linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v2 0/4] Asynchronous raid6 acceleration (part 2 of 3)
@ 2009-08-31 15:40 Dan Williams
  2009-08-31 15:40 ` [PATCH v2 1/4] iop-adma: cleanup iop_adma_run_tx_complete_actions Dan Williams
                   ` (3 more replies)
  0 siblings, 4 replies; 5+ messages in thread
From: Dan Williams @ 2009-08-31 15:40 UTC (permalink / raw)
  To: linux-arm-kernel

RAID6 acceleration support for the iop13xx ADMA engines.  The iop13xx platform
has three Application DMA engines providing support for offloading raid
operations.  Channel0 and Channel1 support memcpy and xor.  Channel2 adds pq
support.

The combined series is available at:

  git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx.git raid6

The changes since v1 are:
1/ Fix the case for continuing a p-disabled pq calculation
2/ A lockdep annotation for teaching lockdep about iop-adma cross channel
   dependency submissions.
---

Dan Williams (4):
      iop-adma: P+Q self test
      iop-adma: P+Q support for iop13xx adma engines
      iop-adma: fix lockdep false positive
      iop-adma: cleanup iop_adma_run_tx_complete_actions


 arch/arm/include/asm/hardware/iop3xx-adma.h |   76 +++++
 arch/arm/include/asm/hardware/iop_adma.h    |    1 
 arch/arm/mach-iop13xx/include/mach/adma.h   |  107 +++++++
 drivers/dma/iop-adma.c                      |  437 ++++++++++++++++++++++++---
 4 files changed, 577 insertions(+), 44 deletions(-)

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH v2 1/4] iop-adma: cleanup iop_adma_run_tx_complete_actions
  2009-08-31 15:40 [PATCH v2 0/4] Asynchronous raid6 acceleration (part 2 of 3) Dan Williams
@ 2009-08-31 15:40 ` Dan Williams
  2009-08-31 15:40 ` [PATCH v2 2/4] iop-adma: fix lockdep false positive Dan Williams
                   ` (2 subsequent siblings)
  3 siblings, 0 replies; 5+ messages in thread
From: Dan Williams @ 2009-08-31 15:40 UTC (permalink / raw)
  To: linux-arm-kernel

Replace 'desc->async_tx.' with 'tx->'

[ Impact: pure cleanup ]

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/dma/iop-adma.c |   19 ++++++++++---------
 1 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index 4496bc6..ce45f3f 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -61,17 +61,18 @@ static dma_cookie_t
 iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc,
 	struct iop_adma_chan *iop_chan, dma_cookie_t cookie)
 {
-	BUG_ON(desc->async_tx.cookie < 0);
-	if (desc->async_tx.cookie > 0) {
-		cookie = desc->async_tx.cookie;
-		desc->async_tx.cookie = 0;
+	struct dma_async_tx_descriptor *tx = &desc->async_tx;
+
+	BUG_ON(tx->cookie < 0);
+	if (tx->cookie > 0) {
+		cookie = tx->cookie;
+		tx->cookie = 0;
 
 		/* call the callback (must not sleep or submit new
 		 * operations to this channel)
 		 */
-		if (desc->async_tx.callback)
-			desc->async_tx.callback(
-				desc->async_tx.callback_param);
+		if (tx->callback)
+			tx->callback(tx->callback_param);
 
 		/* unmap dma addresses
 		 * (unmap_single vs unmap_page?)
@@ -81,7 +82,7 @@ iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc,
 			struct device *dev =
 				&iop_chan->device->pdev->dev;
 			u32 len = unmap->unmap_len;
-			enum dma_ctrl_flags flags = desc->async_tx.flags;
+			enum dma_ctrl_flags flags = tx->flags;
 			u32 src_cnt;
 			dma_addr_t addr;
 			dma_addr_t dest;
@@ -115,7 +116,7 @@ iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc,
 	}
 
 	/* run dependent operations */
-	dma_run_dependencies(&desc->async_tx);
+	dma_run_dependencies(tx);
 
 	return cookie;
 }

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH v2 2/4] iop-adma: fix lockdep false positive
  2009-08-31 15:40 [PATCH v2 0/4] Asynchronous raid6 acceleration (part 2 of 3) Dan Williams
  2009-08-31 15:40 ` [PATCH v2 1/4] iop-adma: cleanup iop_adma_run_tx_complete_actions Dan Williams
@ 2009-08-31 15:40 ` Dan Williams
  2009-08-31 15:40 ` [PATCH v2 3/4] iop-adma: P+Q support for iop13xx adma engines Dan Williams
  2009-08-31 15:40 ` [PATCH v2 4/4] iop-adma: P+Q self test Dan Williams
  3 siblings, 0 replies; 5+ messages in thread
From: Dan Williams @ 2009-08-31 15:40 UTC (permalink / raw)
  To: linux-arm-kernel

lockdep correctly identifies a potential recursive locking case for
iop_chan->lock, but in the dependency submission case we expect that the same
class will be acquired for both the parent dependency and the child channel.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/dma/iop-adma.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index ce45f3f..9c752bd 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -288,7 +288,12 @@ static void iop_adma_tasklet(unsigned long data)
 {
 	struct iop_adma_chan *iop_chan = (struct iop_adma_chan *) data;
 
-	spin_lock(&iop_chan->lock);
+	/* lockdep will flag depedency submissions as potentially
+	 * recursive locking, this is not the case as a dependency
+	 * submission will never recurse a channels submit routine.
+	 * There are checks in async_tx.c to prevent this.
+	 */
+	spin_lock_nested(&iop_chan->lock, SINGLE_DEPTH_NESTING);
 	__iop_adma_slot_cleanup(iop_chan);
 	spin_unlock(&iop_chan->lock);
 }

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH v2 3/4] iop-adma: P+Q support for iop13xx adma engines
  2009-08-31 15:40 [PATCH v2 0/4] Asynchronous raid6 acceleration (part 2 of 3) Dan Williams
  2009-08-31 15:40 ` [PATCH v2 1/4] iop-adma: cleanup iop_adma_run_tx_complete_actions Dan Williams
  2009-08-31 15:40 ` [PATCH v2 2/4] iop-adma: fix lockdep false positive Dan Williams
@ 2009-08-31 15:40 ` Dan Williams
  2009-08-31 15:40 ` [PATCH v2 4/4] iop-adma: P+Q self test Dan Williams
  3 siblings, 0 replies; 5+ messages in thread
From: Dan Williams @ 2009-08-31 15:40 UTC (permalink / raw)
  To: linux-arm-kernel

iop33x support is not included because that engine is a bit more awkward
to handle in that it can either be in xor mode or pq mode.  The
dmaengine/async_tx layers currently only comprehend static capabilities.

Note iop13xx does not support hardware PQ continuation so the driver
must handle the DMA_PREP_CONTINUE flag for operations across > 16
sources. From the comment for dma_maxpq:

/* When an engine does not support native continuation we need 3 extra
 * source slots to reuse P and Q with the following coefficients:
 * 1/ {00} * P : remove P from Q', but use it as a source for P'
 * 2/ {01} * Q : use Q to continue Q' calculation
 * 3/ {00} * Q : subtract Q from P' to cancel (2)
 */

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 arch/arm/include/asm/hardware/iop3xx-adma.h |   76 +++++++++
 arch/arm/include/asm/hardware/iop_adma.h    |    1 
 arch/arm/mach-iop13xx/include/mach/adma.h   |  107 +++++++++++++
 drivers/dma/iop-adma.c                      |  231 +++++++++++++++++++++++----
 4 files changed, 381 insertions(+), 34 deletions(-)

diff --git a/arch/arm/include/asm/hardware/iop3xx-adma.h b/arch/arm/include/asm/hardware/iop3xx-adma.h
index 26eefea..1a8c727 100644
--- a/arch/arm/include/asm/hardware/iop3xx-adma.h
+++ b/arch/arm/include/asm/hardware/iop3xx-adma.h
@@ -187,11 +187,74 @@ union iop3xx_desc {
 	void *ptr;
 };
 
+/* No support for p+q operations */
+static inline int
+iop_chan_pq_slot_count(size_t len, int src_cnt, int *slots_per_op)
+{
+	BUG();
+	return 0;
+}
+
+static inline void
+iop_desc_init_pq(struct iop_adma_desc_slot *desc, int src_cnt,
+		  unsigned long flags)
+{
+	BUG();
+}
+
+static inline void
+iop_desc_set_pq_addr(struct iop_adma_desc_slot *desc, dma_addr_t *addr)
+{
+	BUG();
+}
+
+static inline void
+iop_desc_set_pq_src_addr(struct iop_adma_desc_slot *desc, int src_idx,
+			 dma_addr_t addr, unsigned char coef)
+{
+	BUG();
+}
+
+static inline int
+iop_chan_pq_zero_sum_slot_count(size_t len, int src_cnt, int *slots_per_op)
+{
+	BUG();
+	return 0;
+}
+
+static inline void
+iop_desc_init_pq_zero_sum(struct iop_adma_desc_slot *desc, int src_cnt,
+			  unsigned long flags)
+{
+	BUG();
+}
+
+static inline void
+iop_desc_set_pq_zero_sum_byte_count(struct iop_adma_desc_slot *desc, u32 len)
+{
+	BUG();
+}
+
+#define iop_desc_set_pq_zero_sum_src_addr iop_desc_set_pq_src_addr
+
+static inline void
+iop_desc_set_pq_zero_sum_addr(struct iop_adma_desc_slot *desc, int pq_idx,
+			      dma_addr_t *src)
+{
+	BUG();
+}
+
 static inline int iop_adma_get_max_xor(void)
 {
 	return 32;
 }
 
+static inline int iop_adma_get_max_pq(void)
+{
+	BUG();
+	return 0;
+}
+
 static inline u32 iop_chan_get_current_descriptor(struct iop_adma_chan *chan)
 {
 	int id = chan->device->id;
@@ -332,6 +395,11 @@ static inline int iop_chan_zero_sum_slot_count(size_t len, int src_cnt,
 	return slot_cnt;
 }
 
+static inline int iop_desc_is_pq(struct iop_adma_desc_slot *desc)
+{
+	return 0;
+}
+
 static inline u32 iop_desc_get_dest_addr(struct iop_adma_desc_slot *desc,
 					struct iop_adma_chan *chan)
 {
@@ -349,6 +417,14 @@ static inline u32 iop_desc_get_dest_addr(struct iop_adma_desc_slot *desc,
 	return 0;
 }
 
+
+static inline u32 iop_desc_get_qdest_addr(struct iop_adma_desc_slot *desc,
+					  struct iop_adma_chan *chan)
+{
+	BUG();
+	return 0;
+}
+
 static inline u32 iop_desc_get_byte_count(struct iop_adma_desc_slot *desc,
 					struct iop_adma_chan *chan)
 {
diff --git a/arch/arm/include/asm/hardware/iop_adma.h b/arch/arm/include/asm/hardware/iop_adma.h
index 385c6e8..bbe8a04 100644
--- a/arch/arm/include/asm/hardware/iop_adma.h
+++ b/arch/arm/include/asm/hardware/iop_adma.h
@@ -106,6 +106,7 @@ struct iop_adma_desc_slot {
 	union {
 		u32 *xor_check_result;
 		u32 *crc32_result;
+		u32 *pq_check_result;
 	};
 };
 
diff --git a/arch/arm/mach-iop13xx/include/mach/adma.h b/arch/arm/mach-iop13xx/include/mach/adma.h
index 1cd31df..6d3782d 100644
--- a/arch/arm/mach-iop13xx/include/mach/adma.h
+++ b/arch/arm/mach-iop13xx/include/mach/adma.h
@@ -150,6 +150,8 @@ static inline int iop_adma_get_max_xor(void)
 	return 16;
 }
 
+#define iop_adma_get_max_pq iop_adma_get_max_xor
+
 static inline u32 iop_chan_get_current_descriptor(struct iop_adma_chan *chan)
 {
 	return __raw_readl(ADMA_ADAR(chan));
@@ -211,7 +213,10 @@ iop_chan_xor_slot_count(size_t len, int src_cnt, int *slots_per_op)
 #define IOP_ADMA_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT
 #define IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT
 #define IOP_ADMA_XOR_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT
+#define IOP_ADMA_PQ_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT
 #define iop_chan_zero_sum_slot_count(l, s, o) iop_chan_xor_slot_count(l, s, o)
+#define iop_chan_pq_slot_count iop_chan_xor_slot_count
+#define iop_chan_pq_zero_sum_slot_count iop_chan_xor_slot_count
 
 static inline u32 iop_desc_get_dest_addr(struct iop_adma_desc_slot *desc,
 					struct iop_adma_chan *chan)
@@ -220,6 +225,13 @@ static inline u32 iop_desc_get_dest_addr(struct iop_adma_desc_slot *desc,
 	return hw_desc->dest_addr;
 }
 
+static inline u32 iop_desc_get_qdest_addr(struct iop_adma_desc_slot *desc,
+					  struct iop_adma_chan *chan)
+{
+	struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc;
+	return hw_desc->q_dest_addr;
+}
+
 static inline u32 iop_desc_get_byte_count(struct iop_adma_desc_slot *desc,
 					struct iop_adma_chan *chan)
 {
@@ -319,6 +331,58 @@ iop_desc_init_zero_sum(struct iop_adma_desc_slot *desc, int src_cnt,
 	return 1;
 }
 
+static inline void
+iop_desc_init_pq(struct iop_adma_desc_slot *desc, int src_cnt,
+		  unsigned long flags)
+{
+	struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc;
+	union {
+		u32 value;
+		struct iop13xx_adma_desc_ctrl field;
+	} u_desc_ctrl;
+
+	u_desc_ctrl.value = 0;
+	u_desc_ctrl.field.src_select = src_cnt - 1;
+	u_desc_ctrl.field.xfer_dir = 3; /* local to internal bus */
+	u_desc_ctrl.field.pq_xfer_en = 1;
+	u_desc_ctrl.field.p_xfer_dis = !!(flags & DMA_PREP_PQ_DISABLE_P);
+	u_desc_ctrl.field.int_en = flags & DMA_PREP_INTERRUPT;
+	hw_desc->desc_ctrl = u_desc_ctrl.value;
+}
+
+static inline int iop_desc_is_pq(struct iop_adma_desc_slot *desc)
+{
+	struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc;
+	union {
+		u32 value;
+		struct iop13xx_adma_desc_ctrl field;
+	} u_desc_ctrl;
+
+	u_desc_ctrl.value = hw_desc->desc_ctrl;
+	return u_desc_ctrl.field.pq_xfer_en;
+}
+
+static inline void
+iop_desc_init_pq_zero_sum(struct iop_adma_desc_slot *desc, int src_cnt,
+			  unsigned long flags)
+{
+	struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc;
+	union {
+		u32 value;
+		struct iop13xx_adma_desc_ctrl field;
+	} u_desc_ctrl;
+
+	u_desc_ctrl.value = 0;
+	u_desc_ctrl.field.src_select = src_cnt - 1;
+	u_desc_ctrl.field.xfer_dir = 3; /* local to internal bus */
+	u_desc_ctrl.field.zero_result = 1;
+	u_desc_ctrl.field.status_write_back_en = 1;
+	u_desc_ctrl.field.pq_xfer_en = 1;
+	u_desc_ctrl.field.p_xfer_dis = !!(flags & DMA_PREP_PQ_DISABLE_P);
+	u_desc_ctrl.field.int_en = flags & DMA_PREP_INTERRUPT;
+	hw_desc->desc_ctrl = u_desc_ctrl.value;
+}
+
 static inline void iop_desc_set_byte_count(struct iop_adma_desc_slot *desc,
 					struct iop_adma_chan *chan,
 					u32 byte_count)
@@ -351,6 +415,7 @@ iop_desc_set_zero_sum_byte_count(struct iop_adma_desc_slot *desc, u32 len)
 	}
 }
 
+#define iop_desc_set_pq_zero_sum_byte_count iop_desc_set_zero_sum_byte_count
 
 static inline void iop_desc_set_dest_addr(struct iop_adma_desc_slot *desc,
 					struct iop_adma_chan *chan,
@@ -361,6 +426,16 @@ static inline void iop_desc_set_dest_addr(struct iop_adma_desc_slot *desc,
 	hw_desc->upper_dest_addr = 0;
 }
 
+static inline void
+iop_desc_set_pq_addr(struct iop_adma_desc_slot *desc, dma_addr_t *addr)
+{
+	struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc;
+
+	hw_desc->dest_addr = addr[0];
+	hw_desc->q_dest_addr = addr[1];
+	hw_desc->upper_dest_addr = 0;
+}
+
 static inline void iop_desc_set_memcpy_src_addr(struct iop_adma_desc_slot *desc,
 					dma_addr_t addr)
 {
@@ -389,6 +464,29 @@ static inline void iop_desc_set_xor_src_addr(struct iop_adma_desc_slot *desc,
 }
 
 static inline void
+iop_desc_set_pq_src_addr(struct iop_adma_desc_slot *desc, int src_idx,
+			 dma_addr_t addr, unsigned char coef)
+{
+	int slot_cnt = desc->slot_cnt, slots_per_op = desc->slots_per_op;
+	struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc, *iter;
+	struct iop13xx_adma_src *src;
+	int i = 0;
+
+	do {
+		iter = iop_hw_desc_slot_idx(hw_desc, i);
+		src = &iter->src[src_idx];
+		src->src_addr = addr;
+		src->pq_upper_src_addr = 0;
+		src->pq_dmlt = coef;
+		slot_cnt -= slots_per_op;
+		if (slot_cnt) {
+			i += slots_per_op;
+			addr += IOP_ADMA_PQ_MAX_BYTE_COUNT;
+		}
+	} while (slot_cnt);
+}
+
+static inline void
 iop_desc_init_interrupt(struct iop_adma_desc_slot *desc,
 	struct iop_adma_chan *chan)
 {
@@ -399,6 +497,15 @@ iop_desc_init_interrupt(struct iop_adma_desc_slot *desc,
 }
 
 #define iop_desc_set_zero_sum_src_addr iop_desc_set_xor_src_addr
+#define iop_desc_set_pq_zero_sum_src_addr iop_desc_set_pq_src_addr
+
+static inline void
+iop_desc_set_pq_zero_sum_addr(struct iop_adma_desc_slot *desc, int pq_idx,
+			      dma_addr_t *src)
+{
+	iop_desc_set_xor_src_addr(desc, pq_idx, src[pq_idx]);
+	iop_desc_set_xor_src_addr(desc, pq_idx+1, src[pq_idx+1]);
+}
 
 static inline void iop_desc_set_next_desc(struct iop_adma_desc_slot *desc,
 					u32 next_desc_addr)
diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index 9c752bd..5a0f4fe 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -57,6 +57,80 @@ static void iop_adma_free_slots(struct iop_adma_desc_slot *slot)
 	}
 }
 
+static void
+iop_desc_unmap(struct iop_adma_chan *iop_chan, struct iop_adma_desc_slot *desc)
+{
+	struct dma_async_tx_descriptor *tx = &desc->async_tx;
+	struct iop_adma_desc_slot *unmap = desc->group_head;
+	struct device *dev = &iop_chan->device->pdev->dev;
+	u32 len = unmap->unmap_len;
+	enum dma_ctrl_flags flags = tx->flags;
+	u32 src_cnt;
+	dma_addr_t addr;
+	dma_addr_t dest;
+
+	src_cnt = unmap->unmap_src_cnt;
+	dest = iop_desc_get_dest_addr(unmap, iop_chan);
+	if (!(flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
+		enum dma_data_direction dir;
+
+		if (src_cnt > 1) /* is xor? */
+			dir = DMA_BIDIRECTIONAL;
+		else
+			dir = DMA_FROM_DEVICE;
+
+		dma_unmap_page(dev, dest, len, dir);
+	}
+
+	if (!(flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
+		while (src_cnt--) {
+			addr = iop_desc_get_src_addr(unmap, iop_chan, src_cnt);
+			if (addr == dest)
+				continue;
+			dma_unmap_page(dev, addr, len, DMA_TO_DEVICE);
+		}
+	}
+	desc->group_head = NULL;
+}
+
+static void
+iop_desc_unmap_pq(struct iop_adma_chan *iop_chan, struct iop_adma_desc_slot *desc)
+{
+	struct dma_async_tx_descriptor *tx = &desc->async_tx;
+	struct iop_adma_desc_slot *unmap = desc->group_head;
+	struct device *dev = &iop_chan->device->pdev->dev;
+	u32 len = unmap->unmap_len;
+	enum dma_ctrl_flags flags = tx->flags;
+	u32 src_cnt = unmap->unmap_src_cnt;
+	dma_addr_t pdest = iop_desc_get_dest_addr(unmap, iop_chan);
+	dma_addr_t qdest = iop_desc_get_qdest_addr(unmap, iop_chan);
+	int i;
+
+	if (tx->flags & DMA_PREP_CONTINUE)
+		src_cnt -= 3;
+
+	if (!(flags & DMA_COMPL_SKIP_DEST_UNMAP) && !desc->pq_check_result) {
+		dma_unmap_page(dev, pdest, len, DMA_BIDIRECTIONAL);
+		dma_unmap_page(dev, qdest, len, DMA_BIDIRECTIONAL);
+	}
+
+	if (!(flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
+		dma_addr_t addr;
+
+		for (i = 0; i < src_cnt; i++) {
+			addr = iop_desc_get_src_addr(unmap, iop_chan, i);
+			dma_unmap_page(dev, addr, len, DMA_TO_DEVICE);
+		}
+		if (desc->pq_check_result) {
+			dma_unmap_page(dev, pdest, len, DMA_TO_DEVICE);
+			dma_unmap_page(dev, qdest, len, DMA_TO_DEVICE);
+		}
+	}
+
+	desc->group_head = NULL;
+}
+
+
 static dma_cookie_t
 iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc,
 	struct iop_adma_chan *iop_chan, dma_cookie_t cookie)
@@ -78,40 +152,10 @@ iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc,
 		 * (unmap_single vs unmap_page?)
 		 */
 		if (desc->group_head && desc->unmap_len) {
-			struct iop_adma_desc_slot *unmap = desc->group_head;
-			struct device *dev =
-				&iop_chan->device->pdev->dev;
-			u32 len = unmap->unmap_len;
-			enum dma_ctrl_flags flags = tx->flags;
-			u32 src_cnt;
-			dma_addr_t addr;
-			dma_addr_t dest;
-
-			src_cnt = unmap->unmap_src_cnt;
-			dest = iop_desc_get_dest_addr(unmap, iop_chan);
-			if (!(flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
-				enum dma_data_direction dir;
-
-				if (src_cnt > 1) /* is xor? */
-					dir = DMA_BIDIRECTIONAL;
-				else
-					dir = DMA_FROM_DEVICE;
-
-				dma_unmap_page(dev, dest, len, dir);
-			}
-
-			if (!(flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
-				while (src_cnt--) {
-					addr = iop_desc_get_src_addr(unmap,
-								     iop_chan,
-								     src_cnt);
-					if (addr == dest)
-						continue;
-					dma_unmap_page(dev, addr, len,
-						       DMA_TO_DEVICE);
-				}
-			}
-			desc->group_head = NULL;
+			if (iop_desc_is_pq(desc))
+				iop_desc_unmap_pq(iop_chan, desc);
+			else
+				iop_desc_unmap(iop_chan, desc);
 		}
 	}
 
@@ -702,6 +746,118 @@ iop_adma_prep_dma_xor_val(struct dma_chan *chan, dma_addr_t *dma_src,
 	return sw_desc ? &sw_desc->async_tx : NULL;
 }
 
+static struct dma_async_tx_descriptor *
+iop_adma_prep_dma_pq(struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
+		     unsigned int src_cnt, const unsigned char *scf, size_t len,
+		     unsigned long flags)
+{
+	struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan);
+	struct iop_adma_desc_slot *sw_desc, *g;
+	int slot_cnt, slots_per_op;
+	int continue_srcs;
+
+	if (unlikely(!len))
+		return NULL;
+	BUG_ON(len > IOP_ADMA_XOR_MAX_BYTE_COUNT);
+
+	dev_dbg(iop_chan->device->common.dev,
+		"%s src_cnt: %d len: %u flags: %lx\n",
+		__func__, src_cnt, len, flags);
+
+	if (dmaf_p_disabled_continue(flags))
+		continue_srcs = 1+src_cnt;
+	else if (dmaf_continue(flags))
+		continue_srcs = 3+src_cnt;
+	else
+		continue_srcs = 0+src_cnt;
+
+	spin_lock_bh(&iop_chan->lock);
+	slot_cnt = iop_chan_pq_slot_count(len, continue_srcs, &slots_per_op);
+	sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op);
+	if (sw_desc) {
+		int i;
+
+		g = sw_desc->group_head;
+		iop_desc_set_byte_count(g, iop_chan, len);
+
+		/* even if P is disabled its destination address (bits
+		 * [3:0]) must match Q.  It is ok if P points to an
+		 * invalid address, it won't be written.
+		 */
+		if (flags & DMA_PREP_PQ_DISABLE_P)
+			dst[0] = dst[1] & 0x7;
+
+		iop_desc_set_pq_addr(g, dst);
+		sw_desc->unmap_src_cnt = src_cnt;
+		sw_desc->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+		for (i = 0; i < src_cnt; i++)
+			iop_desc_set_pq_src_addr(g, i, src[i], scf[i]);
+
+		/* if we are continuing a previous operation factor in
+		 * the old p and q values, see the comment for dma_maxpq
+		 * in include/linux/dmaengine.h
+		 */
+		if (dmaf_p_disabled_continue(flags))
+			iop_desc_set_pq_src_addr(g, i++, dst[1], 1);
+		else if (dmaf_continue(flags)) {
+			iop_desc_set_pq_src_addr(g, i++, dst[0], 0);
+			iop_desc_set_pq_src_addr(g, i++, dst[1], 1);
+			iop_desc_set_pq_src_addr(g, i++, dst[1], 0);
+		}
+		iop_desc_init_pq(g, i, flags);
+	}
+	spin_unlock_bh(&iop_chan->lock);
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+static struct dma_async_tx_descriptor *
+iop_adma_prep_dma_pq_val(struct dma_chan *chan, dma_addr_t *pq, dma_addr_t *src,
+			 unsigned int src_cnt, const unsigned char *scf,
+			 size_t len, enum sum_check_flags *pqres,
+			 unsigned long flags)
+{
+	struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan);
+	struct iop_adma_desc_slot *sw_desc, *g;
+	int slot_cnt, slots_per_op;
+
+	if (unlikely(!len))
+		return NULL;
+	BUG_ON(len > IOP_ADMA_XOR_MAX_BYTE_COUNT);
+
+	dev_dbg(iop_chan->device->common.dev, "%s src_cnt: %d len: %u\n",
+		__func__, src_cnt, len);
+
+	spin_lock_bh(&iop_chan->lock);
+	slot_cnt = iop_chan_pq_zero_sum_slot_count(len, src_cnt + 2, &slots_per_op);
+	sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op);
+	if (sw_desc) {
+		/* for validate operations p and q are tagged onto the
+		 * end of the source list
+		 */
+		int pq_idx = src_cnt;
+
+		g = sw_desc->group_head;
+		iop_desc_init_pq_zero_sum(g, src_cnt+2, flags);
+		iop_desc_set_pq_zero_sum_byte_count(g, len);
+		g->pq_check_result = pqres;
+		pr_debug("\t%s: g->pq_check_result: %p\n",
+			__func__, g->pq_check_result);
+		sw_desc->unmap_src_cnt = src_cnt+2;
+		sw_desc->unmap_len = len;
+		sw_desc->async_tx.flags = flags;
+		while (src_cnt--)
+			iop_desc_set_pq_zero_sum_src_addr(g, src_cnt,
+							  src[src_cnt],
+							  scf[src_cnt]);
+		iop_desc_set_pq_zero_sum_addr(g, pq_idx, src);
+	}
+	spin_unlock_bh(&iop_chan->lock);
+
+	return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
 static void iop_adma_free_chan_resources(struct dma_chan *chan)
 {
 	struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan);
@@ -1201,6 +1357,13 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 	if (dma_has_cap(DMA_XOR_VAL, dma_dev->cap_mask))
 		dma_dev->device_prep_dma_xor_val =
 			iop_adma_prep_dma_xor_val;
+	if (dma_has_cap(DMA_PQ, dma_dev->cap_mask)) {
+		dma_set_maxpq(dma_dev, iop_adma_get_max_pq(), 0);
+		dma_dev->device_prep_dma_pq = iop_adma_prep_dma_pq;
+	}
+	if (dma_has_cap(DMA_PQ_VAL, dma_dev->cap_mask))
+		dma_dev->device_prep_dma_pq_val =
+			iop_adma_prep_dma_pq_val;
 	if (dma_has_cap(DMA_INTERRUPT, dma_dev->cap_mask))
 		dma_dev->device_prep_dma_interrupt =
 			iop_adma_prep_dma_interrupt;

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH v2 4/4] iop-adma: P+Q self test
  2009-08-31 15:40 [PATCH v2 0/4] Asynchronous raid6 acceleration (part 2 of 3) Dan Williams
                   ` (2 preceding siblings ...)
  2009-08-31 15:40 ` [PATCH v2 3/4] iop-adma: P+Q support for iop13xx adma engines Dan Williams
@ 2009-08-31 15:40 ` Dan Williams
  3 siblings, 0 replies; 5+ messages in thread
From: Dan Williams @ 2009-08-31 15:40 UTC (permalink / raw)
  To: linux-arm-kernel

Even though the intent is to extend dmatest with P+Q tests there is
still value in having an always-on sanity check to prevent an
unintentionally broken driver from registering.

This depends on raid6_pq.ko for verification, the side effect being that
PQ capable channels will fail to register when raid6 is disabled.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/dma/iop-adma.c |  182 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 181 insertions(+), 1 deletions(-)

diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index 5a0f4fe..f4c59e5 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -31,6 +31,7 @@
 #include <linux/platform_device.h>
 #include <linux/memory.h>
 #include <linux/ioport.h>
+#include <linux/raid/pq.h>
 
 #include <mach/adma.h>
 
@@ -1267,6 +1268,170 @@ out:
 	return err;
 }
 
+#ifdef CONFIG_MD_RAID6_PQ
+static int __devinit
+iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
+{
+	/* combined sources, software pq results, and extra hw pq results */
+	struct page *pq[IOP_ADMA_NUM_SRC_TEST+2+2];
+	/* ptr to the extra hw pq buffers defined above */
+	struct page **pq_hw = &pq[IOP_ADMA_NUM_SRC_TEST+2];
+	/* address conversion buffers (dma_map / page_address) */
+	void *pq_sw[IOP_ADMA_NUM_SRC_TEST+2];
+	dma_addr_t pq_src[IOP_ADMA_NUM_SRC_TEST];
+	dma_addr_t pq_dest[2];
+
+	int i;
+	struct dma_async_tx_descriptor *tx;
+	struct dma_chan *dma_chan;
+	dma_cookie_t cookie;
+	u32 zero_sum_result;
+	int err = 0;
+	struct device *dev;
+
+	dev_dbg(device->common.dev, "%s\n", __func__);
+
+	for (i = 0; i < ARRAY_SIZE(pq); i++) {
+		pq[i] = alloc_page(GFP_KERNEL);
+		if (!pq[i]) {
+			while (i--)
+				__free_page(pq[i]);
+			return -ENOMEM;
+		}
+	}
+
+	/* Fill in src buffers */
+	for (i = 0; i < IOP_ADMA_NUM_SRC_TEST; i++) {
+		pq_sw[i] = page_address(pq[i]);
+		memset(pq_sw[i], 0x11111111 * (1<<i), PAGE_SIZE);
+	}
+	pq_sw[i] = page_address(pq[i]);
+	pq_sw[i+1] = page_address(pq[i+1]);
+
+	dma_chan = container_of(device->common.channels.next,
+				struct dma_chan,
+				device_node);
+	if (iop_adma_alloc_chan_resources(dma_chan) < 1) {
+		err = -ENODEV;
+		goto out;
+	}
+
+	dev = dma_chan->device->dev;
+
+	/* initialize the dests */
+	memset(page_address(pq_hw[0]), 0 , PAGE_SIZE);
+	memset(page_address(pq_hw[1]), 0 , PAGE_SIZE);
+
+	/* test pq */
+	pq_dest[0] = dma_map_page(dev, pq_hw[0], 0, PAGE_SIZE, DMA_FROM_DEVICE);
+	pq_dest[1] = dma_map_page(dev, pq_hw[1], 0, PAGE_SIZE, DMA_FROM_DEVICE);
+	for (i = 0; i < IOP_ADMA_NUM_SRC_TEST; i++)
+		pq_src[i] = dma_map_page(dev, pq[i], 0, PAGE_SIZE,
+					 DMA_TO_DEVICE);
+
+	tx = iop_adma_prep_dma_pq(dma_chan, pq_dest, pq_src,
+				  IOP_ADMA_NUM_SRC_TEST, (u8 *)raid6_gfexp,
+				  PAGE_SIZE,
+				  DMA_PREP_INTERRUPT |
+				  DMA_CTRL_ACK);
+
+	cookie = iop_adma_tx_submit(tx);
+	iop_adma_issue_pending(dma_chan);
+	msleep(8);
+
+	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+		DMA_SUCCESS) {
+		dev_err(dev, "Self-test pq timed out, disabling\n");
+		err = -ENODEV;
+		goto free_resources;
+	}
+
+	raid6_call.gen_syndrome(IOP_ADMA_NUM_SRC_TEST+2, PAGE_SIZE, pq_sw);
+
+	if (memcmp(pq_sw[IOP_ADMA_NUM_SRC_TEST],
+		   page_address(pq_hw[0]), PAGE_SIZE) != 0) {
+		dev_err(dev, "Self-test p failed compare, disabling\n");
+		err = -ENODEV;
+		goto free_resources;
+	}
+	if (memcmp(pq_sw[IOP_ADMA_NUM_SRC_TEST+1],
+		   page_address(pq_hw[1]), PAGE_SIZE) != 0) {
+		dev_err(dev, "Self-test q failed compare, disabling\n");
+		err = -ENODEV;
+		goto free_resources;
+	}
+
+	/* test correct zero sum using the software generated pq values */
+	for (i = 0; i < IOP_ADMA_NUM_SRC_TEST + 2; i++)
+		pq_src[i] = dma_map_page(dev, pq[i], 0, PAGE_SIZE,
+					 DMA_TO_DEVICE);
+
+	zero_sum_result = ~0;
+	tx = iop_adma_prep_dma_pq_val(dma_chan, &pq_src[IOP_ADMA_NUM_SRC_TEST],
+				      pq_src, IOP_ADMA_NUM_SRC_TEST,
+				      raid6_gfexp, PAGE_SIZE, &zero_sum_result,
+				      DMA_PREP_INTERRUPT|DMA_CTRL_ACK);
+
+	cookie = iop_adma_tx_submit(tx);
+	iop_adma_issue_pending(dma_chan);
+	msleep(8);
+
+	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+		DMA_SUCCESS) {
+		dev_err(dev, "Self-test pq-zero-sum timed out, disabling\n");
+		err = -ENODEV;
+		goto free_resources;
+	}
+
+	if (zero_sum_result != 0) {
+		dev_err(dev, "Self-test pq-zero-sum failed to validate: %x\n",
+			zero_sum_result);
+		err = -ENODEV;
+		goto free_resources;
+	}
+
+	/* test incorrect zero sum */
+	i = IOP_ADMA_NUM_SRC_TEST;
+	memset(pq_sw[i] + 100, 0, 100);
+	memset(pq_sw[i+1] + 200, 0, 200);
+	for (i = 0; i < IOP_ADMA_NUM_SRC_TEST + 2; i++)
+		pq_src[i] = dma_map_page(dev, pq[i], 0, PAGE_SIZE,
+					 DMA_TO_DEVICE);
+
+	zero_sum_result = 0;
+	tx = iop_adma_prep_dma_pq_val(dma_chan, &pq_src[IOP_ADMA_NUM_SRC_TEST],
+				      pq_src, IOP_ADMA_NUM_SRC_TEST,
+				      raid6_gfexp, PAGE_SIZE, &zero_sum_result,
+				      DMA_PREP_INTERRUPT|DMA_CTRL_ACK);
+
+	cookie = iop_adma_tx_submit(tx);
+	iop_adma_issue_pending(dma_chan);
+	msleep(8);
+
+	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+		DMA_SUCCESS) {
+		dev_err(dev, "Self-test !pq-zero-sum timed out, disabling\n");
+		err = -ENODEV;
+		goto free_resources;
+	}
+
+	if (zero_sum_result != (SUM_CHECK_P_RESULT | SUM_CHECK_Q_RESULT)) {
+		dev_err(dev, "Self-test !pq-zero-sum failed to validate: %x\n",
+			zero_sum_result);
+		err = -ENODEV;
+		goto free_resources;
+	}
+
+free_resources:
+	iop_adma_free_chan_resources(dma_chan);
+out:
+	i = ARRAY_SIZE(pq);
+	while (i--)
+		__free_page(pq[i]);
+	return err;
+}
+#endif
+
 static int __devexit iop_adma_remove(struct platform_device *dev)
 {
 	struct iop_adma_device *device = platform_get_drvdata(dev);
@@ -1417,13 +1582,28 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 	}
 
 	if (dma_has_cap(DMA_XOR, dma_dev->cap_mask) ||
-		dma_has_cap(DMA_MEMSET, dma_dev->cap_mask)) {
+	    dma_has_cap(DMA_MEMSET, dma_dev->cap_mask)) {
 		ret = iop_adma_xor_val_self_test(adev);
 		dev_dbg(&pdev->dev, "xor self test returned %d\n", ret);
 		if (ret)
 			goto err_free_iop_chan;
 	}
 
+	if (dma_has_cap(DMA_PQ, dma_dev->cap_mask) &&
+	    dma_has_cap(DMA_PQ_VAL, dma_dev->cap_mask)) {
+		#ifdef CONFIG_MD_RAID6_PQ
+		ret = iop_adma_pq_zero_sum_self_test(adev);
+		dev_dbg(&pdev->dev, "pq self test returned %d\n", ret);
+		#else
+		/* can not test raid6, so do not publish capability */
+		dma_cap_clear(DMA_PQ, dma_dev->cap_mask);
+		dma_cap_clear(DMA_PQ_VAL, dma_dev->cap_mask);
+		ret = 0;
+		#endif
+		if (ret)
+			goto err_free_iop_chan;
+	}
+
 	dev_printk(KERN_INFO, &pdev->dev, "Intel(R) IOP: "
 	  "( %s%s%s%s%s%s%s%s%s%s)\n",
 	  dma_has_cap(DMA_PQ, dma_dev->cap_mask) ? "pq " : "",

^ permalink raw reply related	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2009-08-31 15:40 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-08-31 15:40 [PATCH v2 0/4] Asynchronous raid6 acceleration (part 2 of 3) Dan Williams
2009-08-31 15:40 ` [PATCH v2 1/4] iop-adma: cleanup iop_adma_run_tx_complete_actions Dan Williams
2009-08-31 15:40 ` [PATCH v2 2/4] iop-adma: fix lockdep false positive Dan Williams
2009-08-31 15:40 ` [PATCH v2 3/4] iop-adma: P+Q support for iop13xx adma engines Dan Williams
2009-08-31 15:40 ` [PATCH v2 4/4] iop-adma: P+Q self test Dan Williams

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).