xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
From: Oleksandr Tyshchenko <oleksandr.tyshchenko@globallogic.com>
To: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: xen-devel@lists.xen.org
Subject: Re: [PATCH v1] xen/arm: Fix deadlock in gic_set_guest_irq()
Date: Tue, 4 Feb 2014 17:36:38 +0200	[thread overview]
Message-ID: <CAJEb2DFY9m7DPzZD5Kzt-EMqwsEKP0X__7K_hAYP0Aep5M8keA@mail.gmail.com> (raw)
In-Reply-To: <alpine.DEB.2.02.1402041519570.4373@kaball.uk.xensource.com>

Thanks.
Due to this thread "xen/arm: maintenance_interrupt SMP fix", I found
out a lot of interesting things.

On Tue, Feb 4, 2014 at 5:21 PM, Stefano Stabellini
<stefano.stabellini@eu.citrix.com> wrote:
> On Mon, 3 Feb 2014, Oleksandr Tyshchenko wrote:
>> The possible deadlock scenario is explained below:
>>
>> non interrupt context:    interrupt contex        interrupt context
>>                           (CPU0):                 (CPU1):
>> vgic_distr_mmio_write()   do_trap_irq()           do_softirq()
>>   |                         |                       |
>>   vgic_disable_irqs()       ...                     ...
>>     |                         |                       |
>>     gic_remove_from_queues()  vgic_vcpu_inject_irq()  vgic_vcpu_inject_irq()
>>     |  ...                      |                       |
>>     |  spin_lock(...)           gic_set_guest_irq()     gic_set_guest_irq()
>>     |  ...                        ...                     ...
>>     |  ... <----------------.---- spin_lock_irqsave(...)  ...
>>     |  ... <----------------.-.---------------------------spin_lock_irqsave(...)
>>     |  ...                  . .       Oops! The lock has already taken.
>>     |  spin_unlock(...)     . .
>>     |  ...                  . .
>>     gic_irq_disable()       . .
>>        ...                  . .
>>        spin_lock(...)       . .
>>        ...                  . .
>>        ... <----------------. .
>>        ... <------------------.
>>        ...
>>        spin_unlock(...)
>>
>> Since the gic_remove_from_queues() and gic_irq_disable() called from
>> non interrupt context and they acquire the same lock as gic_set_guest_irq()
>> which called from interrupt context we must disable interrupts in these
>> functions to avoid possible deadlocks.
>>
>> Change-Id: Ia354d87bb44418956e30cd7e49cc76616c359cc9
>> Signed-off-by: Oleksandr Tyshchenko <oleksandr.tyshchenko@globallogic.com>
>
> nice work Oleksandr!
>
> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
>
>
>>  xen/arch/arm/gic.c |   10 ++++++----
>>  1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/xen/arch/arm/gic.c b/xen/arch/arm/gic.c
>> index c44a4d0..7d83b0c 100644
>> --- a/xen/arch/arm/gic.c
>> +++ b/xen/arch/arm/gic.c
>> @@ -147,14 +147,15 @@ static void gic_irq_enable(struct irq_desc *desc)
>>  static void gic_irq_disable(struct irq_desc *desc)
>>  {
>>      int irq = desc->irq;
>> +    unsigned long flags;
>>
>> -    spin_lock(&desc->lock);
>> +    spin_lock_irqsave(&desc->lock, flags);
>>      spin_lock(&gic.lock);
>>      /* Disable routing */
>>      GICD[GICD_ICENABLER + irq / 32] = (1u << (irq % 32));
>>      desc->status |= IRQ_DISABLED;
>>      spin_unlock(&gic.lock);
>> -    spin_unlock(&desc->lock);
>> +    spin_unlock_irqrestore(&desc->lock, flags);
>>  }
>>
>>  static unsigned int gic_irq_startup(struct irq_desc *desc)
>> @@ -658,11 +659,12 @@ static inline void gic_add_to_lr_pending(struct vcpu *v, unsigned int irq,
>>  void gic_remove_from_queues(struct vcpu *v, unsigned int virtual_irq)
>>  {
>>      struct pending_irq *p = irq_to_pending(v, virtual_irq);
>> +    unsigned long flags;
>>
>> -    spin_lock(&gic.lock);
>> +    spin_lock_irqsave(&gic.lock, flags);
>>      if ( !list_empty(&p->lr_queue) )
>>          list_del_init(&p->lr_queue);
>> -    spin_unlock(&gic.lock);
>> +    spin_unlock_irqrestore(&gic.lock, flags);
>>  }
>>
>>  void gic_set_guest_irq(struct vcpu *v, unsigned int virtual_irq,
>> --
>> 1.7.9.5
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xen.org
>> http://lists.xen.org/xen-devel
>>



-- 

Name | Title
GlobalLogic
P +x.xxx.xxx.xxxx  M +x.xxx.xxx.xxxx  S skype
www.globallogic.com

http://www.globallogic.com/email_disclaimer.txt

  reply	other threads:[~2014-02-04 15:36 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-02-03 17:33 [PATCH v1] xen/arm: Fix deadlock in gic_set_guest_irq() Oleksandr Tyshchenko
2014-02-03 17:51 ` Julien Grall
2014-02-03 17:59   ` Oleksandr Tyshchenko
2014-02-04 15:52   ` Julien Grall
2014-02-05 14:38     ` George Dunlap
2014-02-04 15:21 ` Stefano Stabellini
2014-02-04 15:36   ` Oleksandr Tyshchenko [this message]
2014-02-06 13:07 ` Ian Campbell

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAJEb2DFY9m7DPzZD5Kzt-EMqwsEKP0X__7K_hAYP0Aep5M8keA@mail.gmail.com \
    --to=oleksandr.tyshchenko@globallogic.com \
    --cc=stefano.stabellini@eu.citrix.com \
    --cc=xen-devel@lists.xen.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).