[PATCH AUTOSEL 5.10 06/21] dmaengine: xilinx: dpdma: Fix descriptor issuing on video group

[PATCH AUTOSEL 5.10 06/21] dmaengine: xilinx: dpdma: Fix descriptor issuing on video group

[Sasha Levin]

From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

[ Upstream commit 1cbd44666216278bbb6a55bcb6b9283702171c77 ]

When multiple channels are part of a video group, the transfer is
triggered only when all channels in the group are ready. The logic to do
so is incorrect, as it causes the descriptors for all channels but the
last one in a group to not being pushed to the hardware. Fix it.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://lore.kernel.org/r/20210307040629.29308-2-laurent.pinchart@ideasonboard.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/dma/xilinx/xilinx_dpdma.c | 28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

diff --git a/drivers/dma/xilinx/xilinx_dpdma.c b/drivers/dma/xilinx/xilinx_dpdma.c
index 55df63dead8d..d504112c609e 100644
--- a/drivers/dma/xilinx/xilinx_dpdma.c
+++ b/drivers/dma/xilinx/xilinx_dpdma.c
@@ -839,6 +839,7 @@ static void xilinx_dpdma_chan_queue_transfer(struct xilinx_dpdma_chan *chan)
 	struct xilinx_dpdma_tx_desc *desc;
 	struct virt_dma_desc *vdesc;
 	u32 reg, channels;
+	bool first_frame;
 
 	lockdep_assert_held(&chan->lock);
 
@@ -852,14 +853,6 @@ static void xilinx_dpdma_chan_queue_transfer(struct xilinx_dpdma_chan *chan)
 		chan->running = true;
 	}
 
-	if (chan->video_group)
-		channels = xilinx_dpdma_chan_video_group_ready(chan);
-	else
-		channels = BIT(chan->id);
-
-	if (!channels)
-		return;
-
 	vdesc = vchan_next_desc(&chan->vchan);
 	if (!vdesc)
 		return;
@@ -884,13 +877,26 @@ static void xilinx_dpdma_chan_queue_transfer(struct xilinx_dpdma_chan *chan)
 			    FIELD_PREP(XILINX_DPDMA_CH_DESC_START_ADDRE_MASK,
 				       upper_32_bits(sw_desc->dma_addr)));
 
-	if (chan->first_frame)
+	first_frame = chan->first_frame;
+	chan->first_frame = false;
+
+	if (chan->video_group) {
+		channels = xilinx_dpdma_chan_video_group_ready(chan);
+		/*
+		 * Trigger the transfer only when all channels in the group are
+		 * ready.
+		 */
+		if (!channels)
+			return;
+	} else {
+		channels = BIT(chan->id);
+	}
+
+	if (first_frame)
 		reg = XILINX_DPDMA_GBL_TRIG_MASK(channels);
 	else
 		reg = XILINX_DPDMA_GBL_RETRIG_MASK(channels);
 
-	chan->first_frame = false;
-
 	dpdma_write(xdev->reg, XILINX_DPDMA_GBL, reg);
 }
 
-- 
2.30.2

[PATCH AUTOSEL 5.10 07/21] dmaengine: xilinx: dpdma: Fix race condition in done IRQ

[Sasha Levin]

From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

[ Upstream commit 868833fbffbe51c487df4f95d4de9194264a4b30 ]

The active descriptor pointer is accessed from different contexts,
including different interrupt handlers, and its access must be protected
by the channel's lock. This wasn't done in the done IRQ handler. Fix it.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://lore.kernel.org/r/20210307040629.29308-3-laurent.pinchart@ideasonboard.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/dma/xilinx/xilinx_dpdma.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/dma/xilinx/xilinx_dpdma.c b/drivers/dma/xilinx/xilinx_dpdma.c
index d504112c609e..70b29bd079c9 100644
--- a/drivers/dma/xilinx/xilinx_dpdma.c
+++ b/drivers/dma/xilinx/xilinx_dpdma.c
@@ -1048,13 +1048,14 @@ static int xilinx_dpdma_chan_stop(struct xilinx_dpdma_chan *chan)
  */
 static void xilinx_dpdma_chan_done_irq(struct xilinx_dpdma_chan *chan)
 {
-	struct xilinx_dpdma_tx_desc *active = chan->desc.active;
+	struct xilinx_dpdma_tx_desc *active;
 	unsigned long flags;
 
 	spin_lock_irqsave(&chan->lock, flags);
 
 	xilinx_dpdma_debugfs_desc_done_irq(chan);
 
+	active = chan->desc.active;
 	if (active)
 		vchan_cyclic_callback(&active->vdesc);
 	else
-- 
2.30.2

[PATCH AUTOSEL 5.10 10/21] dmaengine: tegra20: Fix runtime PM imbalance on error

[Sasha Levin]

From: Dinghao Liu <dinghao.liu@zju.edu.cn>

[ Upstream commit 917a3200b9f467a154999c7572af345f2470aaf4 ]

pm_runtime_get_sync() will increase the runtime PM counter
even it returns an error. Thus a pairing decrement is needed
to prevent refcount leak. Fix this by replacing this API with
pm_runtime_resume_and_get(), which will not change the runtime
PM counter on error.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Acked-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20210409082805.23643-1-dinghao.liu@zju.edu.cn
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/dma/tegra20-apb-dma.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c
index 71827d9b0aa1..b7260749e8ee 100644
--- a/drivers/dma/tegra20-apb-dma.c
+++ b/drivers/dma/tegra20-apb-dma.c
@@ -723,7 +723,7 @@ static void tegra_dma_issue_pending(struct dma_chan *dc)
 		goto end;
 	}
 	if (!tdc->busy) {
-		err = pm_runtime_get_sync(tdc->tdma->dev);
+		err = pm_runtime_resume_and_get(tdc->tdma->dev);
 		if (err < 0) {
 			dev_err(tdc2dev(tdc), "Failed to enable DMA\n");
 			goto end;
@@ -818,7 +818,7 @@ static void tegra_dma_synchronize(struct dma_chan *dc)
 	struct tegra_dma_channel *tdc = to_tegra_dma_chan(dc);
 	int err;
 
-	err = pm_runtime_get_sync(tdc->tdma->dev);
+	err = pm_runtime_resume_and_get(tdc->tdma->dev);
 	if (err < 0) {
 		dev_err(tdc2dev(tdc), "Failed to synchronize DMA: %d\n", err);
 		return;
-- 
2.30.2