dri-devel May 2017

dri-devel@lists.freedesktop.org

207 participants
1452 discussions

Resizeable PCI BAR support V4
by Christian König 04 May '17

04 May '17

Hi everyone, This is the fourth incarnation of this set of patches. It enables device drivers to resize and most likely also relocate the PCI BAR of devices they manage to allow the CPU to access all of the device local memory at once. This is very useful for GFX device drivers where the default PCI BAR is only about 256MB in size for compatibility reasons, but the device easily have multiple gigabyte of local memory. Some changes since V3: 1. A lot of minor style cleanups. 2. Make internal functions for changing BARs directly private to the PCI subsystem. 3. Fail if any BAR is still in use when we try to change it. 4. Handle intermediate bridges as well. 5. Print some more messages when changing something. Please review and/or comment, Christian.

3 21

[PATCH] drm/amd/amdgpu: get rid of else branch
by Nikola Pajkovsky 04 May '17

04 May '17

else branch is pointless if it's right at the end of function and use unlikely() on err path. Signed-off-by: Nikola Pajkovsky <npajkovsky(a)suse.cz> --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 45 +++++++++++++++++++------------------ 1 file changed, 23 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index c1b913541739..8ab250d4761a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1594,30 +1594,31 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, void *sr unsigned occupied, chunk1, chunk2; void *dst; - if (ring->count_dw < count_dw) { + if (unlikely(ring->count_dw < count_dw)) { DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); - } else { - occupied = ring->wptr & ring->ptr_mask; - dst = (void *)&ring->ring[occupied]; - chunk1 = ring->ptr_mask + 1 - occupied; - chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1; - chunk2 = count_dw - chunk1; - chunk1 <<= 2; - chunk2 <<= 2; - - if (chunk1) - memcpy(dst, src, chunk1); - - if (chunk2) { - src += chunk1; - dst = (void *)ring->ring; - memcpy(dst, src, chunk2); - } - - ring->wptr += count_dw; - ring->wptr &= ring->ptr_mask; - ring->count_dw -= count_dw; + return; } + + occupied = ring->wptr & ring->ptr_mask; + dst = (void *)&ring->ring[occupied]; + chunk1 = ring->ptr_mask + 1 - occupied; + chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1; + chunk2 = count_dw - chunk1; + chunk1 <<= 2; + chunk2 <<= 2; + + if (chunk1) + memcpy(dst, src, chunk1); + + if (chunk2) { + src += chunk1; + dst = (void *)ring->ring; + memcpy(dst, src, chunk2); + } + + ring->wptr += count_dw; + ring->wptr &= ring->ptr_mask; + ring->count_dw -= count_dw; } static inline struct amdgpu_sdma_instance * -- 2.12.2

3 2

[RFC] drm/amd/amdgpu: get rid of else branch
by Nikola Pajkovsky 04 May '17

04 May '17

This is super simple elimination of else branch and I should probably even use unlikely in if (ring->count_dw < count_dw) { However, amdgpu_ring_write() has similar if condition, but does not return after DRM_ERROR and it looks suspicious. On error, we still adding v to ring and keeping count_dw-- below zero. if (ring->count_dw <= 0) DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); ring->ring[ring->wptr++] = v; ring->wptr &= ring->ptr_mask; ring->count_dw--; I can obviously be totaly wrong. Hmm? --------8<--------8<--------8<--------8<--------8<--------8<--------8<--------8<--------8< diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index c1b913541739..c6f4f874ea68 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1596,28 +1596,29 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, void *sr if (ring->count_dw < count_dw) { DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); - } else { - occupied = ring->wptr & ring->ptr_mask; - dst = (void *)&ring->ring[occupied]; - chunk1 = ring->ptr_mask + 1 - occupied; - chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1; - chunk2 = count_dw - chunk1; - chunk1 <<= 2; - chunk2 <<= 2; - - if (chunk1) - memcpy(dst, src, chunk1); - - if (chunk2) { - src += chunk1; - dst = (void *)ring->ring; - memcpy(dst, src, chunk2); - } - - ring->wptr += count_dw; - ring->wptr &= ring->ptr_mask; - ring->count_dw -= count_dw; + return; } + + occupied = ring->wptr & ring->ptr_mask; + dst = (void *)&ring->ring[occupied]; + chunk1 = ring->ptr_mask + 1 - occupied; + chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1; + chunk2 = count_dw - chunk1; + chunk1 <<= 2; + chunk2 <<= 2; + + if (chunk1) + memcpy(dst, src, chunk1); + + if (chunk2) { + src += chunk1; + dst = (void *)ring->ring; + memcpy(dst, src, chunk2); + } + + ring->wptr += count_dw; + ring->wptr &= ring->ptr_mask; + ring->count_dw -= count_dw; } static inline struct amdgpu_sdma_instance * -- 2.12.2

2 3

[PATCH 0/2] Introduce crtc->mode_valid() callback
by Jose Abreu 04 May '17

04 May '17

This patchset introduces a new callback for crtc, called mode_valid() that is responsible to limit the number of probbed modes. Just like connector->mode_valid(), this new callback is called at mode probbing stage so that we can validate the mode. This is specially useful because arcpgu crtc is responsible to set a clock value in the commit() stage but unfortunatelly this clock does not support all the needed ranges. This way we can restrict the number of modes that are handed to userspace, so that we dont hand a mode that will fail the commit() stage. I guess this can also happen for other drivers so we introduce the callback in the core. The behaviour remains the same for crtcs that don't have the callback. Also, for a given set of crtcs that can be bound to the connector, if at least one can display the mode then the mode will be probbed. For more info about why this is needed in arcpgu, please refer here [1]. [1] https://patchwork.kernel.org/patch/9694177/ Jose Abreu (2): drm: Introduce crtc->mode_valid() callback drm: arcpgu: Use crtc->mode_valid() callback Cc: Carlos Palminha <palminha(a)synopsys.com> Cc: Alexey Brodkin <abrodkin(a)synopsys.com> Cc: Ville Syrjälä <ville.syrjala(a)linux.intel.com> Cc: Daniel Vetter <daniel.vetter(a)ffwll.ch> Cc: Dave Airlie <airlied(a)linux.ie> drivers/gpu/drm/arc/arcpgu_crtc.c | 14 ++++++++++ drivers/gpu/drm/drm_probe_helper.c | 44 ++++++++++++++++++++++++++++++++ include/drm/drm_modeset_helper_vtables.h | 26 +++++++++++++++++++ 3 files changed, 84 insertions(+) -- 1.9.1

4 19

[PATCH 0/4] IPu tiled scanout support
by Lucas Stach 04 May '17

04 May '17

This adds the required bits to the ipu-v3 and imx-drm driver to make scanout of the Vivante (super-)tiled single buffer format work on i.MX6 QuadPlus. A userspace taking advantage of this support is able to skip the linear resolve step in the GPU, saving quite a lot of memory bandwidth especially with higher resolution displays. The first 3 patches can go in as-is, while the last patch depends on Ben Widawsky's series to add a format modifier blob property to the DRM planes. Regards, Lucas Lucas Stach (4): gpu: ipu-v3: pre: add tiled prefetch support gpu: ipu-v3: prg: add modifier support drm/imx: add FB modifier support drm/imx: advertise supported plane format modifiers drivers/gpu/drm/imx/imx-drm-core.c | 1 + drivers/gpu/drm/imx/ipuv3-plane.c | 92 +++++++++++++++++++++++++++++++++++--- drivers/gpu/ipu-v3/ipu-pre.c | 24 +++++++++- drivers/gpu/ipu-v3/ipu-prg.c | 13 ++++-- drivers/gpu/ipu-v3/ipu-prv.h | 4 +- include/video/imx-ipu-v3.h | 2 +- 6 files changed, 122 insertions(+), 14 deletions(-) -- 2.11.0

3 10

[PATCH] gpu: ipu-v3: prg: remove counter load enable
by Lucas Stach 04 May '17

04 May '17

The counter load enable bit has no effect when the shadow register set is activated. As we always operate the PRG with shadow enabled it is safe to remove this. Signed-off-by: Lucas Stach <l.stach(a)pengutronix.de> --- drivers/gpu/ipu-v3/ipu-prg.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/gpu/ipu-v3/ipu-prg.c b/drivers/gpu/ipu-v3/ipu-prg.c index caca57febbd6..ecc9ea44dc50 100644 --- a/drivers/gpu/ipu-v3/ipu-prg.c +++ b/drivers/gpu/ipu-v3/ipu-prg.c @@ -318,8 +318,6 @@ int ipu_prg_channel_configure(struct ipuv3_channel *ipu_chan, writel(val, prg->regs + IPU_PRG_BADDR(prg_chan)); val = readl(prg->regs + IPU_PRG_CTL); - /* counter load enable */ - val |= IPU_PRG_CTL_CNT_LOAD_EN(prg_chan); /* config AXI ID */ val &= ~(IPU_PRG_CTL_SOFT_ARID_MASK << IPU_PRG_CTL_SOFT_ARID_SHIFT(prg_chan)); -- 2.11.0

2 1

[PATCH] drm/mm: Split up long running selftests with cond_resched()
by Chris Wilson 04 May '17

04 May '17

Scatter a few cond_resched() in between phases of the drm_mm selftests to try and prevent us incurring the wrath of the NMI watchdog. Signed-off-by: Chris Wilson <chris(a)chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen(a)linux.intel.com> --- drivers/gpu/drm/selftests/test-drm_mm.c | 28 ++++++++++++++++++++++++++++ 1 file changed, 28 insertions(+) diff --git a/drivers/gpu/drm/selftests/test-drm_mm.c b/drivers/gpu/drm/selftests/test-drm_mm.c index fa356f5dae27..dfdd858eda0a 100644 --- a/drivers/gpu/drm/selftests/test-drm_mm.c +++ b/drivers/gpu/drm/selftests/test-drm_mm.c @@ -514,6 +514,8 @@ static int igt_reserve(void *ignored) ret = __igt_reserve(count, size + 1); if (ret) return ret; + + cond_resched(); } return 0; @@ -712,6 +714,10 @@ static int igt_insert(void *ignored) return ret; ret = __igt_insert(count, size + 1, false); + if (ret) + return ret; + + cond_resched(); } return 0; @@ -741,6 +747,10 @@ static int igt_replace(void *ignored) return ret; ret = __igt_insert(count, size + 1, true); + if (ret) + return ret; + + cond_resched(); } return 0; @@ -1011,6 +1021,8 @@ static int igt_insert_range(void *ignored) ret = __igt_insert_range(count, size, max/4+1, 3*max/4-1); if (ret) return ret; + + cond_resched(); } return 0; @@ -1056,6 +1068,7 @@ static int igt_align(void *ignored) drm_mm_for_each_node_safe(node, next, &mm) drm_mm_remove_node(node); DRM_MM_BUG_ON(!drm_mm_clean(&mm)); + cond_resched(); } ret = 0; @@ -1097,6 +1110,8 @@ static int igt_align_pot(int max) align, bit); goto out; } + + cond_resched(); } ret = 0; @@ -1471,6 +1486,8 @@ static int igt_evict(void *ignored) goto out; } } + + cond_resched(); } ret = 0; @@ -1566,6 +1583,8 @@ static int igt_evict_range(void *ignored) goto out; } } + + cond_resched(); } ret = 0; @@ -1683,6 +1702,7 @@ static int igt_topdown(void *ignored) drm_mm_for_each_node_safe(node, next, &mm) drm_mm_remove_node(node); DRM_MM_BUG_ON(!drm_mm_clean(&mm)); + cond_resched(); } ret = 0; @@ -1783,6 +1803,7 @@ static int igt_bottomup(void *ignored) drm_mm_for_each_node_safe(node, next, &mm) drm_mm_remove_node(node); DRM_MM_BUG_ON(!drm_mm_clean(&mm)); + cond_resched(); } ret = 0; @@ -1970,6 +1991,8 @@ static int igt_color(void *ignored) drm_mm_remove_node(node); kfree(node); } + + cond_resched(); } ret = 0; @@ -2047,6 +2070,7 @@ static int evict_color(struct drm_mm *mm, } } + cond_resched(); return 0; } @@ -2132,6 +2156,8 @@ static int igt_color_evict(void *ignored) goto out; } } + + cond_resched(); } ret = 0; @@ -2231,6 +2257,8 @@ static int igt_color_evict_range(void *ignored) goto out; } } + + cond_resched(); } ret = 0; -- 2.11.0

3 3

linux-next: build failure after merge of the drm-misc tree
by Stephen Rothwell 03 May '17

03 May '17

Hi all, After merging the drm-misc tree, today's linux-next build (x86_64 allmodconfig) failed like this: drivers/tee/tee_shm.c:87:2: error: unknown field 'kmap_atomic' specified in initializer .kmap_atomic = tee_shm_op_kmap_atomic, ^ drivers/tee/tee_shm.c:87:17: error: initialization from incompatible pointer type [-Werror=incompatible-pointer-types] .kmap_atomic = tee_shm_op_kmap_atomic, ^ drivers/tee/tee_shm.c:87:17: note: (near initialization for 'tee_shm_dma_buf_ops.begin_cpu_access') drivers/tee/tee_shm.c:88:2: error: unknown field 'kmap' specified in initializer .kmap = tee_shm_op_kmap, ^ drivers/tee/tee_shm.c:88:10: error: initialization from incompatible pointer type [-Werror=incompatible-pointer-types] .kmap = tee_shm_op_kmap, ^ drivers/tee/tee_shm.c:88:10: note: (near initialization for 'tee_shm_dma_buf_ops.end_cpu_access') Caused by commit f9b67f0014cb ("dma-buf: Rename dma-ops to prevent conflict with kunmap_atomic macro") interacting with commit 967c9cca2cc5 ("tee: generic TEE subsystem") from the arm-soc tree. I applied the following merge fix patch for today: From: Stephen Rothwell <sfr(a)canb.auug.org.au> Date: Fri, 21 Apr 2017 12:06:32 +1000 Subject: [PATCH] tee: merge fix for dma-ops field name changes Signed-off-by: Stephen Rothwell <sfr(a)canb.auug.org.au> --- drivers/tee/tee_shm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/tee/tee_shm.c b/drivers/tee/tee_shm.c index 0be1e3e93bee..4e14c9c9cb1c 100644 --- a/drivers/tee/tee_shm.c +++ b/drivers/tee/tee_shm.c @@ -84,8 +84,8 @@ static struct dma_buf_ops tee_shm_dma_buf_ops = { .map_dma_buf = tee_shm_op_map_dma_buf, .unmap_dma_buf = tee_shm_op_unmap_dma_buf, .release = tee_shm_op_release, - .kmap_atomic = tee_shm_op_kmap_atomic, - .kmap = tee_shm_op_kmap, + .map_atomic = tee_shm_op_kmap_atomic, + .map = tee_shm_op_kmap, .mmap = tee_shm_op_mmap, }; -- 2.11.0 -- Cheers, Stephen Rothwell

5 7

[PATCH v2 0/2] Introduce crtc->mode_valid() callback
by Jose Abreu 03 May '17

03 May '17

This patchset introduces a new callback for crtc, called mode_valid() that is responsible to limit the number of probbed modes. Just like connector->mode_valid(), this new callback is called at mode probbing stage so that we can validate the mode. This is specially useful because arcpgu crtc is responsible to set a clock value in the commit() stage but unfortunatelly this clock does not support all the needed ranges. This way we can restrict the number of modes that are handed to userspace, so that we dont hand a mode that will fail the commit() stage. I guess this can also happen for other drivers so we introduce the callback in the core. The behaviour remains the same for crtcs that don't have the callback. Also, for a given set of crtcs that can be bound to the connector, if at least one can display the mode then the mode will be probbed. For more info about why this is needed in arcpgu, please refer here [1]. [1] https://patchwork.kernel.org/patch/9694177/ Jose Abreu (2): drm: Introduce crtc->mode_valid() callback drm: arcpgu: Use crtc->mode_valid() callback Cc: Carlos Palminha <palminha(a)synopsys.com> Cc: Alexey Brodkin <abrodkin(a)synopsys.com> Cc: Ville Syrjälä <ville.syrjala(a)linux.intel.com> Cc: Daniel Vetter <daniel.vetter(a)ffwll.ch> Cc: Dave Airlie <airlied(a)linux.ie> Cc: Andrzej Hajda <a.hajda(a)samsung.com> drivers/gpu/drm/arc/arcpgu_crtc.c | 28 ++++++++++++++++-- drivers/gpu/drm/drm_probe_helper.c | 50 ++++++++++++++++++++++++++++++-- include/drm/drm_modeset_helper_vtables.h | 26 +++++++++++++++++ 3 files changed, 99 insertions(+), 5 deletions(-) -- 1.9.1

2 3

[PATCH v2 0/4] Add ZTE VGA driver support
by Shawn Guo 03 May '17

03 May '17

From: Shawn Guo <shawn.guo(a)linaro.org> The series adds the driver for ZTE VGA device, which becomes the third VOU output device we support, besides the existing HDMI and TV Encoder. Changes for v2: - Add return check for encoder and connector init function calls. - Improve comments in function zx_vga_irq_handler() to avoid confusion. - Add lock for VGA_AUTO_DETECT_SEL register access to avoid race condition between irq handler and .get_modes hook. - Copy device tree maintainer and list for bindings doc. Shawn Guo (4): drm: zte: do not enable clock auto-gating by default drm: zte: move CSC register definitions into a common header dt-bindings: display: add support for ZTE VGA device drm: zte: add VGA driver support .../devicetree/bindings/display/zte,vou.txt | 21 + drivers/gpu/drm/zte/Makefile | 1 + drivers/gpu/drm/zte/zx_common_regs.h | 31 ++ drivers/gpu/drm/zte/zx_drm_drv.c | 1 + drivers/gpu/drm/zte/zx_drm_drv.h | 1 + drivers/gpu/drm/zte/zx_plane.c | 1 + drivers/gpu/drm/zte/zx_plane_regs.h | 18 - drivers/gpu/drm/zte/zx_vga.c | 530 +++++++++++++++++++++ drivers/gpu/drm/zte/zx_vga_regs.h | 36 ++ drivers/gpu/drm/zte/zx_vou.c | 36 +- drivers/gpu/drm/zte/zx_vou_regs.h | 12 +- 11 files changed, 663 insertions(+), 25 deletions(-) create mode 100644 drivers/gpu/drm/zte/zx_common_regs.h create mode 100644 drivers/gpu/drm/zte/zx_vga.c create mode 100644 drivers/gpu/drm/zte/zx_vga_regs.h -- 1.9.1

3 12

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

dri-devel May 2017