Re: [GIT PULL] Hisilicon fixes for v5.2

[John Garry]

On 25/06/2019 14:31, John Garry wrote:
> On 25/06/2019 14:03, Olof Johansson wrote:
> > > > are available in the Git repository at:
> > > > > >
> > > > > >   git://github.com/hisilicon/linux-hisi.git tags/hisi-fixes-for-5.2
> > > > > >
> > > > > > for you to fetch changes up to
> > > > 07c811af1c00d7b4212eac86900b023b6405a954:
> > > > > >
> > > > > >   lib: logic_pio: Enforce LOGIC_PIO_INDIRECT region ops are set
> > > > at registration (2019-06-25 09:40:42 +0100)
> > > > > >
> > > > > > ----------------------------------------------------------------
> > > > > > Hisilicon fixes for v5.2-rc
> > > > > >
> > > > > > - fixed RCU usage in logical PIO
> > > > > > - Added a function to unregister a logical PIO range in logical PIO
> > > > > >   to support the fixes in the hisi-lpc driver
> > > > > > - fixed and optimized hisi-lpc driver to avoid potential
> > > > use-after-free
> > > > > >   and driver unbind crash
> > > >
> > > > Merged to fixes, thanks.
> >
> > This broke arm64 allmodconfig:
> >
> >        arm64.allmodconfig:
> > drivers/bus/hisi_lpc.c:656:3: error: implicit declaration of function
> > 'hisi_lpc_acpi_remove'; did you mean 'hisi_lpc_acpi_probe'?
> > [-Werror=implicit-function-declaration]

As an aside, I find it a little strange that arm64 allmodconfig does not 
have CONFIG_ACPI set. It used to have it set, and this patch stopped that:

5bcd44083a082f314032969cd6db1eb8275ac77a is the first bad commit
commit 5bcd44083a082f314032969cd6db1eb8275ac77a
Author: AKASHI Takahiro <takahiro.akashi@xxxxxxxxxx>
Date:   Mon Jul 23 10:57:29 2018 +0900

    drivers: acpi: add dependency of EFI for arm64

    As Ard suggested, CONFIG_ACPI && !CONFIG_EFI doesn't make sense on 
arm64,
    while CONFIG_ACPI and CONFIG_CPU_BIG_ENDIAN doesn't make sense either.

    As CONFIG_EFI already has a dependency of !CONFIG_CPU_BIG_ENDIAN, it is
    good enough to add a dependency of CONFIG_EFI to avoid any useless
    combination of configuration.

    This bug, reported by Will, will be revealed when my patch series,
    "arm64: kexec,kdump: fix boot failures on acpi-only system," is applied
    and the kernel is built under allmodconfig.

    Signed-off-by: AKASHI Takahiro <takahiro.akashi@xxxxxxxxxx>
    Suggested-by: Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>
    Signed-off-by: Will Deacon <will.deacon@xxxxxxx>

That patch stopped many configs being set for allmodconfig.

With this change, CONFIG_EFI is not set. I think that this is because 
CONFIG_CPU_BIG_ENDIAN is set for arm64 allmodconfig.

Any opinion on this? Could we change CONFIG_CPU_BIG_ENDIAN to be unset 
for arm64?

> >
>
> Uhhh, that's my fault - I didn't provide a stub for !ACPI. Sorry. I'll
> send a fixed v3 series.
>
> > Please build and test your branches before you send pull requests, Wei.
> >
> > I've dropped the branch again; please re-submit when fixed. I think
> > it's probably 5.3 material now.
>
> Thanks,
> John
>
> > -Olof
> >
> > .