iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

drm/tegra: Implement correct DMA-BUF semantics

Browse Source 
DMA-BUF requires that each device that accesses a DMA-BUF attaches to it
separately. To do so the host1x_bo_pin() and host1x_bo_unpin() functions
need to be reimplemented so that they can return a mapping, which either
represents an attachment or a map of the driver's own GEM object.

Signed-off-by: Thierry Reding <treding@nvidia.com>
This commit is contained in:
Thierry Reding 2021-09-09 15:51:24 +02:00
parent 7a56783850
commit c6aeaf56f4
9 changed files with 272 additions and 292 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

167
drivers/gpu/drm/tegra/gem.c
View File

@ -23,6 +23,31 @@
MODULE_IMPORT_NS(DMA_BUF); MODULE_IMPORT_NS(DMA_BUF);
static unsigned int sg_dma_count_chunks(struct scatterlist *sgl, unsigned int nents)
{
dma_addr_t next = ~(dma_addr_t)0;
unsigned int count = 0, i;
struct scatterlist *s;
for_each_sg(sgl, s, nents, i) {
/* sg_dma_address(s) is only valid for entries that have sg_dma_len(s) != 0. */
if (!sg_dma_len(s))
continue;
if (sg_dma_address(s) != next) {
next = sg_dma_address(s) + sg_dma_len(s);
count++;
}
}
return count;
}
static inline unsigned int sgt_dma_count_chunks(struct sg_table *sgt)
{
return sg_dma_count_chunks(sgt->sgl, sgt->nents);
}
static void tegra_bo_put(struct host1x_bo *bo) static void tegra_bo_put(struct host1x_bo *bo)
{ {
struct tegra_bo *obj = host1x_to_tegra_bo(bo); struct tegra_bo *obj = host1x_to_tegra_bo(bo);
@ -30,79 +55,64 @@ static void tegra_bo_put(struct host1x_bo *bo)
drm_gem_object_put(&obj->gem); drm_gem_object_put(&obj->gem);
} }
/* XXX move this into lib/scatterlist.c? */ static struct host1x_bo_mapping *tegra_bo_pin(struct device *dev, struct host1x_bo *bo,
static int sg_alloc_table_from_sg(struct sg_table *sgt, struct scatterlist *sg, enum dma_data_direction direction)
unsigned int nents, gfp_t gfp_mask)
{
struct scatterlist *dst;
unsigned int i;
int err;
err = sg_alloc_table(sgt, nents, gfp_mask);
if (err < 0)
return err;
dst = sgt->sgl;
for (i = 0; i < nents; i++) {
sg_set_page(dst, sg_page(sg), sg->length, 0);
dst = sg_next(dst);
sg = sg_next(sg);
}
return 0;
}
static struct sg_table *tegra_bo_pin(struct device *dev, struct host1x_bo *bo,
dma_addr_t *phys)
{ {
struct tegra_bo *obj = host1x_to_tegra_bo(bo); struct tegra_bo *obj = host1x_to_tegra_bo(bo);
struct sg_table *sgt; struct drm_gem_object *gem = &obj->gem;
struct host1x_bo_mapping *map;
int err; int err;
map = kzalloc(sizeof(*map), GFP_KERNEL);
if (!map)
return ERR_PTR(-ENOMEM);
map->bo = host1x_bo_get(bo);
map->direction = direction;
map->dev = dev;
/* /*
* If we've manually mapped the buffer object through the IOMMU, make * Imported buffers need special treatment to satisfy the semantics of DMA-BUF.
* sure to return the IOVA address of our mapping.
*
* Similarly, for buffers that have been allocated by the DMA API the
* physical address can be used for devices that are not attached to
* an IOMMU. For these devices, callers must pass a valid pointer via
* the @phys argument.
*
* Imported buffers were also already mapped at import time, so the
* existing mapping can be reused.
*/ */
if (phys) { if (gem->import_attach) {
*phys = obj->iova; struct dma_buf *buf = gem->import_attach->dmabuf;
return NULL;
map->attach = dma_buf_attach(buf, dev);
if (IS_ERR(map->attach)) {
err = PTR_ERR(map->attach);
goto free;
}
map->sgt = dma_buf_map_attachment(map->attach, direction);
if (IS_ERR(map->sgt)) {
dma_buf_detach(buf, map->attach);
err = PTR_ERR(map->sgt);
goto free;
}
err = sgt_dma_count_chunks(map->sgt);
map->size = gem->size;
goto out;
} }
/* /*
* If we don't have a mapping for this buffer yet, return an SG table * If we don't have a mapping for this buffer yet, return an SG table
* so that host1x can do the mapping for us via the DMA API. * so that host1x can do the mapping for us via the DMA API.
*/ */
sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); map->sgt = kzalloc(sizeof(*map->sgt), GFP_KERNEL);
if (!sgt) if (!map->sgt) {
return ERR_PTR(-ENOMEM); err = -ENOMEM;
goto free;
}
if (obj->pages) { if (obj->pages) {
/* /*
* If the buffer object was allocated from the explicit IOMMU * If the buffer object was allocated from the explicit IOMMU
* API code paths, construct an SG table from the pages. * API code paths, construct an SG table from the pages.
*/ */
err = sg_alloc_table_from_pages(sgt, obj->pages, obj->num_pages, err = sg_alloc_table_from_pages(map->sgt, obj->pages, obj->num_pages, 0, gem->size,
0, obj->gem.size, GFP_KERNEL); GFP_KERNEL);
if (err < 0)
goto free;
} else if (obj->sgt) {
/*
* If the buffer object already has an SG table but no pages
* were allocated for it, it means the buffer was imported and
* the SG table needs to be copied to avoid overwriting any
* other potential users of the original SG table.
*/
err = sg_alloc_table_from_sg(sgt, obj->sgt->sgl,
obj->sgt->orig_nents, GFP_KERNEL);
if (err < 0) if (err < 0)
goto free; goto free;
} else { } else {
@ -111,25 +121,56 @@ static struct sg_table *tegra_bo_pin(struct device *dev, struct host1x_bo *bo,
* not imported, it had to be allocated with the DMA API, so * not imported, it had to be allocated with the DMA API, so
* the DMA API helper can be used. * the DMA API helper can be used.
*/ */
err = dma_get_sgtable(dev, sgt, obj->vaddr, obj->iova, err = dma_get_sgtable(dev, map->sgt, obj->vaddr, obj->iova, gem->size);
obj->gem.size);
if (err < 0) if (err < 0)
goto free; goto free;
} }
return sgt; err = dma_map_sgtable(dev, map->sgt, direction, 0);
if (err)
goto free_sgt;
out:
/*
* If we've manually mapped the buffer object through the IOMMU, make sure to return the
* existing IOVA address of our mapping.
*/
if (!obj->mm) {
map->phys = sg_dma_address(map->sgt->sgl);
map->chunks = err;
} else {
map->phys = obj->iova;
map->chunks = 1;
}
map->size = gem->size;
return map;
free_sgt:
sg_free_table(map->sgt);
free: free:
kfree(sgt); kfree(map->sgt);
kfree(map);
return ERR_PTR(err); return ERR_PTR(err);
} }
static void tegra_bo_unpin(struct device *dev, struct sg_table *sgt) static void tegra_bo_unpin(struct host1x_bo_mapping *map)
{ {
if (sgt) { if (!map)
sg_free_table(sgt); return;
kfree(sgt);
if (map->attach) {
dma_buf_unmap_attachment(map->attach, map->sgt, map->direction);
dma_buf_detach(map->attach->dmabuf, map->attach);
} else {
dma_unmap_sgtable(map->dev, map->sgt, map->direction, 0);
sg_free_table(map->sgt);
kfree(map->sgt);
} }
host1x_bo_put(map->bo);
kfree(map);
} }
static void *tegra_bo_mmap(struct host1x_bo *bo) static void *tegra_bo_mmap(struct host1x_bo *bo)

60
drivers/gpu/drm/tegra/plane.c
View File

@ -74,7 +74,7 @@ tegra_plane_atomic_duplicate_state(struct drm_plane *plane)
for (i = 0; i < 3; i++) { for (i = 0; i < 3; i++) {
copy->iova[i] = DMA_MAPPING_ERROR; copy->iova[i] = DMA_MAPPING_ERROR;
copy->sgt[i] = NULL; copy->map[i] = NULL;
} }
return &copy->base; return &copy->base;
@ -138,55 +138,37 @@ const struct drm_plane_funcs tegra_plane_funcs = {
static int tegra_dc_pin(struct tegra_dc *dc, struct tegra_plane_state *state) static int tegra_dc_pin(struct tegra_dc *dc, struct tegra_plane_state *state)
{ {
struct iommu_domain *domain = iommu_get_domain_for_dev(dc->dev);
unsigned int i; unsigned int i;
int err; int err;
for (i = 0; i < state->base.fb->format->num_planes; i++) { for (i = 0; i < state->base.fb->format->num_planes; i++) {
struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i); struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i);
dma_addr_t phys_addr, *phys; struct host1x_bo_mapping *map;
struct sg_table *sgt;
/* map = host1x_bo_pin(dc->dev, &bo->base, DMA_TO_DEVICE);
* If we're not attached to a domain, we already stored the if (IS_ERR(map)) {
* physical address when the buffer was allocated. If we're err = PTR_ERR(map);
* part of a group that's shared between all display
* controllers, we've also already mapped the framebuffer
* through the SMMU. In both cases we can short-circuit the
* code below and retrieve the stored IOV address.
*/
if (!domain || dc->client.group)
phys = &phys_addr;
else
phys = NULL;
sgt = host1x_bo_pin(dc->dev, &bo->base, phys);
if (IS_ERR(sgt)) {
err = PTR_ERR(sgt);
goto unpin; goto unpin;
} }
if (sgt) { if (!dc->client.group) {
err = dma_map_sgtable(dc->dev, sgt, DMA_TO_DEVICE, 0);
if (err)
goto unpin;
/* /*
* The display controller needs contiguous memory, so * The display controller needs contiguous memory, so
* fail if the buffer is discontiguous and we fail to * fail if the buffer is discontiguous and we fail to
* map its SG table to a single contiguous chunk of * map its SG table to a single contiguous chunk of
* I/O virtual memory. * I/O virtual memory.
*/ */
if (sgt->nents > 1) { if (map->chunks > 1) {
err = -EINVAL; err = -EINVAL;
goto unpin; goto unpin;
} }
state->iova[i] = sg_dma_address(sgt->sgl); state->iova[i] = map->phys;
state->sgt[i] = sgt;
} else { } else {
state->iova[i] = phys_addr; state->iova[i] = bo->iova;
} }
state->map[i] = map;
} }
return 0; return 0;
@ -195,15 +177,9 @@ unpin:
dev_err(dc->dev, "failed to map plane %u: %d\n", i, err); dev_err(dc->dev, "failed to map plane %u: %d\n", i, err);
while (i--) { while (i--) {
struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i); host1x_bo_unpin(state->map[i]);
struct sg_table *sgt = state->sgt[i];
if (sgt)
dma_unmap_sgtable(dc->dev, sgt, DMA_TO_DEVICE, 0);
host1x_bo_unpin(dc->dev, &bo->base, sgt);
state->iova[i] = DMA_MAPPING_ERROR; state->iova[i] = DMA_MAPPING_ERROR;
state->sgt[i] = NULL; state->map[i] = NULL;
} }
return err; return err;
@ -214,15 +190,9 @@ static void tegra_dc_unpin(struct tegra_dc *dc, struct tegra_plane_state *state)
unsigned int i; unsigned int i;
for (i = 0; i < state->base.fb->format->num_planes; i++) { for (i = 0; i < state->base.fb->format->num_planes; i++) {
struct tegra_bo *bo = tegra_fb_get_plane(state->base.fb, i); host1x_bo_unpin(state->map[i]);
struct sg_table *sgt = state->sgt[i];
if (sgt)
dma_unmap_sgtable(dc->dev, sgt, DMA_TO_DEVICE, 0);
host1x_bo_unpin(dc->dev, &bo->base, sgt);
state->iova[i] = DMA_MAPPING_ERROR; state->iova[i] = DMA_MAPPING_ERROR;
state->sgt[i] = NULL; state->map[i] = NULL;
} }
} }

2
drivers/gpu/drm/tegra/plane.h
View File

@ -43,7 +43,7 @@ struct tegra_plane_legacy_blending_state {
struct tegra_plane_state { struct tegra_plane_state {
struct drm_plane_state base; struct drm_plane_state base;
struct sg_table *sgt[3]; struct host1x_bo_mapping *map[3];
dma_addr_t iova[3]; dma_addr_t iova[3];
struct tegra_bo_tiling tiling; struct tegra_bo_tiling tiling;

62
drivers/gpu/drm/tegra/submit.c
View File

@ -64,33 +64,61 @@ static void gather_bo_put(struct host1x_bo *host_bo)
kref_put(&bo->ref, gather_bo_release); kref_put(&bo->ref, gather_bo_release);
} }
static struct sg_table * static struct host1x_bo_mapping *
gather_bo_pin(struct device *dev, struct host1x_bo *host_bo, dma_addr_t *phys) gather_bo_pin(struct device *dev, struct host1x_bo *bo, enum dma_data_direction direction)
{ {
struct gather_bo *bo = container_of(host_bo, struct gather_bo, base); struct gather_bo *gather = container_of(bo, struct gather_bo, base);
struct sg_table *sgt; struct host1x_bo_mapping *map;
int err; int err;
sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); map = kzalloc(sizeof(*map), GFP_KERNEL);
if (!sgt) if (!map)
return ERR_PTR(-ENOMEM); return ERR_PTR(-ENOMEM);
err = dma_get_sgtable(bo->dev, sgt, bo->gather_data, bo->gather_data_dma, map->bo = host1x_bo_get(bo);
bo->gather_data_words * 4); map->direction = direction;
if (err) { map->dev = dev;
kfree(sgt);
return ERR_PTR(err); map->sgt = kzalloc(sizeof(*map->sgt), GFP_KERNEL);
if (!map->sgt) {
err = -ENOMEM;
goto free;
} }
return sgt; err = dma_get_sgtable(gather->dev, map->sgt, gather->gather_data, gather->gather_data_dma,
gather->gather_data_words * 4);
if (err)
goto free_sgt;
err = dma_map_sgtable(dev, map->sgt, direction, 0);
if (err)
goto free_sgt;
map->phys = sg_dma_address(map->sgt->sgl);
map->size = gather->gather_data_words * 4;
map->chunks = err;
return map;
free_sgt:
sg_free_table(map->sgt);
kfree(map->sgt);
free:
kfree(map);
return ERR_PTR(err);
} }
static void gather_bo_unpin(struct device *dev, struct sg_table *sgt) static void gather_bo_unpin(struct host1x_bo_mapping *map)
{ {
if (sgt) { if (!map)
sg_free_table(sgt); return;
kfree(sgt);
} dma_unmap_sgtable(map->dev, map->sgt, map->direction, 0);
sg_free_table(map->sgt);
kfree(map->sgt);
host1x_bo_put(map->bo);
kfree(map);
} }
static void *gather_bo_mmap(struct host1x_bo *host_bo) static void *gather_bo_mmap(struct host1x_bo *host_bo)

70
drivers/gpu/drm/tegra/uapi.c
View File

@ -17,11 +17,7 @@ static void tegra_drm_mapping_release(struct kref *ref)
struct tegra_drm_mapping *mapping = struct tegra_drm_mapping *mapping =
container_of(ref, struct tegra_drm_mapping, ref); container_of(ref, struct tegra_drm_mapping, ref);
if (mapping->sgt) host1x_bo_unpin(mapping->map);
dma_unmap_sgtable(mapping->dev, mapping->sgt, mapping->direction,
DMA_ATTR_SKIP_CPU_SYNC);
host1x_bo_unpin(mapping->dev, mapping->bo, mapping->sgt);
host1x_bo_put(mapping->bo); host1x_bo_put(mapping->bo);
kfree(mapping); kfree(mapping);
@ -159,6 +155,7 @@ int tegra_drm_ioctl_channel_map(struct drm_device *drm, void *data, struct drm_f
struct drm_tegra_channel_map *args = data; struct drm_tegra_channel_map *args = data;
struct tegra_drm_mapping *mapping; struct tegra_drm_mapping *mapping;
struct tegra_drm_context *context; struct tegra_drm_context *context;
enum dma_data_direction direction;
int err = 0; int err = 0;
if (args->flags & ~DRM_TEGRA_CHANNEL_MAP_READ_WRITE) if (args->flags & ~DRM_TEGRA_CHANNEL_MAP_READ_WRITE)
@ -180,68 +177,53 @@ int tegra_drm_ioctl_channel_map(struct drm_device *drm, void *data, struct drm_f
kref_init(&mapping->ref); kref_init(&mapping->ref);
mapping->dev = context->client->base.dev;
mapping->bo = tegra_gem_lookup(file, args->handle); mapping->bo = tegra_gem_lookup(file, args->handle);
if (!mapping->bo) { if (!mapping->bo) {
err = -EINVAL; err = -EINVAL;
goto unlock; goto free;
} }
if (context->client->base.group) { switch (args->flags & DRM_TEGRA_CHANNEL_MAP_READ_WRITE) {
/* IOMMU domain managed directly using IOMMU API */ case DRM_TEGRA_CHANNEL_MAP_READ_WRITE:
host1x_bo_pin(mapping->dev, mapping->bo, &mapping->iova); direction = DMA_BIDIRECTIONAL;
} else { break;
switch (args->flags & DRM_TEGRA_CHANNEL_MAP_READ_WRITE) {
case DRM_TEGRA_CHANNEL_MAP_READ_WRITE:
mapping->direction = DMA_BIDIRECTIONAL;
break;
case DRM_TEGRA_CHANNEL_MAP_WRITE: case DRM_TEGRA_CHANNEL_MAP_WRITE:
mapping->direction = DMA_FROM_DEVICE; direction = DMA_FROM_DEVICE;
break; break;
case DRM_TEGRA_CHANNEL_MAP_READ: case DRM_TEGRA_CHANNEL_MAP_READ:
mapping->direction = DMA_TO_DEVICE; direction = DMA_TO_DEVICE;
break; break;
default: default:
return -EINVAL; err = -EINVAL;
} goto put_gem;
mapping->sgt = host1x_bo_pin(mapping->dev, mapping->bo, NULL);
if (IS_ERR(mapping->sgt)) {
err = PTR_ERR(mapping->sgt);
goto put_gem;
}
err = dma_map_sgtable(mapping->dev, mapping->sgt, mapping->direction,
DMA_ATTR_SKIP_CPU_SYNC);
if (err)
goto unpin;
mapping->iova = sg_dma_address(mapping->sgt->sgl);
} }
mapping->map = host1x_bo_pin(context->client->base.dev, mapping->bo, direction);
if (IS_ERR(mapping->map)) {
err = PTR_ERR(mapping->map);
goto put_gem;
}
mapping->iova = mapping->map->phys;
mapping->iova_end = mapping->iova + host1x_to_tegra_bo(mapping->bo)->gem.size; mapping->iova_end = mapping->iova + host1x_to_tegra_bo(mapping->bo)->gem.size;
err = xa_alloc(&context->mappings, &args->mapping, mapping, XA_LIMIT(1, U32_MAX), err = xa_alloc(&context->mappings, &args->mapping, mapping, XA_LIMIT(1, U32_MAX),
GFP_KERNEL); GFP_KERNEL);
if (err < 0) if (err < 0)
goto unmap; goto unpin;
mutex_unlock(&fpriv->lock); mutex_unlock(&fpriv->lock);
return 0; return 0;
unmap:
if (mapping->sgt) {
dma_unmap_sgtable(mapping->dev, mapping->sgt, mapping->direction,
DMA_ATTR_SKIP_CPU_SYNC);
}
unpin: unpin:
host1x_bo_unpin(mapping->dev, mapping->bo, mapping->sgt); host1x_bo_unpin(mapping->map);
put_gem: put_gem:
host1x_bo_put(mapping->bo); host1x_bo_put(mapping->bo);
free:
kfree(mapping); kfree(mapping);
unlock: unlock:
mutex_unlock(&fpriv->lock); mutex_unlock(&fpriv->lock);

5
drivers/gpu/drm/tegra/uapi.h
View File

@ -27,10 +27,9 @@ struct tegra_drm_file {
struct tegra_drm_mapping { struct tegra_drm_mapping {
struct kref ref; struct kref ref;
struct device *dev; struct host1x_bo_mapping *map;
struct host1x_bo *bo; struct host1x_bo *bo;
struct sg_table *sgt;
enum dma_data_direction direction;
dma_addr_t iova; dma_addr_t iova;
dma_addr_t iova_end; dma_addr_t iova_end;
}; };

162
drivers/gpu/host1x/job.c
View File

@ -134,20 +134,20 @@ EXPORT_SYMBOL(host1x_job_add_wait);
static unsigned int pin_job(struct host1x *host, struct host1x_job *job) static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
{ {
unsigned long mask = HOST1X_RELOC_READ | HOST1X_RELOC_WRITE;
struct host1x_client *client = job->client; struct host1x_client *client = job->client;
struct device *dev = client->dev; struct device *dev = client->dev;
struct host1x_job_gather *g; struct host1x_job_gather *g;
struct iommu_domain *domain;
struct sg_table *sgt;
unsigned int i; unsigned int i;
int err; int err;
domain = iommu_get_domain_for_dev(dev);
job->num_unpins = 0; job->num_unpins = 0;
for (i = 0; i < job->num_relocs; i++) { for (i = 0; i < job->num_relocs; i++) {
struct host1x_reloc *reloc = &job->relocs[i]; struct host1x_reloc *reloc = &job->relocs[i];
dma_addr_t phys_addr, *phys; enum dma_data_direction direction;
struct host1x_bo_mapping *map;
struct host1x_bo *bo;
reloc->target.bo = host1x_bo_get(reloc->target.bo); reloc->target.bo = host1x_bo_get(reloc->target.bo);
if (!reloc->target.bo) { if (!reloc->target.bo) {
@ -155,64 +155,44 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
goto unpin; goto unpin;
} }
/* bo = reloc->target.bo;
* If the client device is not attached to an IOMMU, the
* physical address of the buffer object can be used.
*
* Similarly, when an IOMMU domain is shared between all
* host1x clients, the IOVA is already available, so no
* need to map the buffer object again.
*
* XXX Note that this isn't always safe to do because it
* relies on an assumption that no cache maintenance is
* needed on the buffer objects.
*/
if (!domain || client->group)
phys = &phys_addr;
else
phys = NULL;
sgt = host1x_bo_pin(dev, reloc->target.bo, phys); switch (reloc->flags & mask) {
if (IS_ERR(sgt)) { case HOST1X_RELOC_READ:
err = PTR_ERR(sgt); direction = DMA_TO_DEVICE;
break;
case HOST1X_RELOC_WRITE:
direction = DMA_FROM_DEVICE;
break;
case HOST1X_RELOC_READ | HOST1X_RELOC_WRITE:
direction = DMA_BIDIRECTIONAL;
break;
default:
err = -EINVAL;
goto unpin; goto unpin;
} }
if (sgt) { map = host1x_bo_pin(dev, bo, direction);
unsigned long mask = HOST1X_RELOC_READ | if (IS_ERR(map)) {
HOST1X_RELOC_WRITE; err = PTR_ERR(map);
enum dma_data_direction dir; goto unpin;
switch (reloc->flags & mask) {
case HOST1X_RELOC_READ:
dir = DMA_TO_DEVICE;
break;
case HOST1X_RELOC_WRITE:
dir = DMA_FROM_DEVICE;
break;
case HOST1X_RELOC_READ | HOST1X_RELOC_WRITE:
dir = DMA_BIDIRECTIONAL;
break;
default:
err = -EINVAL;
goto unpin;
}
err = dma_map_sgtable(dev, sgt, dir, 0);
if (err)
goto unpin;
job->unpins[job->num_unpins].dev = dev;
job->unpins[job->num_unpins].dir = dir;
phys_addr = sg_dma_address(sgt->sgl);
} }
job->addr_phys[job->num_unpins] = phys_addr; /*
job->unpins[job->num_unpins].bo = reloc->target.bo; * host1x clients are generally not able to do scatter-gather themselves, so fail
job->unpins[job->num_unpins].sgt = sgt; * if the buffer is discontiguous and we fail to map its SG table to a single
* contiguous chunk of I/O virtual memory.
*/
if (map->chunks > 1) {
err = -EINVAL;
goto unpin;
}
job->addr_phys[job->num_unpins] = map->phys;
job->unpins[job->num_unpins].map = map;
job->num_unpins++; job->num_unpins++;
} }
@ -224,12 +204,11 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
return 0; return 0;
for (i = 0; i < job->num_cmds; i++) { for (i = 0; i < job->num_cmds; i++) {
struct host1x_bo_mapping *map;
size_t gather_size = 0; size_t gather_size = 0;
struct scatterlist *sg; struct scatterlist *sg;
dma_addr_t phys_addr;
unsigned long shift; unsigned long shift;
struct iova *alloc; struct iova *alloc;
dma_addr_t *phys;
unsigned int j; unsigned int j;
if (job->cmds[i].is_wait) if (job->cmds[i].is_wait)
@ -243,25 +222,16 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
goto unpin; goto unpin;
} }
/** map = host1x_bo_pin(host->dev, g->bo, DMA_TO_DEVICE);
* If the host1x is not attached to an IOMMU, there is no need if (IS_ERR(map)) {
* to map the buffer object for the host1x, since the physical err = PTR_ERR(map);
* address can simply be used. goto unpin;
*/
if (!iommu_get_domain_for_dev(host->dev))
phys = &phys_addr;
else
phys = NULL;
sgt = host1x_bo_pin(host->dev, g->bo, phys);
if (IS_ERR(sgt)) {
err = PTR_ERR(sgt);
goto put;
} }
if (host->domain) { if (host->domain) {
for_each_sgtable_sg(sgt, sg, j) for_each_sgtable_sg(map->sgt, sg, j)
gather_size += sg->length; gather_size += sg->length;
gather_size = iova_align(&host->iova, gather_size); gather_size = iova_align(&host->iova, gather_size);
shift = iova_shift(&host->iova); shift = iova_shift(&host->iova);
@ -272,33 +242,23 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
goto put; goto put;
} }
err = iommu_map_sgtable(host->domain, err = iommu_map_sgtable(host->domain, iova_dma_addr(&host->iova, alloc),
iova_dma_addr(&host->iova, alloc), map->sgt, IOMMU_READ);
sgt, IOMMU_READ);
if (err == 0) { if (err == 0) {
__free_iova(&host->iova, alloc); __free_iova(&host->iova, alloc);
err = -EINVAL; err = -EINVAL;
goto put; goto put;
} }
job->unpins[job->num_unpins].size = gather_size; map->phys = iova_dma_addr(&host->iova, alloc);
phys_addr = iova_dma_addr(&host->iova, alloc); map->size = gather_size;
} else if (sgt) {
err = dma_map_sgtable(host->dev, sgt, DMA_TO_DEVICE, 0);
if (err)
goto put;
job->unpins[job->num_unpins].dir = DMA_TO_DEVICE;
job->unpins[job->num_unpins].dev = host->dev;
phys_addr = sg_dma_address(sgt->sgl);
} }
job->addr_phys[job->num_unpins] = phys_addr; job->addr_phys[job->num_unpins] = map->phys;
job->gather_addr_phys[i] = phys_addr; job->unpins[job->num_unpins].map = map;
job->unpins[job->num_unpins].bo = g->bo;
job->unpins[job->num_unpins].sgt = sgt;
job->num_unpins++; job->num_unpins++;
job->gather_addr_phys[i] = map->phys;
} }
return 0; return 0;
@ -690,22 +650,16 @@ void host1x_job_unpin(struct host1x_job *job)
unsigned int i; unsigned int i;
for (i = 0; i < job->num_unpins; i++) { for (i = 0; i < job->num_unpins; i++) {
struct host1x_job_unpin_data *unpin = &job->unpins[i]; struct host1x_bo_mapping *map = job->unpins[i].map;
struct device *dev = unpin->dev ?: host->dev; struct host1x_bo *bo = map->bo;
struct sg_table *sgt = unpin->sgt;
if (!job->enable_firewall && unpin->size && host->domain) { if (!job->enable_firewall && map->size && host->domain) {
iommu_unmap(host->domain, job->addr_phys[i], iommu_unmap(host->domain, job->addr_phys[i], map->size);
unpin->size); free_iova(&host->iova, iova_pfn(&host->iova, job->addr_phys[i]));
free_iova(&host->iova,
iova_pfn(&host->iova, job->addr_phys[i]));
} }
if (unpin->dev && sgt) host1x_bo_unpin(map);
dma_unmap_sgtable(unpin->dev, sgt, unpin->dir, 0); host1x_bo_put(bo);
host1x_bo_unpin(dev, unpin->bo, sgt);
host1x_bo_put(unpin->bo);
} }
job->num_unpins = 0; job->num_unpins = 0;

6
drivers/gpu/host1x/job.h
View File

@ -35,11 +35,7 @@ struct host1x_job_cmd {
}; };
struct host1x_job_unpin_data { struct host1x_job_unpin_data {
struct host1x_bo *bo; struct host1x_bo_mapping *map;
struct sg_table *sgt;
struct device *dev;
size_t size;
enum dma_data_direction dir;
}; };
/* /*

30
include/linux/host1x.h
View File

@ -7,6 +7,7 @@
#define __LINUX_HOST1X_H #define __LINUX_HOST1X_H
#include <linux/device.h> #include <linux/device.h>
#include <linux/dma-direction.h>
#include <linux/types.h> #include <linux/types.h>
enum host1x_class { enum host1x_class {
@ -82,12 +83,23 @@ struct host1x_client {
struct host1x_bo; struct host1x_bo;
struct sg_table; struct sg_table;
struct host1x_bo_mapping {
struct dma_buf_attachment *attach;
enum dma_data_direction direction;
struct host1x_bo *bo;
struct sg_table *sgt;
unsigned int chunks;
struct device *dev;
dma_addr_t phys;
size_t size;
};
struct host1x_bo_ops { struct host1x_bo_ops {
struct host1x_bo *(*get)(struct host1x_bo *bo); struct host1x_bo *(*get)(struct host1x_bo *bo);
void (*put)(struct host1x_bo *bo); void (*put)(struct host1x_bo *bo);
struct sg_table *(*pin)(struct device *dev, struct host1x_bo *bo, struct host1x_bo_mapping *(*pin)(struct device *dev, struct host1x_bo *bo,
dma_addr_t *phys); enum dma_data_direction dir);
void (*unpin)(struct device *dev, struct sg_table *sgt); void (*unpin)(struct host1x_bo_mapping *map);
void *(*mmap)(struct host1x_bo *bo); void *(*mmap)(struct host1x_bo *bo);
void (*munmap)(struct host1x_bo *bo, void *addr); void (*munmap)(struct host1x_bo *bo, void *addr);
}; };
@ -112,17 +124,15 @@ static inline void host1x_bo_put(struct host1x_bo *bo)
bo->ops->put(bo); bo->ops->put(bo);
} }
static inline struct sg_table *host1x_bo_pin(struct device *dev, static inline struct host1x_bo_mapping *host1x_bo_pin(struct device *dev, struct host1x_bo *bo,
struct host1x_bo *bo, enum dma_data_direction dir)
dma_addr_t *phys)
{ {
return bo->ops->pin(dev, bo, phys); return bo->ops->pin(dev, bo, dir);
} }
static inline void host1x_bo_unpin(struct device *dev, struct host1x_bo *bo, static inline void host1x_bo_unpin(struct host1x_bo_mapping *map)
struct sg_table *sgt)
{ {
bo->ops->unpin(dev, sgt); map->bo->ops->unpin(map);
} }
static inline void *host1x_bo_mmap(struct host1x_bo *bo) static inline void *host1x_bo_mmap(struct host1x_bo *bo)