Skip to contentSkip to conversation

drm/msm for 4.15 (resend)

PATCH New Jordan Crouse 2 participants 10 messages Oct 20, 2017 1 message not in the archive

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

v2 cover drm/msm for 4.15 (resend)

(Resending for more visibility)

Here are the refreshed submitqueue/ringbuffer/preemption changes for 4.15. These are the original changes with bug fixes, improvements and suggestions squashed in:

- Moved SUBMITQUEUE_CLOSE param to a u32 instead of reusing the struct
- Changed to use per-ring fence contexts
- Squashed in Rob's change to the driver version
- Squshed in fix for preemption race condition

These apply against drm-msm-next-2017-08-22 from ~robclark/linux.

Jordan Crouse (9):
drm/msm: Add per-instance submit queues
drm/msm: Move memptrs to msm_gpu
drm/msm: Support multiple ringbuffers
drm/msm: Add a parameter query for the number of ringbuffers drm/msm: Shadow current pointer in the ring until command is complete drm/msm: Make the value of RB_CNTL (almost) generic drm/msm: Implement preemption for A5XX targets
drm/msm: Removed unused struct_mutex_task
drm/msm: dump a rd GPUADDR header for all buffers in the command

 drivers/gpu/drm/msm/Makefile              |   4 +-
 drivers/gpu/drm/msm/adreno/a3xx_gpu.c     |  10 +-
 drivers/gpu/drm/msm/adreno/a4xx_gpu.c     |  10 +-
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c     | 222 +++++++++++++++++++---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.h     | 109 ++++++++++-
 drivers/gpu/drm/msm/adreno/a5xx_power.c   |   6 +-
 drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 305 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/msm/adreno/adreno_gpu.c   | 183 +++++++++---------
 drivers/gpu/drm/msm/adreno/adreno_gpu.h   |  43 ++---
 drivers/gpu/drm/msm/msm_drv.c             |  71 ++++++-
 drivers/gpu/drm/msm/msm_drv.h             |  28 +--
 drivers/gpu/drm/msm/msm_fence.c           |   2 +-
 drivers/gpu/drm/msm/msm_fence.h           |   2 +-
 drivers/gpu/drm/msm/msm_gem.h             |   5 +-
 drivers/gpu/drm/msm/msm_gem_submit.c      |  26 ++-
 drivers/gpu/drm/msm/msm_gpu.c             | 174 ++++++++++++-----
 drivers/gpu/drm/msm/msm_gpu.h             |  51 +++--
 drivers/gpu/drm/msm/msm_rd.c              |  30 +--
 drivers/gpu/drm/msm/msm_ringbuffer.c      |  37 +++-
 drivers/gpu/drm/msm/msm_ringbuffer.h      |  33 +++-
 drivers/gpu/drm/msm/msm_submitqueue.c     | 152 +++++++++++++++
 include/uapi/drm/msm_drm.h                |  24 +++
 22 files changed, 1249 insertions(+), 278 deletions(-)
 create mode 100644 drivers/gpu/drm/msm/adreno/a5xx_preempt.c
 create mode 100644 drivers/gpu/drm/msm/msm_submitqueue.c
Signature · 1 line
1.9.1
#1 lore raw

PATCH v2 1/9 drm/msm: Add per-instance submit queues

Currently the behavior of a command stream is provided by the user application during submission and the application is expected to internally maintain the settings for each 'context' or 'rendering queue' and specify the correct ones.

This works okay for simple cases but as applications become more complex we will want to set context specific flags and do various permission checks to allow certain contexts to enable additional privileges.

Add kernel-side submit queues to be analogous to 'contexts' or 'rendering queues' on the application side. Each file descriptor instance will maintain its own list of queues. Queues cannot be shared between file descriptors.

For backwards compatibility context id '0' is defined as a default context specifying no priority and no special flags. This is intended to be the usual configuration for 99% of applications so that a garden variety application can function correctly without creating a queue. Only those applications requiring the specific benefit of different queues need create one.

---
 drivers/gpu/drm/msm/Makefile          |   3 +-
 drivers/gpu/drm/msm/msm_drv.c         |  56 ++++++++++++--
 drivers/gpu/drm/msm/msm_drv.h         |  20 +++--
 drivers/gpu/drm/msm/msm_gem.h         |   1 +
 drivers/gpu/drm/msm/msm_gem_submit.c  |  14 +++-
 drivers/gpu/drm/msm/msm_gpu.h         |  15 ++++
 drivers/gpu/drm/msm/msm_submitqueue.c | 135 ++++++++++++++++++++++++++++++++++
 include/uapi/drm/msm_drm.h            |  22 ++++++
 8 files changed, 250 insertions(+), 16 deletions(-)
 create mode 100644 drivers/gpu/drm/msm/msm_submitqueue.c
drivers/gpu/drm/msm/Makefile +2 -1
index 33008fa..3c234e7 100644
@@ -57,7 +57,8 @@ msm-y := \
 	msm_iommu.o \
 	msm_perf.o \
 	msm_rd.o \
-	msm_ringbuffer.o
+	msm_ringbuffer.o \
+	msm_submitqueue.o
 
 msm-$(CONFIG_DRM_FBDEV_EMULATION) += msm_fbdev.o
 msm-$(CONFIG_COMMON_CLK) += mdp/mdp4/mdp4_lvds_pll.o
drivers/gpu/drm/msm/msm_drv.c +48 -8

100 lines not on this page · raw message

drivers/gpu/drm/msm/msm_drv.h +15 -5
index 5e8109c..b3b8f20 100644
@@ -56,11 +56,9 @@
 struct msm_gem_vma;
 
 struct msm_file_private {
-	/* currently we don't do anything useful with this.. but when
-	 * per-context address spaces are supported we'd keep track of
-	 * the context's page-tables here.
-	 */
-	int dummy;
+	rwlock_t queuelock;
+	struct list_head submitqueues;
+	int queueid;
 };
 
 enum msm_mdp_plane_property {
@@ -319,6 +317,18 @@ void __iomem *msm_ioremap(struct platform_device *pdev, const char *name,
 void msm_writel(u32 data, void __iomem *addr);
 u32 msm_readl(const void __iomem *addr);
 
+struct msm_gpu_submitqueue;
+int msm_submitqueue_init(struct msm_file_private *ctx);
+struct msm_gpu_submitqueue *msm_submitqueue_get(struct msm_file_private *ctx,
+		u32 id);
+int msm_submitqueue_create(struct msm_file_private *ctx, u32 prio,
+		u32 flags, u32 *id);
+int msm_submitqueue_remove(struct msm_file_private *ctx, u32 id);
+void msm_submitqueue_close(struct msm_file_private *ctx);
+
+void msm_submitqueue_destroy(struct kref *kref);
+
+
 #define DBG(fmt, ...) DRM_DEBUG_DRIVER(fmt"\n", ##__VA_ARGS__)
 #define VERB(fmt, ...) if (0) DRM_DEBUG_DRIVER(fmt"\n", ##__VA_ARGS__)
 
drivers/gpu/drm/msm/msm_gem.h +1 -0
index 91c210d..17f8a6c 100644
@@ -142,6 +142,7 @@ struct msm_gem_submit {
 	struct list_head bo_list;
 	struct ww_acquire_ctx ticket;
 	struct dma_fence *fence;
+	struct msm_gpu_submitqueue *queue;
 	struct pid *pid;    /* submitting process */
 	bool valid;         /* true if no cmdstream patching needed */
 	unsigned int nr_cmds;
drivers/gpu/drm/msm/msm_gem_submit.c +12 -2
index 8a75c0b..0973884 100644
@@ -31,7 +31,8 @@
 #define BO_PINNED   0x2000
 
 static struct msm_gem_submit *submit_create(struct drm_device *dev,
-		struct msm_gpu *gpu, uint32_t nr_bos, uint32_t nr_cmds)
+		struct msm_gpu *gpu, struct msm_gpu_submitqueue *queue,
+		uint32_t nr_bos, uint32_t nr_cmds)
 {
 	struct msm_gem_submit *submit;
 	uint64_t sz = sizeof(*submit) + ((u64)nr_bos * sizeof(submit->bos[0])) +
@@ -49,6 +50,7 @@ static struct msm_gem_submit *submit_create(struct drm_device *dev,
 	submit->fence = NULL;
 	submit->pid = get_pid(task_pid(current));
 	submit->cmd = (void *)&submit->bos[nr_bos];
+	submit->queue = queue;
 
 	/* initially, until copy_from_user() and bo lookup succeeds: */
 	submit->nr_bos = 0;
@@ -66,6 +68,8 @@ void msm_gem_submit_free(struct msm_gem_submit *submit)
 	dma_fence_put(submit->fence);
 	list_del(&submit->node);
 	put_pid(submit->pid);
+	msm_submitqueue_put(submit->queue);
+
 	kfree(submit);
 }
 
@@ -391,6 +395,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	struct msm_gpu *gpu = priv->gpu;
 	struct dma_fence *in_fence = NULL;
 	struct sync_file *sync_file = NULL;
+	struct msm_gpu_submitqueue *queue;
 	int out_fence_fd = -1;
 	unsigned i;
 	int ret;
@@ -407,6 +412,10 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	if (MSM_PIPE_FLAGS(args->flags) & ~MSM_SUBMIT_FLAGS)
 		return -EINVAL;
 
+	queue = msm_submitqueue_get(ctx, args->queueid);
+	if (!queue)
+		return -ENOENT;
+
 	if (args->flags & MSM_SUBMIT_FENCE_FD_IN) {
 		in_fence = sync_file_get_fence(args->fence_fd);
 
@@ -437,7 +446,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	}
 	priv->struct_mutex_task = current;
 
-	submit = submit_create(dev, gpu, args->nr_bos, args->nr_cmds);
+	submit = submit_create(dev, gpu, queue, args->nr_bos, args->nr_cmds);
 	if (!submit) {
 		ret = -ENOMEM;
 		goto out_unlock;
@@ -523,6 +532,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	submit->nr_cmds = i;
 
 	submit->fence = msm_fence_alloc(gpu->fctx);
+
 	if (IS_ERR(submit->fence)) {
 		ret = PTR_ERR(submit->fence);
 		submit->fence = NULL;
drivers/gpu/drm/msm/msm_gpu.h +15 -0
index df4e277..a890176 100644
@@ -150,6 +150,15 @@ struct msm_gpu_perfcntr {
 	const char *name;
 };
 
+struct msm_gpu_submitqueue {
+	int id;
+	u32 flags;
+	u32 prio;
+	int faults;
+	struct list_head node;
+	struct kref ref;
+};
+
 static inline void gpu_write(struct msm_gpu *gpu, u32 reg, u32 data)
 {
 	msm_writel(data, gpu->mmio + (reg << 2));
@@ -223,4 +232,10 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 void __init adreno_register(void);
 void __exit adreno_unregister(void);
 
+static inline void msm_submitqueue_put(struct msm_gpu_submitqueue *queue)
+{
+	if (queue)
+		kref_put(&queue->ref, msm_submitqueue_destroy);
+}
+
 #endif /* __MSM_GPU_H__ */
drivers/gpu/drm/msm/msm_submitqueue.c new file +135 -0

135 lines not on this page · raw message

include/uapi/drm/msm_drm.h +22 -0
index ad4eb28..9fb4b81 100644
@@ -218,6 +218,7 @@ struct drm_msm_gem_submit {
 	__u64 bos;            /* in, ptr to array of submit_bo's */
 	__u64 cmds;           /* in, ptr to array of submit_cmd's */
 	__s32 fence_fd;       /* in/out fence fd (see MSM_SUBMIT_FENCE_FD_IN/OUT) */
+	__u32 queueid;         /* in, submitqueue id */
 };
 
 /* The normal way to synchronize with the GPU is just to CPU_PREP on
@@ -254,6 +255,20 @@ struct drm_msm_gem_madvise {
 	__u32 retained;       /* out, whether backing store still exists */
 };
 
+/*
+ * Draw queues allow the user to set specific submission parameter. Command
+ * submissions specify a specific submitqueue to use.  ID 0 is reserved for
+ * backwards compatibility as a "default" submitqueue
+ */
+
+#define MSM_SUBMITQUEUE_FLAGS (0)
+
+struct drm_msm_submitqueue {
+	__u32 flags;   /* in, MSM_SUBMITQUEUE_x */
+	__u32 prio;    /* in, Priority level */
+	__u32 id;      /* out, identifier */
+};
+
 #define DRM_MSM_GET_PARAM              0x00
 /* placeholder:
 #define DRM_MSM_SET_PARAM              0x01
@@ -265,6 +280,11 @@ struct drm_msm_gem_madvise {
 #define DRM_MSM_GEM_SUBMIT             0x06
 #define DRM_MSM_WAIT_FENCE             0x07
 #define DRM_MSM_GEM_MADVISE            0x08
+/* placeholder:
+#define DRM_MSM_GEM_SVM_NEW              0x09
+ */
+#define DRM_MSM_SUBMITQUEUE_NEW          0x0A
+#define DRM_MSM_SUBMITQUEUE_CLOSE        0x0B
 
 #define DRM_IOCTL_MSM_GET_PARAM        DRM_IOWR(DRM_COMMAND_BASE + DRM_MSM_GET_PARAM, struct drm_msm_param)
 #define DRM_IOCTL_MSM_GEM_NEW          DRM_IOWR(DRM_COMMAND_BASE + DRM_MSM_GEM_NEW, struct drm_msm_gem_new)
@@ -274,6 +294,8 @@ struct drm_msm_gem_madvise {
 #define DRM_IOCTL_MSM_GEM_SUBMIT       DRM_IOWR(DRM_COMMAND_BASE + DRM_MSM_GEM_SUBMIT, struct drm_msm_gem_submit)
 #define DRM_IOCTL_MSM_WAIT_FENCE       DRM_IOW (DRM_COMMAND_BASE + DRM_MSM_WAIT_FENCE, struct drm_msm_wait_fence)
 #define DRM_IOCTL_MSM_GEM_MADVISE      DRM_IOWR(DRM_COMMAND_BASE + DRM_MSM_GEM_MADVISE, struct drm_msm_gem_madvise)
+#define DRM_IOCTL_MSM_SUBMITQUEUE_NEW    DRM_IOWR(DRM_COMMAND_BASE + DRM_MSM_SUBMITQUEUE_NEW, struct drm_msm_submitqueue)
+#define DRM_IOCTL_MSM_SUBMITQUEUE_CLOSE  DRM_IOW (DRM_COMMAND_BASE + DRM_MSM_SUBMITQUEUE_CLOSE, __u32)
 
 #if defined(__cplusplus)
 }
Signature · 6 lines
1.9.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#2 lore raw

PATCH v2 2/9 drm/msm: Move memptrs to msm_gpu

In reply to Jordan Crouse

When we move to multiple ringbuffers we're going to store the data in the memptrs on a per-ring basis. In order to prepare for that move the current memptrs from the adreno namespace into msm_gpu. This is way cleaner and immediately lets us kill off some sub functions so there is much less cost later when we do move to per-ring structs.

---
 drivers/gpu/drm/msm/adreno/a3xx_gpu.c   |  1 -
 drivers/gpu/drm/msm/adreno/a4xx_gpu.c   |  1 -
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c   |  8 ++---
 drivers/gpu/drm/msm/adreno/adreno_gpu.c | 52 ++++++++-------------------------
 drivers/gpu/drm/msm/adreno/adreno_gpu.h | 16 ----------
 drivers/gpu/drm/msm/msm_gpu.c           | 30 +++++++++++++++++--
 drivers/gpu/drm/msm/msm_gpu.h           | 17 +++++++++--
 7 files changed, 57 insertions(+), 68 deletions(-)
drivers/gpu/drm/msm/adreno/a3xx_gpu.c +0 -1
index 7791313..789f7fb 100644
@@ -444,7 +444,6 @@ static void a3xx_dump(struct msm_gpu *gpu)
 		.pm_suspend = msm_gpu_pm_suspend,
 		.pm_resume = msm_gpu_pm_resume,
 		.recover = a3xx_recover,
-		.last_fence = adreno_last_fence,
 		.submit = adreno_submit,
 		.flush = adreno_flush,
 		.irq = a3xx_irq,
drivers/gpu/drm/msm/adreno/a4xx_gpu.c +0 -1
index 58341ef..f87c4312 100644
@@ -532,7 +532,6 @@ static int a4xx_get_timestamp(struct msm_gpu *gpu, uint64_t *value)
 		.pm_suspend = a4xx_pm_suspend,
 		.pm_resume = a4xx_pm_resume,
 		.recover = a4xx_recover,
-		.last_fence = adreno_last_fence,
 		.submit = adreno_submit,
 		.flush = adreno_flush,
 		.irq = a4xx_irq,
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +3 -5
index 17c59d8..32a5b55 100644
@@ -98,7 +98,6 @@ static int zap_shader_load_mdt(struct device *dev, const char *fwname)
 static void a5xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 	struct msm_file_private *ctx)
 {
-	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
 	struct msm_drm_private *priv = gpu->dev->dev_private;
 	struct msm_ringbuffer *ring = gpu->rb;
 	unsigned int i, ibs = 0;
@@ -125,8 +124,8 @@ static void a5xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
 	OUT_PKT7(ring, CP_EVENT_WRITE, 4);
 	OUT_RING(ring, CACHE_FLUSH_TS | (1 << 31));
-	OUT_RING(ring, lower_32_bits(rbmemptr(adreno_gpu, fence)));
-	OUT_RING(ring, upper_32_bits(rbmemptr(adreno_gpu, fence)));
+	OUT_RING(ring, lower_32_bits(rbmemptr(gpu, fence)));
+	OUT_RING(ring, upper_32_bits(rbmemptr(gpu, fence)));
 	OUT_RING(ring, submit->fence->seqno);
 
 	gpu->funcs->flush(gpu);
@@ -804,7 +803,7 @@ static void a5xx_fault_detect_irq(struct msm_gpu *gpu)
 	struct msm_drm_private *priv = dev->dev_private;
 
 	dev_err(dev->dev, "gpu fault fence %x status %8.8X rb %4.4x/%4.4x ib1 %16.16llX/%4.4x ib2 %16.16llX/%4.4x\n",
-		gpu->funcs->last_fence(gpu),
+		gpu->memptrs->fence,
 		gpu_read(gpu, REG_A5XX_RBBM_STATUS),
 		gpu_read(gpu, REG_A5XX_CP_RB_RPTR),
 		gpu_read(gpu, REG_A5XX_CP_RB_WPTR),
@@ -992,7 +991,6 @@ static void a5xx_show(struct msm_gpu *gpu, struct seq_file *m)
 		.pm_suspend = a5xx_pm_suspend,
 		.pm_resume = a5xx_pm_resume,
 		.recover = a5xx_recover,
-		.last_fence = adreno_last_fence,
 		.submit = a5xx_submit,
 		.flush = adreno_flush,
 		.irq = a5xx_irq,
drivers/gpu/drm/msm/adreno/adreno_gpu.c +12 -40

106 lines not on this page · raw message

drivers/gpu/drm/msm/adreno/adreno_gpu.h +0 -16
index 4d9165f..5ee1d6a 100644
@@ -82,14 +82,6 @@ struct adreno_info {
 
 const struct adreno_info *adreno_info(struct adreno_rev rev);
 
-#define rbmemptr(adreno_gpu, member)  \
-	((adreno_gpu)->memptrs_iova + offsetof(struct adreno_rbmemptrs, member))
-
-struct adreno_rbmemptrs {
-	volatile uint32_t rptr;
-	volatile uint32_t fence;
-};
-
 struct adreno_gpu {
 	struct msm_gpu base;
 	struct adreno_rev rev;
@@ -104,13 +96,6 @@ struct adreno_gpu {
 	/* firmware: */
 	const struct firmware *pm4, *pfp;
 
-	/* ringbuffer rptr/wptr: */
-	// TODO should this be in msm_ringbuffer?  I think it would be
-	// different for z180..
-	struct adreno_rbmemptrs *memptrs;
-	struct drm_gem_object *memptrs_bo;
-	uint64_t memptrs_iova;
-
 	/*
 	 * Register offsets are different between some GPUs.
 	 * GPU specific offsets will be exported by GPU specific
@@ -197,7 +182,6 @@ static inline int adreno_is_a530(struct adreno_gpu *gpu)
 
 int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value);
 int adreno_hw_init(struct msm_gpu *gpu);
-uint32_t adreno_last_fence(struct msm_gpu *gpu);
 void adreno_recover(struct msm_gpu *gpu);
 void adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 		struct msm_file_private *ctx);
drivers/gpu/drm/msm/msm_gpu.c +27 -3
index ffbff27..9e645ca 100644
@@ -228,7 +228,7 @@ static void recover_worker(struct work_struct *work)
 	struct msm_gpu *gpu = container_of(work, struct msm_gpu, recover_work);
 	struct drm_device *dev = gpu->dev;
 	struct msm_gem_submit *submit;
-	uint32_t fence = gpu->funcs->last_fence(gpu);
+	uint32_t fence = gpu->memptrs->fence;
 
 	msm_update_fence(gpu->fctx, fence + 1);
 
@@ -281,7 +281,7 @@ static void hangcheck_handler(unsigned long data)
 	struct msm_gpu *gpu = (struct msm_gpu *)data;
 	struct drm_device *dev = gpu->dev;
 	struct msm_drm_private *priv = dev->dev_private;
-	uint32_t fence = gpu->funcs->last_fence(gpu);
+	uint32_t fence = gpu->memptrs->fence;
 
 	if (fence != gpu->hangcheck_fence) {
 		/* some progress has been made.. ya! */
@@ -449,7 +449,7 @@ static void retire_worker(struct work_struct *work)
 {
 	struct msm_gpu *gpu = container_of(work, struct msm_gpu, retire_work);
 	struct drm_device *dev = gpu->dev;
-	uint32_t fence = gpu->funcs->last_fence(gpu);
+	uint32_t fence = gpu->memptrs->fence;
 
 	msm_update_fence(gpu->fctx, fence);
 
@@ -689,6 +689,17 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		goto fail;
 	}
 
+	gpu->memptrs = msm_gem_kernel_new(drm, sizeof(*gpu->memptrs_bo),
+		MSM_BO_UNCACHED, gpu->aspace, &gpu->memptrs_bo,
+		&gpu->memptrs_iova);
+
+	if (IS_ERR(gpu->memptrs)) {
+		ret = PTR_ERR(gpu->memptrs);
+		gpu->memptrs = NULL;
+		dev_err(drm->dev, "could not allocate memptrs: %d\n", ret);
+		goto fail;
+	}
+
 	/* Create ringbuffer: */
 	gpu->rb = msm_ringbuffer_new(gpu, config->ringsz);
 	if (IS_ERR(gpu->rb)) {
@@ -701,6 +712,12 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	return 0;
 
 fail:
+	if (gpu->memptrs_bo) {
+		msm_gem_put_vaddr(gpu->memptrs_bo);
+		msm_gem_put_iova(gpu->memptrs_bo, gpu->aspace);
+		drm_gem_object_unreference_unlocked(gpu->memptrs_bo);
+	}
+
 	platform_set_drvdata(pdev, NULL);
 	return ret;
 }
@@ -718,6 +735,13 @@ void msm_gpu_cleanup(struct msm_gpu *gpu)
 			msm_gem_put_iova(gpu->rb->bo, gpu->aspace);
 		msm_ringbuffer_destroy(gpu->rb);
 	}
+
+	if (gpu->memptrs_bo) {
+		msm_gem_put_vaddr(gpu->memptrs_bo);
+		msm_gem_put_iova(gpu->memptrs_bo, gpu->aspace);
+		drm_gem_object_unreference_unlocked(gpu->memptrs_bo);
+	}
+
 	if (gpu->aspace) {
 		gpu->aspace->mmu->funcs->detach(gpu->aspace->mmu,
 			NULL, 0);
drivers/gpu/drm/msm/msm_gpu.h +15 -2
index a890176..8ddda05 100644
@@ -59,7 +59,6 @@ struct msm_gpu_funcs {
 			struct msm_file_private *ctx);
 	void (*flush)(struct msm_gpu *gpu);
 	irqreturn_t (*irq)(struct msm_gpu *irq);
-	uint32_t (*last_fence)(struct msm_gpu *gpu);
 	void (*recover)(struct msm_gpu *gpu);
 	void (*destroy)(struct msm_gpu *gpu);
 #ifdef CONFIG_DEBUG_FS
@@ -68,6 +67,14 @@ struct msm_gpu_funcs {
 #endif
 };
 
+#define rbmemptr(gpu, member)  \
+	((gpu)->memptrs_iova + offsetof(struct msm_rbmemptrs, member))
+
+struct msm_rbmemptrs {
+	volatile uint32_t rptr;
+	volatile uint32_t fence;
+};
+
 struct msm_gpu {
 	const char *name;
 	struct drm_device *dev;
@@ -130,11 +137,17 @@ struct msm_gpu {
 	struct work_struct recover_work;
 
 	struct list_head submit_list;
+
+	struct msm_rbmemptrs *memptrs;
+	struct drm_gem_object *memptrs_bo;
+	uint64_t memptrs_iova;
+
+
 };
 
 static inline bool msm_gpu_active(struct msm_gpu *gpu)
 {
-	return gpu->fctx->last_fence > gpu->funcs->last_fence(gpu);
+	return gpu->fctx->last_fence > gpu->memptrs->fence;
 }
 
 /* Perf-Counters:
Signature · 1 line
1.9.1
#4 lore raw

PATCH 3/9 drm/msm: Support multiple ringbuffers

In reply to a message not in the archive

Add the infrastructure to support the idea of multiple ringbuffers. Assign each ringbuffer an id and use that as an index for the various ring specific operations.

The biggest delta is to support legacy fences. Each fence gets its own sequence number but the legacy functions expect to use a unique integer. To handle this we return a unique identifier for each submission but map it to a specific ring/sequence under the covers. Newer users use a dma_fence pointer anyway so they don't care about the actual sequence ID or ring.

The actual mechanics for multiple ringbuffers are very target specific so this code just allows for the possibility but still only defines one ringbuffer for each target family.

---
 drivers/gpu/drm/msm/adreno/a3xx_gpu.c   |   9 +-
 drivers/gpu/drm/msm/adreno/a4xx_gpu.c   |   9 +-
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c   |  54 ++++++-----
 drivers/gpu/drm/msm/adreno/a5xx_gpu.h   |   2 +-
 drivers/gpu/drm/msm/adreno/a5xx_power.c |   6 +-
 drivers/gpu/drm/msm/adreno/adreno_gpu.c | 137 ++++++++++++++++-----------
 drivers/gpu/drm/msm/adreno/adreno_gpu.h |  20 ++--
 drivers/gpu/drm/msm/msm_drv.c           |  23 +++--
 drivers/gpu/drm/msm/msm_drv.h           |   8 +-
 drivers/gpu/drm/msm/msm_fence.c         |   2 +-
 drivers/gpu/drm/msm/msm_fence.h         |   2 +-
 drivers/gpu/drm/msm/msm_gem.h           |   4 +-
 drivers/gpu/drm/msm/msm_gem_submit.c    |  12 ++-
 drivers/gpu/drm/msm/msm_gpu.c           | 163 ++++++++++++++++++++++----------
 drivers/gpu/drm/msm/msm_gpu.h           |  42 ++++----
 drivers/gpu/drm/msm/msm_ringbuffer.c    |  35 +++++--
 drivers/gpu/drm/msm/msm_ringbuffer.h    |  20 +++-
 drivers/gpu/drm/msm/msm_submitqueue.c   |  27 +++++-
 include/uapi/drm/msm_drm.h              |   1 +
 19 files changed, 366 insertions(+), 210 deletions(-)
drivers/gpu/drm/msm/adreno/a3xx_gpu.c +5 -4
index 789f7fb..4baef27 100644
@@ -44,7 +44,7 @@
 
 static bool a3xx_me_init(struct msm_gpu *gpu)
 {
-	struct msm_ringbuffer *ring = gpu->rb;
+	struct msm_ringbuffer *ring = gpu->rb[0];
 
 	OUT_PKT3(ring, CP_ME_INIT, 17);
 	OUT_RING(ring, 0x000003f7);
@@ -65,7 +65,7 @@ static bool a3xx_me_init(struct msm_gpu *gpu)
 	OUT_RING(ring, 0x00000000);
 	OUT_RING(ring, 0x00000000);
 
-	gpu->funcs->flush(gpu);
+	gpu->funcs->flush(gpu, ring);
 	return a3xx_idle(gpu);
 }
 
@@ -339,7 +339,7 @@ static void a3xx_destroy(struct msm_gpu *gpu)
 static bool a3xx_idle(struct msm_gpu *gpu)
 {
 	/* wait for ringbuffer to drain: */
-	if (!adreno_idle(gpu))
+	if (!adreno_idle(gpu, gpu->rb[0]))
 		return false;
 
 	/* then wait for GPU to finish: */
@@ -446,6 +446,7 @@ static void a3xx_dump(struct msm_gpu *gpu)
 		.recover = a3xx_recover,
 		.submit = adreno_submit,
 		.flush = adreno_flush,
+		.active_ring = adreno_active_ring,
 		.irq = a3xx_irq,
 		.destroy = a3xx_destroy,
 #ifdef CONFIG_DEBUG_FS
@@ -491,7 +492,7 @@ struct msm_gpu *a3xx_gpu_init(struct drm_device *dev)
 	adreno_gpu->registers = a3xx_registers;
 	adreno_gpu->reg_offsets = a3xx_register_offsets;
 
-	ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs);
+	ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs, 1);
 	if (ret)
 		goto fail;
 
drivers/gpu/drm/msm/adreno/a4xx_gpu.c +5 -4
index f87c4312..8199a4b 100644
@@ -116,7 +116,7 @@ static void a4xx_enable_hwcg(struct msm_gpu *gpu)
 
 static bool a4xx_me_init(struct msm_gpu *gpu)
 {
-	struct msm_ringbuffer *ring = gpu->rb;
+	struct msm_ringbuffer *ring = gpu->rb[0];
 
 	OUT_PKT3(ring, CP_ME_INIT, 17);
 	OUT_RING(ring, 0x000003f7);
@@ -137,7 +137,7 @@ static bool a4xx_me_init(struct msm_gpu *gpu)
 	OUT_RING(ring, 0x00000000);
 	OUT_RING(ring, 0x00000000);
 
-	gpu->funcs->flush(gpu);
+	gpu->funcs->flush(gpu, ring);
 	return a4xx_idle(gpu);
 }
 
@@ -337,7 +337,7 @@ static void a4xx_destroy(struct msm_gpu *gpu)
 static bool a4xx_idle(struct msm_gpu *gpu)
 {
 	/* wait for ringbuffer to drain: */
-	if (!adreno_idle(gpu))
+	if (!adreno_idle(gpu, gpu->rb[0]))
 		return false;
 
 	/* then wait for GPU to finish: */
@@ -534,6 +534,7 @@ static int a4xx_get_timestamp(struct msm_gpu *gpu, uint64_t *value)
 		.recover = a4xx_recover,
 		.submit = adreno_submit,
 		.flush = adreno_flush,
+		.active_ring = adreno_active_ring,
 		.irq = a4xx_irq,
 		.destroy = a4xx_destroy,
 #ifdef CONFIG_DEBUG_FS
@@ -573,7 +574,7 @@ struct msm_gpu *a4xx_gpu_init(struct drm_device *dev)
 	adreno_gpu->registers = a4xx_registers;
 	adreno_gpu->reg_offsets = a4xx_register_offsets;
 
-	ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs);
+	ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs, 1);
 	if (ret)
 		goto fail;
 
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +28 -26

128 lines not on this page · raw message

drivers/gpu/drm/msm/adreno/a5xx_gpu.h +1 -1
index e944516..44db48d 100644
@@ -55,7 +55,7 @@ static inline int spin_usecs(struct msm_gpu *gpu, uint32_t usecs,
 	return -ETIMEDOUT;
 }
 
-bool a5xx_idle(struct msm_gpu *gpu);
+bool a5xx_idle(struct msm_gpu *gpu, struct msm_ringbuffer *ring);
 void a5xx_set_hwcg(struct msm_gpu *gpu, bool state);
 
 #endif /* __A5XX_GPU_H__ */
drivers/gpu/drm/msm/adreno/a5xx_power.c +3 -3
index 04aab1d..d09a716 100644
@@ -173,7 +173,7 @@ static int a5xx_gpmu_init(struct msm_gpu *gpu)
 {
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
 	struct a5xx_gpu *a5xx_gpu = to_a5xx_gpu(adreno_gpu);
-	struct msm_ringbuffer *ring = gpu->rb;
+	struct msm_ringbuffer *ring = gpu->rb[0];
 
 	if (!a5xx_gpu->gpmu_dwords)
 		return 0;
@@ -192,9 +192,9 @@ static int a5xx_gpmu_init(struct msm_gpu *gpu)
 	OUT_PKT7(ring, CP_SET_PROTECTED_MODE, 1);
 	OUT_RING(ring, 1);
 
-	gpu->funcs->flush(gpu);
+	gpu->funcs->flush(gpu, ring);
 
-	if (!a5xx_idle(gpu)) {
+	if (!a5xx_idle(gpu, ring)) {
 		DRM_ERROR("%s: Unable to load GPMU firmware. GPMU will not be active\n",
 			gpu->name);
 		return -EINVAL;
drivers/gpu/drm/msm/adreno/adreno_gpu.c +83 -54

255 lines not on this page · raw message

drivers/gpu/drm/msm/adreno/adreno_gpu.h +11 -9
index 5ee1d6a..2a51ac4 100644
@@ -185,17 +185,19 @@ static inline int adreno_is_a530(struct adreno_gpu *gpu)
 void adreno_recover(struct msm_gpu *gpu);
 void adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 		struct msm_file_private *ctx);
-void adreno_flush(struct msm_gpu *gpu);
-bool adreno_idle(struct msm_gpu *gpu);
+void adreno_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring);
+bool adreno_idle(struct msm_gpu *gpu, struct msm_ringbuffer *ring);
 #ifdef CONFIG_DEBUG_FS
 void adreno_show(struct msm_gpu *gpu, struct seq_file *m);
 #endif
 void adreno_dump_info(struct msm_gpu *gpu);
 void adreno_dump(struct msm_gpu *gpu);
-void adreno_wait_ring(struct msm_gpu *gpu, uint32_t ndwords);
+void adreno_wait_ring(struct msm_ringbuffer *ring, uint32_t ndwords);
+struct msm_ringbuffer *adreno_active_ring(struct msm_gpu *gpu);
 
 int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
-		struct adreno_gpu *gpu, const struct adreno_gpu_funcs *funcs);
+		struct adreno_gpu *gpu, const struct adreno_gpu_funcs *funcs,
+		int nr_rings);
 void adreno_gpu_cleanup(struct adreno_gpu *gpu);
 
 
@@ -204,7 +206,7 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 static inline void
 OUT_PKT0(struct msm_ringbuffer *ring, uint16_t regindx, uint16_t cnt)
 {
-	adreno_wait_ring(ring->gpu, cnt+1);
+	adreno_wait_ring(ring, cnt+1);
 	OUT_RING(ring, CP_TYPE0_PKT | ((cnt-1) << 16) | (regindx & 0x7FFF));
 }
 
@@ -212,14 +214,14 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 static inline void
 OUT_PKT2(struct msm_ringbuffer *ring)
 {
-	adreno_wait_ring(ring->gpu, 1);
+	adreno_wait_ring(ring, 1);
 	OUT_RING(ring, CP_TYPE2_PKT);
 }
 
 static inline void
 OUT_PKT3(struct msm_ringbuffer *ring, uint8_t opcode, uint16_t cnt)
 {
-	adreno_wait_ring(ring->gpu, cnt+1);
+	adreno_wait_ring(ring, cnt+1);
 	OUT_RING(ring, CP_TYPE3_PKT | ((cnt-1) << 16) | ((opcode & 0xFF) << 8));
 }
 
@@ -241,14 +243,14 @@ static inline u32 PM4_PARITY(u32 val)
 static inline void
 OUT_PKT4(struct msm_ringbuffer *ring, uint16_t regindx, uint16_t cnt)
 {
-	adreno_wait_ring(ring->gpu, cnt + 1);
+	adreno_wait_ring(ring, cnt + 1);
 	OUT_RING(ring, PKT4(regindx, cnt));
 }
 
 static inline void
 OUT_PKT7(struct msm_ringbuffer *ring, uint8_t opcode, uint16_t cnt)
 {
-	adreno_wait_ring(ring->gpu, cnt + 1);
+	adreno_wait_ring(ring, cnt + 1);
 	OUT_RING(ring, CP_TYPE7_PKT | (cnt << 0) | (PM4_PARITY(cnt) << 15) |
 		((opcode & 0x7F) << 16) | (PM4_PARITY(opcode) << 23));
 }
drivers/gpu/drm/msm/msm_drv.c +17 -6
index cf743c7..45f1170 100644
@@ -517,7 +517,7 @@ static void load_gpu(struct drm_device *dev)
 	mutex_unlock(&init_lock);
 }
 
-static int context_init(struct drm_file *file)
+static int context_init(struct drm_device *dev, struct drm_file *file)
 {
 	struct msm_file_private *ctx;
 
@@ -525,7 +525,7 @@ static int context_init(struct drm_file *file)
 	if (!ctx)
 		return -ENOMEM;
 
-	msm_submitqueue_init(ctx);
+	msm_submitqueue_init(dev, ctx);
 
 	file->driver_priv = ctx;
 
@@ -539,7 +539,7 @@ static int msm_open(struct drm_device *dev, struct drm_file *file)
 	 */
 	load_gpu(dev);
 
-	return context_init(file);
+	return context_init(dev, file);
 }
 
 static void context_close(struct msm_file_private *ctx)
@@ -753,16 +753,27 @@ static int msm_ioctl_wait_fence(struct drm_device *dev, void *data,
 	struct msm_drm_private *priv = dev->dev_private;
 	struct drm_msm_wait_fence *args = data;
 	ktime_t timeout = to_ktime(args->timeout);
+	struct msm_gpu_submitqueue *queue;
+	struct msm_gpu *gpu = priv->gpu;
+	int ret;
 
 	if (args->pad) {
 		DRM_ERROR("invalid pad: %08x\n", args->pad);
 		return -EINVAL;
 	}
 
-	if (!priv->gpu)
+	if (!gpu)
 		return 0;
 
-	return msm_wait_fence(priv->gpu->fctx, args->fence, &timeout, true);
+	queue = msm_submitqueue_get(file->driver_priv, args->queueid);
+	if (!queue)
+		return -ENOENT;
+
+	ret = msm_wait_fence(gpu->rb[queue->prio]->fctx, args->fence, &timeout,
+		true);
+
+	msm_submitqueue_put(queue);
+	return ret;
 }
 
 static int msm_ioctl_gem_madvise(struct drm_device *dev, void *data,
@@ -812,7 +823,7 @@ static int msm_ioctl_submitqueue_new(struct drm_device *dev, void *data,
 	if (args->flags & ~MSM_SUBMITQUEUE_FLAGS)
 		return -EINVAL;
 
-	return msm_submitqueue_create(file->driver_priv, args->prio,
+	return msm_submitqueue_create(dev, file->driver_priv, args->prio,
 		args->flags, &args->id);
 }
 
drivers/gpu/drm/msm/msm_drv.h +5 -3
index b3b8f20..2821f57 100644
@@ -74,6 +74,8 @@ struct msm_vblank_ctrl {
 	spinlock_t lock;
 };
 
+#define MSM_GPU_MAX_RINGS 1
+
 struct msm_drm_private {
 
 	struct drm_device *dev;
@@ -318,11 +320,11 @@ void __iomem *msm_ioremap(struct platform_device *pdev, const char *name,
 u32 msm_readl(const void __iomem *addr);
 
 struct msm_gpu_submitqueue;
-int msm_submitqueue_init(struct msm_file_private *ctx);
+int msm_submitqueue_init(struct drm_device *drm, struct msm_file_private *ctx);
 struct msm_gpu_submitqueue *msm_submitqueue_get(struct msm_file_private *ctx,
 		u32 id);
-int msm_submitqueue_create(struct msm_file_private *ctx, u32 prio,
-		u32 flags, u32 *id);
+int msm_submitqueue_create(struct drm_device *drm, struct msm_file_private *ctx,
+		u32 prio, u32 flags, u32 *id);
 int msm_submitqueue_remove(struct msm_file_private *ctx, u32 id);
 void msm_submitqueue_close(struct msm_file_private *ctx);
 
drivers/gpu/drm/msm/msm_fence.c +1 -1
index a2f89ba..349c12f 100644
@@ -31,7 +31,7 @@ struct msm_fence_context *
 		return ERR_PTR(-ENOMEM);
 
 	fctx->dev = dev;
-	fctx->name = name;
+	strncpy(fctx->name, name, sizeof(fctx->name));
 	fctx->context = dma_fence_context_alloc(1);
 	init_waitqueue_head(&fctx->event);
 	spin_lock_init(&fctx->spinlock);
drivers/gpu/drm/msm/msm_fence.h +1 -1
index 56061aa..1aa6a4c 100644
@@ -22,7 +22,7 @@
 
 struct msm_fence_context {
 	struct drm_device *dev;
-	const char *name;
+	char name[32];
 	unsigned context;
 	/* last_fence == completed_fence --> no pending work */
 	uint32_t last_fence;          /* last assigned fence */
drivers/gpu/drm/msm/msm_gem.h +3 -1
index 17f8a6c..9320e18 100644
@@ -138,13 +138,15 @@ enum msm_gem_lock {
 struct msm_gem_submit {
 	struct drm_device *dev;
 	struct msm_gpu *gpu;
-	struct list_head node;   /* node in gpu submit_list */
+	struct list_head node;   /* node in ring submit list */
 	struct list_head bo_list;
 	struct ww_acquire_ctx ticket;
+	uint32_t seqno;		/* Sequence number of the submit on the ring */
 	struct dma_fence *fence;
 	struct msm_gpu_submitqueue *queue;
 	struct pid *pid;    /* submitting process */
 	bool valid;         /* true if no cmdstream patching needed */
+	struct msm_ringbuffer *ring;
 	unsigned int nr_cmds;
 	unsigned int nr_bos;
 	struct {
drivers/gpu/drm/msm/msm_gem_submit.c +8 -4
index 0973884..9e087cf 100644
@@ -51,6 +51,7 @@ static struct msm_gem_submit *submit_create(struct drm_device *dev,
 	submit->pid = get_pid(task_pid(current));
 	submit->cmd = (void *)&submit->bos[nr_bos];
 	submit->queue = queue;
+	submit->ring = gpu->rb[queue->prio];
 
 	/* initially, until copy_from_user() and bo lookup succeeds: */
 	submit->nr_bos = 0;
@@ -233,7 +234,8 @@ static int submit_fence_sync(struct msm_gem_submit *submit)
 		struct msm_gem_object *msm_obj = submit->bos[i].obj;
 		bool write = submit->bos[i].flags & MSM_SUBMIT_BO_WRITE;
 
-		ret = msm_gem_sync_object(&msm_obj->base, submit->gpu->fctx, write);
+		ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,
+			write);
 		if (ret)
 			break;
 	}
@@ -396,6 +398,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	struct dma_fence *in_fence = NULL;
 	struct sync_file *sync_file = NULL;
 	struct msm_gpu_submitqueue *queue;
+	struct msm_ringbuffer *ring;
 	int out_fence_fd = -1;
 	unsigned i;
 	int ret;
@@ -416,6 +419,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	if (!queue)
 		return -ENOENT;
 
+	ring = gpu->rb[queue->prio];
+
 	if (args->flags & MSM_SUBMIT_FENCE_FD_IN) {
 		in_fence = sync_file_get_fence(args->fence_fd);
 
@@ -426,7 +431,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 		 * Wait if the fence is from a foreign context, or if the fence
 		 * array contains any fence from a foreign context.
 		 */
-		if (!dma_fence_match_context(in_fence, gpu->fctx->context)) {
+		if (!dma_fence_match_context(in_fence, ring->fctx->context)) {
 			ret = dma_fence_wait(in_fence, true);
 			if (ret)
 				return ret;
@@ -531,8 +536,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 
 	submit->nr_cmds = i;
 
-	submit->fence = msm_fence_alloc(gpu->fctx);
-
+	submit->fence = msm_fence_alloc(ring->fctx);
 	if (IS_ERR(submit->fence)) {
 		ret = PTR_ERR(submit->fence);
 		submit->fence = NULL;
drivers/gpu/drm/msm/msm_gpu.c +111 -52

277 lines not on this page · raw message

drivers/gpu/drm/msm/msm_gpu.h +18 -24

84 lines not on this page · raw message

drivers/gpu/drm/msm/msm_ringbuffer.c +25 -10
index bf065a5..febf2541 100644
@@ -18,13 +18,15 @@
 #include "msm_ringbuffer.h"
 #include "msm_gpu.h"
 
-struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size)
+struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
+		void *memptrs, uint64_t memptrs_iova)
 {
 	struct msm_ringbuffer *ring;
+	char name[32];
 	int ret;
 
-	if (WARN_ON(!is_power_of_2(size)))
-		return ERR_PTR(-EINVAL);
+	/* We assume everwhere that MSM_GPU_RINGBUFFER_SZ is a power of 2 */
+	BUILD_BUG_ON(!is_power_of_2(MSM_GPU_RINGBUFFER_SZ));
 
 	ring = kzalloc(sizeof(*ring), GFP_KERNEL);
 	if (!ring) {
@@ -33,32 +35,45 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size)
 	}
 
 	ring->gpu = gpu;
-
+	ring->id = id;
 	/* Pass NULL for the iova pointer - we will map it later */
-	ring->start = msm_gem_kernel_new(gpu->dev, size, MSM_BO_WC,
-		gpu->aspace, &ring->bo, NULL);
+	ring->start = msm_gem_kernel_new(gpu->dev, MSM_GPU_RINGBUFFER_SZ,
+		MSM_BO_WC, gpu->aspace, &ring->bo, NULL);
 
 	if (IS_ERR(ring->start)) {
 		ret = PTR_ERR(ring->start);
 		ring->start = 0;
 		goto fail;
 	}
-	ring->end   = ring->start + (size / 4);
+	ring->end   = ring->start + (MSM_GPU_RINGBUFFER_SZ >> 2);
 	ring->cur   = ring->start;
 
-	ring->size = size;
+	ring->memptrs = memptrs;
+	ring->memptrs_iova = memptrs_iova;
+
+	INIT_LIST_HEAD(&ring->submits);
+
+	snprintf(name, sizeof(name), "gpu-ring-%d", ring->id);
+
+	/* FIXME: need struct drm_device */
+	ring->fctx = msm_fence_context_alloc(gpu->dev, name);
 
 	return ring;
 
 fail:
-	if (ring)
-		msm_ringbuffer_destroy(ring);
+	msm_ringbuffer_destroy(ring);
 	return ERR_PTR(ret);
 }
 
 void msm_ringbuffer_destroy(struct msm_ringbuffer *ring)
 {
+	if (IS_ERR_OR_NULL(ring))
+		return;
+
+	msm_fence_context_free(ring->fctx);
+
 	if (ring->bo) {
+		msm_gem_put_iova(ring->bo, ring->gpu->aspace);
 		msm_gem_put_vaddr(ring->bo);
 		drm_gem_object_unreference_unlocked(ring->bo);
 	}
drivers/gpu/drm/msm/msm_ringbuffer.h +18 -2
index 6e0e104..ec44251 100644
@@ -20,14 +20,30 @@
 
 #include "msm_drv.h"
 
+#define rbmemptr(ring, member)  \
+	((ring)->memptrs_iova + offsetof(struct msm_rbmemptrs, member))
+
+struct msm_rbmemptrs {
+	volatile uint32_t rptr;
+	volatile uint32_t fence;
+};
+
 struct msm_ringbuffer {
 	struct msm_gpu *gpu;
-	int size;
+	int id;
 	struct drm_gem_object *bo;
 	uint32_t *start, *end, *cur;
+	struct list_head submits;
+	uint64_t iova;
+	uint32_t seqno;
+	uint32_t hangcheck_fence;
+	struct msm_rbmemptrs *memptrs;
+	uint64_t memptrs_iova;
+	struct msm_fence_context *fctx;
 };
 
-struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size);
+struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
+		void *memptrs, uint64_t memptrs_iova);
 void msm_ringbuffer_destroy(struct msm_ringbuffer *ring);
 
 /* ringbuffer helpers (the parts that are same for a3xx/a2xx/z180..) */
drivers/gpu/drm/msm/msm_submitqueue.c +22 -5
index 593c3b5..5115f75 100644
@@ -60,9 +60,10 @@ void msm_submitqueue_close(struct msm_file_private *ctx)
 		msm_submitqueue_put(entry);
 }
 
-int msm_submitqueue_create(struct msm_file_private *ctx, u32 prio, u32 flags,
-		u32 *id)
+int msm_submitqueue_create(struct drm_device *drm, struct msm_file_private *ctx,
+		u32 prio, u32 flags, u32 *id)
 {
+	struct msm_drm_private *priv = drm->dev_private;
 	struct msm_gpu_submitqueue *queue;
 
 	if (!ctx)
@@ -75,7 +76,13 @@ int msm_submitqueue_create(struct msm_file_private *ctx, u32 prio, u32 flags,
 
 	kref_init(&queue->ref);
 	queue->flags = flags;
-	queue->prio = prio;
+
+	if (priv->gpu) {
+		if (prio >= priv->gpu->nr_rings)
+			return -EINVAL;
+
+		queue->prio = prio;
+	}
 
 	write_lock(&ctx->queuelock);
 
@@ -91,16 +98,26 @@ int msm_submitqueue_create(struct msm_file_private *ctx, u32 prio, u32 flags,
 	return 0;
 }
 
-int msm_submitqueue_init(struct msm_file_private *ctx)
+int msm_submitqueue_init(struct drm_device *drm, struct msm_file_private *ctx)
 {
+	struct msm_drm_private *priv = drm->dev_private;
+	int default_prio;
+
 	if (!ctx)
 		return 0;
 
+	/*
+	 * Select priority 2 as the "default priority" unless nr_rings is less
+	 * than 2 and then pick the lowest pirority
+	 */
+	default_prio = priv->gpu ?
+		clamp_t(uint32_t, 2, 0, priv->gpu->nr_rings - 1) : 0;
+
 	INIT_LIST_HEAD(&ctx->submitqueues);
 
 	rwlock_init(&ctx->queuelock);
 
-	return msm_submitqueue_create(ctx, 2, 0, NULL);
+	return msm_submitqueue_create(drm, ctx, default_prio, 0, NULL);
 }
 
 int msm_submitqueue_remove(struct msm_file_private *ctx, u32 id)
include/uapi/drm/msm_drm.h +1 -0
index 9fb4b81..42fa781 100644
@@ -232,6 +232,7 @@ struct drm_msm_wait_fence {
 	__u32 fence;          /* in */
 	__u32 pad;
 	struct drm_msm_timespec timeout;   /* in */
+	__u32 queueid;         /* in, submitqueue id */
 };
 
 /* madvise provides a way to tell the kernel in case a buffers contents
Signature · 6 lines
1.9.1

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno
#8 lore raw

PATCH v2 4/9 drm/msm: Add a parameter query for the number of ringbuffers

In reply to Jordan Crouse

In order to manage ringbuffer priority to its fullest userspace should know how many ringbuffers it has to work with. Add a parameter to return the number of active rings.

---
 drivers/gpu/drm/msm/adreno/adreno_gpu.c | 3 +++
 include/uapi/drm/msm_drm.h              | 1 +
 2 files changed, 4 insertions(+)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +3 -0
index fad946946..cf91840 100644
@@ -57,6 +57,9 @@ int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value)
 			return ret;
 		}
 		return -EINVAL;
+	case MSM_PARAM_NR_RINGS:
+		*value = gpu->nr_rings;
+		return 0;
 	default:
 		DBG("%s: invalid param: %u", gpu->name, param);
 		return -EINVAL;
include/uapi/drm/msm_drm.h +1 -0
index 42fa781..6363d94 100644
@@ -73,6 +73,7 @@ struct drm_msm_timespec {
 #define MSM_PARAM_MAX_FREQ   0x04
 #define MSM_PARAM_TIMESTAMP  0x05
 #define MSM_PARAM_GMEM_BASE  0x06
+#define MSM_PARAM_NR_RINGS   0x07
 
 struct drm_msm_param {
 	__u32 pipe;           /* in, MSM_PIPE_x */
Signature · 1 line
1.9.1
#5 lore raw

PATCH 5/9 drm/msm: Shadow current pointer in the ring until command is complete

In reply to a message not in the archive

Add a shadow pointer to track the current command being written into the ring. Don't commit it as 'cur' until the command is submitted. Because 'cur' is used to construct the software copy of the wptr this ensures that somebody peeking in on the ring doesn't assume that a command is inflight while it is being written. This isn't a huge deal with a single ring (though technically the hangcheck could assume the system is prematurely busy when it isn't) but it will be rather important for preemption where the decision to preempt is based on a non-empty ringbuffer. Without a shadow an aggressive preemption scheme could assume that the ringbuffer is non empty and switch to it before the CPU is done writing the command and boom.

Even though preemption won't be supported for all targets because of the way the code is organized it is simpler to make this generic for all targets. The extra load for non-preemption targets should be minimal.

---
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  9 +++++++--
 drivers/gpu/drm/msm/msm_ringbuffer.c    |  1 +
 drivers/gpu/drm/msm/msm_ringbuffer.h    | 12 ++++++++----
 3 files changed, 16 insertions(+), 6 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +7 -2
index cf91840..77eaa46 100644
@@ -89,6 +89,7 @@ int adreno_hw_init(struct msm_gpu *gpu)
 		}
 
 		ring->cur = ring->start;
+		ring->next = ring->start;
 
 		/* reset completed fence seqno: */
 		ring->memptrs->fence = ring->seqno;
@@ -233,12 +234,15 @@ void adreno_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring)
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
 	uint32_t wptr;
 
+	/* Copy the shadow to the actual register */
+	ring->cur = ring->next;
+
 	/*
 	 * Mask wptr value that we calculate to fit in the HW range. This is
 	 * to account for the possibility that the last command fit exactly into
 	 * the ringbuffer and rb->next hasn't wrapped to zero yet
 	 */
-	wptr = get_wptr(ring) % (MSM_GPU_RINGBUFFER_SZ >> 2);
+	wptr = (ring->cur - ring->start) % (MSM_GPU_RINGBUFFER_SZ >> 2);
 
 	/* ensure writes to ringbuffer have hit system memory: */
 	mb();
@@ -350,7 +354,8 @@ static uint32_t ring_freewords(struct msm_ringbuffer *ring)
 {
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(ring->gpu);
 	uint32_t size = MSM_GPU_RINGBUFFER_SZ >> 2;
-	uint32_t wptr = get_wptr(ring);
+	/* Use ring->next to calculate free size */
+	uint32_t wptr = ring->next - ring->start;
 	uint32_t rptr = get_rptr(adreno_gpu, ring);
 	return (rptr + (size - 1) - wptr) % size;
 }
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -0
index febf2541..6015959 100644
@@ -46,6 +46,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
 		goto fail;
 	}
 	ring->end   = ring->start + (MSM_GPU_RINGBUFFER_SZ >> 2);
+	ring->next  = ring->start;
 	ring->cur   = ring->start;
 
 	ring->memptrs = memptrs;
drivers/gpu/drm/msm/msm_ringbuffer.h +8 -4
index ec44251..3749764 100644
@@ -32,7 +32,7 @@ struct msm_ringbuffer {
 	struct msm_gpu *gpu;
 	int id;
 	struct drm_gem_object *bo;
-	uint32_t *start, *end, *cur;
+	uint32_t *start, *end, *cur, *next;
 	struct list_head submits;
 	uint64_t iova;
 	uint32_t seqno;
@@ -51,9 +51,13 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
 static inline void
 OUT_RING(struct msm_ringbuffer *ring, uint32_t data)
 {
-	if (ring->cur == ring->end)
-		ring->cur = ring->start;
-	*(ring->cur++) = data;
+	/*
+	 * ring->next points to the current command being written - it won't be
+	 * committed as ring->cur until the flush
+	 */
+	if (ring->next == ring->end)
+		ring->next = ring->start;
+	*(ring->next++) = data;
 }
 
 #endif /* __MSM_RINGBUFFER_H__ */
Signature · 6 lines
1.9.1

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno
#9 lore raw

PATCH v2 6/9 drm/msm: Make the value of RB_CNTL (almost) generic

In reply to Jordan Crouse

We use a global ringbuffer size and block size for all targets and at least for 5XX preemption we need to know the value the RB_CNTL in several locations so it makes sense to calculate it once and use it everywhere.

The only monkey wrench is that we need to disable the RPTR shadow for A430 targets but that only needs to be done once and doesn't affect A5XX so we can or in the value at init time.

---
 drivers/gpu/drm/msm/adreno/adreno_gpu.c | 12 +++++++-----
 drivers/gpu/drm/msm/msm_gpu.h           |  5 +++++
 2 files changed, 12 insertions(+), 5 deletions(-)
drivers/gpu/drm/msm/adreno/adreno_gpu.c +7 -5
index 77eaa46..72f1132 100644
@@ -21,7 +21,6 @@
 #include "msm_gem.h"
 #include "msm_mmu.h"
 
-#define RB_BLKSIZE 32
 
 int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value)
 {
@@ -96,11 +95,14 @@ int adreno_hw_init(struct msm_gpu *gpu)
 		ring->memptrs->rptr = 0;
 	}
 
-	/* Setup REG_CP_RB_CNTL: */
+	/*
+	 * Setup REG_CP_RB_CNTL.  The same value is used across targets (with
+	 * the excpetion of A430 that disables the RPTR shadow) - the cacluation
+	 * for the ringbuffer size and block size is moved to msm_gpu.h for the
+	 * pre-processor to deal with and the A430 variant is ORed in here
+	 */
 	adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_CNTL,
-		/* size is log2(quad-words): */
-		AXXX_CP_RB_CNTL_BUFSZ(ilog2(MSM_GPU_RINGBUFFER_SZ / 8)) |
-		AXXX_CP_RB_CNTL_BLKSZ(ilog2(RB_BLKSIZE / 8)) |
+		MSM_GPU_RB_CNTL_DEFAULT |
 		(adreno_is_a430(adreno_gpu) ? AXXX_CP_RB_CNTL_NO_UPDATE : 0));
 
 	/* Setup ringbuffer address - use ringbuffer[0] for GPU init */
drivers/gpu/drm/msm/msm_gpu.h +5 -0
index 1be0317..e113d64 100644
@@ -129,6 +129,11 @@ struct msm_gpu {
 
 /* It turns out that all targets use the same ringbuffer size */
 #define MSM_GPU_RINGBUFFER_SZ SZ_32K
+#define MSM_GPU_RINGBUFFER_BLKSIZE 32
+
+#define MSM_GPU_RB_CNTL_DEFAULT \
+		(AXXX_CP_RB_CNTL_BUFSZ(ilog2(MSM_GPU_RINGBUFFER_SZ / 8)) | \
+		AXXX_CP_RB_CNTL_BLKSZ(ilog2(MSM_GPU_RINGBUFFER_BLKSIZE / 8)))
 
 static inline bool msm_gpu_active(struct msm_gpu *gpu)
 {
Signature · 1 line
1.9.1
#10 lore raw

PATCH v2 7/9 drm/msm: Implement preemption for A5XX targets

In reply to Jordan Crouse

Implement preemption for A5XX targets - this allows multiple ringbuffers for different priorities with automatic preemption of a lower priority ringbuffer if a higher one is ready.

---
 drivers/gpu/drm/msm/Makefile              |   1 +
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c     | 176 ++++++++++++++++-
 drivers/gpu/drm/msm/adreno/a5xx_gpu.h     | 107 ++++++++++-
 drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 305 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/msm/adreno/adreno_gpu.c   |  14 +-
 drivers/gpu/drm/msm/adreno/adreno_gpu.h   |   7 +-
 drivers/gpu/drm/msm/msm_drv.h             |   2 +-
 drivers/gpu/drm/msm/msm_gpu.c             |   5 +-
 drivers/gpu/drm/msm/msm_ringbuffer.c      |   1 +
 drivers/gpu/drm/msm/msm_ringbuffer.h      |   1 +
 10 files changed, 599 insertions(+), 20 deletions(-)
 create mode 100644 drivers/gpu/drm/msm/adreno/a5xx_preempt.c
drivers/gpu/drm/msm/Makefile +1 -0
index 3c234e7..d0b26dd 100644
@@ -8,6 +8,7 @@ msm-y := \
 	adreno/a4xx_gpu.o \
 	adreno/a5xx_gpu.o \
 	adreno/a5xx_power.o \
+	adreno/a5xx_preempt.o \
 	hdmi/hdmi.o \
 	hdmi/hdmi_audio.o \
 	hdmi/hdmi_bridge.o \
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +171 -5

270 lines not on this page · raw message

drivers/gpu/drm/msm/adreno/a5xx_gpu.h +106 -1

124 lines not on this page · raw message

drivers/gpu/drm/msm/adreno/a5xx_preempt.c new file +305 -0

305 lines not on this page · raw message

drivers/gpu/drm/msm/adreno/adreno_gpu.c +5 -9
index 72f1132..1b06eef 100644
@@ -118,11 +118,6 @@ int adreno_hw_init(struct msm_gpu *gpu)
 	return 0;
 }
 
-static uint32_t get_wptr(struct msm_ringbuffer *ring)
-{
-	return ring->cur - ring->start;
-}
-
 /* Use this helper to read rptr, since a430 doesn't update rptr in memory */
 static uint32_t get_rptr(struct adreno_gpu *adreno_gpu,
 		struct msm_ringbuffer *ring)
@@ -177,7 +172,7 @@ void adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 		case MSM_SUBMIT_CMD_BUF:
 			OUT_PKT3(ring, adreno_is_a430(adreno_gpu) ?
 				CP_INDIRECT_BUFFER_PFE : CP_INDIRECT_BUFFER_PFD, 2);
-			OUT_RING(ring, submit->cmd[i].iova);
+			OUT_RING(ring, lower_32_bits(submit->cmd[i].iova));
 			OUT_RING(ring, submit->cmd[i].size);
 			OUT_PKT2(ring);
 			break;
@@ -244,7 +239,7 @@ void adreno_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring)
 	 * to account for the possibility that the last command fit exactly into
 	 * the ringbuffer and rb->next hasn't wrapped to zero yet
 	 */
-	wptr = (ring->cur - ring->start) % (MSM_GPU_RINGBUFFER_SZ >> 2);
+	wptr = get_wptr(ring);
 
 	/* ensure writes to ringbuffer have hit system memory: */
 	mb();
@@ -262,8 +257,9 @@ bool adreno_idle(struct msm_gpu *gpu, struct msm_ringbuffer *ring)
 		return true;
 
 	/* TODO maybe we need to reset GPU here to recover from hang? */
-	DRM_ERROR("%s: timeout waiting to drain ringbuffer %d!\n", gpu->name,
-		ring->id);
+	DRM_ERROR("%s: timeout waiting to drain ringbuffer %d rptr/wptr = %X/%X\n",
+		gpu->name, ring->id, get_rptr(adreno_gpu, ring), wptr);
+
 	return false;
 }
 
drivers/gpu/drm/msm/adreno/adreno_gpu.h +6 -1
index 2a51ac4..c81434c 100644
@@ -2,7 +2,7 @@
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <[email protected]>
  *
- * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ * Copyright (c) 2014,2017 The Linux Foundation. All rights reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published by
@@ -309,6 +309,11 @@ static inline void adreno_gpu_write64(struct adreno_gpu *gpu,
 	adreno_gpu_write(gpu, hi, upper_32_bits(data));
 }
 
+static inline uint32_t get_wptr(struct msm_ringbuffer *ring)
+{
+	return (ring->cur - ring->start) % (MSM_GPU_RINGBUFFER_SZ >> 2);
+}
+
 /*
  * Given a register and a count, return a value to program into
  * REG_CP_PROTECT_REG(n) - this will block both reads and writes for _len
drivers/gpu/drm/msm/msm_drv.h +1 -1
index 2821f57..9853e3e 100644
@@ -74,7 +74,7 @@ struct msm_vblank_ctrl {
 	spinlock_t lock;
 };
 
-#define MSM_GPU_MAX_RINGS 1
+#define MSM_GPU_MAX_RINGS 4
 
 struct msm_drm_private {
 
drivers/gpu/drm/msm/msm_gpu.c +2 -3
index 69c62c0..c675439 100644
@@ -295,8 +295,7 @@ static void recover_worker(struct work_struct *work)
 		 * Replay all remaining submits starting with highest priority
 		 * ring
 		 */
-
-		for (i = gpu->nr_rings - 1; i >= 0; i--) {
+		for (i = 0; i < gpu->nr_rings; i++) {
 			struct msm_ringbuffer *ring = gpu->rb[i];
 
 			list_for_each_entry(submit, &ring->submits, node)
@@ -476,7 +475,7 @@ static void retire_submits(struct msm_gpu *gpu)
 	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
 
 	/* Retire the commits starting with highest priority */
-	for (i = gpu->nr_rings - 1; i >= 0; i--) {
+	for (i = 0; i < gpu->nr_rings; i++) {
 		struct msm_ringbuffer *ring = gpu->rb[i];
 
 		list_for_each_entry_safe(submit, tmp, &ring->submits, node) {
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -0
index 6015959..d67955d 100644
@@ -53,6 +53,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
 	ring->memptrs_iova = memptrs_iova;
 
 	INIT_LIST_HEAD(&ring->submits);
+	spin_lock_init(&ring->lock);
 
 	snprintf(name, sizeof(name), "gpu-ring-%d", ring->id);
 
drivers/gpu/drm/msm/msm_ringbuffer.h +1 -0
index 3749764..cffce09 100644
@@ -40,6 +40,7 @@ struct msm_ringbuffer {
 	struct msm_rbmemptrs *memptrs;
 	uint64_t memptrs_iova;
 	struct msm_fence_context *fctx;
+	spinlock_t lock;
 };
 
 struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
Signature · 1 line
1.9.1
#6 lore raw

PATCH 8/9 drm/msm: Removed unused struct_mutex_task

In reply to a message not in the archive

Recent changes to locking have rendered struct_mutex_task unused.

---
 drivers/gpu/drm/msm/msm_drv.h        | 6 ------
 drivers/gpu/drm/msm/msm_gem_submit.c | 2 --
 2 files changed, 8 deletions(-)
drivers/gpu/drm/msm/msm_drv.h +0 -6
index 9853e3e..c46dc12 100644
@@ -154,12 +154,6 @@ struct msm_drm_private {
 	struct shrinker shrinker;
 
 	struct msm_vblank_ctrl vblank_ctrl;
-
-	/* task holding struct_mutex.. currently only used in submit path
-	 * to detect and reject faults from copy_from_user() for submit
-	 * ioctl.
-	 */
-	struct task_struct *struct_mutex_task;
 };
 
 struct msm_format {
drivers/gpu/drm/msm/msm_gem_submit.c +0 -2
index 9e087cf..d68bfa6 100644
@@ -449,7 +449,6 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 			goto out_unlock;
 		}
 	}
-	priv->struct_mutex_task = current;
 
 	submit = submit_create(dev, gpu, queue, args->nr_bos, args->nr_cmds);
 	if (!submit) {
@@ -569,7 +568,6 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 out_unlock:
 	if (ret && (out_fence_fd >= 0))
 		put_unused_fd(out_fence_fd);
-	priv->struct_mutex_task = NULL;
 	mutex_unlock(&dev->struct_mutex);
 	return ret;
 }
Signature · 6 lines
1.9.1

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno
#7 lore raw

PATCH 9/9 drm/msm: dump a rd GPUADDR header for all buffers in the command

In reply to a message not in the archive

Currently the rd dump avoids any buffers marked as WRITE under the assumption that the contents are not interesting. While it is true that the contents are uninteresting we should still print the iova and size for all buffers so that any listening replay tools can correctly construct the submission.

Print the header for all buffers but only dump the contents for buffers marked as READ.

---
 drivers/gpu/drm/msm/msm_rd.c | 30 +++++++++++++++---------------
 1 file changed, 15 insertions(+), 15 deletions(-)
drivers/gpu/drm/msm/msm_rd.c +15 -15
index 0366b80..4c858d8 100644
@@ -268,10 +268,6 @@ static void snapshot_buf(struct msm_rd_state *rd,
 	struct msm_gem_object *obj = submit->bos[idx].obj;
 	const char *buf;
 
-	buf = msm_gem_get_vaddr(&obj->base);
-	if (IS_ERR(buf))
-		return;
-
 	if (iova) {
 		buf += iova - submit->bos[idx].iova;
 	} else {
@@ -279,8 +275,21 @@ static void snapshot_buf(struct msm_rd_state *rd,
 		size = obj->base.size;
 	}
 
+	/*
+	 * Always write the GPUADDR header so can get a complete list of all the
+	 * buffers in the cmd
+	 */
 	rd_write_section(rd, RD_GPUADDR,
 			(uint32_t[3]){ iova, size, iova >> 32 }, 12);
+
+	/* But only dump the contents of buffers marked READ */
+	if (!(submit->bos[idx].flags & MSM_SUBMIT_BO_READ))
+		return;
+
+	buf = msm_gem_get_vaddr(&obj->base);
+	if (IS_ERR(buf))
+		return;
+
 	rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
 	msm_gem_put_vaddr(&obj->base);
@@ -309,17 +318,8 @@ void msm_rd_dump_submit(struct msm_gem_submit *submit)
 
 	rd_write_section(rd, RD_CMD, msg, ALIGN(n, 4));
 
-	if (rd_full) {
-		for (i = 0; i < submit->nr_bos; i++) {
-			/* buffers that are written to probably don't start out
-			 * with anything interesting:
-			 */
-			if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
-				continue;
-
-			snapshot_buf(rd, submit, i, 0, 0);
-		}
-	}
+	for (i = 0; rd_full && i < submit->nr_bos; i++)
+		snapshot_buf(rd, submit, i, 0, 0);
 
 	for (i = 0; i < submit->nr_cmds; i++) {
 		uint64_t iova = submit->cmd[i].iova;
Signature · 6 lines
1.9.1

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno