linux-kernel - Re: [RFC PATCH 1/3] iommu: Add support for out of band flushing

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]

Message-ID: <20150929093257.GE9460@ulmo.nvidia.com>
Date:	Tue, 29 Sep 2015 11:32:57 +0200
From:	Thierry Reding <thierry.reding@...il.com>
To:	Tomasz Figa <tfiga@...omium.org>
Cc:	iommu@...ts.linux-foundation.org, Joerg Roedel <joro@...tes.org>,
	Hiroshi Doyu <hdoyu@...dia.com>,
	Stephen Warren <swarren@...dotorg.org>,
	Alexandre Courbot <gnurou@...il.com>,
	Vince Hsu <vince.h@...dia.com>,
	Russell King <rmk+kernel@....linux.org.uk>,
	Paul Walmsley <paul@...an.com>,
	Mikko Perttunen <mperttunen@...dia.com>,
	Tomeu Vizoso <tomeu.vizoso@...labora.com>,
	Will Deacon <will.deacon@....com>,
	Alex Williamson <alex.williamson@...hat.com>,
	Marek Szyprowski <m.szyprowski@...sung.com>,
	Arnd Bergmann <arnd@...db.de>,
	Antonios Motakis <a.motakis@...tualopensystems.com>,
	Olav Haugan <ohaugan@...eaurora.org>,
	Nicolas Iooss <nicolas.iooss_linux@....org>,
	linux-kernel@...r.kernel.org, linux-tegra@...r.kernel.org,
	Rob Clark <robdclark@...il.com>
Subject: Re: [RFC PATCH 1/3] iommu: Add support for out of band flushing

On Tue, Sep 29, 2015 at 02:25:24PM +0900, Tomasz Figa wrote:
> This patch adds a new "flush" callback to iommu_ops, which is supposed
> to perform any necessary flushes within given IOMMU domain to make any
> changes to mappings of given area [iova; iova + size) be reflected to
> IOMMU clients.
> 
> The purpose is to let IOMMU drivers skip page-by-page flushes and
> replace it with one flush of full address range on devices which support
> it.
> 
> Signed-off-by: Tomasz Figa <tfiga@...omium.org>
> ---
>  drivers/iommu/iommu.c | 33 ++++++++++++++++++++++++++++++---
>  include/linux/iommu.h |  2 ++
>  2 files changed, 32 insertions(+), 3 deletions(-)

I seem to remember that Rob Clark had proposed something like this back
before it was decided to introduce the ->map_sg() callback instead. I
can't find a reference to the discussion, so perhaps I'm misremembering
but adding Rob Clark in case he has any recollection of why the outcome
was what it was.

> diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
[...]
> @@ -1389,6 +1410,7 @@ size_t iommu_unmap(struct iommu_domain *domain, unsigned long iova, size_t size)
>  		unmapped += unmapped_page;
>  	}
>  
> +	iommu_flush(domain, orig_iova, unmapped);
>  	trace_unmap(orig_iova, size, unmapped);
>  	return unmapped;
>  }
> @@ -1419,19 +1441,24 @@ size_t default_iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
>  		if (!IS_ALIGNED(s->offset, min_pagesz))
>  			goto out_err;
>  
> -		ret = iommu_map(domain, iova + mapped, phys, s->length, prot);
> +		ret = __iommu_map(domain, iova + mapped, phys, s->length, prot);
>  		if (ret)
>  			goto out_err;
>  
>  		mapped += s->length;
>  	}
>  
> +	iommu_flush(domain, iova, mapped);
> +
>  	return mapped;
>  
>  out_err:
>  	/* undo mappings already done */
>  	iommu_unmap(domain, iova, mapped);
>  
> +	/* flush in case part of our mapping already got cached */
> +	iommu_flush(domain, iova, mapped);
> +
>  	return 0;
>  
>  }

iommu_unmap() already does an iommu_flush(), so why flush again after
iommu_unmap()?

Thierry

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)