linux/drivers/gpu/drm/nouveau/nv50_fifo.c
Francisco Jerez 3945e47543 drm/nouveau: Refactor context destruction to avoid a lock ordering issue.
The destroy_context() engine hooks call gpuobj management functions to
release the channel resources, these functions use HARDIRQ-unsafe locks
whereas destroy_context() is called with the HARDIRQ-safe
context_switch_lock held, that's a lock ordering violation.

Push the engine-specific channel destruction logic into destroy_context()
and let the hardware-specific code lock and unlock when it's actually
needed. Change the engine destruction order to avoid a race in the small
gap between pgraph and pfifo context uninitialization.

Reported-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:06:35 +10:00

490 lines
14 KiB
C

/*
* Copyright (C) 2007 Ben Skeggs.
* All Rights Reserved.
*
* 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 (including the
* next paragraph) 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 OWNER(S) AND/OR ITS SUPPLIERS 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.
*
*/
#include "drmP.h"
#include "drm.h"
#include "nouveau_drv.h"
#include "nouveau_ramht.h"
static void
nv50_fifo_playlist_update(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
struct nouveau_gpuobj *cur;
int i, nr;
NV_DEBUG(dev, "\n");
cur = pfifo->playlist[pfifo->cur_playlist];
pfifo->cur_playlist = !pfifo->cur_playlist;
/* We never schedule channel 0 or 127 */
for (i = 1, nr = 0; i < 127; i++) {
if (dev_priv->channels.ptr[i] &&
dev_priv->channels.ptr[i]->ramfc) {
nv_wo32(cur, (nr * 4), i);
nr++;
}
}
dev_priv->engine.instmem.flush(dev);
nv_wr32(dev, 0x32f4, cur->vinst >> 12);
nv_wr32(dev, 0x32ec, nr);
nv_wr32(dev, 0x2500, 0x101);
}
static void
nv50_fifo_channel_enable(struct drm_device *dev, int channel)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_channel *chan = dev_priv->channels.ptr[channel];
uint32_t inst;
NV_DEBUG(dev, "ch%d\n", channel);
if (dev_priv->chipset == 0x50)
inst = chan->ramfc->vinst >> 12;
else
inst = chan->ramfc->vinst >> 8;
nv_wr32(dev, NV50_PFIFO_CTX_TABLE(channel), inst |
NV50_PFIFO_CTX_TABLE_CHANNEL_ENABLED);
}
static void
nv50_fifo_channel_disable(struct drm_device *dev, int channel)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
uint32_t inst;
NV_DEBUG(dev, "ch%d\n", channel);
if (dev_priv->chipset == 0x50)
inst = NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G80;
else
inst = NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G84;
nv_wr32(dev, NV50_PFIFO_CTX_TABLE(channel), inst);
}
static void
nv50_fifo_init_reset(struct drm_device *dev)
{
uint32_t pmc_e = NV_PMC_ENABLE_PFIFO;
NV_DEBUG(dev, "\n");
nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) & ~pmc_e);
nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) | pmc_e);
}
static void
nv50_fifo_init_intr(struct drm_device *dev)
{
NV_DEBUG(dev, "\n");
nv_wr32(dev, NV03_PFIFO_INTR_0, 0xFFFFFFFF);
nv_wr32(dev, NV03_PFIFO_INTR_EN_0, 0xFFFFFFFF);
}
static void
nv50_fifo_init_context_table(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
int i;
NV_DEBUG(dev, "\n");
for (i = 0; i < NV50_PFIFO_CTX_TABLE__SIZE; i++) {
if (dev_priv->channels.ptr[i])
nv50_fifo_channel_enable(dev, i);
else
nv50_fifo_channel_disable(dev, i);
}
nv50_fifo_playlist_update(dev);
}
static void
nv50_fifo_init_regs__nv(struct drm_device *dev)
{
NV_DEBUG(dev, "\n");
nv_wr32(dev, 0x250c, 0x6f3cfc34);
}
static void
nv50_fifo_init_regs(struct drm_device *dev)
{
NV_DEBUG(dev, "\n");
nv_wr32(dev, 0x2500, 0);
nv_wr32(dev, 0x3250, 0);
nv_wr32(dev, 0x3220, 0);
nv_wr32(dev, 0x3204, 0);
nv_wr32(dev, 0x3210, 0);
nv_wr32(dev, 0x3270, 0);
/* Enable dummy channels setup by nv50_instmem.c */
nv50_fifo_channel_enable(dev, 0);
nv50_fifo_channel_enable(dev, 127);
}
int
nv50_fifo_init(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
int ret;
NV_DEBUG(dev, "\n");
if (pfifo->playlist[0]) {
pfifo->cur_playlist = !pfifo->cur_playlist;
goto just_reset;
}
ret = nouveau_gpuobj_new(dev, NULL, 128*4, 0x1000,
NVOBJ_FLAG_ZERO_ALLOC,
&pfifo->playlist[0]);
if (ret) {
NV_ERROR(dev, "error creating playlist 0: %d\n", ret);
return ret;
}
ret = nouveau_gpuobj_new(dev, NULL, 128*4, 0x1000,
NVOBJ_FLAG_ZERO_ALLOC,
&pfifo->playlist[1]);
if (ret) {
nouveau_gpuobj_ref(NULL, &pfifo->playlist[0]);
NV_ERROR(dev, "error creating playlist 1: %d\n", ret);
return ret;
}
just_reset:
nv50_fifo_init_reset(dev);
nv50_fifo_init_intr(dev);
nv50_fifo_init_context_table(dev);
nv50_fifo_init_regs__nv(dev);
nv50_fifo_init_regs(dev);
dev_priv->engine.fifo.enable(dev);
dev_priv->engine.fifo.reassign(dev, true);
return 0;
}
void
nv50_fifo_takedown(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
NV_DEBUG(dev, "\n");
if (!pfifo->playlist[0])
return;
nouveau_gpuobj_ref(NULL, &pfifo->playlist[0]);
nouveau_gpuobj_ref(NULL, &pfifo->playlist[1]);
}
int
nv50_fifo_channel_id(struct drm_device *dev)
{
return nv_rd32(dev, NV03_PFIFO_CACHE1_PUSH1) &
NV50_PFIFO_CACHE1_PUSH1_CHID_MASK;
}
int
nv50_fifo_create_context(struct nouveau_channel *chan)
{
struct drm_device *dev = chan->dev;
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_gpuobj *ramfc = NULL;
unsigned long flags;
int ret;
NV_DEBUG(dev, "ch%d\n", chan->id);
if (dev_priv->chipset == 0x50) {
ret = nouveau_gpuobj_new_fake(dev, chan->ramin->pinst,
chan->ramin->vinst, 0x100,
NVOBJ_FLAG_ZERO_ALLOC |
NVOBJ_FLAG_ZERO_FREE,
&chan->ramfc);
if (ret)
return ret;
ret = nouveau_gpuobj_new_fake(dev, chan->ramin->pinst + 0x0400,
chan->ramin->vinst + 0x0400,
4096, 0, &chan->cache);
if (ret)
return ret;
} else {
ret = nouveau_gpuobj_new(dev, chan, 0x100, 256,
NVOBJ_FLAG_ZERO_ALLOC |
NVOBJ_FLAG_ZERO_FREE, &chan->ramfc);
if (ret)
return ret;
ret = nouveau_gpuobj_new(dev, chan, 4096, 1024,
0, &chan->cache);
if (ret)
return ret;
}
ramfc = chan->ramfc;
spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
nv_wo32(ramfc, 0x48, chan->pushbuf->cinst >> 4);
nv_wo32(ramfc, 0x80, ((chan->ramht->bits - 9) << 27) |
(4 << 24) /* SEARCH_FULL */ |
(chan->ramht->gpuobj->cinst >> 4));
nv_wo32(ramfc, 0x44, 0x2101ffff);
nv_wo32(ramfc, 0x60, 0x7fffffff);
nv_wo32(ramfc, 0x40, 0x00000000);
nv_wo32(ramfc, 0x7c, 0x30000001);
nv_wo32(ramfc, 0x78, 0x00000000);
nv_wo32(ramfc, 0x3c, 0x403f6078);
nv_wo32(ramfc, 0x50, chan->pushbuf_base + chan->dma.ib_base * 4);
nv_wo32(ramfc, 0x54, drm_order(chan->dma.ib_max + 1) << 16);
if (dev_priv->chipset != 0x50) {
nv_wo32(chan->ramin, 0, chan->id);
nv_wo32(chan->ramin, 4, chan->ramfc->vinst >> 8);
nv_wo32(ramfc, 0x88, chan->cache->vinst >> 10);
nv_wo32(ramfc, 0x98, chan->ramin->vinst >> 12);
}
dev_priv->engine.instmem.flush(dev);
nv50_fifo_channel_enable(dev, chan->id);
nv50_fifo_playlist_update(dev);
spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
return 0;
}
void
nv50_fifo_destroy_context(struct nouveau_channel *chan)
{
struct drm_device *dev = chan->dev;
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
struct nouveau_gpuobj *ramfc = NULL;
unsigned long flags;
NV_DEBUG(dev, "ch%d\n", chan->id);
spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
pfifo->reassign(dev, false);
/* Unload the context if it's the currently active one */
if (pfifo->channel_id(dev) == chan->id) {
pfifo->disable(dev);
pfifo->unload_context(dev);
pfifo->enable(dev);
}
/* This will ensure the channel is seen as disabled. */
nouveau_gpuobj_ref(chan->ramfc, &ramfc);
nouveau_gpuobj_ref(NULL, &chan->ramfc);
nv50_fifo_channel_disable(dev, chan->id);
/* Dummy channel, also used on ch 127 */
if (chan->id == 0)
nv50_fifo_channel_disable(dev, 127);
nv50_fifo_playlist_update(dev);
pfifo->reassign(dev, true);
spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
/* Free the channel resources */
nouveau_gpuobj_ref(NULL, &ramfc);
nouveau_gpuobj_ref(NULL, &chan->cache);
}
int
nv50_fifo_load_context(struct nouveau_channel *chan)
{
struct drm_device *dev = chan->dev;
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_gpuobj *ramfc = chan->ramfc;
struct nouveau_gpuobj *cache = chan->cache;
int ptr, cnt;
NV_DEBUG(dev, "ch%d\n", chan->id);
nv_wr32(dev, 0x3330, nv_ro32(ramfc, 0x00));
nv_wr32(dev, 0x3334, nv_ro32(ramfc, 0x04));
nv_wr32(dev, 0x3240, nv_ro32(ramfc, 0x08));
nv_wr32(dev, 0x3320, nv_ro32(ramfc, 0x0c));
nv_wr32(dev, 0x3244, nv_ro32(ramfc, 0x10));
nv_wr32(dev, 0x3328, nv_ro32(ramfc, 0x14));
nv_wr32(dev, 0x3368, nv_ro32(ramfc, 0x18));
nv_wr32(dev, 0x336c, nv_ro32(ramfc, 0x1c));
nv_wr32(dev, 0x3370, nv_ro32(ramfc, 0x20));
nv_wr32(dev, 0x3374, nv_ro32(ramfc, 0x24));
nv_wr32(dev, 0x3378, nv_ro32(ramfc, 0x28));
nv_wr32(dev, 0x337c, nv_ro32(ramfc, 0x2c));
nv_wr32(dev, 0x3228, nv_ro32(ramfc, 0x30));
nv_wr32(dev, 0x3364, nv_ro32(ramfc, 0x34));
nv_wr32(dev, 0x32a0, nv_ro32(ramfc, 0x38));
nv_wr32(dev, 0x3224, nv_ro32(ramfc, 0x3c));
nv_wr32(dev, 0x324c, nv_ro32(ramfc, 0x40));
nv_wr32(dev, 0x2044, nv_ro32(ramfc, 0x44));
nv_wr32(dev, 0x322c, nv_ro32(ramfc, 0x48));
nv_wr32(dev, 0x3234, nv_ro32(ramfc, 0x4c));
nv_wr32(dev, 0x3340, nv_ro32(ramfc, 0x50));
nv_wr32(dev, 0x3344, nv_ro32(ramfc, 0x54));
nv_wr32(dev, 0x3280, nv_ro32(ramfc, 0x58));
nv_wr32(dev, 0x3254, nv_ro32(ramfc, 0x5c));
nv_wr32(dev, 0x3260, nv_ro32(ramfc, 0x60));
nv_wr32(dev, 0x3264, nv_ro32(ramfc, 0x64));
nv_wr32(dev, 0x3268, nv_ro32(ramfc, 0x68));
nv_wr32(dev, 0x326c, nv_ro32(ramfc, 0x6c));
nv_wr32(dev, 0x32e4, nv_ro32(ramfc, 0x70));
nv_wr32(dev, 0x3248, nv_ro32(ramfc, 0x74));
nv_wr32(dev, 0x2088, nv_ro32(ramfc, 0x78));
nv_wr32(dev, 0x2058, nv_ro32(ramfc, 0x7c));
nv_wr32(dev, 0x2210, nv_ro32(ramfc, 0x80));
cnt = nv_ro32(ramfc, 0x84);
for (ptr = 0; ptr < cnt; ptr++) {
nv_wr32(dev, NV40_PFIFO_CACHE1_METHOD(ptr),
nv_ro32(cache, (ptr * 8) + 0));
nv_wr32(dev, NV40_PFIFO_CACHE1_DATA(ptr),
nv_ro32(cache, (ptr * 8) + 4));
}
nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, cnt << 2);
nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0);
/* guessing that all the 0x34xx regs aren't on NV50 */
if (dev_priv->chipset != 0x50) {
nv_wr32(dev, 0x340c, nv_ro32(ramfc, 0x88));
nv_wr32(dev, 0x3400, nv_ro32(ramfc, 0x8c));
nv_wr32(dev, 0x3404, nv_ro32(ramfc, 0x90));
nv_wr32(dev, 0x3408, nv_ro32(ramfc, 0x94));
nv_wr32(dev, 0x3410, nv_ro32(ramfc, 0x98));
}
nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, chan->id | (1<<16));
return 0;
}
int
nv50_fifo_unload_context(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
struct nouveau_gpuobj *ramfc, *cache;
struct nouveau_channel *chan = NULL;
int chid, get, put, ptr;
NV_DEBUG(dev, "\n");
chid = pfifo->channel_id(dev);
if (chid < 1 || chid >= dev_priv->engine.fifo.channels - 1)
return 0;
chan = dev_priv->channels.ptr[chid];
if (!chan) {
NV_ERROR(dev, "Inactive channel on PFIFO: %d\n", chid);
return -EINVAL;
}
NV_DEBUG(dev, "ch%d\n", chan->id);
ramfc = chan->ramfc;
cache = chan->cache;
nv_wo32(ramfc, 0x00, nv_rd32(dev, 0x3330));
nv_wo32(ramfc, 0x04, nv_rd32(dev, 0x3334));
nv_wo32(ramfc, 0x08, nv_rd32(dev, 0x3240));
nv_wo32(ramfc, 0x0c, nv_rd32(dev, 0x3320));
nv_wo32(ramfc, 0x10, nv_rd32(dev, 0x3244));
nv_wo32(ramfc, 0x14, nv_rd32(dev, 0x3328));
nv_wo32(ramfc, 0x18, nv_rd32(dev, 0x3368));
nv_wo32(ramfc, 0x1c, nv_rd32(dev, 0x336c));
nv_wo32(ramfc, 0x20, nv_rd32(dev, 0x3370));
nv_wo32(ramfc, 0x24, nv_rd32(dev, 0x3374));
nv_wo32(ramfc, 0x28, nv_rd32(dev, 0x3378));
nv_wo32(ramfc, 0x2c, nv_rd32(dev, 0x337c));
nv_wo32(ramfc, 0x30, nv_rd32(dev, 0x3228));
nv_wo32(ramfc, 0x34, nv_rd32(dev, 0x3364));
nv_wo32(ramfc, 0x38, nv_rd32(dev, 0x32a0));
nv_wo32(ramfc, 0x3c, nv_rd32(dev, 0x3224));
nv_wo32(ramfc, 0x40, nv_rd32(dev, 0x324c));
nv_wo32(ramfc, 0x44, nv_rd32(dev, 0x2044));
nv_wo32(ramfc, 0x48, nv_rd32(dev, 0x322c));
nv_wo32(ramfc, 0x4c, nv_rd32(dev, 0x3234));
nv_wo32(ramfc, 0x50, nv_rd32(dev, 0x3340));
nv_wo32(ramfc, 0x54, nv_rd32(dev, 0x3344));
nv_wo32(ramfc, 0x58, nv_rd32(dev, 0x3280));
nv_wo32(ramfc, 0x5c, nv_rd32(dev, 0x3254));
nv_wo32(ramfc, 0x60, nv_rd32(dev, 0x3260));
nv_wo32(ramfc, 0x64, nv_rd32(dev, 0x3264));
nv_wo32(ramfc, 0x68, nv_rd32(dev, 0x3268));
nv_wo32(ramfc, 0x6c, nv_rd32(dev, 0x326c));
nv_wo32(ramfc, 0x70, nv_rd32(dev, 0x32e4));
nv_wo32(ramfc, 0x74, nv_rd32(dev, 0x3248));
nv_wo32(ramfc, 0x78, nv_rd32(dev, 0x2088));
nv_wo32(ramfc, 0x7c, nv_rd32(dev, 0x2058));
nv_wo32(ramfc, 0x80, nv_rd32(dev, 0x2210));
put = (nv_rd32(dev, NV03_PFIFO_CACHE1_PUT) & 0x7ff) >> 2;
get = (nv_rd32(dev, NV03_PFIFO_CACHE1_GET) & 0x7ff) >> 2;
ptr = 0;
while (put != get) {
nv_wo32(cache, ptr + 0,
nv_rd32(dev, NV40_PFIFO_CACHE1_METHOD(get)));
nv_wo32(cache, ptr + 4,
nv_rd32(dev, NV40_PFIFO_CACHE1_DATA(get)));
get = (get + 1) & 0x1ff;
ptr += 8;
}
/* guessing that all the 0x34xx regs aren't on NV50 */
if (dev_priv->chipset != 0x50) {
nv_wo32(ramfc, 0x84, ptr >> 3);
nv_wo32(ramfc, 0x88, nv_rd32(dev, 0x340c));
nv_wo32(ramfc, 0x8c, nv_rd32(dev, 0x3400));
nv_wo32(ramfc, 0x90, nv_rd32(dev, 0x3404));
nv_wo32(ramfc, 0x94, nv_rd32(dev, 0x3408));
nv_wo32(ramfc, 0x98, nv_rd32(dev, 0x3410));
}
dev_priv->engine.instmem.flush(dev);
/*XXX: probably reload ch127 (NULL) state back too */
nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, 127);
return 0;
}
void
nv50_fifo_tlb_flush(struct drm_device *dev)
{
nv50_vm_flush(dev, 5);
}