2017-06-12 11:28:16

by Arnd Bergmann

[permalink] [raw]
Subject: Re: stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)

On Mon, Jun 12, 2017 at 12:59 PM, kernelci.org bot <[email protected]> wrote:
> stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)
>
>
> arm: gcc version 5.3.1 20160412 (Linaro GCC 5.3-2016.05)
>
> ep93xx_defconfig: 1 error
>
> mips: gcc version 6.3.0 (GCC)
>
> decstation_defconfig: 1 error
> defconfig+CONFIG_LKDTM=y: 1 error
> ip22_defconfig: 1 error
> jazz_defconfig: 1 error
> malta_defconfig: 1 error
> malta_kvm_defconfig: 1 error
> malta_kvm_guest_defconfig: 1 error
> maltaup_xpa_defconfig: 1 error
> nlm_xlp_defconfig: 1 error
> nlm_xlr_defconfig: 1 error
> rm200_defconfig: 1 error
>
> Errors summary:
>
> 12 fs/ufs/balloc.c:501:26: error: implicit declaration of function 'i_blocksize' [-Werror=implicit-function-declaration]

Same as stable-rc/linux-4.9.y, this needs a backport of 93407472a21b ("fs: add
i_blocksize()")

> 1 drivers/dma/ep93xx_dma.c:1357:9: error: 'struct dma_device' has no member named 'device_synchronize'

The error was introudced by the backport of 98f9de366fcc ("dmaengine: ep93xx:
Don't drain the transfers in terminate_all()"), which requires the dmaengine
synchronize callback.

device_synchronize was introduced by b36f09c3c441 ("dmaengine: Add transfer
termination synchronization support") in linux-4.5. That is not a
bugfix but would
cleanly apply on top of 4.4 and has no dependencies or side-effects.

We can either revert 98f9de366fcc or backport b36f09c3c441.

Arnd


2017-06-12 11:28:50

by Arnd Bergmann

[permalink] [raw]
Subject: Re: stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)

[+cc GregKH]

On Mon, Jun 12, 2017 at 1:28 PM, Arnd Bergmann <[email protected]> wrote:
> On Mon, Jun 12, 2017 at 12:59 PM, kernelci.org bot <[email protected]> wrote:
>> stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)
>>
>>
>> arm: gcc version 5.3.1 20160412 (Linaro GCC 5.3-2016.05)
>>
>> ep93xx_defconfig: 1 error
>>
>> mips: gcc version 6.3.0 (GCC)
>>
>> decstation_defconfig: 1 error
>> defconfig+CONFIG_LKDTM=y: 1 error
>> ip22_defconfig: 1 error
>> jazz_defconfig: 1 error
>> malta_defconfig: 1 error
>> malta_kvm_defconfig: 1 error
>> malta_kvm_guest_defconfig: 1 error
>> maltaup_xpa_defconfig: 1 error
>> nlm_xlp_defconfig: 1 error
>> nlm_xlr_defconfig: 1 error
>> rm200_defconfig: 1 error
>>
>> Errors summary:
>>
>> 12 fs/ufs/balloc.c:501:26: error: implicit declaration of function 'i_blocksize' [-Werror=implicit-function-declaration]
>
> Same as stable-rc/linux-4.9.y, this needs a backport of 93407472a21b ("fs: add
> i_blocksize()")
>
>> 1 drivers/dma/ep93xx_dma.c:1357:9: error: 'struct dma_device' has no member named 'device_synchronize'
>
> The error was introudced by the backport of 98f9de366fcc ("dmaengine: ep93xx:
> Don't drain the transfers in terminate_all()"), which requires the dmaengine
> synchronize callback.
>
> device_synchronize was introduced by b36f09c3c441 ("dmaengine: Add transfer
> termination synchronization support") in linux-4.5. That is not a
> bugfix but would
> cleanly apply on top of 4.4 and has no dependencies or side-effects.
>
> We can either revert 98f9de366fcc or backport b36f09c3c441.
>
> Arnd

2017-06-12 11:59:54

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)

On Mon, Jun 12, 2017 at 01:28:48PM +0200, Arnd Bergmann wrote:
> [+cc GregKH]
>
> On Mon, Jun 12, 2017 at 1:28 PM, Arnd Bergmann <[email protected]> wrote:
> > On Mon, Jun 12, 2017 at 12:59 PM, kernelci.org bot <[email protected]> wrote:
> >> stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)
> >>
> >>
> >> arm: gcc version 5.3.1 20160412 (Linaro GCC 5.3-2016.05)
> >>
> >> ep93xx_defconfig: 1 error
> >>
> >> mips: gcc version 6.3.0 (GCC)
> >>
> >> decstation_defconfig: 1 error
> >> defconfig+CONFIG_LKDTM=y: 1 error
> >> ip22_defconfig: 1 error
> >> jazz_defconfig: 1 error
> >> malta_defconfig: 1 error
> >> malta_kvm_defconfig: 1 error
> >> malta_kvm_guest_defconfig: 1 error
> >> maltaup_xpa_defconfig: 1 error
> >> nlm_xlp_defconfig: 1 error
> >> nlm_xlr_defconfig: 1 error
> >> rm200_defconfig: 1 error
> >>
> >> Errors summary:
> >>
> >> 12 fs/ufs/balloc.c:501:26: error: implicit declaration of function 'i_blocksize' [-Werror=implicit-function-declaration]
> >
> > Same as stable-rc/linux-4.9.y, this needs a backport of 93407472a21b ("fs: add
> > i_blocksize()")

This is now fixed.

> >> 1 drivers/dma/ep93xx_dma.c:1357:9: error: 'struct dma_device' has no member named 'device_synchronize'
> >
> > The error was introudced by the backport of 98f9de366fcc ("dmaengine: ep93xx:
> > Don't drain the transfers in terminate_all()"), which requires the dmaengine
> > synchronize callback.
> >
> > device_synchronize was introduced by b36f09c3c441 ("dmaengine: Add transfer
> > termination synchronization support") in linux-4.5. That is not a
> > bugfix but would
> > cleanly apply on top of 4.4 and has no dependencies or side-effects.
> >
> > We can either revert 98f9de366fcc or backport b36f09c3c441.

I'll just go drop 98f9de366fcc as that's best here. I don't think
anyone really cares about this driver in 4.4, and if they do, then they
can speak up :)

thanks,

greg k-h

2017-06-12 12:44:08

by Arnd Bergmann

[permalink] [raw]
Subject: Re: stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)

On Mon, Jun 12, 2017 at 1:59 PM, gregkh <[email protected]> wrote:
> On Mon, Jun 12, 2017 at 01:28:48PM +0200, Arnd Bergmann wrote:
>> >> 1 drivers/dma/ep93xx_dma.c:1357:9: error: 'struct dma_device' has no member named 'device_synchronize'
>> >
>> > The error was introudced by the backport of 98f9de366fcc ("dmaengine: ep93xx:
>> > Don't drain the transfers in terminate_all()"), which requires the dmaengine
>> > synchronize callback.
>> >
>> > device_synchronize was introduced by b36f09c3c441 ("dmaengine: Add transfer
>> > termination synchronization support") in linux-4.5. That is not a
>> > bugfix but would
>> > cleanly apply on top of 4.4 and has no dependencies or side-effects.
>> >
>> > We can either revert 98f9de366fcc or backport b36f09c3c441.
>
> I'll just go drop 98f9de366fcc as that's best here. I don't think
> anyone really cares about this driver in 4.4, and if they do, then they
> can speak up :)

Sounds fine. ep93xx was once an important platform and the chips are still
being sold, but I guess not many users remain. Alexander Sverdlin just stepped
up as the platform maintainer, so he can decide whether he wants his fix
in linux-4.4.y.

Arnd

2017-06-12 13:32:38

by Alexander Sverdlin

[permalink] [raw]
Subject: Re: stable-rc/linux-4.4.y build: 199 builds: 13 failed, 186 passed, 13 errors, 4 warnings (v4.4.71-63-g409ebd600970)

Hello!

On Mon Jun 12 14:44:04 2017 Arnd Bergmann <[email protected]> wrote:
> On Mon, Jun 12, 2017 at 1:59 PM, gregkh <[email protected]>
> wrote:
> > On Mon, Jun 12, 2017 at 01:28:48PM +0200, Arnd Bergmann wrote:
> > > > > 1  drivers/dma/ep93xx_dma.c:1357:9: error: 'struct dma_device'
> > > > > has no member named 'device_synchronize'
> > > >
> > > > The error was introudced by the backport of 98f9de366fcc
> > > > ("dmaengine: ep93xx: Don't drain the transfers in
> > > > terminate_all()"), which requires the dmaengine synchronize
> > > > callback.
> > > >
> > > > device_synchronize was introduced by b36f09c3c441 ("dmaengine: Add
> > > > transfer termination synchronization support") in linux-4.5. That
> > > > is not a bugfix but would
> > > > cleanly apply on top of 4.4 and has no dependencies or
> > > > side-effects.
> > > >
> > > > We can either revert 98f9de366fcc or backport b36f09c3c441.
> >
> > I'll just go drop 98f9de366fcc as that's best here.  I don't think
> > anyone really cares about this driver in 4.4, and if they do, then they
> > can speak up :)
>
> Sounds fine. ep93xx was once an important platform and the chips are
> still being sold, but I guess not many users remain. Alexander Sverdlin
> just stepped up as the platform maintainer, so he can decide whether he
> wants his fix in linux-4.4.y.

Fair enough! While I'm aware of several thousands devices still in service, there is no "enterprise Linux distribution" for this platform, these embedded devices are able to switch to the fresh kernel.

So, it doesn't worth the effort, repairing v4.4.x is meaningless.

--
Thanks,
Alexander.