2009-12-16 15:34:58

by Vishnu Suresh

[permalink] [raw]
Subject: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

Expose Talitos's XOR functionality to be used for
RAID Parity calculation via the Async_tx layer.

Known Issue:
When used with fsldma, random crashes are observed
on some platforms. Hence, inter-operability with fsldma
is currently disabled

Thanks to Surender Kumar and Lee Nipper for their help in
realising this driver

Signed-off-by: Kim Phillips <[email protected]>
Signed-off-by: Dipen Dudhat <[email protected]>
Signed-off-by: Maneesh Gupta <[email protected]>
Signed-off-by: Vishnu Suresh <[email protected]>
---
Changes with respect to v1 as per comments received
o. Rebased to linux-next as of 20091216
o. The selection is based exclusive of fsldma
o. Intoduced a new Kernel Configuration variable
*. This enables selecting the Cryptographic functionality
of Talitos along with fsldma.
*. Disables the XOR parity calculation offload, if fsldma enabled
either as kernel in-built or as a module
*. Once the inter-operability with fsldma is resolved, this option
can be removed

drivers/crypto/Kconfig | 9 +
drivers/crypto/talitos.c | 402 +++++++++++++++++++++++++++++++++++++++++++++-
drivers/crypto/talitos.h | 2 +
3 files changed, 412 insertions(+), 1 deletions(-)

diff --git a/drivers/crypto/Kconfig b/drivers/crypto/Kconfig
index b08403d..f8a6376 100644
--- a/drivers/crypto/Kconfig
+++ b/drivers/crypto/Kconfig
@@ -203,6 +203,15 @@ config CRYPTO_DEV_TALITOS
To compile this driver as a module, choose M here: the module
will be called talitos.

+config CRYPTO_DEV_TALITOS_RAIDXOR
+ bool "Talitos RAID5 XOR Calculation Offload"
+ select DMA_ENGINE
+ depends on CRYPTO_DEV_TALITOS
+ depends on FSL_DMA=n
+ help
+ Say 'Y' here to use the Freescale Security Engine (SEC) to
+ offload RAID XOR parity Calculation
+
config CRYPTO_DEV_IXP4XX
tristate "Driver for IXP4xx crypto hardware acceleration"
depends on ARCH_IXP4XX
diff --git a/drivers/crypto/talitos.c b/drivers/crypto/talitos.c
index c47ffe8..e63b25a 100644
--- a/drivers/crypto/talitos.c
+++ b/drivers/crypto/talitos.c
@@ -1,7 +1,7 @@
/*
* talitos - Freescale Integrated Security Engine (SEC) device driver
*
- * Copyright (c) 2008 Freescale Semiconductor, Inc.
+ * Copyright (c) 2008-2009 Freescale Semiconductor, Inc.
*
* Scatterlist Crypto API glue code copied from files with the following:
* Copyright (c) 2006-2007 Herbert Xu <[email protected]>
@@ -37,6 +37,8 @@
#include <linux/io.h>
#include <linux/spinlock.h>
#include <linux/rtnetlink.h>
+#include <linux/dmaengine.h>
+#include <linux/raid/xor.h>

#include <crypto/algapi.h>
#include <crypto/aes.h>
@@ -140,6 +142,10 @@ struct talitos_private {

/* hwrng device */
struct hwrng rng;
+#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
+ /* XOR Device */
+ struct dma_device dma_dev_common;
+#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
};

/* .features flag */
@@ -684,6 +690,375 @@ static void talitos_unregister_rng(struct device *dev)
hwrng_unregister(&priv->rng);
}

+#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
+/*
+ * async_tx interface for XOR-capable SECs
+ *
+ * Dipen Dudhat <[email protected]>
+ * Maneesh Gupta <[email protected]>
+ * Vishnu Suresh <[email protected]>
+ */
+
+/**
+ * talitos_xor_chan - context management for the async_tx channel
+ * @completed_cookie: the last completed cookie
+ * @desc_lock: lock for tx queue
+ * @total_desc: number of descriptors allocated
+ * @submit_q: queue of submitted descriptors
+ * @pending_q: queue of pending descriptors
+ * @in_progress_q: queue of descriptors in progress
+ * @free_desc: queue of unused descriptors
+ * @dev: talitos device implementing this channel
+ * @common: the corresponding xor channel in async_tx
+ */
+struct talitos_xor_chan {
+ dma_cookie_t completed_cookie;
+ spinlock_t desc_lock;
+ unsigned int total_desc;
+ struct list_head submit_q;
+ struct list_head pending_q;
+ struct list_head in_progress_q;
+ struct list_head free_desc;
+ struct device *dev;
+ struct dma_chan common;
+};
+
+/**
+ * talitos_xor_desc - software xor descriptor
+ * @async_tx: the referring async_tx descriptor
+ * @node:
+ * @hwdesc: h/w descriptor
+ */
+struct talitos_xor_desc {
+ struct dma_async_tx_descriptor async_tx;
+ struct list_head tx_list;
+ struct list_head node;
+ struct talitos_desc hwdesc;
+};
+
+static enum dma_status talitos_is_tx_complete(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ dma_cookie_t *done,
+ dma_cookie_t *used)
+{
+ struct talitos_xor_chan *xor_chan;
+ dma_cookie_t last_used;
+ dma_cookie_t last_complete;
+
+ xor_chan = container_of(chan, struct talitos_xor_chan, common);
+
+ last_used = chan->cookie;
+ last_complete = xor_chan->completed_cookie;
+
+ if (done)
+ *done = last_complete;
+
+ if (used)
+ *used = last_used;
+
+ return dma_async_is_complete(cookie, last_complete, last_used);
+}
+
+static void talitos_release_xor(struct device *dev, struct talitos_desc *hwdesc,
+ void *context, int error)
+{
+ struct talitos_xor_desc *desc = context;
+ struct talitos_xor_chan *xor_chan;
+ dma_async_tx_callback callback;
+ void *callback_param;
+
+ if (unlikely(error)) {
+ dev_err(dev, "xor operation: talitos error %d\n", error);
+ BUG();
+ }
+
+ xor_chan = container_of(desc->async_tx.chan, struct talitos_xor_chan,
+ common);
+ spin_lock_bh(&xor_chan->desc_lock);
+ if (xor_chan->completed_cookie < desc->async_tx.cookie)
+ xor_chan->completed_cookie = desc->async_tx.cookie;
+
+ callback = desc->async_tx.callback;
+ callback_param = desc->async_tx.callback_param;
+ list_del(&desc->node);
+ list_add_tail(&desc->node, &xor_chan->free_desc);
+ spin_unlock_bh(&xor_chan->desc_lock);
+
+ if (callback)
+ callback(callback_param);
+
+ /* run dependent operations */
+ dma_run_dependencies(&desc->async_tx);
+}
+
+static void talitos_process_pending(struct talitos_xor_chan *xor_chan)
+{
+ struct talitos_xor_desc *desc, *_desc;
+
+ spin_lock_bh(&xor_chan->desc_lock);
+ list_for_each_entry_safe(desc, _desc, &xor_chan->pending_q, node) {
+ if (talitos_submit(xor_chan->dev, &desc->hwdesc,
+ talitos_release_xor, desc) != -EINPROGRESS)
+ break;
+
+ list_del(&desc->node);
+ list_add_tail(&desc->node, &xor_chan->in_progress_q);
+ }
+ spin_unlock_bh(&xor_chan->desc_lock);
+}
+
+/**
+ * talitos_issue_pending - move the descriptors in submit
+ * queue to pending queue and submit them for processing
+ * @chan: DMA channel
+ */
+static void talitos_issue_pending(struct dma_chan *chan)
+{
+ struct talitos_xor_chan *xor_chan;
+
+ xor_chan = container_of(chan, struct talitos_xor_chan, common);
+ spin_lock_bh(&xor_chan->desc_lock);
+ list_splice_tail_init(&xor_chan->submit_q,
+ &xor_chan->pending_q);
+ spin_unlock_bh(&xor_chan->desc_lock);
+ talitos_process_pending(xor_chan);
+}
+
+static dma_cookie_t talitos_async_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct talitos_xor_desc *desc;
+ struct talitos_xor_chan *xor_chan;
+ dma_cookie_t cookie;
+
+ desc = container_of(tx, struct talitos_xor_desc, async_tx);
+ xor_chan = container_of(tx->chan, struct talitos_xor_chan, common);
+
+ cookie = xor_chan->common.cookie + 1;
+ if (cookie < 0)
+ cookie = 1;
+
+ desc->async_tx.cookie = cookie;
+ xor_chan->common.cookie = desc->async_tx.cookie;
+
+ list_splice_tail_init(&desc->tx_list,
+ &xor_chan->submit_q);
+
+ return cookie;
+}
+
+static struct talitos_xor_desc *talitos_xor_alloc_descriptor(
+ struct talitos_xor_chan *xor_chan, gfp_t flags)
+{
+ struct talitos_xor_desc *desc;
+
+ desc = kmalloc(sizeof(*desc), flags);
+ if (desc) {
+ xor_chan->total_desc++;
+ memset(desc, 0, sizeof(*desc));
+ dma_async_tx_descriptor_init(&desc->async_tx, &xor_chan->common);
+ desc->async_tx.tx_submit = talitos_async_tx_submit;
+ INIT_LIST_HEAD(&desc->node);
+ INIT_LIST_HEAD(&desc->tx_list);
+ }
+
+ return desc;
+}
+
+static void talitos_free_chan_resources(struct dma_chan *chan)
+{
+ struct talitos_xor_chan *xor_chan;
+ struct talitos_xor_desc *desc, *_desc;
+
+ xor_chan = container_of(chan, struct talitos_xor_chan, common);
+
+ list_for_each_entry_safe(desc, _desc, &xor_chan->submit_q, node) {
+ list_del(&desc->node);
+ xor_chan->total_desc--;
+ kfree(desc);
+ }
+ list_for_each_entry_safe(desc, _desc, &xor_chan->pending_q, node) {
+ list_del(&desc->node);
+ xor_chan->total_desc--;
+ kfree(desc);
+ }
+ list_for_each_entry_safe(desc, _desc, &xor_chan->in_progress_q, node) {
+ list_del(&desc->node);
+ xor_chan->total_desc--;
+ kfree(desc);
+ }
+ list_for_each_entry_safe(desc, _desc, &xor_chan->free_desc, node) {
+ list_del(&desc->node);
+ xor_chan->total_desc--;
+ kfree(desc);
+ }
+ BUG_ON(unlikely(xor_chan->total_desc)); /* Some descriptor not freed? */
+}
+
+static int talitos_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct talitos_xor_chan *xor_chan;
+ struct talitos_xor_desc *desc;
+ LIST_HEAD(tmp_list);
+ int i;
+
+ xor_chan = container_of(chan, struct talitos_xor_chan, common);
+
+ if (!list_empty(&xor_chan->free_desc))
+ return xor_chan->total_desc;
+
+ /* 256 initial descriptors */
+ for (i = 0; i < 256; i++) {
+ desc = talitos_xor_alloc_descriptor(xor_chan, GFP_KERNEL);
+ if (!desc) {
+ dev_err(xor_chan->common.device->dev,
+ "Only %d initial descriptors\n", i);
+ break;
+ }
+ list_add_tail(&desc->node, &tmp_list);
+ }
+
+ if (!i)
+ return -ENOMEM;
+
+ /* At least one desc is allocated */
+ list_splice_init(&tmp_list, &xor_chan->free_desc);
+
+ return xor_chan->total_desc;
+}
+
+static struct dma_async_tx_descriptor * talitos_prep_dma_xor(
+ struct dma_chan *chan, dma_addr_t dest, dma_addr_t *src,
+ unsigned int src_cnt, size_t len, unsigned long flags)
+{
+ struct talitos_xor_chan *xor_chan;
+ struct talitos_xor_desc *new;
+ struct talitos_desc *desc;
+ int i, j;
+
+ BUG_ON(unlikely(len > TALITOS_MAX_DATA_LEN));
+
+ xor_chan = container_of(chan, struct talitos_xor_chan, common);
+
+ if (!list_empty(&xor_chan->free_desc)) {
+ new = container_of(xor_chan->free_desc.next,
+ struct talitos_xor_desc, node);
+ list_del(&new->node);
+ } else {
+ new = talitos_xor_alloc_descriptor(xor_chan, GFP_KERNEL);
+ }
+
+ if (!new) {
+ dev_err(xor_chan->common.device->dev,
+ "No free memory for XOR DMA descriptor\n");
+ return NULL;
+ }
+
+ desc = &new->hwdesc;
+ /* Set destination: Last pointer pair */
+ to_talitos_ptr(&desc->ptr[6], dest);
+ desc->ptr[6].len = cpu_to_be16(len);
+ desc->ptr[6].j_extent = 0;
+
+ /* Set Sources: End loading from second-last pointer pair */
+ for (i = 5, j = 0; (j < src_cnt) && (i > 0); i--, j++) {
+ to_talitos_ptr(&desc->ptr[i], src[j]);
+ desc->ptr[i].len = cpu_to_be16(len);
+ desc->ptr[i].j_extent = 0;
+ }
+
+ /*
+ * documentation states first 0 ptr/len combo marks end of sources
+ * yet device produces scatter boundary error unless all subsequent
+ * sources are zeroed out
+ */
+ for (; i >= 0; i--) {
+ to_talitos_ptr(&desc->ptr[i], 0);
+ desc->ptr[i].len = 0;
+ desc->ptr[i].j_extent = 0;
+ }
+
+ desc->hdr = DESC_HDR_SEL0_AESU | DESC_HDR_MODE0_AESU_XOR
+ | DESC_HDR_TYPE_RAID_XOR;
+
+ list_add_tail(&new->node, &new->tx_list);
+
+ new->async_tx.flags = flags;
+ new->async_tx.cookie = -EBUSY;
+
+ return &new->async_tx;
+}
+
+static void talitos_unregister_async_xor(struct device *dev)
+{
+ struct talitos_private *priv = dev_get_drvdata(dev);
+ struct talitos_xor_chan *xor_chan;
+ struct dma_chan *chan;
+
+ if (priv->dma_dev_common.chancnt)
+ dma_async_device_unregister(&priv->dma_dev_common);
+
+ list_for_each_entry(chan, &priv->dma_dev_common.channels, device_node) {
+ xor_chan = container_of(chan, struct talitos_xor_chan, common);
+ list_del(&chan->device_node);
+ priv->dma_dev_common.chancnt--;
+ kfree(xor_chan);
+ }
+}
+
+/**
+ * talitos_register_dma_async - Initialize the Freescale XOR ADMA device
+ * It is registered as a DMA device with the capability to perform
+ * XOR operation with the Async_tx layer.
+ * The various queues and channel resources are also allocated.
+ */
+static int talitos_register_async_tx(struct device *dev, int max_xor_srcs)
+{
+ struct talitos_private *priv = dev_get_drvdata(dev);
+ struct dma_device *dma_dev = &priv->dma_dev_common;
+ struct talitos_xor_chan *xor_chan;
+ int err;
+
+ xor_chan = kzalloc(sizeof(struct talitos_xor_chan), GFP_KERNEL);
+ if (!xor_chan) {
+ dev_err(dev, "unable to allocate xor channel\n");
+ return -ENOMEM;
+ }
+
+ dma_dev->dev = dev;
+ dma_dev->device_alloc_chan_resources = talitos_alloc_chan_resources;
+ dma_dev->device_free_chan_resources = talitos_free_chan_resources;
+ dma_dev->device_prep_dma_xor = talitos_prep_dma_xor;
+ dma_dev->max_xor = max_xor_srcs;
+ dma_dev->device_is_tx_complete = talitos_is_tx_complete;
+ dma_dev->device_issue_pending = talitos_issue_pending;
+ INIT_LIST_HEAD(&dma_dev->channels);
+ dma_cap_set(DMA_XOR, dma_dev->cap_mask);
+
+ xor_chan->dev = dev;
+ xor_chan->common.device = dma_dev;
+ xor_chan->total_desc = 0;
+ INIT_LIST_HEAD(&xor_chan->submit_q);
+ INIT_LIST_HEAD(&xor_chan->pending_q);
+ INIT_LIST_HEAD(&xor_chan->in_progress_q);
+ INIT_LIST_HEAD(&xor_chan->free_desc);
+ spin_lock_init(&xor_chan->desc_lock);
+
+ list_add_tail(&xor_chan->common.device_node, &dma_dev->channels);
+ dma_dev->chancnt++;
+
+ err = dma_async_device_register(dma_dev);
+ if (err) {
+ dev_err(dev, "Unable to register XOR with Async_tx\n");
+ goto err_out;
+ }
+
+ return err;
+
+err_out:
+ talitos_unregister_async_xor(dev);
+ return err;
+}
+#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
/*
* crypto alg
*/
@@ -1768,6 +2143,10 @@ static int talitos_remove(struct of_device *ofdev)
tasklet_kill(&priv->done_task);

iounmap(priv->reg);
+#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
+ if (priv->dma_dev_common.chancnt)
+ talitos_unregister_async_xor(dev);
+#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */

dev_set_drvdata(dev, NULL);

@@ -1926,6 +2305,27 @@ static int talitos_probe(struct of_device *ofdev,
dev_info(dev, "hwrng\n");
}

+#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
+ /*
+ * register with async_tx xor, if capable
+ * SEC 2.x support up to 3 RAID sources,
+ * SEC 3.x support up to 6
+ */
+ if (hw_supports(dev, DESC_HDR_SEL0_AESU | DESC_HDR_TYPE_RAID_XOR)) {
+ int max_xor_srcs = 3;
+ if (of_device_is_compatible(np, "fsl,sec3.0"))
+ max_xor_srcs = 6;
+
+ err = talitos_register_async_tx(dev, max_xor_srcs);
+ if (err) {
+ dev_err(dev, "failed to register async_tx xor: %d\n",
+ err);
+ goto err_out;
+ }
+ dev_info(dev, "max_xor_srcs %d\n", max_xor_srcs);
+ }
+#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
+
/* register crypto algorithms the device supports */
for (i = 0; i < ARRAY_SIZE(driver_algs); i++) {
if (hw_supports(dev, driver_algs[i].desc_hdr_template)) {
diff --git a/drivers/crypto/talitos.h b/drivers/crypto/talitos.h
index ff5a145..b6197bc 100644
--- a/drivers/crypto/talitos.h
+++ b/drivers/crypto/talitos.h
@@ -155,6 +155,7 @@
/* primary execution unit mode (MODE0) and derivatives */
#define DESC_HDR_MODE0_ENCRYPT cpu_to_be32(0x00100000)
#define DESC_HDR_MODE0_AESU_CBC cpu_to_be32(0x00200000)
+#define DESC_HDR_MODE0_AESU_XOR cpu_to_be32(0x0c600000)
#define DESC_HDR_MODE0_DEU_CBC cpu_to_be32(0x00400000)
#define DESC_HDR_MODE0_DEU_3DES cpu_to_be32(0x00200000)
#define DESC_HDR_MODE0_MDEU_INIT cpu_to_be32(0x01000000)
@@ -202,6 +203,7 @@
#define DESC_HDR_TYPE_IPSEC_ESP cpu_to_be32(1 << 3)
#define DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU cpu_to_be32(2 << 3)
#define DESC_HDR_TYPE_HMAC_SNOOP_NO_AFEU cpu_to_be32(4 << 3)
+#define DESC_HDR_TYPE_RAID_XOR cpu_to_be32(21 << 3)

/* link table extent field bits */
#define DESC_PTR_LNKTBL_JUMP 0x80
--
1.6.4.2



2009-12-16 22:36:58

by Kim Phillips

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

On Wed, 16 Dec 2009 21:04:58 +0530
Vishnu Suresh <[email protected]> wrote:

> Expose Talitos's XOR functionality to be used for
> RAID Parity calculation via the Async_tx layer.
>
> Known Issue:
> When used with fsldma, random crashes are observed
> on some platforms. Hence, inter-operability with fsldma
> is currently disabled
>
> Thanks to Surender Kumar and Lee Nipper for their help in
> realising this driver
>
> Signed-off-by: Kim Phillips <[email protected]>
> Signed-off-by: Dipen Dudhat <[email protected]>
> Signed-off-by: Maneesh Gupta <[email protected]>
> Signed-off-by: Vishnu Suresh <[email protected]>
> ---
> Changes with respect to v1 as per comments received
> o. Rebased to linux-next as of 20091216
> o. The selection is based exclusive of fsldma
> o. Intoduced a new Kernel Configuration variable
> *. This enables selecting the Cryptographic functionality
> of Talitos along with fsldma.
> *. Disables the XOR parity calculation offload, if fsldma enabled
> either as kernel in-built or as a module
> *. Once the inter-operability with fsldma is resolved, this option
> can be removed

wait, why can't the interoperability bug be fixed in the first place?

Kim

2009-12-16 22:39:10

by Kumar Gala

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload


On Dec 16, 2009, at 4:41 PM, Kim Phillips wrote:

> On Wed, 16 Dec 2009 21:04:58 +0530
> Vishnu Suresh <[email protected]> wrote:
>
>> Expose Talitos's XOR functionality to be used for
>> RAID Parity calculation via the Async_tx layer.
>>
>> Known Issue:
>> When used with fsldma, random crashes are observed
>> on some platforms. Hence, inter-operability with fsldma
>> is currently disabled
>>
>> Thanks to Surender Kumar and Lee Nipper for their help in
>> realising this driver
>>
>> Signed-off-by: Kim Phillips <[email protected]>
>> Signed-off-by: Dipen Dudhat <[email protected]>
>> Signed-off-by: Maneesh Gupta <[email protected]>
>> Signed-off-by: Vishnu Suresh <[email protected]>
>> ---
>> Changes with respect to v1 as per comments received
>> o. Rebased to linux-next as of 20091216
>> o. The selection is based exclusive of fsldma
>> o. Intoduced a new Kernel Configuration variable
>> *. This enables selecting the Cryptographic functionality
>> of Talitos along with fsldma.
>> *. Disables the XOR parity calculation offload, if fsldma enabled
>> either as kernel in-built or as a module
>> *. Once the inter-operability with fsldma is resolved, this option
>> can be removed
>
> wait, why can't the interoperability bug be fixed in the first place?

I agree w/Kim. We need to better understand what the bug is and how to reproduce it so we can get to the root cause.

Paper taping over it by disabling fsldma is not the right solution.

- k

2009-12-16 22:47:50

by Williams, Dan J

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

Kumar Gala wrote:
>>> Changes with respect to v1 as per comments received
>>> o. Rebased to linux-next as of 20091216
>>> o. The selection is based exclusive of fsldma
>>> o. Intoduced a new Kernel Configuration variable
>>> *. This enables selecting the Cryptographic functionality
>>> of Talitos along with fsldma.
>>> *. Disables the XOR parity calculation offload, if fsldma enabled
>>> either as kernel in-built or as a module
>>> *. Once the inter-operability with fsldma is resolved, this option
>>> can be removed
>> wait, why can't the interoperability bug be fixed in the first place?
>
> I agree w/Kim. We need to better understand what the bug is and how to reproduce it so we can get to the root cause.
>
> Paper taping over it by disabling fsldma is not the right solution.

Hopefully this prompts fsldma authors to get involved because the
interoperability issue has been out there without comment*, just
band-aids, since October.

--
Dan

* well one comment from Ira saying the interrupt functionality worked
for him.

2009-12-17 17:10:00

by Ira W. Snyder

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

On Wed, Dec 16, 2009 at 03:47:48PM -0700, Dan Williams wrote:
> Kumar Gala wrote:
> >>> Changes with respect to v1 as per comments received
> >>> o. Rebased to linux-next as of 20091216
> >>> o. The selection is based exclusive of fsldma
> >>> o. Intoduced a new Kernel Configuration variable
> >>> *. This enables selecting the Cryptographic functionality
> >>> of Talitos along with fsldma.
> >>> *. Disables the XOR parity calculation offload, if fsldma enabled
> >>> either as kernel in-built or as a module
> >>> *. Once the inter-operability with fsldma is resolved, this option
> >>> can be removed
> >> wait, why can't the interoperability bug be fixed in the first place?
> >
> > I agree w/Kim. We need to better understand what the bug is and how to reproduce it so we can get to the root cause.
> >
> > Paper taping over it by disabling fsldma is not the right solution.
>
> Hopefully this prompts fsldma authors to get involved because the
> interoperability issue has been out there without comment*, just
> band-aids, since October.
>
> --
> Dan
>
> * well one comment from Ira saying the interrupt functionality worked
> for him.

Yes, I have used the device_prep_dma_interrupt() functionality quite a
while back. However, I found it to be pretty much useless. Any
functionality I need is covered by adding a callback to the last DMA
memcpy() operation. Since the operations happen in-order, I can be sure
that the entire set of memcpy()s cas completed. I never needed the
capability to generate an interrupt without a memcpy().

I agree that the fsldma driver could use some love. There are places
where I am still not confident in the locking. Perhaps I can find some
time over Christmas to work on it, but I need someone with 85xx/86xx
hardware to test the changes. I only have 83xx hardware.

Ira

2009-12-17 17:29:26

by Kumar Gala

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload


On Dec 17, 2009, at 11:09 AM, Ira W. Snyder wrote:

> On Wed, Dec 16, 2009 at 03:47:48PM -0700, Dan Williams wrote:
>> Kumar Gala wrote:
>>>>> Changes with respect to v1 as per comments received
>>>>> o. Rebased to linux-next as of 20091216
>>>>> o. The selection is based exclusive of fsldma
>>>>> o. Intoduced a new Kernel Configuration variable
>>>>> *. This enables selecting the Cryptographic functionality
>>>>> of Talitos along with fsldma.
>>>>> *. Disables the XOR parity calculation offload, if fsldma enabled
>>>>> either as kernel in-built or as a module
>>>>> *. Once the inter-operability with fsldma is resolved, this option
>>>>> can be removed
>>>> wait, why can't the interoperability bug be fixed in the first place?
>>>
>>> I agree w/Kim. We need to better understand what the bug is and how to reproduce it so we can get to the root cause.
>>>
>>> Paper taping over it by disabling fsldma is not the right solution.
>>
>> Hopefully this prompts fsldma authors to get involved because the
>> interoperability issue has been out there without comment*, just
>> band-aids, since October.
>>
>> --
>> Dan
>>
>> * well one comment from Ira saying the interrupt functionality worked
>> for him.
>
> Yes, I have used the device_prep_dma_interrupt() functionality quite a
> while back. However, I found it to be pretty much useless. Any
> functionality I need is covered by adding a callback to the last DMA
> memcpy() operation. Since the operations happen in-order, I can be sure
> that the entire set of memcpy()s cas completed. I never needed the
> capability to generate an interrupt without a memcpy().
>
> I agree that the fsldma driver could use some love. There are places
> where I am still not confident in the locking. Perhaps I can find some
> time over Christmas to work on it, but I need someone with 85xx/86xx
> hardware to test the changes. I only have 83xx hardware.

I can test on 85xx/86xx if you work up some patches.

- k

2009-12-17 17:44:26

by Williams, Dan J

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

Ira W. Snyder wrote:
> Yes, I have used the device_prep_dma_interrupt() functionality quite a
> while back. However, I found it to be pretty much useless.

The specific case it is needed for Talitos/raid is a channel switch
interrupt. The interrupt causes the cleanup operation to be run which
will kick off any pending dependent operations on the xor channel. In
the raid case we only have callbacks at the end of a chain, so we need
the interrupt to kick the engine in an operation chain like
xor->copy->xor->callback.

--
Dan

2009-12-17 18:45:21

by Kumar Gala

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload


On Dec 17, 2009, at 11:44 AM, Dan Williams wrote:

> Ira W. Snyder wrote:
>> Yes, I have used the device_prep_dma_interrupt() functionality quite a
>> while back. However, I found it to be pretty much useless.
>
> The specific case it is needed for Talitos/raid is a channel switch interrupt. The interrupt causes the cleanup operation to be run which will kick off any pending dependent operations on the xor channel. In the raid case we only have callbacks at the end of a chain, so we need the interrupt to kick the engine in an operation chain like xor->copy->xor->callback.

Ok, I'm still confused as to how the DMA interrupt interacts with the Talitos/raid side of things. The should be completely independent (separate interrupts, separate IP blocks).

- k

2009-12-18 00:58:23

by Williams, Dan J

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

On Thu, Dec 17, 2009 at 11:45 AM, Kumar Gala <[email protected]> wrote:
>> The specific case it is needed for Talitos/raid is a channel switch interrupt. ?The interrupt causes the cleanup operation to be run which will kick off any pending dependent operations on the xor channel. ?In the raid case we only have callbacks at the end of a chain, so we need the interrupt to kick the engine in an operation chain like xor->copy->xor->callback.
>
> Ok, I'm still confused as to how the DMA interrupt interacts with the Talitos/raid side of things. ?The should be completely independent (separate interrupts, separate IP blocks).
>

To keep hardware implementation details out of md/raid the async_tx
api provides support for managing cross-channel dependency chains.
When the raid5 code submits a xor->copy->xor chain the api prepares
all the descriptors across all the involved channels but then delays
submission as needed to maintain ordering. So at a minimum we need
two interrupts in this scenario one from Talitos to kick the
submission of the copy-descriptor to fsldma when the first xor
completes, and another one to kick the submission of the second
xor-descriptor on Talitos when fsldma completes the copy. Needless to
say it is more efficient when a channel has all the capabilities, but
this channel switch mechanism has proven effective on iop3xx and
ppc4xx.

--
Dan

2009-12-18 14:46:27

by Li Yang-R58472

[permalink] [raw]
Subject: RE: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload


>Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for
>Async_tx XOR offload
>
>On Wed, Dec 16, 2009 at 03:47:48PM -0700, Dan Williams wrote:
>> Kumar Gala wrote:
>> >>> Changes with respect to v1 as per comments received o.
>Rebased to
>> >>> linux-next as of 20091216 o. The selection is based exclusive of
>> >>> fsldma o. Intoduced a new Kernel Configuration variable
>> >>> *. This enables selecting the Cryptographic functionality
>> >>> of Talitos along with fsldma.
>> >>> *. Disables the XOR parity calculation offload, if
>fsldma enabled
>> >>> either as kernel in-built or as a module
>> >>> *. Once the inter-operability with fsldma is resolved,
>this option
>> >>> can be removed
>> >> wait, why can't the interoperability bug be fixed in the
>first place?
>> >
>> > I agree w/Kim. We need to better understand what the bug
>is and how to reproduce it so we can get to the root cause.
>> >
>> > Paper taping over it by disabling fsldma is not the right solution.
>>
>> Hopefully this prompts fsldma authors to get involved because the
>> interoperability issue has been out there without comment*, just
>> band-aids, since October.
>>
>> --
>> Dan
>>
>> * well one comment from Ira saying the interrupt
>functionality worked
>> for him.
>
>Yes, I have used the device_prep_dma_interrupt() functionality
>quite a while back. However, I found it to be pretty much
>useless. Any functionality I need is covered by adding a
>callback to the last DMA
>memcpy() operation. Since the operations happen in-order, I
>can be sure that the entire set of memcpy()s cas completed. I
>never needed the capability to generate an interrupt without a
>memcpy().
>
>I agree that the fsldma driver could use some love. There are
>places where I am still not confident in the locking. Perhaps
>I can find some time over Christmas to work on it, but I need
>someone with 85xx/86xx hardware to test the changes. I only
>have 83xx hardware.

I can also help with the 85xx testing when I finish the busy project
soon.

- Leo

2009-12-18 15:02:22

by Li Yang-R58472

[permalink] [raw]
Subject: RE: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload


>Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for
>Async_tx XOR offload
>
>Ira W. Snyder wrote:
>> Yes, I have used the device_prep_dma_interrupt()
>functionality quite a
>> while back. However, I found it to be pretty much useless.
>
>The specific case it is needed for Talitos/raid is a channel
>switch interrupt. The interrupt causes the cleanup operation
>to be run which will kick off any pending dependent operations
>on the xor channel. In the raid case we only have callbacks
>at the end of a chain, so we need the interrupt to kick the
>engine in an operation chain like
>xor->copy->xor->callback.

I am wondering if can use more callbacks to kick off pending dependent operations?
Like xor->callback->copy->callback->xor->callback?

- Leo

2009-12-18 22:17:42

by Williams, Dan J

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

On Fri, Dec 18, 2009 at 8:02 AM, Li Yang-R58472 <[email protected]> wrote:
>
>>Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for
>>Async_tx XOR offload
>>
>>Ira W. Snyder wrote:
>>> Yes, I have used the device_prep_dma_interrupt()
>>functionality quite a
>>> while back. However, I found it to be pretty much useless.
>>
>>The specific case it is needed for Talitos/raid is a channel
>>switch interrupt. ?The interrupt causes the cleanup operation
>>to be run which will kick off any pending dependent operations
>>on the xor channel. ?In the raid case we only have callbacks
>>at the end of a chain, so we need the interrupt to kick the
>>engine in an operation chain like
>>xor->copy->xor->callback.
>
> I am wondering if can use more callbacks to kick off pending dependent operations?
> Like xor->callback->copy->callback->xor->callback?
>

No, the callback field is reserved for clients of the api. What you want is:

xor->cleanupT->copy->cleanupF->xor->cleanupT->callback

Where cleanupT is the Talitos descriptor cleanup routine and cleanupF
is from fsldma. The assumption is that the interrupt kicks the
cleanup routine and that calls dma_run_dependencies().

--
Dan

2009-12-26 21:41:46

by Ira W. Snyder

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

On Fri, Dec 18, 2009 at 03:17:42PM -0700, Dan Williams wrote:
> On Fri, Dec 18, 2009 at 8:02 AM, Li Yang-R58472 <[email protected]> wrote:
> >
> >>Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for
> >>Async_tx XOR offload
> >>
> >>Ira W. Snyder wrote:
> >>> Yes, I have used the device_prep_dma_interrupt()
> >>functionality quite a
> >>> while back. However, I found it to be pretty much useless.
> >>
> >>The specific case it is needed for Talitos/raid is a channel
> >>switch interrupt. ?The interrupt causes the cleanup operation
> >>to be run which will kick off any pending dependent operations
> >>on the xor channel. ?In the raid case we only have callbacks
> >>at the end of a chain, so we need the interrupt to kick the
> >>engine in an operation chain like
> >>xor->copy->xor->callback.
> >
> > I am wondering if can use more callbacks to kick off pending dependent operations?
> > Like xor->callback->copy->callback->xor->callback?
> >
>
> No, the callback field is reserved for clients of the api. What you want is:
>
> xor->cleanupT->copy->cleanupF->xor->cleanupT->callback
>
> Where cleanupT is the Talitos descriptor cleanup routine and cleanupF
> is from fsldma. The assumption is that the interrupt kicks the
> cleanup routine and that calls dma_run_dependencies().
>

Hello Dan,

I guess it is not clear to driver authors that they should call
dma_run_dependencies() for each dma_async_tx_descriptor that is
processed. Without a careful re-reading of this email, I would not have
known. I guess anyone reviewing the driver missed it too. Judging by the
code in other drivers, it should be called immediately after calling the
callback function. The fsldma driver doesn't even call the function at
the moment.

To the people testing fsldma with talitos: you should probably try
adding a call to dma_run_dependencies() in the fsl_chan_ld_cleanup()
function. Then run your tests again, and see if the interoperability
problems are fixed.

I'm still working through a cleanup patch series. There are some places
where the locking doesn't seem right to me, and I'll be attempting to
fix those as I go through the driver.

Ira

2010-07-14 03:58:00

by hank peng

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] Crypto: Talitos: Support for Async_tx XOR offload

Hi:
I used your patch and test it on MPC8548 board. Today, I found a
problem when doing raid5 recovering.

talitos e0030000.crypto: master data transfer error
talitos e0030000.crypto: xor operation: talitos error -22
------------[ cut here ]------------
Kernel BUG at c02dcb6c [verbose debug info unavailable]
Oops: Exception in kernel mode, sig: 5 [#1]
MPC85xx CDS
Modules linked in: iscsi_trgt
NIP: c02dcb6c LR: c02dcb6c CTR: c023e7a8
REGS: e8b8d820 TRAP: 0700 Not tainted (2.6.31.6)
MSR: 00029000 <EE,ME,CE> CR: 22008022 XER: 20000000
TASK = ef8cb2e0[1897] 'md2_raid5' THREAD: e8b8c000
GPR00: c02dcb6c e8b8d8d0 ef8cb2e0 00000040 00007c99 ffffffff c023bccc 00007c5f
GPR08: c04a5c60 c049b5fc 00007c99 00004000 80008022 00000000 3fff5700 c3374754
GPR16: c337474c 00000000 00000000 ffffffea 00000001 00000000 000186a0 00000000
GPR24: ffffffea ffffffea ef81f850 0000000c 00029000 ffffffea ef81f850 efb1d580
NIP [c02dcb6c] talitos_release_xor+0xfc/0x104
LR [c02dcb6c] talitos_release_xor+0xfc/0x104
Call Trace:
[e8b8d8d0] [c02dcb6c] talitos_release_xor+0xfc/0x104 (unreliable)
[e8b8d8f0] [c02db928] flush_channel+0x11c/0x178
[e8b8d920] [c02dd344] talitos_interrupt+0x320/0x9e8
[e8b8d970] [c0060b3c] handle_IRQ_event+0x5c/0x140
[e8b8d990] [c006280c] handle_fasteoi_irq+0x68/0x118
[e8b8d9a0] [c0004f08] do_IRQ+0x94/0xb0
[e8b8d9c0] [c000fe00] ret_from_except+0x0/0x18
[e8b8da80] [c02b47a0] release_stripe+0x24/0x3c
[e8b8da90] [c02ba4ec] raid5_end_read_request+0x160/0x3f8
[e8b8dae0] [c00ba36c] bio_endio+0x48/0x6c
[e8b8daf0] [c01f80e0] req_bio_endio+0xa4/0x128
[e8b8db10] [c01f81f0] blk_update_request+0x8c/0x43c
[e8b8db40] [c01f85c0] blk_update_bidi_request+0x20/0x88
[e8b8db60] [c01f92c4] blk_end_bidi_request+0x1c/0x58
[e8b8db80] [c01f9314] blk_end_request+0x14/0x24
[e8b8db90] [c025ece0] scsi_io_completion+0x8c/0x4ac
[e8b8dbd0] [c0257fd0] scsi_finish_command+0xd0/0xf4
[e8b8dbf0] [c025f1c8] scsi_softirq_done+0xc8/0x150
[e8b8dc10] [c01fe114] blk_done_softirq+0x80/0xa0
[e8b8dc30] [c003a0d8] __do_softirq+0xa8/0x128
[e8b8dc70] [c0004e70] do_softirq+0x54/0x58
[e8b8dc80] [c0039f4c] irq_exit+0x94/0x98
[e8b8dc90] [c0004f0c] do_IRQ+0x98/0xb0
[e8b8dcb0] [c000fe00] ret_from_except+0x0/0x18
[e8b8dd70] [c00679c8] mempool_alloc_slab+0x1c/0x2c
[e8b8ddb0] [c02b6814] ops_run_io+0x1ac/0x2b8
[e8b8ddf0] [c02b93e4] handle_stripe5+0xa80/0x15c0
[e8b8de70] [c02bb408] handle_stripe+0x34/0x12d4
[e8b8df10] [c02bc8ec] raid5d+0x244/0x458
[e8b8df70] [c02c9624] md_thread+0x5c/0x124
[e8b8dfc0] [c004ab24] kthread+0x78/0x7c
[e8b8dff0] [c000f52c] kernel_thread+0x4c/0x68
Instruction dump:
bb61000c 38210020 7c0803a6 4bffefac 4bf63bc9 80be0008 7c641b78 3c60c042
7fa6eb78 38631ccc 4cc63182 4bd58b9d <0fe00000> 48000000 9421ffd0 7c0802a6
Kernel panic - not syncing: Fatal exception in interrupt
Rebooting in 1 seconds..
------------[ cut here ]------------
Badness at c0223124 [verbose debug info unavailable]
NIP: c0223124 LR: c0223308 CTR: 00000000
REGS: e8b8d580 TRAP: 0700 Tainted: G D (2.6.31.6)
MSR: 00021000 <ME,CE> CR: 82008088 XER: 20000000
TASK = ef8cb2e0[1897] 'md2_raid5' THREAD: e8b8c000
GPR00: 00000001 e8b8d630 ef8cb2e0 e84f1d60 00000000 e84f1d60 e84f1d7c c04a1364
GPR08: c04a5c60 e8b8c000 0000873b e8b8d650 82008088 00000000 3fff5700 c3374754
GPR16: c337474c 00000000 00000000 ffffffea 00000001 00000000 000186a0 00000000
GPR24: ffffffea ffffffea 00000000 ffffffff ffffffff 00000000 00001106 e84f1d60
NIP [c0223124] pci_get_dev_by_id+0x34/0x98
LR [c0223308] pci_get_subsys+0x64/0xa4
Call Trace:
[e8b8d630] [c021e378] no_pci_devices+0x34/0x50 (unreliable)
[e8b8d650] [c0223308] pci_get_subsys+0x64/0xa4
[e8b8d670] [c0017be0] mpc85xx_cds_restart+0x24/0x90
[e8b8d690] [c000e7d0] machine_restart+0x34/0x4c
[e8b8d6a0] [c00447e0] emergency_restart+0x14/0x24
[e8b8d6b0] [c003473c] panic+0x118/0x158
[e8b8d700] [c000cf24] die+0x160/0x16c
[e8b8d720] [c000d1b0] _exception+0x12c/0x154
[e8b8d810] [c000fdb4] ret_from_except_full+0x0/0x4c
[e8b8d8d0] [c02dcb6c] talitos_release_xor+0xfc/0x104
[e8b8d8f0] [c02db928] flush_channel+0x11c/0x178
[e8b8d920] [c02dd344] talitos_interrupt+0x320/0x9e8
[e8b8d970] [c0060b3c] handle_IRQ_event+0x5c/0x140
[e8b8d990] [c006280c] handle_fasteoi_irq+0x68/0x118
[e8b8d9a0] [c0004f08] do_IRQ+0x94/0xb0
[e8b8d9c0] [c000fe00] ret_from_except+0x0/0x18
[e8b8da80] [c02b47a0] release_stripe+0x24/0x3c
[e8b8da90] [c02ba4ec] raid5_end_read_request+0x160/0x3f8
[e8b8dae0] [c00ba36c] bio_endio+0x48/0x6c
[e8b8daf0] [c01f80e0] req_bio_endio+0xa4/0x128
[e8b8db10] [c01f81f0] blk_update_request+0x8c/0x43c
[e8b8db40] [c01f85c0] blk_update_bidi_request+0x20/0x88
[e8b8db60] [c01f92c4] blk_end_bidi_request+0x1c/0x58
[e8b8db80] [c01f9314] blk_end_request+0x14/0x24
[e8b8db90] [c025ece0] scsi_io_completion+0x8c/0x4ac
[e8b8dbd0] [c0257fd0] scsi_finish_command+0xd0/0xf4
[e8b8dbf0] [c025f1c8] scsi_softirq_done+0xc8/0x150
[e8b8dc10] [c01fe114] blk_done_softirq+0x80/0xa0
[e8b8dc30] [c003a0d8] __do_softirq+0xa8/0x128
[e8b8dc70] [c0004e70] do_softirq+0x54/0x58
[e8b8dc80] [c0039f4c] irq_exit+0x94/0x98
[e8b8dc90] [c0004f0c] do_IRQ+0x98/0xb0
[e8b8dcb0] [c000fe00] ret_from_except+0x0/0x18
[e8b8dd70] [c00679c8] mempool_alloc_slab+0x1c/0x2c
[e8b8ddb0] [c02b6814] ops_run_io+0x1ac/0x2b8
[e8b8ddf0] [c02b93e4] handle_stripe5+0xa80/0x15c0
[e8b8de70] [c02bb408] handle_stripe+0x34/0x12d4
[e8b8df10] [c02bc8ec] raid5d+0x244/0x458
[e8b8df70] [c02c9624] md_thread+0x5c/0x124
[e8b8dfc0] [c004ab24] kthread+0x78/0x7c
[e8b8dff0] [c000f52c] kernel_thread+0x4c/0x68
Instruction dump:
7c0802a6 7d800026 7c651b78 bf810010 54290024 90010024 7c9d2378 9181000c
8009000c 5400016e 7c0000d0 54000ffe <0f000000> 3b800000 2e040000 3cc0c022


This oops is invoked by the following code:

static void talitos_release_xor(struct device *dev, struct talitos_desc *hwdesc,
void *context, int error)
{
struct talitos_xor_desc *desc = context;
struct talitos_xor_chan *xor_chan;
dma_async_tx_callback callback;
void *callback_param;

if (unlikely(error)) {
dev_err(dev, "xor operation: talitos error %d\n", error);
BUG();
<-----------------------------------------------------------------
here
}

I wonder here why a tailitos error occured, BUG is invoked? I think we
can do something to error recovery other than invoking BUG.

2009/12/16 Vishnu Suresh <[email protected]>:
> Expose Talitos's XOR functionality to be used for
> RAID Parity calculation via the Async_tx layer.
>
> Known Issue:
> When used with fsldma, random crashes are observed
> on some platforms. Hence, inter-operability with fsldma
> is currently disabled
>
> Thanks to Surender Kumar and Lee Nipper for their help in
> realising this driver
>
> Signed-off-by: Kim Phillips <[email protected]>
> Signed-off-by: Dipen Dudhat <[email protected]>
> Signed-off-by: Maneesh Gupta <[email protected]>
> Signed-off-by: Vishnu Suresh <[email protected]>
> ---
> Changes with respect to v1 as per comments received
> o. Rebased to linux-next as of 20091216
> o. The selection is based exclusive of fsldma
> o. Intoduced a new Kernel Configuration variable
>   *. This enables selecting the Cryptographic functionality
>      of Talitos along with fsldma.
>   *. Disables the XOR parity calculation offload, if fsldma enabled
>      either as kernel in-built or as a module
>   *. Once the inter-operability with fsldma is resolved, this option
>      can be removed
>
>  drivers/crypto/Kconfig   |    9 +
>  drivers/crypto/talitos.c |  402 +++++++++++++++++++++++++++++++++++++++++++++-
>  drivers/crypto/talitos.h |    2 +
>  3 files changed, 412 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/crypto/Kconfig b/drivers/crypto/Kconfig
> index b08403d..f8a6376 100644
> --- a/drivers/crypto/Kconfig
> +++ b/drivers/crypto/Kconfig
> @@ -203,6 +203,15 @@ config CRYPTO_DEV_TALITOS
>          To compile this driver as a module, choose M here: the module
>          will be called talitos.
>
> +config CRYPTO_DEV_TALITOS_RAIDXOR
> +       bool "Talitos RAID5 XOR Calculation Offload"
> +       select DMA_ENGINE
> +       depends on CRYPTO_DEV_TALITOS
> +       depends on FSL_DMA=n
> +       help
> +         Say 'Y' here to use the Freescale Security Engine (SEC) to
> +         offload RAID XOR parity Calculation
> +
>  config CRYPTO_DEV_IXP4XX
>        tristate "Driver for IXP4xx crypto hardware acceleration"
>        depends on ARCH_IXP4XX
> diff --git a/drivers/crypto/talitos.c b/drivers/crypto/talitos.c
> index c47ffe8..e63b25a 100644
> --- a/drivers/crypto/talitos.c
> +++ b/drivers/crypto/talitos.c
> @@ -1,7 +1,7 @@
>  /*
>  * talitos - Freescale Integrated Security Engine (SEC) device driver
>  *
> - * Copyright (c) 2008 Freescale Semiconductor, Inc.
> + * Copyright (c) 2008-2009 Freescale Semiconductor, Inc.
>  *
>  * Scatterlist Crypto API glue code copied from files with the following:
>  * Copyright (c) 2006-2007 Herbert Xu <[email protected]>
> @@ -37,6 +37,8 @@
>  #include <linux/io.h>
>  #include <linux/spinlock.h>
>  #include <linux/rtnetlink.h>
> +#include <linux/dmaengine.h>
> +#include <linux/raid/xor.h>
>
>  #include <crypto/algapi.h>
>  #include <crypto/aes.h>
> @@ -140,6 +142,10 @@ struct talitos_private {
>
>        /* hwrng device */
>        struct hwrng rng;
> +#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
> +       /* XOR Device */
> +       struct dma_device dma_dev_common;
> +#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
>  };
>
>  /* .features flag */
> @@ -684,6 +690,375 @@ static void talitos_unregister_rng(struct device *dev)
>        hwrng_unregister(&priv->rng);
>  }
>
> +#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
> +/*
> + * async_tx interface for XOR-capable SECs
> + *
> + * Dipen Dudhat <[email protected]>
> + * Maneesh Gupta <[email protected]>
> + * Vishnu Suresh <[email protected]>
> + */
> +
> +/**
> + * talitos_xor_chan - context management for the async_tx channel
> + * @completed_cookie: the last completed cookie
> + * @desc_lock: lock for tx queue
> + * @total_desc: number of descriptors allocated
> + * @submit_q: queue of submitted descriptors
> + * @pending_q: queue of pending descriptors
> + * @in_progress_q: queue of descriptors in progress
> + * @free_desc: queue of unused descriptors
> + * @dev: talitos device implementing this channel
> + * @common: the corresponding xor channel in async_tx
> + */
> +struct talitos_xor_chan {
> +       dma_cookie_t completed_cookie;
> +       spinlock_t desc_lock;
> +       unsigned int total_desc;
> +       struct list_head submit_q;
> +       struct list_head pending_q;
> +       struct list_head in_progress_q;
> +       struct list_head free_desc;
> +       struct device *dev;
> +       struct dma_chan common;
> +};
> +
> +/**
> + * talitos_xor_desc - software xor descriptor
> + * @async_tx: the referring async_tx descriptor
> + * @node:
> + * @hwdesc: h/w descriptor
> + */
> +struct talitos_xor_desc {
> +       struct dma_async_tx_descriptor async_tx;
> +       struct list_head tx_list;
> +       struct list_head node;
> +       struct talitos_desc hwdesc;
> +};
> +
> +static enum dma_status talitos_is_tx_complete(struct dma_chan *chan,
> +                                             dma_cookie_t cookie,
> +                                             dma_cookie_t *done,
> +                                             dma_cookie_t *used)
> +{
> +       struct talitos_xor_chan *xor_chan;
> +       dma_cookie_t last_used;
> +       dma_cookie_t last_complete;
> +
> +       xor_chan = container_of(chan, struct talitos_xor_chan, common);
> +
> +       last_used = chan->cookie;
> +       last_complete = xor_chan->completed_cookie;
> +
> +       if (done)
> +               *done = last_complete;
> +
> +       if (used)
> +               *used = last_used;
> +
> +       return dma_async_is_complete(cookie, last_complete, last_used);
> +}
> +
> +static void talitos_release_xor(struct device *dev, struct talitos_desc *hwdesc,
> +                               void *context, int error)
> +{
> +       struct talitos_xor_desc *desc = context;
> +       struct talitos_xor_chan *xor_chan;
> +       dma_async_tx_callback callback;
> +       void *callback_param;
> +
> +       if (unlikely(error)) {
> +               dev_err(dev, "xor operation: talitos error %d\n", error);
> +               BUG();
> +       }
> +
> +       xor_chan = container_of(desc->async_tx.chan, struct talitos_xor_chan,
> +                               common);
> +       spin_lock_bh(&xor_chan->desc_lock);
> +       if (xor_chan->completed_cookie < desc->async_tx.cookie)
> +               xor_chan->completed_cookie = desc->async_tx.cookie;
> +
> +       callback = desc->async_tx.callback;
> +       callback_param = desc->async_tx.callback_param;
> +       list_del(&desc->node);
> +       list_add_tail(&desc->node, &xor_chan->free_desc);
> +       spin_unlock_bh(&xor_chan->desc_lock);
> +
> +       if (callback)
> +               callback(callback_param);
> +
> +       /* run dependent operations */
> +       dma_run_dependencies(&desc->async_tx);
> +}
> +
> +static void talitos_process_pending(struct talitos_xor_chan *xor_chan)
> +{
> +       struct talitos_xor_desc *desc, *_desc;
> +
> +       spin_lock_bh(&xor_chan->desc_lock);
> +       list_for_each_entry_safe(desc, _desc, &xor_chan->pending_q, node) {
> +               if (talitos_submit(xor_chan->dev, &desc->hwdesc,
> +                                  talitos_release_xor, desc) != -EINPROGRESS)
> +                       break;
> +
> +               list_del(&desc->node);
> +               list_add_tail(&desc->node, &xor_chan->in_progress_q);
> +       }
> +       spin_unlock_bh(&xor_chan->desc_lock);
> +}
> +
> +/**
> + * talitos_issue_pending - move the descriptors in submit
> + * queue to pending queue and submit them for processing
> + * @chan: DMA channel
> + */
> +static void talitos_issue_pending(struct dma_chan *chan)
> +{
> +       struct talitos_xor_chan *xor_chan;
> +
> +       xor_chan = container_of(chan, struct talitos_xor_chan, common);
> +       spin_lock_bh(&xor_chan->desc_lock);
> +       list_splice_tail_init(&xor_chan->submit_q,
> +                                &xor_chan->pending_q);
> +       spin_unlock_bh(&xor_chan->desc_lock);
> +       talitos_process_pending(xor_chan);
> +}
> +
> +static dma_cookie_t talitos_async_tx_submit(struct dma_async_tx_descriptor *tx)
> +{
> +       struct talitos_xor_desc *desc;
> +       struct talitos_xor_chan *xor_chan;
> +       dma_cookie_t cookie;
> +
> +       desc = container_of(tx, struct talitos_xor_desc, async_tx);
> +       xor_chan = container_of(tx->chan, struct talitos_xor_chan, common);
> +
> +       cookie = xor_chan->common.cookie + 1;
> +       if (cookie < 0)
> +               cookie = 1;
> +
> +       desc->async_tx.cookie = cookie;
> +       xor_chan->common.cookie = desc->async_tx.cookie;
> +
> +       list_splice_tail_init(&desc->tx_list,
> +                                &xor_chan->submit_q);
> +
> +       return cookie;
> +}
> +
> +static struct talitos_xor_desc *talitos_xor_alloc_descriptor(
> +                               struct talitos_xor_chan *xor_chan, gfp_t flags)
> +{
> +       struct talitos_xor_desc *desc;
> +
> +       desc = kmalloc(sizeof(*desc), flags);
> +       if (desc) {
> +               xor_chan->total_desc++;
> +               memset(desc, 0, sizeof(*desc));
> +               dma_async_tx_descriptor_init(&desc->async_tx, &xor_chan->common);
> +               desc->async_tx.tx_submit = talitos_async_tx_submit;
> +               INIT_LIST_HEAD(&desc->node);
> +               INIT_LIST_HEAD(&desc->tx_list);
> +       }
> +
> +       return desc;
> +}
> +
> +static void talitos_free_chan_resources(struct dma_chan *chan)
> +{
> +       struct talitos_xor_chan *xor_chan;
> +       struct talitos_xor_desc *desc, *_desc;
> +
> +       xor_chan = container_of(chan, struct talitos_xor_chan, common);
> +
> +       list_for_each_entry_safe(desc, _desc, &xor_chan->submit_q, node) {
> +               list_del(&desc->node);
> +               xor_chan->total_desc--;
> +               kfree(desc);
> +       }
> +       list_for_each_entry_safe(desc, _desc, &xor_chan->pending_q, node) {
> +               list_del(&desc->node);
> +               xor_chan->total_desc--;
> +               kfree(desc);
> +       }
> +       list_for_each_entry_safe(desc, _desc, &xor_chan->in_progress_q, node) {
> +               list_del(&desc->node);
> +               xor_chan->total_desc--;
> +               kfree(desc);
> +       }
> +       list_for_each_entry_safe(desc, _desc, &xor_chan->free_desc, node) {
> +               list_del(&desc->node);
> +               xor_chan->total_desc--;
> +               kfree(desc);
> +       }
> +       BUG_ON(unlikely(xor_chan->total_desc)); /* Some descriptor not freed? */
> +}
> +
> +static int talitos_alloc_chan_resources(struct dma_chan *chan)
> +{
> +       struct talitos_xor_chan *xor_chan;
> +       struct talitos_xor_desc *desc;
> +       LIST_HEAD(tmp_list);
> +       int i;
> +
> +       xor_chan = container_of(chan, struct talitos_xor_chan, common);
> +
> +       if (!list_empty(&xor_chan->free_desc))
> +               return xor_chan->total_desc;
> +
> +       /* 256 initial descriptors */
> +       for (i = 0; i < 256; i++) {
> +               desc = talitos_xor_alloc_descriptor(xor_chan, GFP_KERNEL);
> +               if (!desc) {
> +                       dev_err(xor_chan->common.device->dev,
> +                               "Only %d initial descriptors\n", i);
> +                       break;
> +               }
> +               list_add_tail(&desc->node, &tmp_list);
> +       }
> +
> +       if (!i)
> +               return -ENOMEM;
> +
> +       /* At least one desc is allocated */
> +       list_splice_init(&tmp_list, &xor_chan->free_desc);
> +
> +       return xor_chan->total_desc;
> +}
> +
> +static struct dma_async_tx_descriptor * talitos_prep_dma_xor(
> +                       struct dma_chan *chan, dma_addr_t dest, dma_addr_t *src,
> +                       unsigned int src_cnt, size_t len, unsigned long flags)
> +{
> +       struct talitos_xor_chan *xor_chan;
> +       struct talitos_xor_desc *new;
> +       struct talitos_desc *desc;
> +       int i, j;
> +
> +       BUG_ON(unlikely(len > TALITOS_MAX_DATA_LEN));
> +
> +       xor_chan = container_of(chan, struct talitos_xor_chan, common);
> +
> +       if (!list_empty(&xor_chan->free_desc)) {
> +               new = container_of(xor_chan->free_desc.next,
> +                                  struct talitos_xor_desc, node);
> +               list_del(&new->node);
> +       } else {
> +               new = talitos_xor_alloc_descriptor(xor_chan, GFP_KERNEL);
> +       }
> +
> +       if (!new) {
> +               dev_err(xor_chan->common.device->dev,
> +                       "No free memory for XOR DMA descriptor\n");
> +               return NULL;
> +       }
> +
> +       desc = &new->hwdesc;
> +       /* Set destination: Last pointer pair */
> +       to_talitos_ptr(&desc->ptr[6], dest);
> +       desc->ptr[6].len = cpu_to_be16(len);
> +       desc->ptr[6].j_extent = 0;
> +
> +       /* Set Sources: End loading from second-last pointer pair */
> +       for (i = 5, j = 0; (j < src_cnt) && (i > 0); i--, j++) {
> +               to_talitos_ptr(&desc->ptr[i], src[j]);
> +               desc->ptr[i].len = cpu_to_be16(len);
> +               desc->ptr[i].j_extent = 0;
> +       }
> +
> +       /*
> +        * documentation states first 0 ptr/len combo marks end of sources
> +        * yet device produces scatter boundary error unless all subsequent
> +        * sources are zeroed out
> +        */
> +       for (; i >= 0; i--) {
> +               to_talitos_ptr(&desc->ptr[i], 0);
> +               desc->ptr[i].len = 0;
> +               desc->ptr[i].j_extent = 0;
> +       }
> +
> +       desc->hdr = DESC_HDR_SEL0_AESU | DESC_HDR_MODE0_AESU_XOR
> +                   | DESC_HDR_TYPE_RAID_XOR;
> +
> +       list_add_tail(&new->node, &new->tx_list);
> +
> +       new->async_tx.flags = flags;
> +       new->async_tx.cookie = -EBUSY;
> +
> +       return &new->async_tx;
> +}
> +
> +static void talitos_unregister_async_xor(struct device *dev)
> +{
> +       struct talitos_private *priv = dev_get_drvdata(dev);
> +       struct talitos_xor_chan *xor_chan;
> +       struct dma_chan *chan;
> +
> +       if (priv->dma_dev_common.chancnt)
> +               dma_async_device_unregister(&priv->dma_dev_common);
> +
> +       list_for_each_entry(chan, &priv->dma_dev_common.channels, device_node) {
> +               xor_chan = container_of(chan, struct talitos_xor_chan, common);
> +               list_del(&chan->device_node);
> +               priv->dma_dev_common.chancnt--;
> +               kfree(xor_chan);
> +       }
> +}
> +
> +/**
> + * talitos_register_dma_async - Initialize the Freescale XOR ADMA device
> + * It is registered as a DMA device with the capability to perform
> + * XOR operation with the Async_tx layer.
> + * The various queues and channel resources are also allocated.
> + */
> +static int talitos_register_async_tx(struct device *dev, int max_xor_srcs)
> +{
> +       struct talitos_private *priv = dev_get_drvdata(dev);
> +       struct dma_device *dma_dev = &priv->dma_dev_common;
> +       struct talitos_xor_chan *xor_chan;
> +       int err;
> +
> +       xor_chan = kzalloc(sizeof(struct talitos_xor_chan), GFP_KERNEL);
> +       if (!xor_chan) {
> +               dev_err(dev, "unable to allocate xor channel\n");
> +               return -ENOMEM;
> +       }
> +
> +       dma_dev->dev = dev;
> +       dma_dev->device_alloc_chan_resources = talitos_alloc_chan_resources;
> +       dma_dev->device_free_chan_resources = talitos_free_chan_resources;
> +       dma_dev->device_prep_dma_xor = talitos_prep_dma_xor;
> +       dma_dev->max_xor = max_xor_srcs;
> +       dma_dev->device_is_tx_complete = talitos_is_tx_complete;
> +       dma_dev->device_issue_pending = talitos_issue_pending;
> +       INIT_LIST_HEAD(&dma_dev->channels);
> +       dma_cap_set(DMA_XOR, dma_dev->cap_mask);
> +
> +       xor_chan->dev = dev;
> +       xor_chan->common.device = dma_dev;
> +       xor_chan->total_desc = 0;
> +       INIT_LIST_HEAD(&xor_chan->submit_q);
> +       INIT_LIST_HEAD(&xor_chan->pending_q);
> +       INIT_LIST_HEAD(&xor_chan->in_progress_q);
> +       INIT_LIST_HEAD(&xor_chan->free_desc);
> +       spin_lock_init(&xor_chan->desc_lock);
> +
> +       list_add_tail(&xor_chan->common.device_node, &dma_dev->channels);
> +       dma_dev->chancnt++;
> +
> +       err = dma_async_device_register(dma_dev);
> +       if (err) {
> +               dev_err(dev, "Unable to register XOR with Async_tx\n");
> +               goto err_out;
> +       }
> +
> +       return err;
> +
> +err_out:
> +       talitos_unregister_async_xor(dev);
> +       return err;
> +}
> +#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
>  /*
>  * crypto alg
>  */
> @@ -1768,6 +2143,10 @@ static int talitos_remove(struct of_device *ofdev)
>        tasklet_kill(&priv->done_task);
>
>        iounmap(priv->reg);
> +#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
> +       if (priv->dma_dev_common.chancnt)
> +               talitos_unregister_async_xor(dev);
> +#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
>
>        dev_set_drvdata(dev, NULL);
>
> @@ -1926,6 +2305,27 @@ static int talitos_probe(struct of_device *ofdev,
>                        dev_info(dev, "hwrng\n");
>        }
>
> +#ifdef CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR
> +       /*
> +        * register with async_tx xor, if capable
> +        * SEC 2.x support up to 3 RAID sources,
> +        * SEC 3.x support up to 6
> +        */
> +       if (hw_supports(dev, DESC_HDR_SEL0_AESU | DESC_HDR_TYPE_RAID_XOR)) {
> +               int max_xor_srcs = 3;
> +               if (of_device_is_compatible(np, "fsl,sec3.0"))
> +                       max_xor_srcs = 6;
> +
> +               err = talitos_register_async_tx(dev, max_xor_srcs);
> +               if (err) {
> +                       dev_err(dev, "failed to register async_tx xor: %d\n",
> +                               err);
> +                       goto err_out;
> +               }
> +               dev_info(dev, "max_xor_srcs %d\n", max_xor_srcs);
> +       }
> +#endif /* CONFIG_CRYPTO_DEV_TALITOS_RAIDXOR */
> +
>        /* register crypto algorithms the device supports */
>        for (i = 0; i < ARRAY_SIZE(driver_algs); i++) {
>                if (hw_supports(dev, driver_algs[i].desc_hdr_template)) {
> diff --git a/drivers/crypto/talitos.h b/drivers/crypto/talitos.h
> index ff5a145..b6197bc 100644
> --- a/drivers/crypto/talitos.h
> +++ b/drivers/crypto/talitos.h
> @@ -155,6 +155,7 @@
>  /* primary execution unit mode (MODE0) and derivatives */
>  #define        DESC_HDR_MODE0_ENCRYPT          cpu_to_be32(0x00100000)
>  #define        DESC_HDR_MODE0_AESU_CBC         cpu_to_be32(0x00200000)
> +#define        DESC_HDR_MODE0_AESU_XOR         cpu_to_be32(0x0c600000)
>  #define        DESC_HDR_MODE0_DEU_CBC          cpu_to_be32(0x00400000)
>  #define        DESC_HDR_MODE0_DEU_3DES         cpu_to_be32(0x00200000)
>  #define        DESC_HDR_MODE0_MDEU_INIT        cpu_to_be32(0x01000000)
> @@ -202,6 +203,7 @@
>  #define DESC_HDR_TYPE_IPSEC_ESP                        cpu_to_be32(1 << 3)
>  #define DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU  cpu_to_be32(2 << 3)
>  #define DESC_HDR_TYPE_HMAC_SNOOP_NO_AFEU       cpu_to_be32(4 << 3)
> +#define DESC_HDR_TYPE_RAID_XOR                 cpu_to_be32(21 << 3)
>
>  /* link table extent field bits */
>  #define DESC_PTR_LNKTBL_JUMP                   0x80
> --
> 1.6.4.2
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to [email protected]
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>



--
The simplest is not all best but the best is surely the simplest!
_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://lists.ozlabs.org/listinfo/linuxppc-dev