[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <80f362fd-8e69-1dcd-ba7a-abfc55894d61@arm.com>
Date: Tue, 31 May 2022 17:42:36 +0100
From: Robin Murphy <robin.murphy@....com>
To: Jason Gunthorpe <jgg@...dia.com>
Cc: Baolu Lu <baolu.lu@...ux.intel.com>,
Kevin Tian <kevin.tian@...el.com>,
Ashok Raj <ashok.raj@...el.com>, linux-kernel@...r.kernel.org,
Christoph Hellwig <hch@...radead.org>,
iommu@...ts.linux-foundation.org,
Jacob jun Pan <jacob.jun.pan@...el.com>,
Will Deacon <will@...nel.org>
Subject: Re: [PATCH 01/12] iommu/vt-d: Use iommu_get_domain_for_dev() in
debugfs
On 2022-05-31 16:59, Jason Gunthorpe wrote:
> On Tue, May 31, 2022 at 02:52:28PM +0100, Robin Murphy wrote:
>
>>> + break;
>>> + pgtable_walk_level(m, phys_to_virt(phys_addr),
>>
>> Also, obligatory reminder that pfn_valid() only means that pfn_to_page()
>> gets you a valid struct page. Whether that page is direct-mapped kernel
>> memory or not is a different matter.
>
> Even though this is debugfs, if the operation is sketchy like that and
> can theortically crash the kernel the driver should test capabilities,
> CAP_SYS_RAWIO or something may be appropriate. I don't think we have a
> better cap for 'userspace may crash the kernel'
It shouldn't be insurmountable to make this safe, it just needs a bit
more than pfn_valid(), which can still return true off the ends of the
memory map if they're not perfectly section-aligned, and for random
reserved holes in the middle.
Robin.
Powered by blists - more mailing lists