RE: [PATCH] intel-iommu: intel_iommu_domain_alloc should check'dmar_disabled'

[Bhavesh Davda]

At least with 2.6.28, you could trigger the Oops as follows:

1. Boot with intel_iommu=off
2. Call the exported symbol intel_iommu_domain_alloc() from a module which knows nothing about the fact that dmar_disabled is true

Looks like intel_iommu_domain_alloc is not exported any more at git head, and I see a bunch of commits to git head from Joerg Roedel on Jan 3, 2009, that happened sometime between 2.6.28 and 2.6.29-rc1, so I guess we're good.

Thanks

- Bhavesh
 
Bhavesh P. Davda

> -----Original Message-----
> From: Kyle McMartin [mailto:kyle@xxxxxxxxxxxxx] On Behalf Of Kyle
> McMartin
> Sent: Friday, January 30, 2009 8:45 PM
> To: Bhavesh Davda
> Cc: David.Woodhouse@xxxxxxxxx; mingo@xxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH] intel-iommu: intel_iommu_domain_alloc should check
> 'dmar_disabled'
> 
> On Fri, Jan 16, 2009 at 03:54:12PM -0800, Bhavesh Davda wrote:
> >
> > intel_iommu_domain_alloc() will try and iommu_alloc_domain(iommu),
> > which calls alloc_domain_mem() to allocate out of
> 'iommu_domain_cache'.
> > Unfortunately iommu_domain_cache will not be created if booted with
> > "intel_iommu=off", i.e. dmar_disabled = 1.
> >
> 
> Hi Bhavesh,
> 
> What's the callchain that can result in intel_iommu_domain_alloc
> being called? It looks like most possible callsites are checking
> intel_iommu_found, which will be 0 since we check dmar_disabled before
> calling init_dmars to enumerate the iommus.
> 
> (I'm looking at 2.6.28.y right now, is there an interim between when
> you
>  stumbled upon this two weeks ago and git head where it's no longer an
>  issue? I guess it probably got missed due to folks travelling for
> LCA.)
> 
> regards, Kyle
--
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/