2013-06-25 08:55:07

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 0/6] Various ux500 crypto updates

Hello Herbert,

these are some fixes to issue pointed out by sparse and dmaengine API
udpates on the recently enabled ux500 hw crypto drivers.

These are based on recent linux-next, and can probably be applied on an
architecture specific branch with your ack, or just skip to the next
merge window.

Thanks,
Fabio


Fabio Baltieri (6):
crypto: ux500/hash: use readl on iomem addresses
crypto: ux500/hash: add missing static qualifiers
crypto: ux500/crypt: add missing __iomem qualifiers
crypto: ux500: use dmaengine_device_control API
crypto: ux500: use dmaengine_prep_slave_sg API
crypto: ux500: use dmaengine_submit API

drivers/crypto/ux500/cryp/cryp.c | 4 ++--
drivers/crypto/ux500/cryp/cryp_core.c | 26 +++++++++++++-------------
drivers/crypto/ux500/hash/hash_core.c | 33 ++++++++++++++++-----------------
3 files changed, 31 insertions(+), 32 deletions(-)

--
1.8.2


2013-06-25 08:54:44

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 1/6] crypto: ux500/hash: use readl on iomem addresses

Always use readl when reading memory mapped registers.

Signed-off-by: Fabio Baltieri <[email protected]>
---
drivers/crypto/ux500/hash/hash_core.c | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/drivers/crypto/ux500/hash/hash_core.c b/drivers/crypto/ux500/hash/hash_core.c
index f89fe8a..154f437 100644
--- a/drivers/crypto/ux500/hash/hash_core.c
+++ b/drivers/crypto/ux500/hash/hash_core.c
@@ -473,12 +473,12 @@ static void hash_hw_write_key(struct hash_device_data *device_data,
HASH_SET_DIN(&word, nwords);
}

- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();

HASH_SET_DCAL;

- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();
}

@@ -661,7 +661,7 @@ static void hash_messagepad(struct hash_device_data *device_data,
if (index_bytes)
HASH_SET_DIN(message, nwords);

- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();

/* num_of_bytes == 0 => NBLW <- 0 (32 bits valid in DATAIN) */
@@ -676,7 +676,7 @@ static void hash_messagepad(struct hash_device_data *device_data,
(int)(readl_relaxed(&device_data->base->str) &
HASH_STR_NBLW_MASK));

- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();
}

@@ -776,7 +776,7 @@ void hash_begin(struct hash_device_data *device_data, struct hash_ctx *ctx)
/* HW and SW initializations */
/* Note: there is no need to initialize buffer and digest members */

- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();

/*
@@ -962,7 +962,7 @@ static int hash_dma_final(struct ahash_request *req)
wait_for_completion(&ctx->device->dma.complete);
hash_dma_done(ctx);

- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();

if (ctx->config.oper_mode == HASH_OPER_MODE_HMAC && ctx->key) {
@@ -1060,7 +1060,7 @@ int hash_hw_final(struct ahash_request *req)
req_ctx->state.index);
} else {
HASH_SET_DCAL;
- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();
}

@@ -1189,7 +1189,7 @@ int hash_resume_state(struct hash_device_data *device_data,
temp_cr = device_state->temp_cr;
writel_relaxed(temp_cr & HASH_CR_RESUME_MASK, &device_data->base->cr);

- if (device_data->base->cr & HASH_CR_MODE_MASK)
+ if (readl(&device_data->base->cr) & HASH_CR_MODE_MASK)
hash_mode = HASH_OPER_MODE_HMAC;
else
hash_mode = HASH_OPER_MODE_HASH;
@@ -1233,7 +1233,7 @@ int hash_save_state(struct hash_device_data *device_data,
* actually makes sure that there isn't any ongoing calculation in the
* hardware.
*/
- while (device_data->base->str & HASH_STR_DCAL_MASK)
+ while (readl(&device_data->base->str) & HASH_STR_DCAL_MASK)
cpu_relax();

temp_cr = readl_relaxed(&device_data->base->cr);
@@ -1242,7 +1242,7 @@ int hash_save_state(struct hash_device_data *device_data,

device_state->din_reg = readl_relaxed(&device_data->base->din);

- if (device_data->base->cr & HASH_CR_MODE_MASK)
+ if (readl(&device_data->base->cr) & HASH_CR_MODE_MASK)
hash_mode = HASH_OPER_MODE_HMAC;
else
hash_mode = HASH_OPER_MODE_HASH;
--
1.8.2

2013-06-25 08:55:44

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 2/6] crypto: ux500/hash: add missing static qualifiers

Add missing static qualifiers to hash_process_data and hash_hw_final.

Signed-off-by: Fabio Baltieri <[email protected]>
---
drivers/crypto/ux500/hash/hash_core.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/crypto/ux500/hash/hash_core.c b/drivers/crypto/ux500/hash/hash_core.c
index 154f437..dbf7b63 100644
--- a/drivers/crypto/ux500/hash/hash_core.c
+++ b/drivers/crypto/ux500/hash/hash_core.c
@@ -792,8 +792,7 @@ void hash_begin(struct hash_device_data *device_data, struct hash_ctx *ctx)
HASH_CLEAR_BITS(&device_data->base->str, HASH_STR_NBLW_MASK);
}

-int hash_process_data(
- struct hash_device_data *device_data,
+static int hash_process_data(struct hash_device_data *device_data,
struct hash_ctx *ctx, struct hash_req_ctx *req_ctx,
int msg_length, u8 *data_buffer, u8 *buffer, u8 *index)
{
@@ -992,7 +991,7 @@ out:
* hash_hw_final - The final hash calculation function
* @req: The hash request for the job.
*/
-int hash_hw_final(struct ahash_request *req)
+static int hash_hw_final(struct ahash_request *req)
{
int ret = 0;
struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
--
1.8.2

2013-06-25 08:54:47

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 4/6] crypto: ux500: use dmaengine_device_control API

Use dmaengine_device_control inline function instead of going through the
structures manually.

Signed-off-by: Fabio Baltieri <[email protected]>
---
drivers/crypto/ux500/cryp/cryp_core.c | 4 ++--
drivers/crypto/ux500/hash/hash_core.c | 2 +-
2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/crypto/ux500/cryp/cryp_core.c b/drivers/crypto/ux500/cryp/cryp_core.c
index b675bbd..1957c18 100644
--- a/drivers/crypto/ux500/cryp/cryp_core.c
+++ b/drivers/crypto/ux500/cryp/cryp_core.c
@@ -607,12 +607,12 @@ static void cryp_dma_done(struct cryp_ctx *ctx)
dev_dbg(ctx->device->dev, "[%s]: ", __func__);

chan = ctx->device->dma.chan_mem2cryp;
- chan->device->device_control(chan, DMA_TERMINATE_ALL, 0);
+ dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
dma_unmap_sg(chan->device->dev, ctx->device->dma.sg_src,
ctx->device->dma.sg_src_len, DMA_TO_DEVICE);

chan = ctx->device->dma.chan_cryp2mem;
- chan->device->device_control(chan, DMA_TERMINATE_ALL, 0);
+ dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
dma_unmap_sg(chan->device->dev, ctx->device->dma.sg_dst,
ctx->device->dma.sg_dst_len, DMA_FROM_DEVICE);
}
diff --git a/drivers/crypto/ux500/hash/hash_core.c b/drivers/crypto/ux500/hash/hash_core.c
index dbf7b63..8c2817804 100644
--- a/drivers/crypto/ux500/hash/hash_core.c
+++ b/drivers/crypto/ux500/hash/hash_core.c
@@ -203,7 +203,7 @@ static void hash_dma_done(struct hash_ctx *ctx)
struct dma_chan *chan;

chan = ctx->device->dma.chan_mem2hash;
- chan->device->device_control(chan, DMA_TERMINATE_ALL, 0);
+ dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
dma_unmap_sg(chan->device->dev, ctx->device->dma.sg,
ctx->device->dma.sg_len, DMA_TO_DEVICE);

--
1.8.2

2013-06-25 08:54:49

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 6/6] crypto: ux500: use dmaengine_submit API

Use dmaengine_submit instead of calling desc->tx_submit manually.

Signed-off-by: Fabio Baltieri <[email protected]>
---
drivers/crypto/ux500/cryp/cryp_core.c | 2 +-
drivers/crypto/ux500/hash/hash_core.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/crypto/ux500/cryp/cryp_core.c b/drivers/crypto/ux500/cryp/cryp_core.c
index 0cd89df..a999f53 100644
--- a/drivers/crypto/ux500/cryp/cryp_core.c
+++ b/drivers/crypto/ux500/cryp/cryp_core.c
@@ -594,7 +594,7 @@ static int cryp_set_dma_transfer(struct cryp_ctx *ctx,
return -EFAULT;
}

- cookie = desc->tx_submit(desc);
+ cookie = dmaengine_submit(desc);
dma_async_issue_pending(channel);

return 0;
diff --git a/drivers/crypto/ux500/hash/hash_core.c b/drivers/crypto/ux500/hash/hash_core.c
index 3527777..496ae6a 100644
--- a/drivers/crypto/ux500/hash/hash_core.c
+++ b/drivers/crypto/ux500/hash/hash_core.c
@@ -192,7 +192,7 @@ static int hash_set_dma_transfer(struct hash_ctx *ctx, struct scatterlist *sg,
desc->callback = hash_dma_callback;
desc->callback_param = ctx;

- cookie = desc->tx_submit(desc);
+ cookie = dmaengine_submit(desc);
dma_async_issue_pending(channel);

return 0;
--
1.8.2

2013-06-25 08:54:48

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 5/6] crypto: ux500: use dmaengine_prep_slave_sg API

Use dmaengine_prep_slave_sg inline function instead of going through the
structures manually.

Signed-off-by: Fabio Baltieri <[email protected]>
---
drivers/crypto/ux500/cryp/cryp_core.c | 20 ++++++++++----------
drivers/crypto/ux500/hash/hash_core.c | 4 ++--
2 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/drivers/crypto/ux500/cryp/cryp_core.c b/drivers/crypto/ux500/cryp/cryp_core.c
index 1957c18..0cd89df 100644
--- a/drivers/crypto/ux500/cryp/cryp_core.c
+++ b/drivers/crypto/ux500/cryp/cryp_core.c
@@ -553,10 +553,10 @@ static int cryp_set_dma_transfer(struct cryp_ctx *ctx,
dev_dbg(ctx->device->dev, "[%s]: Setting up DMA for buffer "
"(TO_DEVICE)", __func__);

- desc = channel->device->device_prep_slave_sg(channel,
- ctx->device->dma.sg_src,
- ctx->device->dma.sg_src_len,
- direction, DMA_CTRL_ACK, NULL);
+ desc = dmaengine_prep_slave_sg(channel,
+ ctx->device->dma.sg_src,
+ ctx->device->dma.sg_src_len,
+ direction, DMA_CTRL_ACK);
break;

case DMA_FROM_DEVICE:
@@ -577,12 +577,12 @@ static int cryp_set_dma_transfer(struct cryp_ctx *ctx,
dev_dbg(ctx->device->dev, "[%s]: Setting up DMA for buffer "
"(FROM_DEVICE)", __func__);

- desc = channel->device->device_prep_slave_sg(channel,
- ctx->device->dma.sg_dst,
- ctx->device->dma.sg_dst_len,
- direction,
- DMA_CTRL_ACK |
- DMA_PREP_INTERRUPT, NULL);
+ desc = dmaengine_prep_slave_sg(channel,
+ ctx->device->dma.sg_dst,
+ ctx->device->dma.sg_dst_len,
+ direction,
+ DMA_CTRL_ACK |
+ DMA_PREP_INTERRUPT);

desc->callback = cryp_dma_out_callback;
desc->callback_param = ctx;
diff --git a/drivers/crypto/ux500/hash/hash_core.c b/drivers/crypto/ux500/hash/hash_core.c
index 8c2817804..3527777 100644
--- a/drivers/crypto/ux500/hash/hash_core.c
+++ b/drivers/crypto/ux500/hash/hash_core.c
@@ -180,9 +180,9 @@ static int hash_set_dma_transfer(struct hash_ctx *ctx, struct scatterlist *sg,

dev_dbg(ctx->device->dev, "[%s]: Setting up DMA for buffer "
"(TO_DEVICE)", __func__);
- desc = channel->device->device_prep_slave_sg(channel,
+ desc = dmaengine_prep_slave_sg(channel,
ctx->device->dma.sg, ctx->device->dma.sg_len,
- direction, DMA_CTRL_ACK | DMA_PREP_INTERRUPT, NULL);
+ direction, DMA_CTRL_ACK | DMA_PREP_INTERRUPT);
if (!desc) {
dev_err(ctx->device->dev,
"[%s]: device_prep_slave_sg() failed!", __func__);
--
1.8.2

2013-06-25 08:55:48

by Fabio Baltieri

[permalink] [raw]
Subject: [PATCH 3/6] crypto: ux500/crypt: add missing __iomem qualifiers

Add missing __iomem to struct cryp_register pointers, this solve some
"incorrect type in initializer (different address spaces)" sparse
warnings.

Signed-off-by: Fabio Baltieri <[email protected]>
---
drivers/crypto/ux500/cryp/cryp.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/crypto/ux500/cryp/cryp.c b/drivers/crypto/ux500/cryp/cryp.c
index 3eafa90..43a0c8a 100644
--- a/drivers/crypto/ux500/cryp/cryp.c
+++ b/drivers/crypto/ux500/cryp/cryp.c
@@ -291,7 +291,7 @@ void cryp_save_device_context(struct cryp_device_data *device_data,
int cryp_mode)
{
enum cryp_algo_mode algomode;
- struct cryp_register *src_reg = device_data->base;
+ struct cryp_register __iomem *src_reg = device_data->base;
struct cryp_config *config =
(struct cryp_config *)device_data->current_ctx;

@@ -349,7 +349,7 @@ void cryp_save_device_context(struct cryp_device_data *device_data,
void cryp_restore_device_context(struct cryp_device_data *device_data,
struct cryp_device_context *ctx)
{
- struct cryp_register *reg = device_data->base;
+ struct cryp_register __iomem *reg = device_data->base;
struct cryp_config *config =
(struct cryp_config *)device_data->current_ctx;

--
1.8.2

2013-06-25 09:20:50

by Linus Walleij

[permalink] [raw]
Subject: Re: [PATCH 0/6] Various ux500 crypto updates

On Tue, Jun 25, 2013 at 10:54 AM, Fabio Baltieri
<[email protected]> wrote:

> these are some fixes to issue pointed out by sparse and dmaengine API
> udpates on the recently enabled ux500 hw crypto drivers.

The series:
Acked-by: Linus Walleij <[email protected]>

> These are based on recent linux-next, and can probably be applied on an
> architecture specific branch with your ack, or just skip to the next
> merge window.

I think these depend on the DMA40 fixes that reside in the ARM SoC
tree. If I get Herbert's ACK I can queue them.

They will have to go through that tree or wait for the next merge
window, I think they are not regressions so no hurry.

Yours,
Linus Walleij

2013-06-25 11:17:25

by Herbert Xu

[permalink] [raw]
Subject: Re: [PATCH 0/6] Various ux500 crypto updates

On Tue, Jun 25, 2013 at 11:20:50AM +0200, Linus Walleij wrote:
> On Tue, Jun 25, 2013 at 10:54 AM, Fabio Baltieri
> <[email protected]> wrote:
>
> > these are some fixes to issue pointed out by sparse and dmaengine API
> > udpates on the recently enabled ux500 hw crypto drivers.
>
> The series:
> Acked-by: Linus Walleij <[email protected]>
>
> > These are based on recent linux-next, and can probably be applied on an
> > architecture specific branch with your ack, or just skip to the next
> > merge window.
>
> I think these depend on the DMA40 fixes that reside in the ARM SoC
> tree. If I get Herbert's ACK I can queue them.

Sure,

Acked-by: Herbert Xu <[email protected]>

Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

2013-06-25 12:34:05

by Linus Walleij

[permalink] [raw]
Subject: Re: [PATCH 0/6] Various ux500 crypto updates

On Tue, Jun 25, 2013 at 1:17 PM, Herbert Xu <[email protected]> wrote:
> On Tue, Jun 25, 2013 at 11:20:50AM +0200, Linus Walleij wrote:
>> On Tue, Jun 25, 2013 at 10:54 AM, Fabio Baltieri
>> <[email protected]> wrote:
>>
>> > these are some fixes to issue pointed out by sparse and dmaengine API
>> > udpates on the recently enabled ux500 hw crypto drivers.
>>
>> The series:
>> Acked-by: Linus Walleij <[email protected]>
>>
>> > These are based on recent linux-next, and can probably be applied on an
>> > architecture specific branch with your ack, or just skip to the next
>> > merge window.
>>
>> I think these depend on the DMA40 fixes that reside in the ARM SoC
>> tree. If I get Herbert's ACK I can queue them.
>
> Sure,
>
> Acked-by: Herbert Xu <[email protected]>

OK I'll attempt to queue this up and send it to ARM SoC.

Yours,
Linus Walleij