Merging upstream version 6.7.7.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

5 files changed, 93 insertions, 18 deletions

diff --git a/drivers/gpu/host1x/channel.c b/drivers/gpu/host1x/channel.c
index 2d0051d631..08077afe4c 100644
--- a/drivers/gpu/host1x/channel.c
+++ b/drivers/gpu/host1x/channel.c
@@ -27,6 +27,8 @@ int host1x_channel_list_init(struct host1x_channel_list *chlist,
 		return -ENOMEM;
 	}
 
+	mutex_init(&chlist->lock);
+
 	return 0;
 }
 
@@ -79,6 +81,25 @@ void host1x_channel_stop(struct host1x_channel *channel)
 }
 EXPORT_SYMBOL(host1x_channel_stop);
 
+/**
+ * host1x_channel_stop_all() - disable CDMA on allocated channels
+ * @host: host1x instance
+ *
+ * Stop CDMA on allocated channels
+ */
+void host1x_channel_stop_all(struct host1x *host)
+{
+	struct host1x_channel_list *chlist = &host->channel_list;
+	int bit;
+
+	mutex_lock(&chlist->lock);
+
+	for_each_set_bit(bit, chlist->allocated_channels, host->info->nb_channels)
+		host1x_channel_stop(&chlist->channels[bit]);
+
+	mutex_unlock(&chlist->lock);
+}
+
 static void release_channel(struct kref *kref)
 {
 	struct host1x_channel *channel =
@@ -104,8 +125,11 @@ static struct host1x_channel *acquire_unused_channel(struct host1x *host)
 	unsigned int max_channels = host->info->nb_channels;
 	unsigned int index;
 
+	mutex_lock(&chlist->lock);
+
 	index = find_first_zero_bit(chlist->allocated_channels, max_channels);
 	if (index >= max_channels) {
+		mutex_unlock(&chlist->lock);
 		dev_err(host->dev, "failed to find free channel\n");
 		return NULL;
 	}
@@ -114,6 +138,8 @@ static struct host1x_channel *acquire_unused_channel(struct host1x *host)
 
 	set_bit(index, chlist->allocated_channels);
 
+	mutex_unlock(&chlist->lock);
+
 	return &chlist->channels[index];
 }
 
diff --git a/drivers/gpu/host1x/channel.h b/drivers/gpu/host1x/channel.h
index 39044ff6c3..d7aede204d 100644
--- a/drivers/gpu/host1x/channel.h
+++ b/drivers/gpu/host1x/channel.h
@@ -10,6 +10,7 @@
 
 #include <linux/io.h>
 #include <linux/kref.h>
+#include <linux/mutex.h>
 
 #include "cdma.h"
 
@@ -18,6 +19,8 @@ struct host1x_channel;
 
 struct host1x_channel_list {
 	struct host1x_channel *channels;
+
+	struct mutex lock;
 	unsigned long *allocated_channels;
 };
 
@@ -37,5 +40,6 @@ int host1x_channel_list_init(struct host1x_channel_list *chlist,
 void host1x_channel_list_free(struct host1x_channel_list *chlist);
 struct host1x_channel *host1x_channel_get_index(struct host1x *host,
 						unsigned int index);
+void host1x_channel_stop_all(struct host1x *host);
 
 #endif
diff --git a/drivers/gpu/host1x/dev.c b/drivers/gpu/host1x/dev.c
index 7c6699aed7..42fd504abb 100644
--- a/drivers/gpu/host1x/dev.c
+++ b/drivers/gpu/host1x/dev.c
@@ -488,7 +488,7 @@ static int host1x_get_resets(struct host1x *host)
 static int host1x_probe(struct platform_device *pdev)
 {
 	struct host1x *host;
-	int err;
+	int err, i;
 
 	host = devm_kzalloc(&pdev->dev, sizeof(*host), GFP_KERNEL);
 	if (!host)
@@ -516,9 +516,30 @@ static int host1x_probe(struct platform_device *pdev)
 			return PTR_ERR(host->regs);
 	}
 
-	host->syncpt_irq = platform_get_irq(pdev, 0);
-	if (host->syncpt_irq < 0)
-		return host->syncpt_irq;
+	for (i = 0; i < ARRAY_SIZE(host->syncpt_irqs); i++) {
+		char irq_name[] = "syncptX";
+
+		sprintf(irq_name, "syncpt%d", i);
+
+		err = platform_get_irq_byname_optional(pdev, irq_name);
+		if (err == -ENXIO)
+			break;
+		if (err < 0)
+			return err;
+
+		host->syncpt_irqs[i] = err;
+	}
+
+	host->num_syncpt_irqs = i;
+
+	/* Device tree without irq names */
+	if (i == 0) {
+		host->syncpt_irqs[0] = platform_get_irq(pdev, 0);
+		if (host->syncpt_irqs[0] < 0)
+			return host->syncpt_irqs[0];
+
+		host->num_syncpt_irqs = 1;
+	}
 
 	mutex_init(&host->devices_lock);
 	INIT_LIST_HEAD(&host->devices);
@@ -655,6 +676,7 @@ static int __maybe_unused host1x_runtime_suspend(struct device *dev)
 	struct host1x *host = dev_get_drvdata(dev);
 	int err;
 
+	host1x_channel_stop_all(host);
 	host1x_intr_stop(host);
 	host1x_syncpt_save(host);
 
@@ -719,7 +741,7 @@ release_reset:
 static const struct dev_pm_ops host1x_pm_ops = {
 	SET_RUNTIME_PM_OPS(host1x_runtime_suspend, host1x_runtime_resume,
 			   NULL)
-	/* TODO: add system suspend-resume once driver will be ready for that */
+	SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend, pm_runtime_force_resume)
 };
 
 static struct platform_driver tegra_host1x_driver = {
diff --git a/drivers/gpu/host1x/dev.h b/drivers/gpu/host1x/dev.h
index 75de50fe03..c8e302de76 100644
--- a/drivers/gpu/host1x/dev.h
+++ b/drivers/gpu/host1x/dev.h
@@ -124,7 +124,8 @@ struct host1x {
 	void __iomem *regs;
 	void __iomem *hv_regs; /* hypervisor region */
 	void __iomem *common_regs;
-	int syncpt_irq;
+	int syncpt_irqs[8];
+	int num_syncpt_irqs;
 	struct host1x_syncpt *syncpt;
 	struct host1x_syncpt_base *bases;
 	struct device *dev;
diff --git a/drivers/gpu/host1x/hw/intr_hw.c b/drivers/gpu/host1x/hw/intr_hw.c
index b915ef7d03..9880e0c472 100644
--- a/drivers/gpu/host1x/hw/intr_hw.c
+++ b/drivers/gpu/host1x/hw/intr_hw.c
@@ -13,13 +13,20 @@
 #include "../intr.h"
 #include "../dev.h"
 
+struct host1x_intr_irq_data {
+	struct host1x *host;
+	u32 offset;
+};
+
 static irqreturn_t syncpt_thresh_isr(int irq, void *dev_id)
 {
-	struct host1x *host = dev_id;
+	struct host1x_intr_irq_data *irq_data = dev_id;
+	struct host1x *host = irq_data->host;
 	unsigned long reg;
 	unsigned int i, id;
 
-	for (i = 0; i < DIV_ROUND_UP(host->info->nb_pts, 32); i++) {
+	for (i = irq_data->offset; i < DIV_ROUND_UP(host->info->nb_pts, 32);
+	     i += host->num_syncpt_irqs) {
 		reg = host1x_sync_readl(host,
 			HOST1X_SYNC_SYNCPT_THRESH_CPU0_INT_STATUS(i));
 
@@ -67,26 +74,41 @@ static void intr_hw_init(struct host1x *host, u32 cpm)
 
 	/*
 	 * Program threshold interrupt destination among 8 lines per VM,
-	 * per syncpoint. For now, just direct all to the first interrupt
-	 * line.
+	 * per syncpoint. For each group of 32 syncpoints (corresponding to one
+	 * interrupt status register), direct to one interrupt line, going
+	 * around in a round robin fashion.
 	 */
-	for (id = 0; id < host->info->nb_pts; id++)
-		host1x_sync_writel(host, 0, HOST1X_SYNC_SYNCPT_INTR_DEST(id));
+	for (id = 0; id < host->info->nb_pts; id++) {
+		u32 reg_offset = id / 32;
+		u32 irq_index = reg_offset % host->num_syncpt_irqs;
+
+		host1x_sync_writel(host, irq_index, HOST1X_SYNC_SYNCPT_INTR_DEST(id));
+	}
 #endif
 }
 
 static int
 host1x_intr_init_host_sync(struct host1x *host, u32 cpm)
 {
-	int err;
+	int err, i;
+	struct host1x_intr_irq_data *irq_data;
+
+	irq_data = devm_kcalloc(host->dev, host->num_syncpt_irqs, sizeof(irq_data[0]), GFP_KERNEL);
+	if (!irq_data)
+		return -ENOMEM;
 
 	host1x_hw_intr_disable_all_syncpt_intrs(host);
 
-	err = devm_request_irq(host->dev, host->syncpt_irq,
-			       syncpt_thresh_isr, IRQF_SHARED,
-			       "host1x_syncpt", host);
-	if (err < 0)
-		return err;
+	for (i = 0; i < host->num_syncpt_irqs; i++) {
+		irq_data[i].host = host;
+		irq_data[i].offset = i;
+
+		err = devm_request_irq(host->dev, host->syncpt_irqs[i],
+				       syncpt_thresh_isr, IRQF_SHARED,
+				       "host1x_syncpt", &irq_data[i]);
+		if (err < 0)
+			return err;
+	}
 
 	intr_hw_init(host, cpm);