[PATCH 36/48] staging: etnaviv: add support for GEM_WAIT ioctl
Lucas Stach
l.stach at pengutronix.de
Fri Sep 25 04:57:48 PDT 2015
From: Russell King <rmk+kernel at arm.linux.org.uk>
Add support for the new GEM_WAIT ioctl, which waits for the kernel to
finish with the GEM object (all render complete, and has been retired).
This is different from the WAIT_FENCE ioctl, which just waits for
rendering to complete.
This is an important distinction, as when a gem object is retired and
freed, the kernel will invalidate the cache for this object, which
effectively changes the data userspace sees. If userspace merely
waits for rendering to complete, there is a race condition where a
userptr BO can be free()d and re-used, and the kernel's invalidation
then corrupts the malloc() free lists.
GEM_WAIT is not fully race free: if the BO is re-submitted via a
different thread, we can't report this (indeed, this could happen
after the call has returned.) It is up to userspace to ensure that
such behaviour does not occur.
Signed-off-by: Russell King <rmk+kernel at arm.linux.org.uk>
---
drivers/staging/etnaviv/etnaviv_drv.c | 29 +++++++++++++
drivers/staging/etnaviv/etnaviv_gem.c | 8 ++++
drivers/staging/etnaviv/etnaviv_gem.h | 2 +
drivers/staging/etnaviv/etnaviv_gpu.c | 76 +++++++++++++++++++++++++++--------
drivers/staging/etnaviv/etnaviv_gpu.h | 8 ++++
include/uapi/drm/etnaviv_drm.h | 10 ++++-
6 files changed, 116 insertions(+), 17 deletions(-)
diff --git a/drivers/staging/etnaviv/etnaviv_drv.c b/drivers/staging/etnaviv/etnaviv_drv.c
index d7133e4450c6..85dbe3813859 100644
--- a/drivers/staging/etnaviv/etnaviv_drv.c
+++ b/drivers/staging/etnaviv/etnaviv_drv.c
@@ -19,6 +19,7 @@
#include "etnaviv_drv.h"
#include "etnaviv_gpu.h"
+#include "etnaviv_gem.h"
#include "etnaviv_mmu.h"
#ifdef CONFIG_DRM_ETNAVIV_REGISTER_LOGGING
@@ -451,6 +452,33 @@ static int etnaviv_ioctl_gem_userptr(struct drm_device *dev, void *data,
&args->handle);
}
+static int etnaviv_ioctl_gem_wait(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct etnaviv_drm_private *priv = dev->dev_private;
+ struct drm_etnaviv_gem_wait *args = data;
+ struct drm_gem_object *obj;
+ struct etnaviv_gpu *gpu;
+ int ret;
+
+ if (args->pipe >= ETNA_MAX_PIPES)
+ return -EINVAL;
+
+ gpu = priv->gpu[args->pipe];
+ if (!gpu)
+ return -ENXIO;
+
+ obj = drm_gem_object_lookup(dev, file, args->handle);
+ if (!obj)
+ return -ENOENT;
+
+ ret = etnaviv_gem_wait_bo(gpu, obj, &TS(args->timeout));
+
+ drm_gem_object_unreference_unlocked(obj);
+
+ return ret;
+}
+
static const struct drm_ioctl_desc etnaviv_ioctls[] = {
#define ETNA_IOCTL(n, func, flags) \
DRM_IOCTL_DEF_DRV(ETNAVIV_##n, etnaviv_ioctl_##func, flags)
@@ -462,6 +490,7 @@ static const struct drm_ioctl_desc etnaviv_ioctls[] = {
ETNA_IOCTL(GEM_SUBMIT, gem_submit, DRM_UNLOCKED|DRM_AUTH|DRM_RENDER_ALLOW),
ETNA_IOCTL(WAIT_FENCE, wait_fence, DRM_UNLOCKED|DRM_AUTH|DRM_RENDER_ALLOW),
ETNA_IOCTL(GEM_USERPTR, gem_userptr, DRM_UNLOCKED|DRM_RENDER_ALLOW),
+ ETNA_IOCTL(GEM_WAIT, gem_wait, DRM_UNLOCKED|DRM_AUTH|DRM_RENDER_ALLOW),
};
static const struct vm_operations_struct vm_ops = {
diff --git a/drivers/staging/etnaviv/etnaviv_gem.c b/drivers/staging/etnaviv/etnaviv_gem.c
index 1c003afc9ab1..2518f897fdd8 100644
--- a/drivers/staging/etnaviv/etnaviv_gem.c
+++ b/drivers/staging/etnaviv/etnaviv_gem.c
@@ -457,6 +457,14 @@ int etnaviv_gem_cpu_fini(struct drm_gem_object *obj)
return 0;
}
+int etnaviv_gem_wait_bo(struct etnaviv_gpu *gpu, struct drm_gem_object *obj,
+ struct timespec *timeout)
+{
+ struct etnaviv_gem_object *etnaviv_obj = to_etnaviv_bo(obj);
+
+ return etnaviv_gpu_wait_obj_inactive(gpu, etnaviv_obj, timeout);
+}
+
#ifdef CONFIG_DEBUG_FS
static void etnaviv_gem_describe(struct drm_gem_object *obj, struct seq_file *m)
{
diff --git a/drivers/staging/etnaviv/etnaviv_gem.h b/drivers/staging/etnaviv/etnaviv_gem.h
index f3136d86d73e..4d5455a3fe39 100644
--- a/drivers/staging/etnaviv/etnaviv_gem.h
+++ b/drivers/staging/etnaviv/etnaviv_gem.h
@@ -130,6 +130,8 @@ struct etnaviv_gem_submit {
} bos[0];
};
+int etnaviv_gem_wait_bo(struct etnaviv_gpu *gpu, struct drm_gem_object *obj,
+ struct timespec *timeout);
struct etnaviv_vram_mapping *
etnaviv_gem_get_vram_mapping(struct etnaviv_gem_object *obj,
struct etnaviv_iommu *mmu);
diff --git a/drivers/staging/etnaviv/etnaviv_gpu.c b/drivers/staging/etnaviv/etnaviv_gpu.c
index 137c09293375..f89fa869885e 100644
--- a/drivers/staging/etnaviv/etnaviv_gpu.c
+++ b/drivers/staging/etnaviv/etnaviv_gpu.c
@@ -743,7 +743,7 @@ static void hangcheck_timer_reset(struct etnaviv_gpu *gpu)
static void hangcheck_handler(unsigned long data)
{
struct etnaviv_gpu *gpu = (struct etnaviv_gpu *)data;
- u32 fence = gpu->retired_fence;
+ u32 fence = gpu->completed_fence;
bool progress = false;
if (fence != gpu->hangcheck_fence) {
@@ -837,7 +837,7 @@ static void retire_worker(struct work_struct *work)
struct etnaviv_gpu *gpu = container_of(work, struct etnaviv_gpu,
retire_work);
struct drm_device *dev = gpu->drm;
- u32 fence = gpu->retired_fence;
+ u32 fence = gpu->completed_fence;
mutex_lock(&dev->struct_mutex);
@@ -860,11 +860,27 @@ static void retire_worker(struct work_struct *work)
}
}
+ gpu->retired_fence = fence;
+
mutex_unlock(&dev->struct_mutex);
wake_up_all(&gpu->fence_event);
}
+static unsigned long etnaviv_timeout_to_jiffies(struct timespec *timeout)
+{
+ unsigned long timeout_jiffies = timespec_to_jiffies(timeout);
+ unsigned long start_jiffies = jiffies;
+ unsigned long remaining_jiffies;
+
+ if (time_after(start_jiffies, timeout_jiffies))
+ remaining_jiffies = 0;
+ else
+ remaining_jiffies = timeout_jiffies - start_jiffies;
+
+ return remaining_jiffies;
+}
+
int etnaviv_gpu_wait_fence_interruptible(struct etnaviv_gpu *gpu,
u32 fence, struct timespec *timeout)
{
@@ -880,21 +896,15 @@ int etnaviv_gpu_wait_fence_interruptible(struct etnaviv_gpu *gpu,
/* No timeout was requested: just test for completion */
ret = fence_completed(gpu, fence) ? 0 : -EBUSY;
} else {
- unsigned long timeout_jiffies = timespec_to_jiffies(timeout);
- unsigned long start_jiffies = jiffies;
- unsigned long remaining_jiffies;
-
- if (time_after(start_jiffies, timeout_jiffies))
- remaining_jiffies = 0;
- else
- remaining_jiffies = timeout_jiffies - start_jiffies;
+ unsigned long remaining = etnaviv_timeout_to_jiffies(timeout);
ret = wait_event_interruptible_timeout(gpu->fence_event,
fence_completed(gpu, fence),
- remaining_jiffies);
+ remaining);
if (ret == 0) {
- DBG("timeout waiting for fence: %u (completed: %u)",
- fence, gpu->retired_fence);
+ DBG("timeout waiting for fence: %u (retired: %u completed: %u)",
+ fence, gpu->retired_fence,
+ gpu->completed_fence);
ret = -ETIMEDOUT;
} else if (ret != -ERESTARTSYS) {
ret = 0;
@@ -904,6 +914,39 @@ int etnaviv_gpu_wait_fence_interruptible(struct etnaviv_gpu *gpu,
return ret;
}
+/*
+ * Wait for an object to become inactive. This, on it's own, is not race
+ * free: the object is moved by the retire worker off the active list, and
+ * then the iova is put. Moreover, the object could be re-submitted just
+ * after we notice that it's become inactive.
+ *
+ * Although the retirement happens under the struct_mutex, we don't want
+ * to hold that lock in this function. Instead, the caller is responsible
+ * for ensuring that the retire worker has finished (which will happen, eg,
+ * when we unreference the object, an action which takes the struct_mutex.)
+ */
+int etnaviv_gpu_wait_obj_inactive(struct etnaviv_gpu *gpu,
+ struct etnaviv_gem_object *etnaviv_obj, struct timespec *timeout)
+{
+ unsigned long remaining;
+ long ret;
+
+ if (!timeout)
+ return !is_active(etnaviv_obj) ? 0 : -EBUSY;
+
+ remaining = etnaviv_timeout_to_jiffies(timeout);
+
+ ret = wait_event_interruptible_timeout(gpu->fence_event,
+ !is_active(etnaviv_obj),
+ remaining);
+ if (ret > 0)
+ return 0;
+ else if (ret == -ERESTARTSYS)
+ return -ERESTARTSYS;
+ else
+ return -ETIMEDOUT;
+}
+
int etnaviv_gpu_pm_get_sync(struct etnaviv_gpu *gpu)
{
return pm_runtime_get_sync(gpu->dev);
@@ -1021,8 +1064,9 @@ static irqreturn_t irq_handler(int irq, void *data)
* - event 1 and event 0 complete
* we can end up processing event 0 first, then 1.
*/
- if (fence_after(gpu->event[event].fence, gpu->retired_fence))
- gpu->retired_fence = gpu->event[event].fence;
+ if (fence_after(gpu->event[event].fence,
+ gpu->completed_fence))
+ gpu->completed_fence = gpu->event[event].fence;
event_free(gpu, event);
/*
@@ -1304,7 +1348,7 @@ static int etnaviv_gpu_rpm_suspend(struct device *dev)
u32 idle, mask;
/* If we have outstanding fences, we're not idle */
- if (gpu->retired_fence != gpu->submitted_fence)
+ if (gpu->completed_fence != gpu->submitted_fence)
return -EBUSY;
/* Check whether the hardware (except FE) is idle */
diff --git a/drivers/staging/etnaviv/etnaviv_gpu.h b/drivers/staging/etnaviv/etnaviv_gpu.h
index b88340302571..374f37f3665f 100644
--- a/drivers/staging/etnaviv/etnaviv_gpu.h
+++ b/drivers/staging/etnaviv/etnaviv_gpu.h
@@ -107,6 +107,7 @@ struct etnaviv_gpu {
/* Fencing support */
u32 submitted_fence;
+ u32 completed_fence;
u32 retired_fence;
wait_queue_head_t fence_event;
@@ -144,6 +145,11 @@ static inline u32 gpu_read(struct etnaviv_gpu *gpu, u32 reg)
static inline bool fence_completed(struct etnaviv_gpu *gpu, u32 fence)
{
+ return fence_after_eq(gpu->completed_fence, fence);
+}
+
+static inline bool fence_retired(struct etnaviv_gpu *gpu, u32 fence)
+{
return fence_after_eq(gpu->retired_fence, fence);
}
@@ -158,6 +164,8 @@ int etnaviv_gpu_debugfs(struct etnaviv_gpu *gpu, struct seq_file *m);
void etnaviv_gpu_retire(struct etnaviv_gpu *gpu);
int etnaviv_gpu_wait_fence_interruptible(struct etnaviv_gpu *gpu,
u32 fence, struct timespec *timeout);
+int etnaviv_gpu_wait_obj_inactive(struct etnaviv_gpu *gpu,
+ struct etnaviv_gem_object *etnaviv_obj, struct timespec *timeout);
int etnaviv_gpu_submit(struct etnaviv_gpu *gpu,
struct etnaviv_gem_submit *submit, struct etnaviv_file_private *ctx);
int etnaviv_gpu_pm_get_sync(struct etnaviv_gpu *gpu);
diff --git a/include/uapi/drm/etnaviv_drm.h b/include/uapi/drm/etnaviv_drm.h
index 0aca6b189e63..7cc749c30970 100644
--- a/include/uapi/drm/etnaviv_drm.h
+++ b/include/uapi/drm/etnaviv_drm.h
@@ -199,6 +199,12 @@ struct drm_etnaviv_gem_userptr {
__u32 handle; /* out, non-zero handle */
};
+struct drm_etnaviv_gem_wait {
+ __u32 pipe; /* in */
+ __u32 handle; /* in, bo to be waited for */
+ struct drm_etnaviv_timespec timeout; /* in */
+};
+
#define DRM_ETNAVIV_GET_PARAM 0x00
/* placeholder:
#define DRM_ETNAVIV_SET_PARAM 0x01
@@ -210,7 +216,8 @@ struct drm_etnaviv_gem_userptr {
#define DRM_ETNAVIV_GEM_SUBMIT 0x06
#define DRM_ETNAVIV_WAIT_FENCE 0x07
#define DRM_ETNAVIV_GEM_USERPTR 0x08
-#define DRM_ETNAVIV_NUM_IOCTLS 0x09
+#define DRM_ETNAVIV_GEM_WAIT 0x09
+#define DRM_ETNAVIV_NUM_IOCTLS 0x0a
#define DRM_IOCTL_ETNAVIV_GET_PARAM DRM_IOWR(DRM_COMMAND_BASE + DRM_ETNAVIV_GET_PARAM, struct drm_etnaviv_param)
#define DRM_IOCTL_ETNAVIV_GEM_NEW DRM_IOWR(DRM_COMMAND_BASE + DRM_ETNAVIV_GEM_NEW, struct drm_etnaviv_gem_new)
@@ -220,5 +227,6 @@ struct drm_etnaviv_gem_userptr {
#define DRM_IOCTL_ETNAVIV_GEM_SUBMIT DRM_IOWR(DRM_COMMAND_BASE + DRM_ETNAVIV_GEM_SUBMIT, struct drm_etnaviv_gem_submit)
#define DRM_IOCTL_ETNAVIV_WAIT_FENCE DRM_IOW(DRM_COMMAND_BASE + DRM_ETNAVIV_WAIT_FENCE, struct drm_etnaviv_wait_fence)
#define DRM_IOCTL_ETNAVIV_GEM_USERPTR DRM_IOWR(DRM_COMMAND_BASE + DRM_ETNAVIV_GEM_USERPTR, struct drm_etnaviv_gem_userptr)
+#define DRM_IOCTL_ETNAVIV_GEM_WAIT DRM_IOW(DRM_COMMAND_BASE + DRM_ETNAVIV_GEM_WAIT, struct drm_etnaviv_gem_wait)
#endif /* __ETNAVIV_DRM_H__ */
--
2.5.1
More information about the dri-devel
mailing list