Skip to content

Commit

Permalink
drm/nv84/fence: access fences with full virtual address, not offset
Browse files Browse the repository at this point in the history 
Allows most of the code to be shared between nv84/nvc0 implementations,
and paves the way for doing emit/sync on non-VRAM buffers (multi-gpu,
dma-buf).

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
  • Loading branch information
Ben Skeggs committed Feb 20, 2013
1 parent fa531bc commit a34caf7
Show file tree
Hide file tree
Showing 4 changed files with 149 additions and 191 deletions.
21 changes: 20 additions & 1 deletion drivers/gpu/drm/nouveau/nouveau_fence.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -68,8 +68,27 @@ void nv17_fence_resume(struct nouveau_drm *drm);
int nv50_fence_create(struct nouveau_drm *);
int nv84_fence_create(struct nouveau_drm *);
int nvc0_fence_create(struct nouveau_drm *);
u64 nvc0_fence_crtc(struct nouveau_channel *, int crtc);

int nouveau_flip_complete(void *chan);

struct nv84_fence_chan {
struct nouveau_fence_chan base;
struct nouveau_vma vma;
struct nouveau_vma dispc_vma[4];
};

struct nv84_fence_priv {
struct nouveau_fence_priv base;
struct nouveau_bo *bo;
u32 *suspend;
};

u64 nv84_fence_crtc(struct nouveau_channel *, int);
u32 nv84_fence_read(struct nouveau_channel *);
int nv84_fence_context_new(struct nouveau_channel *);
void nv84_fence_context_del(struct nouveau_channel *);
bool nv84_fence_suspend(struct nouveau_drm *);
void nv84_fence_resume(struct nouveau_drm *);
void nv84_fence_destroy(struct nouveau_drm *);

#endif
28 changes: 20 additions & 8 deletions drivers/gpu/drm/nouveau/nv50_display.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -502,7 +502,7 @@ nv50_display_flip_next(struct drm_crtc *crtc, struct drm_framebuffer *fb,
if (ret)
return ret;

if (nv_mclass(chan->object) < NVC0_CHANNEL_IND_CLASS) {
if (nv_mclass(chan->object) < NV84_CHANNEL_IND_CLASS) {
BEGIN_NV04(chan, 0, NV11_SUBCHAN_DMA_SEMAPHORE, 2);
OUT_RING (chan, NvEvoSema0 + nv_crtc->index);
OUT_RING (chan, sync->sem.offset);
Expand All @@ -512,24 +512,36 @@ nv50_display_flip_next(struct drm_crtc *crtc, struct drm_framebuffer *fb,
OUT_RING (chan, sync->sem.offset ^ 0x10);
OUT_RING (chan, 0x74b1e000);
BEGIN_NV04(chan, 0, NV11_SUBCHAN_DMA_SEMAPHORE, 1);
if (nv_mclass(chan->object) < NV84_CHANNEL_DMA_CLASS)
OUT_RING (chan, NvSema);
else
OUT_RING (chan, chan->vram);
OUT_RING (chan, NvSema);
} else
if (nv_mclass(chan->object) < NVC0_CHANNEL_IND_CLASS) {
u64 offset = nv84_fence_crtc(chan, nv_crtc->index);
offset += sync->sem.offset;

BEGIN_NV04(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 4);
OUT_RING (chan, upper_32_bits(offset));
OUT_RING (chan, lower_32_bits(offset));
OUT_RING (chan, 0xf00d0000 | sync->sem.value);
OUT_RING (chan, 0x00000002);
BEGIN_NV04(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 4);
OUT_RING (chan, upper_32_bits(offset));
OUT_RING (chan, lower_32_bits(offset ^ 0x10));
OUT_RING (chan, 0x74b1e000);
OUT_RING (chan, 0x00000001);
} else {
u64 offset = nvc0_fence_crtc(chan, nv_crtc->index);
u64 offset = nv84_fence_crtc(chan, nv_crtc->index);
offset += sync->sem.offset;

BEGIN_NVC0(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 4);
OUT_RING (chan, upper_32_bits(offset));
OUT_RING (chan, lower_32_bits(offset));
OUT_RING (chan, 0xf00d0000 | sync->sem.value);
OUT_RING (chan, 0x1002);
OUT_RING (chan, 0x00001002);
BEGIN_NVC0(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 4);
OUT_RING (chan, upper_32_bits(offset));
OUT_RING (chan, lower_32_bits(offset ^ 0x10));
OUT_RING (chan, 0x74b1e000);
OUT_RING (chan, 0x1001);
OUT_RING (chan, 0x00001001);
}

FIRE_RING (chan);
Expand Down
153 changes: 99 additions & 54 deletions drivers/gpu/drm/nouveau/nv84_fence.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@
*/

#include <core/object.h>
#include <core/client.h>
#include <core/class.h>

#include <engine/fifo.h>
Expand All @@ -33,80 +34,96 @@

#include "nv50_display.h"

struct nv84_fence_chan {
struct nouveau_fence_chan base;
};

struct nv84_fence_priv {
struct nouveau_fence_priv base;
struct nouveau_gpuobj *mem;
};
u64
nv84_fence_crtc(struct nouveau_channel *chan, int crtc)
{
struct nv84_fence_chan *fctx = chan->fence;
return fctx->dispc_vma[crtc].offset;
}

static int
nv84_fence_emit(struct nouveau_fence *fence)
{
struct nouveau_channel *chan = fence->channel;
struct nv84_fence_chan *fctx = chan->fence;
struct nouveau_fifo_chan *fifo = (void *)chan->object;
int ret = RING_SPACE(chan, 8);
u64 addr = fctx->vma.offset + fifo->chid * 16;
int ret;

ret = RING_SPACE(chan, 8);
if (ret == 0) {
BEGIN_NV04(chan, 0, NV11_SUBCHAN_DMA_SEMAPHORE, 1);
OUT_RING (chan, NvSema);
OUT_RING (chan, chan->vram);
BEGIN_NV04(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 5);
OUT_RING (chan, upper_32_bits(fifo->chid * 16));
OUT_RING (chan, lower_32_bits(fifo->chid * 16));
OUT_RING (chan, upper_32_bits(addr));
OUT_RING (chan, lower_32_bits(addr));
OUT_RING (chan, fence->sequence);
OUT_RING (chan, NV84_SUBCHAN_SEMAPHORE_TRIGGER_WRITE_LONG);
OUT_RING (chan, 0x00000000);
FIRE_RING (chan);
}

return ret;
}


static int
nv84_fence_sync(struct nouveau_fence *fence,
struct nouveau_channel *prev, struct nouveau_channel *chan)
{
struct nv84_fence_chan *fctx = chan->fence;
struct nouveau_fifo_chan *fifo = (void *)prev->object;
int ret = RING_SPACE(chan, 7);
u64 addr = fctx->vma.offset + fifo->chid * 16;
int ret;

ret = RING_SPACE(chan, 7);
if (ret == 0) {
BEGIN_NV04(chan, 0, NV11_SUBCHAN_DMA_SEMAPHORE, 1);
OUT_RING (chan, NvSema);
OUT_RING (chan, chan->vram);
BEGIN_NV04(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 4);
OUT_RING (chan, upper_32_bits(fifo->chid * 16));
OUT_RING (chan, lower_32_bits(fifo->chid * 16));
OUT_RING (chan, upper_32_bits(addr));
OUT_RING (chan, lower_32_bits(addr));
OUT_RING (chan, fence->sequence);
OUT_RING (chan, NV84_SUBCHAN_SEMAPHORE_TRIGGER_ACQUIRE_GEQUAL);
FIRE_RING (chan);
}

return ret;
}

static u32
u32
nv84_fence_read(struct nouveau_channel *chan)
{
struct nouveau_fifo_chan *fifo = (void *)chan->object;
struct nv84_fence_priv *priv = chan->drm->fence;
return nv_ro32(priv->mem, fifo->chid * 16);
return nouveau_bo_rd32(priv->bo, fifo->chid * 16/4);
}

static void
void
nv84_fence_context_del(struct nouveau_channel *chan)
{
struct drm_device *dev = chan->drm->dev;
struct nv84_fence_priv *priv = chan->drm->fence;
struct nv84_fence_chan *fctx = chan->fence;
int i;

for (i = 0; i < dev->mode_config.num_crtc; i++) {
struct nouveau_bo *bo = nv50_display_crtc_sema(dev, i);
nouveau_bo_vma_del(bo, &fctx->dispc_vma[i]);
}

nouveau_bo_vma_del(priv->bo, &fctx->vma);
nouveau_fence_context_del(&fctx->base);
chan->fence = NULL;
kfree(fctx);
}

static int
int
nv84_fence_context_new(struct nouveau_channel *chan)
{
struct drm_device *dev = chan->drm->dev;
struct nouveau_fifo_chan *fifo = (void *)chan->object;
struct nouveau_client *client = nouveau_client(fifo);
struct nv84_fence_priv *priv = chan->drm->fence;
struct nv84_fence_chan *fctx;
struct nouveau_object *object;
int ret, i;

fctx = chan->fence = kzalloc(sizeof(*fctx), GFP_KERNEL);
Expand All @@ -115,43 +132,59 @@ nv84_fence_context_new(struct nouveau_channel *chan)

nouveau_fence_context_new(&fctx->base);

ret = nouveau_object_new(nv_object(chan->cli), chan->handle,
NvSema, 0x0002,
&(struct nv_dma_class) {
.flags = NV_DMA_TARGET_VRAM |
NV_DMA_ACCESS_RDWR,
.start = priv->mem->addr,
.limit = priv->mem->addr +
priv->mem->size - 1,
}, sizeof(struct nv_dma_class),
&object);

/* dma objects for display sync channel semaphore blocks */
for (i = 0; !ret && i < dev->mode_config.num_crtc; i++) {
struct nouveau_bo *bo = nv50_display_crtc_sema(dev, i);
ret = nouveau_bo_vma_add(priv->bo, client->vm, &fctx->vma);
if (ret)
nv84_fence_context_del(chan);

ret = nouveau_object_new(nv_object(chan->cli), chan->handle,
NvEvoSema0 + i, 0x003d,
&(struct nv_dma_class) {
.flags = NV_DMA_TARGET_VRAM |
NV_DMA_ACCESS_RDWR,
.start = bo->bo.offset,
.limit = bo->bo.offset + 0xfff,
}, sizeof(struct nv_dma_class),
&object);
/* map display semaphore buffers into channel's vm */
for (i = 0; !ret && i < chan->drm->dev->mode_config.num_crtc; i++) {
struct nouveau_bo *bo = nv50_display_crtc_sema(chan->drm->dev, i);
ret = nouveau_bo_vma_add(bo, client->vm, &fctx->dispc_vma[i]);
}

if (ret)
nv84_fence_context_del(chan);
nv_wo32(priv->mem, fifo->chid * 16, 0x00000000);
nouveau_bo_wr32(priv->bo, fifo->chid * 16/4, 0x00000000);
return ret;
}

static void
bool
nv84_fence_suspend(struct nouveau_drm *drm)
{
struct nouveau_fifo *pfifo = nouveau_fifo(drm->device);
struct nv84_fence_priv *priv = drm->fence;
int i;

priv->suspend = vmalloc((pfifo->max + 1) * sizeof(u32));
if (priv->suspend) {
for (i = 0; i <= pfifo->max; i++)
priv->suspend[i] = nouveau_bo_rd32(priv->bo, i*4);
}

return priv->suspend != NULL;
}

void
nv84_fence_resume(struct nouveau_drm *drm)
{
struct nouveau_fifo *pfifo = nouveau_fifo(drm->device);
struct nv84_fence_priv *priv = drm->fence;
int i;

if (priv->suspend) {
for (i = 0; i <= pfifo->max; i++)
nouveau_bo_wr32(priv->bo, i*4, priv->suspend[i]);
vfree(priv->suspend);
priv->suspend = NULL;
}
}

void
nv84_fence_destroy(struct nouveau_drm *drm)
{
struct nv84_fence_priv *priv = drm->fence;
nouveau_gpuobj_ref(NULL, &priv->mem);
nouveau_bo_unmap(priv->bo);
if (priv->bo)
nouveau_bo_unpin(priv->bo);
nouveau_bo_ref(NULL, &priv->bo);
drm->fence = NULL;
kfree(priv);
}
Expand All @@ -161,14 +194,15 @@ nv84_fence_create(struct nouveau_drm *drm)
{
struct nouveau_fifo *pfifo = nouveau_fifo(drm->device);
struct nv84_fence_priv *priv;
u32 chan = pfifo->max + 1;
int ret;

priv = drm->fence = kzalloc(sizeof(*priv), GFP_KERNEL);
if (!priv)
return -ENOMEM;

priv->base.dtor = nv84_fence_destroy;
priv->base.suspend = nv84_fence_suspend;
priv->base.resume = nv84_fence_resume;
priv->base.context_new = nv84_fence_context_new;
priv->base.context_del = nv84_fence_context_del;
priv->base.emit = nv84_fence_emit;
Expand All @@ -178,8 +212,19 @@ nv84_fence_create(struct nouveau_drm *drm)
init_waitqueue_head(&priv->base.waiting);
priv->base.uevent = true;

ret = nouveau_gpuobj_new(drm->device, NULL, chan * 16, 0x1000, 0,
&priv->mem);
ret = nouveau_bo_new(drm->dev, 16 * (pfifo->max + 1), 0,
TTM_PL_FLAG_VRAM, 0, 0, NULL, &priv->bo);
if (ret == 0) {
ret = nouveau_bo_pin(priv->bo, TTM_PL_FLAG_VRAM);
if (ret == 0) {
ret = nouveau_bo_map(priv->bo);
if (ret)
nouveau_bo_unpin(priv->bo);
}
if (ret)
nouveau_bo_ref(NULL, &priv->bo);
}

if (ret)
nv84_fence_destroy(drm);
return ret;
Expand Down
Loading

0 comments on commit a34caf7

Please sign in to comment.