Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1542803 > unrolled thread

[PATCH] dmaengine: Convert ID allocation to an IDA

Back to article view | Back to linux.kernel

Contents

  [PATCH] dmaengine: Convert ID allocation to an IDA Matthew Wilcox <mawilcox@linuxonhyperv.com> - 2016-12-15 16:10 +0100
    Re: [PATCH] dmaengine: Convert ID allocation to an IDA Vinod Koul <vinod.koul@intel.com> - 2017-01-02 05:20 +0100

#1542803 — [PATCH] dmaengine: Convert ID allocation to an IDA

From: Matthew Wilcox <mawilcox@microsoft.com>

dmaengine currently uses an IDR to allocate DMA IDs, but it only needs
to know whether IDs are in use or not; the ID to pointer functionality
of the IDR is unused.  That means it can use the more space-efficient IDA.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 drivers/dma/dmaengine.c | 21 +++++++++++----------
 1 file changed, 11 insertions(+), 10 deletions(-)

diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 6b53526..24e0221 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -65,7 +65,7 @@
 #include <linux/mempool.h>
 
 static DEFINE_MUTEX(dma_list_mutex);
-static DEFINE_IDR(dma_idr);
+static DEFINE_IDA(dma_ida);
 static LIST_HEAD(dma_device_list);
 static long dmaengine_ref_count;
 
@@ -162,7 +162,7 @@ static void chan_dev_release(struct device *dev)
 	chan_dev = container_of(dev, typeof(*chan_dev), device);
 	if (atomic_dec_and_test(chan_dev->idr_ref)) {
 		mutex_lock(&dma_list_mutex);
-		idr_remove(&dma_idr, chan_dev->dev_id);
+		ida_remove(&dma_ida, chan_dev->dev_id);
 		mutex_unlock(&dma_list_mutex);
 		kfree(chan_dev->idr_ref);
 	}
@@ -898,14 +898,15 @@ static int get_dma_id(struct dma_device *device)
 {
 	int rc;
 
-	mutex_lock(&dma_list_mutex);
-
-	rc = idr_alloc(&dma_idr, NULL, 0, 0, GFP_KERNEL);
-	if (rc >= 0)
-		device->dev_id = rc;
+	do {
+		if (!ida_pre_get(&dma_ida, GFP_KERNEL))
+			return -ENOMEM;
+		mutex_lock(&dma_list_mutex);
+		rc = ida_get_new(&dma_ida, &device->dev_id);
+		mutex_unlock(&dma_list_mutex);
+	} while (rc == -EAGAIN);
 
-	mutex_unlock(&dma_list_mutex);
-	return rc < 0 ? rc : 0;
+	return rc;
 }
 
 /**
@@ -1035,7 +1036,7 @@ int dma_async_device_register(struct dma_device *device)
 	/* if we never registered a channel just release the idr */
 	if (atomic_read(idr_ref) == 0) {
 		mutex_lock(&dma_list_mutex);
-		idr_remove(&dma_idr, device->dev_id);
+		ida_remove(&dma_ida, device->dev_id);
 		mutex_unlock(&dma_list_mutex);
 		kfree(idr_ref);
 		return rc;
-- 
2.10.2

[toc] | [next] | [standalone]

#1548950

On Thu, Dec 15, 2016 at 08:57:51AM -0800, Matthew Wilcox wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>
> 
> dmaengine currently uses an IDR to allocate DMA IDs, but it only needs
> to know whether IDs are in use or not; the ID to pointer functionality
> of the IDR is unused.  That means it can use the more space-efficient IDA.

Applied, thanks

-- 
~Vinod

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web