linux/drivers/gpu/drm/msm/msm_gem.c

876 lines
21 KiB
C
Raw Normal View History

/*
* Copyright (C) 2013 Red Hat
* Author: Rob Clark <robdclark@gmail.com>
*
* 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/>.
*/
#include <linux/spinlock.h>
#include <linux/shmem_fs.h>
#include <linux/dma-buf.h>
#include <linux/pfn_t.h>
#include "msm_drv.h"
#include "msm_fence.h"
#include "msm_gem.h"
#include "msm_gpu.h"
#include "msm_mmu.h"
static dma_addr_t physaddr(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
struct msm_drm_private *priv = obj->dev->dev_private;
return (((dma_addr_t)msm_obj->vram_node->start) << PAGE_SHIFT) +
priv->vram.paddr;
}
static bool use_pages(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
return !msm_obj->vram_node;
}
/* allocate pages from VRAM carveout, used when no IOMMU: */
static struct page **get_pages_vram(struct drm_gem_object *obj,
int npages)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
struct msm_drm_private *priv = obj->dev->dev_private;
dma_addr_t paddr;
struct page **p;
int ret, i;
p = drm_malloc_ab(npages, sizeof(struct page *));
if (!p)
return ERR_PTR(-ENOMEM);
drm: Improve drm_mm search (and fix topdown allocation) with rbtrees The drm_mm range manager claimed to support top-down insertion, but it was neither searching for the top-most hole that could fit the allocation request nor fitting the request to the hole correctly. In order to search the range efficiently, we create a secondary index for the holes using either their size or their address. This index allows us to find the smallest hole or the hole at the bottom or top of the range efficiently, whilst keeping the hole stack to rapidly service evictions. v2: Search for holes both high and low. Rename flags to mode. v3: Discover rb_entry_safe() and use it! v4: Kerneldoc for enum drm_mm_insert_mode. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: "Christian König" <christian.koenig@amd.com> Cc: David Airlie <airlied@linux.ie> Cc: Russell King <rmk+kernel@armlinux.org.uk> Cc: Daniel Vetter <daniel.vetter@intel.com> Cc: Jani Nikula <jani.nikula@linux.intel.com> Cc: Sean Paul <seanpaul@chromium.org> Cc: Lucas Stach <l.stach@pengutronix.de> Cc: Christian Gmeiner <christian.gmeiner@gmail.com> Cc: Rob Clark <robdclark@gmail.com> Cc: Thierry Reding <thierry.reding@gmail.com> Cc: Stephen Warren <swarren@wwwdotorg.org> Cc: Alexandre Courbot <gnurou@gmail.com> Cc: Eric Anholt <eric@anholt.net> Cc: Sinclair Yeh <syeh@vmware.com> Cc: Thomas Hellstrom <thellstrom@vmware.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Sinclair Yeh <syeh@vmware.com> # vmwgfx Reviewed-by: Lucas Stach <l.stach@pengutronix.de> #etnaviv Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20170202210438.28702-1-chris@chris-wilson.co.uk
2017-02-03 05:04:38 +08:00
ret = drm_mm_insert_node(&priv->vram.mm, msm_obj->vram_node, npages);
if (ret) {
drm_free_large(p);
return ERR_PTR(ret);
}
paddr = physaddr(obj);
for (i = 0; i < npages; i++) {
p[i] = phys_to_page(paddr);
paddr += PAGE_SIZE;
}
return p;
}
/* called with dev->struct_mutex held */
static struct page **get_pages(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
if (!msm_obj->pages) {
struct drm_device *dev = obj->dev;
struct page **p;
int npages = obj->size >> PAGE_SHIFT;
if (use_pages(obj))
p = drm_gem_get_pages(obj);
else
p = get_pages_vram(obj, npages);
if (IS_ERR(p)) {
dev_err(dev->dev, "could not get pages: %ld\n",
PTR_ERR(p));
return p;
}
msm_obj->sgt = drm_prime_pages_to_sg(p, npages);
if (IS_ERR(msm_obj->sgt)) {
dev_err(dev->dev, "failed to allocate sgt\n");
return ERR_CAST(msm_obj->sgt);
}
msm_obj->pages = p;
/* For non-cached buffers, ensure the new pages are clean
* because display controller, GPU, etc. are not coherent:
*/
if (msm_obj->flags & (MSM_BO_WC|MSM_BO_UNCACHED))
dma_map_sg(dev->dev, msm_obj->sgt->sgl,
msm_obj->sgt->nents, DMA_BIDIRECTIONAL);
}
return msm_obj->pages;
}
static void put_pages(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
if (msm_obj->pages) {
/* For non-cached buffers, ensure the new pages are clean
* because display controller, GPU, etc. are not coherent:
*/
if (msm_obj->flags & (MSM_BO_WC|MSM_BO_UNCACHED))
dma_unmap_sg(obj->dev->dev, msm_obj->sgt->sgl,
msm_obj->sgt->nents, DMA_BIDIRECTIONAL);
sg_free_table(msm_obj->sgt);
kfree(msm_obj->sgt);
if (use_pages(obj))
drm_gem_put_pages(obj, msm_obj->pages, true, false);
else {
drm_mm_remove_node(msm_obj->vram_node);
drm_free_large(msm_obj->pages);
}
msm_obj->pages = NULL;
}
}
struct page **msm_gem_get_pages(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct page **p;
mutex_lock(&dev->struct_mutex);
p = get_pages(obj);
mutex_unlock(&dev->struct_mutex);
return p;
}
void msm_gem_put_pages(struct drm_gem_object *obj)
{
/* when we start tracking the pin count, then do something here */
}
int msm_gem_mmap_obj(struct drm_gem_object *obj,
struct vm_area_struct *vma)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
vma->vm_flags &= ~VM_PFNMAP;
vma->vm_flags |= VM_MIXEDMAP;
if (msm_obj->flags & MSM_BO_WC) {
vma->vm_page_prot = pgprot_writecombine(vm_get_page_prot(vma->vm_flags));
} else if (msm_obj->flags & MSM_BO_UNCACHED) {
vma->vm_page_prot = pgprot_noncached(vm_get_page_prot(vma->vm_flags));
} else {
/*
* Shunt off cached objs to shmem file so they have their own
* address_space (so unmap_mapping_range does what we want,
* in particular in the case of mmap'd dmabufs)
*/
fput(vma->vm_file);
get_file(obj->filp);
vma->vm_pgoff = 0;
vma->vm_file = obj->filp;
vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
}
return 0;
}
int msm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
{
int ret;
ret = drm_gem_mmap(filp, vma);
if (ret) {
DBG("mmap failed: %d", ret);
return ret;
}
return msm_gem_mmap_obj(vma->vm_private_data, vma);
}
int msm_gem_fault(struct vm_fault *vmf)
{
struct vm_area_struct *vma = vmf->vma;
struct drm_gem_object *obj = vma->vm_private_data;
struct drm_device *dev = obj->dev;
struct msm_drm_private *priv = dev->dev_private;
struct page **pages;
unsigned long pfn;
pgoff_t pgoff;
int ret;
/* This should only happen if userspace tries to pass a mmap'd
* but unfaulted gem bo vaddr into submit ioctl, triggering
* a page fault while struct_mutex is already held. This is
* not a valid use-case so just bail.
*/
if (priv->struct_mutex_task == current)
return VM_FAULT_SIGBUS;
/* Make sure we don't parallel update on a fault, nor move or remove
* something from beneath our feet
*/
ret = mutex_lock_interruptible(&dev->struct_mutex);
if (ret)
goto out;
/* make sure we have pages attached now */
pages = get_pages(obj);
if (IS_ERR(pages)) {
ret = PTR_ERR(pages);
goto out_unlock;
}
/* We don't use vmf->pgoff since that has the fake offset: */
pgoff = (vmf->address - vma->vm_start) >> PAGE_SHIFT;
pfn = page_to_pfn(pages[pgoff]);
VERB("Inserting %p pfn %lx, pa %lx", (void *)vmf->address,
pfn, pfn << PAGE_SHIFT);
ret = vm_insert_mixed(vma, vmf->address, __pfn_to_pfn_t(pfn, PFN_DEV));
out_unlock:
mutex_unlock(&dev->struct_mutex);
out:
switch (ret) {
case -EAGAIN:
case 0:
case -ERESTARTSYS:
case -EINTR:
case -EBUSY:
/*
* EBUSY is ok: this just means that another thread
* already did the job.
*/
return VM_FAULT_NOPAGE;
case -ENOMEM:
return VM_FAULT_OOM;
default:
return VM_FAULT_SIGBUS;
}
}
/** get mmap offset */
static uint64_t mmap_offset(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
int ret;
WARN_ON(!mutex_is_locked(&dev->struct_mutex));
/* Make it mmapable */
ret = drm_gem_create_mmap_offset(obj);
if (ret) {
dev_err(dev->dev, "could not allocate mmap offset\n");
return 0;
}
return drm_vma_node_offset_addr(&obj->vma_node);
}
uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
{
uint64_t offset;
mutex_lock(&obj->dev->struct_mutex);
offset = mmap_offset(obj);
mutex_unlock(&obj->dev->struct_mutex);
return offset;
}
static void
put_iova(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct msm_drm_private *priv = obj->dev->dev_private;
struct msm_gem_object *msm_obj = to_msm_bo(obj);
int id;
WARN_ON(!mutex_is_locked(&dev->struct_mutex));
for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) {
if (!priv->aspace[id])
continue;
msm_gem_unmap_vma(priv->aspace[id],
&msm_obj->domain[id], msm_obj->sgt);
}
}
/* should be called under struct_mutex.. although it can be called
* from atomic context without struct_mutex to acquire an extra
* iova ref if you know one is already held.
*
* That means when I do eventually need to add support for unpinning
* the refcnt counter needs to be atomic_t.
*/
int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
uint64_t *iova)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
int ret = 0;
if (!msm_obj->domain[id].iova) {
struct msm_drm_private *priv = obj->dev->dev_private;
struct page **pages = get_pages(obj);
if (IS_ERR(pages))
return PTR_ERR(pages);
if (iommu_present(&platform_bus_type)) {
ret = msm_gem_map_vma(priv->aspace[id], &msm_obj->domain[id],
msm_obj->sgt, obj->size >> PAGE_SHIFT);
} else {
msm_obj->domain[id].iova = physaddr(obj);
}
}
if (!ret)
*iova = msm_obj->domain[id].iova;
return ret;
}
/* get iova, taking a reference. Should have a matching put */
int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
int ret;
/* this is safe right now because we don't unmap until the
* bo is deleted:
*/
if (msm_obj->domain[id].iova) {
*iova = msm_obj->domain[id].iova;
return 0;
}
mutex_lock(&obj->dev->struct_mutex);
ret = msm_gem_get_iova_locked(obj, id, iova);
mutex_unlock(&obj->dev->struct_mutex);
return ret;
}
/* get iova without taking a reference, used in places where you have
* already done a 'msm_gem_get_iova()'.
*/
uint64_t msm_gem_iova(struct drm_gem_object *obj, int id)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!msm_obj->domain[id].iova);
return msm_obj->domain[id].iova;
}
void msm_gem_put_iova(struct drm_gem_object *obj, int id)
{
// XXX TODO ..
// NOTE: probably don't need a _locked() version.. we wouldn't
// normally unmap here, but instead just mark that it could be
// unmapped (if the iova refcnt drops to zero), but then later
// if another _get_iova_locked() fails we can start unmapping
// things that are no longer needed..
}
int msm_gem_dumb_create(struct drm_file *file, struct drm_device *dev,
struct drm_mode_create_dumb *args)
{
args->pitch = align_pitch(args->width, args->bpp);
args->size = PAGE_ALIGN(args->pitch * args->height);
return msm_gem_new_handle(dev, file, args->size,
MSM_BO_SCANOUT | MSM_BO_WC, &args->handle);
}
int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
uint32_t handle, uint64_t *offset)
{
struct drm_gem_object *obj;
int ret = 0;
/* GEM does all our handle to object mapping */
obj = drm_gem_object_lookup(file, handle);
if (obj == NULL) {
ret = -ENOENT;
goto fail;
}
*offset = msm_gem_mmap_offset(obj);
drm_gem_object_unreference_unlocked(obj);
fail:
return ret;
}
void *msm_gem_get_vaddr_locked(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
if (!msm_obj->vaddr) {
struct page **pages = get_pages(obj);
if (IS_ERR(pages))
return ERR_CAST(pages);
msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
VM_MAP, pgprot_writecombine(PAGE_KERNEL));
if (msm_obj->vaddr == NULL)
return ERR_PTR(-ENOMEM);
}
msm_obj->vmap_count++;
return msm_obj->vaddr;
}
void *msm_gem_get_vaddr(struct drm_gem_object *obj)
{
void *ret;
mutex_lock(&obj->dev->struct_mutex);
ret = msm_gem_get_vaddr_locked(obj);
mutex_unlock(&obj->dev->struct_mutex);
return ret;
}
void msm_gem_put_vaddr_locked(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
WARN_ON(msm_obj->vmap_count < 1);
msm_obj->vmap_count--;
}
void msm_gem_put_vaddr(struct drm_gem_object *obj)
{
mutex_lock(&obj->dev->struct_mutex);
msm_gem_put_vaddr_locked(obj);
mutex_unlock(&obj->dev->struct_mutex);
}
/* Update madvise status, returns true if not purged, else
* false or -errno.
*/
int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
if (msm_obj->madv != __MSM_MADV_PURGED)
msm_obj->madv = madv;
return (msm_obj->madv != __MSM_MADV_PURGED);
}
void msm_gem_purge(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!mutex_is_locked(&dev->struct_mutex));
WARN_ON(!is_purgeable(msm_obj));
WARN_ON(obj->import_attach);
put_iova(obj);
msm_gem_vunmap(obj);
put_pages(obj);
msm_obj->madv = __MSM_MADV_PURGED;
drm_vma_node_unmap(&obj->vma_node, dev->anon_inode->i_mapping);
drm_gem_free_mmap_offset(obj);
/* Our goal here is to return as much of the memory as
* is possible back to the system as we are called from OOM.
* To do this we must instruct the shmfs to drop all of its
* backing pages, *now*.
*/
shmem_truncate_range(file_inode(obj->filp), 0, (loff_t)-1);
invalidate_mapping_pages(file_inode(obj->filp)->i_mapping,
0, (loff_t)-1);
}
void msm_gem_vunmap(struct drm_gem_object *obj)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
if (!msm_obj->vaddr || WARN_ON(!is_vunmapable(msm_obj)))
return;
vunmap(msm_obj->vaddr);
msm_obj->vaddr = NULL;
}
/* must be called before _move_to_active().. */
int msm_gem_sync_object(struct drm_gem_object *obj,
struct msm_fence_context *fctx, bool exclusive)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
struct reservation_object_list *fobj;
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
struct dma_fence *fence;
int i, ret;
if (!exclusive) {
/* NOTE: _reserve_shared() must happen before _add_shared_fence(),
* which makes this a slightly strange place to call it. OTOH this
* is a convenient can-fail point to hook it in. (And similar to
* how etnaviv and nouveau handle this.)
*/
ret = reservation_object_reserve_shared(msm_obj->resv);
if (ret)
return ret;
}
fobj = reservation_object_get_list(msm_obj->resv);
if (!fobj || (fobj->shared_count == 0)) {
fence = reservation_object_get_excl(msm_obj->resv);
/* don't need to wait on our own fences, since ring is fifo */
if (fence && (fence->context != fctx->context)) {
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
ret = dma_fence_wait(fence, true);
if (ret)
return ret;
}
}
if (!exclusive || !fobj)
return 0;
for (i = 0; i < fobj->shared_count; i++) {
fence = rcu_dereference_protected(fobj->shared[i],
reservation_object_held(msm_obj->resv));
if (fence->context != fctx->context) {
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
ret = dma_fence_wait(fence, true);
if (ret)
return ret;
}
}
return 0;
}
void msm_gem_move_to_active(struct drm_gem_object *obj,
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
struct msm_gpu *gpu, bool exclusive, struct dma_fence *fence)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED);
msm_obj->gpu = gpu;
if (exclusive)
reservation_object_add_excl_fence(msm_obj->resv, fence);
else
reservation_object_add_shared_fence(msm_obj->resv, fence);
list_del_init(&msm_obj->mm_list);
list_add_tail(&msm_obj->mm_list, &gpu->active_list);
}
void msm_gem_move_to_inactive(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct msm_drm_private *priv = dev->dev_private;
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!mutex_is_locked(&dev->struct_mutex));
msm_obj->gpu = NULL;
list_del_init(&msm_obj->mm_list);
list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
}
int msm_gem_cpu_prep(struct drm_gem_object *obj, uint32_t op, ktime_t *timeout)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
bool write = !!(op & MSM_PREP_WRITE);
unsigned long remain =
op & MSM_PREP_NOSYNC ? 0 : timeout_to_jiffies(timeout);
long ret;
ret = reservation_object_wait_timeout_rcu(msm_obj->resv, write,
true, remain);
if (ret == 0)
return remain == 0 ? -EBUSY : -ETIMEDOUT;
else if (ret < 0)
return ret;
/* TODO cache maintenance */
return 0;
}
int msm_gem_cpu_fini(struct drm_gem_object *obj)
{
/* TODO cache maintenance */
return 0;
}
#ifdef CONFIG_DEBUG_FS
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
static void describe_fence(struct dma_fence *fence, const char *type,
struct seq_file *m)
{
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
if (!dma_fence_is_signaled(fence))
seq_printf(m, "\t%9s: %s %s seq %u\n", type,
fence->ops->get_driver_name(fence),
fence->ops->get_timeline_name(fence),
fence->seqno);
}
void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)
{
struct msm_gem_object *msm_obj = to_msm_bo(obj);
struct reservation_object *robj = msm_obj->resv;
struct reservation_object_list *fobj;
struct msm_drm_private *priv = obj->dev->dev_private;
dma-buf: Rename struct fence to dma_fence I plan to usurp the short name of struct fence for a core kernel struct, and so I need to rename the specialised fence/timeline for DMA operations to make room. A consensus was reached in https://lists.freedesktop.org/archives/dri-devel/2016-July/113083.html that making clear this fence applies to DMA operations was a good thing. Since then the patch has grown a bit as usage increases, so hopefully it remains a good thing! (v2...: rebase, rerun spatch) v3: Compile on msm, spotted a manual fixup that I broke. v4: Try again for msm, sorry Daniel coccinelle script: @@ @@ - struct fence + struct dma_fence @@ @@ - struct fence_ops + struct dma_fence_ops @@ @@ - struct fence_cb + struct dma_fence_cb @@ @@ - struct fence_array + struct dma_fence_array @@ @@ - enum fence_flag_bits + enum dma_fence_flag_bits @@ @@ ( - fence_init + dma_fence_init | - fence_release + dma_fence_release | - fence_free + dma_fence_free | - fence_get + dma_fence_get | - fence_get_rcu + dma_fence_get_rcu | - fence_put + dma_fence_put | - fence_signal + dma_fence_signal | - fence_signal_locked + dma_fence_signal_locked | - fence_default_wait + dma_fence_default_wait | - fence_add_callback + dma_fence_add_callback | - fence_remove_callback + dma_fence_remove_callback | - fence_enable_sw_signaling + dma_fence_enable_sw_signaling | - fence_is_signaled_locked + dma_fence_is_signaled_locked | - fence_is_signaled + dma_fence_is_signaled | - fence_is_later + dma_fence_is_later | - fence_later + dma_fence_later | - fence_wait_timeout + dma_fence_wait_timeout | - fence_wait_any_timeout + dma_fence_wait_any_timeout | - fence_wait + dma_fence_wait | - fence_context_alloc + dma_fence_context_alloc | - fence_array_create + dma_fence_array_create | - to_fence_array + to_dma_fence_array | - fence_is_array + dma_fence_is_array | - trace_fence_emit + trace_dma_fence_emit | - FENCE_TRACE + DMA_FENCE_TRACE | - FENCE_WARN + DMA_FENCE_WARN | - FENCE_ERR + DMA_FENCE_ERR ) ( ... ) Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20161025120045.28839-1-chris@chris-wilson.co.uk
2016-10-25 20:00:45 +08:00
struct dma_fence *fence;
uint64_t off = drm_vma_node_start(&obj->vma_node);
const char *madv;
unsigned id;
WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
switch (msm_obj->madv) {
case __MSM_MADV_PURGED:
madv = " purged";
break;
case MSM_MADV_DONTNEED:
madv = " purgeable";
break;
case MSM_MADV_WILLNEED:
default:
madv = "";
break;
}
seq_printf(m, "%08x: %c %2d (%2d) %08llx %p\t",
msm_obj->flags, is_active(msm_obj) ? 'A' : 'I',
obj->name, kref_read(&obj->refcount),
off, msm_obj->vaddr);
for (id = 0; id < priv->num_aspaces; id++)
seq_printf(m, " %08llx", msm_obj->domain[id].iova);
seq_printf(m, " %zu%s\n", obj->size, madv);
rcu_read_lock();
fobj = rcu_dereference(robj->fence);
if (fobj) {
unsigned int i, shared_count = fobj->shared_count;
for (i = 0; i < shared_count; i++) {
fence = rcu_dereference(fobj->shared[i]);
describe_fence(fence, "Shared", m);
}
}
fence = rcu_dereference(robj->fence_excl);
if (fence)
describe_fence(fence, "Exclusive", m);
rcu_read_unlock();
}
void msm_gem_describe_objects(struct list_head *list, struct seq_file *m)
{
struct msm_gem_object *msm_obj;
int count = 0;
size_t size = 0;
list_for_each_entry(msm_obj, list, mm_list) {
struct drm_gem_object *obj = &msm_obj->base;
seq_printf(m, " ");
msm_gem_describe(obj, m);
count++;
size += obj->size;
}
seq_printf(m, "Total %d objects, %zu bytes\n", count, size);
}
#endif
void msm_gem_free_object(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct msm_gem_object *msm_obj = to_msm_bo(obj);
WARN_ON(!mutex_is_locked(&dev->struct_mutex));
/* object should not be on active list: */
WARN_ON(is_active(msm_obj));
list_del(&msm_obj->mm_list);
put_iova(obj);
if (obj->import_attach) {
if (msm_obj->vaddr)
dma_buf_vunmap(obj->import_attach->dmabuf, msm_obj->vaddr);
/* Don't drop the pages for imported dmabuf, as they are not
* ours, just free the array we allocated:
*/
if (msm_obj->pages)
drm_free_large(msm_obj->pages);
drm_prime_gem_destroy(obj, msm_obj->sgt);
} else {
msm_gem_vunmap(obj);
put_pages(obj);
}
if (msm_obj->resv == &msm_obj->_resv)
reservation_object_fini(msm_obj->resv);
drm_gem_object_release(obj);
kfree(msm_obj);
}
/* convenience method to construct a GEM buffer object, and userspace handle */
int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
uint32_t size, uint32_t flags, uint32_t *handle)
{
struct drm_gem_object *obj;
int ret;
ret = mutex_lock_interruptible(&dev->struct_mutex);
if (ret)
return ret;
obj = msm_gem_new(dev, size, flags);
mutex_unlock(&dev->struct_mutex);
if (IS_ERR(obj))
return PTR_ERR(obj);
ret = drm_gem_handle_create(file, obj, handle);
/* drop reference from allocate - handle holds it now */
drm_gem_object_unreference_unlocked(obj);
return ret;
}
static int msm_gem_new_impl(struct drm_device *dev,
uint32_t size, uint32_t flags,
struct reservation_object *resv,
struct drm_gem_object **obj)
{
struct msm_drm_private *priv = dev->dev_private;
struct msm_gem_object *msm_obj;
bool use_vram = false;
switch (flags & MSM_BO_CACHE_MASK) {
case MSM_BO_UNCACHED:
case MSM_BO_CACHED:
case MSM_BO_WC:
break;
default:
dev_err(dev->dev, "invalid cache flag: %x\n",
(flags & MSM_BO_CACHE_MASK));
return -EINVAL;
}
if (!iommu_present(&platform_bus_type))
use_vram = true;
else if ((flags & MSM_BO_STOLEN) && priv->vram.size)
use_vram = true;
if (WARN_ON(use_vram && !priv->vram.size))
return -EINVAL;
msm_obj = kzalloc(sizeof(*msm_obj), GFP_KERNEL);
if (!msm_obj)
return -ENOMEM;
if (use_vram)
msm_obj->vram_node = &msm_obj->domain[0].node;
msm_obj->flags = flags;
msm_obj->madv = MSM_MADV_WILLNEED;
if (resv) {
msm_obj->resv = resv;
} else {
msm_obj->resv = &msm_obj->_resv;
reservation_object_init(msm_obj->resv);
}
INIT_LIST_HEAD(&msm_obj->submit_entry);
list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
*obj = &msm_obj->base;
return 0;
}
struct drm_gem_object *msm_gem_new(struct drm_device *dev,
uint32_t size, uint32_t flags)
{
struct drm_gem_object *obj = NULL;
int ret;
WARN_ON(!mutex_is_locked(&dev->struct_mutex));
size = PAGE_ALIGN(size);
ret = msm_gem_new_impl(dev, size, flags, NULL, &obj);
if (ret)
goto fail;
if (use_pages(obj)) {
ret = drm_gem_object_init(dev, obj, size);
if (ret)
goto fail;
} else {
drm_gem_private_object_init(dev, obj, size);
}
return obj;
fail:
drm_gem_object_unreference(obj);
return ERR_PTR(ret);
}
struct drm_gem_object *msm_gem_import(struct drm_device *dev,
struct dma_buf *dmabuf, struct sg_table *sgt)
{
struct msm_gem_object *msm_obj;
struct drm_gem_object *obj;
uint32_t size;
int ret, npages;
/* if we don't have IOMMU, don't bother pretending we can import: */
if (!iommu_present(&platform_bus_type)) {
dev_err(dev->dev, "cannot import without IOMMU\n");
return ERR_PTR(-EINVAL);
}
size = PAGE_ALIGN(dmabuf->size);
ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj);
if (ret)
goto fail;
drm_gem_private_object_init(dev, obj, size);
npages = size / PAGE_SIZE;
msm_obj = to_msm_bo(obj);
msm_obj->sgt = sgt;
msm_obj->pages = drm_malloc_ab(npages, sizeof(struct page *));
if (!msm_obj->pages) {
ret = -ENOMEM;
goto fail;
}
ret = drm_prime_sg_to_page_addr_arrays(sgt, msm_obj->pages, NULL, npages);
if (ret)
goto fail;
return obj;
fail:
drm_gem_object_unreference_unlocked(obj);
return ERR_PTR(ret);
}