Re: [PATCH V8] ACPI, PCI, irq: support IRQ numbers greater than 256

[Bjorn Helgaas]

On Thu, Dec 03, 2015 at 01:58:55PM -0500, Sinan Kaya wrote:
> The ACPI compiler uses the extended format when used interrupt numbers
> are greater than 15. The extended IRQ is 32 bits according to the ACPI
> spec. The code supports parsing the extended interrupt numbers. However,
> due to used data structure type; the code silently truncates interrupt
> numbers greater than 256.
> 
> First, this patch changes the interrupt number type to 32 bits. Next, the
> penalty array has been limited to 16 for ISA IRQs. Finally, a new penalty
> linklist has been added for all other interrupts greater than 16. If an IRQ
> is not found in the link list, an IRQ info structure will be dynamically
> allocated on the first access and will be placed on the list for further
> reuse. The list will grow by the number of supported interrupts in the
> ACPI table rather than having a 256 hard limitation.

Can you split this into two patches?  One to replace the penalty
storage scheme, and a second to change the interrupt number types
from u8 to u32?

Generally looks good to me.  Tracking all the penalty information
still seems clunky, but I don't have any great ideas of better ways.
I have a few minor comments below; when you address them, you can add
my:

Acked-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>

> Signed-off-by: Sinan Kaya <okaya@xxxxxxxxxxxxxx>
> ---
>  drivers/acpi/pci_link.c | 134 +++++++++++++++++++++++++++++++++++++-----------
>  1 file changed, 104 insertions(+), 30 deletions(-)
> 
> diff --git a/drivers/acpi/pci_link.c b/drivers/acpi/pci_link.c
> index 7c8408b..e10661f 100644
> --- a/drivers/acpi/pci_link.c
> +++ b/drivers/acpi/pci_link.c
> @@ -4,6 +4,7 @@
>   *  Copyright (C) 2001, 2002 Andy Grover <andrew.grover@xxxxxxxxx>
>   *  Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@xxxxxxxxx>
>   *  Copyright (C) 2002       Dominik Brodowski <devel@xxxxxxxx>
> + *  Copyright (c) 2015, The Linux Foundation. All rights reserved.
>   *
>   * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>   *
> @@ -67,12 +68,12 @@ static struct acpi_scan_handler pci_link_handler = {
>   * later even the link is disable. Instead, we just repick the active irq
>   */
>  struct acpi_pci_link_irq {
> -	u8 active;		/* Current IRQ */
> +	u32 active;		/* Current IRQ */
>  	u8 triggering;		/* All IRQs */
>  	u8 polarity;		/* All IRQs */
>  	u8 resource_type;
>  	u8 possible_count;
> -	u8 possible[ACPI_PCI_LINK_MAX_POSSIBLE];
> +	u32 possible[ACPI_PCI_LINK_MAX_POSSIBLE];
>  	u8 initialized:1;
>  	u8 reserved:7;
>  };
> @@ -437,8 +438,7 @@ static int acpi_pci_link_set(struct acpi_pci_link *link, int irq)
>   * enabled system.
>   */
>  
> -#define ACPI_MAX_IRQS		256
> -#define ACPI_MAX_ISA_IRQ	16
> + #define ACPI_MAX_ISA_IRQ	16

Extra leading space here.

>  #define PIRQ_PENALTY_PCI_AVAILABLE	(0)
>  #define PIRQ_PENALTY_PCI_POSSIBLE	(16*16)
> @@ -447,7 +447,7 @@ static int acpi_pci_link_set(struct acpi_pci_link *link, int irq)
>  #define PIRQ_PENALTY_ISA_USED		(16*16*16*16*16)
>  #define PIRQ_PENALTY_ISA_ALWAYS		(16*16*16*16*16*16)
>  
> -static int acpi_irq_penalty[ACPI_MAX_IRQS] = {
> +static int acpi_irq_isa_penalty[ACPI_MAX_ISA_IRQ] = {
>  	PIRQ_PENALTY_ISA_ALWAYS,	/* IRQ0 timer */
>  	PIRQ_PENALTY_ISA_ALWAYS,	/* IRQ1 keyboard */
>  	PIRQ_PENALTY_ISA_ALWAYS,	/* IRQ2 cascade */
> @@ -464,9 +464,61 @@ static int acpi_irq_penalty[ACPI_MAX_IRQS] = {
>  	PIRQ_PENALTY_ISA_USED,		/* IRQ13 fpe, sometimes */
>  	PIRQ_PENALTY_ISA_USED,		/* IRQ14 ide0 */
>  	PIRQ_PENALTY_ISA_USED,		/* IRQ15 ide1 */
> -	/* >IRQ15 */
>  };
>  
> +struct irq_penalty_info {
> +	unsigned int irq;
> +	int penalty;
> +	struct list_head node;
> +};
> +
> +LIST_HEAD(acpi_irq_penalty_list);

Should be static.

> +static int acpi_irq_get_penalty(int irq)
> +{
> +	struct irq_penalty_info *irq_info;
> +
> +	if (irq < ACPI_MAX_ISA_IRQ)
> +		return acpi_irq_isa_penalty[irq];
> +
> +	list_for_each_entry(irq_info, &acpi_irq_penalty_list, node) {
> +		if (irq_info->irq == irq)
> +			return irq_info->penalty;
> +	}
> +
> +	return 0;
> +}
> +
> +static int acpi_irq_set_penalty(int irq, unsigned int new_penalty)

"int new_penalty" to match irq_info->penalty and acpi_irq_get_penalty()
return type.

> +{
> +	struct irq_penalty_info *irq_info;
> +
> +	/* see if this is a ISA IRQ */
> +	if (irq < ACPI_MAX_ISA_IRQ) {
> +		acpi_irq_isa_penalty[irq] = new_penalty;
> +		return 0;
> +	}
> +
> +	/* next, try to locate from the dynamic list */
> +	list_for_each_entry(irq_info, &acpi_irq_penalty_list, node) {
> +		if (irq_info->irq == irq) {
> +			irq_info->penalty  = new_penalty;
> +			return 0;
> +		}
> +	}
> +
> +	/* nope, let's allocate a slot for this IRQ */
> +	irq_info = kzalloc(sizeof(*irq_info), GFP_KERNEL);
> +	if (!irq_info)
> +		return -ENOMEM;
> +
> +	irq_info->irq = irq;
> +	irq_info->penalty = new_penalty;
> +	list_add_tail(&irq_info->node, &acpi_irq_penalty_list);
> +
> +	return 0;
> +}

An "acpi_irq_add_penalty(int irq, int penalty)" here would simplify
most of the calls below:

  static void acpi_irq_add_penalty(int irq, int penalty)
  {
    int current = acpi_irq_get_penalty(irq);

    acpi_irq_set_penalty(irq, current + penalty);
  }

> +
>  int __init acpi_irq_penalty_init(void)
>  {
>  	struct acpi_pci_link *link;
> @@ -487,15 +539,22 @@ int __init acpi_irq_penalty_init(void)
>  			    link->irq.possible_count;
>  
>  			for (i = 0; i < link->irq.possible_count; i++) {
> -				if (link->irq.possible[i] < ACPI_MAX_ISA_IRQ)
> -					acpi_irq_penalty[link->irq.
> -							 possible[i]] +=
> -					    penalty;
> +				if (link->irq.possible[i] < ACPI_MAX_ISA_IRQ) {
> +					int irqpos = link->irq.possible[i];
> +					int curpen;
> +
> +					curpen = acpi_irq_get_penalty(irqpos);
> +					curpen += penalty;
> +					acpi_irq_set_penalty(irqpos, curpen);

	acpi_irq_add_penalty(link->irq.possible[i], penalty);

> +				}
>  			}
>  
>  		} else if (link->irq.active) {

You didn't change this, but the "else" here looks wrong to me: if we
got any IRQs from _PRS, we never add PIRQ_PENALTY_PCI_POSSIBLE to the
active IRQ.

It also seems wrong that we loop through everything on acpi_link_list.
It would be better if we could do this for each link as it is
enumerated in acpi_pci_link_add(), so any hot-added links would be
handled the same way.

These are both pre-existing issues/questions, so I don't think you're
obligated to address them.

> -			acpi_irq_penalty[link->irq.active] +=
> -			    PIRQ_PENALTY_PCI_POSSIBLE;
> +			int curpen;
> +
> +			curpen = acpi_irq_get_penalty(link->irq.active);
> +			curpen += PIRQ_PENALTY_PCI_POSSIBLE;
> +			acpi_irq_set_penalty(link->irq.active, curpen);
>  		}
>  	}
>  
> @@ -547,12 +606,12 @@ static int acpi_pci_link_allocate(struct acpi_pci_link *link)
>  		 * the use of IRQs 9, 10, 11, and >15.
>  		 */
>  		for (i = (link->irq.possible_count - 1); i >= 0; i--) {
> -			if (acpi_irq_penalty[irq] >
> -			    acpi_irq_penalty[link->irq.possible[i]])
> +			if (acpi_irq_get_penalty(irq) >
> +			    acpi_irq_get_penalty(link->irq.possible[i]))
>  				irq = link->irq.possible[i];
>  		}
>  	}
> -	if (acpi_irq_penalty[irq] >= PIRQ_PENALTY_ISA_ALWAYS) {
> +	if (acpi_irq_get_penalty(irq) >= PIRQ_PENALTY_ISA_ALWAYS) {
>  		printk(KERN_ERR PREFIX "No IRQ available for %s [%s]. "
>  			    "Try pci=noacpi or acpi=off\n",
>  			    acpi_device_name(link->device),
> @@ -568,7 +627,12 @@ static int acpi_pci_link_allocate(struct acpi_pci_link *link)
>  			    acpi_device_bid(link->device));
>  		return -ENODEV;
>  	} else {
> -		acpi_irq_penalty[link->irq.active] += PIRQ_PENALTY_PCI_USING;
> +		int curpen;
> +
> +		curpen = acpi_irq_get_penalty(link->irq.active);
> +		curpen += PIRQ_PENALTY_PCI_USING;
> +		acpi_irq_set_penalty(link->irq.active, curpen);
> +
>  		printk(KERN_WARNING PREFIX "%s [%s] enabled at IRQ %d\n",
>  		       acpi_device_name(link->device),
>  		       acpi_device_bid(link->device), link->irq.active);
> @@ -778,7 +842,7 @@ static void acpi_pci_link_remove(struct acpi_device *device)
>  }
>  
>  /*
> - * modify acpi_irq_penalty[] from cmdline
> + * modify penalty from cmdline
>   */
>  static int __init acpi_irq_penalty_update(char *str, int used)
>  {
> @@ -796,13 +860,15 @@ static int __init acpi_irq_penalty_update(char *str, int used)
>  		if (irq < 0)
>  			continue;
>  
> -		if (irq >= ARRAY_SIZE(acpi_irq_penalty))
> -			continue;
> +		if (used) {
> +			int curpen;
>  
> -		if (used)
> -			acpi_irq_penalty[irq] += PIRQ_PENALTY_ISA_USED;
> +			curpen = acpi_irq_get_penalty(irq);
> +			curpen += PIRQ_PENALTY_ISA_USED;
> +			acpi_irq_set_penalty(irq, curpen);
> +		}
>  		else
> -			acpi_irq_penalty[irq] = PIRQ_PENALTY_PCI_AVAILABLE;
> +			acpi_irq_set_penalty(irq, PIRQ_PENALTY_PCI_AVAILABLE);
>  
>  		if (retval != 2)	/* no next number */
>  			break;
> @@ -819,18 +885,22 @@ static int __init acpi_irq_penalty_update(char *str, int used)
>   */
>  void acpi_penalize_isa_irq(int irq, int active)
>  {
> -	if (irq >= 0 && irq < ARRAY_SIZE(acpi_irq_penalty)) {
> +	if (irq >= 0) {

I would structure this as:

  if (irq < 0)
    return;

  if (active)
    acpi_irq_add_penalty(irq, PIRQ_PENALTY_ISA_USED);
  else
    acpi_irq_add_penalty(irq, PIRQ_PENALTY_PCI_USING);

But that might be just my personal preference.  Similarly in
acpi_penalize_sci_irq() below.

> +		int curpen;
> +
> +		curpen = acpi_irq_get_penalty(irq);
>  		if (active)
> -			acpi_irq_penalty[irq] += PIRQ_PENALTY_ISA_USED;
> +			curpen += PIRQ_PENALTY_ISA_USED;
>  		else
> -			acpi_irq_penalty[irq] += PIRQ_PENALTY_PCI_USING;
> +			curpen += PIRQ_PENALTY_PCI_USING;
> +		acpi_irq_set_penalty(irq, curpen);
>  	}
>  }
>  
>  bool acpi_isa_irq_available(int irq)
>  {
> -	return irq >= 0 && (irq >= ARRAY_SIZE(acpi_irq_penalty) ||
> -			    acpi_irq_penalty[irq] < PIRQ_PENALTY_ISA_ALWAYS);
> +	return irq >= 0 &&
> +		(acpi_irq_get_penalty(irq) < PIRQ_PENALTY_ISA_ALWAYS);
>  }
>  
>  /*
> @@ -840,12 +910,16 @@ bool acpi_isa_irq_available(int irq)
>   */
>  void acpi_penalize_sci_irq(int irq, int trigger, int polarity)
>  {
> -	if (irq >= 0 && irq < ARRAY_SIZE(acpi_irq_penalty)) {
> +	if (irq >= 0) {
> +		int curpen;
> +
> +		curpen = acpi_irq_get_penalty(irq);
>  		if (trigger != ACPI_MADT_TRIGGER_LEVEL ||
>  		    polarity != ACPI_MADT_POLARITY_ACTIVE_LOW)
> -			acpi_irq_penalty[irq] += PIRQ_PENALTY_ISA_ALWAYS;
> +			curpen += PIRQ_PENALTY_ISA_ALWAYS;
>  		else
> -			acpi_irq_penalty[irq] += PIRQ_PENALTY_PCI_USING;
> +			curpen += PIRQ_PENALTY_PCI_USING;
> +		acpi_irq_set_penalty(irq, curpen);
>  	}
>  }
>  
> -- 
> Qualcomm Technologies, Inc. on behalf of Qualcomm Innovation Center, Inc.
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/