Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Thomas Lindroth]

Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
Gentoo and Arch apply their own patch to fix the problem in their distribution
kernels but I would prefer if a proper fix was applied upstream.

https://bugs.gentoo.org/814200 a gentoo report with the relevant info.

I've searched for any upstream discussions about the problem but I've only found
one message saying the backport needs an additional fix. That was almost a year
ago. https://www.spinics.net/lists/stable/msg438000.html

/Thomas Lindroth

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Greg KH]

On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
> Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
> Gentoo and Arch apply their own patch to fix the problem in their distribution
> kernels but I would prefer if a proper fix was applied upstream.
> 
> https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
> 
> I've searched for any upstream discussions about the problem but I've only found
> one message saying the backport needs an additional fix. That was almost a year
> ago. https://www.spinics.net/lists/stable/msg438000.html

We can not take a patch in a stable kernel release unless it is already
in Linus's tree.  Please work to get a patch accepted there first,
before worrying about older kernel versions.

thanks,

greg k-h

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Thomas Lindroth]

On 12/6/21 14:04, Greg KH wrote:
> On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
>> Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
>> Gentoo and Arch apply their own patch to fix the problem in their distribution
>> kernels but I would prefer if a proper fix was applied upstream.
>>
>> https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
>>
>> I've searched for any upstream discussions about the problem but I've only found
>> one message saying the backport needs an additional fix. That was almost a year
>> ago. https://www.spinics.net/lists/stable/msg438000.html
> 
> We can not take a patch in a stable kernel release unless it is already
> in Linus's tree.  Please work to get a patch accepted there first,
> before worrying about older kernel versions.
> 
> thanks,
> 
> greg k-h
> 

The problem was fixed in Linus tree in commit 67a5a68013056cbcf0a647e36cb6f4622fb6a470
"gcc-plugins: fix gcc 11 indigestion with plugins..." added in v5.11

That's the patch that needed some kind of additional fix before it could be backported
but that fix never materialized.

/Thomas Lindroth

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Greg KH]

On Mon, Dec 06, 2021 at 02:24:29PM +0100, Thomas Lindroth wrote:
> On 12/6/21 14:04, Greg KH wrote:
> > On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
> > > Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
> > > Gentoo and Arch apply their own patch to fix the problem in their distribution
> > > kernels but I would prefer if a proper fix was applied upstream.
> > > 
> > > https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
> > > 
> > > I've searched for any upstream discussions about the problem but I've only found
> > > one message saying the backport needs an additional fix. That was almost a year
> > > ago. https://www.spinics.net/lists/stable/msg438000.html
> > 
> > We can not take a patch in a stable kernel release unless it is already
> > in Linus's tree.  Please work to get a patch accepted there first,
> > before worrying about older kernel versions.
> > 
> > thanks,
> > 
> > greg k-h
> > 
> 
> The problem was fixed in Linus tree in commit 67a5a68013056cbcf0a647e36cb6f4622fb6a470
> "gcc-plugins: fix gcc 11 indigestion with plugins..." added in v5.11
> 
> That's the patch that needed some kind of additional fix before it could be backported
> but that fix never materialized.

If you have a working version, based on a distro's use, please provide
it and I will be glad to pick it up.

thanks,

greg k-h

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Thomas Lindroth]

On 12/6/21 14:27, Greg KH wrote:
> On Mon, Dec 06, 2021 at 02:24:29PM +0100, Thomas Lindroth wrote:
>> On 12/6/21 14:04, Greg KH wrote:
>>> On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
>>>> Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
>>>> Gentoo and Arch apply their own patch to fix the problem in their distribution
>>>> kernels but I would prefer if a proper fix was applied upstream.
>>>>
>>>> https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
>>>>
>>>> I've searched for any upstream discussions about the problem but I've only found
>>>> one message saying the backport needs an additional fix. That was almost a year
>>>> ago. https://www.spinics.net/lists/stable/msg438000.html
>>>
>>> We can not take a patch in a stable kernel release unless it is already
>>> in Linus's tree.  Please work to get a patch accepted there first,
>>> before worrying about older kernel versions.
>>>
>>> thanks,
>>>
>>> greg k-h
>>>
>>
>> The problem was fixed in Linus tree in commit 67a5a68013056cbcf0a647e36cb6f4622fb6a470
>> "gcc-plugins: fix gcc 11 indigestion with plugins..." added in v5.11
>>
>> That's the patch that needed some kind of additional fix before it could be backported
>> but that fix never materialized.
> 
> If you have a working version, based on a distro's use, please provide
> it and I will be glad to pick it up.
> 
> thanks,
> 
> greg k-h
> 

https://dev.gentoo.org/~mpagano/genpatches/trunk/5.10/2910_fix-gcc-detection-method.patch

Here is the patch gentoo applies to 5.10. It seems to be a combination of two upstream
commits:
67a5a68013056cbcf0a647e36cb6f4622fb6a470 "gcc-plugins: fix gcc 11 indigestion with plugins..."
1e860048c53ee77ee9870dcce94847a28544b753 "gcc-plugins: simplify GCC plugin-dev capability test"

I can't vouch for the correctness of that fix. I'm just a regular user who stumbled upon this
problem and found that gentoo bug report. Check with Kees Cook for that "additional fix". I
don't know what fix that is.

/Thomas Lindroth

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Greg KH]

On Mon, Dec 06, 2021 at 02:59:57PM +0100, Thomas Lindroth wrote:
> On 12/6/21 14:27, Greg KH wrote:
> > On Mon, Dec 06, 2021 at 02:24:29PM +0100, Thomas Lindroth wrote:
> > > On 12/6/21 14:04, Greg KH wrote:
> > > > On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
> > > > > Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
> > > > > Gentoo and Arch apply their own patch to fix the problem in their distribution
> > > > > kernels but I would prefer if a proper fix was applied upstream.
> > > > > 
> > > > > https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
> > > > > 
> > > > > I've searched for any upstream discussions about the problem but I've only found
> > > > > one message saying the backport needs an additional fix. That was almost a year
> > > > > ago. https://www.spinics.net/lists/stable/msg438000.html
> > > > 
> > > > We can not take a patch in a stable kernel release unless it is already
> > > > in Linus's tree.  Please work to get a patch accepted there first,
> > > > before worrying about older kernel versions.
> > > > 
> > > > thanks,
> > > > 
> > > > greg k-h
> > > > 
> > > 
> > > The problem was fixed in Linus tree in commit 67a5a68013056cbcf0a647e36cb6f4622fb6a470
> > > "gcc-plugins: fix gcc 11 indigestion with plugins..." added in v5.11
> > > 
> > > That's the patch that needed some kind of additional fix before it could be backported
> > > but that fix never materialized.
> > 
> > If you have a working version, based on a distro's use, please provide
> > it and I will be glad to pick it up.
> > 
> > thanks,
> > 
> > greg k-h
> > 
> 
> https://dev.gentoo.org/~mpagano/genpatches/trunk/5.10/2910_fix-gcc-detection-method.patch
> 
> Here is the patch gentoo applies to 5.10. It seems to be a combination of two upstream
> commits:
> 67a5a68013056cbcf0a647e36cb6f4622fb6a470 "gcc-plugins: fix gcc 11 indigestion with plugins..."
> 1e860048c53ee77ee9870dcce94847a28544b753 "gcc-plugins: simplify GCC plugin-dev capability test"

That's not how I can take a patch, sorry.  Please submit these as
individual emails so that then could be applied.

> I can't vouch for the correctness of that fix. I'm just a regular user who stumbled upon this
> problem and found that gentoo bug report. Check with Kees Cook for that "additional fix". I
> don't know what fix that is.

I don't think either of them differ from what is in mainline, so I do
not know.

thanks,

greg k-h

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Kees Cook]

On Mon, Dec 06, 2021 at 03:41:55PM +0100, Greg KH wrote:
> On Mon, Dec 06, 2021 at 02:59:57PM +0100, Thomas Lindroth wrote:
> > On 12/6/21 14:27, Greg KH wrote:
> > > On Mon, Dec 06, 2021 at 02:24:29PM +0100, Thomas Lindroth wrote:
> > > > On 12/6/21 14:04, Greg KH wrote:
> > > > > On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
> > > > > > Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
> > > > > > Gentoo and Arch apply their own patch to fix the problem in their distribution
> > > > > > kernels but I would prefer if a proper fix was applied upstream.
> > > > > > 
> > > > > > https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
> > > > > > 
> > > > > > I've searched for any upstream discussions about the problem but I've only found
> > > > > > one message saying the backport needs an additional fix. That was almost a year
> > > > > > ago. https://www.spinics.net/lists/stable/msg438000.html
> > > > > 
> > > > > We can not take a patch in a stable kernel release unless it is already
> > > > > in Linus's tree.  Please work to get a patch accepted there first,
> > > > > before worrying about older kernel versions.
> > > > > 
> > > > > thanks,
> > > > > 
> > > > > greg k-h
> > > > > 
> > > > 
> > > > The problem was fixed in Linus tree in commit 67a5a68013056cbcf0a647e36cb6f4622fb6a470
> > > > "gcc-plugins: fix gcc 11 indigestion with plugins..." added in v5.11
> > > > 
> > > > That's the patch that needed some kind of additional fix before it could be backported
> > > > but that fix never materialized.
> > > 
> > > If you have a working version, based on a distro's use, please provide
> > > it and I will be glad to pick it up.
> > > 
> > > thanks,
> > > 
> > > greg k-h
> > > 
> > 
> > https://dev.gentoo.org/~mpagano/genpatches/trunk/5.10/2910_fix-gcc-detection-method.patch
> > 
> > Here is the patch gentoo applies to 5.10. It seems to be a combination of two upstream
> > commits:
> > 67a5a68013056cbcf0a647e36cb6f4622fb6a470 "gcc-plugins: fix gcc 11 indigestion with plugins..."
> > 1e860048c53ee77ee9870dcce94847a28544b753 "gcc-plugins: simplify GCC plugin-dev capability test"
> 
> That's not how I can take a patch, sorry.  Please submit these as
> individual emails so that then could be applied.
> 
> > I can't vouch for the correctness of that fix. I'm just a regular user who stumbled upon this
> > problem and found that gentoo bug report. Check with Kees Cook for that "additional fix". I
> > don't know what fix that is.
> 
> I don't think either of them differ from what is in mainline, so I do
> not know.

Hi!

Yeah, the two fixes that I think are needed for v5.10 are:

1e860048c53ee77ee9870dcce94847a28544b753
67a5a68013056cbcf0a647e36cb6f4622fb6a470

This additionally appears to match the combo patch in gentoo:
https://814200.bugs.gentoo.org/attachment.cgi?id=746715

Local build testing shows the plugins working...

-- 
Kees Cook

Re: Could the fix for broken gcc-plugins with gcc-11 be backported to 5.10?

[Greg KH]

On Mon, Dec 06, 2021 at 12:29:31PM -0800, Kees Cook wrote:
> On Mon, Dec 06, 2021 at 03:41:55PM +0100, Greg KH wrote:
> > On Mon, Dec 06, 2021 at 02:59:57PM +0100, Thomas Lindroth wrote:
> > > On 12/6/21 14:27, Greg KH wrote:
> > > > On Mon, Dec 06, 2021 at 02:24:29PM +0100, Thomas Lindroth wrote:
> > > > > On 12/6/21 14:04, Greg KH wrote:
> > > > > > On Mon, Dec 06, 2021 at 01:59:31PM +0100, Thomas Lindroth wrote:
> > > > > > > Build support for gcc-plugins are not detected with gcc-11 in lts kernels.
> > > > > > > Gentoo and Arch apply their own patch to fix the problem in their distribution
> > > > > > > kernels but I would prefer if a proper fix was applied upstream.
> > > > > > > 
> > > > > > > https://bugs.gentoo.org/814200 a gentoo report with the relevant info.
> > > > > > > 
> > > > > > > I've searched for any upstream discussions about the problem but I've only found
> > > > > > > one message saying the backport needs an additional fix. That was almost a year
> > > > > > > ago. https://www.spinics.net/lists/stable/msg438000.html
> > > > > > 
> > > > > > We can not take a patch in a stable kernel release unless it is already
> > > > > > in Linus's tree.  Please work to get a patch accepted there first,
> > > > > > before worrying about older kernel versions.
> > > > > > 
> > > > > > thanks,
> > > > > > 
> > > > > > greg k-h
> > > > > > 
> > > > > 
> > > > > The problem was fixed in Linus tree in commit 67a5a68013056cbcf0a647e36cb6f4622fb6a470
> > > > > "gcc-plugins: fix gcc 11 indigestion with plugins..." added in v5.11
> > > > > 
> > > > > That's the patch that needed some kind of additional fix before it could be backported
> > > > > but that fix never materialized.
> > > > 
> > > > If you have a working version, based on a distro's use, please provide
> > > > it and I will be glad to pick it up.
> > > > 
> > > > thanks,
> > > > 
> > > > greg k-h
> > > > 
> > > 
> > > https://dev.gentoo.org/~mpagano/genpatches/trunk/5.10/2910_fix-gcc-detection-method.patch
> > > 
> > > Here is the patch gentoo applies to 5.10. It seems to be a combination of two upstream
> > > commits:
> > > 67a5a68013056cbcf0a647e36cb6f4622fb6a470 "gcc-plugins: fix gcc 11 indigestion with plugins..."
> > > 1e860048c53ee77ee9870dcce94847a28544b753 "gcc-plugins: simplify GCC plugin-dev capability test"
> > 
> > That's not how I can take a patch, sorry.  Please submit these as
> > individual emails so that then could be applied.
> > 
> > > I can't vouch for the correctness of that fix. I'm just a regular user who stumbled upon this
> > > problem and found that gentoo bug report. Check with Kees Cook for that "additional fix". I
> > > don't know what fix that is.
> > 
> > I don't think either of them differ from what is in mainline, so I do
> > not know.
> 
> Hi!
> 
> Yeah, the two fixes that I think are needed for v5.10 are:
> 
> 1e860048c53ee77ee9870dcce94847a28544b753
> 67a5a68013056cbcf0a647e36cb6f4622fb6a470
> 
> This additionally appears to match the combo patch in gentoo:
> https://814200.bugs.gentoo.org/attachment.cgi?id=746715
> 
> Local build testing shows the plugins working...

Ok, both now queued up, thanks.

greg k-h