2013-04-18 10:27:33

by Lee Jones

Subject: [PATCH 2/9] crypto: ux500/hash - Set DMA configuration though dma_slave_config()

The DMA controller currently takes configuration information from
information passed though dma_channel_request(), but it shouldn't.
Using the API, the DMA channel should only be configured during
a dma_slave_config() call.

Cc: Herbert Xu <[email protected]>
Cc: David S. Miller <[email protected]>
Cc: Andreas Westin <[email protected]>
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
---
drivers/crypto/ux500/hash/hash_alg.h | 5 ++++-
drivers/crypto/ux500/hash/hash_core.c | 10 ++++++++++
2 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/drivers/crypto/ux500/hash/hash_alg.h b/drivers/crypto/ux500/hash/hash_alg.h
index cd9351c..be6eb54 100644
--- a/drivers/crypto/ux500/hash/hash_alg.h
+++ b/drivers/crypto/ux500/hash/hash_alg.h
@@ -11,6 +11,7 @@
#include <linux/bitops.h>

#define HASH_BLOCK_SIZE 64
+#define HASH_DMA_FIFO 4
#define HASH_DMA_ALIGN_SIZE 4
#define HASH_DMA_PERFORMANCE_MIN_SIZE 1024
#define HASH_BYTES_PER_WORD 4
@@ -347,7 +348,8 @@ struct hash_req_ctx {

/**
* struct hash_device_data - structure for a hash device.
- * @base: Pointer to the hardware base address.
+ * @base: Pointer to virtual base address of the hash device.
+ * @phybase: Pointer to physical memory location of the hash device.
* @list_node: For inclusion in klist.
* @dev: Pointer to the device dev structure.
* @ctx_lock: Spinlock for current_ctx.
@@ -361,6 +363,7 @@ struct hash_req_ctx {
*/
struct hash_device_data {
struct hash_register __iomem *base;
+ phys_addr_t phybase;
struct klist_node list_node;
struct device *dev;
struct spinlock ctx_lock;
diff --git a/drivers/crypto/ux500/hash/hash_core.c b/drivers/crypto/ux500/hash/hash_core.c
index 118386a..f9126f4 100644
--- a/drivers/crypto/ux500/hash/hash_core.c
+++ b/drivers/crypto/ux500/hash/hash_core.c
@@ -123,6 +123,13 @@ static void hash_dma_setup_channel(struct hash_device_data *device_data,
struct device *dev)
{
struct hash_platform_data *platform_data = dev->platform_data;
+ struct dma_slave_config conf = {
+ .direction = DMA_MEM_TO_DEV,
+ .dst_addr = device_data->phybase + HASH_DMA_FIFO,
+ .dst_addr_width = DMA_SLAVE_BUSWIDTH_2_BYTES,
+ .dst_maxburst = 16,
+ };
+
dma_cap_zero(device_data->dma.mask);
dma_cap_set(DMA_SLAVE, device_data->dma.mask);

@@ -132,6 +139,8 @@ static void hash_dma_setup_channel(struct hash_device_data *device_data,
platform_data->dma_filter,
device_data->dma.cfg_mem2hash);

+ dmaengine_slave_config(device_data->dma.chan_mem2hash, &conf);
+
init_completion(&device_data->dma.complete);
}

@@ -1700,6 +1709,7 @@ static int ux500_hash_probe(struct platform_device *pdev)
goto out_kfree;
}

+ device_data->phybase = res->start;
device_data->base = ioremap(res->start, resource_size(res));
if (!device_data->base) {
dev_err(dev, "[%s] ioremap() failed!",
--
1.7.10.4

2013-04-25 11:55:39

by Linus Walleij

Subject: Re: [PATCH 2/9] crypto: ux500/hash - Set DMA configuration though dma_slave_config()

Pls include [email protected] on all these crypto/hash
postings.

On Thu, Apr 18, 2013 at 12:26 PM, Lee Jones <[email protected]> wrote:

> The DMA controller currently takes configuration information from
> information passed though dma_channel_request(), but it shouldn't.
> Using the API, the DMA channel should only be configured during
> a dma_slave_config() call.
>
> Cc: Herbert Xu <[email protected]>
> Cc: David S. Miller <[email protected]>
> Cc: Andreas Westin <[email protected]>
> Cc: [email protected]
> Signed-off-by: Lee Jones <[email protected]>

(...)
> #define HASH_BLOCK_SIZE 64
> +#define HASH_DMA_FIFO 4

This is an address so write 0x0004 here please. Some hex notation atleast.

> /**
> * struct hash_device_data - structure for a hash device.
> - * @base: Pointer to the hardware base address.
> + * @base: Pointer to virtual base address of the hash device.
> + * @phybase: Pointer to physical memory location of the hash device.
> * @list_node: For inclusion in klist.
> * @dev: Pointer to the device dev structure.
> * @ctx_lock: Spinlock for current_ctx.
> @@ -361,6 +363,7 @@ struct hash_req_ctx {
> */
> struct hash_device_data {
> struct hash_register __iomem *base;
> + phys_addr_t phybase;

What you pass to the config function is a dma_addr_t actually.

This is the same thing on the platform, but generally:

phys_addr_t = in the address space as the memory controller sees it.
dma_addr_t = in the address space as the DMA controller sees it.

Often the same. Not always. So use dma_addr_t.

Apart from this a real nice patch!

Yours,
Linus Walleij