Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1740671 > unrolled thread

[PATCH] irq/generic-chip: Fix memory leak of domain->name

Back to article view | Back to linux.kernel

Contents

  [PATCH] irq/generic-chip: Fix memory leak of domain->name Jeffy Chen <jeffy.chen@rock-chips.com> - 2017-09-27 14:30 +0200
    Re: [PATCH] irq/generic-chip: Fix memory leak of domain->name Thomas Gleixner <tglx@linutronix.de> - 2017-09-27 15:40 +0200
      Re: [PATCH] irq/generic-chip: Fix memory leak of domain->name jeffy <jeffy.chen@rock-chips.com> - 2017-09-28 05:20 +0200
        Re: [PATCH] irq/generic-chip: Fix memory leak of domain->name jeffy <jeffy.chen@rock-chips.com> - 2017-09-28 06:40 +0200

#1740671 — [PATCH] irq/generic-chip: Fix memory leak of domain->name

Free domain->name when IRQ_DOMAIN_NAME_ALLOCATED been set.

Fixes: d59f6617eef0 ("genirq: Allow fwnode to carry name information only")
Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
---

 kernel/irq/generic-chip.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/irq/generic-chip.c b/kernel/irq/generic-chip.c
index f7086b78ad6e..174dae8ee7fe 100644
--- a/kernel/irq/generic-chip.c
+++ b/kernel/irq/generic-chip.c
@@ -322,6 +322,10 @@ int __irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip,
 		/* Calc pointer to the next generic chip */
 		tmp += sizeof(*gc) + num_ct * sizeof(struct irq_chip_type);
 	}
+	if (d->flags & IRQ_DOMAIN_NAME_ALLOCATED) {
+		kfree(d->name);
+		d->flags &= ~IRQ_DOMAIN_NAME_ALLOCATED;
+	}
 	d->name = name;
 	return 0;
 }
-- 
2.11.0

[toc] | [next] | [standalone]

#1740729

On Wed, 27 Sep 2017, Jeffy Chen wrote:

> Free domain->name when IRQ_DOMAIN_NAME_ALLOCATED been set.

I can see that from the patch, but you fail to explain what the problem
is.

It's actually more than just a memory leak. If the domain gets destroyed
then the domain free code would try to free d->name which might be a string
constant .....

> +	if (d->flags & IRQ_DOMAIN_NAME_ALLOCATED) {
> +		kfree(d->name);
> +		d->flags &= ~IRQ_DOMAIN_NAME_ALLOCATED;
> +	}
>  	d->name = name;

I don't think that this is the proper thing to do. There is no reason why
the domain should have the same name as the irq chip. So we rather should
do:

	if (!d->name)
		d->name = name;

Along with a proper comment.

Thanks,

	tglx

[toc] | [prev] | [next] | [standalone]

#1741147

Hi Thomas,

Thanks for your review.

On 09/27/2017 09:29 PM, Thomas Gleixner wrote:
> On Wed, 27 Sep 2017, Jeffy Chen wrote:
>
>> Free domain->name when IRQ_DOMAIN_NAME_ALLOCATED been set.
>
> I can see that from the patch, but you fail to explain what the problem
> is.
>
> It's actually more than just a memory leak. If the domain gets destroyed
> then the domain free code would try to free d->name which might be a string
> constant .....
ok, i'll add that to commit msg.
>
>> +	if (d->flags & IRQ_DOMAIN_NAME_ALLOCATED) {
>> +		kfree(d->name);
>> +		d->flags &= ~IRQ_DOMAIN_NAME_ALLOCATED;
>> +	}
>>   	d->name = name;
>
> I don't think that this is the proper thing to do. There is no reason why
> the domain should have the same name as the irq chip. So we rather should
> do:
>
> 	if (!d->name)
> 		d->name = name;
>
> Along with a proper comment.
that is better, will do it in next version :)
>
> Thanks,
>
> 	tglx
>
>
>
>

[toc] | [prev] | [next] | [standalone]

#1741177

Hi Thomas,

On 09/28/2017 11:18 AM, jeffy wrote:
>>
>> I don't think that this is the proper thing to do. There is no reason why
>> the domain should have the same name as the irq chip. So we rather should
>> do:
>>
>>     if (!d->name)
>>         d->name = name;
>>
>> Along with a proper comment.
> that is better, will do it in next version :)

it looks like the __irq_domain_add() would guarantee every domain has a 
valid name, maybe we can just remove this?

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web