mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-28 04:25:09 +07:00
bfe91afaca
For a channel to make use of SVM features, it requires a different GPU MMU configuration than we would normally use, which is not desirable to switch to unless a client is actively going to use SVM. In order to supporting SVM without more extensive changes to the userspace interfaces, the SVM_INIT ioctl needs to replace the previous configuration safely. The only way we can currently do this safely, accounting for some unlikely failure conditions, is to allocate the new VMM without destroying the last one, and prioritising the SVM-enabled configuration in the code that cares. This will get cleaned up again further down the track. Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
223 lines
6.0 KiB
C
223 lines
6.0 KiB
C
/*
|
|
* Copyright 2012 Red Hat Inc.
|
|
*
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
* to deal in the Software without restriction, including without limitation
|
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
|
* and/or sell copies of the Software, and to permit persons to whom the
|
|
* Software is furnished to do so, subject to the following conditions:
|
|
*
|
|
* The above copyright notice and this permission notice shall be included in
|
|
* all copies or substantial portions of the Software.
|
|
*
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
|
|
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
|
|
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
|
|
* OTHER DEALINGS IN THE SOFTWARE.
|
|
*
|
|
* Authors: Ben Skeggs
|
|
*/
|
|
|
|
#include "nouveau_drv.h"
|
|
#include "nouveau_dma.h"
|
|
#include "nouveau_fence.h"
|
|
#include "nouveau_vmm.h"
|
|
|
|
#include "nv50_display.h"
|
|
|
|
static int
|
|
nv84_fence_emit32(struct nouveau_channel *chan, u64 virtual, u32 sequence)
|
|
{
|
|
int ret = RING_SPACE(chan, 8);
|
|
if (ret == 0) {
|
|
BEGIN_NV04(chan, 0, NV11_SUBCHAN_DMA_SEMAPHORE, 1);
|
|
OUT_RING (chan, chan->vram.handle);
|
|
BEGIN_NV04(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 5);
|
|
OUT_RING (chan, upper_32_bits(virtual));
|
|
OUT_RING (chan, lower_32_bits(virtual));
|
|
OUT_RING (chan, sequence);
|
|
OUT_RING (chan, NV84_SUBCHAN_SEMAPHORE_TRIGGER_WRITE_LONG);
|
|
OUT_RING (chan, 0x00000000);
|
|
FIRE_RING (chan);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
static int
|
|
nv84_fence_sync32(struct nouveau_channel *chan, u64 virtual, u32 sequence)
|
|
{
|
|
int ret = RING_SPACE(chan, 7);
|
|
if (ret == 0) {
|
|
BEGIN_NV04(chan, 0, NV11_SUBCHAN_DMA_SEMAPHORE, 1);
|
|
OUT_RING (chan, chan->vram.handle);
|
|
BEGIN_NV04(chan, 0, NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH, 4);
|
|
OUT_RING (chan, upper_32_bits(virtual));
|
|
OUT_RING (chan, lower_32_bits(virtual));
|
|
OUT_RING (chan, sequence);
|
|
OUT_RING (chan, NV84_SUBCHAN_SEMAPHORE_TRIGGER_ACQUIRE_GEQUAL);
|
|
FIRE_RING (chan);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
static int
|
|
nv84_fence_emit(struct nouveau_fence *fence)
|
|
{
|
|
struct nouveau_channel *chan = fence->channel;
|
|
struct nv84_fence_chan *fctx = chan->fence;
|
|
u64 addr = fctx->vma->addr + chan->chid * 16;
|
|
|
|
return fctx->base.emit32(chan, addr, fence->base.seqno);
|
|
}
|
|
|
|
static int
|
|
nv84_fence_sync(struct nouveau_fence *fence,
|
|
struct nouveau_channel *prev, struct nouveau_channel *chan)
|
|
{
|
|
struct nv84_fence_chan *fctx = chan->fence;
|
|
u64 addr = fctx->vma->addr + prev->chid * 16;
|
|
|
|
return fctx->base.sync32(chan, addr, fence->base.seqno);
|
|
}
|
|
|
|
static u32
|
|
nv84_fence_read(struct nouveau_channel *chan)
|
|
{
|
|
struct nv84_fence_priv *priv = chan->drm->fence;
|
|
return nouveau_bo_rd32(priv->bo, chan->chid * 16/4);
|
|
}
|
|
|
|
static void
|
|
nv84_fence_context_del(struct nouveau_channel *chan)
|
|
{
|
|
struct nv84_fence_priv *priv = chan->drm->fence;
|
|
struct nv84_fence_chan *fctx = chan->fence;
|
|
|
|
nouveau_bo_wr32(priv->bo, chan->chid * 16 / 4, fctx->base.sequence);
|
|
mutex_lock(&priv->mutex);
|
|
nouveau_vma_del(&fctx->vma);
|
|
mutex_unlock(&priv->mutex);
|
|
nouveau_fence_context_del(&fctx->base);
|
|
chan->fence = NULL;
|
|
nouveau_fence_context_free(&fctx->base);
|
|
}
|
|
|
|
int
|
|
nv84_fence_context_new(struct nouveau_channel *chan)
|
|
{
|
|
struct nv84_fence_priv *priv = chan->drm->fence;
|
|
struct nv84_fence_chan *fctx;
|
|
int ret;
|
|
|
|
fctx = chan->fence = kzalloc(sizeof(*fctx), GFP_KERNEL);
|
|
if (!fctx)
|
|
return -ENOMEM;
|
|
|
|
nouveau_fence_context_new(chan, &fctx->base);
|
|
fctx->base.emit = nv84_fence_emit;
|
|
fctx->base.sync = nv84_fence_sync;
|
|
fctx->base.read = nv84_fence_read;
|
|
fctx->base.emit32 = nv84_fence_emit32;
|
|
fctx->base.sync32 = nv84_fence_sync32;
|
|
fctx->base.sequence = nv84_fence_read(chan);
|
|
|
|
mutex_lock(&priv->mutex);
|
|
ret = nouveau_vma_new(priv->bo, chan->vmm, &fctx->vma);
|
|
mutex_unlock(&priv->mutex);
|
|
|
|
if (ret)
|
|
nv84_fence_context_del(chan);
|
|
return ret;
|
|
}
|
|
|
|
static bool
|
|
nv84_fence_suspend(struct nouveau_drm *drm)
|
|
{
|
|
struct nv84_fence_priv *priv = drm->fence;
|
|
int i;
|
|
|
|
priv->suspend = vmalloc(array_size(sizeof(u32), drm->chan.nr));
|
|
if (priv->suspend) {
|
|
for (i = 0; i < drm->chan.nr; i++)
|
|
priv->suspend[i] = nouveau_bo_rd32(priv->bo, i*4);
|
|
}
|
|
|
|
return priv->suspend != NULL;
|
|
}
|
|
|
|
static void
|
|
nv84_fence_resume(struct nouveau_drm *drm)
|
|
{
|
|
struct nv84_fence_priv *priv = drm->fence;
|
|
int i;
|
|
|
|
if (priv->suspend) {
|
|
for (i = 0; i < drm->chan.nr; i++)
|
|
nouveau_bo_wr32(priv->bo, i*4, priv->suspend[i]);
|
|
vfree(priv->suspend);
|
|
priv->suspend = NULL;
|
|
}
|
|
}
|
|
|
|
static void
|
|
nv84_fence_destroy(struct nouveau_drm *drm)
|
|
{
|
|
struct nv84_fence_priv *priv = drm->fence;
|
|
nouveau_bo_unmap(priv->bo);
|
|
if (priv->bo)
|
|
nouveau_bo_unpin(priv->bo);
|
|
nouveau_bo_ref(NULL, &priv->bo);
|
|
drm->fence = NULL;
|
|
kfree(priv);
|
|
}
|
|
|
|
int
|
|
nv84_fence_create(struct nouveau_drm *drm)
|
|
{
|
|
struct nv84_fence_priv *priv;
|
|
u32 domain;
|
|
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.uevent = true;
|
|
|
|
mutex_init(&priv->mutex);
|
|
|
|
/* Use VRAM if there is any ; otherwise fallback to system memory */
|
|
domain = drm->client.device.info.ram_size != 0 ? TTM_PL_FLAG_VRAM :
|
|
/*
|
|
* fences created in sysmem must be non-cached or we
|
|
* will lose CPU/GPU coherency!
|
|
*/
|
|
TTM_PL_FLAG_TT | TTM_PL_FLAG_UNCACHED;
|
|
ret = nouveau_bo_new(&drm->client, 16 * drm->chan.nr, 0,
|
|
domain, 0, 0, NULL, NULL, &priv->bo);
|
|
if (ret == 0) {
|
|
ret = nouveau_bo_pin(priv->bo, domain, false);
|
|
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;
|
|
}
|