[PATCH 2.6.21-rc1 0/2] sata_vsc regression fix and vsc_sata_interrupt_cleanup (take2)

[PATCH 2.6.21-rc1 0/2] sata_vsc regression fix and vsc_sata_interrupt_cleanup (take2)

[Dan Williams]

Tejun,

Here are the sata_vsc patches again with the following changes:
* revert changes to vsc_intr_mask_update (vsc_thaw enables all interrupts)
* use unlikely() for the pci-abort and not-our-interrupt cases in
  vsc_sata_interrupt
* move vsc_freeze and vsc_thaw into this patch
* removed definition of invalid fis bit
* let standard ata-error-handling handle the serror register
* clear all unhandled interrupts

Thanks for pointing me in the right direction.

--
Dan

[PATCH 2.6.21-rc1 1/2] sata_vsc: factor the error and normal intr paths into separate routines

[Dan Williams]

Separate sata_vsc interrupt handling into a normal (per-port) path and an
error path with the addition of vsc_port_intr and vsc_error_intr
respectively.  The error path handles interrupt based
hotplug events which requires the definition of vsc_freeze and vsc_thaw.

Note: vsc_port_intr has a workaround for unexpected interrupts that occur
during polled commands.

Changes in take2:
* removed definition of invalid fis bit
* let standard ata-error-handling handle the serror register
* clear all unhandled interrupts

Cc: Jeremy Higdon <jeremy@sgi.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---

 drivers/ata/sata_vsc.c |   30 ++++++++++++++++++++++++++++++
 1 files changed, 30 insertions(+), 0 deletions(-)

diff --git a/drivers/ata/sata_vsc.c b/drivers/ata/sata_vsc.c
index 2fd037b..6a0c9cf 100644
--- a/drivers/ata/sata_vsc.c
+++ b/drivers/ata/sata_vsc.c
@@ -203,6 +203,36 @@ static void vsc_sata_tf_read(struct ata_port *ap, struct ata_taskfile *tf)
         }
 }
 
+static inline void vsc_error_intr(u8 port_status, struct ata_port *ap)
+{
+	if (port_status & (VSC_SATA_INT_PHY_CHANGE | VSC_SATA_INT_ERROR_M))
+		ata_port_freeze(ap);
+	else
+		ata_port_abort(ap);
+}
+
+static void vsc_port_intr(u8 port_status, struct ata_port *ap)
+{
+	struct ata_queued_cmd *qc;
+	int handled = 0;
+
+	if (unlikely(port_status & VSC_SATA_INT_ERROR)) {
+		vsc_error_intr(port_status, ap);
+		return;
+	}
+
+	qc = ata_qc_from_tag(ap, ap->active_tag);
+	if (qc && likely(!(qc->tf.flags & ATA_TFLAG_POLLING)))
+		handled = ata_host_intr(ap, qc);
+
+	/* We received an interrupt during a polled command,
+	 * or some other spurious condition.  Interrupt reporting
+	 * with this hardware is fairly reliable so it is safe to
+	 * simply clear the interrupt
+	 */
+	if (unlikely(!handled))
+		ata_chk_status(ap);
+}
 
 /*
  * vsc_sata_interrupt

[PATCH 2.6.21-rc1 2/2] sata_vsc: clean up vsc_sata_interrupt

[Dan Williams]

1/ Clean up vsc_sata_interrupt and have it call vsc_port_intr to handle the
   interrupt.
2/ Hook up vsc_freeze and vsc_thaw in vsc_sata_ops
3/ Remove the now unnecessary is_vsc_sata_int_err

Tested with an iq3124h PCI-X add-in card on an iop13xx.

Changes in take2
* revert changes to vsc_intr_mask_update (vsc_thaw enables all interrupts)
* use unlikely() for the pci-abort and not-our-interrupt cases in
  vsc_sata_interrupt
* move vsc_freeze and vsc_thaw into this patch

Cc: Jeremy Higdon <jeremy@sgi.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---

 drivers/ata/sata_vsc.c |   91 ++++++++++++++++++++++--------------------------
 1 files changed, 42 insertions(+), 49 deletions(-)

diff --git a/drivers/ata/sata_vsc.c b/drivers/ata/sata_vsc.c
index 6a0c9cf..0c68740 100644
--- a/drivers/ata/sata_vsc.c
+++ b/drivers/ata/sata_vsc.c
@@ -98,10 +98,6 @@ enum {
 			      VSC_SATA_INT_PHY_CHANGE),
 };
 
-#define is_vsc_sata_int_err(port_idx, int_status) \
-	 (int_status & (VSC_SATA_INT_ERROR << (8 * port_idx)))
-
-
 static u32 vsc_sata_scr_read (struct ata_port *ap, unsigned int sc_reg)
 {
 	if (sc_reg > SCR_CONTROL)
@@ -119,6 +115,28 @@ static void vsc_sata_scr_write (struct ata_port *ap, unsigned int sc_reg,
 }
 
 
+static void vsc_freeze(struct ata_port *ap)
+{
+	void __iomem *mask_addr;
+
+	mask_addr = ap->host->iomap[VSC_MMIO_BAR] +
+		VSC_SATA_INT_MASK_OFFSET + ap->port_no;
+
+	writeb(0, mask_addr);
+}
+
+
+static void vsc_thaw(struct ata_port *ap)
+{
+	void __iomem *mask_addr;
+
+	mask_addr = ap->host->iomap[VSC_MMIO_BAR] +
+		VSC_SATA_INT_MASK_OFFSET + ap->port_no;
+
+	writeb(0xff, mask_addr);
+}
+
+
 static void vsc_intr_mask_update(struct ata_port *ap, u8 ctl)
 {
 	void __iomem *mask_addr;
@@ -244,59 +262,34 @@ static irqreturn_t vsc_sata_interrupt (int irq, void *dev_instance)
 	struct ata_host *host = dev_instance;
 	unsigned int i;
 	unsigned int handled = 0;
-	u32 int_status;
-
-	spin_lock(&host->lock);
+	u32 status;
 
-	int_status = readl(host->iomap[VSC_MMIO_BAR] +
-			   VSC_SATA_INT_STAT_OFFSET);
+	status = readl(host->iomap[VSC_MMIO_BAR] + VSC_SATA_INT_STAT_OFFSET);
 
-	for (i = 0; i < host->n_ports; i++) {
-		if (int_status & ((u32) 0xFF << (8 * i))) {
-			struct ata_port *ap;
+	if (unlikely(status == 0xffffffff || status == 0)) {
+		status && printk(KERN_ERR DRV_NAME ": IRQ status == 0xffffffff, "
+			"PCI fault or device removal?\n");
+		goto out;
+	}
 
-			ap = host->ports[i];
+	spin_lock(&host->lock);
 
-			if (is_vsc_sata_int_err(i, int_status)) {
-				u32 err_status;
-				printk(KERN_DEBUG "%s: ignoring interrupt(s)\n", __FUNCTION__);
-				err_status = ap ? vsc_sata_scr_read(ap, SCR_ERROR) : 0;
-				vsc_sata_scr_write(ap, SCR_ERROR, err_status);
-				handled++;
-			}
+	for (i = 0; i < host->n_ports; i++) {
+		u8 port_status = (status >> (8 * i)) & 0xff;
+		if (port_status) {
+			struct ata_port *ap = host->ports[i];
 
 			if (ap && !(ap->flags & ATA_FLAG_DISABLED)) {
-				struct ata_queued_cmd *qc;
-
-				qc = ata_qc_from_tag(ap, ap->active_tag);
-				if (qc && (!(qc->tf.flags & ATA_TFLAG_POLLING)))
-					handled += ata_host_intr(ap, qc);
-				else if (is_vsc_sata_int_err(i, int_status)) {
-					/*
-					 * On some chips (i.e. Intel 31244), an error
-					 * interrupt will sneak in at initialization
-					 * time (phy state changes).  Clearing the SCR
-					 * error register is not required, but it prevents
-					 * the phy state change interrupts from recurring
-					 * later.
-					 */
-					u32 err_status;
-					err_status = vsc_sata_scr_read(ap, SCR_ERROR);
-					printk(KERN_DEBUG "%s: clearing interrupt, "
-					       "status %x; sata err status %x\n",
-					       __FUNCTION__,
-					       int_status, err_status);
-					vsc_sata_scr_write(ap, SCR_ERROR, err_status);
-					/* Clear interrupt status */
-					ata_chk_status(ap);
-					handled++;
-				}
-			}
+				vsc_port_intr(port_status, ap);
+				handled++;
+			} else
+				printk(KERN_ERR DRV_NAME
+					": interrupt from disabled port %d\n", i);
 		}
 	}
 
 	spin_unlock(&host->lock);
-
+out:
 	return IRQ_RETVAL(handled);
 }
 
@@ -334,8 +327,8 @@ static const struct ata_port_operations vsc_sata_ops = {
 	.qc_prep		= ata_qc_prep,
 	.qc_issue		= ata_qc_issue_prot,
 	.data_xfer		= ata_data_xfer,
-	.freeze			= ata_bmdma_freeze,
-	.thaw			= ata_bmdma_thaw,
+	.freeze			= vsc_freeze,
+	.thaw			= vsc_thaw,
 	.error_handler		= ata_bmdma_error_handler,
 	.post_internal_cmd	= ata_bmdma_post_internal_cmd,
 	.irq_handler		= vsc_sata_interrupt,

Re: [PATCH 2.6.21-rc1 1/2] sata_vsc: factor the error and normal intr paths into separate routines

[Tejun Heo]

Hello, Dan.

Dan Williams wrote:
> +static inline void vsc_error_intr(u8 port_status, struct ata_port *ap)
> +{
> +	if (port_status & (VSC_SATA_INT_PHY_CHANGE | VSC_SATA_INT_ERROR_M))
> +		ata_port_freeze(ap);
> +	else
> +		ata_port_abort(ap);
> +}
> +
> +static void vsc_port_intr(u8 port_status, struct ata_port *ap)
> +{
> +	struct ata_queued_cmd *qc;
> +	int handled = 0;
> +
> +	if (unlikely(port_status & VSC_SATA_INT_ERROR)) {
> +		vsc_error_intr(port_status, ap);
> +		return;
> +	}
> +
> +	qc = ata_qc_from_tag(ap, ap->active_tag);
> +	if (qc && likely(!(qc->tf.flags & ATA_TFLAG_POLLING)))
> +		handled = ata_host_intr(ap, qc);
> +
> +	/* We received an interrupt during a polled command,
> +	 * or some other spurious condition.  Interrupt reporting
> +	 * with this hardware is fairly reliable so it is safe to
> +	 * simply clear the interrupt
> +	 */
> +	if (unlikely(!handled))
> +		ata_chk_status(ap);
> +}

Sorry to keep nagging about patch separation and your way could be okay
too, but IMHO this can be done better in one of the following ways.

1. Keep two patches.  One to break down the interrupt handler the other
to improve it.  In this case the first one shouldn't introduce major new
features but just focus on breaking down existing one.

2. Do it in single patch.  It seems the changes are localized enough.
Just name it something like 'reimplement irq handler' and list changes
in the commit message.

After seeing the change, I'm leaning toward #2.  Thanks.  :-)

-- 
tejun

Re: [PATCH 2.6.21-rc1 2/2] sata_vsc: clean up vsc_sata_interrupt

[Tejun Heo]

Hello,

Just one nit.

Dan Williams wrote:
> +	if (unlikely(status == 0xffffffff || status == 0)) {
> +		status && printk(KERN_ERR DRV_NAME ": IRQ status == 0xffffffff, "
> +			"PCI fault or device removal?\n");

People usually don't like using logical operator short circuit for
branching.  Just put another if () and you can also use dev_printk() there.

Other than that, everything looks good.  Thanks for doing this.

-- 
tejun

Re: [PATCH 2.6.21-rc1 1/2] sata_vsc: factor the error and normal intr paths into separate routines

[Jeff Garzik]

Tejun Heo wrote:
> Sorry to keep nagging about patch separation and your way could be okay
> too, but IMHO this can be done better in one of the following ways.
> 
> 1. Keep two patches.  One to break down the interrupt handler the other
> to improve it.  In this case the first one shouldn't introduce major new
> features but just focus on breaking down existing one.
> 
> 2. Do it in single patch.  It seems the changes are localized enough.
> Just name it something like 'reimplement irq handler' and list changes
> in the commit message.
> 
> After seeing the change, I'm leaning toward #2.  Thanks.  :-)


These two sata_vsc changes should be done in a single patch.  "add new 
stuff" and "use new stuff" patches can be combined.  The types of 
patches that normally should be separated are logical changes like 
"clean up irq handling" or "improve EH", changes that can be 
git-bisect'd usefully.

The changes look OK to me.

	Jeff