc->pipe;
+       const struct psb_offset *map = &dev_priv->regmap[pipe];
+       u32 temp;
+
+       /* XXX: When our outputs are all unaware of DPMS modes other than off
+        * and on, we should map those modes to DRM_MODE_DPMS_OFF in the CRTC.
+        */
+
+       if (IS_CDV(dev))
+               dev_priv->ops->disable_sr(dev);
+
+       switch (mode) {
+       case DRM_MODE_DPMS_ON:
+       case DRM_MODE_DPMS_STANDBY:
+       case DRM_MODE_DPMS_SUSPEND:
+               if (gma_crtc->active)
+                       break;
+
+               gma_crtc->active = true;
+
+               /* Enable the DPLL */
+               temp = REG_READ(map->dpll);
+               if ((temp & DPLL_VCO_ENABLE) == 0) {
+                       REG_WRITE(map->dpll, temp);
+                       REG_READ(map->dpll);
+                       /* Wait for the clocks to stabilize. */
+                       udelay(150);
+                       REG_WRITE(map->dpll, temp | DPLL_VCO_ENABLE);
+                       REG_READ(map->dpll);
+                       /* Wait for the clocks to stabilize. */
+                       udelay(150);
+                       REG_WRITE(map->dpll, temp | DPLL_VCO_ENABLE);
+                       REG_READ(map->dpll);
+                       /* Wait for the clocks to stabilize. */
+                       udelay(150);
+               }
+
+               /* Enable the plane */
+               temp = REG_READ(map->cntr);
+               if ((temp & DISPLAY_PLANE_ENABLE) == 0) {
+                       REG_WRITE(map->cntr,
+                                 temp | DISPLAY_PLANE_ENABLE);
+                       /* Flush the plane changes */
+                       REG_WRITE(map->base, REG_READ(map->base));
+               }
+
+               udelay(150);
+
+               /* Enable the pipe */
+               temp = REG_READ(map->conf);
+               if ((temp & PIPEACONF_ENABLE) == 0)
+                       REG_WRITE(map->conf, temp | PIPEACONF_ENABLE);
+
+               temp = REG_READ(map->status);
+               temp &= ~(0xFFFF);
+               temp |= PIPE_FIFO_UNDERRUN;
+               REG_WRITE(map->status, temp);
+               REG_READ(map->status);
+
+               gma_crtc_load_lut(crtc);
+
+               /* Give the overlay scaler a chance to enable
+                * if it's on this pipe */
+               /* psb_intel_crtc_dpms_video(crtc, true); TODO */
+               break;
+       case DRM_MODE_DPMS_OFF:
+               if (!gma_crtc->active)
+                       break;
+
+               gma_crtc->active = false;
+
+               /* Give the overlay scaler a chance to disable
+                * if it's on this pipe */
+               /* psb_intel_crtc_dpms_video(crtc, FALSE); TODO */
+
+               /* Disable the VGA plane that we never use */
+               REG_WRITE(VGACNTRL, VGA_DISP_DISABLE);
+
+               /* Turn off vblank interrupts */
+               drm_vblank_off(dev, pipe);
+
+               /* Wait for vblank for the disable to take effect */
+               gma_wait_for_vblank(dev);
+
+               /* Disable plane */
+               temp = REG_READ(map->cntr);
+               if ((temp & DISPLAY_PLANE_ENABLE) != 0) {
+                       REG_WRITE(map->cntr,
+                                 temp & ~DISPLAY_PLANE_ENABLE);
+                       /* Flush the plane changes */
+                       REG_WRITE(map->base, REG_READ(map->base));
+                       REG_READ(map->base);
+               }
+
+               /* Disable pipe */
+               temp = REG_READ(map->conf);
+               if ((temp & PIPEACONF_ENABLE) != 0) {
+                       REG_WRITE(map->conf, temp & ~PIPEACONF_ENABLE);
+                       REG_READ(map->conf);
+               }
+
+               /* Wait for vblank for the disable to take effect. */
+               gma_wait_for_vblank(dev);
+
+               udelay(150);
+
+               /* Disable DPLL */
+               temp = REG_READ(map->dpll);
+               if ((temp & DPLL_VCO_ENABLE) != 0) {
+                       REG_WRITE(map->dpll, temp & ~DPLL_VCO_ENABLE);
+                       REG_READ(map->dpll);
+               }
+
+               /* Wait for the clocks to turn off. */
+               udelay(150);
+               break;
+       }
+
+       if (IS_CDV(dev))
+               dev_priv->ops->update_wm(dev, crtc);
+
+       /* Set FIFO watermarks */
+       REG_WRITE(DSPARB, 0x3F3E);
+}
+
+int gma_crtc_cursor_set(struct drm_crtc *crtc,
+                       struct drm_file *file_priv,
+                       uint32_t handle,
+                       uint32_t width, uint32_t height)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       int pipe = gma_crtc->pipe;
+       uint32_t control = (pipe == 0) ? CURACNTR : CURBCNTR;
+       uint32_t base = (pipe == 0) ? CURABASE : CURBBASE;
+       uint32_t temp;
+       size_t addr = 0;
+       struct gtt_range *gt;
+       struct gtt_range *cursor_gt = gma_crtc->cursor_gt;
+       struct drm_gem_object *obj;
+       void *tmp_dst, *tmp_src;
+       int ret = 0, i, cursor_pages;
+
+       /* If we didn't get a handle then turn the cursor off */
+       if (!handle) {
+               temp = CURSOR_MODE_DISABLE;
+               mutex_lock(&dev->struct_mutex);
+
+               if (gma_power_begin(dev, false)) {
+                       REG_WRITE(control, temp);
+                       REG_WRITE(base, 0);
+                       gma_power_end(dev);
+               }
+
+               /* Unpin the old GEM object */
+               if (gma_crtc->cursor_obj) {
+                       gt = container_of(gma_crtc->cursor_obj,
+                                         struct gtt_range, gem);
+                       psb_gtt_unpin(gt);
+                       drm_gem_object_unreference(gma_crtc->cursor_obj);
+                       gma_crtc->cursor_obj = NULL;
+               }
+
+               mutex_unlock(&dev->struct_mutex);
+               return 0;
+       }
+
+       /* Currently we only support 64x64 cursors */
+       if (width != 64 || height != 64) {
+               dev_dbg(dev->dev, "We currently only support 64x64 cursors\n");
+               return -EINVAL;
+       }
+
+       mutex_lock(&dev->struct_mutex);
+       obj = drm_gem_object_lookup(dev, file_priv, handle);
+       if (!obj) {
+               ret = -ENOENT;
+               goto unlock;
+       }
+
+       if (obj->size < width * height * 4) {
+               dev_dbg(dev->dev, "Buffer is too small\n");
+               ret = -ENOMEM;
+               goto unref_cursor;
+       }
+
+       gt = container_of(obj, struct gtt_range, gem);
+
+       /* Pin the memory into the GTT */
+       ret = psb_gtt_pin(gt);
+       if (ret) {
+               dev_err(dev->dev, "Can not pin down handle 0x%x\n", handle);
+               goto unref_cursor;
+       }
+
+       if (dev_priv->ops->cursor_needs_phys) {
+               if (cursor_gt == NULL) {
+                       dev_err(dev->dev, "No hardware cursor mem available");
+                       ret = -ENOMEM;
+                       goto unref_cursor;
+               }
+
+               /* Prevent overflow */
+               if (gt->npage > 4)
+                       cursor_pages = 4;
+               else
+                       cursor_pages = gt->npage;
+
+               /* Copy the cursor to cursor mem */
+               tmp_dst = dev_priv->vram_addr + cursor_gt->offset;
+               for (i = 0; i < cursor_pages; i++) {
+                       tmp_src = kmap(gt->pages[i]);
+                       memcpy(tmp_dst, tmp_src, PAGE_SIZE);
+                       kunmap(gt->pages[i]);
+                       tmp_dst += PAGE_SIZE;
+               }
+
+               addr = gma_crtc->cursor_addr;
+       } else {
+               addr = gt->offset;
+               gma_crtc->cursor_addr = addr;
+       }
+
+       temp = 0;
+       /* set the pipe for the cursor */
+       temp |= (pipe << 28);
+       temp |= CURSOR_MODE_64_ARGB_AX | MCURSOR_GAMMA_ENABLE;
+
+       if (gma_power_begin(dev, false)) {
+               REG_WRITE(control, temp);
+               REG_WRITE(base, addr);
+               gma_power_end(dev);
+       }
+
+       /* unpin the old bo */
+       if (gma_crtc->cursor_obj) {
+               gt = container_of(gma_crtc->cursor_obj, struct gtt_range, gem);
+               psb_gtt_unpin(gt);
+               drm_gem_object_unreference(gma_crtc->cursor_obj);
+       }
+
+       gma_crtc->cursor_obj = obj;
+unlock:
+       mutex_unlock(&dev->struct_mutex);
+       return ret;
+
+unref_cursor:
+       drm_gem_object_unreference(obj);
+       mutex_unlock(&dev->struct_mutex);
+       return ret;
+}
+
+int gma_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
+{
+       struct drm_device *dev = crtc->dev;
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       int pipe = gma_crtc->pipe;
+       uint32_t temp = 0;
+       uint32_t addr;
+
+       if (x < 0) {
+               temp |= (CURSOR_POS_SIGN << CURSOR_X_SHIFT);
+               x = -x;
+       }
+       if (y < 0) {
+               temp |= (CURSOR_POS_SIGN << CURSOR_Y_SHIFT);
+               y = -y;
+       }
+
+       temp |= ((x & CURSOR_POS_MASK) << CURSOR_X_SHIFT);
+       temp |= ((y & CURSOR_POS_MASK) << CURSOR_Y_SHIFT);
+
+       addr = gma_crtc->cursor_addr;
+
+       if (gma_power_begin(dev, false)) {
+               REG_WRITE((pipe == 0) ? CURAPOS : CURBPOS, temp);
+               REG_WRITE((pipe == 0) ? CURABASE : CURBBASE, addr);
+               gma_power_end(dev);
+       }
+       return 0;
+}
+
+bool gma_encoder_mode_fixup(struct drm_encoder *encoder,
+                           const struct drm_display_mode *mode,
+                           struct drm_display_mode *adjusted_mode)
+{
+       return true;
+}
+
+bool gma_crtc_mode_fixup(struct drm_crtc *crtc,
+                        const struct drm_display_mode *mode,
+                        struct drm_display_mode *adjusted_mode)
+{
+       return true;
+}
+
+void gma_crtc_prepare(struct drm_crtc *crtc)
+{
+       const struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
+       crtc_funcs->dpms(crtc, DRM_MODE_DPMS_OFF);
+}
+
+void gma_crtc_commit(struct drm_crtc *crtc)
+{
+       const struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
+       crtc_funcs->dpms(crtc, DRM_MODE_DPMS_ON);
+}
+
+void gma_crtc_disable(struct drm_crtc *crtc)
+{
+       struct gtt_range *gt;
+       const struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
+
+       crtc_funcs->dpms(crtc, DRM_MODE_DPMS_OFF);
+
+       if (crtc->primary->fb) {
+               gt = to_psb_fb(crtc->primary->fb)->gtt;
+               psb_gtt_unpin(gt);
+       }
+}
+
+void gma_crtc_destroy(struct drm_crtc *crtc)
+{
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+
+       kfree(gma_crtc->crtc_state);
+       drm_crtc_cleanup(crtc);
+       kfree(gma_crtc);
+}
+
+int gma_crtc_set_config(struct drm_mode_set *set)
+{
+       struct drm_device *dev = set->crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       int ret;
+
+       if (!dev_priv->rpm_enabled)
+               return drm_crtc_helper_set_config(set);
+
+       pm_runtime_forbid(&dev->pdev->dev);
+       ret = drm_crtc_helper_set_config(set);
+       pm_runtime_allow(&dev->pdev->dev);
+
+       return ret;
+}
+
+/**
+ * Save HW states of given crtc
+ */
+void gma_crtc_save(struct drm_crtc *crtc)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       struct gma_crtc *gma_crtc = to_gma_crtc(crtc);
+       struct psb_intel_crtc_state *crtc_state = gma_crtc->crtc_state;
+       const struct psb_offset *map = &dev_priv->regmap[gma_crtc->pipe];
+       uint32_t palette_reg;
+       int i;
+
+       if (!crtc_state) {
+               dev_err(dev->dev, "No CRTC state found\n");
+               return;
+       }
+
+       crtc_state->saveDSPCNTR = REG_READ(map->cntr);
+       crtc_state->savePIPECONF = REG_READ(map->conf);
+       crtc_state->savePIPESRC = REG_READ(map->src);
+       crtc_state->saveFP0 = REG_READ(map->fp0);
+       crtc_state->saveFP1 = REG_READ(map->fp1);
+       crtc_state->saveDPLL = REG_READ(map->dpll);
+       crtc_state->saveHTOTAL = REG_READ(map->htotal);
+       crtc_state->saveHBLANK = REG_READ(map->hblank);
+       crtc_state->saveHSYNC = REG_READ(map->hsync);
+       crtc_state->saveVTOTAL = REG_READ(map->vtotal);
+       crtc_state->saveVBLANK = REG_READ(map->vblank);
+       crtc_state->saveVSYNC = REG_READ(map->vsync);
+       crtc_state->saveDSPSTRIDE = REG_READ(map->stride);
+
+       /* NOTE: DSPSIZE DSPPOS only for psb */
+       crtc_state->saveDSPSIZE = REG_READ(map->size);
+       crtc_state->saveDSPPOS = REG_READ(map->pos);
+
+       crtc_state->saveDSPBASE = REG_READ(map->base);
+
+       palette_reg = map->palette;
+       for (i = 0; i < 256; ++i)
+               crtc_state->savePalette[i] = REG_READ(palette_reg + (i << 2));
+}
+
+/**
+ * Restore HW states of given crtc
+ */
+void gma_crtc_restore(struct drm_crtc *crtc)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_psb_private *dev_priv = dev->dev_private;
+       struct gma_crtc *gma_crtc =  to_gma_crtc(crtc);
+       struct psb_intel_crtc_state *crtc_state = gma_crtc->crtc_state;
+       const struct psb_offset *map = &dev_priv->regmap[gma_crtc->pipe];
+       uint32_t palette_reg;
+       int i;
+
+       if (!crtc_state) {
+               dev_err(dev->dev, "No crtc state\n");
+               return;
+       }
+
+       if (crtc_state->saveDPLL & DPLL_VCO_ENABLE) {
+               REG_WRITE(map->dpll,
+                       crtc_state->saveDPLL & ~DPLL_VCO_ENABLE);
+               REG_READ(map->dpll);
+               udelay(150);
+       }
+
+       REG_WRITE(map->fp0, crtc_state->saveFP0);
+       REG_READ(map->fp0);
+
+       REG_WRITE(map->fp1, crtc_state->saveFP1);
+       REG_READ(map->fp1);
+
+       REG_WRITE(map->dpll, crtc_state->saveDPLL);
+       REG_READ(map->dpll);
+       udelay(150);
+
+       REG_WRITE(map->htotal, crtc_state->saveHTOTAL);
+       REG_WRITE(map->hblank, crtc_state->saveHBLANK);
+       REG_WRITE(map->hsync, crtc_state->saveHSYNC);
+       REG_WRITE(map->vtotal, crtc_state->saveVTOTAL);
+       REG_WRITE(map->vblank, crtc_state->saveVBLANK);
+       REG_WRITE(map->vsync, crtc_state->saveVSYNC);
+       REG_WRITE(map->stride, crtc_state->saveDSPSTRIDE);
+
+       REG_WRITE(map->size, crtc_state->saveDSPSIZE);
+       REG_WRITE(map->pos, crtc_state->saveDSPPOS);
+
+       REG_WRITE(map->src, crtc_state->savePIPESRC);
+       REG_WRITE(map->base, crtc_state->saveDSPBASE);
+       REG_WRITE(map->conf, crtc_state->savePIPECONF);
+
+       gma_wait_for_vblank(dev);
+
+       REG_WRITE(map->cntr, crtc_state->saveDSPCNTR);
+       REG_WRITE(map->base, crtc_state->saveDSPBASE);
+
+       gma_wait_for_vblank(dev);
+
+       palette_reg = map->palette;
+       for (i = 0; i < 256; ++i)
+               REG_WRITE(palette_reg + (i << 2), crtc_state->savePalette[i]);
+}
+
+void gma_encoder_prepare(struct drm_encoder *encoder)
+{
+       const struct drm_encoder_helper_funcs *encoder_funcs =
+           encoder->helper_private;
+       /* lvds has its own version of prepare see psb_intel_lvds_prepare */
+       encoder_funcs->dpms(encoder, DRM_MODE_DPMS_OFF);
+}
+
+void gma_encoder_commit(struct drm_encoder *encoder)
+{
+       const struct drm_encoder_helper_funcs *encoder_funcs =
+           encoder->helper_private;
+       /* lvds has its own version of commit see psb_intel_lvds_commit */
+       encoder_funcs->dpms(encoder, DRM_MODE_DPMS_ON);
+}
+
+void gma_encoder_destroy(struct drm_encoder *encoder)
+{
+       struct gma_encoder *intel_encoder = to_gma_encoder(encoder);
+
+       drm_encoder_cleanup(encoder);
+       kfree(intel_encoder);
+}
+
+/* Currently there is only a 1:1 mapping of encoders and connectors */
+struct drm_encoder *gma_best_encoder(struct drm_connector *connector)
+{
+       struct gma_encoder *gma_encoder = gma_attached_encoder(connector);
+
+       return &gma_encoder->base;
+}
+
+void gma_connector_attach_encoder(struct gma_connector *connector,
+                                 struct gma_encoder *encoder)
+{
+       connector->encoder = encoder;
+       drm_mode_connector_attach_encoder(&connector->base,
+                                         &encoder->base);
+}
+
+#define GMA_PLL_INVALID(s) { /* DRM_ERROR(s); */ return false; }
+
+bool gma_pll_is_valid(struct drm_crtc *crtc,
+                     const struct gma_limit_t *limit,
+                     struct gma_clock_t *clock)
+{
+       if (clock->p1 < limit->p1.min || limit->p1.max < clock->p1)
+               GMA_PLL_INVALID("p1 out of range");
+       if (clock->p < limit->p.min || limit->p.max < clock->p)
+               GMA_PLL_INVALID("p out of range");
+       if (clock->m2 < limit->m2.min || limit->m2.max < clock->m2)
+               GMA_PLL_INVALID("m2 out of range");
+       if (clock->m1 < limit->m1.min || limit->m1.max < clock->m1)
+               GMA_PLL_INVALID("m1 out of range");
+       /* On CDV m1 is always 0 */
+       if (clock->m1 <= clock->m2 && clock->m1 != 0)
+               GMA_PLL_INVALID("m1 <= m2 && m1 != 0");
+       if (clock->m < limit->m.min || limit->m.max < clock->m)
+               GMA_PLL_INVALID("m out of range");
+       if (clock->n < limit->n.min || limit->n.max < clock->n)
+               GMA_PLL_INVALID("n out of range");
+       if (clock->vco < limit->vco.min || limit->vco.max < clock->vco)
+               GMA_PLL_INVALID("vco out of range");
+       /* XXX: We may need to be checking "Dot clock"
+        * depending on the multiplier, connector, etc.,
+        * rather than just a single range.
+        */
+       if (clock->dot < limit->dot.min || limit->dot.max < clock->dot)
+               GMA_PLL_INVALID("dot out of range");
+
+       return true;
+}
+
+bool gma_find_best_pll(const struct gma_limit_t *limit,
+                      struct drm_crtc *crtc, int target, int refclk,
+                      struct gma_clock_t *best_clock)
+{
+       struct drm_device *dev = crtc->dev;
+       const struct gma_clock_funcs *clock_funcs =
+                                               to_gma_crtc(crtc)->clock_funcs;
+       struct gma_clock_t clock;
+       int err = target;
+
+       if (gma_pipe_has_type(crtc, INTEL_OUTPUT_LVDS) &&
+           (REG_READ(LVDS) & LVDS_PORT_EN) != 0) {
+               /*
+                * For LVDS, if the panel is on, just rely on its current
+                * settings for dual-channel.  We haven't figured out how to
+                * reliably set up different single/dual channel state, if we
+                * even can.
+                */
+               if ((REG_READ(LVDS) & LVDS_CLKB_POWER_MASK) ==
+                   LVDS_CLKB_POWER_UP)
+                       clock.p2 = limit->p2.p2_fast;
+               else
+                       clock.p2 = limit->p2.p2_slow;
+       } else {
+               if (target < limit->p2.dot_limit)
+                       clock.p2 = limit->p2.p2_slow;
+               else
+                       clock.p2 = limit->p2.p2_fast;
+       }
+
+       memset(best_clock, 0, sizeof(*best_clock));
+
+       /* m1 is always 0 on CDV so the outmost loop will run just once */
+       for (clock.m1 = limit->m1.min; clock.m1 <= limit->m1.max; clock.m1++) {
+               for (clock.m2 = limit->m2.min;
+                    (clock.m2 < clock.m1 || clock.m1 == 0) &&
+                     clock.m2 <= limit->m2.max; clock.m2++) {
+                       for (clock.n = limit->n.min;
+                            clock.n <= limit->n.max; clock.n++) {
+                               for (clock.p1 = limit->p1.min;
+                                    clock.p1 <= limit->p1.max;
+                                    clock.p1++) {
+                                       int this_err;
+
+                                       clock_funcs->clock(refclk, &clock);
+
+                                       if (!clock_funcs->pll_is_valid(crtc,
+                                                               limit, &clock))
+                                               continue;
+
+                                       this_err = abs(clock.dot - target);
+                                       if (this_err < err) {
+                                               *best_clock = clock;
+                                               err = this_err;
+                                       }
+                               }
+                       }
+               }
+       }
+
+       return err != target;
+}