All of lore.kernel.org
 help / color / mirror / Atom feed
From: Volodymyr Babchuk <Volodymyr_Babchuk@epam.com>
To: Oleksandr <olekstysh@gmail.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>,
	Julien Grall <julien@xen.org>,
	Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>,
	Oleksandr Tyshchenko <Oleksandr_Tyshchenko@epam.com>,
	"xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>,
	Volodymyr Babchuk <Volodymyr_Babchuk@epam.com>
Subject: Re: [Xen-devel] [PATCH] iommu/arm: Don't allow the same micro-TLB to be shared between domains
Date: Thu, 30 Jan 2020 15:05:43 +0000	[thread overview]
Message-ID: <8736bxougo.fsf@epam.com> (raw)
In-Reply-To: <b3cf0ae9-d9ea-5a8a-0f2d-36f158aa680b@gmail.com>


Hi Oleksandr,

Oleksandr writes:

[...]

>>> +
>>> +    /*
>>> +     * We need to prevent the use cases where devices which use the same
>>> +     * micro-TLB are assigned to different Xen domains (micro-TLB cannot be
>>> +     * shared between multiple Xen domains, since it points to the context bank
>>> +     * to use for the page walk).
>>> +     * As each Xen domain uses individual context bank pointed by context_id,
>>> +     * we can potentially recognize that use case by comparing current and new
>>> +     * context_id for already enabled micro-TLB and prevent different context
>>> +     * bank from being set.
>>> +     */
>>> +    data = ipmmu_read(mmu, IMUCTR(utlb));
>> I can see that this code is not covered by spinlock. So, I believe,
>> there can be a race comdition, when this register is being read on two
>> CPUs simultaneously.
>
> I don't think, ipmmu_assign(deassign)_device callbacks take a
> spinlock, so the micro-TLB management routines inside
> are protected.
Yeah, you are right. Somehow I missed this when checked the code yesterday.

>
>>     /* Disable MMU translation for the micro-TLB. */
>> @@ -671,7 +698,12 @@ static int ipmmu_attach_device(struct ipmmu_vmsa_domain *domain,
>>           dev_info(dev, "Reusing IPMMU context %u\n", domain->context_id);
>>         for ( i = 0; i < fwspec->num_ids; ++i )
>> -        ipmmu_utlb_enable(domain, fwspec->ids[i]);
>> +    {
>> +        int ret = ipmmu_utlb_enable(domain, fwspec->ids[i]);
>> +
>> +        if ( ret )
>> +            return ret;
>> I can't see error path where ipmmu_utlb_disable() would be called for
>> already enable uTLBs. Is this normal?
>
> Good question. Indeed, we need to restore previous state in case of error.
>
>
> I will add the following:
>
> diff --git a/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> b/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> index c21d2d7..411fc0f 100644
> --- a/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> +++ b/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> @@ -702,7 +702,12 @@ static int ipmmu_attach_device(struct
> ipmmu_vmsa_domain *domain,
>          int ret = ipmmu_utlb_enable(domain, fwspec->ids[i]);
>
>          if ( ret )
> +        {
> +            while ( i-- )
You will call
> +                ipmmu_utlb_disable(domain, fwspec->ids[i]);
for uTLB that caused the error. Likely, this uTLB right now is assigned
for another domain. So, you will disable active uTLB which belongs to
that domain. 

> +
>              return ret;
> +        }
>      }
>
>      return 0;


--
Volodymyr Babchuk at EPAM
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

  reply	other threads:[~2020-01-30 15:06 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-29 14:55 [Xen-devel] [PATCH] iommu/arm: Don't allow the same micro-TLB to be shared between domains Oleksandr Tyshchenko
2020-01-29 16:07 ` Volodymyr Babchuk
2020-01-30 14:34   ` Oleksandr
2020-01-30 15:05     ` Volodymyr Babchuk [this message]
2020-01-30 15:20       ` Volodymyr Babchuk

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=8736bxougo.fsf@epam.com \
    --to=volodymyr_babchuk@epam.com \
    --cc=Oleksandr_Tyshchenko@epam.com \
    --cc=julien@xen.org \
    --cc=olekstysh@gmail.com \
    --cc=sstabellini@kernel.org \
    --cc=xen-devel@lists.xenproject.org \
    --cc=yoshihiro.shimoda.uh@renesas.com \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.