Patches for -stable to improve support for recent gcc versions

Patches for -stable to improve support for recent gcc versions

[Guenter Roeck]

Hi Greg,

please apply the following patches.

to v4.4-stable and older:

cb36af3e48be sh: New gcc support
	to support gcc 8.1.0 for sh

to v4.9-stable and older:

009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer type
	to support gcc 7.3.0 for sparc

Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any idea ?

Thanks,
Guenter

Re: Patches for -stable to improve support for recent gcc versions

[Greg Kroah-Hartman]

On Mon, May 28, 2018 at 06:48:30AM -0700, Guenter Roeck wrote:
> Hi Greg,
> 
> please apply the following patches.
> 
> to v4.4-stable and older:
> 
> cb36af3e48be sh: New gcc support
> 	to support gcc 8.1.0 for sh

I don't see that commit id in Linus's tree, are you sure it is correct?

> to v4.9-stable and older:
> 
> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer type
> 	to support gcc 7.3.0 for sparc

Now queued up, thanks.

greg k-h

Re: Patches for -stable to improve support for recent gcc versions

[Arnd Bergmann]

On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> Hi Greg,
>
> please apply the following patches.
>
> to v4.4-stable and older:
>
> cb36af3e48be sh: New gcc support
>         to support gcc 8.1.0 for sh
>
> to v4.9-stable and older:
>
> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer
> type
>         to support gcc 7.3.0 for sparc
>
> Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any idea

No, sorry. I've just tried it again and couldn't find any problems with my
gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.

       Arnd

Re: Patches for -stable to improve support for recent gcc versions

[Greg Kroah-Hartman]

On Wed, May 30, 2018 at 11:27:50AM +0200, Arnd Bergmann wrote:
> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> > Hi Greg,
> >
> > please apply the following patches.
> >
> > to v4.4-stable and older:
> >
> > cb36af3e48be sh: New gcc support
> >         to support gcc 8.1.0 for sh
> >
> > to v4.9-stable and older:
> >
> > 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer
> > type
> >         to support gcc 7.3.0 for sparc
> >
> > Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any idea
> 
> No, sorry. I've just tried it again and couldn't find any problems with my
> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.

So you mean it works, or does not work?

And you are seeing loads of build warnings, right?  I tried your 8.1
binary for s390 and had to just give up looking at the warnings :(

thanks,

greg k-h

Re: Patches for -stable to improve support for recent gcc versions

[Arnd Bergmann]

On Wed, May 30, 2018 at 11:44 AM, Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> On Wed, May 30, 2018 at 11:27:50AM +0200, Arnd Bergmann wrote:
>> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
>> > Hi Greg,
>> >
>> > please apply the following patches.
>> >
>> > to v4.4-stable and older:
>> >
>> > cb36af3e48be sh: New gcc support
>> >         to support gcc 8.1.0 for sh
>> >
>> > to v4.9-stable and older:
>> >
>> > 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer
>> > type
>> >         to support gcc 7.3.0 for sparc
>> >
>> > Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any idea
>>
>> No, sorry. I've just tried it again and couldn't find any problems with my
>> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.
>
> So you mean it works, or does not work?
>
> And you are seeing loads of build warnings, right?  I tried your 8.1
> binary for s390 and had to just give up looking at the warnings :(

It does work, but yes, there are many warnings. I still need to revisit
my patch for the system call entry points: I posted one a long time
ago, but then the mainline implementation changed and I never got
around to sending an updated version. Similarly, I planned to send
a patch that disables all the -Wstringop-truncated warnings unless
'make W=1' is used, but that ended up in my long backlog of minor
fixes after my initial plan to fix address those warnings didn't work out.

       Arnd

Re: Patches for -stable to improve support for recent gcc versions

[Greg Kroah-Hartman]

On Wed, May 30, 2018 at 11:52:14AM +0200, Arnd Bergmann wrote:
> On Wed, May 30, 2018 at 11:44 AM, Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> > On Wed, May 30, 2018 at 11:27:50AM +0200, Arnd Bergmann wrote:
> >> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> >> > Hi Greg,
> >> >
> >> > please apply the following patches.
> >> >
> >> > to v4.4-stable and older:
> >> >
> >> > cb36af3e48be sh: New gcc support
> >> >         to support gcc 8.1.0 for sh
> >> >
> >> > to v4.9-stable and older:
> >> >
> >> > 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer
> >> > type
> >> >         to support gcc 7.3.0 for sparc
> >> >
> >> > Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any idea
> >>
> >> No, sorry. I've just tried it again and couldn't find any problems with my
> >> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.
> >
> > So you mean it works, or does not work?
> >
> > And you are seeing loads of build warnings, right?  I tried your 8.1
> > binary for s390 and had to just give up looking at the warnings :(
> 
> It does work, but yes, there are many warnings. I still need to revisit
> my patch for the system call entry points: I posted one a long time
> ago, but then the mainline implementation changed and I never got
> around to sending an updated version. Similarly, I planned to send
> a patch that disables all the -Wstringop-truncated warnings unless
> 'make W=1' is used, but that ended up in my long backlog of minor
> fixes after my initial plan to fix address those warnings didn't work out.

Given the recent mess that just happened yesterday when 0-day tried to
use gcc-8 and started emailing tons of innocent developers, it might be
good to dust those off soon :)

If you have a pointer to them, and are busy with other stuff, I can look
at getting them merged.

thanks,

greg k-h

Re: Patches for -stable to improve support for recent gcc versions

[Guenter Roeck]

On 05/30/2018 01:38 AM, Greg Kroah-Hartman wrote:
> On Mon, May 28, 2018 at 06:48:30AM -0700, Guenter Roeck wrote:
>> Hi Greg,
>>
>> please apply the following patches.
>>
>> to v4.4-stable and older:
>>
>> cb36af3e48be sh: New gcc support
>> 	to support gcc 8.1.0 for sh
> 
> I don't see that commit id in Linus's tree, are you sure it is correct?
> 
Oops, sorry, that was the sha after I test-applied it. Correct is

940d4113f330 sh: New gcc support

Guenter

>> to v4.9-stable and older:
>>
>> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer type
>> 	to support gcc 7.3.0 for sparc
> 
> Now queued up, thanks.
> 
> greg k-h
> 

Re: Patches for -stable to improve support for recent gcc versions

[Guenter Roeck]

On 05/30/2018 02:27 AM, Arnd Bergmann wrote:
> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
>> Hi Greg,
>>
>> please apply the following patches.
>>
>> to v4.4-stable and older:
>>
>> cb36af3e48be sh: New gcc support
>>          to support gcc 8.1.0 for sh
>>
>> to v4.9-stable and older:
>>
>> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible pointer
>> type
>>          to support gcc 7.3.0 for sparc
>>
>> Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any idea
> 
> No, sorry. I've just tried it again and couldn't find any problems with my
> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.
> 

Interesting. With sparc64-linux-gcc (GCC) 8.1.0 from kernel.org:

make ARCH=sparc64 CROSS_COMPILE=sparc64-linux- allmodconfig
make ARCH=sparc64 CROSS_COMPILE=sparc64-linux- arch/sparc/kernel/sys_sparc_64.o

results in:

./include/linux/syscalls.h:233:18: error:
	'sys_sparc64_personality' alias between functions of incompatible types
	'long int(long unsigned int)' and 'long int(long int)' [-Werror=attribute-alias]

and many more.

Presumably this is due to

arch/sparc/kernel/Makefile:ccflags-y := -Werror

Guenter

Re: Patches for -stable to improve support for recent gcc versions

[Greg Kroah-Hartman]

On Wed, May 30, 2018 at 06:18:10AM -0700, Guenter Roeck wrote:
> On 05/30/2018 01:38 AM, Greg Kroah-Hartman wrote:
> > On Mon, May 28, 2018 at 06:48:30AM -0700, Guenter Roeck wrote:
> > > Hi Greg,
> > > 
> > > please apply the following patches.
> > > 
> > > to v4.4-stable and older:
> > > 
> > > cb36af3e48be sh: New gcc support
> > > 	to support gcc 8.1.0 for sh
> > 
> > I don't see that commit id in Linus's tree, are you sure it is correct?
> > 
> Oops, sorry, that was the sha after I test-applied it. Correct is
> 
> 940d4113f330 sh: New gcc support

Thanks, that worked.

greg k-h

Re: Patches for -stable to improve support for recent gcc versions

[Arnd Bergmann]

On Wed, May 30, 2018 at 3:31 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> On 05/30/2018 02:27 AM, Arnd Bergmann wrote:
>>
>> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
>>>
>>> Hi Greg,
>>>
>>> please apply the following patches.
>>>
>>> to v4.4-stable and older:
>>>
>>> cb36af3e48be sh: New gcc support
>>>          to support gcc 8.1.0 for sh
>>>
>>> to v4.9-stable and older:
>>>
>>> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible
>>> pointer
>>> type
>>>          to support gcc 7.3.0 for sparc
>>>
>>> Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any
>>> idea
>>
>>
>> No, sorry. I've just tried it again and couldn't find any problems with my
>> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.
>>
>
> Interesting. With sparc64-linux-gcc (GCC) 8.1.0 from kernel.org:
>
> make ARCH=sparc64 CROSS_COMPILE=sparc64-linux- allmodconfig
> make ARCH=sparc64 CROSS_COMPILE=sparc64-linux-
> arch/sparc/kernel/sys_sparc_64.o
>
> results in:
>
> ./include/linux/syscalls.h:233:18: error:
>         'sys_sparc64_personality' alias between functions of incompatible
> types
>         'long int(long unsigned int)' and 'long int(long int)'
> [-Werror=attribute-alias]
>
> and many more.
>
> Presumably this is due to
>
> arch/sparc/kernel/Makefile:ccflags-y := -Werror

Right, I have disabled that warning locally, so I didn't get
it despite the -Werror.

       Arnd

Re: Patches for -stable to improve support for recent gcc versions

[Guenter Roeck]

On Wed, May 30, 2018 at 03:55:04PM +0200, Arnd Bergmann wrote:
> On Wed, May 30, 2018 at 3:31 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> > On 05/30/2018 02:27 AM, Arnd Bergmann wrote:
> >>
> >> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> >>>
> >>> Hi Greg,
> >>>
> >>> please apply the following patches.
> >>>
> >>> to v4.4-stable and older:
> >>>
> >>> cb36af3e48be sh: New gcc support
> >>>          to support gcc 8.1.0 for sh
> >>>
> >>> to v4.9-stable and older:
> >>>
> >>> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible
> >>> pointer
> >>> type
> >>>          to support gcc 7.3.0 for sparc
> >>>
> >>> Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any
> >>> idea
> >>
> >>
> >> No, sorry. I've just tried it again and couldn't find any problems with my
> >> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.
> >>
> >
> > Interesting. With sparc64-linux-gcc (GCC) 8.1.0 from kernel.org:
> >
> > make ARCH=sparc64 CROSS_COMPILE=sparc64-linux- allmodconfig
> > make ARCH=sparc64 CROSS_COMPILE=sparc64-linux-
> > arch/sparc/kernel/sys_sparc_64.o
> >
> > results in:
> >
> > ./include/linux/syscalls.h:233:18: error:
> >         'sys_sparc64_personality' alias between functions of incompatible
> > types
> >         'long int(long unsigned int)' and 'long int(long int)'
> > [-Werror=attribute-alias]
> >
> > and many more.
> >
> > Presumably this is due to
> >
> > arch/sparc/kernel/Makefile:ccflags-y := -Werror
> 
> Right, I have disabled that warning locally, so I didn't get
> it despite the -Werror.
> 

Would it make sense to add
	KBUILD_CFLAGS += $(call cc-option,-Wno-attribute-alias)
to the sparc Makefile ? It seems to fix the problem for me, and it would
be easier to backport than a more comprehensive fix.

Guenter

Re: Patches for -stable to improve support for recent gcc versions

[Greg Kroah-Hartman]

On Wed, May 30, 2018 at 11:46:45AM -0700, Guenter Roeck wrote:
> On Wed, May 30, 2018 at 03:55:04PM +0200, Arnd Bergmann wrote:
> > On Wed, May 30, 2018 at 3:31 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> > > On 05/30/2018 02:27 AM, Arnd Bergmann wrote:
> > >>
> > >> On Mon, May 28, 2018 at 3:48 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> > >>>
> > >>> Hi Greg,
> > >>>
> > >>> please apply the following patches.
> > >>>
> > >>> to v4.4-stable and older:
> > >>>
> > >>> cb36af3e48be sh: New gcc support
> > >>>          to support gcc 8.1.0 for sh
> > >>>
> > >>> to v4.9-stable and older:
> > >>>
> > >>> 009615ab7fd4 USB: serial: cp210x: use tcflag_t to fix incompatible
> > >>> pointer
> > >>> type
> > >>>          to support gcc 7.3.0 for sparc
> > >>>
> > >>> Copying Arnd - I can't get sparc images to compile with gcc 8.1.0. Any
> > >>> idea
> > >>
> > >>
> > >> No, sorry. I've just tried it again and couldn't find any problems with my
> > >> gcc-8.1 binaries building either sparc32_defconfig or sparc64_defconfig.
> > >>
> > >
> > > Interesting. With sparc64-linux-gcc (GCC) 8.1.0 from kernel.org:
> > >
> > > make ARCH=sparc64 CROSS_COMPILE=sparc64-linux- allmodconfig
> > > make ARCH=sparc64 CROSS_COMPILE=sparc64-linux-
> > > arch/sparc/kernel/sys_sparc_64.o
> > >
> > > results in:
> > >
> > > ./include/linux/syscalls.h:233:18: error:
> > >         'sys_sparc64_personality' alias between functions of incompatible
> > > types
> > >         'long int(long unsigned int)' and 'long int(long int)'
> > > [-Werror=attribute-alias]
> > >
> > > and many more.
> > >
> > > Presumably this is due to
> > >
> > > arch/sparc/kernel/Makefile:ccflags-y := -Werror
> > 
> > Right, I have disabled that warning locally, so I didn't get
> > it despite the -Werror.
> > 
> 
> Would it make sense to add
> 	KBUILD_CFLAGS += $(call cc-option,-Wno-attribute-alias)
> to the sparc Makefile ? It seems to fix the problem for me, and it would
> be easier to backport than a more comprehensive fix.

Why is that not an option enabled for all arches right now for gcc-8?

thanks,

greg k-h

Re: Patches for -stable to improve support for recent gcc versions

[Arnd Bergmann]

On Wed, May 30, 2018 at 9:24 PM, Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> On Wed, May 30, 2018 at 11:46:45AM -0700, Guenter Roeck wrote:
>> On Wed, May 30, 2018 at 03:55:04PM +0200, Arnd Bergmann wrote:
>> > On Wed, May 30, 2018 at 3:31 PM, Guenter Roeck <linux@roeck-us.net> wrote:
>> > > On 05/30/2018 02:27 AM, Arnd Bergmann wrote:

>> > > Presumably this is due to
>> > >
>> > > arch/sparc/kernel/Makefile:ccflags-y := -Werror
>> >
>> > Right, I have disabled that warning locally, so I didn't get
>> > it despite the -Werror.
>> >
>>
>> Would it make sense to add
>>       KBUILD_CFLAGS += $(call cc-option,-Wno-attribute-alias)
>> to the sparc Makefile ? It seems to fix the problem for me, and it would
>> be easier to backport than a more comprehensive fix.
>
> Why is that not an option enabled for all arches right now for gcc-8?

I would still want this warning enabled by default in future kernels,
just disabled
for the system call definitions (until we decide to rework the way
they are defined).

What I'd suggest we do is a series of patches:

1. disable both -Wno-attribute-alias and -Wstringop-truncation by
default, but leave them
    enabled in 'make W=1'. Mark this one for stable backports
2. add a macro to let users disable warnings locally within a file,
based on _Pragma("GCC
    diagnostic ...")
3. change the system call macros to disable -Wno-attribute-alias inside of the
    SYSCALL_DEFINEx() macros
4. turn on -Wno-attribute-alias again by default.

        Arnd

Re: Patches for -stable to improve support for recent gcc versions

[Greg Kroah-Hartman]

On Wed, May 30, 2018 at 09:34:34PM +0200, Arnd Bergmann wrote:
> On Wed, May 30, 2018 at 9:24 PM, Greg Kroah-Hartman
> > Why is that not an option enabled for all arches right now for gcc-8?
> 
> I would still want this warning enabled by default in future kernels,
> just disabled
> for the system call definitions (until we decide to rework the way
> they are defined).
> 
> What I'd suggest we do is a series of patches:
> 
> 1. disable both -Wno-attribute-alias and -Wstringop-truncation by
> default, but leave them
>     enabled in 'make W=1'. Mark this one for stable backports
> 2. add a macro to let users disable warnings locally within a file,
> based on _Pragma("GCC
>     diagnostic ...")
> 3. change the system call macros to disable -Wno-attribute-alias inside of the
>     SYSCALL_DEFINEx() macros
> 4. turn on -Wno-attribute-alias again by default.

That sounds like a good plan to me.

Re: Patches for -stable to improve support for recent gcc versions

[Guenter Roeck]

On Wed, May 30, 2018 at 10:04:51PM +0200, Greg Kroah-Hartman wrote:
> On Wed, May 30, 2018 at 09:34:34PM +0200, Arnd Bergmann wrote:
> > On Wed, May 30, 2018 at 9:24 PM, Greg Kroah-Hartman
> > > Why is that not an option enabled for all arches right now for gcc-8?
> > 
> > I would still want this warning enabled by default in future kernels,
> > just disabled
> > for the system call definitions (until we decide to rework the way
> > they are defined).
> > 
> > What I'd suggest we do is a series of patches:
> > 
> > 1. disable both -Wno-attribute-alias and -Wstringop-truncation by
> > default, but leave them
> >     enabled in 'make W=1'. Mark this one for stable backports
> > 2. add a macro to let users disable warnings locally within a file,
> > based on _Pragma("GCC
> >     diagnostic ...")
> > 3. change the system call macros to disable -Wno-attribute-alias inside of the
> >     SYSCALL_DEFINEx() macros
> > 4. turn on -Wno-attribute-alias again by default.
> 
> That sounds like a good plan to me.
> 
Agreed.

Guenter