linux-kernel - RE: [PATCH] Intel IOMMU Pass Through Support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 20 Apr 2009 10:27:24 -0700
From:	"Yu, Fenghua" <fenghua.yu@...el.com>
To:	'David Woodhouse' <dwmw2@...radead.org>
CC:	'Ingo Molnar' <mingo@...e.hu>,
	'Linus Torvalds' <torvalds@...ux-foundation.org>,
	"Han, Weidong" <weidong.han@...el.com>,
	'LKML' <linux-kernel@...r.kernel.org>,
	'IOMMU' <iommu@...ts.linux-foundation.org>
Subject: RE: [PATCH] Intel IOMMU Pass Through Support

>On Wed, 2009-04-15 at 17:19 -0700, Fenghua Yu wrote:
>> +       if (iommu_pass_through)
>> +               if (!pass_through) {
>> +                       printk(KERN_INFO
>> +                               "Pass Through is not supported by
>hardware.\n");
>> +                       iommu_pass_through = 0;
>> +               }
>
>So if we ask for pass-through and the hardware doesn't support it, we
>end up in full-translation mode? Wouldn't it be better to set up 1:1
>mappings for each device instead?
>
>(We probably want to fix up the code which sets up 1:1 mappings to
>actually share page tables where appropriate, and to use superpages).
>
>Also, did we already have a discussion about whether this should be an
>intel_iommu= parameter, or a generic iommu= one?

For this patch, I would just fall back to "normal" translation if without hardware pass-through support.

I'm doing 1:1 mapping now to speed up mapping and unmapping. When the 1:1 mapping is implemented, the fall back from pass-through will go to 1:1 mapping which will be "normal" translation.

BTW, I think most of (or newer) VT-d hardware supports pass through feature now. So it won't fail anyway if pass through is wanted.

So at this point, this should be fine.

>
>>                 ret = domain_context_mapping_one(domain,
>> -                                                pci_domain_nr(parent-
>>bus),
>> -                                                parent->bus->number,
>> -                                                parent->devfn);
>> +                                               pci_domain_nr(parent-
>>bus),
>> +                                               parent->bus->number,
>> +                                               parent->devfn,
>translation);
>
>Whitespace damage here -- you're modifying continuation lines which used
>to be right underneath the 'domain' parameter, and you're making them
>not line up properly.
>
>You have the same problem elsewhere for newly-added code too, in fact:
>
>> +               ret = domain_context_mapping(domain, pdev,
>> +                                               CONTEXT_TT_PASS_THROUGH);
>

OK. I'll change this to align to the upper parameter.

Thanks.

-Fenghuas
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/