public inbox for linux-scsi@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw()
@ 2015-04-05 14:59 Akinobu Mita
  2015-04-05 14:59 ` [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_* Akinobu Mita
                   ` (2 more replies)
  0 siblings, 3 replies; 9+ messages in thread
From: Akinobu Mita @ 2015-04-05 14:59 UTC (permalink / raw)
  To: target-devel
  Cc: Akinobu Mita, Nicholas Bellinger, Sagi Grimberg,
	Martin K. Petersen, Christoph Hellwig, James E.J. Bottomley,
	linux-scsi

Factor out code duplication in rd_execute_rw() into a helper function
rd_do_prot_rw().  This change is required to minimize the forthcoming
fix in rd_do_prot_rw().

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Nicholas Bellinger <nab@linux-iscsi.org>
Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
Cc: target-devel@vger.kernel.org
Cc: linux-scsi@vger.kernel.org
---
* v2
- Pass dif_verify() function pointer to helper function instead of is_write,
  suggested by Sagi Grimberg.

 drivers/target/target_core_rd.c | 66 ++++++++++++++++++++---------------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
index 98e83ac..ac5e8d2 100644
--- a/drivers/target/target_core_rd.c
+++ b/drivers/target/target_core_rd.c
@@ -382,6 +382,36 @@ static struct rd_dev_sg_table *rd_get_prot_table(struct rd_dev *rd_dev, u32 page
 	return NULL;
 }
 
+typedef sense_reason_t (*dif_verify)(struct se_cmd *, sector_t, unsigned int,
+				     unsigned int, struct scatterlist *, int);
+
+static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
+{
+	struct se_device *se_dev = cmd->se_dev;
+	struct rd_dev *dev = RD_DEV(se_dev);
+	struct rd_dev_sg_table *prot_table;
+	struct scatterlist *prot_sg;
+	u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
+	u32 prot_offset, prot_page;
+	u64 tmp;
+	sense_reason_t rc;
+
+	tmp = cmd->t_task_lba * se_dev->prot_length;
+	prot_offset = do_div(tmp, PAGE_SIZE);
+	prot_page = tmp;
+
+	prot_table = rd_get_prot_table(dev, prot_page);
+	if (!prot_table)
+		return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
+
+	prot_sg = &prot_table->sg_table[prot_page -
+					prot_table->page_start_offset];
+
+	rc = dif_verify(cmd, cmd->t_task_lba, sectors, 0, prot_sg, prot_offset);
+
+	return rc;
+}
+
 static sense_reason_t
 rd_execute_rw(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 	      enum dma_data_direction data_direction)
@@ -420,23 +450,7 @@ rd_execute_rw(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 			cmd->t_task_lba, rd_size, rd_page, rd_offset);
 
 	if (cmd->prot_type && data_direction == DMA_TO_DEVICE) {
-		struct rd_dev_sg_table *prot_table;
-		struct scatterlist *prot_sg;
-		u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
-		u32 prot_offset, prot_page;
-
-		tmp = cmd->t_task_lba * se_dev->prot_length;
-		prot_offset = do_div(tmp, PAGE_SIZE);
-		prot_page = tmp;
-
-		prot_table = rd_get_prot_table(dev, prot_page);
-		if (!prot_table)
-			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
-
-		prot_sg = &prot_table->sg_table[prot_page - prot_table->page_start_offset];
-
-		rc = sbc_dif_verify_write(cmd, cmd->t_task_lba, sectors, 0,
-					  prot_sg, prot_offset);
+		rc = rd_do_prot_rw(cmd, sbc_dif_verify_write);
 		if (rc)
 			return rc;
 	}
@@ -503,23 +517,7 @@ rd_execute_rw(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
 	sg_miter_stop(&m);
 
 	if (cmd->prot_type && data_direction == DMA_FROM_DEVICE) {
-		struct rd_dev_sg_table *prot_table;
-		struct scatterlist *prot_sg;
-		u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
-		u32 prot_offset, prot_page;
-
-		tmp = cmd->t_task_lba * se_dev->prot_length;
-		prot_offset = do_div(tmp, PAGE_SIZE);
-		prot_page = tmp;
-
-		prot_table = rd_get_prot_table(dev, prot_page);
-		if (!prot_table)
-			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
-
-		prot_sg = &prot_table->sg_table[prot_page - prot_table->page_start_offset];
-
-		rc = sbc_dif_verify_read(cmd, cmd->t_task_lba, sectors, 0,
-					 prot_sg, prot_offset);
+		rc = rd_do_prot_rw(cmd, sbc_dif_verify_read);
 		if (rc)
 			return rc;
 	}
-- 
1.9.1

^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_*
  2015-04-05 14:59 [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Akinobu Mita
@ 2015-04-05 14:59 ` Akinobu Mita
  2015-04-06  8:42   ` Sagi Grimberg
  2015-04-08  5:13   ` Nicholas A. Bellinger
  2015-04-06  7:43 ` [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Sagi Grimberg
  2015-04-08  5:03 ` Nicholas A. Bellinger
  2 siblings, 2 replies; 9+ messages in thread
From: Akinobu Mita @ 2015-04-05 14:59 UTC (permalink / raw)
  To: target-devel
  Cc: Akinobu Mita, Nicholas Bellinger, Sagi Grimberg,
	Martin K. Petersen, Christoph Hellwig, James E.J. Bottomley,
	linux-scsi

The scatterlist for protection information which is passed to
sbc_dif_verify_read() or sbc_dif_verify_write() requires that
neighboring scatterlist entries are contiguous or chained so that they
can be iterated by sg_next().

However, the protection information for RD-MCP backends could be located
in the multiple scatterlist arrays when the ramdisk space is too large.
So if the read/write request straddles this boundary, sbc_dif_verify_read()
or sbc_dif_verify_write() can't iterate all scatterlist entries.

This fixes it by allocating temporary scatterlist if it is needed.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Nicholas Bellinger <nab@linux-iscsi.org>
Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
Cc: target-devel@vger.kernel.org
Cc: linux-scsi@vger.kernel.org
---
* No change from v1

 drivers/target/target_core_rd.c | 39 +++++++++++++++++++++++++++++++++++----
 1 file changed, 35 insertions(+), 4 deletions(-)

diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
index ac5e8d2..6e25eaa 100644
--- a/drivers/target/target_core_rd.c
+++ b/drivers/target/target_core_rd.c
@@ -390,11 +390,12 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
 	struct se_device *se_dev = cmd->se_dev;
 	struct rd_dev *dev = RD_DEV(se_dev);
 	struct rd_dev_sg_table *prot_table;
+	bool need_to_release = false;
 	struct scatterlist *prot_sg;
 	u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
-	u32 prot_offset, prot_page;
+	u32 prot_offset, prot_page, prot_npages;
 	u64 tmp;
-	sense_reason_t rc;
+	sense_reason_t rc = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 
 	tmp = cmd->t_task_lba * se_dev->prot_length;
 	prot_offset = do_div(tmp, PAGE_SIZE);
@@ -404,10 +405,40 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
 	if (!prot_table)
 		return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 
-	prot_sg = &prot_table->sg_table[prot_page -
-					prot_table->page_start_offset];
+	prot_npages = DIV_ROUND_UP(prot_offset + sectors * se_dev->prot_length,
+				   PAGE_SIZE);
+
+	/* prot pages straddles multiple scatterlist tables */
+	if (prot_table->page_end_offset < prot_page + prot_npages - 1) {
+		int i;
+
+		prot_sg = kcalloc(prot_npages, sizeof(*prot_sg), GFP_KERNEL);
+		if (!prot_sg)
+			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
+
+		need_to_release = true;
+		sg_init_table(prot_sg, prot_npages);
+
+		for (i = 0; i < prot_npages; i++) {
+			if (prot_page + i > prot_table->page_end_offset) {
+				prot_table = rd_get_prot_table(dev,
+								prot_page + i);
+				if (!prot_table)
+					goto out;
+				sg_unmark_end(&prot_sg[i - 1]);
+			}
+			prot_sg[i] = prot_table->sg_table[prot_page + i -
+						prot_table->page_start_offset];
+		}
+	} else {
+		prot_sg = &prot_table->sg_table[prot_page -
+						prot_table->page_start_offset];
+	}
 
 	rc = dif_verify(cmd, cmd->t_task_lba, sectors, 0, prot_sg, prot_offset);
+out:
+	if (need_to_release)
+		kfree(prot_sg);
 
 	return rc;
 }
-- 
1.9.1

^ permalink raw reply related	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw()
  2015-04-05 14:59 [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Akinobu Mita
  2015-04-05 14:59 ` [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_* Akinobu Mita
@ 2015-04-06  7:43 ` Sagi Grimberg
  2015-04-06 11:16   ` Akinobu Mita
  2015-04-08  5:03 ` Nicholas A. Bellinger
  2 siblings, 1 reply; 9+ messages in thread
From: Sagi Grimberg @ 2015-04-06  7:43 UTC (permalink / raw)
  To: Akinobu Mita, target-devel
  Cc: Nicholas Bellinger, Martin K. Petersen, Christoph Hellwig,
	James E.J. Bottomley, linux-scsi

On 4/5/2015 5:59 PM, Akinobu Mita wrote:
> Factor out code duplication in rd_execute_rw() into a helper function
> rd_do_prot_rw().  This change is required to minimize the forthcoming
> fix in rd_do_prot_rw().
>
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
> Cc: target-devel@vger.kernel.org
> Cc: linux-scsi@vger.kernel.org
> ---
> * v2
> - Pass dif_verify() function pointer to helper function instead of is_write,
>    suggested by Sagi Grimberg.
>
>   drivers/target/target_core_rd.c | 66 ++++++++++++++++++++---------------------
>   1 file changed, 32 insertions(+), 34 deletions(-)
>
> diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
> index 98e83ac..ac5e8d2 100644
> --- a/drivers/target/target_core_rd.c
> +++ b/drivers/target/target_core_rd.c
> @@ -382,6 +382,36 @@ static struct rd_dev_sg_table *rd_get_prot_table(struct rd_dev *rd_dev, u32 page
>   	return NULL;
>   }
>
> +typedef sense_reason_t (*dif_verify)(struct se_cmd *, sector_t, unsigned int,
> +				     unsigned int, struct scatterlist *, int);
> +
> +static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
> +{
> +	struct se_device *se_dev = cmd->se_dev;
> +	struct rd_dev *dev = RD_DEV(se_dev);
> +	struct rd_dev_sg_table *prot_table;
> +	struct scatterlist *prot_sg;
> +	u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
> +	u32 prot_offset, prot_page;
> +	u64 tmp;
> +	sense_reason_t rc;
> +
> +	tmp = cmd->t_task_lba * se_dev->prot_length;
> +	prot_offset = do_div(tmp, PAGE_SIZE);
> +	prot_page = tmp;
> +
> +	prot_table = rd_get_prot_table(dev, prot_page);
> +	if (!prot_table)
> +		return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> +
> +	prot_sg = &prot_table->sg_table[prot_page -
> +					prot_table->page_start_offset];
> +
> +	rc = dif_verify(cmd, cmd->t_task_lba, sectors, 0, prot_sg, prot_offset);
> +
> +	return rc;

Nit, Given there is no action on the returned rc, this can be reduced to:
	return dif_verify();

> +}
> +
>   static sense_reason_t
>   rd_execute_rw(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
>   	      enum dma_data_direction data_direction)
> @@ -420,23 +450,7 @@ rd_execute_rw(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
>   			cmd->t_task_lba, rd_size, rd_page, rd_offset);
>
>   	if (cmd->prot_type && data_direction == DMA_TO_DEVICE) {
> -		struct rd_dev_sg_table *prot_table;
> -		struct scatterlist *prot_sg;
> -		u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
> -		u32 prot_offset, prot_page;
> -
> -		tmp = cmd->t_task_lba * se_dev->prot_length;
> -		prot_offset = do_div(tmp, PAGE_SIZE);
> -		prot_page = tmp;
> -
> -		prot_table = rd_get_prot_table(dev, prot_page);
> -		if (!prot_table)
> -			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> -
> -		prot_sg = &prot_table->sg_table[prot_page - prot_table->page_start_offset];
> -
> -		rc = sbc_dif_verify_write(cmd, cmd->t_task_lba, sectors, 0,
> -					  prot_sg, prot_offset);
> +		rc = rd_do_prot_rw(cmd, sbc_dif_verify_write);
>   		if (rc)
>   			return rc;
>   	}
> @@ -503,23 +517,7 @@ rd_execute_rw(struct se_cmd *cmd, struct scatterlist *sgl, u32 sgl_nents,
>   	sg_miter_stop(&m);
>
>   	if (cmd->prot_type && data_direction == DMA_FROM_DEVICE) {
> -		struct rd_dev_sg_table *prot_table;
> -		struct scatterlist *prot_sg;
> -		u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
> -		u32 prot_offset, prot_page;
> -
> -		tmp = cmd->t_task_lba * se_dev->prot_length;
> -		prot_offset = do_div(tmp, PAGE_SIZE);
> -		prot_page = tmp;
> -
> -		prot_table = rd_get_prot_table(dev, prot_page);
> -		if (!prot_table)
> -			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> -
> -		prot_sg = &prot_table->sg_table[prot_page - prot_table->page_start_offset];
> -
> -		rc = sbc_dif_verify_read(cmd, cmd->t_task_lba, sectors, 0,
> -					 prot_sg, prot_offset);
> +		rc = rd_do_prot_rw(cmd, sbc_dif_verify_read);
>   		if (rc)
>   			return rc;
>   	}
>


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_*
  2015-04-05 14:59 ` [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_* Akinobu Mita
@ 2015-04-06  8:42   ` Sagi Grimberg
  2015-04-08  5:13   ` Nicholas A. Bellinger
  1 sibling, 0 replies; 9+ messages in thread
From: Sagi Grimberg @ 2015-04-06  8:42 UTC (permalink / raw)
  To: Akinobu Mita, target-devel
  Cc: Nicholas Bellinger, Martin K. Petersen, Christoph Hellwig,
	James E.J. Bottomley, linux-scsi

On 4/5/2015 5:59 PM, Akinobu Mita wrote:
> The scatterlist for protection information which is passed to
> sbc_dif_verify_read() or sbc_dif_verify_write() requires that
> neighboring scatterlist entries are contiguous or chained so that they
> can be iterated by sg_next().
>
> However, the protection information for RD-MCP backends could be located
> in the multiple scatterlist arrays when the ramdisk space is too large.
> So if the read/write request straddles this boundary, sbc_dif_verify_read()
> or sbc_dif_verify_write() can't iterate all scatterlist entries.
>
> This fixes it by allocating temporary scatterlist if it is needed.
>
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
> Cc: target-devel@vger.kernel.org
> Cc: linux-scsi@vger.kernel.org
> ---
> * No change from v1
>
>   drivers/target/target_core_rd.c | 39 +++++++++++++++++++++++++++++++++++----
>   1 file changed, 35 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
> index ac5e8d2..6e25eaa 100644
> --- a/drivers/target/target_core_rd.c
> +++ b/drivers/target/target_core_rd.c
> @@ -390,11 +390,12 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
>   	struct se_device *se_dev = cmd->se_dev;
>   	struct rd_dev *dev = RD_DEV(se_dev);
>   	struct rd_dev_sg_table *prot_table;
> +	bool need_to_release = false;
>   	struct scatterlist *prot_sg;
>   	u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
> -	u32 prot_offset, prot_page;
> +	u32 prot_offset, prot_page, prot_npages;
>   	u64 tmp;
> -	sense_reason_t rc;
> +	sense_reason_t rc = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>
>   	tmp = cmd->t_task_lba * se_dev->prot_length;
>   	prot_offset = do_div(tmp, PAGE_SIZE);
> @@ -404,10 +405,40 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
>   	if (!prot_table)
>   		return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>
> -	prot_sg = &prot_table->sg_table[prot_page -
> -					prot_table->page_start_offset];
> +	prot_npages = DIV_ROUND_UP(prot_offset + sectors * se_dev->prot_length,
> +				   PAGE_SIZE);
> +
> +	/* prot pages straddles multiple scatterlist tables */
> +	if (prot_table->page_end_offset < prot_page + prot_npages - 1) {
> +		int i;
> +
> +		prot_sg = kcalloc(prot_npages, sizeof(*prot_sg), GFP_KERNEL);
> +		if (!prot_sg)
> +			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> +
> +		need_to_release = true;
> +		sg_init_table(prot_sg, prot_npages);
> +
> +		for (i = 0; i < prot_npages; i++) {
> +			if (prot_page + i > prot_table->page_end_offset) {
> +				prot_table = rd_get_prot_table(dev,
> +								prot_page + i);
> +				if (!prot_table)
> +					goto out;
> +				sg_unmark_end(&prot_sg[i - 1]);
> +			}
> +			prot_sg[i] = prot_table->sg_table[prot_page + i -
> +						prot_table->page_start_offset];
> +		}
> +	} else {
> +		prot_sg = &prot_table->sg_table[prot_page -
> +						prot_table->page_start_offset];
> +	}
>
>   	rc = dif_verify(cmd, cmd->t_task_lba, sectors, 0, prot_sg, prot_offset);
> +out:
> +	if (need_to_release)
> +		kfree(prot_sg);
>
>   	return rc;
>   }
>

Acked-by: Sagi Grimberg <sagig@mellanox.com>

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw()
  2015-04-06  7:43 ` [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Sagi Grimberg
@ 2015-04-06 11:16   ` Akinobu Mita
  0 siblings, 0 replies; 9+ messages in thread
From: Akinobu Mita @ 2015-04-06 11:16 UTC (permalink / raw)
  To: Sagi Grimberg
  Cc: target-devel, Nicholas Bellinger, Martin K. Petersen,
	Christoph Hellwig, James E.J. Bottomley,
	linux-scsi@vger.kernel.org

2015-04-06 16:43 GMT+09:00 Sagi Grimberg <sagig@dev.mellanox.co.il>:
> On 4/5/2015 5:59 PM, Akinobu Mita wrote:
>>
>> Factor out code duplication in rd_execute_rw() into a helper function
>> rd_do_prot_rw().  This change is required to minimize the forthcoming
>> fix in rd_do_prot_rw().
>>
>> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
>> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
>> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
>> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
>> Cc: Christoph Hellwig <hch@lst.de>
>> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
>> Cc: target-devel@vger.kernel.org
>> Cc: linux-scsi@vger.kernel.org
>> ---
>> * v2
>> - Pass dif_verify() function pointer to helper function instead of
>> is_write,
>>    suggested by Sagi Grimberg.
>>
>>   drivers/target/target_core_rd.c | 66
>> ++++++++++++++++++++---------------------
>>   1 file changed, 32 insertions(+), 34 deletions(-)
>>
>> diff --git a/drivers/target/target_core_rd.c
>> b/drivers/target/target_core_rd.c
>> index 98e83ac..ac5e8d2 100644
>> --- a/drivers/target/target_core_rd.c
>> +++ b/drivers/target/target_core_rd.c
>> @@ -382,6 +382,36 @@ static struct rd_dev_sg_table
>> *rd_get_prot_table(struct rd_dev *rd_dev, u32 page
>>         return NULL;
>>   }
>>
>> +typedef sense_reason_t (*dif_verify)(struct se_cmd *, sector_t, unsigned
>> int,
>> +                                    unsigned int, struct scatterlist *,
>> int);
>> +
>> +static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify
>> dif_verify)
>> +{
>> +       struct se_device *se_dev = cmd->se_dev;
>> +       struct rd_dev *dev = RD_DEV(se_dev);
>> +       struct rd_dev_sg_table *prot_table;
>> +       struct scatterlist *prot_sg;
>> +       u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
>> +       u32 prot_offset, prot_page;
>> +       u64 tmp;
>> +       sense_reason_t rc;
>> +
>> +       tmp = cmd->t_task_lba * se_dev->prot_length;
>> +       prot_offset = do_div(tmp, PAGE_SIZE);
>> +       prot_page = tmp;
>> +
>> +       prot_table = rd_get_prot_table(dev, prot_page);
>> +       if (!prot_table)
>> +               return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>> +
>> +       prot_sg = &prot_table->sg_table[prot_page -
>> +                                       prot_table->page_start_offset];
>> +
>> +       rc = dif_verify(cmd, cmd->t_task_lba, sectors, 0, prot_sg,
>> prot_offset);
>> +
>> +       return rc;
>
>
> Nit, Given there is no action on the returned rc, this can be reduced to:
>         return dif_verify();

You are right,  but the patch 2/2 inserts the lines to release
temporary prot_sg before return statement, so assignment and return
statements will be both required in the end.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw()
  2015-04-05 14:59 [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Akinobu Mita
  2015-04-05 14:59 ` [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_* Akinobu Mita
  2015-04-06  7:43 ` [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Sagi Grimberg
@ 2015-04-08  5:03 ` Nicholas A. Bellinger
  2 siblings, 0 replies; 9+ messages in thread
From: Nicholas A. Bellinger @ 2015-04-08  5:03 UTC (permalink / raw)
  To: Akinobu Mita
  Cc: target-devel, Sagi Grimberg, Martin K. Petersen,
	Christoph Hellwig, James E.J. Bottomley, linux-scsi

On Sun, 2015-04-05 at 23:59 +0900, Akinobu Mita wrote:
> Factor out code duplication in rd_execute_rw() into a helper function
> rd_do_prot_rw().  This change is required to minimize the forthcoming
> fix in rd_do_prot_rw().
> 
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
> Cc: target-devel@vger.kernel.org
> Cc: linux-scsi@vger.kernel.org
> ---
> * v2
> - Pass dif_verify() function pointer to helper function instead of is_write,
>   suggested by Sagi Grimberg.
> 
>  drivers/target/target_core_rd.c | 66 ++++++++++++++++++++---------------------
>  1 file changed, 32 insertions(+), 34 deletions(-)

Nice patch.  Applied to for-next with a bit of fuzz due to the
WRITE_STRIP + READ_INSERT related changes already queued up..

Thanks Akinobu!

--nab

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_*
  2015-04-05 14:59 ` [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_* Akinobu Mita
  2015-04-06  8:42   ` Sagi Grimberg
@ 2015-04-08  5:13   ` Nicholas A. Bellinger
  2015-04-08 14:25     ` Akinobu Mita
  1 sibling, 1 reply; 9+ messages in thread
From: Nicholas A. Bellinger @ 2015-04-08  5:13 UTC (permalink / raw)
  To: Akinobu Mita
  Cc: target-devel, Sagi Grimberg, Martin K. Petersen,
	Christoph Hellwig, James E.J. Bottomley, linux-scsi

On Sun, 2015-04-05 at 23:59 +0900, Akinobu Mita wrote:
> The scatterlist for protection information which is passed to
> sbc_dif_verify_read() or sbc_dif_verify_write() requires that
> neighboring scatterlist entries are contiguous or chained so that they
> can be iterated by sg_next().
> 
> However, the protection information for RD-MCP backends could be located
> in the multiple scatterlist arrays when the ramdisk space is too large.
> So if the read/write request straddles this boundary, sbc_dif_verify_read()
> or sbc_dif_verify_write() can't iterate all scatterlist entries.
> 
> This fixes it by allocating temporary scatterlist if it is needed.
> 
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
> Cc: target-devel@vger.kernel.org
> Cc: linux-scsi@vger.kernel.org
> ---
> * No change from v1
> 
>  drivers/target/target_core_rd.c | 39 +++++++++++++++++++++++++++++++++++----
>  1 file changed, 35 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
> index ac5e8d2..6e25eaa 100644
> --- a/drivers/target/target_core_rd.c
> +++ b/drivers/target/target_core_rd.c
> @@ -390,11 +390,12 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
>  	struct se_device *se_dev = cmd->se_dev;
>  	struct rd_dev *dev = RD_DEV(se_dev);
>  	struct rd_dev_sg_table *prot_table;
> +	bool need_to_release = false;
>  	struct scatterlist *prot_sg;
>  	u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
> -	u32 prot_offset, prot_page;
> +	u32 prot_offset, prot_page, prot_npages;
>  	u64 tmp;
> -	sense_reason_t rc;
> +	sense_reason_t rc = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>  
>  	tmp = cmd->t_task_lba * se_dev->prot_length;
>  	prot_offset = do_div(tmp, PAGE_SIZE);
> @@ -404,10 +405,40 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
>  	if (!prot_table)
>  		return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>  
> -	prot_sg = &prot_table->sg_table[prot_page -
> -					prot_table->page_start_offset];
> +	prot_npages = DIV_ROUND_UP(prot_offset + sectors * se_dev->prot_length,
> +				   PAGE_SIZE);
> +
> +	/* prot pages straddles multiple scatterlist tables */
> +	if (prot_table->page_end_offset < prot_page + prot_npages - 1) {
> +		int i;
> +
> +		prot_sg = kcalloc(prot_npages, sizeof(*prot_sg), GFP_KERNEL);
> +		if (!prot_sg)
> +			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> +
> +		need_to_release = true;
> +		sg_init_table(prot_sg, prot_npages);
> +
> +		for (i = 0; i < prot_npages; i++) {
> +			if (prot_page + i > prot_table->page_end_offset) {
> +				prot_table = rd_get_prot_table(dev,
> +								prot_page + i);
> +				if (!prot_table)
> +					goto out;
> +				sg_unmark_end(&prot_sg[i - 1]);
> +			}
> +			prot_sg[i] = prot_table->sg_table[prot_page + i -
> +						prot_table->page_start_offset];
> +		}
> +	} else {
> +		prot_sg = &prot_table->sg_table[prot_page -
> +						prot_table->page_start_offset];
> +	}
>  

Mmm, how about just explicitly doing sg_link() at prot_table scatterlist
creation time instead..?

That would save the extra allocation above, and AFAICT make for simpler
code.

--nab

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_*
  2015-04-08  5:13   ` Nicholas A. Bellinger
@ 2015-04-08 14:25     ` Akinobu Mita
  2015-04-09  6:41       ` Nicholas A. Bellinger
  0 siblings, 1 reply; 9+ messages in thread
From: Akinobu Mita @ 2015-04-08 14:25 UTC (permalink / raw)
  To: Nicholas A. Bellinger
  Cc: target-devel, Sagi Grimberg, Martin K. Petersen,
	Christoph Hellwig, James E.J. Bottomley,
	linux-scsi@vger.kernel.org

2015-04-08 14:13 GMT+09:00 Nicholas A. Bellinger <nab@linux-iscsi.org>:
> On Sun, 2015-04-05 at 23:59 +0900, Akinobu Mita wrote:
>> The scatterlist for protection information which is passed to
>> sbc_dif_verify_read() or sbc_dif_verify_write() requires that
>> neighboring scatterlist entries are contiguous or chained so that they
>> can be iterated by sg_next().
>>
>> However, the protection information for RD-MCP backends could be located
>> in the multiple scatterlist arrays when the ramdisk space is too large.
>> So if the read/write request straddles this boundary, sbc_dif_verify_read()
>> or sbc_dif_verify_write() can't iterate all scatterlist entries.
>>
>> This fixes it by allocating temporary scatterlist if it is needed.
>>
>> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
>> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
>> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
>> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
>> Cc: Christoph Hellwig <hch@lst.de>
>> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
>> Cc: target-devel@vger.kernel.org
>> Cc: linux-scsi@vger.kernel.org
>> ---
>> * No change from v1
>>
>>  drivers/target/target_core_rd.c | 39 +++++++++++++++++++++++++++++++++++----
>>  1 file changed, 35 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
>> index ac5e8d2..6e25eaa 100644
>> --- a/drivers/target/target_core_rd.c
>> +++ b/drivers/target/target_core_rd.c
>> @@ -390,11 +390,12 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
>>       struct se_device *se_dev = cmd->se_dev;
>>       struct rd_dev *dev = RD_DEV(se_dev);
>>       struct rd_dev_sg_table *prot_table;
>> +     bool need_to_release = false;
>>       struct scatterlist *prot_sg;
>>       u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
>> -     u32 prot_offset, prot_page;
>> +     u32 prot_offset, prot_page, prot_npages;
>>       u64 tmp;
>> -     sense_reason_t rc;
>> +     sense_reason_t rc = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>>
>>       tmp = cmd->t_task_lba * se_dev->prot_length;
>>       prot_offset = do_div(tmp, PAGE_SIZE);
>> @@ -404,10 +405,40 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
>>       if (!prot_table)
>>               return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>>
>> -     prot_sg = &prot_table->sg_table[prot_page -
>> -                                     prot_table->page_start_offset];
>> +     prot_npages = DIV_ROUND_UP(prot_offset + sectors * se_dev->prot_length,
>> +                                PAGE_SIZE);
>> +
>> +     /* prot pages straddles multiple scatterlist tables */
>> +     if (prot_table->page_end_offset < prot_page + prot_npages - 1) {
>> +             int i;
>> +
>> +             prot_sg = kcalloc(prot_npages, sizeof(*prot_sg), GFP_KERNEL);
>> +             if (!prot_sg)
>> +                     return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
>> +
>> +             need_to_release = true;
>> +             sg_init_table(prot_sg, prot_npages);
>> +
>> +             for (i = 0; i < prot_npages; i++) {
>> +                     if (prot_page + i > prot_table->page_end_offset) {
>> +                             prot_table = rd_get_prot_table(dev,
>> +                                                             prot_page + i);
>> +                             if (!prot_table)
>> +                                     goto out;
>> +                             sg_unmark_end(&prot_sg[i - 1]);
>> +                     }
>> +                     prot_sg[i] = prot_table->sg_table[prot_page + i -
>> +                                             prot_table->page_start_offset];
>> +             }
>> +     } else {
>> +             prot_sg = &prot_table->sg_table[prot_page -
>> +                                             prot_table->page_start_offset];
>> +     }
>>
>
> Mmm, how about just explicitly doing sg_link() at prot_table scatterlist
> creation time instead..?

Yeap, that would be an optimal solution.  But some architectures don't
support sg chaining (i.e. !CONFIG_ARCH_HAS_SG_CHAIN).

> That would save the extra allocation above, and AFAICT make for simpler
> code.

Do you prefer fixing it conditionally with using #ifdef ? (i.e.
sg chaining at creating time if CONFIG_ARCH_HAS_SG_CHAIN is defined,
otherwise do the extra allocation above) If so, I'll resubmit the
patch with such changes.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_*
  2015-04-08 14:25     ` Akinobu Mita
@ 2015-04-09  6:41       ` Nicholas A. Bellinger
  0 siblings, 0 replies; 9+ messages in thread
From: Nicholas A. Bellinger @ 2015-04-09  6:41 UTC (permalink / raw)
  To: Akinobu Mita
  Cc: target-devel, Sagi Grimberg, Martin K. Petersen,
	Christoph Hellwig, James E.J. Bottomley,
	linux-scsi@vger.kernel.org

On Wed, 2015-04-08 at 23:25 +0900, Akinobu Mita wrote:
> 2015-04-08 14:13 GMT+09:00 Nicholas A. Bellinger <nab@linux-iscsi.org>:
> > On Sun, 2015-04-05 at 23:59 +0900, Akinobu Mita wrote:
> >> The scatterlist for protection information which is passed to
> >> sbc_dif_verify_read() or sbc_dif_verify_write() requires that
> >> neighboring scatterlist entries are contiguous or chained so that they
> >> can be iterated by sg_next().
> >>
> >> However, the protection information for RD-MCP backends could be located
> >> in the multiple scatterlist arrays when the ramdisk space is too large.
> >> So if the read/write request straddles this boundary, sbc_dif_verify_read()
> >> or sbc_dif_verify_write() can't iterate all scatterlist entries.
> >>
> >> This fixes it by allocating temporary scatterlist if it is needed.
> >>
> >> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> >> Cc: Nicholas Bellinger <nab@linux-iscsi.org>
> >> Cc: Sagi Grimberg <sagig@dev.mellanox.co.il>
> >> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
> >> Cc: Christoph Hellwig <hch@lst.de>
> >> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
> >> Cc: target-devel@vger.kernel.org
> >> Cc: linux-scsi@vger.kernel.org
> >> ---
> >> * No change from v1
> >>
> >>  drivers/target/target_core_rd.c | 39 +++++++++++++++++++++++++++++++++++----
> >>  1 file changed, 35 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/target/target_core_rd.c b/drivers/target/target_core_rd.c
> >> index ac5e8d2..6e25eaa 100644
> >> --- a/drivers/target/target_core_rd.c
> >> +++ b/drivers/target/target_core_rd.c
> >> @@ -390,11 +390,12 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
> >>       struct se_device *se_dev = cmd->se_dev;
> >>       struct rd_dev *dev = RD_DEV(se_dev);
> >>       struct rd_dev_sg_table *prot_table;
> >> +     bool need_to_release = false;
> >>       struct scatterlist *prot_sg;
> >>       u32 sectors = cmd->data_length / se_dev->dev_attrib.block_size;
> >> -     u32 prot_offset, prot_page;
> >> +     u32 prot_offset, prot_page, prot_npages;
> >>       u64 tmp;
> >> -     sense_reason_t rc;
> >> +     sense_reason_t rc = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> >>
> >>       tmp = cmd->t_task_lba * se_dev->prot_length;
> >>       prot_offset = do_div(tmp, PAGE_SIZE);
> >> @@ -404,10 +405,40 @@ static sense_reason_t rd_do_prot_rw(struct se_cmd *cmd, dif_verify dif_verify)
> >>       if (!prot_table)
> >>               return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> >>
> >> -     prot_sg = &prot_table->sg_table[prot_page -
> >> -                                     prot_table->page_start_offset];
> >> +     prot_npages = DIV_ROUND_UP(prot_offset + sectors * se_dev->prot_length,
> >> +                                PAGE_SIZE);
> >> +
> >> +     /* prot pages straddles multiple scatterlist tables */
> >> +     if (prot_table->page_end_offset < prot_page + prot_npages - 1) {
> >> +             int i;
> >> +
> >> +             prot_sg = kcalloc(prot_npages, sizeof(*prot_sg), GFP_KERNEL);
> >> +             if (!prot_sg)
> >> +                     return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
> >> +
> >> +             need_to_release = true;
> >> +             sg_init_table(prot_sg, prot_npages);
> >> +
> >> +             for (i = 0; i < prot_npages; i++) {
> >> +                     if (prot_page + i > prot_table->page_end_offset) {
> >> +                             prot_table = rd_get_prot_table(dev,
> >> +                                                             prot_page + i);
> >> +                             if (!prot_table)
> >> +                                     goto out;
> >> +                             sg_unmark_end(&prot_sg[i - 1]);
> >> +                     }
> >> +                     prot_sg[i] = prot_table->sg_table[prot_page + i -
> >> +                                             prot_table->page_start_offset];
> >> +             }
> >> +     } else {
> >> +             prot_sg = &prot_table->sg_table[prot_page -
> >> +                                             prot_table->page_start_offset];
> >> +     }
> >>
> >
> > Mmm, how about just explicitly doing sg_link() at prot_table scatterlist
> > creation time instead..?
> 
> Yeap, that would be an optimal solution.  But some architectures don't
> support sg chaining (i.e. !CONFIG_ARCH_HAS_SG_CHAIN).
> 
> > That would save the extra allocation above, and AFAICT make for simpler
> > code.
> 
> Do you prefer fixing it conditionally with using #ifdef ? (i.e.
> sg chaining at creating time if CONFIG_ARCH_HAS_SG_CHAIN is defined,
> otherwise do the extra allocation above) If so, I'll resubmit the
> patch with such changes.

Yes please.

It would be nice to use sg_link() for the typical case when available,
as long as the #idefs aren't too ugly and limited to target_core_rd.c
code.

--nab


^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2015-04-09  6:41 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-04-05 14:59 [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Akinobu Mita
2015-04-05 14:59 ` [PATCH v2 2/2] target/rd: Don't pass imcomplete scatterlist entries to sbc_dif_verify_* Akinobu Mita
2015-04-06  8:42   ` Sagi Grimberg
2015-04-08  5:13   ` Nicholas A. Bellinger
2015-04-08 14:25     ` Akinobu Mita
2015-04-09  6:41       ` Nicholas A. Bellinger
2015-04-06  7:43 ` [PATCH v2 1/2] target/rd: reduce code duplication in rd_execute_rw() Sagi Grimberg
2015-04-06 11:16   ` Akinobu Mita
2015-04-08  5:03 ` Nicholas A. Bellinger

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox