/* called from IRQ to update cursor related registers (if needed). The
* cursor registers, other than x/y position, appear not to be double
* buffered, and changing them other than from vblank seems to trigger
* underflow.
*/
static void update_cursor(struct drm_crtc *crtc)
{
struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
struct mdp4_kms *mdp4_kms = get_kms(crtc);
enum mdp4_dma dma = mdp4_crtc->dma;
unsigned long flags;
spin_lock_irqsave(&mdp4_crtc->cursor.lock, flags);
if (mdp4_crtc->cursor.stale) {
struct drm_gem_object *next_bo = mdp4_crtc->cursor.next_bo;
struct drm_gem_object *prev_bo = mdp4_crtc->cursor.scanout_bo;
uint32_t iova = mdp4_crtc->cursor.next_iova;
if (next_bo) {
/* take a obj ref + iova ref when we start scanning out: */
drm_gem_object_reference(next_bo);
msm_gem_get_iova_locked(next_bo, mdp4_kms->id, &iova);
/* enable cursor: */
mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),
MDP4_DMA_CURSOR_SIZE_WIDTH(mdp4_crtc->cursor.width) |
MDP4_DMA_CURSOR_SIZE_HEIGHT(mdp4_crtc->cursor.height));
mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_BASE(dma), iova);
mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_BLEND_CONFIG(dma),
MDP4_DMA_CURSOR_BLEND_CONFIG_FORMAT(CURSOR_ARGB) |
MDP4_DMA_CURSOR_BLEND_CONFIG_CURSOR_EN);
} else {
/* disable cursor: */
mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_BASE(dma),
mdp4_kms->blank_cursor_iova);
}
/* and drop the iova ref + obj rev when done scanning out: */
if (prev_bo)
drm_flip_work_queue(&mdp4_crtc->unref_cursor_work, prev_bo);
mdp4_crtc->cursor.scanout_bo = next_bo;
mdp4_crtc->cursor.stale = false;
}
mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_POS(dma),
MDP4_DMA_CURSOR_POS_X(mdp4_crtc->cursor.x) |
MDP4_DMA_CURSOR_POS_Y(mdp4_crtc->cursor.y));
spin_unlock_irqrestore(&mdp4_crtc->cursor.lock, flags);
}
/* get iova, taking a reference. Should have a matching put */
int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
int ret;
/* this is safe right now because we don't unmap until the
* bo is deleted:
*/
if (msm_obj->domain[id].iova) {
*iova = msm_obj->domain[id].iova;
return 0;
}
mutex_lock(&obj->dev->struct_mutex);
ret = msm_gem_get_iova_locked(obj, id, iova);
mutex_unlock(&obj->dev->struct_mutex);
return ret;
}
/* add bo's to gpu's ring, and kick gpu: */
int msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
struct msm_file_private *ctx)
{
struct drm_device *dev = gpu->dev;
struct msm_drm_private *priv = dev->dev_private;
int i, ret;
mutex_lock(&dev->struct_mutex);
submit->fence = ++priv->next_fence;
gpu->submitted_fence = submit->fence;
ret = gpu->funcs->submit(gpu, submit, ctx);
priv->lastctx = ctx;
for (i = 0; i < submit->nr_bos; i++) {
struct msm_gem_object *msm_obj = submit->bos[i].obj;
/* can't happen yet.. but when we add 2d support we'll have
* to deal w/ cross-ring synchronization:
*/
WARN_ON(is_active(msm_obj) && (msm_obj->gpu != gpu));
if (!is_active(msm_obj)) {
uint32_t iova;
/* ring takes a reference to the bo and iova: */
drm_gem_object_reference(&msm_obj->base);
msm_gem_get_iova_locked(&msm_obj->base,
submit->gpu->id, &iova);
}
if (submit->bos[i].flags & MSM_SUBMIT_BO_READ)
msm_gem_move_to_active(&msm_obj->base, gpu, false, submit->fence);
if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
msm_gem_move_to_active(&msm_obj->base, gpu, true, submit->fence);
}
hangcheck_timer_reset(gpu);
mutex_unlock(&dev->struct_mutex);
return ret;
}
int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
struct msm_gpu *gpu, const struct msm_gpu_funcs *funcs,
const char *name, const char *ioname, const char *irqname, int ringsz)
{
int i, ret;
gpu->dev = drm;
gpu->funcs = funcs;
gpu->name = name;
INIT_LIST_HEAD(&gpu->active_list);
INIT_WORK(&gpu->retire_work, retire_worker);
INIT_WORK(&gpu->recover_work, recover_worker);
setup_timer(&gpu->hangcheck_timer, hangcheck_handler,
(unsigned long)gpu);
BUG_ON(ARRAY_SIZE(clk_names) != ARRAY_SIZE(gpu->grp_clks));
/* Map registers: */
gpu->mmio = msm_ioremap(pdev, ioname, name);
if (IS_ERR(gpu->mmio)) {
ret = PTR_ERR(gpu->mmio);
goto fail;
}
/* Get Interrupt: */
gpu->irq = platform_get_irq_byname(pdev, irqname);
if (gpu->irq < 0) {
ret = gpu->irq;
dev_err(drm->dev, "failed to get irq: %d\n", ret);
goto fail;
}
ret = devm_request_irq(&pdev->dev, gpu->irq, irq_handler,
IRQF_TRIGGER_HIGH, gpu->name, gpu);
if (ret) {
dev_err(drm->dev, "failed to request IRQ%u: %d\n", gpu->irq, ret);
goto fail;
}
/* Acquire clocks: */
for (i = 0; i < ARRAY_SIZE(clk_names); i++) {
gpu->grp_clks[i] = devm_clk_get(&pdev->dev, clk_names[i]);
DBG("grp_clks[%s]: %p", clk_names[i], gpu->grp_clks[i]);
if (IS_ERR(gpu->grp_clks[i]))
gpu->grp_clks[i] = NULL;
}
gpu->ebi1_clk = devm_clk_get(&pdev->dev, "bus_clk");
DBG("ebi1_clk: %p", gpu->ebi1_clk);
if (IS_ERR(gpu->ebi1_clk))
gpu->ebi1_clk = NULL;
/* Acquire regulators: */
gpu->gpu_reg = devm_regulator_get(&pdev->dev, "vdd");
DBG("gpu_reg: %p", gpu->gpu_reg);
if (IS_ERR(gpu->gpu_reg))
gpu->gpu_reg = NULL;
gpu->gpu_cx = devm_regulator_get(&pdev->dev, "vddcx");
DBG("gpu_cx: %p", gpu->gpu_cx);
if (IS_ERR(gpu->gpu_cx))
gpu->gpu_cx = NULL;
/* Setup IOMMU.. eventually we will (I think) do this once per context
* and have separate page tables per context. For now, to keep things
* simple and to get something working, just use a single address space:
*/
gpu->iommu = iommu_domain_alloc(&platform_bus_type);
if (!gpu->iommu) {
dev_err(drm->dev, "failed to allocate IOMMU\n");
ret = -ENOMEM;
goto fail;
}
gpu->id = msm_register_iommu(drm, gpu->iommu);
/* Create ringbuffer: */
gpu->rb = msm_ringbuffer_new(gpu, ringsz);
if (IS_ERR(gpu->rb)) {
ret = PTR_ERR(gpu->rb);
gpu->rb = NULL;
dev_err(drm->dev, "could not create ringbuffer: %d\n", ret);
goto fail;
}
ret = msm_gem_get_iova_locked(gpu->rb->bo, gpu->id, &gpu->rb_iova);
if (ret) {
gpu->rb_iova = 0;
dev_err(drm->dev, "could not map ringbuffer: %d\n", ret);
goto fail;
}
bs_init(gpu, pdev);
return 0;
fail:
return ret;
}