[PATCH 2/5] aic94xx: plumb in I_T_nexus_reset task management function

[PATCH 2/5] aic94xx: plumb in I_T_nexus_reset task management function

[James Bottomley]

Currently aic94xx has no exported I_T_nexus_reset function.  This is a
bit of a huge problem, since sas_ata relies on this function to
perform an ATA phy reset and also it means that if abort fails, we
really have no bigger hammer to hit everything with.

Plumb in the I_T_nexus_reset by quiescing the sequencer, sending the
correct phy reset (link for ATA and hard for SAS) and then carefully
resuming the sequencer again.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
---
 drivers/scsi/aic94xx/aic94xx.h      |    1 +
 drivers/scsi/aic94xx/aic94xx_init.c |    2 +-
 drivers/scsi/aic94xx/aic94xx_tmf.c  |   58 ++++++++++++++++++++++++++++++++--
 3 files changed, 56 insertions(+), 5 deletions(-)

diff --git a/drivers/scsi/aic94xx/aic94xx.h b/drivers/scsi/aic94xx/aic94xx.h
index 32f513b..eb8efdc 100644
--- a/drivers/scsi/aic94xx/aic94xx.h
+++ b/drivers/scsi/aic94xx/aic94xx.h
@@ -102,6 +102,7 @@ int  asd_abort_task_set(struct domain_device *, u8 *lun);
 int  asd_clear_aca(struct domain_device *, u8 *lun);
 int  asd_clear_task_set(struct domain_device *, u8 *lun);
 int  asd_lu_reset(struct domain_device *, u8 *lun);
+int  asd_I_T_nexus_reset(struct domain_device *dev);
 int  asd_query_task(struct sas_task *);
 
 /* ---------- Adapter and Port management ---------- */
diff --git a/drivers/scsi/aic94xx/aic94xx_init.c b/drivers/scsi/aic94xx/aic94xx_init.c
index 5d761eb..88d1e73 100644
--- a/drivers/scsi/aic94xx/aic94xx_init.c
+++ b/drivers/scsi/aic94xx/aic94xx_init.c
@@ -1003,7 +1003,7 @@ static struct sas_domain_function_template aic94xx_transport_functions = {
 	.lldd_abort_task_set	= asd_abort_task_set,
 	.lldd_clear_aca		= asd_clear_aca,
 	.lldd_clear_task_set	= asd_clear_task_set,
-	.lldd_I_T_nexus_reset	= NULL,
+	.lldd_I_T_nexus_reset	= asd_I_T_nexus_reset,
 	.lldd_lu_reset		= asd_lu_reset,
 	.lldd_query_task	= asd_query_task,
 
diff --git a/drivers/scsi/aic94xx/aic94xx_tmf.c b/drivers/scsi/aic94xx/aic94xx_tmf.c
index 144f5ad..d684c74 100644
--- a/drivers/scsi/aic94xx/aic94xx_tmf.c
+++ b/drivers/scsi/aic94xx/aic94xx_tmf.c
@@ -140,8 +140,14 @@ int asd_clear_nexus_port(struct asd_sas_port *port)
 	CLEAR_NEXUS_POST;
 }
 
-#if 0
-static int asd_clear_nexus_I_T(struct domain_device *dev)
+enum clear_nexus_phase {
+	NEXUS_PHASE_PRE,
+	NEXUS_PHASE_POST,
+	NEXUS_PHASE_RESUME,
+};
+
+static int asd_clear_nexus_I_T(struct domain_device *dev,
+			       enum clear_nexus_phase phase)
 {
 	struct asd_ha_struct *asd_ha = dev->port->ha->lldd_ha;
 	struct asd_ascb *ascb;
@@ -150,12 +156,56 @@ static int asd_clear_nexus_I_T(struct domain_device *dev)
 
 	CLEAR_NEXUS_PRE;
 	scb->clear_nexus.nexus = NEXUS_I_T;
-	scb->clear_nexus.flags = SEND_Q | EXEC_Q | NOTINQ;
+	switch (phase) {
+	case NEXUS_PHASE_PRE:
+		scb->clear_nexus.flags = EXEC_Q | SUSPEND_TX;
+		break;
+	case NEXUS_PHASE_POST:
+		scb->clear_nexus.flags = SEND_Q | NOTINQ;
+		break;
+	case NEXUS_PHASE_RESUME:
+		scb->clear_nexus.flags = RESUME_TX;
+	}
 	scb->clear_nexus.conn_handle = cpu_to_le16((u16)(unsigned long)
 						   dev->lldd_dev);
 	CLEAR_NEXUS_POST;
 }
-#endif
+
+int asd_I_T_nexus_reset(struct domain_device *dev)
+{
+	int res, tmp_res, i;
+	struct sas_phy *phy = sas_find_local_phy(dev);
+	/* Standard mandates link reset for ATA  (type 0) and
+	 * hard reset for SSP (type 1) */
+	int reset_type = (dev->dev_type == SATA_DEV ||
+			  (dev->tproto & SAS_PROTOCOL_STP)) ? 0 : 1;
+
+	asd_clear_nexus_I_T(dev, NEXUS_PHASE_PRE);
+	/* send a hard reset */
+	ASD_DPRINTK("sending %s reset to %s\n",
+		    reset_type ? "hard" : "soft", phy->dev.bus_id);
+	res = sas_phy_reset(phy, reset_type);
+	if (res == TMF_RESP_FUNC_COMPLETE) {
+		/* wait for the maximum settle time */
+		msleep(500);
+		/* clear all outstanding commands (keep nexus suspended) */
+		asd_clear_nexus_I_T(dev, NEXUS_PHASE_POST);
+	}
+	for (i = 0 ; i < 3; i++) {
+		tmp_res = asd_clear_nexus_I_T(dev, NEXUS_PHASE_RESUME);
+		if (tmp_res == TC_RESUME)
+			return res;
+		msleep(500);
+	}
+
+	/* This is a bit of a problem:  the sequencer is still suspended
+	 * and is refusing to resume.  Hope it will resume on a bigger hammer
+	 * or the disk is lost */
+	dev_printk(KERN_ERR, &phy->dev,
+		   "Failed to resume nexus after reset 0x%x\n", tmp_res);
+
+	return TMF_RESP_FUNC_FAILED;
+}
 
 static int asd_clear_nexus_I_T_L(struct domain_device *dev, u8 *lun)
 {
-- 
1.5.4.1

Re: [PATCH 2/5] aic94xx: plumb in I_T_nexus_reset task management function

[Jeff Garzik]

James Bottomley wrote:
> Currently aic94xx has no exported I_T_nexus_reset function.  This is a
> bit of a huge problem, since sas_ata relies on this function to
> perform an ATA phy reset and also it means that if abort fails, we
> really have no bigger hammer to hit everything with.
> 
> Plumb in the I_T_nexus_reset by quiescing the sequencer, sending the
> correct phy reset (link for ATA and hard for SAS) and then carefully
> resuming the sequencer again.
> 
> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
> ---
>  drivers/scsi/aic94xx/aic94xx.h      |    1 +
>  drivers/scsi/aic94xx/aic94xx_init.c |    2 +-
>  drivers/scsi/aic94xx/aic94xx_tmf.c  |   58 ++++++++++++++++++++++++++++++++--
>  3 files changed, 56 insertions(+), 5 deletions(-)

Tangent:

Now that patches for ipr and sata_sx4 exist, libsas is the last holdout 
using libata's old error handling paths.

I think Darrick was either scoping that work, or had preliminary patches?

Anyway, that's holding back removal of a lot of libata code, so I wanted 
to make sure the issue remained in people's minds...

	Jeff

Re: [PATCH 2/5] aic94xx: plumb in I_T_nexus_reset task management function

[James Bottomley]

On Sun, 2008-02-24 at 00:51 -0500, Jeff Garzik wrote:
> Tangent:
> 
> Now that patches for ipr and sata_sx4 exist, libsas is the last holdout 
> using libata's old error handling paths.
> 
> I think Darrick was either scoping that work, or had preliminary patches?
> 
> Anyway, that's holding back removal of a lot of libata code, so I wanted 
> to make sure the issue remained in people's minds...

I've not really been following what the issue actually is, so hopefully
Darrick has.

At the moment, we don't use any of libata for error handling in libsas.
(In fact, there's a big issue with ATA error handling in libsas in
general, since the sequencer abort task doesn't know what to do with a
task that's at an ATA drive ... we need to detect that condition and
then apply the SATL abort function, which depends on whether we're NCQ
or not...SATA error handling was next on my list of things to look at).

James