There are some cases where it's fine to not hold the cli->mutex, but in those cases we are the only caller and we should really be able to acquire cli->mutex without blocking.
This patch will prevent future occurrences where commands are sent to a channel without proper locking.
Signed-off-by: Maarten Lankhorstmaarten.lankhorst@canonical.com --- diff --git a/drivers/gpu/drm/nouveau/nouveau_abi16.c b/drivers/gpu/drm/nouveau/nouveau_abi16.c index 900fae01793e..20959e445d05 100644 --- a/drivers/gpu/drm/nouveau/nouveau_abi16.c +++ b/drivers/gpu/drm/nouveau/nouveau_abi16.c @@ -153,10 +153,12 @@ nouveau_abi16_fini(struct nouveau_abi16 *abi16) struct nouveau_cli *cli = (void *)abi16->client; struct nouveau_abi16_chan *chan, *temp;
+ mutex_lock(&cli->mutex); /* cleanup channels */ list_for_each_entry_safe(chan, temp, &abi16->channels, head) { nouveau_abi16_chan_fini(abi16, chan); } + mutex_unlock(&cli->mutex);
/* destroy the device object */ nouveau_object_del(abi16->client, NVDRM_CLIENT, NVDRM_DEVICE); diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.h b/drivers/gpu/drm/nouveau/nouveau_dma.h index 984004d66a6d..82465bd0b59b 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dma.h +++ b/drivers/gpu/drm/nouveau/nouveau_dma.h @@ -107,6 +107,7 @@ RING_SPACE(struct nouveau_channel *chan, int size) { int ret;
+ lockdep_assert_held(&chan->cli->mutex); ret = nouveau_dma_wait(chan, 1, size); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.c b/drivers/gpu/drm/nouveau/nouveau_drm.c index f7a6568cfa6a..a60c404d59e4 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drm.c +++ b/drivers/gpu/drm/nouveau/nouveau_drm.c @@ -476,7 +476,9 @@ nouveau_drm_load(struct drm_device *dev, unsigned long flags)
nouveau_sysfs_init(dev); nouveau_hwmon_init(dev); + mutex_lock(&drm->client.mutex); nouveau_accel_init(drm); + mutex_unlock(&drm->client.mutex); nouveau_fbcon_init(dev);
if (nouveau_runtime_pm != 0) { @@ -562,17 +564,17 @@ nouveau_do_suspend(struct drm_device *dev) ttm_bo_evict_mm(&drm->ttm.bdev, TTM_PL_VRAM);
NV_INFO(drm, "waiting for kernel channels to go idle...\n"); - if (drm->cechan) { + ret = 0; + mutex_lock(&drm->client.mutex); + if (drm->cechan) ret = nouveau_channel_idle(drm->cechan); - if (ret) - return ret; - }
- if (drm->channel) { + if (drm->channel && !ret) ret = nouveau_channel_idle(drm->channel); - if (ret) - return ret; - } + mutex_unlock(&drm->client.mutex); + if (ret) + return ret; +
NV_INFO(drm, "suspending client object trees...\n"); if (drm->fence && nouveau_fence(drm)->suspend) { diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c index 7903e0ed3c75..aec87e1d3855 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c +++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c @@ -357,7 +357,7 @@ nouveau_fbcon_create(struct drm_fb_helper *helper,
mutex_unlock(&dev->struct_mutex);
- if (chan) { + if (chan && !WARN_ON(!mutex_trylock(&drm->client.mutex))) { ret = -ENODEV; if (device->card_type < NV_50) ret = nv04_fbcon_accel_init(info); @@ -366,6 +366,7 @@ nouveau_fbcon_create(struct drm_fb_helper *helper, ret = nv50_fbcon_accel_init(info); else ret = nvc0_fbcon_accel_init(info); + mutex_unlock(&drm->client.mutex);
if (ret == 0) info->fbops = &nouveau_fbcon_ops;