[Nouveau] [RFC PATCH 4/5] channel timeout using repeated sched timeouts
Konsta Hölttä
kholtta at nvidia.com
Wed Aug 5 04:27:52 PDT 2015
Enable the scheduling timeout error interrupt and set it to a low value
to happen periodically, since it can be missed in HW in certain
conditions. Increment a channel-specific counter in software if the
current channel hasn't advanced. Abort the channel once the timeout
limit is hit (with the periodic granularity). The error notifier is set
to NOUVEAU_GEM_CHANNEL_FIFO_ERROR_IDLE_TIMEOUT when this occurs.
A new KEPLER_SET_CHANNEL_TIMEOUT mthd sets the timeout limit, in
milliseconds. The interrupt granularity is set to 100 ms.
Signed-off-by: Konsta Hölttä <kholtta at nvidia.com>
---
drm/nouveau/include/nvif/class.h | 8 ++++
drm/nouveau/nvkm/engine/fifo/gk104.c | 78 +++++++++++++++++++++++++++++++-----
2 files changed, 75 insertions(+), 11 deletions(-)
diff --git a/drm/nouveau/include/nvif/class.h b/drm/nouveau/include/nvif/class.h
index 381c72d..f9a3647 100644
--- a/drm/nouveau/include/nvif/class.h
+++ b/drm/nouveau/include/nvif/class.h
@@ -620,18 +620,26 @@ struct fermi_a_zbc_depth_v0 {
__u8 format;
__u8 index;
__u8 pad03[5];
__u32 ds;
__u32 l2;
};
#define KEPLER_SET_CHANNEL_PRIORITY 0x42 // XXX
+#define KEPLER_SET_CHANNEL_TIMEOUT 0x43 // XXX
+
struct kepler_set_channel_priority_v0 {
__u8 version;
#define KEPLER_SET_CHANNEL_PRIORITY_LOW 0x00
#define KEPLER_SET_CHANNEL_PRIORITY_MEDIUM 0x01
#define KEPLER_SET_CHANNEL_PRIORITY_HIGH 0x02
__u8 priority;
__u8 pad03[6];
};
+struct kepler_set_channel_timeout_v0 {
+ __u8 version;
+ __u8 pad03[3];
+ __u32 timeout_ms;
+};
+
#endif
diff --git a/drm/nouveau/nvkm/engine/fifo/gk104.c b/drm/nouveau/nvkm/engine/fifo/gk104.c
index 2bab45e..15360a6 100644
--- a/drm/nouveau/nvkm/engine/fifo/gk104.c
+++ b/drm/nouveau/nvkm/engine/fifo/gk104.c
@@ -83,18 +83,25 @@ struct gk104_fifo_base {
struct gk104_fifo_chan {
struct nvkm_fifo_chan base;
u32 engine;
enum {
STOPPED,
RUNNING,
KILLED
} state;
+ struct {
+ u32 sum_ms;
+ u32 limit_ms;
+ u32 gpfifo_get;
+ } timeout;
};
+#define GRFIFO_TIMEOUT_CHECK_PERIOD_MS 100
+
/*******************************************************************************
* FIFO channel objects
******************************************************************************/
static void
gk104_fifo_runlist_update(struct gk104_fifo_priv *priv, u32 engine)
{
struct nvkm_bar *bar = nvkm_bar(priv);
@@ -288,16 +295,21 @@ gk104_fifo_chan_ctor(struct nvkm_object *parent, struct nvkm_object *engine,
nv_wo32(base, 0x94, 0x30000001);
nv_wo32(base, 0x9c, 0x00000100);
nv_wo32(base, 0xac, 0x0000001f);
nv_wo32(base, 0xe8, chan->base.chid);
nv_wo32(base, 0xb8, 0xf8000000);
nv_wo32(base, 0xf8, 0x10003080); /* 0x002310 */
nv_wo32(base, 0xfc, 0x10000010); /* 0x002350 */
bar->flush(bar);
+
+ chan->timeout.sum_ms = 0;
+ chan->timeout.limit_ms = -1;
+ chan->timeout.gpfifo_get = 0;
+
return 0;
}
static int
gk104_fifo_chan_init(struct nvkm_object *object)
{
struct nvkm_gpuobj *base = nv_gpuobj(object->parent);
struct gk104_fifo_priv *priv = (void *)object->engine;
@@ -379,21 +391,39 @@ gk104_fifo_chan_set_priority(struct nvkm_object *object, void *data, u32 size)
return -EINVAL;
}
}
return ret;
}
int
+gk104_fifo_chan_set_timeout(struct nvkm_object *object, void *data, u32 size)
+{
+ struct gk104_fifo_chan *chan = (void *)object;
+ union {
+ struct kepler_set_channel_timeout_v0 v0;
+ } *args = data;
+ int ret;
+
+ if (nvif_unpack(args->v0, 0, 0, false)) {
+ chan->timeout.limit_ms = args->v0.timeout_ms;
+ }
+
+ return ret;
+}
+
+int
gk104_fifo_chan_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size)
{
switch (mthd) {
case KEPLER_SET_CHANNEL_PRIORITY:
return gk104_fifo_chan_set_priority(object, data, size);
+ case KEPLER_SET_CHANNEL_TIMEOUT:
+ return gk104_fifo_chan_set_timeout(object, data, size);
default:
break;
}
return -EINVAL;
}
struct nvkm_ofuncs
gk104_fifo_chan_ofuncs = {
@@ -604,61 +634,83 @@ gk104_fifo_intr_bind(struct gk104_fifo_priv *priv)
}
static const struct nvkm_enum
gk104_fifo_sched_reason[] = {
{ 0x0a, "CTXSW_TIMEOUT" },
{}
};
+static bool
+gk104_fifo_update_timeout(struct gk104_fifo_priv *priv,
+ struct gk104_fifo_chan *chan, u32 dt)
+{
+ u32 gpfifo_get = nv_rd32(priv, 34);
+ if (gpfifo_get == chan->timeout.gpfifo_get) {
+ chan->timeout.sum_ms += dt;
+ } else {
+ chan->timeout.sum_ms = dt;
+ }
+
+ chan->timeout.gpfifo_get = gpfifo_get;
+
+ return chan->timeout.sum_ms > chan->timeout.limit_ms;
+}
+
static void
gk104_fifo_intr_sched_ctxsw(struct gk104_fifo_priv *priv)
{
struct nvkm_engine *engine;
struct gk104_fifo_chan *chan;
u32 engn;
for (engn = 0; engn < ARRAY_SIZE(fifo_engine); engn++) {
u32 stat = nv_rd32(priv, 0x002640 + (engn * 0x04));
u32 busy = (stat & 0x80000000);
u32 next = (stat & 0x07ff0000) >> 16;
- u32 chsw = (stat & 0x00008000);
- u32 save = (stat & 0x00004000);
- u32 load = (stat & 0x00002000);
+ u32 cxsw = (stat & 0x0000e000) >> 13;
u32 prev = (stat & 0x000007ff);
- u32 chid = load ? next : prev;
- (void)save;
+ /* if loading context, take next id */
+ u32 chid = cxsw == 5 ? next : prev;
- if (busy && chsw) {
+ nv_error(priv, "ctxsw eng stat: %08x\n", stat);
+ /* doing context switch? */
+ if (busy && (cxsw >= 5 && cxsw <= 7)) {
if (!(chan = (void *)priv->base.channel[chid]))
continue;
if (!(engine = gk104_fifo_engine(priv, engn)))
continue;
- nvkm_fifo_eevent(&priv->base, chid,
- NOUVEAU_GEM_CHANNEL_FIFO_ERROR_IDLE_TIMEOUT);
-
- gk104_fifo_recover(priv, engine, chan);
+ if (gk104_fifo_update_timeout(priv, chan,
+ GRFIFO_TIMEOUT_CHECK_PERIOD_MS)) {
+ nvkm_fifo_eevent(&priv->base, chid,
+ NOUVEAU_GEM_CHANNEL_FIFO_ERROR_IDLE_TIMEOUT);
+ gk104_fifo_recover(priv, engine, chan);
+ } else {
+ nv_debug(priv, "fifo waiting for ctxsw %d ms on ch %d\n",
+ chan->timeout.sum_ms, chid);
+ }
}
}
}
static void
gk104_fifo_intr_sched(struct gk104_fifo_priv *priv)
{
u32 intr = nv_rd32(priv, 0x00254c);
u32 code = intr & 0x000000ff;
const struct nvkm_enum *en;
char enunk[6] = "";
en = nvkm_enum_find(gk104_fifo_sched_reason, code);
if (!en)
snprintf(enunk, sizeof(enunk), "UNK%02x", code);
- nv_error(priv, "SCHED_ERROR [ %s ]\n", en ? en->name : enunk);
+ /* this is a normal situation, not so loud */
+ nv_debug(priv, "SCHED_ERROR [ %s ]\n", en ? en->name : enunk);
switch (code) {
case 0x0a:
gk104_fifo_intr_sched_ctxsw(priv);
break;
default:
break;
}
@@ -1131,18 +1183,22 @@ gk104_fifo_init(struct nvkm_object *object)
/* PBDMA[n].HCE */
for (i = 0; i < priv->spoon_nr; i++) {
nv_wr32(priv, 0x040148 + (i * 0x2000), 0xffffffff); /* INTR */
nv_wr32(priv, 0x04014c + (i * 0x2000), 0xffffffff); /* INTREN */
}
nv_wr32(priv, 0x002254, 0x10000000 | priv->user.bar.offset >> 12);
+ /* enable interrupts */
nv_wr32(priv, 0x002100, 0xffffffff);
nv_wr32(priv, 0x002140, 0x7fffffff);
+
+ /* engine timeout */
+ nv_wr32(priv, 0x002a0c, 0x80000000 | (1000 * GRFIFO_TIMEOUT_CHECK_PERIOD_MS));
return 0;
}
void
gk104_fifo_dtor(struct nvkm_object *object)
{
struct gk104_fifo_priv *priv = (void *)object;
int i;
--
2.1.4
More information about the Nouveau
mailing list