linux-kernel - Re: [PATCH] iommu/dma: limit the IOVA allocated to dma-ranges region

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: <3aeeb13e-045f-4cde-2cfb-62d4a879d207@arm.com>
Date:   Fri, 1 Sep 2017 10:43:05 +0100
From:   Robin Murphy <robin.murphy@....com>
To:     Joerg Roedel <joro@...tes.org>, Krishna Reddy <vdumpa@...dia.com>
Cc:     iommu@...ts.linux-foundation.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] iommu/dma: limit the IOVA allocated to dma-ranges region

On 01/09/17 10:26, Joerg Roedel wrote:
> Adding Robin for review.
> 
> On Thu, Aug 31, 2017 at 03:08:21PM -0700, Krishna Reddy wrote:
>> Limit the IOVA allocated to dma-ranges specified for the device.
>> This is necessary to ensure that IOVA allocated is addressable
>> by device.

Why? IOVA allocation is already constrained as much as it should be - if
the device's DMA mask is wrong that's another problem, and this isn't
the right place to fix it.

dma_32bit_pfn means nothing more than an internal detail of IOVA
allocator caching, which is subject to change[1]. As-is, on some
platforms this patch will effectively force all allocations to fail already.

Robin.

[1]:https://www.mail-archive.com/iommu@lists.linux-foundation.org/msg19694.html
>> Signed-off-by: Krishna Reddy <vdumpa@...dia.com>
>> ---
>>  drivers/iommu/dma-iommu.c | 5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
>> index 9d1cebe7f6cb..e8a8320b571b 100644
>> --- a/drivers/iommu/dma-iommu.c
>> +++ b/drivers/iommu/dma-iommu.c
>> @@ -364,6 +364,7 @@ static dma_addr_t iommu_dma_alloc_iova(struct iommu_domain *domain,
>>  	struct iommu_dma_cookie *cookie = domain->iova_cookie;
>>  	struct iova_domain *iovad = &cookie->iovad;
>>  	unsigned long shift, iova_len, iova = 0;
>> +	dma_addr_t dma_end_addr;
>>  
>>  	if (cookie->type == IOMMU_DMA_MSI_COOKIE) {
>>  		cookie->msi_iova += size;
>> @@ -381,6 +382,10 @@ static dma_addr_t iommu_dma_alloc_iova(struct iommu_domain *domain,
>>  	if (iova_len < (1 << (IOVA_RANGE_CACHE_MAX_SIZE - 1)))
>>  		iova_len = roundup_pow_of_two(iova_len);
>>  
>> +	/* Limit IOVA allocated to device addressable dma-ranges region. */
>> +	dma_end_addr = (dma_addr_t)iovad->dma_32bit_pfn << shift;
>> +	dma_limit = dma_limit > dma_end_addr ? dma_end_addr : dma_limit;
> 
> This looks like a good use-case for min().
> 
>> +
>>  	if (domain->geometry.force_aperture)
>>  		dma_limit = min(dma_limit, domain->geometry.aperture_end);
>>  
>> -- 
>> 2.1.4