Issue with pxa_dma driver on linux 4.4.y

Issue with pxa_dma driver on linux 4.4.y

[Robert Jarzmik]

Dear maintainers,

In the tagging of fixing commits, I have submitted this patch which was merged
into the stable tree :
 - 6bab1c6afdca ("dmaengine: pxa_dma: fix the maximum requestor line")
which translated into this commit in the stable tree :
 - 94d75e190f19 ("dmaengine: pxa_dma: fix the maximum requestor line")

Unfortunately I have not tagged the dependant patches this one relies on, and
more specifically :
 - build breakage fix :
   - c283e41ef324 ("dmaengine: mmp-pdma: add number of requestors")
 - functionality breakage fix :
   - 72b195cb7162 ("ARM: pxa: add the number of DMA requestor lines")
   - 4c35430ad18f ("ARM: pxa: fix the number of DMA requestor lines")

I see 2 ways out of this situation :
 - either we revert stable commit 94d75e190f19
   => functionnality will remain broken, but only for a subset of DMA lines
 - or apply the dependant patches, ie. c283e41ef324, 72b195cb7162, 4c35430ad18f

It's up to you, but if you could choose one and apply it, the build breakage
will be gone.

I must mention that there is no defconfig file so far with covers this specific
build, and this was seen by Andrea who reported the issue.

Cheers.

-- 
Robert

PS: Andrea, please correct me if I misunderstood something.

Re: Issue with pxa_dma driver on linux 4.4.y

[Greg KH]

On Mon, Sep 25, 2017 at 11:33:42PM +0200, Robert Jarzmik wrote:
> Dear maintainers,
> 
> In the tagging of fixing commits, I have submitted this patch which was merged
> into the stable tree :
>  - 6bab1c6afdca ("dmaengine: pxa_dma: fix the maximum requestor line")
> which translated into this commit in the stable tree :
>  - 94d75e190f19 ("dmaengine: pxa_dma: fix the maximum requestor line")
> 
> Unfortunately I have not tagged the dependant patches this one relies on, and
> more specifically :
>  - build breakage fix :
>    - c283e41ef324 ("dmaengine: mmp-pdma: add number of requestors")
>  - functionality breakage fix :
>    - 72b195cb7162 ("ARM: pxa: add the number of DMA requestor lines")
>    - 4c35430ad18f ("ARM: pxa: fix the number of DMA requestor lines")
> 
> I see 2 ways out of this situation :
>  - either we revert stable commit 94d75e190f19
>    => functionnality will remain broken, but only for a subset of DMA lines
>  - or apply the dependant patches, ie. c283e41ef324, 72b195cb7162, 4c35430ad18f
> 
> It's up to you, but if you could choose one and apply it, the build breakage
> will be gone.

I've applied the extra patches, thanks for letting me know.

> I must mention that there is no defconfig file so far with covers this specific
> build, and this was seen by Andrea who reported the issue.

Ah, that's why none of the build systems caught this.  Perhaps we should
add a defconfig for this?  Is there one in Linus's tree yet?

thanks,

greg k-h

Re: Issue with pxa_dma driver on linux 4.4.y

[Andrea Adami]

On Tue, Oct 3, 2017 at 10:58 AM, Greg KH <gregkh@linuxfoundation.org> wrote:
> On Mon, Sep 25, 2017 at 11:33:42PM +0200, Robert Jarzmik wrote:
>> Dear maintainers,
>>
>> In the tagging of fixing commits, I have submitted this patch which was merged
>> into the stable tree :
>>  - 6bab1c6afdca ("dmaengine: pxa_dma: fix the maximum requestor line")
>> which translated into this commit in the stable tree :
>>  - 94d75e190f19 ("dmaengine: pxa_dma: fix the maximum requestor line")
>>
>> Unfortunately I have not tagged the dependant patches this one relies on, and
>> more specifically :
>>  - build breakage fix :
>>    - c283e41ef324 ("dmaengine: mmp-pdma: add number of requestors")
>>  - functionality breakage fix :
>>    - 72b195cb7162 ("ARM: pxa: add the number of DMA requestor lines")
>>    - 4c35430ad18f ("ARM: pxa: fix the number of DMA requestor lines")
>>
>> I see 2 ways out of this situation :
>>  - either we revert stable commit 94d75e190f19
>>    => functionnality will remain broken, but only for a subset of DMA lines
>>  - or apply the dependant patches, ie. c283e41ef324, 72b195cb7162, 4c35430ad18f
>>
>> It's up to you, but if you could choose one and apply it, the build breakage
>> will be gone.
>
> I've applied the extra patches, thanks for letting me know.
>
>> I must mention that there is no defconfig file so far with covers this specific
>> build, and this was seen by Andrea who reported the issue.
>
> Ah, that's why none of the build systems caught this.  Perhaps we should
> add a defconfig for this?  Is there one in Linus's tree yet?
>
> thanks,
>
> greg k-h

Hi,

I see two candidates here:

corgi_defconfig (pxa255)
spitz_defconfig (pxa270)

I'd say spitz because there is the corresponding qemu machine.

The config seems a bit outdated though...spitz at least uses PATA_PXA
but none of them has CONFIG_PXA_DMA. Ouch...

We do have our defconfigs in meta-handheld [1] but maybe these are too
heavy, able to boot from every media, many filesystems, with perf and
other stuff.
Robert, whith your help we can maybe patch the ones upstream?

Cheers
Andrea

[1] http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-handheld-4.4

Re: Issue with pxa_dma driver on linux 4.4.y

[Robert Jarzmik]

Greg and Andrea write :

>> Ah, that's why none of the build systems caught this.  Perhaps we should
>> add a defconfig for this?  Is there one in Linus's tree yet?
There is one, but added only later in v4.5 here :
3168eb330b86 ("ARM: pxa: add defconfig covering all the boards")

...zip to Andrea ...
> corgi_defconfig (pxa255)
> spitz_defconfig (pxa270)
>
> I'd say spitz because there is the corresponding qemu machine.
>
> The config seems a bit outdated though...spitz at least uses PATA_PXA
> but none of them has CONFIG_PXA_DMA. Ouch...
>
> We do have our defconfigs in meta-handheld [1] but maybe these are too
> heavy, able to boot from every media, many filesystems, with perf and
> other stuff.
> Robert, whith your help we can maybe patch the ones upstream?
As long as they are pretty tiny yes. I'll be received with a machine gun if I
have a big defconfig patch to push upstream ...

Cheers.

-- 
Robert