dri-devel July 2012

dri-devel@lists.freedesktop.org

115 participants
688 discussions

[PATCH 1/3] drm/radeon: move ring locking out of reset path
by Christian König 02 Jul '12

02 Jul '12

Hold the ring lock the whole time the reset is in progress, otherwise another process can submit new jobs. Signed-off-by: Christian König <deathsimple(a)vodafone.de> --- drivers/gpu/drm/radeon/evergreen.c | 8 ++++---- drivers/gpu/drm/radeon/ni.c | 8 ++++---- drivers/gpu/drm/radeon/r100.c | 8 ++++---- drivers/gpu/drm/radeon/r300.c | 4 ++-- drivers/gpu/drm/radeon/r420.c | 8 ++++---- drivers/gpu/drm/radeon/r600.c | 8 … [View More]++++---- drivers/gpu/drm/radeon/radeon_device.c | 14 ++++++++------ drivers/gpu/drm/radeon/rv515.c | 4 ++-- drivers/gpu/drm/radeon/si.c | 12 ++++++------ 9 files changed, 38 insertions(+), 36 deletions(-) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index f716e08..09e6da8 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -1418,7 +1418,7 @@ static int evergreen_cp_start(struct radeon_device *rdev) int r, i; uint32_t cp_me; - r = radeon_ring_lock(rdev, ring, 7); + r = radeon_ring_alloc(rdev, ring, 7); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -1430,12 +1430,12 @@ static int evergreen_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1)); radeon_ring_write(ring, 0); radeon_ring_write(ring, 0); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); cp_me = 0xff; WREG32(CP_ME_CNTL, cp_me); - r = radeon_ring_lock(rdev, ring, evergreen_default_size + 19); + r = radeon_ring_alloc(rdev, ring, evergreen_default_size + 19); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -1473,7 +1473,7 @@ static int evergreen_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */ radeon_ring_write(ring, 0x00000010); /* */ - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); return 0; } diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 2366be3..7e516ce 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -918,7 +918,7 @@ static int cayman_cp_start(struct radeon_device *rdev) struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; int r, i; - r = radeon_ring_lock(rdev, ring, 7); + r = radeon_ring_alloc(rdev, ring, 7); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -930,11 +930,11 @@ static int cayman_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1)); radeon_ring_write(ring, 0); radeon_ring_write(ring, 0); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); cayman_cp_enable(rdev, true); - r = radeon_ring_lock(rdev, ring, cayman_default_size + 19); + r = radeon_ring_alloc(rdev, ring, cayman_default_size + 19); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -972,7 +972,7 @@ static int cayman_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */ radeon_ring_write(ring, 0x00000010); /* */ - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); /* XXX init other rings */ diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index 35825bf..3901a68 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -955,7 +955,7 @@ void r100_ring_start(struct radeon_device *rdev, struct radeon_ring *ring) { int r; - r = radeon_ring_lock(rdev, ring, 2); + r = radeon_ring_alloc(rdev, ring, 2); if (r) { return; } @@ -965,7 +965,7 @@ void r100_ring_start(struct radeon_device *rdev, struct radeon_ring *ring) RADEON_ISYNC_ANY3D_IDLE2D | RADEON_ISYNC_WAIT_IDLEGUI | RADEON_ISYNC_CPSCRATCH_IDLEGUI); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); } @@ -3631,7 +3631,7 @@ int r100_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) return r; } WREG32(scratch, 0xCAFEDEAD); - r = radeon_ring_lock(rdev, ring, 2); + r = radeon_ring_alloc(rdev, ring, 2); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); radeon_scratch_free(rdev, scratch); @@ -3639,7 +3639,7 @@ int r100_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) } radeon_ring_write(ring, PACKET0(scratch, 0)); radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); for (i = 0; i < rdev->usec_timeout; i++) { tmp = RREG32(scratch); if (tmp == 0xDEADBEEF) { diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index 97722a3..32d3510 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -229,7 +229,7 @@ void r300_ring_start(struct radeon_device *rdev, struct radeon_ring *ring) break; } - r = radeon_ring_lock(rdev, ring, 64); + r = radeon_ring_alloc(rdev, ring, 64); if (r) { return; } @@ -293,7 +293,7 @@ void r300_ring_start(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_write(ring, R300_GEOMETRY_ROUND_NEAREST | R300_COLOR_ROUND_NEAREST); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); } void r300_errata(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c index 99137be..b9a4560 100644 --- a/drivers/gpu/drm/radeon/r420.c +++ b/drivers/gpu/drm/radeon/r420.c @@ -208,11 +208,11 @@ static void r420_cp_errata_init(struct radeon_device *rdev) * of the CP init, apparently. */ radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch); - radeon_ring_lock(rdev, ring, 8); + radeon_ring_alloc(rdev, ring, 8); radeon_ring_write(ring, PACKET0(R300_CP_RESYNC_ADDR, 1)); radeon_ring_write(ring, rdev->config.r300.resync_scratch); radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); } static void r420_cp_errata_fini(struct radeon_device *rdev) @@ -222,10 +222,10 @@ static void r420_cp_errata_fini(struct radeon_device *rdev) /* Catch the RESYNC we dispatched all the way back, * at the very beginning of the CP init. */ - radeon_ring_lock(rdev, ring, 8); + radeon_ring_alloc(rdev, ring, 8); radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); radeon_ring_write(ring, R300_RB3D_DC_FINISH); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); radeon_scratch_free(rdev, rdev->config.r300.resync_scratch); } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 43d0c41..fe60bf8 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2066,7 +2066,7 @@ int r600_cp_start(struct radeon_device *rdev) int r; uint32_t cp_me; - r = radeon_ring_lock(rdev, ring, 7); + r = radeon_ring_alloc(rdev, ring, 7); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -2083,7 +2083,7 @@ int r600_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1)); radeon_ring_write(ring, 0); radeon_ring_write(ring, 0); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); cp_me = 0xff; WREG32(R_0086D8_CP_ME_CNTL, cp_me); @@ -2198,7 +2198,7 @@ int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) return r; } WREG32(scratch, 0xCAFEDEAD); - r = radeon_ring_lock(rdev, ring, 3); + r = radeon_ring_alloc(rdev, ring, 3); if (r) { DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", ridx, r); radeon_scratch_free(rdev, scratch); @@ -2207,7 +2207,7 @@ int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2)); radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); for (i = 0; i < rdev->usec_timeout; i++) { tmp = RREG32(scratch); if (tmp == 0xDEADBEEF) diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index f654ba8..9df4130 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -991,26 +991,28 @@ int radeon_gpu_reset(struct radeon_device *rdev) radeon_save_bios_scratch_regs(rdev); /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&rdev->mman.bdev); - radeon_suspend(rdev); + mutex_lock(&rdev->ring_lock); + radeon_suspend(rdev); r = radeon_asic_reset(rdev); if (!r) { dev_info(rdev->dev, "GPU reset succeed\n"); 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); } + mutex_unlock(&rdev->ring_lock); + + /* no matter what restore video mode */ + radeon_restore_bios_scratch_regs(rdev); + drm_helper_resume_force_mode(rdev->ddev); + ttm_bo_unlock_delayed_workqueue(&rdev->mman.bdev, resched); if (r) { /* bad news, how to tell it to userspace ? */ dev_info(rdev->dev, "GPU reset failed\n"); } - return r; } - /* * Debugfs */ diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c index 7f08ced..48441ce 100644 --- a/drivers/gpu/drm/radeon/rv515.c +++ b/drivers/gpu/drm/radeon/rv515.c @@ -57,7 +57,7 @@ void rv515_ring_start(struct radeon_device *rdev, struct radeon_ring *ring) { int r; - r = radeon_ring_lock(rdev, ring, 64); + r = radeon_ring_alloc(rdev, ring, 64); if (r) { return; } @@ -118,7 +118,7 @@ void rv515_ring_start(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_write(ring, GEOMETRY_ROUND_NEAREST | COLOR_ROUND_NEAREST); radeon_ring_write(ring, PACKET0(0x20C8, 0)); radeon_ring_write(ring, 0); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); } int rv515_mc_wait_for_idle(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 34603b3c8..e38a360 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -1851,7 +1851,7 @@ static int si_cp_start(struct radeon_device *rdev) struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; int r, i; - r = radeon_ring_lock(rdev, ring, 7 + 4); + r = radeon_ring_alloc(rdev, ring, 7 + 4); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -1870,11 +1870,11 @@ static int si_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, PACKET3_BASE_INDEX(CE_PARTITION_BASE)); radeon_ring_write(ring, 0xc000); radeon_ring_write(ring, 0xe000); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); si_cp_enable(rdev, true); - r = radeon_ring_lock(rdev, ring, si_default_size + 10); + r = radeon_ring_alloc(rdev, ring, si_default_size + 10); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; @@ -1899,17 +1899,17 @@ static int si_cp_start(struct radeon_device *rdev) radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */ radeon_ring_write(ring, 0x00000010); /* VGT_OUT_DEALLOC_CNTL */ - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); for (i = RADEON_RING_TYPE_GFX_INDEX; i <= CAYMAN_RING_TYPE_CP2_INDEX; ++i) { ring = &rdev->ring[i]; - r = radeon_ring_lock(rdev, ring, 2); + r = radeon_ring_alloc(rdev, ring, 2); /* clear the compute context state */ radeon_ring_write(ring, PACKET3_COMPUTE(PACKET3_CLEAR_STATE, 0)); radeon_ring_write(ring, 0); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_commit(rdev, ring); } return 0; -- 1.7.9.5 [View Less]

3 11

[RFC v2] implicit drm synchronization wip with dma-buf
by Maarten Lankhorst 02 Jul '12

02 Jul '12

Well, V2 time! I was hinted to look at ttm_execbuf_util, and it does indeed contain some nice code. My goal was to extend dma-buf in a generic way now, some elements from v1 are remaining, most notably a dma-buf used for syncing. However it is expected now that dma-buf syncing will work in a very specific way now, slightly more strict than in v1. Instead of each buffer having their own dma-buf I put in 1 per command submission. This submission hasn't been run-time tested yet, but I expect … [View More]the api to go something like this. Intended to be used like this: list_init(&head); add_list_tail(&validate1->entry, &head); add_list_tail(&validate2->entry, &head); add_list_tail(&validate3->entry, &head); r = dmabufmgr_eu_reserve_buffers(&head); if (r) return r; // add waits on cpu or gpu list_for_each_entry(validate, ...) { if (!validate->sync_buf) continue; // Check attachments to see if we already imported sync_buf // somewhere, if not attach to it. // waiting until cur_seq - dmabuf->sync_val >= 0 on either cpu // or as command submitted to gpu // sync_buf itself is a dma-buf, so it should be trivial // TODO: sync_buf should NEVER be validated, add is_sync_buf to dma_buf? // If this step fails: dmabufmgr_eu_backoff_reservation // else: // dmabufmgr_eu_fence_buffer_objects(our_own_sync_buf, // hwchannel * max(minhwalign, 4), ++counter[hwchannel]); // XXX: Do we still require a minimum alignment? I set up 16 for nouveau, // but this is no longer needed in this design since it only matters // for writes for which nouveau would already control the offset. } // Some time after execbuffer was executed, doesn't have to be right away but before // getting in the danger of our own counter wrapping around: // grab dmabufmgr.lru_lock, and cleanup by unreffing sync_buf when // sync_buf == ownbuf, sync_ofs == ownofs, and sync_val == saved_counter // In the meantime someone else or even us might have reserved this dma_buf // again, which is why all those checks are needed before unreffing. diff --git a/drivers/base/Makefile b/drivers/base/Makefile index 5aa2d70..86e7598 100644 --- a/drivers/base/Makefile +++ b/drivers/base/Makefile @@ -10,7 +10,7 @@ obj-$(CONFIG_CMA) += dma-contiguous.o obj-y += power/ obj-$(CONFIG_HAS_DMA) += dma-mapping.o obj-$(CONFIG_HAVE_GENERIC_DMA_COHERENT) += dma-coherent.o -obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o +obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o dma-buf-mgr.o dma-buf-mgr-eu.o obj-$(CONFIG_ISA) += isa.o obj-$(CONFIG_FW_LOADER) += firmware_class.o obj-$(CONFIG_NUMA) += node.o diff --git a/drivers/base/dma-buf-mgr-eu.c b/drivers/base/dma-buf-mgr-eu.c new file mode 100644 index 0000000..27ebc68 --- /dev/null +++ b/drivers/base/dma-buf-mgr-eu.c @@ -0,0 +1,170 @@ +/* + * Copyright (C) 2012 Canonical Ltd + * + * Based on ttm_bo.c which bears the following copyright notice, + * but is dual licensed: + * + * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <linux/dma-buf-mgr.h> +#include <linux/sched.h> +#include <linux/export.h> + +static void dmabufmgr_eu_backoff_reservation_locked(struct list_head *list) +{ + struct dmabufmgr_validate *entry; + + list_for_each_entry(entry, list, head) { + struct dma_buf *bo = entry->bo; + if (!entry->reserved) + continue; + + entry->reserved = false; + atomic_set(&bo->reserved, 0); + wake_up_all(&bo->event_queue); + if (entry->sync_buf) + dma_buf_put(entry->sync_buf); + entry->sync_buf = NULL; + } +} + +static int +dmabufmgr_eu_wait_unreserved_locked(struct list_head *list, + struct dma_buf *bo) +{ + int ret; + + spin_unlock(&dmabufmgr.lru_lock); + ret = dmabufmgr_bo_wait_unreserved(bo, true); + spin_lock(&dmabufmgr.lru_lock); + if (unlikely(ret != 0)) + dmabufmgr_eu_backoff_reservation_locked(list); + return ret; +} + +void +dmabufmgr_eu_backoff_reservation(struct list_head *list) +{ + struct dmabufmgr_validate *entry; + + if (list_empty(list)) + return; + + entry = list_first_entry(list, struct dmabufmgr_validate, head); + spin_lock(&dmabufmgr.lru_lock); + dmabufmgr_eu_backoff_reservation_locked(list); + spin_unlock(&dmabufmgr.lru_lock); +} +EXPORT_SYMBOL_GPL(dmabufmgr_eu_backoff_reservation); + +int +dmabufmgr_eu_reserve_buffers(struct list_head *list) +{ + struct dmabufmgr_validate *entry; + int ret; + u32 val_seq; + + if (list_empty(list)) + return 0; + + list_for_each_entry(entry, list, head) { + entry->reserved = false; + entry->sync_buf = NULL; + } + +retry: + spin_lock(&dmabufmgr.lru_lock); + val_seq = dmabufmgr.counter++; + + list_for_each_entry(entry, list, head) { + struct dma_buf *bo = entry->bo; + +retry_this_bo: + ret = dmabufmgr_bo_reserve_locked(bo, true, true, true, val_seq); + switch (ret) { + case 0: + break; + case -EBUSY: + ret = dmabufmgr_eu_wait_unreserved_locked(list, bo); + if (unlikely(ret != 0)) { + spin_unlock(&dmabufmgr.lru_lock); + return ret; + } + goto retry_this_bo; + case -EAGAIN: + dmabufmgr_eu_backoff_reservation_locked(list); + spin_unlock(&dmabufmgr.lru_lock); + ret = dmabufmgr_bo_wait_unreserved(bo, true); + if (unlikely(ret != 0)) + return ret; + goto retry; + default: + dmabufmgr_eu_backoff_reservation_locked(list); + spin_unlock(&dmabufmgr.lru_lock); + return ret; + } + + entry->reserved = true; + if (bo->sync_buf) + get_dma_buf(bo->sync_buf); + entry->sync_buf = bo->sync_buf; + entry->sync_ofs = bo->sync_ofs; + entry->sync_val = bo->sync_val; + } + spin_unlock(&dmabufmgr.lru_lock); + + return 0; +} +EXPORT_SYMBOL_GPL(dmabufmgr_eu_reserve_buffers); + +void +dmabufmgr_eu_fence_buffer_objects(struct dma_buf *sync_buf, u32 ofs, u32 seq, struct list_head *list) +{ + struct dmabufmgr_validate *entry; + struct dma_buf *bo; + + if (list_empty(list) || WARN_ON(!sync_buf)) + return; + + spin_lock(&dmabufmgr.lru_lock); + + list_for_each_entry(entry, list, head) { + bo = entry->bo; + dmabufmgr_bo_unreserve_locked(bo); + entry->reserved = false; + if (entry->sync_buf) + dma_buf_put(entry->sync_buf); + entry->sync_buf = NULL; + + get_dma_buf(sync_buf); + bo->sync_buf = sync_buf; + bo->sync_ofs = ofs; + bo->sync_val = seq; + } + + spin_unlock(&dmabufmgr.lru_lock); +} +EXPORT_SYMBOL_GPL(dmabufmgr_eu_fence_buffer_objects); diff --git a/drivers/base/dma-buf-mgr.c b/drivers/base/dma-buf-mgr.c new file mode 100644 index 0000000..14756ff --- /dev/null +++ b/drivers/base/dma-buf-mgr.c @@ -0,0 +1,149 @@ +/* + * Copyright (C) 2012 Canonical Ltd + * + * Based on ttm_bo.c which bears the following copyright notice, + * but is dual licensed: + * + * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +/* + * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com> + */ + + +#include <linux/fs.h> +#include <linux/slab.h> +#include <linux/dma-buf-mgr.h> +#include <linux/anon_inodes.h> +#include <linux/export.h> +#include <linux/sched.h> +#include <linux/list.h> + +/* Based on ttm_bo.c with vm_lock and fence_lock removed + * lru_lock takes care of fence_lock as well + */ +struct dmabufmgr dmabufmgr = { + .lru_lock = __SPIN_LOCK_UNLOCKED(dmabufmgr.lru_lock), + .counter = 1, +}; + +int +dmabufmgr_bo_reserve_locked(struct dma_buf *bo, + bool interruptible, bool no_wait, + bool use_sequence, u32 sequence) +{ + int ret; + + while (unlikely(atomic_cmpxchg(&bo->reserved, 0, 1) != 0)) { + /** + * Deadlock avoidance for multi-bo reserving. + */ + if (use_sequence && bo->seq_valid) { + /** + * We've already reserved this one. + */ + if (unlikely(sequence == bo->val_seq)) + return -EDEADLK; + /** + * Already reserved by a thread that will not back + * off for us. We need to back off. + */ + if (unlikely(sequence - bo->val_seq < (1 << 31))) + return -EAGAIN; + } + + if (no_wait) + return -EBUSY; + + spin_unlock(&dmabufmgr.lru_lock); + ret = dmabufmgr_bo_wait_unreserved(bo, interruptible); + spin_lock(&dmabufmgr.lru_lock); + + if (unlikely(ret)) + return ret; + } + + if (use_sequence) { + /** + * Wake up waiters that may need to recheck for deadlock, + * if we decreased the sequence number. + */ + if (unlikely((bo->val_seq - sequence < (1 << 31)) + || !bo->seq_valid)) + wake_up_all(&bo->event_queue); + + bo->val_seq = sequence; + bo->seq_valid = true; + } else { + bo->seq_valid = false; + } + + return 0; +} +EXPORT_SYMBOL_GPL(dmabufmgr_bo_reserve_locked); + +int +dmabufmgr_bo_reserve(struct dma_buf *bo, + bool interruptible, bool no_wait, + bool use_sequence, u32 sequence) +{ + int ret; + + spin_lock(&dmabufmgr.lru_lock); + ret = dmabufmgr_bo_reserve_locked(bo, interruptible, no_wait, + use_sequence, sequence); + spin_unlock(&dmabufmgr.lru_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(dmabufmgr_bo_reserve); + +int +dmabufmgr_bo_wait_unreserved(struct dma_buf *bo, bool interruptible) +{ + if (interruptible) { + return wait_event_interruptible(bo->event_queue, + atomic_read(&bo->reserved) == 0); + } else { + wait_event(bo->event_queue, atomic_read(&bo->reserved) == 0); + return 0; + } +} +EXPORT_SYMBOL_GPL(dmabufmgr_bo_wait_unreserved); + +void dmabufmgr_bo_unreserve_locked(struct dma_buf *bo) +{ + atomic_set(&bo->reserved, 0); + wake_up_all(&bo->event_queue); +} +EXPORT_SYMBOL_GPL(dmabufmgr_bo_unreserve_locked); + +void dmabufmgr_bo_unreserve(struct dma_buf *bo) +{ + spin_lock(&dmabufmgr.lru_lock); + dmabufmgr_bo_unreserve_locked(bo); + spin_unlock(&dmabufmgr.lru_lock); +} +EXPORT_SYMBOL_GPL(dmabufmgr_bo_unreserve); diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c index 24e88fe..01c4f71 100644 --- a/drivers/base/dma-buf.c +++ b/drivers/base/dma-buf.c @@ -40,6 +40,9 @@ static int dma_buf_release(struct inode *inode, struct file *file) dmabuf = file->private_data; dmabuf->ops->release(dmabuf); + BUG_ON(waitqueue_active(&dmabuf->event_queue)); + if (dmabuf->sync_buf) + dma_buf_put(dmabuf->sync_buf); kfree(dmabuf); return 0; } @@ -119,6 +122,7 @@ struct dma_buf *dma_buf_export(void *priv, const struct dma_buf_ops *ops, mutex_init(&dmabuf->lock); INIT_LIST_HEAD(&dmabuf->attachments); + init_waitqueue_head(&dmabuf->event_queue); return dmabuf; } diff --git a/include/linux/dma-buf-mgr.h b/include/linux/dma-buf-mgr.h new file mode 100644 index 0000000..b26462e --- /dev/null +++ b/include/linux/dma-buf-mgr.h @@ -0,0 +1,84 @@ +/* + * Header file for dma buffer sharing framework. + * + * Copyright(C) 2011 Linaro Limited. All rights reserved. + * Author: Sumit Semwal <sumit.semwal(a)ti.com> + * + * Many thanks to linaro-mm-sig list, and specially + * Arnd Bergmann <arnd(a)arndb.de>, Rob Clark <rob(a)ti.com> and + * Daniel Vetter <daniel(a)ffwll.ch> for their support in creation and + * refining of this idea. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that 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, see <http://www.gnu.org/licenses/>. + */ +#ifndef __DMA_BUF_MGR_H__ +#define __DMA_BUF_MGR_H__ + +#include <linux/dma-buf.h> +#include <linux/list.h> + +/** Size of each hwcontext in synchronization dma-buf */ +#define DMABUFMGR_HWCONTEXT_SYNC_ALIGN 16 + +struct dmabufmgr { + spinlock_t lru_lock; + + u32 counter; +}; +extern struct dmabufmgr dmabufmgr; + +extern int +dmabufmgr_bo_reserve_locked(struct dma_buf *bo, + bool interruptible, bool no_wait, + bool use_sequence, u32 sequence); + +extern int +dmabufmgr_bo_reserve(struct dma_buf *bo, + bool interruptible, bool no_wait, + bool use_sequence, u32 sequence); + +extern void +dmabufmgr_bo_unreserve_locked(struct dma_buf *bo); + +extern void +dmabufmgr_bo_unreserve(struct dma_buf *bo); + +extern int +dmabufmgr_bo_wait_unreserved(struct dma_buf *bo, bool interruptible); + +/* execbuf util support for reservations + * matches ttm_execbuf_util + */ +struct dmabufmgr_validate { + struct list_head head; + struct dma_buf *bo; + bool reserved; + + /* If non-null, check for attachments */ + struct dma_buf *sync_buf; + u32 sync_ofs, sync_val; +}; + +/** reserve a linked list of struct dmabufmgr_validate entries */ +extern int +dmabufmgr_eu_reserve_buffers(struct list_head *list); + +/** Undo reservation */ +extern void +dmabufmgr_eu_backoff_reservation(struct list_head *list); + +/** Commit reservation */ +extern void +dmabufmgr_eu_fence_buffer_objects(struct dma_buf *sync_buf, u32 ofs, u32 val, struct list_head *list); + +#endif /* __DMA_BUF_MGR_H__ */ diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h index eb48f38..b2ab395 100644 --- a/include/linux/dma-buf.h +++ b/include/linux/dma-buf.h @@ -113,6 +113,8 @@ struct dma_buf_ops { * @attachments: list of dma_buf_attachment that denotes all devices attached. * @ops: dma_buf_ops associated with this buffer object. * @priv: exporter specific private data for this buffer object. + * @bufmgr_entry: used by dmabufmgr + * @bufdev: used by dmabufmgr */ struct dma_buf { size_t size; @@ -122,6 +124,24 @@ struct dma_buf { /* mutex to serialize list manipulation and attach/detach */ struct mutex lock; void *priv; + + /** dmabufmgr members */ + wait_queue_head_t event_queue; + + /** + * dmabufmgr members protected by the dmabufmgr::lru_lock. + */ + u32 val_seq; + bool seq_valid; + + struct dma_buf *sync_buf; + u32 sync_ofs, sync_val; + + /** + * dmabufmgr members protected by the dmabufmgr::lru_lock + * only when written to. + */ + atomic_t reserved; }; /** [View Less]

1 0

[PATCH] staging: drm/omap: add rotation properties
by Rob Clark 02 Jul '12

02 Jul '12

From: Rob Clark <rob(a)ti.com> Use tiled buffers for rotated/reflected scanout, with CRTC and plane properties as the interface for userspace to configure rotation. Signed-off-by: Rob Clark <rob(a)ti.com> --- drivers/staging/omapdrm/omap_crtc.c | 17 +++++ drivers/staging/omapdrm/omap_dmm_tiler.c | 47 ++++++++++++-- drivers/staging/omapdrm/omap_dmm_tiler.h | 17 ++++- drivers/staging/omapdrm/omap_drv.c | 17 +++++ drivers/staging/omapdrm/omap_drv.h | … [View More]32 +++++++++- drivers/staging/omapdrm/omap_fb.c | 99 +++++++++++++++++++++++++----- drivers/staging/omapdrm/omap_gem.c | 43 ++++++++++++- drivers/staging/omapdrm/omap_plane.c | 92 ++++++++++++++++++++------- 8 files changed, 318 insertions(+), 46 deletions(-) diff --git a/drivers/staging/omapdrm/omap_crtc.c b/drivers/staging/omapdrm/omap_crtc.c index 8b864af..7479dcb 100644 --- a/drivers/staging/omapdrm/omap_crtc.c +++ b/drivers/staging/omapdrm/omap_crtc.c @@ -191,10 +191,25 @@ static int omap_crtc_page_flip_locked(struct drm_crtc *crtc, return 0; } +static int omap_crtc_set_property(struct drm_crtc *crtc, + struct drm_property *property, uint64_t val) +{ + struct omap_crtc *omap_crtc = to_omap_crtc(crtc); + struct omap_drm_private *priv = crtc->dev->dev_private; + + if (property == priv->rotation_prop) { + crtc->invert_dimensions = + !!(val & ((1LL << DRM_ROTATE_90) | (1LL << DRM_ROTATE_270))); + } + + return omap_plane_set_property(omap_crtc->plane, property, val); +} + static const struct drm_crtc_funcs omap_crtc_funcs = { .set_config = drm_crtc_helper_set_config, .destroy = omap_crtc_destroy, .page_flip = omap_crtc_page_flip_locked, + .set_property = omap_crtc_set_property, }; static const struct drm_crtc_helper_funcs omap_crtc_helper_funcs = { @@ -231,6 +246,8 @@ struct drm_crtc *omap_crtc_init(struct drm_device *dev, drm_crtc_init(dev, crtc, &omap_crtc_funcs); drm_crtc_helper_add(crtc, &omap_crtc_helper_funcs); + omap_plane_install_properties(omap_crtc->plane, &crtc->base); + return crtc; fail: diff --git a/drivers/staging/omapdrm/omap_dmm_tiler.c b/drivers/staging/omapdrm/omap_dmm_tiler.c index 9d83060..68da290 100644 --- a/drivers/staging/omapdrm/omap_dmm_tiler.c +++ b/drivers/staging/omapdrm/omap_dmm_tiler.c @@ -401,8 +401,26 @@ int tiler_release(struct tiler_block *block) * Utils */ -/* calculate the tiler space address of a pixel in a view orientation */ -static u32 tiler_get_address(u32 orient, enum tiler_fmt fmt, u32 x, u32 y) +/* calculate the tiler space address of a pixel in a view orientation... + * below description copied from the display subsystem section of TRM: + * + * When the TILER is addressed, the bits: + * [28:27] = 0x0 for 8-bit tiled + * 0x1 for 16-bit tiled + * 0x2 for 32-bit tiled + * 0x3 for page mode + * [31:29] = 0x0 for 0-degree view + * 0x1 for 180-degree view + mirroring + * 0x2 for 0-degree view + mirroring + * 0x3 for 180-degree view + * 0x4 for 270-degree view + mirroring + * 0x5 for 270-degree view + * 0x6 for 90-degree view + * 0x7 for 90-degree view + mirroring + * Otherwise the bits indicated the corresponding bit address to access + * the SDRAM. + */ +static u32 tiler_get_address(enum tiler_fmt fmt, u32 orient, u32 x, u32 y) { u32 x_bits, y_bits, tmp, x_mask, y_mask, alignment; @@ -414,8 +432,11 @@ static u32 tiler_get_address(u32 orient, enum tiler_fmt fmt, u32 x, u32 y) x_mask = MASK(x_bits); y_mask = MASK(y_bits); - if (x < 0 || x > x_mask || y < 0 || y > y_mask) + if (x < 0 || x > x_mask || y < 0 || y > y_mask) { + DBG("invalid coords: %u < 0 || %u > %u || %u < 0 || %u > %u", + x, x, x_mask, y, y, y_mask); return 0; + } /* account for mirroring */ if (orient & MASK_X_INVERT) @@ -436,11 +457,22 @@ dma_addr_t tiler_ssptr(struct tiler_block *block) { BUG_ON(!validfmt(block->fmt)); - return TILVIEW_8BIT + tiler_get_address(0, block->fmt, + return TILVIEW_8BIT + tiler_get_address(block->fmt, 0, block->area.p0.x * geom[block->fmt].slot_w, block->area.p0.y * geom[block->fmt].slot_h); } +dma_addr_t tiler_tsptr(struct tiler_block *block, uint32_t orient, + uint32_t x, uint32_t y) +{ + struct tcm_pt *p = &block->area.p0; + BUG_ON(!validfmt(block->fmt)); + + return tiler_get_address(block->fmt, orient, + (p->x * geom[block->fmt].slot_w) + x, + (p->y * geom[block->fmt].slot_h) + y); +} + void tiler_align(enum tiler_fmt fmt, uint16_t *w, uint16_t *h) { BUG_ON(!validfmt(fmt)); @@ -448,11 +480,14 @@ void tiler_align(enum tiler_fmt fmt, uint16_t *w, uint16_t *h) *h = round_up(*h, geom[fmt].slot_h); } -uint32_t tiler_stride(enum tiler_fmt fmt) +uint32_t tiler_stride(enum tiler_fmt fmt, uint32_t orient) { BUG_ON(!validfmt(fmt)); - return 1 << (CONT_WIDTH_BITS + geom[fmt].y_shft); + if (orient & MASK_XY_FLIP) + return 1 << (CONT_HEIGHT_BITS + geom[fmt].x_shft); + else + return 1 << (CONT_WIDTH_BITS + geom[fmt].y_shft); } size_t tiler_size(enum tiler_fmt fmt, uint16_t w, uint16_t h) diff --git a/drivers/staging/omapdrm/omap_dmm_tiler.h b/drivers/staging/omapdrm/omap_dmm_tiler.h index 7b1052a..740911d 100644 --- a/drivers/staging/omapdrm/omap_dmm_tiler.h +++ b/drivers/staging/omapdrm/omap_dmm_tiler.h @@ -54,7 +54,18 @@ struct tiler_block { #define TILER_WIDTH (1 << (CONT_WIDTH_BITS - SLOT_WIDTH_BITS)) #define TILER_HEIGHT (1 << (CONT_HEIGHT_BITS - SLOT_HEIGHT_BITS)) -/* tiler space addressing bitfields */ +/* +Table 15-11. Coding and Description of TILER Orientations +S Y X Description Alternate description +0 0 0 0-degree view Natural view +0 0 1 0-degree view with vertical mirror 180-degree view with horizontal mirror +0 1 0 0-degree view with horizontal mirror 180-degree view with vertical mirror +0 1 1 180-degree view +1 0 0 90-degree view with vertical mirror 270-degree view with horizontal mirror +1 0 1 270-degree view +1 1 0 90-degree view +1 1 1 90-degree view with horizontal mirror 270-degree view with vertical mirror + */ #define MASK_XY_FLIP (1 << 31) #define MASK_Y_INVERT (1 << 30) #define MASK_X_INVERT (1 << 29) @@ -90,7 +101,9 @@ int tiler_release(struct tiler_block *block); /* utilities */ dma_addr_t tiler_ssptr(struct tiler_block *block); -uint32_t tiler_stride(enum tiler_fmt fmt); +dma_addr_t tiler_tsptr(struct tiler_block *block, uint32_t orient, + uint32_t x, uint32_t y); +uint32_t tiler_stride(enum tiler_fmt fmt, uint32_t orient); size_t tiler_size(enum tiler_fmt fmt, uint16_t w, uint16_t h); size_t tiler_vsize(enum tiler_fmt fmt, uint16_t w, uint16_t h); void tiler_align(enum tiler_fmt fmt, uint16_t *w, uint16_t *h); diff --git a/drivers/staging/omapdrm/omap_drv.c b/drivers/staging/omapdrm/omap_drv.c index 4beab94..342645a 100644 --- a/drivers/staging/omapdrm/omap_drv.c +++ b/drivers/staging/omapdrm/omap_drv.c @@ -649,6 +649,8 @@ static int dev_firstopen(struct drm_device *dev) */ static void dev_lastclose(struct drm_device *dev) { + int i; + /* we don't support vga-switcheroo.. so just make sure the fbdev * mode is active */ @@ -657,6 +659,21 @@ static void dev_lastclose(struct drm_device *dev) DBG("lastclose: dev=%p", dev); + /* need to restore default rotation state.. not sure if there is + * a cleaner way to restore properties to default state? Maybe + * a flag that properties should automatically be restored to + * default state on lastclose? + */ + for (i = 0; i < priv->num_crtcs; i++) { + drm_object_property_set_value(&priv->crtcs[i]->base, + priv->rotation_prop, 0); + } + + for (i = 0; i < priv->num_planes; i++) { + drm_object_property_set_value(&priv->planes[i]->base, + priv->rotation_prop, 0); + } + ret = drm_fb_helper_restore_fbdev_mode(priv->fbdev); if (ret) DBG("failed to restore crtc mode"); diff --git a/drivers/staging/omapdrm/omap_drv.h b/drivers/staging/omapdrm/omap_drv.h index f238d57..bfdf9d3 100644 --- a/drivers/staging/omapdrm/omap_drv.h +++ b/drivers/staging/omapdrm/omap_drv.h @@ -59,6 +59,26 @@ struct omap_drm_private { struct list_head obj_list; bool has_dmm; + + /* properties: */ + struct drm_property *rotation_prop; +}; + +/* this should probably be in drm-core to standardize amongst drivers */ +#define DRM_ROTATE_0 0 +#define DRM_ROTATE_90 1 +#define DRM_ROTATE_180 2 +#define DRM_ROTATE_270 3 +#define DRM_REFLECT_X 4 +#define DRM_REFLECT_Y 5 + +/* parameters which describe (unrotated) coordinates of scanout within a fb: */ +struct omap_drm_window { + uint32_t rotation; + int32_t crtc_x, crtc_y; /* signed because can be offscreen */ + uint32_t crtc_w, crtc_h; + uint32_t src_x, src_y; + uint32_t src_w, src_h; }; #ifdef CONFIG_DEBUG_FS @@ -87,6 +107,10 @@ int omap_plane_mode_set(struct drm_plane *plane, uint32_t src_w, uint32_t src_h); void omap_plane_on_endwin(struct drm_plane *plane, void (*fxn)(void *), void *arg); +void omap_plane_install_properties(struct drm_plane *plane, + struct drm_mode_object *obj); +int omap_plane_set_property(struct drm_plane *plane, + struct drm_property *property, uint64_t val); struct drm_encoder *omap_encoder_init(struct drm_device *dev, struct omap_overlay_manager *mgr); @@ -114,8 +138,8 @@ struct drm_gem_object *omap_framebuffer_bo(struct drm_framebuffer *fb, int p); int omap_framebuffer_replace(struct drm_framebuffer *a, struct drm_framebuffer *b, void *arg, void (*unpin)(void *arg, struct drm_gem_object *bo)); -void omap_framebuffer_update_scanout(struct drm_framebuffer *fb, int x, int y, - struct omap_overlay_info *info); +void omap_framebuffer_update_scanout(struct drm_framebuffer *fb, + struct omap_drm_window *win, struct omap_overlay_info *info); struct drm_connector *omap_framebuffer_get_next_connector( struct drm_framebuffer *fb, struct drm_connector *from); void omap_framebuffer_flush(struct drm_framebuffer *fb, @@ -157,8 +181,12 @@ int omap_gem_get_pages(struct drm_gem_object *obj, struct page ***pages, bool remap); int omap_gem_put_pages(struct drm_gem_object *obj); uint32_t omap_gem_flags(struct drm_gem_object *obj); +int omap_gem_rotated_paddr(struct drm_gem_object *obj, uint32_t orient, + int x, int y, dma_addr_t *paddr); uint64_t omap_gem_mmap_offset(struct drm_gem_object *obj); size_t omap_gem_mmap_size(struct drm_gem_object *obj); +int omap_gem_tiled_size(struct drm_gem_object *obj, uint16_t *w, uint16_t *h); +int omap_gem_tiled_stride(struct drm_gem_object *obj, uint32_t orient); struct dma_buf * omap_gem_prime_export(struct drm_device *dev, struct drm_gem_object *obj, int flags); diff --git a/drivers/staging/omapdrm/omap_fb.c b/drivers/staging/omapdrm/omap_fb.c index 74260f0..446801d 100644 --- a/drivers/staging/omapdrm/omap_fb.c +++ b/drivers/staging/omapdrm/omap_fb.c @@ -18,6 +18,7 @@ */ #include "omap_drv.h" +#include "omap_dmm_tiler.h" #include "drm_crtc.h" #include "drm_crtc_helper.h" @@ -137,30 +138,100 @@ static const struct drm_framebuffer_funcs omap_framebuffer_funcs = { .dirty = omap_framebuffer_dirty, }; +static uint32_t get_linear_addr(struct plane *plane, + const struct format *format, int n, int x, int y) +{ + uint32_t offset; + + offset = plane->offset + + (x * format->planes[n].stride_bpp) + + (y * plane->pitch / format->planes[n].sub_y); + + return plane->paddr + offset; +} + /* update ovl info for scanout, handles cases of multi-planar fb's, etc. */ -void omap_framebuffer_update_scanout(struct drm_framebuffer *fb, int x, int y, - struct omap_overlay_info *info) +void omap_framebuffer_update_scanout(struct drm_framebuffer *fb, + struct omap_drm_window *win, struct omap_overlay_info *info) { struct omap_framebuffer *omap_fb = to_omap_framebuffer(fb); const struct format *format = omap_fb->format; struct plane *plane = &omap_fb->planes[0]; - unsigned int offset; + uint32_t x, y, orient = 0; + + info->color_mode = format->dss_format; + + info->pos_x = win->crtc_x; + info->pos_y = win->crtc_y; + info->out_width = win->crtc_w; + info->out_height = win->crtc_h; + info->width = win->src_w; + info->height = win->src_h; + + x = win->src_x; + y = win->src_y; + + if (omap_gem_flags(plane->bo) & OMAP_BO_TILED) { + uint32_t w = win->src_w; + uint32_t h = win->src_h; + + switch (win->rotation & 0xf) { + default: + dev_err(fb->dev->dev, "invalid rotation: %02x", + (uint32_t)win->rotation); + /* fallthru to default to no rotation */ + case 0: + case BIT(DRM_ROTATE_0): + orient = 0; + break; + case BIT(DRM_ROTATE_90): + orient = MASK_XY_FLIP | MASK_X_INVERT; + break; + case BIT(DRM_ROTATE_180): + orient = MASK_X_INVERT | MASK_Y_INVERT; + break; + case BIT(DRM_ROTATE_270): + orient = MASK_XY_FLIP | MASK_Y_INVERT; + break; + } - offset = plane->offset + - (x * format->planes[0].stride_bpp) + - (y * plane->pitch / format->planes[0].sub_y); + if (win->rotation & BIT(DRM_REFLECT_X)) + orient ^= MASK_X_INVERT; + + if (win->rotation & BIT(DRM_REFLECT_Y)) + orient ^= MASK_Y_INVERT; + + /* adjust x,y offset for flip/invert: */ + if (orient & MASK_XY_FLIP) + swap(w, h); + if (orient & MASK_Y_INVERT) + y += h - 1; + if (orient & MASK_X_INVERT) + x += w - 1; - info->color_mode = format->dss_format; - info->paddr = plane->paddr + offset; - info->screen_width = plane->pitch / format->planes[0].stride_bpp; + omap_gem_rotated_paddr(plane->bo, orient, x, y, &info->paddr); + info->rotation_type = OMAP_DSS_ROT_TILER; + info->screen_width = omap_gem_tiled_stride(plane->bo, orient); + } else { + info->paddr = get_linear_addr(plane, format, 0, x, y); + info->rotation_type = OMAP_DSS_ROT_DMA; + info->screen_width = plane->pitch; + } + + /* convert to pixels: */ + info->screen_width /= format->planes[0].stride_bpp; if (format->dss_format == OMAP_DSS_COLOR_NV12) { plane = &omap_fb->planes[1]; - offset = plane->offset + - (x * format->planes[1].stride_bpp) + - (y * plane->pitch / format->planes[1].sub_y); - info->p_uv_addr = plane->paddr + offset; + + if (info->rotation_type == OMAP_DSS_ROT_TILER) { + WARN_ON(!(omap_gem_flags(plane->bo) & OMAP_BO_TILED)); + omap_gem_rotated_paddr(plane->bo, orient, + x/2, y/2, &info->p_uv_addr); + } else { + info->p_uv_addr = get_linear_addr(plane, format, 1, x, y); + } } else { info->p_uv_addr = 0; } @@ -377,7 +448,7 @@ struct drm_framebuffer *omap_framebuffer_init(struct drm_device *dev, size = pitch * mode_cmd->height / format->planes[i].sub_y; - if (size > (bos[i]->size - mode_cmd->offsets[i])) { + if (size > (omap_gem_mmap_size(bos[i]) - mode_cmd->offsets[i])) { dev_err(dev->dev, "provided buffer object is too small! %d < %d\n", bos[i]->size - mode_cmd->offsets[i], size); ret = -EINVAL; diff --git a/drivers/staging/omapdrm/omap_gem.c b/drivers/staging/omapdrm/omap_gem.c index 3a0d035..74082aa 100644 --- a/drivers/staging/omapdrm/omap_gem.c +++ b/drivers/staging/omapdrm/omap_gem.c @@ -339,6 +339,17 @@ size_t omap_gem_mmap_size(struct drm_gem_object *obj) return size; } +/* get tiled size, returns -EINVAL if not tiled buffer */ +int omap_gem_tiled_size(struct drm_gem_object *obj, uint16_t *w, uint16_t *h) +{ + struct omap_gem_object *omap_obj = to_omap_bo(obj); + if (omap_obj->flags & OMAP_BO_TILED) { + *w = omap_obj->width; + *h = omap_obj->height; + return 0; + } + return -EINVAL; +} /* Normal handling for the case of faulting in non-tiled buffers */ static int fault_1d(struct drm_gem_object *obj, @@ -832,6 +843,36 @@ fail: return ret; } +/* Get rotated scanout address (only valid if already pinned), at the + * specified orientation and x,y offset from top-left corner of buffer + * (only valid for tiled 2d buffers) + */ +int omap_gem_rotated_paddr(struct drm_gem_object *obj, uint32_t orient, + int x, int y, dma_addr_t *paddr) +{ + struct omap_gem_object *omap_obj = to_omap_bo(obj); + int ret = -EINVAL; + + mutex_lock(&obj->dev->struct_mutex); + if ((omap_obj->paddr_cnt > 0) && omap_obj->block && + (omap_obj->flags & OMAP_BO_TILED)) { + *paddr = tiler_tsptr(omap_obj->block, orient, x, y); + ret = 0; + } + mutex_unlock(&obj->dev->struct_mutex); + return ret; +} + +/* Get tiler stride for the buffer (only valid for 2d tiled buffers) */ +int omap_gem_tiled_stride(struct drm_gem_object *obj, uint32_t orient) +{ + struct omap_gem_object *omap_obj = to_omap_bo(obj); + int ret = -EINVAL; + if (omap_obj->flags & OMAP_BO_TILED) + ret = tiler_stride(gem2fmt(omap_obj->flags), orient); + return ret; +} + /* acquire pages when needed (for example, for DMA where physically * contiguous buffer is not required */ @@ -1402,7 +1443,7 @@ void omap_gem_init(struct drm_device *dev) */ usergart[i].height = h; usergart[i].height_shift = ilog2(h); - usergart[i].stride_pfn = tiler_stride(fmts[i]) >> PAGE_SHIFT; + usergart[i].stride_pfn = tiler_stride(fmts[i], 0) >> PAGE_SHIFT; usergart[i].slot_shift = ilog2((PAGE_SIZE / h) >> i); for (j = 0; j < NUM_USERGART_ENTRIES; j++) { struct usergart_entry *entry = &usergart[i].entry[j]; diff --git a/drivers/staging/omapdrm/omap_plane.c b/drivers/staging/omapdrm/omap_plane.c index 7997be7..6931d06 100644 --- a/drivers/staging/omapdrm/omap_plane.c +++ b/drivers/staging/omapdrm/omap_plane.c @@ -20,6 +20,7 @@ #include <linux/kfifo.h> #include "omap_drv.h" +#include "omap_dmm_tiler.h" /* some hackery because omapdss has an 'enum omap_plane' (which would be * better named omap_plane_id).. and compiler seems unhappy about having @@ -43,10 +44,9 @@ struct omap_plane { struct omap_overlay *ovl; struct omap_overlay_info info; - /* Source values, converted to integers because we don't support - * fractional positions: - */ - unsigned int src_x, src_y; + /* position/orientation of scanout within the fb: */ + struct omap_drm_window win; + /* last fb that we pinned: */ struct drm_framebuffer *pinned_fb; @@ -289,6 +289,7 @@ static void update_scanout(struct drm_plane *plane) { struct omap_plane *omap_plane = to_omap_plane(plane); struct omap_overlay_info *info = &omap_plane->info; + struct omap_drm_window *win = &omap_plane->win; int ret; ret = update_pin(plane, plane->fb); @@ -299,11 +300,10 @@ static void update_scanout(struct drm_plane *plane) return; } - omap_framebuffer_update_scanout(plane->fb, - omap_plane->src_x, omap_plane->src_y, info); + omap_framebuffer_update_scanout(plane->fb, win, info); DBG("%s: %d,%d: %08x %08x (%d)", omap_plane->ovl->name, - omap_plane->src_x, omap_plane->src_y, + win->src_x, win->src_y, (u32)info->paddr, (u32)info->p_uv_addr, info->screen_width); } @@ -316,21 +316,18 @@ int omap_plane_mode_set(struct drm_plane *plane, uint32_t src_w, uint32_t src_h) { struct omap_plane *omap_plane = to_omap_plane(plane); + struct omap_drm_window *win = &omap_plane->win; + + win->crtc_x = crtc_x; + win->crtc_y = crtc_y; + win->crtc_w = crtc_w; + win->crtc_h = crtc_h; /* src values are in Q16 fixed point, convert to integer: */ - src_x = src_x >> 16; - src_y = src_y >> 16; - src_w = src_w >> 16; - src_h = src_h >> 16; - - omap_plane->info.pos_x = crtc_x; - omap_plane->info.pos_y = crtc_y; - omap_plane->info.out_width = crtc_w; - omap_plane->info.out_height = crtc_h; - omap_plane->info.width = src_w; - omap_plane->info.height = src_h; - omap_plane->src_x = src_x; - omap_plane->src_y = src_y; + win->src_x = src_x >> 16; + win->src_y = src_y >> 16; + win->src_w = src_w >> 16; + win->src_h = src_h >> 16; /* note: this is done after this fxn returns.. but if we need * to do a commit/update_scanout, etc before this returns we @@ -359,6 +356,8 @@ static int omap_plane_update(struct drm_plane *plane, static int omap_plane_disable(struct drm_plane *plane) { + struct omap_plane *omap_plane = to_omap_plane(plane); + omap_plane->win.rotation = BIT(DRM_ROTATE_0); return omap_plane_dpms(plane, DRM_MODE_DPMS_OFF); } @@ -409,10 +408,60 @@ void omap_plane_on_endwin(struct drm_plane *plane, install_irq(plane); } +/* helper to install properties which are common to planes and crtcs */ +void omap_plane_install_properties(struct drm_plane *plane, + struct drm_mode_object *obj) +{ + struct drm_device *dev = plane->dev; + struct omap_drm_private *priv = dev->dev_private; + struct drm_property *prop; + + prop = priv->rotation_prop; + if (!prop) { + const struct drm_prop_enum_list props[] = { + { DRM_ROTATE_0, "rotate-0" }, + { DRM_ROTATE_90, "rotate-90" }, + { DRM_ROTATE_180, "rotate-180" }, + { DRM_ROTATE_270, "rotate-270" }, + { DRM_REFLECT_X, "reflect-x" }, + { DRM_REFLECT_Y, "reflect-y" }, + }; + prop = drm_property_create_bitmask(dev, 0, "rotation", + props, ARRAY_SIZE(props)); + if (prop == NULL) + return; + priv->rotation_prop = prop; + } + drm_object_attach_property(obj, prop, 0); +} + +int omap_plane_set_property(struct drm_plane *plane, + struct drm_property *property, uint64_t val) +{ + struct omap_plane *omap_plane = to_omap_plane(plane); + struct omap_drm_private *priv = plane->dev->dev_private; + int ret = -EINVAL; + + if (property == priv->rotation_prop) { + struct omap_overlay *ovl = omap_plane->ovl; + + DBG("%s: rotation: %02x", ovl->name, (uint32_t)val); + omap_plane->win.rotation = val; + + if (ovl->is_enabled(ovl)) + ret = omap_plane_dpms(plane, DRM_MODE_DPMS_ON); + else + ret = 0; + } + + return ret; +} + static const struct drm_plane_funcs omap_plane_funcs = { .update_plane = omap_plane_update, .disable_plane = omap_plane_disable, .destroy = omap_plane_destroy, + .set_property = omap_plane_set_property, }; /* initialize plane */ @@ -455,6 +504,8 @@ struct drm_plane *omap_plane_init(struct drm_device *dev, drm_plane_init(dev, plane, possible_crtcs, &omap_plane_funcs, omap_plane->formats, omap_plane->nformats, priv); + omap_plane_install_properties(plane, &plane->base); + /* get our starting configuration, set defaults for parameters * we don't currently use, etc: */ @@ -463,7 +514,6 @@ struct drm_plane *omap_plane_init(struct drm_device *dev, omap_plane->info.rotation = OMAP_DSS_ROT_0; omap_plane->info.global_alpha = 0xff; omap_plane->info.mirror = 0; - omap_plane->info.mirror = 0; /* Set defaults depending on whether we are a CRTC or overlay * layer. -- 1.7.9.5 [View Less]

4 6

[PATCH 00/10] Start documenting the radeon drm better
by alexdeucher＠gmail.com 02 Jul '12

02 Jul '12

From: Alex Deucher <alexander.deucher(a)amd.com> This is something I've been wanting to do for a while and I finally spent a little time getting a start on it. There is still a lot to do and not all of my descriptions are great, but I think we can document the rest in bits and pieces. I also added a note about what asics the function is applicable to. I tried to start with the more common and complex code. I was thinking it would make sense to have an informal documentation policy … [View More]

5 27

[PATCH] drm/i915: kick any firmware framebuffers before claiming the gtt
by Daniel Vetter 01 Jul '12

01 Jul '12

Especially vesafb likes to map everything as uc- (yikes), and if that mapping hangs around still while we try to map the gtt as wc the kernel will downgrade our request to uc-, resulting in abyssal performance. Unfortunately we can't do this as early as readon does (i.e. as the first thing we do when initializing the hw) because our fb/mmio space region moves around on a per-gen basis. So I've had to move it below the gtt initialization, but that seems to work, too. The important thing is that … [View More]we do this before we set up the gtt wc mapping. Now an altogether different question is why people compile their kernels with vesafb enabled, but I guess making things just work isn't bad per se ... v2: - s/radeondrmfb/inteldrmfb/ - fix up error handling v3: Kill #ifdef X86, this is Intel after all. Noticed by Ben Widawsky. v4: Jani Nikula complained about the pointless bool primary initialization. v5: Don't oops if we can't allocate, noticed by Chris Wilson. v6: Resolve conflicts with agp rework and fixup whitespace. This is commit e188719a2891f01b3100d in drm-next. Backport to 3.5 -fixes queue requested by Dave Airlie - due to grub using vesa on fedora their initrd seems to load vesafb before loading the real kms driver. So tons more people actually experience a dead-slow gpu. Hence also the Cc: stable. Cc: Dave Airlie <airlied(a)linux.ie> Cc: stable(a)vger.kernel.org Reported-and-tested-by: "Kilarski, Bernard R" <bernard.r.kilarski(a)intel.com> Reviewed-by: Chris Wilson <chris(a)chris-wilson.co.uk> Signed-Off-by: Daniel Vetter <daniel.vetter(a)ffwll.ch> --- drivers/gpu/drm/i915/i915_dma.c | 37 ++++++++++++++++++++++++++++++------- 1 file changed, 30 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index f947926..36822b9 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1401,6 +1401,27 @@ i915_mtrr_setup(struct drm_i915_private *dev_priv, unsigned long base, } } +static void i915_kick_out_firmware_fb(struct drm_i915_private *dev_priv) +{ + struct apertures_struct *ap; + struct pci_dev *pdev = dev_priv->dev->pdev; + bool primary; + + ap = alloc_apertures(1); + if (!ap) + return; + + ap->ranges[0].base = dev_priv->dev->agp->base; + ap->ranges[0].size = + dev_priv->mm.gtt->gtt_mappable_entries << PAGE_SHIFT; + primary = + pdev->resource[PCI_ROM_RESOURCE].flags & IORESOURCE_ROM_SHADOW; + + remove_conflicting_framebuffers(ap, "inteldrmfb", primary); + + kfree(ap); +} + /** * i915_driver_load - setup chip and create an initial config * @dev: DRM device @@ -1446,6 +1467,15 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) goto free_priv; } + dev_priv->mm.gtt = intel_gtt_get(); + if (!dev_priv->mm.gtt) { + DRM_ERROR("Failed to initialize GTT\n"); + ret = -ENODEV; + goto put_bridge; + } + + i915_kick_out_firmware_fb(dev_priv); + pci_set_master(dev->pdev); /* overlay on gen2 is broken and can't address above 1G */ @@ -1471,13 +1501,6 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) goto put_bridge; } - dev_priv->mm.gtt = intel_gtt_get(); - if (!dev_priv->mm.gtt) { - DRM_ERROR("Failed to initialize GTT\n"); - ret = -ENODEV; - goto out_rmmap; - } - aperture_size = dev_priv->mm.gtt->gtt_mappable_entries << PAGE_SHIFT; dev_priv->mm.gtt_mapping = -- 1.7.10 [View Less]

1 0

[Bug 43448] New: No native 3D acceleration with Radeon FirePro M7740 (m97 / rv740) -> llvm pipe
by bugzilla-daemon＠freedesktop.org 01 Jul '12

01 Jul '12

https://bugs.freedesktop.org/show_bug.cgi?id=43448 Bug #: 43448 Summary: No native 3D acceleration with Radeon FirePro M7740 (m97 / rv740) -> llvm pipe Classification: Unclassified Product: Mesa Version: 7.11 Platform: x86-64 (AMD64) OS/Version: Linux (All) Status: NEW Severity: major Priority: medium Component: Drivers/DRI/R600 AssignedTo: dri-devel(a)… [View More]

1 31

[Bug 43367] New: radeon: NULL pointer dereference on switching to dynpm power method
by bugzilla-daemon＠bugzilla.kernel.org 01 Jul '12

01 Jul '12

https://bugzilla.kernel.org/show_bug.cgi?id=43367 Summary: radeon: NULL pointer dereference on switching to dynpm power method Product: Drivers Version: 2.5 Kernel Version: 3.5-rc2 Platform: All OS/Version: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: Video(DRI - non Intel) AssignedTo: drivers_video-dri(a)kernel-bugs.osdl.org … [View More]

1 3

[Bug 43353] New: nouveau will not load or boot
by bugzilla-daemon＠bugzilla.kernel.org 01 Jul '12

01 Jul '12

https://bugzilla.kernel.org/show_bug.cgi?id=43353 Summary: nouveau will not load or boot Product: Drivers Version: 2.5 Kernel Version: 3.4 Platform: All OS/Version: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: Video(DRI - non Intel) AssignedTo: drivers_video-dri(a)kernel-bugs.osdl.org ReportedBy: kvs(a)inbox.ru Regression: Yes … [View More]

1 2

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

dri-devel July 2012