* [PATCH] media: vimc: Add support for contiguous DMA buffers @ 2021-07-30 0:19 Laurent Pinchart 2021-07-30 12:35 ` Dafna Hirschfeld 0 siblings, 1 reply; 10+ messages in thread From: Laurent Pinchart @ 2021-07-30 0:19 UTC (permalink / raw) To: linux-media; +Cc: linux-renesas-soc, Helen Koike, Shuah Khan The vimc driver is used for testing purpose, and some test use cases involve sharing buffers with a consumer device. Consumers often require DMA contiguous memory, which vimc doesn't currently support. This leads in the best case to usage of bounce buffers, which is very slow, and in the worst case in a complete failure. Add support for the dma-contig allocator in vimc to support those use cases properly. The allocator is selected through a new "allocator" module parameter, which defaults to vmalloc. Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> --- drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ 3 files changed, 19 insertions(+), 2 deletions(-) diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c index 5e9fd902cd37..92b69a6529fb 100644 --- a/drivers/media/test-drivers/vimc/vimc-capture.c +++ b/drivers/media/test-drivers/vimc/vimc-capture.c @@ -7,6 +7,7 @@ #include <media/v4l2-ioctl.h> #include <media/videobuf2-core.h> +#include <media/videobuf2-dma-contig.h> #include <media/videobuf2-vmalloc.h> #include "vimc-common.h" @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, /* Initialize the vb2 queue */ q = &vcap->queue; q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; + q->io_modes = VB2_MMAP | VB2_DMABUF; + if (vimc_allocator != 1) + q->io_modes |= VB2_USERPTR; q->drv_priv = vcap; q->buf_struct_size = sizeof(struct vimc_cap_buffer); q->ops = &vimc_cap_qops; - q->mem_ops = &vb2_vmalloc_memops; + q->mem_ops = vimc_allocator == 1 + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; q->min_buffers_needed = 2; q->lock = &vcap->lock; + q->dev = v4l2_dev->dev; ret = vb2_queue_init(q); if (ret) { diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h index a289434e75ba..b77939123501 100644 --- a/drivers/media/test-drivers/vimc/vimc-common.h +++ b/drivers/media/test-drivers/vimc/vimc-common.h @@ -35,6 +35,8 @@ #define VIMC_PIX_FMT_MAX_CODES 8 +extern unsigned int vimc_allocator; + /** * vimc_colorimetry_clamp - Adjust colorimetry parameters * diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c index 4b0ae6f51d76..7badcecb7aed 100644 --- a/drivers/media/test-drivers/vimc/vimc-core.c +++ b/drivers/media/test-drivers/vimc/vimc-core.c @@ -5,6 +5,7 @@ * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> */ +#include <linux/dma-mapping.h> #include <linux/font.h> #include <linux/init.h> #include <linux/module.h> @@ -15,6 +16,12 @@ #include "vimc-common.h" +unsigned int vimc_allocator; +module_param_named(allocator, vimc_allocator, uint, 0444); +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" + "\t\t 0 == vmalloc\n" + "\t\t 1 == dma-contig"); + #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) tpg_set_font(font->data); + if (vimc_allocator == 1) + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); + vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); if (!vimc) return -ENOMEM; -- Regards, Laurent Pinchart ^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 0:19 [PATCH] media: vimc: Add support for contiguous DMA buffers Laurent Pinchart @ 2021-07-30 12:35 ` Dafna Hirschfeld 2021-07-30 13:11 ` Laurent Pinchart 2021-07-30 16:13 ` Ezequiel Garcia 0 siblings, 2 replies; 10+ messages in thread From: Dafna Hirschfeld @ 2021-07-30 12:35 UTC (permalink / raw) To: Laurent Pinchart, linux-media; +Cc: linux-renesas-soc, Helen Koike, Shuah Khan Hi On 30.07.21 02:19, Laurent Pinchart wrote: > The vimc driver is used for testing purpose, and some test use cases > involve sharing buffers with a consumer device. Consumers often require > DMA contiguous memory, which vimc doesn't currently support. This leads > in the best case to usage of bounce buffers, which is very slow, and in > the worst case in a complete failure. > > Add support for the dma-contig allocator in vimc to support those use > cases properly. The allocator is selected through a new "allocator" > module parameter, which defaults to vmalloc. > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > --- > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > 3 files changed, 19 insertions(+), 2 deletions(-) > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > index 5e9fd902cd37..92b69a6529fb 100644 > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > @@ -7,6 +7,7 @@ > > #include <media/v4l2-ioctl.h> > #include <media/videobuf2-core.h> > +#include <media/videobuf2-dma-contig.h> > #include <media/videobuf2-vmalloc.h> > > #include "vimc-common.h" > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > /* Initialize the vb2 queue */ > q = &vcap->queue; > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > + q->io_modes = VB2_MMAP | VB2_DMABUF; maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > + if (vimc_allocator != 1) maybe define a macro instead of `1` ? > + q->io_modes |= VB2_USERPTR; > q->drv_priv = vcap; > q->buf_struct_size = sizeof(struct vimc_cap_buffer); > q->ops = &vimc_cap_qops; > - q->mem_ops = &vb2_vmalloc_memops; > + q->mem_ops = vimc_allocator == 1 > + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; > q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; > q->min_buffers_needed = 2; > q->lock = &vcap->lock; > + q->dev = v4l2_dev->dev; > > ret = vb2_queue_init(q); > if (ret) { > diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h > index a289434e75ba..b77939123501 100644 > --- a/drivers/media/test-drivers/vimc/vimc-common.h > +++ b/drivers/media/test-drivers/vimc/vimc-common.h > @@ -35,6 +35,8 @@ > > #define VIMC_PIX_FMT_MAX_CODES 8 > > +extern unsigned int vimc_allocator; > + > /** > * vimc_colorimetry_clamp - Adjust colorimetry parameters > * > diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c > index 4b0ae6f51d76..7badcecb7aed 100644 > --- a/drivers/media/test-drivers/vimc/vimc-core.c > +++ b/drivers/media/test-drivers/vimc/vimc-core.c > @@ -5,6 +5,7 @@ > * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> > */ > > +#include <linux/dma-mapping.h> > #include <linux/font.h> > #include <linux/init.h> > #include <linux/module.h> > @@ -15,6 +16,12 @@ > > #include "vimc-common.h" > > +unsigned int vimc_allocator; > +module_param_named(allocator, vimc_allocator, uint, 0444); > +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" > + "\t\t 0 == vmalloc\n" > + "\t\t 1 == dma-contig"); > + There is a section 'Module options' in vimc.rst. So a doc should be added there. Thanks, Dafna > #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" > > #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ > @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) > > tpg_set_font(font->data); > > + if (vimc_allocator == 1) > + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); > + > vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); > if (!vimc) > return -ENOMEM; > ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 12:35 ` Dafna Hirschfeld @ 2021-07-30 13:11 ` Laurent Pinchart 2021-07-30 14:08 ` Dafna Hirschfeld 2021-07-30 15:56 ` Nicolas Dufresne 2021-07-30 16:13 ` Ezequiel Garcia 1 sibling, 2 replies; 10+ messages in thread From: Laurent Pinchart @ 2021-07-30 13:11 UTC (permalink / raw) To: Dafna Hirschfeld; +Cc: linux-media, linux-renesas-soc, Helen Koike, Shuah Khan Hi Dafna, On Fri, Jul 30, 2021 at 02:35:20PM +0200, Dafna Hirschfeld wrote: > On 30.07.21 02:19, Laurent Pinchart wrote: > > The vimc driver is used for testing purpose, and some test use cases > > involve sharing buffers with a consumer device. Consumers often require > > DMA contiguous memory, which vimc doesn't currently support. This leads > > in the best case to usage of bounce buffers, which is very slow, and in > > the worst case in a complete failure. > > > > Add support for the dma-contig allocator in vimc to support those use > > cases properly. The allocator is selected through a new "allocator" > > module parameter, which defaults to vmalloc. > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > > --- > > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > > 3 files changed, 19 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > > index 5e9fd902cd37..92b69a6529fb 100644 > > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > > @@ -7,6 +7,7 @@ > > > > #include <media/v4l2-ioctl.h> > > #include <media/videobuf2-core.h> > > +#include <media/videobuf2-dma-contig.h> > > #include <media/videobuf2-vmalloc.h> > > > > #include "vimc-common.h" > > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > > /* Initialize the vb2 queue */ > > q = &vcap->queue; > > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > > + q->io_modes = VB2_MMAP | VB2_DMABUF; > > maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? Why so ? vb2-vmalloc can import dma-bufs. > > + if (vimc_allocator != 1) > > maybe define a macro instead of `1` ? Good idea. > > + q->io_modes |= VB2_USERPTR; > > q->drv_priv = vcap; > > q->buf_struct_size = sizeof(struct vimc_cap_buffer); > > q->ops = &vimc_cap_qops; > > - q->mem_ops = &vb2_vmalloc_memops; > > + q->mem_ops = vimc_allocator == 1 > > + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; > > q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; > > q->min_buffers_needed = 2; > > q->lock = &vcap->lock; > > + q->dev = v4l2_dev->dev; > > > > ret = vb2_queue_init(q); > > if (ret) { > > diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h > > index a289434e75ba..b77939123501 100644 > > --- a/drivers/media/test-drivers/vimc/vimc-common.h > > +++ b/drivers/media/test-drivers/vimc/vimc-common.h > > @@ -35,6 +35,8 @@ > > > > #define VIMC_PIX_FMT_MAX_CODES 8 > > > > +extern unsigned int vimc_allocator; > > + > > /** > > * vimc_colorimetry_clamp - Adjust colorimetry parameters > > * > > diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c > > index 4b0ae6f51d76..7badcecb7aed 100644 > > --- a/drivers/media/test-drivers/vimc/vimc-core.c > > +++ b/drivers/media/test-drivers/vimc/vimc-core.c > > @@ -5,6 +5,7 @@ > > * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> > > */ > > > > +#include <linux/dma-mapping.h> > > #include <linux/font.h> > > #include <linux/init.h> > > #include <linux/module.h> > > @@ -15,6 +16,12 @@ > > > > #include "vimc-common.h" > > > > +unsigned int vimc_allocator; > > +module_param_named(allocator, vimc_allocator, uint, 0444); > > +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" > > + "\t\t 0 == vmalloc\n" > > + "\t\t 1 == dma-contig"); > > + > > There is a section 'Module options' in vimc.rst. So a doc should be added there. OK, I'll update that. > > #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" > > > > #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ > > @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) > > > > tpg_set_font(font->data); > > > > + if (vimc_allocator == 1) > > + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); > > + > > vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); > > if (!vimc) > > return -ENOMEM; > > -- Regards, Laurent Pinchart ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 13:11 ` Laurent Pinchart @ 2021-07-30 14:08 ` Dafna Hirschfeld 2021-07-30 14:15 ` Laurent Pinchart 2021-07-30 15:56 ` Nicolas Dufresne 1 sibling, 1 reply; 10+ messages in thread From: Dafna Hirschfeld @ 2021-07-30 14:08 UTC (permalink / raw) To: Laurent Pinchart; +Cc: linux-media, linux-renesas-soc, Helen Koike, Shuah Khan On 30.07.21 15:11, Laurent Pinchart wrote: > Hi Dafna, > > On Fri, Jul 30, 2021 at 02:35:20PM +0200, Dafna Hirschfeld wrote: >> On 30.07.21 02:19, Laurent Pinchart wrote: >>> The vimc driver is used for testing purpose, and some test use cases >>> involve sharing buffers with a consumer device. Consumers often require >>> DMA contiguous memory, which vimc doesn't currently support. This leads >>> in the best case to usage of bounce buffers, which is very slow, and in >>> the worst case in a complete failure. >>> >>> Add support for the dma-contig allocator in vimc to support those use >>> cases properly. The allocator is selected through a new "allocator" >>> module parameter, which defaults to vmalloc. >>> >>> Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> >>> --- >>> drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- >>> drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ >>> drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ >>> 3 files changed, 19 insertions(+), 2 deletions(-) >>> >>> diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c >>> index 5e9fd902cd37..92b69a6529fb 100644 >>> --- a/drivers/media/test-drivers/vimc/vimc-capture.c >>> +++ b/drivers/media/test-drivers/vimc/vimc-capture.c >>> @@ -7,6 +7,7 @@ >>> >>> #include <media/v4l2-ioctl.h> >>> #include <media/videobuf2-core.h> >>> +#include <media/videobuf2-dma-contig.h> >>> #include <media/videobuf2-vmalloc.h> >>> >>> #include "vimc-common.h" >>> @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, >>> /* Initialize the vb2 queue */ >>> q = &vcap->queue; >>> q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; >>> - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; >>> + q->io_modes = VB2_MMAP | VB2_DMABUF; >> >> maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > Why so ? vb2-vmalloc can import dma-bufs. oh, I meant that exporting should not be supported. I see that vimc set ".vidioc_expbuf = vb2_ioctl_expbuf", maybe remove that if allocator is vmalloc? Thanks, Dafna > >>> + if (vimc_allocator != 1) >> >> maybe define a macro instead of `1` ? > > Good idea. > >>> + q->io_modes |= VB2_USERPTR; >>> q->drv_priv = vcap; >>> q->buf_struct_size = sizeof(struct vimc_cap_buffer); >>> q->ops = &vimc_cap_qops; >>> - q->mem_ops = &vb2_vmalloc_memops; >>> + q->mem_ops = vimc_allocator == 1 >>> + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; >>> q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; >>> q->min_buffers_needed = 2; >>> q->lock = &vcap->lock; >>> + q->dev = v4l2_dev->dev; >>> >>> ret = vb2_queue_init(q); >>> if (ret) { >>> diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h >>> index a289434e75ba..b77939123501 100644 >>> --- a/drivers/media/test-drivers/vimc/vimc-common.h >>> +++ b/drivers/media/test-drivers/vimc/vimc-common.h >>> @@ -35,6 +35,8 @@ >>> >>> #define VIMC_PIX_FMT_MAX_CODES 8 >>> >>> +extern unsigned int vimc_allocator; >>> + >>> /** >>> * vimc_colorimetry_clamp - Adjust colorimetry parameters >>> * >>> diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c >>> index 4b0ae6f51d76..7badcecb7aed 100644 >>> --- a/drivers/media/test-drivers/vimc/vimc-core.c >>> +++ b/drivers/media/test-drivers/vimc/vimc-core.c >>> @@ -5,6 +5,7 @@ >>> * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> >>> */ >>> >>> +#include <linux/dma-mapping.h> >>> #include <linux/font.h> >>> #include <linux/init.h> >>> #include <linux/module.h> >>> @@ -15,6 +16,12 @@ >>> >>> #include "vimc-common.h" >>> >>> +unsigned int vimc_allocator; >>> +module_param_named(allocator, vimc_allocator, uint, 0444); >>> +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" >>> + "\t\t 0 == vmalloc\n" >>> + "\t\t 1 == dma-contig"); >>> + >> >> There is a section 'Module options' in vimc.rst. So a doc should be added there. > > OK, I'll update that. > >>> #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" >>> >>> #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ >>> @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) >>> >>> tpg_set_font(font->data); >>> >>> + if (vimc_allocator == 1) >>> + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); >>> + >>> vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); >>> if (!vimc) >>> return -ENOMEM; >>> > ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 14:08 ` Dafna Hirschfeld @ 2021-07-30 14:15 ` Laurent Pinchart 2021-07-30 15:59 ` Nicolas Dufresne 0 siblings, 1 reply; 10+ messages in thread From: Laurent Pinchart @ 2021-07-30 14:15 UTC (permalink / raw) To: Dafna Hirschfeld; +Cc: linux-media, linux-renesas-soc, Helen Koike, Shuah Khan Hi Dafna, On Fri, Jul 30, 2021 at 04:08:11PM +0200, Dafna Hirschfeld wrote: > On 30.07.21 15:11, Laurent Pinchart wrote: > > On Fri, Jul 30, 2021 at 02:35:20PM +0200, Dafna Hirschfeld wrote: > >> On 30.07.21 02:19, Laurent Pinchart wrote: > >>> The vimc driver is used for testing purpose, and some test use cases > >>> involve sharing buffers with a consumer device. Consumers often require > >>> DMA contiguous memory, which vimc doesn't currently support. This leads > >>> in the best case to usage of bounce buffers, which is very slow, and in > >>> the worst case in a complete failure. > >>> > >>> Add support for the dma-contig allocator in vimc to support those use > >>> cases properly. The allocator is selected through a new "allocator" > >>> module parameter, which defaults to vmalloc. > >>> > >>> Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > >>> --- > >>> drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > >>> drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > >>> drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > >>> 3 files changed, 19 insertions(+), 2 deletions(-) > >>> > >>> diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > >>> index 5e9fd902cd37..92b69a6529fb 100644 > >>> --- a/drivers/media/test-drivers/vimc/vimc-capture.c > >>> +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > >>> @@ -7,6 +7,7 @@ > >>> > >>> #include <media/v4l2-ioctl.h> > >>> #include <media/videobuf2-core.h> > >>> +#include <media/videobuf2-dma-contig.h> > >>> #include <media/videobuf2-vmalloc.h> > >>> > >>> #include "vimc-common.h" > >>> @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > >>> /* Initialize the vb2 queue */ > >>> q = &vcap->queue; > >>> q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > >>> - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > >>> + q->io_modes = VB2_MMAP | VB2_DMABUF; > >> > >> maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > > > Why so ? vb2-vmalloc can import dma-bufs. > > oh, I meant that exporting should not be supported. > I see that vimc set ".vidioc_expbuf = vb2_ioctl_expbuf", maybe remove that if allocator is vmalloc? If the importer support non-contiguous buffers, vb2-vmalloc can be used as an exporter. I've successfully used this to test sharing buffers between vimc in vmalloc mode and the R-Car H3 display driver with an IOMMU. > >>> + if (vimc_allocator != 1) > >> > >> maybe define a macro instead of `1` ? > > > > Good idea. > > > >>> + q->io_modes |= VB2_USERPTR; > >>> q->drv_priv = vcap; > >>> q->buf_struct_size = sizeof(struct vimc_cap_buffer); > >>> q->ops = &vimc_cap_qops; > >>> - q->mem_ops = &vb2_vmalloc_memops; > >>> + q->mem_ops = vimc_allocator == 1 > >>> + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; > >>> q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; > >>> q->min_buffers_needed = 2; > >>> q->lock = &vcap->lock; > >>> + q->dev = v4l2_dev->dev; > >>> > >>> ret = vb2_queue_init(q); > >>> if (ret) { > >>> diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h > >>> index a289434e75ba..b77939123501 100644 > >>> --- a/drivers/media/test-drivers/vimc/vimc-common.h > >>> +++ b/drivers/media/test-drivers/vimc/vimc-common.h > >>> @@ -35,6 +35,8 @@ > >>> > >>> #define VIMC_PIX_FMT_MAX_CODES 8 > >>> > >>> +extern unsigned int vimc_allocator; > >>> + > >>> /** > >>> * vimc_colorimetry_clamp - Adjust colorimetry parameters > >>> * > >>> diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c > >>> index 4b0ae6f51d76..7badcecb7aed 100644 > >>> --- a/drivers/media/test-drivers/vimc/vimc-core.c > >>> +++ b/drivers/media/test-drivers/vimc/vimc-core.c > >>> @@ -5,6 +5,7 @@ > >>> * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> > >>> */ > >>> > >>> +#include <linux/dma-mapping.h> > >>> #include <linux/font.h> > >>> #include <linux/init.h> > >>> #include <linux/module.h> > >>> @@ -15,6 +16,12 @@ > >>> > >>> #include "vimc-common.h" > >>> > >>> +unsigned int vimc_allocator; > >>> +module_param_named(allocator, vimc_allocator, uint, 0444); > >>> +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" > >>> + "\t\t 0 == vmalloc\n" > >>> + "\t\t 1 == dma-contig"); > >>> + > >> > >> There is a section 'Module options' in vimc.rst. So a doc should be added there. > > > > OK, I'll update that. > > > >>> #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" > >>> > >>> #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ > >>> @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) > >>> > >>> tpg_set_font(font->data); > >>> > >>> + if (vimc_allocator == 1) > >>> + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); > >>> + > >>> vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); > >>> if (!vimc) > >>> return -ENOMEM; > >>> -- Regards, Laurent Pinchart ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 14:15 ` Laurent Pinchart @ 2021-07-30 15:59 ` Nicolas Dufresne 2021-07-30 16:56 ` Laurent Pinchart 0 siblings, 1 reply; 10+ messages in thread From: Nicolas Dufresne @ 2021-07-30 15:59 UTC (permalink / raw) To: Laurent Pinchart, Dafna Hirschfeld Cc: linux-media, linux-renesas-soc, Helen Koike, Shuah Khan Le vendredi 30 juillet 2021 à 17:15 +0300, Laurent Pinchart a écrit : > Hi Dafna, > > On Fri, Jul 30, 2021 at 04:08:11PM +0200, Dafna Hirschfeld wrote: > > On 30.07.21 15:11, Laurent Pinchart wrote: > > > On Fri, Jul 30, 2021 at 02:35:20PM +0200, Dafna Hirschfeld wrote: > > > > On 30.07.21 02:19, Laurent Pinchart wrote: > > > > > The vimc driver is used for testing purpose, and some test use cases > > > > > involve sharing buffers with a consumer device. Consumers often require > > > > > DMA contiguous memory, which vimc doesn't currently support. This leads > > > > > in the best case to usage of bounce buffers, which is very slow, and in > > > > > the worst case in a complete failure. > > > > > > > > > > Add support for the dma-contig allocator in vimc to support those use > > > > > cases properly. The allocator is selected through a new "allocator" > > > > > module parameter, which defaults to vmalloc. > > > > > > > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > > > > > --- > > > > > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > > > > > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > > > > > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > > > > > 3 files changed, 19 insertions(+), 2 deletions(-) > > > > > > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > > > > > index 5e9fd902cd37..92b69a6529fb 100644 > > > > > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > > > > > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > > > > > @@ -7,6 +7,7 @@ > > > > > > > > > > #include <media/v4l2-ioctl.h> > > > > > #include <media/videobuf2-core.h> > > > > > +#include <media/videobuf2-dma-contig.h> > > > > > #include <media/videobuf2-vmalloc.h> > > > > > > > > > > #include "vimc-common.h" > > > > > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > > > > > /* Initialize the vb2 queue */ > > > > > q = &vcap->queue; > > > > > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > > > > > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > > > > > + q->io_modes = VB2_MMAP | VB2_DMABUF; > > > > > > > > maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > > > > > Why so ? vb2-vmalloc can import dma-bufs. > > > > oh, I meant that exporting should not be supported. > > I see that vimc set ".vidioc_expbuf = vb2_ioctl_expbuf", maybe remove that if allocator is vmalloc? > > If the importer support non-contiguous buffers, vb2-vmalloc can be used > as an exporter. I've successfully used this to test sharing buffers > between vimc in vmalloc mode and the R-Car H3 display driver with an > IOMMU. Having an IOMMU is not sufficient, as this is shown with Intel DRM. The DRM driver needs to support CPU cache. Note that in GStreamer this is used a lot from UVC camera to GL (but it breaks, corrupted images, with kmssink). > > > > > > + if (vimc_allocator != 1) > > > > > > > > maybe define a macro instead of `1` ? > > > > > > Good idea. > > > > > > > > + q->io_modes |= VB2_USERPTR; > > > > > q->drv_priv = vcap; > > > > > q->buf_struct_size = sizeof(struct vimc_cap_buffer); > > > > > q->ops = &vimc_cap_qops; > > > > > - q->mem_ops = &vb2_vmalloc_memops; > > > > > + q->mem_ops = vimc_allocator == 1 > > > > > + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; > > > > > q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; > > > > > q->min_buffers_needed = 2; > > > > > q->lock = &vcap->lock; > > > > > + q->dev = v4l2_dev->dev; > > > > > > > > > > ret = vb2_queue_init(q); > > > > > if (ret) { > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h > > > > > index a289434e75ba..b77939123501 100644 > > > > > --- a/drivers/media/test-drivers/vimc/vimc-common.h > > > > > +++ b/drivers/media/test-drivers/vimc/vimc-common.h > > > > > @@ -35,6 +35,8 @@ > > > > > > > > > > #define VIMC_PIX_FMT_MAX_CODES 8 > > > > > > > > > > +extern unsigned int vimc_allocator; > > > > > + > > > > > /** > > > > > * vimc_colorimetry_clamp - Adjust colorimetry parameters > > > > > * > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c > > > > > index 4b0ae6f51d76..7badcecb7aed 100644 > > > > > --- a/drivers/media/test-drivers/vimc/vimc-core.c > > > > > +++ b/drivers/media/test-drivers/vimc/vimc-core.c > > > > > @@ -5,6 +5,7 @@ > > > > > * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> > > > > > */ > > > > > > > > > > +#include <linux/dma-mapping.h> > > > > > #include <linux/font.h> > > > > > #include <linux/init.h> > > > > > #include <linux/module.h> > > > > > @@ -15,6 +16,12 @@ > > > > > > > > > > #include "vimc-common.h" > > > > > > > > > > +unsigned int vimc_allocator; > > > > > +module_param_named(allocator, vimc_allocator, uint, 0444); > > > > > +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" > > > > > + "\t\t 0 == vmalloc\n" > > > > > + "\t\t 1 == dma-contig"); > > > > > + > > > > > > > > There is a section 'Module options' in vimc.rst. So a doc should be added there. > > > > > > OK, I'll update that. > > > > > > > > #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" > > > > > > > > > > #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ > > > > > @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) > > > > > > > > > > tpg_set_font(font->data); > > > > > > > > > > + if (vimc_allocator == 1) > > > > > + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); > > > > > + > > > > > vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); > > > > > if (!vimc) > > > > > return -ENOMEM; > > > > > > ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 15:59 ` Nicolas Dufresne @ 2021-07-30 16:56 ` Laurent Pinchart 0 siblings, 0 replies; 10+ messages in thread From: Laurent Pinchart @ 2021-07-30 16:56 UTC (permalink / raw) To: Nicolas Dufresne Cc: Dafna Hirschfeld, linux-media, linux-renesas-soc, Helen Koike, Shuah Khan Hi Nicolas, On Fri, Jul 30, 2021 at 11:59:27AM -0400, Nicolas Dufresne wrote: > Le vendredi 30 juillet 2021 à 17:15 +0300, Laurent Pinchart a écrit : > > On Fri, Jul 30, 2021 at 04:08:11PM +0200, Dafna Hirschfeld wrote: > > > On 30.07.21 15:11, Laurent Pinchart wrote: > > > > On Fri, Jul 30, 2021 at 02:35:20PM +0200, Dafna Hirschfeld wrote: > > > > > On 30.07.21 02:19, Laurent Pinchart wrote: > > > > > > The vimc driver is used for testing purpose, and some test use cases > > > > > > involve sharing buffers with a consumer device. Consumers often require > > > > > > DMA contiguous memory, which vimc doesn't currently support. This leads > > > > > > in the best case to usage of bounce buffers, which is very slow, and in > > > > > > the worst case in a complete failure. > > > > > > > > > > > > Add support for the dma-contig allocator in vimc to support those use > > > > > > cases properly. The allocator is selected through a new "allocator" > > > > > > module parameter, which defaults to vmalloc. > > > > > > > > > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > > > > > > --- > > > > > > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > > > > > > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > > > > > > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > > > > > > 3 files changed, 19 insertions(+), 2 deletions(-) > > > > > > > > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > > > > > > index 5e9fd902cd37..92b69a6529fb 100644 > > > > > > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > > > > > > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > > > > > > @@ -7,6 +7,7 @@ > > > > > > > > > > > > #include <media/v4l2-ioctl.h> > > > > > > #include <media/videobuf2-core.h> > > > > > > +#include <media/videobuf2-dma-contig.h> > > > > > > #include <media/videobuf2-vmalloc.h> > > > > > > > > > > > > #include "vimc-common.h" > > > > > > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > > > > > > /* Initialize the vb2 queue */ > > > > > > q = &vcap->queue; > > > > > > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > > > > > > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > > > > > > + q->io_modes = VB2_MMAP | VB2_DMABUF; > > > > > > > > > > maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > > > > > > > Why so ? vb2-vmalloc can import dma-bufs. > > > > > > oh, I meant that exporting should not be supported. > > > I see that vimc set ".vidioc_expbuf = vb2_ioctl_expbuf", maybe remove that if allocator is vmalloc? > > > > If the importer support non-contiguous buffers, vb2-vmalloc can be used > > as an exporter. I've successfully used this to test sharing buffers > > between vimc in vmalloc mode and the R-Car H3 display driver with an > > IOMMU. > > Having an IOMMU is not sufficient, as this is shown with Intel DRM. The DRM > driver needs to support CPU cache. Note that in GStreamer this is used a lot > from UVC camera to GL (but it breaks, corrupted images, with kmssink). That depends on the platform. When the sink device isn't cache-coherent, it needs to perform a cache sync operation. V4L2 handles this automatically for codecs (technically speaking for cameras as well, but cameras are rarely sinks :-)). In DRM/KMS, there's still quite a bit of work to do. It's not an issue in vimc in any case. > > > > > > + if (vimc_allocator != 1) > > > > > > > > > > maybe define a macro instead of `1` ? > > > > > > > > Good idea. > > > > > > > > > > + q->io_modes |= VB2_USERPTR; > > > > > > q->drv_priv = vcap; > > > > > > q->buf_struct_size = sizeof(struct vimc_cap_buffer); > > > > > > q->ops = &vimc_cap_qops; > > > > > > - q->mem_ops = &vb2_vmalloc_memops; > > > > > > + q->mem_ops = vimc_allocator == 1 > > > > > > + ? &vb2_dma_contig_memops : &vb2_vmalloc_memops; > > > > > > q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC; > > > > > > q->min_buffers_needed = 2; > > > > > > q->lock = &vcap->lock; > > > > > > + q->dev = v4l2_dev->dev; > > > > > > > > > > > > ret = vb2_queue_init(q); > > > > > > if (ret) { > > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h > > > > > > index a289434e75ba..b77939123501 100644 > > > > > > --- a/drivers/media/test-drivers/vimc/vimc-common.h > > > > > > +++ b/drivers/media/test-drivers/vimc/vimc-common.h > > > > > > @@ -35,6 +35,8 @@ > > > > > > > > > > > > #define VIMC_PIX_FMT_MAX_CODES 8 > > > > > > > > > > > > +extern unsigned int vimc_allocator; > > > > > > + > > > > > > /** > > > > > > * vimc_colorimetry_clamp - Adjust colorimetry parameters > > > > > > * > > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c > > > > > > index 4b0ae6f51d76..7badcecb7aed 100644 > > > > > > --- a/drivers/media/test-drivers/vimc/vimc-core.c > > > > > > +++ b/drivers/media/test-drivers/vimc/vimc-core.c > > > > > > @@ -5,6 +5,7 @@ > > > > > > * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> > > > > > > */ > > > > > > > > > > > > +#include <linux/dma-mapping.h> > > > > > > #include <linux/font.h> > > > > > > #include <linux/init.h> > > > > > > #include <linux/module.h> > > > > > > @@ -15,6 +16,12 @@ > > > > > > > > > > > > #include "vimc-common.h" > > > > > > > > > > > > +unsigned int vimc_allocator; > > > > > > +module_param_named(allocator, vimc_allocator, uint, 0444); > > > > > > +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" > > > > > > + "\t\t 0 == vmalloc\n" > > > > > > + "\t\t 1 == dma-contig"); > > > > > > + > > > > > > > > > > There is a section 'Module options' in vimc.rst. So a doc should be added there. > > > > > > > > OK, I'll update that. > > > > > > > > > > #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" > > > > > > > > > > > > #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ > > > > > > @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) > > > > > > > > > > > > tpg_set_font(font->data); > > > > > > > > > > > > + if (vimc_allocator == 1) > > > > > > + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); > > > > > > + > > > > > > vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); > > > > > > if (!vimc) > > > > > > return -ENOMEM; > > > > > > -- Regards, Laurent Pinchart ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 13:11 ` Laurent Pinchart 2021-07-30 14:08 ` Dafna Hirschfeld @ 2021-07-30 15:56 ` Nicolas Dufresne 1 sibling, 0 replies; 10+ messages in thread From: Nicolas Dufresne @ 2021-07-30 15:56 UTC (permalink / raw) To: Laurent Pinchart, Dafna Hirschfeld Cc: linux-media, linux-renesas-soc, Helen Koike, Shuah Khan Le vendredi 30 juillet 2021 à 16:11 +0300, Laurent Pinchart a écrit : > Hi Dafna, > > On Fri, Jul 30, 2021 at 02:35:20PM +0200, Dafna Hirschfeld wrote: > > On 30.07.21 02:19, Laurent Pinchart wrote: > > > The vimc driver is used for testing purpose, and some test use cases > > > involve sharing buffers with a consumer device. Consumers often require > > > DMA contiguous memory, which vimc doesn't currently support. This leads > > > in the best case to usage of bounce buffers, which is very slow, and in > > > the worst case in a complete failure. > > > > > > Add support for the dma-contig allocator in vimc to support those use > > > cases properly. The allocator is selected through a new "allocator" > > > module parameter, which defaults to vmalloc. > > > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > > > --- > > > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > > > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > > > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > > > 3 files changed, 19 insertions(+), 2 deletions(-) > > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > > > index 5e9fd902cd37..92b69a6529fb 100644 > > > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > > > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > > > @@ -7,6 +7,7 @@ > > > > > > #include <media/v4l2-ioctl.h> > > > #include <media/videobuf2-core.h> > > > +#include <media/videobuf2-dma-contig.h> > > > #include <media/videobuf2-vmalloc.h> > > > > > > #include "vimc-common.h" > > > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > > > /* Initialize the vb2 queue */ > > > q = &vcap->queue; > > > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > > > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > > > + q->io_modes = VB2_MMAP | VB2_DMABUF; > > > > maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > Why so ? vb2-vmalloc can import dma-bufs. Indeed, should be safe with both allocator, the CMA one will validate that the pages are contiguous and fail synchronously as expected for CMA. The known issues are mostly for reading importers (consumers), for writing importer (producers) it is likely still a bit buggy on cache management. > > > if (ret) { > > > diff --git a/drivers/media/test-drivers/vimc/vimc-common.h b/drivers/media/test-drivers/vimc/vimc-common.h > > > index a289434e75ba..b77939123501 100644 > > > --- a/drivers/media/test-drivers/vimc/vimc-common.h > > > +++ b/drivers/media/test-drivers/vimc/vimc-common.h > > > @@ -35,6 +35,8 @@ > > > > > > #define VIMC_PIX_FMT_MAX_CODES 8 > > > > > > +extern unsigned int vimc_allocator; > > > + > > > /** > > > * vimc_colorimetry_clamp - Adjust colorimetry parameters > > > * > > > diff --git a/drivers/media/test-drivers/vimc/vimc-core.c b/drivers/media/test-drivers/vimc/vimc-core.c > > > index 4b0ae6f51d76..7badcecb7aed 100644 > > > --- a/drivers/media/test-drivers/vimc/vimc-core.c > > > +++ b/drivers/media/test-drivers/vimc/vimc-core.c > > > @@ -5,6 +5,7 @@ > > > * Copyright (C) 2015-2017 Helen Koike <helen.fornazier@gmail.com> > > > */ > > > > > > +#include <linux/dma-mapping.h> > > > #include <linux/font.h> > > > #include <linux/init.h> > > > #include <linux/module.h> > > > @@ -15,6 +16,12 @@ > > > > > > #include "vimc-common.h" > > > > > > +unsigned int vimc_allocator; > > > +module_param_named(allocator, vimc_allocator, uint, 0444); > > > +MODULE_PARM_DESC(allocator, " memory allocator selection, default is 0.\n" > > > + "\t\t 0 == vmalloc\n" > > > + "\t\t 1 == dma-contig"); > > > + > > > > There is a section 'Module options' in vimc.rst. So a doc should be added there. > > OK, I'll update that. > > > > #define VIMC_MDEV_MODEL_NAME "VIMC MDEV" > > > > > > #define VIMC_ENT_LINK(src, srcpad, sink, sinkpad, link_flags) { \ > > > @@ -278,6 +285,9 @@ static int vimc_probe(struct platform_device *pdev) > > > > > > tpg_set_font(font->data); > > > > > > + if (vimc_allocator == 1) > > > + dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); > > > + > > > vimc = kzalloc(sizeof(*vimc), GFP_KERNEL); > > > if (!vimc) > > > return -ENOMEM; > > > > ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 12:35 ` Dafna Hirschfeld 2021-07-30 13:11 ` Laurent Pinchart @ 2021-07-30 16:13 ` Ezequiel Garcia 2021-07-30 16:50 ` Laurent Pinchart 1 sibling, 1 reply; 10+ messages in thread From: Ezequiel Garcia @ 2021-07-30 16:13 UTC (permalink / raw) To: Dafna Hirschfeld Cc: Laurent Pinchart, linux-media, Linux-Renesas, Helen Koike, Shuah Khan Hi Laurent, Dafna, On Fri, 30 Jul 2021 at 09:35, Dafna Hirschfeld <dafna.hirschfeld@collabora.com> wrote: > > Hi > > On 30.07.21 02:19, Laurent Pinchart wrote: > > The vimc driver is used for testing purpose, and some test use cases > > involve sharing buffers with a consumer device. Consumers often require > > DMA contiguous memory, which vimc doesn't currently support. This leads > > in the best case to usage of bounce buffers, which is very slow, and in > > the worst case in a complete failure. > > > > Add support for the dma-contig allocator in vimc to support those use > > cases properly. The allocator is selected through a new "allocator" > > module parameter, which defaults to vmalloc. > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > > --- > > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > > 3 files changed, 19 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > > index 5e9fd902cd37..92b69a6529fb 100644 > > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > > @@ -7,6 +7,7 @@ > > > > #include <media/v4l2-ioctl.h> > > #include <media/videobuf2-core.h> > > +#include <media/videobuf2-dma-contig.h> > > #include <media/videobuf2-vmalloc.h> > > > > #include "vimc-common.h" > > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > > /* Initialize the vb2 queue */ > > q = &vcap->queue; > > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > > + q->io_modes = VB2_MMAP | VB2_DMABUF; > > maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > > + if (vimc_allocator != 1) > > maybe define a macro instead of `1` ? > This is maybe an overkill, but you can make the parameter accept strings instead of integers, which makes it "modprobe vimc allocator=vmalloc", and improves a bit user-friendlyness. See drivers/media/pci/tw686x/tw686x-core.c. For a test driver, it is worth the trouble, maybe? Thanks, Ezequiel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] media: vimc: Add support for contiguous DMA buffers 2021-07-30 16:13 ` Ezequiel Garcia @ 2021-07-30 16:50 ` Laurent Pinchart 0 siblings, 0 replies; 10+ messages in thread From: Laurent Pinchart @ 2021-07-30 16:50 UTC (permalink / raw) To: Ezequiel Garcia Cc: Dafna Hirschfeld, linux-media, Linux-Renesas, Helen Koike, Shuah Khan Hi Ezequiel, On Fri, Jul 30, 2021 at 01:13:20PM -0300, Ezequiel Garcia wrote: > On Fri, 30 Jul 2021 at 09:35, Dafna Hirschfeld wrote: > > On 30.07.21 02:19, Laurent Pinchart wrote: > > > The vimc driver is used for testing purpose, and some test use cases > > > involve sharing buffers with a consumer device. Consumers often require > > > DMA contiguous memory, which vimc doesn't currently support. This leads > > > in the best case to usage of bounce buffers, which is very slow, and in > > > the worst case in a complete failure. > > > > > > Add support for the dma-contig allocator in vimc to support those use > > > cases properly. The allocator is selected through a new "allocator" > > > module parameter, which defaults to vmalloc. > > > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> > > > --- > > > drivers/media/test-drivers/vimc/vimc-capture.c | 9 +++++++-- > > > drivers/media/test-drivers/vimc/vimc-common.h | 2 ++ > > > drivers/media/test-drivers/vimc/vimc-core.c | 10 ++++++++++ > > > 3 files changed, 19 insertions(+), 2 deletions(-) > > > > > > diff --git a/drivers/media/test-drivers/vimc/vimc-capture.c b/drivers/media/test-drivers/vimc/vimc-capture.c > > > index 5e9fd902cd37..92b69a6529fb 100644 > > > --- a/drivers/media/test-drivers/vimc/vimc-capture.c > > > +++ b/drivers/media/test-drivers/vimc/vimc-capture.c > > > @@ -7,6 +7,7 @@ > > > > > > #include <media/v4l2-ioctl.h> > > > #include <media/videobuf2-core.h> > > > +#include <media/videobuf2-dma-contig.h> > > > #include <media/videobuf2-vmalloc.h> > > > > > > #include "vimc-common.h" > > > @@ -423,14 +424,18 @@ static struct vimc_ent_device *vimc_cap_add(struct vimc_device *vimc, > > > /* Initialize the vb2 queue */ > > > q = &vcap->queue; > > > q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE; > > > - q->io_modes = VB2_MMAP | VB2_DMABUF | VB2_USERPTR; > > > + q->io_modes = VB2_MMAP | VB2_DMABUF; > > > > maybe to be on the safe side VB2_DMABUF should be set only if vimc_allocator==1 ? > > > > > + if (vimc_allocator != 1) > > > > maybe define a macro instead of `1` ? > > This is maybe an overkill, but you can make the parameter accept strings > instead of integers, which makes it "modprobe vimc allocator=vmalloc", > and improves a bit user-friendlyness. > > See drivers/media/pci/tw686x/tw686x-core.c. > > For a test driver, it is worth the trouble, maybe? I copied the values from vivid, which uses an array of integers. As vimc needs a single parameter only, a string could make more sense. I'll submit a v3 with a string if there's a consensus for that. -- Regards, Laurent Pinchart ^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2021-07-30 16:57 UTC | newest] Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2021-07-30 0:19 [PATCH] media: vimc: Add support for contiguous DMA buffers Laurent Pinchart 2021-07-30 12:35 ` Dafna Hirschfeld 2021-07-30 13:11 ` Laurent Pinchart 2021-07-30 14:08 ` Dafna Hirschfeld 2021-07-30 14:15 ` Laurent Pinchart 2021-07-30 15:59 ` Nicolas Dufresne 2021-07-30 16:56 ` Laurent Pinchart 2021-07-30 15:56 ` Nicolas Dufresne 2021-07-30 16:13 ` Ezequiel Garcia 2021-07-30 16:50 ` Laurent Pinchart
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).