gpu: host1x: When out of memory contexts, wait for free context

Instead of immediately returning an error when trying to allocate
a context device and none are free, log a warning and stall the
submit until a usable context becomes available.

Bug 4712458

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Change-Id: I964c906e94a8624ded04a5f3851f8c3dbbe76d8b
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3227750
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Tested-by: Dane Liu <danel@nvidia.com>
Reviewed-by: Bibek Basu <bbasu@nvidia.com>

drivers/gpu/host1x/context.c

@@ -3,8 +3,10 @@
  * Copyright (c) 2021-2024, NVIDIA Corporation.
  */
 
+#include <linux/completion.h>
 #include <linux/device.h>
 #include <linux/kref.h>
+#include <linux/list.h>
 #include <linux/of.h>
 #include <linux/of_device.h>
 #include <linux/pid.h>
@@ -24,6 +26,7 @@ int host1x_memory_context_list_init(struct host1x *host1x)
 	cdl->devs = NULL;
 	cdl->len = 0;
 	mutex_init(&cdl->lock);
+	INIT_LIST_HEAD(&cdl->waiters);
 
 	err = of_property_count_u32_elems(node, "iommu-map");
 	if (err < 0)
@@ -103,6 +106,11 @@ void host1x_memory_context_list_free(struct host1x_memory_context_list *cdl)
 	cdl->len = 0;
 }
 
+static bool hw_usable_for_dev(struct host1x_hw_memory_context *hw, struct device *dev)
+{
+	return hw->dev.iommu->iommu_dev == dev->iommu->iommu_dev;
+}
+
 static struct host1x_hw_memory_context *host1x_memory_context_alloc_hw_locked(struct host1x *host1x,
 							  struct device *dev,
 							  struct pid *pid)
@@ -118,7 +126,7 @@ static struct host1x_hw_memory_context *host1x_memory_context_alloc_hw_locked(st
 	for (i = 0; i < cdl->len; i++) {
 		struct host1x_hw_memory_context *cd = &cdl->devs[i];
 
-		if (cd->dev.iommu->iommu_dev != dev->iommu->iommu_dev)
+		if (!hw_usable_for_dev(cd, dev))
 			continue;
 
 		if (cd->owner == pid) {
@@ -136,10 +144,8 @@ static struct host1x_hw_memory_context *host1x_memory_context_alloc_hw_locked(st
 
 	/* Steal */
 
-	if (!can_steal) {
+	if (!can_steal)
-		dev_warn(dev, "all context devices are busy\n");
 		return ERR_PTR(-EBUSY);
-	}
 
 	list_for_each_entry(ctx, &can_steal->owners, entry) {
 		struct host1x_context_mapping *mapping;
@@ -197,17 +203,47 @@ struct host1x_memory_context *host1x_memory_context_alloc(
 }
 EXPORT_SYMBOL_GPL(host1x_memory_context_alloc);
 
+struct hw_alloc_waiter {
+	struct completion wait; /* Completion to wait for free hw context */
+	struct list_head entry;
+	struct device *dev;
+};
+
 int host1x_memory_context_active(struct host1x_memory_context *ctx)
 {
 	struct host1x_memory_context_list *cdl = &ctx->host->context_list;
 	struct host1x_context_mapping *mapping;
 	struct host1x_hw_memory_context *hw;
+	struct hw_alloc_waiter waiter;
+	bool retrying = false;
 	int err = 0;
 
 	mutex_lock(&cdl->lock);
 
+retry:
 	if (!ctx->hw) {
 		hw = host1x_memory_context_alloc_hw_locked(ctx->host, ctx->dev, ctx->pid);
+		if (PTR_ERR(hw) == -EBUSY) {
+			/* All contexts busy. Wait for free context. */
+			if (!retrying)
+				dev_warn(ctx->dev, "%s: all memory contexts are busy, waiting\n",
+					current->comm);
+
+			init_completion(&waiter.wait);
+			waiter.dev = ctx->dev;
+			list_add(&waiter.entry, &cdl->waiters);
+
+			mutex_unlock(&cdl->lock);
+			err = wait_for_completion_interruptible(&waiter.wait);
+			mutex_lock(&cdl->lock);
+
+			list_del(&waiter.entry);
+			if (err)
+				goto unlock;
+
+			retrying = true;
+			goto retry;
+		}
 		if (IS_ERR(hw)) {
 			err = PTR_ERR(hw);
 			goto unlock;
@@ -306,10 +342,24 @@ EXPORT_SYMBOL_GPL(host1x_memory_context_unmap);
 void host1x_memory_context_inactive(struct host1x_memory_context *ctx)
 {
 	struct host1x_memory_context_list *cdl = &ctx->host->context_list;
+	struct hw_alloc_waiter *waiter;
 
 	mutex_lock(&cdl->lock);
 
-	ctx->hw->active--;
+	if (--ctx->hw->active == 0) {
+		/* Hardware context becomes eligible for stealing */
+		list_for_each_entry(waiter, &cdl->waiters, entry) {
+			if (!hw_usable_for_dev(ctx->hw, waiter->dev))
+				continue;
+
+			complete(&waiter->wait);
+
+			/*
+			 * Need to wake up all waiters -- there could be multiple from
+			 * the same process that can use the same freed hardware context.
+			 */
+		}
+	}
 
 	mutex_unlock(&cdl->lock);
 }

drivers/gpu/host1x/context.h

@@ -19,6 +19,7 @@ struct host1x_memory_context_list {
 	struct mutex lock;
 	struct host1x_hw_memory_context *devs;
 	unsigned int len;
+	struct list_head waiters;
 };
 
 struct host1x_hw_memory_context {