drm/vmwgfx: Stabilize the command buffer submission code

This commit addresses some stability problems with the command buffer
submission code recently introduced:

1) Make the vmw_cmdbuf_man_process() function handle reruns internally to
avoid losing interrupts if the caller forgets to rerun on -EAGAIN.
2) Handle default command buffer allocations using inline command buffers.
This avoids rare allocation deadlocks.
3) In case of command buffer errors we might lose fence submissions.
Therefore send a new fence after each command buffer error. This will help
avoid lengthy fence waits.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c
index 8a76821..6377e81 100644 (file)
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c
@@ -415,16 +415,16 @@ static void vmw_cmdbuf_ctx_process(struct vmw_cmdbuf_man *man,
  *
  * Calls vmw_cmdbuf_ctx_process() on all contexts. If any context has
  * command buffers left that are not submitted to hardware, Make sure
- * IRQ handling is turned on. Otherwise, make sure it's turned off. This
- * function may return -EAGAIN to indicate it should be rerun due to
- * possibly missed IRQs if IRQs has just been turned on.
+ * IRQ handling is turned on. Otherwise, make sure it's turned off.
  */
-static int vmw_cmdbuf_man_process(struct vmw_cmdbuf_man *man)
+static void vmw_cmdbuf_man_process(struct vmw_cmdbuf_man *man)
 {
-       int notempty = 0;
+       int notempty;
        struct vmw_cmdbuf_context *ctx;
        int i;
 
+retry:
+       notempty = 0;
        for_each_cmdbuf_ctx(man, i, ctx)
                vmw_cmdbuf_ctx_process(man, ctx, &notempty);
 
@@ -440,10 +440,8 @@ static int vmw_cmdbuf_man_process(struct vmw_cmdbuf_man *man)
                man->irq_on = true;
 
                /* Rerun in case we just missed an irq. */
-               return -EAGAIN;
+               goto retry;
        }
-
-       return 0;
 }
 
 /**
@@ -468,8 +466,7 @@ static void vmw_cmdbuf_ctx_add(struct vmw_cmdbuf_man *man,
        header->cb_context = cb_context;
        list_add_tail(&header->list, &man->ctx[cb_context].submitted);
 
-       if (vmw_cmdbuf_man_process(man) == -EAGAIN)
-               vmw_cmdbuf_man_process(man);
+       vmw_cmdbuf_man_process(man);
 }
 
 /**
@@ -488,8 +485,7 @@ static void vmw_cmdbuf_man_tasklet(unsigned long data)
        struct vmw_cmdbuf_man *man = (struct vmw_cmdbuf_man *) data;
 
        spin_lock(&man->lock);
-       if (vmw_cmdbuf_man_process(man) == -EAGAIN)
-               (void) vmw_cmdbuf_man_process(man);
+       vmw_cmdbuf_man_process(man);
        spin_unlock(&man->lock);
 }
 
@@ -507,6 +503,7 @@ static void vmw_cmdbuf_work_func(struct work_struct *work)
        struct vmw_cmdbuf_man *man =
                container_of(work, struct vmw_cmdbuf_man, work);
        struct vmw_cmdbuf_header *entry, *next;
+       uint32_t dummy;
        bool restart = false;
 
        spin_lock_bh(&man->lock);
@@ -523,6 +520,8 @@ static void vmw_cmdbuf_work_func(struct work_struct *work)
        if (restart && vmw_cmdbuf_startstop(man, true))
                DRM_ERROR("Failed restarting command buffer context 0.\n");
 
+       /* Send a new fence in case one was removed */
+       vmw_fifo_send_fence(man->dev_priv, &dummy);
 }
 
 /**
@@ -682,7 +681,7 @@ static bool vmw_cmdbuf_try_alloc(struct vmw_cmdbuf_man *man,
                                         DRM_MM_SEARCH_DEFAULT,
                                         DRM_MM_CREATE_DEFAULT);
        if (ret) {
-               (void) vmw_cmdbuf_man_process(man);
+               vmw_cmdbuf_man_process(man);
                ret = drm_mm_insert_node_generic(&man->mm, info->node,
                                                 info->page_size, 0, 0,
                                                 DRM_MM_SEARCH_DEFAULT,
@@ -1168,7 +1167,14 @@ int vmw_cmdbuf_set_pool_size(struct vmw_cmdbuf_man *man,
        drm_mm_init(&man->mm, 0, size >> PAGE_SHIFT);
 
        man->has_pool = true;
-       man->default_size = default_size;
+
+       /*
+        * For now, set the default size to VMW_CMDBUF_INLINE_SIZE to
+        * prevent deadlocks from happening when vmw_cmdbuf_space_pool()
+        * needs to wait for space and we block on further command
+        * submissions to be able to free up space.
+        */
+       man->default_size = VMW_CMDBUF_INLINE_SIZE;
        DRM_INFO("Using command buffers with %s pool.\n",
                 (man->using_mob) ? "MOB" : "DMA");