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] [day] [month] [year] [list]
Message-ID: <3562ac98-544b-ea83-423f-0a91ff471bc4@arm.com>
Date:   Mon, 16 Oct 2017 12:14:46 +0100
From:   Robin Murphy <robin.murphy@....com>
To:     Marian Mihailescu <marian.mihailescu@...laide.edu.au>
Cc:     Vladimir Murzin <vladimir.murzin@....com>,
        Randy Dunlap <rdunlap@...radead.org>,
        linux-kernel@...r.kernel.org,
        Bart Van Assche <bart.vanassche@...disk.com>,
        iommu@...ts.linux-foundation.org, Christoph Hellwig <hch@....de>
Subject: Re: CONFIG_DMA_NOOP_OPS breaks ARM arch

On 16/10/17 09:12, Vladimir Murzin wrote:
> + Robin and Christoph
> 
> On 16/10/17 06:27, Marian Mihailescu wrote:
>> I am using 4.14-rc4 with a patch on top that includes
>> arch/arm/include/asm/dma-mapping.h in a module.
>>
>> I have MMU enabled, so
>> select DMA_NOOP_OPS if !MMU
>> does nothing for me, and I get a compile error because dma_noop_ops is unknown.
> 
> Can you post an error message here, please?
> 
>>
>> Maybe I should include linux/dma-mapping.h?

That is the *only* dma-mapping header you should be referring to.
Including the arch/*/asm/ version directly is just broken and asking for
trouble.

Robin.

> Where to include? In your driver or what?
> 
>>
>> Thanks for the quick reply.
> 
> with CONFIG_MMU compiler should optimise out dma_noop_ops in:
> 
> return IS_ENABLED(CONFIG_MMU) ? &arm_dma_ops : &dma_noop_ops;
> 
> What toolchain are you using?
> 
> Cheers
> Vladimir
> 
>>
>>
>> On Mon, Oct 16, 2017 at 2:28 PM, Randy Dunlap <rdunlap@...radead.org> wrote:
>>> On 10/15/17 20:29, Randy Dunlap wrote:
>>>> On 10/15/17 20:27, Randy Dunlap wrote:
>>>>> On 10/15/17 19:27, Marian Mihailescu wrote:
>>>>>> After commit 7844572c633964c864d9f32dc3f2a8ffe5d70371, dma_noop_ops
>>>>>> are built only for architectures that use it.
>>>>>>
>>>>>> For ARM architecture, CONFIG_DMA_NOOP_OPS is not selected, and cannot
>>>>>> be selected.
>>>
>>> What kernel version are you looking at?
>>> I see that it is selected:
>>>
>>> --- a/arch/arm/Kconfig
>>> +++ b/arch/arm/Kconfig
>>> @@ -22,6 +22,7 @@ config ARM
>>>         select CLONE_BACKWARDS
>>>         select CPU_PM if (SUSPEND || CPU_IDLE)
>>>         select DCACHE_WORD_ACCESS if HAVE_EFFICIENT_UNALIGNED_ACCESS
>>> +       select DMA_NOOP_OPS if !MMU
>>>         select EDAC_SUPPORT
>>>         select EDAC_ATOMIC_SCRUB
>>>         select GENERIC_ALLOCATOR
>>>
>>>
>>> That's in commit ID 1c51c429f30ea10428337f3a33c12059ba59f668 from May 24, 2017.
>>>
>>>>>> However, arch/arm/include/asm/dma-mapping.h is referencing dma_noop_ops:
>>>>>>
>>>>>> static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
>>>>>> {
>>>>>> return IS_ENABLED(CONFIG_MMU) ? &arm_dma_ops : &dma_noop_ops;
>>>>>> }
>>>>>>
>>>>>> I will let a maintainer suggest the best resolution for this :)
>>>>>>
>>>>>
>>>>> add Bart and iommu mailing list.
>>>>>
>>>>
>>>> and add Vladimir.
>>>>
>>>>
>>>
>>>
>>> --
>>> ~Randy
> 

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ