drm/nouveau/fifo/gf100-: protect channel preempt with subdev mutex

commit b27add13f500469127afdf011dbcc9c649e16e54 upstream.

This avoids an issue that occurs when we're attempting to preempt multiple
channels simultaneously.  HW seems to ignore preempt requests while it's
still processing a previous one, which, well, makes sense.

Fixes random "fifo: SCHED_ERROR 0d []" + GPCCS page faults during parallel
piglit runs on (at least) GM107.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c
index e7cbc13..89976ff 100644 (file)
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c
@@ -59,6 +59,7 @@ gf100_fifo_gpfifo_engine_fini(struct nvkm_fifo_chan *base,
        struct nvkm_gpuobj *inst = chan->base.inst;
        int ret = 0;
 
+       mutex_lock(&subdev->mutex);
        nvkm_wr32(device, 0x002634, chan->base.chid);
        if (nvkm_msec(device, 2000,
                if (nvkm_rd32(device, 0x002634) == chan->base.chid)
@@ -66,10 +67,12 @@ gf100_fifo_gpfifo_engine_fini(struct nvkm_fifo_chan *base,
        ) < 0) {
                nvkm_error(subdev, "channel %d [%s] kick timeout\n",
                           chan->base.chid, chan->base.object.client->name);
-               ret = -EBUSY;
-               if (suspend)
-                       return ret;
+               ret = -ETIMEDOUT;
        }
+       mutex_unlock(&subdev->mutex);
+
+       if (ret && suspend)
+               return ret;
 
        if (offset) {
                nvkm_kmap(inst);

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c
index 0b81754..aa1692e 100644 (file)
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c
@@ -39,7 +39,9 @@ gk104_fifo_gpfifo_kick(struct gk104_fifo_chan *chan)
        struct nvkm_subdev *subdev = &fifo->base.engine.subdev;
        struct nvkm_device *device = subdev->device;
        struct nvkm_client *client = chan->base.object.client;
+       int ret = 0;
 
+       mutex_lock(&subdev->mutex);
        nvkm_wr32(device, 0x002634, chan->base.chid);
        if (nvkm_msec(device, 2000,
                if (!(nvkm_rd32(device, 0x002634) & 0x00100000))
@@ -47,10 +49,10 @@ gk104_fifo_gpfifo_kick(struct gk104_fifo_chan *chan)
        ) < 0) {
                nvkm_error(subdev, "channel %d [%s] kick timeout\n",
                           chan->base.chid, client->name);
-               return -EBUSY;
+               ret = -ETIMEDOUT;
        }
-
-       return 0;
+       mutex_unlock(&subdev->mutex);
+       return ret;
 }
 
 static u32