diff mbox series

[1/2] scsi: target: core: Add sense reason for space allocation errors

Message ID 20211020184319.588002-2-k.shelekhin@yadro.com (mailing list archive)
State New, archived
Headers show
Series scsi: target: iblock: Report space allocation errors | expand

Commit Message

Konstantin Shelekhin Oct. 20, 2021, 6:43 p.m. UTC 
According to SBC-3 4.7.3.6 this sense reason shall be used in situations
where thin provisioned logical unit cannot satisfy the write request due
to the lack of free blocks.

Signed-off-by: Konstantin Shelekhin <k.shelekhin@yadro.com>
Reviewed-by: Dmitry Bogdanov <d.bogdanov@yadro.com>
---
 drivers/target/target_core_transport.c | 6 ++++++
 include/target/target_core_base.h      | 1 +
 2 files changed, 7 insertions(+)

Comments

Martin K. Petersen Oct. 21, 2021, 3:21 a.m. UTC | #1 
Konstantin,

> According to SBC-3 4.7.3.6 this sense reason shall be used in situations
> where thin provisioned logical unit cannot satisfy the write request due
> to the lack of free blocks.

> +	[TCM_SPACE_ALLOCATION_FAILED] = {
> +		.key = DATA_PROTECT,
> +		.asc = 0x27,
> +		.ascq = 0x07, /* SPACE ALLOCATION FAILED WRITE PROTECT */
> +	},

How do we know this is a permanent condition and not a temporary space
exhaustion?
Konstantin Shelekhin Oct. 21, 2021, 8:55 a.m. UTC | #2 
On Wed, Oct 20, 2021 at 11:21:54PM -0400, Martin K. Petersen wrote:
> 
> Konstantin,
> 
> > According to SBC-3 4.7.3.6 this sense reason shall be used in situations
> > where thin provisioned logical unit cannot satisfy the write request due
> > to the lack of free blocks.
> 
> > +	[TCM_SPACE_ALLOCATION_FAILED] = {
> > +		.key = DATA_PROTECT,
> > +		.asc = 0x27,
> > +		.ascq = 0x07, /* SPACE ALLOCATION FAILED WRITE PROTECT */
> > +	},
> 
> How do we know this is a permanent condition and not a temporary space
> exhaustion?

By permanent condition SBC-3 means that an initiator should not resend
the command immediately as it will fail again. Kernel tries hard not to
fail with BLK_STS_NOSPC:

  /*
   * We're holding onto IO to allow userland time to react.  After the
   * timeout either the pool will have been resized (and thus back in
   * PM_WRITE mode), or we degrade to PM_OUT_OF_DATA_SPACE w/ error_if_no_space.
   */
  static void do_no_space_timeout(struct work_struct *ws)

So BLK_STS_NOSPC means that we are stuck with this condition and some
out-of-scope actions (like running fs-trim on the initiator) are
required.
diff mbox series

Patch

diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c
index 14c6f2bb1b01..d8261bd1cb5c 100644
--- a/drivers/target/target_core_transport.c
+++ b/drivers/target/target_core_transport.c
@@ -2025,6 +2025,7 @@  void transport_generic_request_failure(struct se_cmd *cmd,
 	case TCM_ALUA_TG_PT_UNAVAILABLE:
 	case TCM_ALUA_STATE_TRANSITION:
 	case TCM_ALUA_OFFLINE:
+	case TCM_SPACE_ALLOCATION_FAILED:
 		break;
 	case TCM_OUT_OF_RESOURCES:
 		cmd->scsi_status = SAM_STAT_TASK_SET_FULL;
@@ -3369,6 +3370,11 @@  static const struct sense_detail sense_detail_table[] = {
 		.asc = 0x04,
 		.ascq = ASCQ_04H_ALUA_OFFLINE,
 	},
+	[TCM_SPACE_ALLOCATION_FAILED] = {
+		.key = DATA_PROTECT,
+		.asc = 0x27,
+		.ascq = 0x07, /* SPACE ALLOCATION FAILED WRITE PROTECT */
+	},
 };
 
 /**
diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
index fb11c7693b25..fe2f1c5bb1b8 100644
--- a/include/target/target_core_base.h
+++ b/include/target/target_core_base.h
@@ -192,6 +192,7 @@  enum tcm_sense_reason_table {
 	TCM_ALUA_TG_PT_UNAVAILABLE		= R(0x21),
 	TCM_ALUA_STATE_TRANSITION		= R(0x22),
 	TCM_ALUA_OFFLINE			= R(0x23),
+	TCM_SPACE_ALLOCATION_FAILED		= R(0x24),
 #undef R
 };