Skip to contentSkip to conversation

(g_)NCR5380: Improve IRQ probing and some fixes

PATCH In review Ondrej Zary 2 participants 22 messages Oct 31, 2016 – Dec 3, 2016

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover (g_)NCR5380: Improve IRQ probing and some fixes

Hello,
this patch series improves IRQ probing and moves it from NCR5380 to g_NCR5380, adds IRQ auto-configuration for HP C2502 and enables all this by default. It also adds IRQ and base address checks to prevent hangs when wrong values are specified by user. There's also a small release region fix in error handling.

#1 lore raw

PATCH 1/6 NCR5380: Use probe_irq_*() for IRQ probing

Use standard probe_irq_on() and probe_irq_off() functions instead of own implementation.
This prevents warning messages like this in the kernel log:
genirq: Flags mismatch irq 1. 00000000 (NCR-probe) vs. 00000080 (i8042)

Move the IRQ trigger code to a separate function so it can be used for other purposes (testing if the IRQ works) and move the code from NCR5380 to g_NCR5380.

Also add missing IRQ reset before and after the probe.

---
 drivers/scsi/NCR5380.c   |   77 +---------------------------------------------
 drivers/scsi/NCR5380.h   |    1 -
 drivers/scsi/g_NCR5380.c |   50 +++++++++++++++++++++++++++++-
 3 files changed, 50 insertions(+), 78 deletions(-)
drivers/scsi/NCR5380.c +1 -76
index d849ffa..4f5ca79 100644
@@ -97,9 +97,6 @@
  * and macros and include this file in your driver.
  *
  * These macros control options :
- * AUTOPROBE_IRQ - if defined, the NCR5380_probe_irq() function will be
- * defined.
- *
  * AUTOSENSE - if defined, REQUEST SENSE will be performed automatically
  * for commands that return with a CHECK CONDITION status.
  *
@@ -127,9 +124,7 @@
  * NCR5380_dma_residual   - residual byte count
  *
  * The generic driver is initialized by calling NCR5380_init(instance),
- * after setting the appropriate host specific fields and ID.  If the
- * driver wishes to autoprobe for an IRQ line, the NCR5380_probe_irq(instance,
- * possible) function may be used.
+ * after setting the appropriate host specific fields and ID.
  */
 
 #ifndef NCR5380_io_delay
@@ -351,76 +346,6 @@ static void NCR5380_print_phase(struct Scsi_Host *instance)
 }
 #endif
 
-
-static int probe_irq;
-
-/**
- * probe_intr	-	helper for IRQ autoprobe
- * @irq: interrupt number
- * @dev_id: unused
- * @regs: unused
- *
- * Set a flag to indicate the IRQ in question was received. This is
- * used by the IRQ probe code.
- */
-
-static irqreturn_t probe_intr(int irq, void *dev_id)
-{
-	probe_irq = irq;
-	return IRQ_HANDLED;
-}
-
-/**
- * NCR5380_probe_irq	-	find the IRQ of an NCR5380
- * @instance: NCR5380 controller
- * @possible: bitmask of ISA IRQ lines
- *
- * Autoprobe for the IRQ line used by the NCR5380 by triggering an IRQ
- * and then looking to see what interrupt actually turned up.
- */
-
-static int __maybe_unused NCR5380_probe_irq(struct Scsi_Host *instance,
-						int possible)
-{
-	struct NCR5380_hostdata *hostdata = shost_priv(instance);
-	unsigned long timeout;
-	int trying_irqs, i, mask;
-
-	for (trying_irqs = 0, i = 1, mask = 2; i < 16; ++i, mask <<= 1)
-		if ((mask & possible) && (request_irq(i, &probe_intr, 0, "NCR-probe", NULL) == 0))
-			trying_irqs |= mask;
-
-	timeout = jiffies + msecs_to_jiffies(250);
-	probe_irq = NO_IRQ;
-
-	/*
-	 * A interrupt is triggered whenever BSY = false, SEL = true
-	 * and a bit set in the SELECT_ENABLE_REG is asserted on the
-	 * SCSI bus.
-	 *
-	 * Note that the bus is only driven when the phase control signals
-	 * (I/O, C/D, and MSG) match those in the TCR, so we must reset that
-	 * to zero.
-	 */
-
-	NCR5380_write(TARGET_COMMAND_REG, 0);
-	NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
-	NCR5380_write(OUTPUT_DATA_REG, hostdata->id_mask);
-	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE | ICR_ASSERT_DATA | ICR_ASSERT_SEL);
-
-	while (probe_irq == NO_IRQ && time_before(jiffies, timeout))
-		schedule_timeout_uninterruptible(1);
-
-	NCR5380_write(SELECT_ENABLE_REG, 0);
-	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
-
-	for (i = 1, mask = 2; i < 16; ++i, mask <<= 1)
-		if (trying_irqs & mask)
-			free_irq(i, NULL);
-
-	return probe_irq;
-}
-
 /**
  * NCR58380_info - report driver and host information
  * @instance: relevant scsi host instance
drivers/scsi/NCR5380.h +0 -1
index 3c6ce54..4724558 100644
@@ -290,7 +290,6 @@ static inline struct scsi_cmnd *NCR5380_to_scmd(struct NCR5380_cmd *ncmd_ptr)
 #define NCR5380_dprint_phase(flg, arg) do {} while (0)
 #endif
 
-static int NCR5380_probe_irq(struct Scsi_Host *instance, int possible);
 static int NCR5380_init(struct Scsi_Host *instance, int flags);
 static int NCR5380_maybe_reset_bus(struct Scsi_Host *);
 static void NCR5380_exit(struct Scsi_Host *instance);
drivers/scsi/g_NCR5380.c +49 -1

62 lines not on this page · raw message

Signature · 1 line
Ondrej Zary
#5 lore raw

PATCH 2/6 g_NCR5380: Test the IRQ before accepting it

In reply to Ondrej Zary

Trigger an IRQ first with a test IRQ handler to find out if it really works. Disable the IRQ if not.

This prevents hang when incorrect IRQ was specified by user.

---
 drivers/scsi/g_NCR5380.c |   44 +++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 41 insertions(+), 3 deletions(-)
drivers/scsi/g_NCR5380.c +41 -3
index 09c660b..0d1f6ad 100644
@@ -115,6 +115,32 @@ static int NCR5380_probe_irq(struct Scsi_Host *instance)
 	return irq;
 }
 
+static bool irq_working;
+
+static irqreturn_t test_irq(int irq, void *dev_id)
+{
+	irq_working = true;
+	return IRQ_HANDLED;
+}
+
+/* test if the IRQ is working */
+static int NCR5380_test_irq(struct Scsi_Host *instance, int irq)
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+
+	irq_working = false;
+	if (request_irq(irq, test_irq, 0, "NCR5380-irqtest", NULL))
+		return -EBUSY;
+	NCR5380_trigger_irq(instance);
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	free_irq(irq, NULL);
+
+	if (!irq_working)
+		return -EIO;
+
+	return 0;
+}
+
 /*
  * Configure I/O address of 53C400A or DTC436 by writing magic numbers
  * to ports 0x779 and 0x379.
@@ -323,9 +349,21 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 		/* set IRQ for HP C2502 */
 		if (board == BOARD_HP_C2502)
 			magic_configure(port_idx, instance->irq, magic);
-		if (request_irq(instance->irq, generic_NCR5380_intr,
-				0, "NCR5380", instance)) {
-			printk(KERN_WARNING "scsi%d : IRQ%d not free, interrupts disabled\n", instance->host_no, instance->irq);
+		ret = NCR5380_test_irq(instance, instance->irq);
+		if (ret) {
+			printk(KERN_WARNING "scsi%d : IRQ%d not %s, interrupts disabled\n",
+			       instance->host_no, instance->irq,
+			       (ret == -EBUSY) ? "free" : "working");
+			instance->irq = NO_IRQ;
+		}
+	}
+
+	if (instance->irq != NO_IRQ) {
+		if (request_irq(instance->irq, generic_NCR5380_intr, 0,
+				"NCR5380", instance)) {
+			printk(KERN_WARNING "scsi%d : IRQ%d not free, interrupts disabled\n",
+			       instance->host_no,
+			       instance->irq);
 			instance->irq = NO_IRQ;
 		}
 	}
Signature · 1 line
Ondrej Zary
#11 lore raw

PATCH 4/6 g_NCR5380: Add IRQ auto-configuration for HP C2502

In reply to Ondrej Zary

Find free and working IRQ automatically on HP C2502 cards. Also allow IRQ 9 to work (aliases to IRQ 2 on the card).

---
 drivers/scsi/g_NCR5380.c |   29 +++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)
drivers/scsi/g_NCR5380.c +27 -2
index e713dba..27fc499 100644
@@ -156,6 +156,8 @@ static void magic_configure(int idx, u8 irq, u8 magic[])
 	outb(magic[4], 0x379);
 
 	/* allowed IRQs for HP C2502 */
+	if (irq == 9)
+		irq = 2;
 	if (irq != 2 && irq != 3 && irq != 4 && irq != 5 && irq != 7)
 		irq = 0;
 	if (idx >= 0 && idx <= 7)
@@ -163,6 +165,21 @@ static void magic_configure(int idx, u8 irq, u8 magic[])
 	outb(cfg, 0x379);
 }
 
+/* find a free and working IRQ (for HP C2502) */
+static int NCR5380_find_irq(struct Scsi_Host *instance, u8 irqs[],
+			    int port_idx, u8 magic[])
+{
+	int i;
+
+	for (i = 0; irqs[i]; i++) {
+		magic_configure(port_idx, irqs[i], magic);
+		if (NCR5380_test_irq(instance, irqs[i]) == 0)
+			return irqs[i];
+	}
+	magic_configure(port_idx, 0, magic);
+	return NO_IRQ;
+}
+
 static unsigned int ncr_53c400a_ports[] = {
 	0x280, 0x290, 0x300, 0x310, 0x330, 0x340, 0x348, 0x350, 0
 };
@@ -175,6 +192,9 @@ static void magic_configure(int idx, u8 irq, u8 magic[])
 static u8 hp_c2502_magic[] = {	/* HP C2502 */
 	0x0f, 0x22, 0xf0, 0x20, 0x80
 };
+static u8 hp_c2502_irqs[] = {
+	9, 5, 7, 3, 4, 0
+};
 
 static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 			struct device *pdev, int base, int irq, int board)
@@ -345,8 +365,13 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 
 	if (irq != IRQ_AUTO)
 		instance->irq = irq;
-	else
-		instance->irq = NCR5380_probe_irq(instance);
+	else {
+		if (board == BOARD_HP_C2502)
+			instance->irq = NCR5380_find_irq(instance,
+						hp_c2502_irqs, port_idx, magic);
+		else
+			instance->irq = NCR5380_probe_irq(instance);
+	}
 
 	/* Compatibility with documented NCR5380 kernel parameters */
 	if (instance->irq == 255)
Signature · 1 line
Ondrej Zary
#17 lore raw

PATCH 5/6 g_NCR5380: Autoprobe IRQ by default

In reply to Ondrej Zary

IRQ probing seems to work fine now. Default to autoprobe for IRQ instead of disabling it.

---
 drivers/scsi/g_NCR5380.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/scsi/g_NCR5380.c +2 -2
index 27fc499..6a08d3e 100644
@@ -52,9 +52,9 @@
 module_param(dtc_3181e, int, 0);
 module_param(hp_c2502, int, 0);
 
-static int irq[] = { 0, 0, 0, 0, 0, 0, 0, 0 };
+static int irq[] = { IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO };
 module_param_array(irq, int, NULL, 0);
-MODULE_PARM_DESC(irq, "IRQ number(s)");
+MODULE_PARM_DESC(irq, "IRQ number(s) (0=disable, 254=auto [default])");
 
 static int base[] = { 0, 0, 0, 0, 0, 0, 0, 0 };
 module_param_array(base, int, NULL, 0);
Signature · 1 line
Ondrej Zary
#20 lore raw

PATCH 6/6 g_NCR5380: Fix release region in error handling

In reply to Ondrej Zary

When a SW-configurable card is specified but not found, the driver releases wrong region, causing the following message in kernel log:
Trying to free nonexistent resource <0000000000000000-000000000000000f>

Fix it by assigning base earlier.

---
 drivers/scsi/g_NCR5380.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/scsi/g_NCR5380.c +3 -3
index 6a08d3e..d33e71f 100644
@@ -258,12 +258,12 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 		if (ports[i]) {
 			/* At this point we have our region reserved */
 			magic_configure(i, 0, magic); /* no IRQ yet */
-			outb(0xc0, ports[i] + 9);
-			if (inb(ports[i] + 9) != 0x80) {
+			base = ports[i];
+			outb(0xc0, base + 9);
+			if (inb(base + 9) != 0x80) {
 				ret = -ENODEV;
 				goto out_release;
 			}
-			base = ports[i];
 			port_idx = i;
 		} else
 			return -EINVAL;
Signature · 1 line
Ondrej Zary
#9 lore raw

PATCH 3/6 g_NCR5380: Check for chip presence before calling NCR5380_init()

In reply to Ondrej Zary

Write and read back MODE_REG to check if the chip is really there before doing more initialization.

This prevents hang when incorrect I/O address was specified by user (in the most common case where no device is present there so all reads result in 0xff).

---
 drivers/scsi/g_NCR5380.c |    7 +++++++
 1 file changed, 7 insertions(+)
drivers/scsi/g_NCR5380.c +7 -0
index 0d1f6ad..e713dba 100644
@@ -322,6 +322,13 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 		}
 	}
 
+	/* check if the chip is really there */
+	NCR5380_write(MODE_REG, 0);
+	if (NCR5380_read(MODE_REG) != 0) {
+		ret = -ENODEV;
+		goto out_unregister;
+	}
+
 	ret = NCR5380_init(instance, flags | FLAG_LATE_DMA_SETUP);
 	if (ret)
 		goto out_unregister;
Signature · 1 line
Ondrej Zary
#2 lore raw

Re: [PATCH 1/6] NCR5380: Use probe_irq_*() for IRQ probing

In reply to Ondrej Zary
Quoted 137 lines · Ondrej Zary

Use standard probe_irq_on() and probe_irq_off() functions instead of own implementation. This prevents warning messages like this in the kernel
log: genirq: Flags mismatch irq 1. 00000000 (NCR-probe) vs. 00000080 (i8042)

Move the IRQ trigger code to a separate function so it can be used for other purposes (testing if the IRQ works) and move the code from NCR5380 to g_NCR5380.

Also add missing IRQ reset before and after the probe.

---
 drivers/scsi/NCR5380.c   |   77 +---------------------------------------------
 drivers/scsi/NCR5380.h   |    1 -
 drivers/scsi/g_NCR5380.c |   50 +++++++++++++++++++++++++++++-
 3 files changed, 50 insertions(+), 78 deletions(-)
drivers/scsi/NCR5380.c +1 -76
index d849ffa..4f5ca79 100644
@@ -97,9 +97,6 @@
  * and macros and include this file in your driver.
  *
  * These macros control options :
- * AUTOPROBE_IRQ - if defined, the NCR5380_probe_irq() function will be
- * defined.
- *
  * AUTOSENSE - if defined, REQUEST SENSE will be performed automatically
  * for commands that return with a CHECK CONDITION status.
  *
@@ -127,9 +124,7 @@
  * NCR5380_dma_residual   - residual byte count
  *
  * The generic driver is initialized by calling NCR5380_init(instance),
- * after setting the appropriate host specific fields and ID.  If the
- * driver wishes to autoprobe for an IRQ line, the NCR5380_probe_irq(instance,
- * possible) function may be used.
+ * after setting the appropriate host specific fields and ID.
  */
 
 #ifndef NCR5380_io_delay
@@ -351,76 +346,6 @@ static void NCR5380_print_phase(struct Scsi_Host *instance)
 }
 #endif
 
-
-static int probe_irq;
-
-/**
- * probe_intr	-	helper for IRQ autoprobe
- * @irq: interrupt number
- * @dev_id: unused
- * @regs: unused
- *
- * Set a flag to indicate the IRQ in question was received. This is
- * used by the IRQ probe code.
- */
-
-static irqreturn_t probe_intr(int irq, void *dev_id)
-{
-	probe_irq = irq;
-	return IRQ_HANDLED;
-}
-
-/**
- * NCR5380_probe_irq	-	find the IRQ of an NCR5380
- * @instance: NCR5380 controller
- * @possible: bitmask of ISA IRQ lines
- *
- * Autoprobe for the IRQ line used by the NCR5380 by triggering an IRQ
- * and then looking to see what interrupt actually turned up.
- */
-
-static int __maybe_unused NCR5380_probe_irq(struct Scsi_Host *instance,
-						int possible)
-{
-	struct NCR5380_hostdata *hostdata = shost_priv(instance);
-	unsigned long timeout;
-	int trying_irqs, i, mask;
-
-	for (trying_irqs = 0, i = 1, mask = 2; i < 16; ++i, mask <<= 1)
-		if ((mask & possible) && (request_irq(i, &probe_intr, 0, "NCR-probe", NULL) == 0))
-			trying_irqs |= mask;
-
-	timeout = jiffies + msecs_to_jiffies(250);
-	probe_irq = NO_IRQ;
-
-	/*
-	 * A interrupt is triggered whenever BSY = false, SEL = true
-	 * and a bit set in the SELECT_ENABLE_REG is asserted on the
-	 * SCSI bus.
-	 *
-	 * Note that the bus is only driven when the phase control signals
-	 * (I/O, C/D, and MSG) match those in the TCR, so we must reset that
-	 * to zero.
-	 */
-
-	NCR5380_write(TARGET_COMMAND_REG, 0);
-	NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
-	NCR5380_write(OUTPUT_DATA_REG, hostdata->id_mask);
-	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE | ICR_ASSERT_DATA | ICR_ASSERT_SEL);
-
-	while (probe_irq == NO_IRQ && time_before(jiffies, timeout))
-		schedule_timeout_uninterruptible(1);
-
-	NCR5380_write(SELECT_ENABLE_REG, 0);
-	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
-
-	for (i = 1, mask = 2; i < 16; ++i, mask <<= 1)
-		if (trying_irqs & mask)
-			free_irq(i, NULL);
-
-	return probe_irq;
-}
-
 /**
  * NCR58380_info - report driver and host information
  * @instance: relevant scsi host instance
drivers/scsi/NCR5380.h +0 -1
index 3c6ce54..4724558 100644
@@ -290,7 +290,6 @@ static inline struct scsi_cmnd *NCR5380_to_scmd(struct NCR5380_cmd *ncmd_ptr)
 #define NCR5380_dprint_phase(flg, arg) do {} while (0)
 #endif
 
-static int NCR5380_probe_irq(struct Scsi_Host *instance, int possible);
 static int NCR5380_init(struct Scsi_Host *instance, int flags);
 static int NCR5380_maybe_reset_bus(struct Scsi_Host *);
 static void NCR5380_exit(struct Scsi_Host *instance);
drivers/scsi/g_NCR5380.c +0 -0
index 7299ad9..09c660b 100644

drivers/scsi/g_NCR5380.c line 70

Quoted 4 lines
drivers/scsi/g_NCR5380.c +1 -0
@@ -67,6 +67,54 @@
 MODULE_ALIAS("g_NCR5380_mmio");
 MODULE_LICENSE("GPL");
 
+static void NCR5380_trigger_irq(struct Scsi_Host *instance)

That should be 'g_NCR5380_trigger_irq' (if it needs a prefix).

drivers/scsi/g_NCR5380.c line 83

Quoted 13 lines
drivers/scsi/g_NCR5380.c +13 -0
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+
+	/*
+	 * An interrupt is triggered whenever BSY = false, SEL = true
+	 * and a bit set in the SELECT_ENABLE_REG is asserted on the
+	 * SCSI bus.
+	 *
+	 * Note that the bus is only driven when the phase control signals
+	 * (I/O, C/D, and MSG) match those in the TCR, so we must reset that
+	 * to zero.
+	 */
+	NCR5380_write(TARGET_COMMAND_REG, 0);

That assumes bus free phase, which we hope is the case but NCR5380_maybe_reset_bus() doesn't guarantee it. This is more robust:

	 * Note that the bus is only driven when the phase control signals
	 * (I/O, C/D, and MSG) match those in the TCR.
	 */
	NCR5380_write(TARGET_COMMAND_REG,
		PHASE_SR_TO_TCR(NCR5380_read(STATUS_REG) & PHASE_MASK));

drivers/scsi/g_NCR5380.c line 88

Quoted 5 lines
drivers/scsi/g_NCR5380.c +5 -0
+	NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
+	NCR5380_write(OUTPUT_DATA_REG, hostdata->id_mask);
+	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE | ICR_ASSERT_DATA | ICR_ASSERT_SEL);
+
+	usleep_range(1000, 20000);

Again, msleep(1) would be more appropriate here.

drivers/scsi/g_NCR5380.c line 95

Quoted 7 lines
drivers/scsi/g_NCR5380.c +7 -0
+
+	NCR5380_write(SELECT_ENABLE_REG, 0);
+	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
+}
+
+/**
+ * NCR5380_probe_irq	-	find the IRQ of an NCR5380

This should be 'g_NCR5380_probe_irq' (if it needs a prefix).

Quoted 35 lines
drivers/scsi/g_NCR5380.c +23 -1
+ * @instance: NCR5380 controller
+ *
+ * Autoprobe for the IRQ line used by the NCR5380 by triggering an IRQ
+ * and then looking to see what interrupt actually turned up.
+ */
+
+static int NCR5380_probe_irq(struct Scsi_Host *instance)
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+	int irq_mask, irq;
+
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	irq_mask = probe_irq_on();
+	NCR5380_trigger_irq(instance);
+	irq = probe_irq_off(irq_mask);
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	if (irq < 0)
+		irq = NO_IRQ;
+
+	return irq;
+}
+
 /*
  * Configure I/O address of 53C400A or DTC436 by writing magic numbers
  * to ports 0x779 and 0x379.
@@ -265,7 +313,7 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 	if (irq != IRQ_AUTO)
 		instance->irq = irq;
 	else
-		instance->irq = NCR5380_probe_irq(instance, 0xffff);
+		instance->irq = NCR5380_probe_irq(instance);
 
 	/* Compatibility with documented NCR5380 kernel parameters */
 	if (instance->irq == 255)
#6 lore raw

Re: [PATCH 2/6] g_NCR5380: Test the IRQ before accepting it

In reply to Ondrej Zary
Quoted 5 lines · Ondrej Zary

Trigger an IRQ first with a test IRQ handler to find out if it really works. Disable the IRQ if not.

This prevents hang when incorrect IRQ was specified by user.

Once again, how does it cause a hang?

If the user specifies an IRQ, we should trust them. If they don't specify an IRQ then probe (as in patch 5/6).

Quoted 29 lines
---
 drivers/scsi/g_NCR5380.c |   44 +++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 41 insertions(+), 3 deletions(-)
drivers/scsi/g_NCR5380.c +16 -0
index 09c660b..0d1f6ad 100644
@@ -115,6 +115,32 @@ static int NCR5380_probe_irq(struct Scsi_Host *instance)
 	return irq;
 }
 
+static bool irq_working;
+
+static irqreturn_t test_irq(int irq, void *dev_id)
+{
+	irq_working = true;
+	return IRQ_HANDLED;
+}
+
+/* test if the IRQ is working */
+static int NCR5380_test_irq(struct Scsi_Host *instance, int irq)
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+
+	irq_working = false;
+	if (request_irq(irq, test_irq, 0, "NCR5380-irqtest", NULL))
+		return -EBUSY;

drivers/scsi/g_NCR5380.c static int NCR5380_probe_irq(struct Scsi_Host *instance) · line 143

Quoted 10 lines
drivers/scsi/g_NCR5380.c +10 -0
+	NCR5380_trigger_irq(instance);
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	free_irq(irq, NULL);
+
+	if (!irq_working)
+		return -EIO;
+
+	return 0;
+}
+

IMO the extra complexity is not justified by the possibility of machines with misconfigured BIOS or incorrect module parameters.

I don't want another irq probing mechanism. We just got rid of one.

But I won't mind if the SCSI maintainers who know ISA better than I do would like to review this.

Quoted 29 lines
drivers/scsi/g_NCR5380.c +15 -3
 /*
  * Configure I/O address of 53C400A or DTC436 by writing magic numbers
  * to ports 0x779 and 0x379.
@@ -323,9 +349,21 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 		/* set IRQ for HP C2502 */
 		if (board == BOARD_HP_C2502)
 			magic_configure(port_idx, instance->irq, magic);
-		if (request_irq(instance->irq, generic_NCR5380_intr,
-				0, "NCR5380", instance)) {
-			printk(KERN_WARNING "scsi%d : IRQ%d not free, interrupts disabled\n", instance->host_no, instance->irq);
+		ret = NCR5380_test_irq(instance, instance->irq);
+		if (ret) {
+			printk(KERN_WARNING "scsi%d : IRQ%d not %s, interrupts disabled\n",
+			       instance->host_no, instance->irq,
+			       (ret == -EBUSY) ? "free" : "working");
+			instance->irq = NO_IRQ;
+		}
+	}
+
+	if (instance->irq != NO_IRQ) {
+		if (request_irq(instance->irq, generic_NCR5380_intr, 0,
+				"NCR5380", instance)) {
+			printk(KERN_WARNING "scsi%d : IRQ%d not free, interrupts disabled\n",
+			       instance->host_no,
+			       instance->irq);
 			instance->irq = NO_IRQ;
 		}
 	}
#10 lore raw

Re: [PATCH 3/6] g_NCR5380: Check for chip presence before calling NCR5380_init()

In reply to Ondrej Zary
Quoted 17 lines · Ondrej Zary

Write and read back MODE_REG to check if the chip is really there before doing more initialization.

This prevents hang when incorrect I/O address was specified by user (in the most common case where no device is present there so all reads result in 0xff).

---
 drivers/scsi/g_NCR5380.c |    7 +++++++
 1 file changed, 7 insertions(+)
drivers/scsi/g_NCR5380.c +0 -0
index 0d1f6ad..e713dba 100644

drivers/scsi/g_NCR5380.c static int generic_NCR5380_init_one(struct scsi_host_template *tpnt, · line 327

Quoted 6 lines
drivers/scsi/g_NCR5380.c +3 -0
@@ -322,6 +322,13 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 		}
 	}
 
+	/* check if the chip is really there */
+	NCR5380_write(MODE_REG, 0);
+	if (NCR5380_read(MODE_REG) != 0) {

Wouldn't it be more accurate to write,

	/* Check for some kind of device. A vacant slot reads 0xff. */

?

Quoted 8 lines
drivers/scsi/g_NCR5380.c +4 -0
+		ret = -ENODEV;
+		goto out_unregister;
+	}
+
 	ret = NCR5380_init(instance, flags | FLAG_LATE_DMA_SETUP);
 	if (ret)
 		goto out_unregister;
#12 lore raw

Re: [PATCH 4/6] g_NCR5380: Add IRQ auto-configuration for HP C2502

In reply to Ondrej Zary
Quoted 25 lines · Ondrej Zary

Find free and working IRQ automatically on HP C2502 cards. Also allow IRQ 9 to work (aliases to IRQ 2 on the card).

---
 drivers/scsi/g_NCR5380.c |   29 +++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)
drivers/scsi/g_NCR5380.c +2 -0
index e713dba..27fc499 100644
@@ -156,6 +156,8 @@ static void magic_configure(int idx, u8 irq, u8 magic[])
 	outb(magic[4], 0x379);
 
 	/* allowed IRQs for HP C2502 */
+	if (irq == 9)
+		irq = 2;
 	if (irq != 2 && irq != 3 && irq != 4 && irq != 5 && irq != 7)
 		irq = 0;
 	if (idx >= 0 && idx <= 7)
@@ -163,6 +165,21 @@ static void magic_configure(int idx, u8 irq, u8 magic[])
 	outb(cfg, 0x379);
 }
 

drivers/scsi/g_NCR5380.c static void magic_configure(int idx, u8 irq, u8 magic[]) · line 177

Quoted 10 lines
drivers/scsi/g_NCR5380.c +10 -0
+/* find a free and working IRQ (for HP C2502) */
+static int NCR5380_find_irq(struct Scsi_Host *instance, u8 irqs[],
+			    int port_idx, u8 magic[])
+{
+	int i;
+
+	for (i = 0; irqs[i]; i++) {
+		magic_configure(port_idx, irqs[i], magic);
+		if (NCR5380_test_irq(instance, irqs[i]) == 0)
+			return irqs[i];

The NCR5380_test_irq routine in patch 2/6 doesn't work for shared irqs, so you may not get the IRQ you would expect. (The core driver does support shared irqs, BTW.)

Also, you've ignored the irq module parameters. From the user's point of view, surely the least surprising thing is to attempt to configure the card for whatever irq the user asked for.

If the specified irq isn't supported by the board, just log an error and fail. If you want to be user friendly, print a message to tell them what irqs the card supports.

If the user asks for IRQ_AUTO, just configure the board for a hard-coded default, say 9, and print a warning message to say so.

Either way, if request_irq fails just continue with NO_IRQ, as per usual.

To me that's the most flexible and least surprising behaviour. But again, if someone with more ISA knowledge wishes to weigh in, that's fine too.

Quoted 35 lines
drivers/scsi/g_NCR5380.c +15 -2
+	}
+	magic_configure(port_idx, 0, magic);
+	return NO_IRQ;
+}
+
 static unsigned int ncr_53c400a_ports[] = {
 	0x280, 0x290, 0x300, 0x310, 0x330, 0x340, 0x348, 0x350, 0
 };
@@ -175,6 +192,9 @@ static void magic_configure(int idx, u8 irq, u8 magic[])
 static u8 hp_c2502_magic[] = {	/* HP C2502 */
 	0x0f, 0x22, 0xf0, 0x20, 0x80
 };
+static u8 hp_c2502_irqs[] = {
+	9, 5, 7, 3, 4, 0
+};
 
 static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 			struct device *pdev, int base, int irq, int board)
@@ -345,8 +365,13 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 
 	if (irq != IRQ_AUTO)
 		instance->irq = irq;
-	else
-		instance->irq = NCR5380_probe_irq(instance);
+	else {
+		if (board == BOARD_HP_C2502)
+			instance->irq = NCR5380_find_irq(instance,
+						hp_c2502_irqs, port_idx, magic);
+		else
+			instance->irq = NCR5380_probe_irq(instance);
+	}
 
 	/* Compatibility with documented NCR5380 kernel parameters */
 	if (instance->irq == 255)
#18 lore raw

Re: [PATCH 5/6] g_NCR5380: Autoprobe IRQ by default

In reply to Ondrej Zary
Quoted 15 lines · Ondrej Zary

IRQ probing seems to work fine now. Default to autoprobe for IRQ instead of disabling it.

---
 drivers/scsi/g_NCR5380.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/scsi/g_NCR5380.c +0 -0
index 27fc499..6a08d3e 100644
@@ -52,9 +52,9 @@
 module_param(dtc_3181e, int, 0);

drivers/scsi/g_NCR5380.c line 60

Quoted 10 lines
drivers/scsi/g_NCR5380.c +2 -2
 module_param(hp_c2502, int, 0);
 
-static int irq[] = { 0, 0, 0, 0, 0, 0, 0, 0 };
+static int irq[] = { IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO, IRQ_AUTO };
 module_param_array(irq, int, NULL, 0);
-MODULE_PARM_DESC(irq, "IRQ number(s)");
+MODULE_PARM_DESC(irq, "IRQ number(s) (0=disable, 254=auto [default])");
 
 static int base[] = { 0, 0, 0, 0, 0, 0, 0, 0 };
 module_param_array(base, int, NULL, 0);

I think this patch is incomplete and you should add these changes:

drivers/scsi/g_NCR5380.c +2 -2
index 7299ad9..0bf0322 100644
@@ -44,7 +44,7 @@ static int ncr_53c400;
 static int ncr_53c400a;
 static int dtc_3181e;
 static int hp_c2502;
-module_param(ncr_irq, int, 0);
+module_param(ncr_irq, int, IRQ_AUTO);
 module_param(ncr_addr, int, 0);
 module_param(ncr_5380, int, 0);
 module_param(ncr_53c400, int, 0);
@@ -597,7 +597,7 @@ static int __init generic_NCR5380_init(void)
 	int ret = 0;
 
 	/* compatibility with old-style parameters */
-	if (irq[0] == 0 && base[0] == 0 && card[0] == -1) {
+	if (irq[0] == IRQ_AUTO && base[0] == 0 && card[0] == -1) {
 		irq[0] = ncr_irq;
 		base[0] = ncr_addr;
 		if (ncr_5380)
Acked-by on v1 6/6
#21 lore raw

Re: [PATCH 6/6] g_NCR5380: Fix release region in error handling

In reply to Ondrej Zary
Quoted 7 lines · Ondrej Zary

When a SW-configurable card is specified but not found, the driver releases wrong region, causing the following message in kernel log:
Trying to free nonexistent resource <0000000000000000-000000000000000f>

Fix it by assigning base earlier.

This fix is not related to the others in this series. It could go at the beginning where it is easy to cherry-pick.

Fixes a8cfbcaec0c1 ("scsi: g_NCR5380: Stop using scsi_module.c")
Quoted 25 lines
---
 drivers/scsi/g_NCR5380.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/scsi/g_NCR5380.c +3 -3
index 6a08d3e..d33e71f 100644
@@ -258,12 +258,12 @@ static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 		if (ports[i]) {
 			/* At this point we have our region reserved */
 			magic_configure(i, 0, magic); /* no IRQ yet */
-			outb(0xc0, ports[i] + 9);
-			if (inb(ports[i] + 9) != 0x80) {
+			base = ports[i];
+			outb(0xc0, base + 9);
+			if (inb(base + 9) != 0x80) {
 				ret = -ENODEV;
 				goto out_release;
 			}
-			base = ports[i];
 			port_idx = i;
 		} else
 			return -EINVAL;
#3 lore raw

Re: [PATCH 1/6] NCR5380: Use probe_irq_*() for IRQ probing

In reply to Finn Thain
Quoted 170 lines · Finn Thain
Quoted 139 lines · Ondrej Zary

Use standard probe_irq_on() and probe_irq_off() functions instead of own implementation. This prevents warning messages like this in the kernel
log: genirq: Flags mismatch irq 1. 00000000 (NCR-probe) vs. 00000080 (i8042)

Move the IRQ trigger code to a separate function so it can be used for other purposes (testing if the IRQ works) and move the code from NCR5380 to g_NCR5380.

Also add missing IRQ reset before and after the probe.

---
 drivers/scsi/NCR5380.c   |   77
+--------------------------------------------- drivers/scsi/NCR5380.h   |
   1 -
 drivers/scsi/g_NCR5380.c |   50 +++++++++++++++++++++++++++++-
 3 files changed, 50 insertions(+), 78 deletions(-)
drivers/scsi/NCR5380.c +0 -5
index d849ffa..4f5ca79 100644
@@ -97,9 +97,6 @@
  * and macros and include this file in your driver.
  *
  * These macros control options :
- * AUTOPROBE_IRQ - if defined, the NCR5380_probe_irq() function will be
- * defined.
- *
  * AUTOSENSE - if defined, REQUEST SENSE will be performed automatically
  * for commands that return with a CHECK CONDITION status.
  *
@@ -127,9 +124,7 @@
  * NCR5380_dma_residual   - residual byte count
  *
  * The generic driver is initialized by calling NCR5380_init(instance),
- * after setting the appropriate host specific fields and ID.  If the
- * driver wishes to autoprobe for an IRQ line, the

NCR5380_probe_irq(instance, - * possible) function may be used.
+ * after setting the appropriate host specific fields and ID. */

#ifndef NCR5380_io_delay

drivers/scsi/NCR5380.c +0 -55
@@ -351,76 +346,6 @@ static void NCR5380_print_phase(struct Scsi_Host
 *instance) }
 #endif
 
-
-static int probe_irq;
-
-/**
- * probe_intr	-	helper for IRQ autoprobe
- * @irq: interrupt number
- * @dev_id: unused
- * @regs: unused
- *
- * Set a flag to indicate the IRQ in question was received. This is
- * used by the IRQ probe code.
- */
-
-static irqreturn_t probe_intr(int irq, void *dev_id)
-{
-	probe_irq = irq;
-	return IRQ_HANDLED;
-}
-
-/**
- * NCR5380_probe_irq	-	find the IRQ of an NCR5380
- * @instance: NCR5380 controller
- * @possible: bitmask of ISA IRQ lines
- *
- * Autoprobe for the IRQ line used by the NCR5380 by triggering an IRQ
- * and then looking to see what interrupt actually turned up.
- */
-
-static int __maybe_unused NCR5380_probe_irq(struct Scsi_Host *instance,
-						int possible)
-{
-	struct NCR5380_hostdata *hostdata = shost_priv(instance);
-	unsigned long timeout;
-	int trying_irqs, i, mask;
-
-	for (trying_irqs = 0, i = 1, mask = 2; i < 16; ++i, mask <<= 1)
-		if ((mask & possible) && (request_irq(i, &probe_intr, 0, "NCR-probe",
 NULL) == 0)) -			trying_irqs |= mask;
-
-	timeout = jiffies + msecs_to_jiffies(250);
-	probe_irq = NO_IRQ;
-
-	/*
-	 * A interrupt is triggered whenever BSY = false, SEL = true
-	 * and a bit set in the SELECT_ENABLE_REG is asserted on the
-	 * SCSI bus.
-	 *
-	 * Note that the bus is only driven when the phase control signals
-	 * (I/O, C/D, and MSG) match those in the TCR, so we must reset that
-	 * to zero.
-	 */
-
-	NCR5380_write(TARGET_COMMAND_REG, 0);
-	NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
-	NCR5380_write(OUTPUT_DATA_REG, hostdata->id_mask);
-	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE | ICR_ASSERT_DATA |

ICR_ASSERT_SEL); -

drivers/scsi/NCR5380.c +0 -13
-	while (probe_irq == NO_IRQ && time_before(jiffies, timeout))
-		schedule_timeout_uninterruptible(1);
-
-	NCR5380_write(SELECT_ENABLE_REG, 0);
-	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
-
-	for (i = 1, mask = 2; i < 16; ++i, mask <<= 1)
-		if (trying_irqs & mask)
-			free_irq(i, NULL);
-
-	return probe_irq;
-}
-
 /**
  * NCR58380_info - report driver and host information
  * @instance: relevant scsi host instance
drivers/scsi/NCR5380.h +0 -1
index 3c6ce54..4724558 100644
@@ -290,7 +290,6 @@ static inline struct scsi_cmnd
 *NCR5380_to_scmd(struct NCR5380_cmd *ncmd_ptr) #define
 NCR5380_dprint_phase(flg, arg) do {} while (0)
 #endif
 
-static int NCR5380_probe_irq(struct Scsi_Host *instance, int possible);
 static int NCR5380_init(struct Scsi_Host *instance, int flags);
 static int NCR5380_maybe_reset_bus(struct Scsi_Host *);

static void NCR5380_exit(struct Scsi_Host *instance);

drivers/scsi/g_NCR5380.c +0 -0
index 7299ad9..09c660b 100644

drivers/scsi/g_NCR5380.c line 70

Quoted 4 lines
drivers/scsi/g_NCR5380.c +1 -0
@@ -67,6 +67,54 @@
 MODULE_ALIAS("g_NCR5380_mmio");
 MODULE_LICENSE("GPL");
 
+static void NCR5380_trigger_irq(struct Scsi_Host *instance)

That should be 'g_NCR5380_trigger_irq' (if it needs a prefix).

drivers/scsi/g_NCR5380.c line 83

Quoted 13 lines
drivers/scsi/g_NCR5380.c +13 -0
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+
+	/*
+	 * An interrupt is triggered whenever BSY = false, SEL = true
+	 * and a bit set in the SELECT_ENABLE_REG is asserted on the
+	 * SCSI bus.
+	 *
+	 * Note that the bus is only driven when the phase control signals
+	 * (I/O, C/D, and MSG) match those in the TCR, so we must reset that
+	 * to zero.
+	 */
+	NCR5380_write(TARGET_COMMAND_REG, 0);

That assumes bus free phase, which we hope is the case but NCR5380_maybe_reset_bus() doesn't guarantee it. This is more robust:

	 * Note that the bus is only driven when the phase control signals
	 * (I/O, C/D, and MSG) match those in the TCR.
	 */
	NCR5380_write(TARGET_COMMAND_REG,
		PHASE_SR_TO_TCR(NCR5380_read(STATUS_REG) & PHASE_MASK));
Quoted 7 lines

drivers/scsi/g_NCR5380.c line 88

Quoted 5 lines
drivers/scsi/g_NCR5380.c +3 -0
+	NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
+	NCR5380_write(OUTPUT_DATA_REG, hostdata->id_mask);
+	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE | ICR_ASSERT_DATA |

ICR_ASSERT_SEL); +

drivers/scsi/g_NCR5380.c +1 -0
+	usleep_range(1000, 20000);

Again, msleep(1) would be more appropriate here.

WARNING: msleep < 20ms can sleep for up to 20ms; see Documentation/timers/timers-howto.txt #164: FILE: drivers/scsi/g_NCR5380.c:88:
+ msleep(1);

Quoted 44 lines

drivers/scsi/g_NCR5380.c line 95

Quoted 7 lines
drivers/scsi/g_NCR5380.c +7 -0
+
+	NCR5380_write(SELECT_ENABLE_REG, 0);
+	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE);
+}
+
+/**
+ * NCR5380_probe_irq	-	find the IRQ of an NCR5380

This should be 'g_NCR5380_probe_irq' (if it needs a prefix).

Quoted 34 lines

+ * @instance: NCR5380 controller
+ *
+ * Autoprobe for the IRQ line used by the NCR5380 by triggering an IRQ
+ * and then looking to see what interrupt actually turned up.
+ */

drivers/scsi/g_NCR5380.c +18 -1
+
+static int NCR5380_probe_irq(struct Scsi_Host *instance)
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+	int irq_mask, irq;
+
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	irq_mask = probe_irq_on();
+	NCR5380_trigger_irq(instance);
+	irq = probe_irq_off(irq_mask);
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	if (irq < 0)
+		irq = NO_IRQ;
+
+	return irq;
+}
+
 /*
  * Configure I/O address of 53C400A or DTC436 by writing magic numbers
  * to ports 0x779 and 0x379.
@@ -265,7 +313,7 @@ static int generic_NCR5380_init_one(struct
 scsi_host_template *tpnt, if (irq != IRQ_AUTO)
 		instance->irq = irq;
 	else
-		instance->irq = NCR5380_probe_irq(instance, 0xffff);
+		instance->irq = NCR5380_probe_irq(instance);
 
 	/* Compatibility with documented NCR5380 kernel parameters */
 	if (instance->irq == 255)
Signature · 1 line
Ondrej Zary
#13 lore raw

Re: [PATCH 4/6] g_NCR5380: Add IRQ auto-configuration for HP C2502

In reply to Finn Thain
Quoted 26 lines · Finn Thain
Quoted 25 lines · Ondrej Zary

Find free and working IRQ automatically on HP C2502 cards. Also allow IRQ 9 to work (aliases to IRQ 2 on the card).

---
 drivers/scsi/g_NCR5380.c |   29 +++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)
drivers/scsi/g_NCR5380.c +2 -0
index e713dba..27fc499 100644
@@ -156,6 +156,8 @@ static void magic_configure(int idx, u8 irq, u8
 magic[]) outb(magic[4], 0x379);
 
 	/* allowed IRQs for HP C2502 */
+	if (irq == 9)
+		irq = 2;
 	if (irq != 2 && irq != 3 && irq != 4 && irq != 5 && irq != 7)
 		irq = 0;
 	if (idx >= 0 && idx <= 7)
@@ -163,6 +165,21 @@ static void magic_configure(int idx, u8 irq, u8
 magic[]) outb(cfg, 0x379);
 }
 
Quoted 14 lines

drivers/scsi/g_NCR5380.c static void magic_configure(int idx, u8 irq, u8 magic[]) · line 177

Quoted 10 lines
drivers/scsi/g_NCR5380.c +10 -0
+/* find a free and working IRQ (for HP C2502) */
+static int NCR5380_find_irq(struct Scsi_Host *instance, u8 irqs[],
+			    int port_idx, u8 magic[])
+{
+	int i;
+
+	for (i = 0; irqs[i]; i++) {
+		magic_configure(port_idx, irqs[i], magic);
+		if (NCR5380_test_irq(instance, irqs[i]) == 0)
+			return irqs[i];

The NCR5380_test_irq routine in patch 2/6 doesn't work for shared irqs, so you may not get the IRQ you would expect. (The core driver does support shared irqs, BTW.)

ISA bus does not support IRQ sharing.

Quoted 3 lines

Also, you've ignored the irq module parameters. From the user's point of view, surely the least surprising thing is to attempt to configure the card for whatever irq the user asked for.

I haven't. NCR5380_find_irq is only called when irq is set to IRQ_AUTO.

Quoted 3 lines

If the specified irq isn't supported by the board, just log an error and fail. If you want to be user friendly, print a message to tell them what irqs the card supports.

If the IRQ is not supported (or does not work), user gets a warning and the driver continues with IRQ disabled.

Quoted 2 lines

If the user asks for IRQ_AUTO, just configure the board for a hard-coded default, say 9, and print a warning message to say so.

The card is almost Plug&Play. The base address is already configured automatically by the driver so doing the same for IRQ makes sense.

Quoted 40 lines

Either way, if request_irq fails just continue with NO_IRQ, as per usual.

To me that's the most flexible and least surprising behaviour. But again, if someone with more ISA knowledge wishes to weigh in, that's fine too.

Quoted 35 lines
drivers/scsi/g_NCR5380.c +5 -0
+	}
+	magic_configure(port_idx, 0, magic);
+	return NO_IRQ;
+}
+
 static unsigned int ncr_53c400a_ports[] = {
 	0x280, 0x290, 0x300, 0x310, 0x330, 0x340, 0x348, 0x350, 0
 };
@@ -175,6 +192,9 @@ static void magic_configure(int idx, u8 irq, u8

magic[]) static u8 hp_c2502_magic[] = { /* HP C2502 */

drivers/scsi/g_NCR5380.c +10 -2
 	0x0f, 0x22, 0xf0, 0x20, 0x80
 };
+static u8 hp_c2502_irqs[] = {
+	9, 5, 7, 3, 4, 0
+};
 
 static int generic_NCR5380_init_one(struct scsi_host_template *tpnt,
 			struct device *pdev, int base, int irq, int board)
@@ -345,8 +365,13 @@ static int generic_NCR5380_init_one(struct
 scsi_host_template *tpnt,
 
 	if (irq != IRQ_AUTO)
 		instance->irq = irq;
-	else
-		instance->irq = NCR5380_probe_irq(instance);
+	else {
+		if (board == BOARD_HP_C2502)
+			instance->irq = NCR5380_find_irq(instance,
+						hp_c2502_irqs, port_idx, magic);
+		else
+			instance->irq = NCR5380_probe_irq(instance);
+	}
 
 	/* Compatibility with documented NCR5380 kernel parameters */

if (instance->irq == 255)

Signature · 1 line
Ondrej Zary
#7 lore raw

Re: [PATCH 2/6] g_NCR5380: Test the IRQ before accepting it

In reply to Finn Thain
Quoted 7 lines · On Wednesday 02 November 2016 08:45:26 Finn Thain wrote:
Quoted 4 lines · Ondrej Zary

Trigger an IRQ first with a test IRQ handler to find out if it really works. Disable the IRQ if not.

This prevents hang when incorrect IRQ was specified by user.

Once again, how does it cause a hang?

Kernel scans the bus, finds a HDD, then attempts to read MBR. modprobe process is stuck but the system is still running. Then the transfer probably times out and everything locks up hard, even fbcon cursor stops blinking. I guess that kernel is trying to abort or reset.
BTW. rescan-scsi-bus also causes hang, anytime, even without IRQ.

Quoted 81 lines

If the user specifies an IRQ, we should trust them. If they don't specify an IRQ then probe (as in patch 5/6).

Quoted 30 lines
---
 drivers/scsi/g_NCR5380.c |   44

+++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 41 insertions(+), 3 deletions(-)

drivers/scsi/g_NCR5380.c +16 -0
index 09c660b..0d1f6ad 100644
@@ -115,6 +115,32 @@ static int NCR5380_probe_irq(struct Scsi_Host
 *instance) return irq;
 }
 
+static bool irq_working;
+
+static irqreturn_t test_irq(int irq, void *dev_id)
+{
+	irq_working = true;
+	return IRQ_HANDLED;
+}
+
+/* test if the IRQ is working */
+static int NCR5380_test_irq(struct Scsi_Host *instance, int irq)
+{
+	struct NCR5380_hostdata *hostdata = shost_priv(instance);
+
+	irq_working = false;
+	if (request_irq(irq, test_irq, 0, "NCR5380-irqtest", NULL))
+		return -EBUSY;

drivers/scsi/g_NCR5380.c static int NCR5380_probe_irq(struct Scsi_Host · line 143

Quoted 10 lines
drivers/scsi/g_NCR5380.c +10 -0
+	NCR5380_trigger_irq(instance);
+	NCR5380_read(RESET_PARITY_INTERRUPT_REG);
+	free_irq(irq, NULL);
+
+	if (!irq_working)
+		return -EIO;
+
+	return 0;
+}
+

IMO the extra complexity is not justified by the possibility of machines with misconfigured BIOS or incorrect module parameters.

I don't want another irq probing mechanism. We just got rid of one.

But I won't mind if the SCSI maintainers who know ISA better than I do would like to review this.

Quoted 29 lines
drivers/scsi/g_NCR5380.c +0 -0
 /*
  * Configure I/O address of 53C400A or DTC436 by writing magic numbers
  * to ports 0x779 and 0x379.
@@ -323,9 +349,21 @@ static int generic_NCR5380_init_one(struct

scsi_host_template *tpnt, /* set IRQ for HP C2502 */

drivers/scsi/g_NCR5380.c +0 -3
 		if (board == BOARD_HP_C2502)
 			magic_configure(port_idx, instance->irq, magic);
-		if (request_irq(instance->irq, generic_NCR5380_intr,
-				0, "NCR5380", instance)) {
-			printk(KERN_WARNING "scsi%d : IRQ%d not free, interrupts disabled\n",

instance->host_no, instance->irq); + ret = NCR5380_test_irq(instance,
instance->irq);

drivers/scsi/g_NCR5380.c +14 -0
+		if (ret) {
+			printk(KERN_WARNING "scsi%d : IRQ%d not %s, interrupts disabled\n",
+			       instance->host_no, instance->irq,
+			       (ret == -EBUSY) ? "free" : "working");
+			instance->irq = NO_IRQ;
+		}
+	}
+
+	if (instance->irq != NO_IRQ) {
+		if (request_irq(instance->irq, generic_NCR5380_intr, 0,
+				"NCR5380", instance)) {
+			printk(KERN_WARNING "scsi%d : IRQ%d not free, interrupts disabled\n",
+			       instance->host_no,
+			       instance->irq);
 			instance->irq = NO_IRQ;
 		}
 	}
Signature · 1 line
Ondrej Zary
#4 lore raw

Re: [PATCH 1/6] NCR5380: Use probe_irq_*() for IRQ probing

In reply to Ondrej Zary
Quoted 11 lines · Ondrej Zary
Quoted 9 lines · Finn Thain

drivers/scsi/g_NCR5380.c line 88

Quoted 5 lines · Ondrej Zary
drivers/scsi/g_NCR5380.c +5 -0
+	NCR5380_write(SELECT_ENABLE_REG, hostdata->id_mask);
+	NCR5380_write(OUTPUT_DATA_REG, hostdata->id_mask);
+	NCR5380_write(INITIATOR_COMMAND_REG, ICR_BASE | ICR_ASSERT_DATA | ICR_ASSERT_SEL);
+ 
+	usleep_range(1000, 20000);

Again, msleep(1) would be more appropriate here.

Quoted 4 lines

WARNING: msleep < 20ms can sleep for up to 20ms; see Documentation/timers/timers-howto.txt #164: FILE: drivers/scsi/g_NCR5380.c:88:
+ msleep(1);

That link is almost 10 years old. I wonder if it is still accurate.

Anyway, I take it that you chose usleep_range() so as to make the checkpatch warning go away. Why not use checkpatch.conf for that, or just ignore it? The upper bound for the delay is unimportant and so the warning isn't applicable.

Since the lower bound is some unknown number of microseconds, I think msleep(1) nicely expresses the two constraints. Whereas, usleep_range(1000, 20000) misrepresents them.

We really don't need three significant figures. If you want precision, surely you would have to measure the time it takes for the IRQ to fire, and derive a worst case (lower bound) from the measurements.

#8 lore raw

Re: [PATCH 2/6] g_NCR5380: Test the IRQ before accepting it

In reply to Ondrej Zary
Quoted 5 lines · Ondrej Zary
Quoted 4 lines · On Wednesday 02 November 2016 08:45:26 Finn Thain wrote:
Quoted 3 lines · Ondrej Zary

Trigger an IRQ first with a test IRQ handler to find out if it really works. Disable the IRQ if not.

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

This prevents hang when incorrect IRQ was specified by user.

Once again, how does it cause a hang?

Kernel scans the bus, finds a HDD, then attempts to read MBR. modprobe process is stuck but the system is still running. Then the transfer probably times out and everything locks up hard, even fbcon cursor stops blinking. I guess that kernel is trying to abort or reset.

I don't think this issue relates to the patch, because the chip irq is not needed for exception handling.

A backtrace from the soft lockup detector should help explain this.

Quoted 1 line

BTW. rescan-scsi-bus also causes hang, anytime, even without IRQ.

I would try "scsi_logging_level -s -a 7" to find out what is going on during the bus scan (for modprobe or rescan-scsi-bus).

The polling loops in generic_NCR5380_pread/pwrite can cause a lockup because they lack timeouts. Better to call NCR5380_poll_politely, as in macscsi_pread/pwrite.

#14 lore raw

Re: [PATCH 4/6] g_NCR5380: Add IRQ auto-configuration for HP C2502

In reply to Ondrej Zary
Quoted 6 lines · Ondrej Zary
Quoted 3 lines

Also, you've ignored the irq module parameters. From the user's point of view, surely the least surprising thing is to attempt to configure the card for whatever irq the user asked for.

I haven't. NCR5380_find_irq is only called when irq is set to IRQ_AUTO.

My mistake.

Quoted 4 lines
Quoted 3 lines

If the specified irq isn't supported by the board, just log an error and fail. If you want to be user friendly, print a message to tell them what irqs the card supports.

Quoted 8 lines

If the IRQ is not supported (or does not work), user gets a warning and the driver continues with IRQ disabled.

Quoted 2 lines

If the user asks for IRQ_AUTO, just configure the board for a hard-coded default, say 9, and print a warning message to say so.

The card is almost Plug&Play. The base address is already configured automatically by the driver so doing the same for IRQ makes sense.

Why don't we see any other drivers doing this?

If the card was really plug and play, I expect we would just call pnp_irq(), as the other PNP drivers do.

Quoted 7 lines
Quoted 7 lines

Either way, if request_irq fails just continue with NO_IRQ, as per usual.

To me that's the most flexible and least surprising behaviour. But again, if someone with more ISA knowledge wishes to weigh in, that's fine too.

#15 lore raw

Re: [PATCH 4/6] g_NCR5380: Add IRQ auto-configuration for HP C2502

Quoted 16 lines · Finn Thain
Quoted 5 lines · Ondrej Zary
Quoted 3 lines

Also, you've ignored the irq module parameters. From the user's point of view, surely the least surprising thing is to attempt to configure the card for whatever irq the user asked for.

I haven't. NCR5380_find_irq is only called when irq is set to IRQ_AUTO.

My mistake.

Quoted 7 lines
Quoted 3 lines

If the specified irq isn't supported by the board, just log an error and fail. If you want to be user friendly, print a message to tell them what irqs the card supports.

If the IRQ is not supported (or does not work), user gets a warning and the driver continues with IRQ disabled.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

If the user asks for IRQ_AUTO, just configure the board for a hard-coded default, say 9, and print a warning message to say so.

The card is almost Plug&Play. The base address is already configured automatically by the driver so doing the same for IRQ makes sense.

Why don't we see any other drivers doing this?

Many ISA sound card drivers do this - there's even a function for this:
static int snd_legacy_find_free_irq(int *irq_table)

Unfortunately, it's defined in ALSA headers and even protected with an #ifdef.

Quoted 2 lines

If the card was really plug and play, I expect we would just call pnp_irq(), as the other PNP drivers do.

The card predates the PnP standard so we can't.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

Either way, if request_irq fails just continue with NO_IRQ, as per usual.

To me that's the most flexible and least surprising behaviour. But again, if someone with more ISA knowledge wishes to weigh in, that's fine too.

Signature · 1 line
Ondrej Zary
#16 lore raw

Re: [PATCH 4/6] g_NCR5380: Add IRQ auto-configuration for HP C2502

Quoted 10 lines · Ondrej Zary
Quoted 6 lines · Finn Thain
Quoted 3 lines · Ondrej Zary

The card is almost Plug&Play. The base address is already configured automatically by the driver so doing the same for IRQ makes sense.

Why don't we see any other drivers doing this?

Many ISA sound card drivers do this - there's even a function for this:
static int snd_legacy_find_free_irq(int *irq_table)

Well, if snd_legacy_find_free_irq() is good enough for the ALSA maintainers, it is good enough for me.

But I still don't like this patch. I would prefer to cut-and-paste snd_legacy_find_free_irq() as g_NCR5380_find_free_irq(), and snd_legacy_empty_irq_handler() as g_NCR5380_empty_irq_handler(), to allow for refactoring all that into a common header file later on.

When the user asks for board == BOARD_HP_C2502 and irq == IRQ_AUTO, just use the first result from g_NCR5380_find_free_irq() to program the card. If no irq is free, program the card for irq 0.

Then just proceed with the usual IRQ_AUTO probing, like you would for any other board. That way, there is no need for NCR5380_test_irq() at all, and you can drop patch 2/5 (at least until we figure out why a command timeout during modprobe hangs your system).

#19 lore raw

Re: [PATCH 5/6] g_NCR5380: Autoprobe IRQ by default

In reply to Finn Thain

Hi Ondrej,

drivers/scsi/g_NCR5380.c static int ncr_53c400; · line 47

Quoted 13 lines · On Wed, 2 Nov 2016, I wrote:

I think this patch is incomplete and you should add these changes:

drivers/scsi/g_NCR5380.c +1 -1
index 7299ad9..0bf0322 100644
@@ -44,7 +44,7 @@ static int ncr_53c400;
 static int ncr_53c400a;
 static int dtc_3181e;
 static int hp_c2502;
-module_param(ncr_irq, int, 0);
+module_param(ncr_irq, int, IRQ_AUTO);

Oops, this doesn't even compile! Sorry about that.

What I was trying to achieve was,

-static int ncr_irq;
+static int ncr_irq = IRQ_AUTO;

I will update my patch queue, compile test it, and ask you to review.