From d9fd0c7d259e6e041890764f21f3033d248e0ac8 Mon Sep 17 00:00:00 2001 From: Lucas Stach Date: Thu, 7 Jan 2016 12:43:15 +0100 Subject: drm/etnaviv: move runtime PM balance into retire worker The retire worker is kicked for each fence, either the normal way by signaling the fence from the event completion interrupt or by the recover worker if the GPU got stuck. Moving the RPM put into the retire worker allows us to have it in a single place for both cases. This also shaves off quite a bit of the CPU time spent in hardirq context, as arming the autosuspend timer when the RPM refcount drops to 0 is a relatively costly operation. Tested-by: Russell King Acked-by: Russell King Signed-off-by: Lucas Stach --- drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 22 ++++++++-------------- 1 file changed, 8 insertions(+), 14 deletions(-) (limited to 'drivers/gpu/drm/etnaviv/etnaviv_gpu.c') diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index a33162cf4f4c..40f2a37f56e3 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -871,11 +871,6 @@ static void recover_worker(struct work_struct *work) gpu->event[i].fence = NULL; gpu->event[i].used = false; complete(&gpu->event_free); - /* - * Decrement the PM count for each stuck event. This is safe - * even in atomic context as we use ASYNC RPM here. - */ - pm_runtime_put_autosuspend(gpu->dev); } spin_unlock_irqrestore(&gpu->event_spinlock, flags); gpu->completed_fence = gpu->active_fence; @@ -1158,6 +1153,14 @@ static void retire_worker(struct work_struct *work) } etnaviv_gpu_cmdbuf_free(cmdbuf); + /* + * We need to balance the runtime PM count caused by + * each submission. Upon submission, we increment + * the runtime PM counter, and allocate one event. + * So here, we put the runtime PM count for each + * completed event. + */ + pm_runtime_put_autosuspend(gpu->dev); } gpu->retired_fence = fence; @@ -1378,15 +1381,6 @@ static irqreturn_t irq_handler(int irq, void *data) gpu->completed_fence = fence->seqno; event_free(gpu, event); - - /* - * We need to balance the runtime PM count caused by - * each submission. Upon submission, we increment - * the runtime PM counter, and allocate one event. - * So here, we put the runtime PM count for each - * completed event. - */ - pm_runtime_put_autosuspend(gpu->dev); } /* Retire the buffer objects in a work */ -- cgit From 471070abd2f53e579ebeb362e78ce62d04287f49 Mon Sep 17 00:00:00 2001 From: Lucas Stach Date: Mon, 25 Jan 2016 17:03:09 +0100 Subject: drm/etnaviv: move GPU linear window to end of DMA window If the end of the system DMA window is farther away from the start of physical RAM than the size of the GPU linear window, move the linear window so that it ends at the same address than the system DMA window. This allows to map command buffer from CMA, which is likely to reside at the end of the system DMA window, while also overlapping as much RAM as possible, in order to optimize regular buffer mappings through the linear window. Signed-off-by: Lucas Stach --- drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/etnaviv/etnaviv_gpu.c') diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index 40f2a37f56e3..e9e66b99ab7c 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -1563,6 +1563,7 @@ static int etnaviv_gpu_platform_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; struct etnaviv_gpu *gpu; + u32 dma_mask; int err = 0; gpu = devm_kzalloc(dev, sizeof(*gpu), GFP_KERNEL); @@ -1573,12 +1574,16 @@ static int etnaviv_gpu_platform_probe(struct platform_device *pdev) mutex_init(&gpu->lock); /* - * Set the GPU base address to the start of physical memory. This - * ensures that if we have up to 2GB, the v1 MMU can address the - * highest memory. This is important as command buffers may be - * allocated outside of this limit. + * Set the GPU linear window to be at the end of the DMA window, where + * the CMA area is likely to reside. This ensures that we are able to + * map the command buffers while having the linear window overlap as + * much RAM as possible, so we can optimize mappings for other buffers. */ - gpu->memory_base = PHYS_OFFSET; + dma_mask = (u32)dma_get_required_mask(dev); + if (dma_mask < PHYS_OFFSET + SZ_2G) + gpu->memory_base = PHYS_OFFSET; + else + gpu->memory_base = dma_mask - SZ_2G + 1; /* Map registers: */ gpu->mmio = etnaviv_ioremap(pdev, NULL, dev_name(gpu->dev)); -- cgit From f60863116b4026713fba1810927f8639bfd6ae80 Mon Sep 17 00:00:00 2001 From: Russell King Date: Thu, 21 Jan 2016 15:20:19 +0000 Subject: drm: etnaviv: track current execution state Add tracking of the current execution state (iow, active GPU pipe). Signed-off-by: Russell King Signed-off-by: Lucas Stach --- drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/etnaviv/etnaviv_gpu.c') diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index e9e66b99ab7c..2adee13058df 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -628,6 +628,7 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) /* Now program the hardware */ mutex_lock(&gpu->lock); etnaviv_gpu_hw_init(gpu); + gpu->exec_state = -1; mutex_unlock(&gpu->lock); pm_runtime_mark_last_busy(gpu->dev); @@ -877,6 +878,7 @@ static void recover_worker(struct work_struct *work) etnaviv_gpu_hw_init(gpu); gpu->switch_context = true; + gpu->exec_state = -1; mutex_unlock(&gpu->lock); pm_runtime_mark_last_busy(gpu->dev); @@ -1475,6 +1477,7 @@ static int etnaviv_gpu_hw_resume(struct etnaviv_gpu *gpu) etnaviv_gpu_hw_init(gpu); gpu->switch_context = true; + gpu->exec_state = -1; mutex_unlock(&gpu->lock); -- cgit From b6325f409959c7e1065ef1537f2e54cf4d7ab465 Mon Sep 17 00:00:00 2001 From: Russell King Date: Thu, 21 Jan 2016 15:20:50 +0000 Subject: drm: etnaviv: clean up vram_mapping submission/retire path Currently, we scan the list of mappings each time we want to operate on the vram_mapping struct. Rather than repeatedly scanning these, look them up once in the submission path, and then use _reference and _unreference methods as necessary to manage this object. Signed-off-by: Russell King Signed-off-by: Lucas Stach --- drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/etnaviv/etnaviv_gpu.c') diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index 2adee13058df..d13303ce530d 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -1103,7 +1103,7 @@ struct etnaviv_cmdbuf *etnaviv_gpu_cmdbuf_new(struct etnaviv_gpu *gpu, u32 size, size_t nr_bos) { struct etnaviv_cmdbuf *cmdbuf; - size_t sz = size_vstruct(nr_bos, sizeof(cmdbuf->bo[0]), + size_t sz = size_vstruct(nr_bos, sizeof(cmdbuf->bo_map[0]), sizeof(*cmdbuf)); cmdbuf = kzalloc(sz, GFP_KERNEL); @@ -1147,11 +1147,12 @@ static void retire_worker(struct work_struct *work) fence_put(cmdbuf->fence); for (i = 0; i < cmdbuf->nr_bos; i++) { - struct etnaviv_gem_object *etnaviv_obj = cmdbuf->bo[i]; + struct etnaviv_vram_mapping *mapping = cmdbuf->bo_map[i]; + struct etnaviv_gem_object *etnaviv_obj = mapping->object; atomic_dec(&etnaviv_obj->gpu_active); /* drop the refcount taken in etnaviv_gpu_submit */ - etnaviv_gem_put_iova(gpu, &etnaviv_obj->base); + etnaviv_gem_mapping_unreference(mapping); } etnaviv_gpu_cmdbuf_free(cmdbuf); @@ -1309,11 +1310,10 @@ int etnaviv_gpu_submit(struct etnaviv_gpu *gpu, for (i = 0; i < submit->nr_bos; i++) { struct etnaviv_gem_object *etnaviv_obj = submit->bos[i].obj; - u32 iova; - /* Each cmdbuf takes a refcount on the iova */ - etnaviv_gem_get_iova(gpu, &etnaviv_obj->base, &iova); - cmdbuf->bo[i] = etnaviv_obj; + /* Each cmdbuf takes a refcount on the mapping */ + etnaviv_gem_mapping_reference(submit->bos[i].mapping); + cmdbuf->bo_map[i] = submit->bos[i].mapping; atomic_inc(&etnaviv_obj->gpu_active); if (submit->bos[i].flags & ETNA_SUBMIT_BO_WRITE) -- cgit