devicetree.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH 1/2] dmaengine: nbpfaxi: fix a theoretical race
@ 2014-08-03 17:13 Guennadi Liakhovetski
       [not found] ` <Pine.LNX.4.64.1408031909420.13820-0199iw4Nj15frtckUFj5Ag@public.gmane.org>
  0 siblings, 1 reply; 2+ messages in thread
From: Guennadi Liakhovetski @ 2014-08-03 17:13 UTC (permalink / raw)
  To: dmaengine-u79uwXL29TY76Z2rM5mHXA
  Cc: Vinod Koul, Shevchenko, Andriy, devicetree-u79uwXL29TY76Z2rM5mHXA,
	Grant Likely, Rob Herring, Phil Edworthy, Arnd Bergmann

A race possibility exists if a DMA slave driver tries to free channel
resources witout waiting for all transfers to complete and without
explicitly terminating all requests. In such a case the IRQ processing
thread can race with .device_free_chan_resources(). To fix this race empty
all descriptor lists before freeing descriptor cache.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---
 drivers/dma/nbpfaxi.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/dma/nbpfaxi.c b/drivers/dma/nbpfaxi.c
index 77c5a89..5b40ac8 100644
--- a/drivers/dma/nbpfaxi.c
+++ b/drivers/dma/nbpfaxi.c
@@ -1054,6 +1054,7 @@ static void nbpf_free_chan_resources(struct dma_chan *dchan)
 	dev_dbg(dchan->device->dev, "Entry %s()\n", __func__);
 
 	nbpf_chan_halt(chan);
+	nbpf_chan_idle(chan);
 	/* Clean up for if a channel is re-used for MEMCPY after slave DMA */
 	nbpf_chan_prepare_default(chan);
 
-- 
1.9.3

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 2+ messages in thread

* [PATCH 2/2] dmaengine: nbpfaxi: convert to tasklet
       [not found] ` <Pine.LNX.4.64.1408031909420.13820-0199iw4Nj15frtckUFj5Ag@public.gmane.org>
@ 2014-08-03 17:13   ` Guennadi Liakhovetski
  0 siblings, 0 replies; 2+ messages in thread
From: Guennadi Liakhovetski @ 2014-08-03 17:13 UTC (permalink / raw)
  To: dmaengine-u79uwXL29TY76Z2rM5mHXA
  Cc: Vinod Koul, Shevchenko, Andriy, devicetree-u79uwXL29TY76Z2rM5mHXA,
	Grant Likely, Rob Herring, Phil Edworthy, Arnd Bergmann

It is common among dmaengine drivers to use a tasklet for bottom half
interrupt processing. Convert nbpfaxi to do the same.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski-Mmb7MZpHnFY@public.gmane.org>
---
 drivers/dma/nbpfaxi.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

diff --git a/drivers/dma/nbpfaxi.c b/drivers/dma/nbpfaxi.c
index 5b40ac8..cea6a37 100644
--- a/drivers/dma/nbpfaxi.c
+++ b/drivers/dma/nbpfaxi.c
@@ -197,6 +197,7 @@ struct nbpf_desc_page {
  */
 struct nbpf_channel {
 	struct dma_chan dma_chan;
+	struct tasklet_struct tasklet;
 	void __iomem *base;
 	struct nbpf_device *nbpf;
 	char name[16];
@@ -1111,9 +1112,9 @@ static struct dma_chan *nbpf_of_xlate(struct of_phandle_args *dma_spec,
 	return dchan;
 }
 
-static irqreturn_t nbpf_chan_irqt(int irq, void *dev)
+static void nbpf_chan_tasklet(unsigned long data)
 {
-	struct nbpf_channel *chan = dev;
+	struct nbpf_channel *chan = (struct nbpf_channel *)data;
 	struct nbpf_desc *desc, *tmp;
 	dma_async_tx_callback callback;
 	void *param;
@@ -1176,8 +1177,6 @@ static irqreturn_t nbpf_chan_irqt(int irq, void *dev)
 		if (must_put)
 			nbpf_desc_put(desc);
 	}
-
-	return IRQ_HANDLED;
 }
 
 static irqreturn_t nbpf_chan_irq(int irq, void *dev)
@@ -1186,6 +1185,7 @@ static irqreturn_t nbpf_chan_irq(int irq, void *dev)
 	bool done = nbpf_status_get(chan);
 	struct nbpf_desc *desc;
 	irqreturn_t ret;
+	bool bh = false;
 
 	if (!done)
 		return IRQ_NONE;
@@ -1200,7 +1200,8 @@ static irqreturn_t nbpf_chan_irq(int irq, void *dev)
 		ret = IRQ_NONE;
 		goto unlock;
 	} else {
-		ret = IRQ_WAKE_THREAD;
+		ret = IRQ_HANDLED;
+		bh = true;
 	}
 
 	list_move_tail(&desc->node, &chan->done);
@@ -1216,6 +1217,9 @@ static irqreturn_t nbpf_chan_irq(int irq, void *dev)
 unlock:
 	spin_unlock(&chan->lock);
 
+	if (bh)
+		tasklet_schedule(&chan->tasklet);
+
 	return ret;
 }
 
@@ -1258,8 +1262,9 @@ static int nbpf_chan_probe(struct nbpf_device *nbpf, int n)
 
 	snprintf(chan->name, sizeof(chan->name), "nbpf %d", n);
 
-	ret = devm_request_threaded_irq(dma_dev->dev, chan->irq,
-			nbpf_chan_irq, nbpf_chan_irqt, IRQF_SHARED,
+	tasklet_init(&chan->tasklet, nbpf_chan_tasklet, (unsigned long)chan);
+	ret = devm_request_irq(dma_dev->dev, chan->irq,
+			nbpf_chan_irq, IRQF_SHARED,
 			chan->name, chan);
 	if (ret < 0)
 		return ret;
-- 
1.9.3

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2014-08-03 17:13 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-08-03 17:13 [PATCH 1/2] dmaengine: nbpfaxi: fix a theoretical race Guennadi Liakhovetski
     [not found] ` <Pine.LNX.4.64.1408031909420.13820-0199iw4Nj15frtckUFj5Ag@public.gmane.org>
2014-08-03 17:13   ` [PATCH 2/2] dmaengine: nbpfaxi: convert to tasklet Guennadi Liakhovetski

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).