linux-next: build warning after merge of the bpf-next tree

linux-next: build warning after merge of the bpf-next tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 389 bytes --]

Hi all,

After merging the bpf-next tree, today's linux-next build (powerpc
ppc64_defconfig) produced this warning:

WARNING: 2 bad relocations
c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end

Introduced by commit

  8580ac9404f6 ("bpf: Process in-kernel BTF")

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build warning after merge of the bpf-next tree

[Alexei Starovoitov]

On Fri, Oct 18, 2019 at 10:56:57AM +1100, Stephen Rothwell wrote:
> Hi all,
> 
> After merging the bpf-next tree, today's linux-next build (powerpc
> ppc64_defconfig) produced this warning:
> 
> WARNING: 2 bad relocations
> c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
> c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end

Can ppc folks help me figure out what this warning means?

Re: linux-next: build warning after merge of the bpf-next tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 566 bytes --]

Hi all,

On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
> Hi all,
> 
> After merging the bpf-next tree, today's linux-next build (powerpc
> ppc64_defconfig) produced this warning:
> 
> WARNING: 2 bad relocations
> c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
> c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end
> 
> Introduced by commit
> 
>   8580ac9404f6 ("bpf: Process in-kernel BTF")

This warning now appears in the net-next tree build.

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: Re: linux-next: build warning after merge of the bpf-next tree

[Alexandre Ghiti]

Hi guys,

On 10/27/19 8:02 PM, Stephen Rothwell wrote:
> Hi all,
>
> On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>> Hi all,
>>
>> After merging the bpf-next tree, today's linux-next build (powerpc
>> ppc64_defconfig) produced this warning:
>>
>> WARNING: 2 bad relocations
>> c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
>> c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end
>>
>> Introduced by commit
>>
>>    8580ac9404f6 ("bpf: Process in-kernel BTF")
> This warning now appears in the net-next tree build.
>
>
I bump that thread up because Zong also noticed that 2 new relocations for
those symbols appeared in my riscv relocatable kernel branch following 
that commit.

I also noticed 2 new relocations R_AARCH64_ABS64 appearing in arm64 kernel.

Those 2 weak undefined symbols have existed since commit
341dfcf8d78e ("btf: expose BTF info through sysfs") but this is the fact
to declare those symbols into btf.c that produced those relocations.

I'm not sure what this all means, but this is not something I expected 
for riscv for
a kernel linked with -shared/-fpie. Maybe should we just leave them to 
zero ?

I think that deserves a deeper look if someone understands all this 
better than I do.

Alex

Re: Re: linux-next: build warning after merge of the bpf-next tree

[Alexei Starovoitov]

On Fri, Jan 10, 2020 at 2:28 PM Alexandre Ghiti <alexandre@ghiti.fr> wrote:
>
> Hi guys,
>
> On 10/27/19 8:02 PM, Stephen Rothwell wrote:
> > Hi all,
> >
> > On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> >> Hi all,
> >>
> >> After merging the bpf-next tree, today's linux-next build (powerpc
> >> ppc64_defconfig) produced this warning:
> >>
> >> WARNING: 2 bad relocations
> >> c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
> >> c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end
> >>
> >> Introduced by commit
> >>
> >>    8580ac9404f6 ("bpf: Process in-kernel BTF")
> > This warning now appears in the net-next tree build.
> >
> >
> I bump that thread up because Zong also noticed that 2 new relocations for
> those symbols appeared in my riscv relocatable kernel branch following
> that commit.
>
> I also noticed 2 new relocations R_AARCH64_ABS64 appearing in arm64 kernel.
>
> Those 2 weak undefined symbols have existed since commit
> 341dfcf8d78e ("btf: expose BTF info through sysfs") but this is the fact
> to declare those symbols into btf.c that produced those relocations.
>
> I'm not sure what this all means, but this is not something I expected
> for riscv for
> a kernel linked with -shared/-fpie. Maybe should we just leave them to
> zero ?
>
> I think that deserves a deeper look if someone understands all this
> better than I do.

Are you saying there is a warning for arm64 as well?
Can ppc folks explain the above warning?
What does it mean "2 bad relocations"?
The code is doing:
extern char __weak _binary__btf_vmlinux_bin_start[];
extern char __weak _binary__btf_vmlinux_bin_end[];
Since they are weak they should be zero when not defined.
What's the issue?

Re: linux-next: build warning after merge of the bpf-next tree

[Alexandre Ghiti]

On 1/10/20 6:18 PM, Alexei Starovoitov wrote:
> On Fri, Jan 10, 2020 at 2:28 PM Alexandre Ghiti <alexandre@ghiti.fr> wrote:
>> Hi guys,
>>
>> On 10/27/19 8:02 PM, Stephen Rothwell wrote:
>>> Hi all,
>>>
>>> On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>>>> Hi all,
>>>>
>>>> After merging the bpf-next tree, today's linux-next build (powerpc
>>>> ppc64_defconfig) produced this warning:
>>>>
>>>> WARNING: 2 bad relocations
>>>> c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
>>>> c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end
>>>>
>>>> Introduced by commit
>>>>
>>>>     8580ac9404f6 ("bpf: Process in-kernel BTF")
>>> This warning now appears in the net-next tree build.
>>>
>>>
>> I bump that thread up because Zong also noticed that 2 new relocations for
>> those symbols appeared in my riscv relocatable kernel branch following
>> that commit.
>>
>> I also noticed 2 new relocations R_AARCH64_ABS64 appearing in arm64 kernel.
>>
>> Those 2 weak undefined symbols have existed since commit
>> 341dfcf8d78e ("btf: expose BTF info through sysfs") but this is the fact
>> to declare those symbols into btf.c that produced those relocations.
>>
>> I'm not sure what this all means, but this is not something I expected
>> for riscv for
>> a kernel linked with -shared/-fpie. Maybe should we just leave them to
>> zero ?
>>
>> I think that deserves a deeper look if someone understands all this
>> better than I do.
> Are you saying there is a warning for arm64 as well?


Nop.


> Can ppc folks explain the above warning?
> What does it mean "2 bad relocations"?


This is what I'd like to understand too, it is not clear in
the ppc tool that outputs this message why it is considered 'bad'.


> The code is doing:
> extern char __weak _binary__btf_vmlinux_bin_start[];
> extern char __weak _binary__btf_vmlinux_bin_end[];
> Since they are weak they should be zero when not defined.
> What's the issue?


There likely is no issue, I just want to make sure those relocations
are legitimate and I want to understand what we should do with those.

At the moment arm64 does not relocate those at runtime and purely
ignore them: is this the right thing to do ?

Re: Re: linux-next: build warning after merge of the bpf-next tree

[Palmer Dabbelt]

On Fri, 10 Jan 2020 14:28:17 PST (-0800), alexandre@ghiti.fr wrote:
> Hi guys,
>
> On 10/27/19 8:02 PM, Stephen Rothwell wrote:
>> Hi all,
>>
>> On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>>> Hi all,
>>>
>>> After merging the bpf-next tree, today's linux-next build (powerpc
>>> ppc64_defconfig) produced this warning:
>>>
>>> WARNING: 2 bad relocations
>>> c000000001998a48 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_start
>>> c000000001998a50 R_PPC64_ADDR64    _binary__btf_vmlinux_bin_end
>>>
>>> Introduced by commit
>>>
>>>    8580ac9404f6 ("bpf: Process in-kernel BTF")
>> This warning now appears in the net-next tree build.
>>
>>
> I bump that thread up because Zong also noticed that 2 new relocations for
> those symbols appeared in my riscv relocatable kernel branch following
> that commit.
>
> I also noticed 2 new relocations R_AARCH64_ABS64 appearing in arm64 kernel.
>
> Those 2 weak undefined symbols have existed since commit
> 341dfcf8d78e ("btf: expose BTF info through sysfs") but this is the fact
> to declare those symbols into btf.c that produced those relocations.
>
> I'm not sure what this all means, but this is not something I expected
> for riscv for
> a kernel linked with -shared/-fpie. Maybe should we just leave them to
> zero ?
>
> I think that deserves a deeper look if someone understands all this
> better than I do.

Can you give me a pointer to your tree and how to build a relocatable kernel?
Weak undefined symbols have the absolute value 0, but the kernel is linked at
an address such that 0 can't be reached by normal means.  When I added support
to binutils for this I did it in a way that required almost no code --
essetially I just stopped dissallowing x0 as a possible base register for PCREL
relocations, which results in 0 always being accessible.  I just wanted to get
the kernel to build again, so I didn't worry about chasing around all the
addressing modes.  The PIC/PIE support generates different relocations and I
wouldn't be surprised if I just missed one (or more likely all) of them.

It's probably a simple fix, though I feel like every time I say that about the
linker I end up spending a month in there...

Re: linux-next: build warning after merge of the bpf-next tree

[Alexandre Ghiti]

On 1/10/20 7:20 PM, Palmer Dabbelt wrote:
> On Fri, 10 Jan 2020 14:28:17 PST (-0800), alexandre@ghiti.fr wrote:
>> Hi guys,
>>
>> On 10/27/19 8:02 PM, Stephen Rothwell wrote:
>>> Hi all,
>>>
>>> On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell 
>>> <sfr@canb.auug.org.au> wrote:
>>>> Hi all,
>>>>
>>>> After merging the bpf-next tree, today's linux-next build (powerpc
>>>> ppc64_defconfig) produced this warning:
>>>>
>>>> WARNING: 2 bad relocations
>>>> c000000001998a48 R_PPC64_ADDR64 _binary__btf_vmlinux_bin_start
>>>> c000000001998a50 R_PPC64_ADDR64 _binary__btf_vmlinux_bin_end
>>>>
>>>> Introduced by commit
>>>>
>>>>    8580ac9404f6 ("bpf: Process in-kernel BTF")
>>> This warning now appears in the net-next tree build.
>>>
>>>
>> I bump that thread up because Zong also noticed that 2 new 
>> relocations for
>> those symbols appeared in my riscv relocatable kernel branch following
>> that commit.
>>
>> I also noticed 2 new relocations R_AARCH64_ABS64 appearing in arm64 
>> kernel.
>>
>> Those 2 weak undefined symbols have existed since commit
>> 341dfcf8d78e ("btf: expose BTF info through sysfs") but this is the fact
>> to declare those symbols into btf.c that produced those relocations.
>>
>> I'm not sure what this all means, but this is not something I expected
>> for riscv for
>> a kernel linked with -shared/-fpie. Maybe should we just leave them to
>> zero ?
>>
>> I think that deserves a deeper look if someone understands all this
>> better than I do.
>
> Can you give me a pointer to your tree and how to build a relocatable 
> kernel?
> Weak undefined symbols have the absolute value 0,


So according to you the 2 new relocations R_RISCV_64 are normal and 
should not
be modified at runtime right ?


> but the kernel is linked at
> an address such that 0 can't be reached by normal means.  When I added 
> support
> to binutils for this I did it in a way that required almost no code --
> essetially I just stopped dissallowing x0 as a possible base register 
> for PCREL
> relocations, which results in 0 always being accessible.  I just 
> wanted to get
> the kernel to build again, so I didn't worry about chasing around all the
> addressing modes.  The PIC/PIE support generates different relocations 
> and I
> wouldn't be surprised if I just missed one (or more likely all) of them.
>
> It's probably a simple fix, though I feel like every time I say that 
> about the
> linker I end up spending a month in there...

You can find it here:

https://github.com/AlexGhiti/riscv-linux/tree/int/alex/riscv_relocatable_v1

Zong fixed the bug introduced by those 2 new relocations and everything 
works
like a charm, so I'm not sure you have to dig in the linker :)

Alex

Re: linux-next: build warning after merge of the bpf-next tree

[Zong Li]

On Sat, Jan 11, 2020 at 10:31 PM Alexandre Ghiti <alexandre@ghiti.fr> wrote:
>
>
> On 1/10/20 7:20 PM, Palmer Dabbelt wrote:
> > On Fri, 10 Jan 2020 14:28:17 PST (-0800), alexandre@ghiti.fr wrote:
> >> Hi guys,
> >>
> >> On 10/27/19 8:02 PM, Stephen Rothwell wrote:
> >>> Hi all,
> >>>
> >>> On Fri, 18 Oct 2019 10:56:57 +1100 Stephen Rothwell
> >>> <sfr@canb.auug.org.au> wrote:
> >>>> Hi all,
> >>>>
> >>>> After merging the bpf-next tree, today's linux-next build (powerpc
> >>>> ppc64_defconfig) produced this warning:
> >>>>
> >>>> WARNING: 2 bad relocations
> >>>> c000000001998a48 R_PPC64_ADDR64 _binary__btf_vmlinux_bin_start
> >>>> c000000001998a50 R_PPC64_ADDR64 _binary__btf_vmlinux_bin_end
> >>>>
> >>>> Introduced by commit
> >>>>
> >>>>    8580ac9404f6 ("bpf: Process in-kernel BTF")
> >>> This warning now appears in the net-next tree build.
> >>>
> >>>
> >> I bump that thread up because Zong also noticed that 2 new
> >> relocations for
> >> those symbols appeared in my riscv relocatable kernel branch following
> >> that commit.
> >>
> >> I also noticed 2 new relocations R_AARCH64_ABS64 appearing in arm64
> >> kernel.
> >>
> >> Those 2 weak undefined symbols have existed since commit
> >> 341dfcf8d78e ("btf: expose BTF info through sysfs") but this is the fact
> >> to declare those symbols into btf.c that produced those relocations.
> >>
> >> I'm not sure what this all means, but this is not something I expected
> >> for riscv for
> >> a kernel linked with -shared/-fpie. Maybe should we just leave them to
> >> zero ?
> >>
> >> I think that deserves a deeper look if someone understands all this
> >> better than I do.
> >
> > Can you give me a pointer to your tree and how to build a relocatable
> > kernel?
> > Weak undefined symbols have the absolute value 0,
>
>
> So according to you the 2 new relocations R_RISCV_64 are normal and
> should not
> be modified at runtime right ?
>
>
> > but the kernel is linked at
> > an address such that 0 can't be reached by normal means.  When I added
> > support
> > to binutils for this I did it in a way that required almost no code --
> > essetially I just stopped dissallowing x0 as a possible base register
> > for PCREL
> > relocations, which results in 0 always being accessible.  I just
> > wanted to get
> > the kernel to build again, so I didn't worry about chasing around all the
> > addressing modes.  The PIC/PIE support generates different relocations
> > and I
> > wouldn't be surprised if I just missed one (or more likely all) of them.
> >
> > It's probably a simple fix, though I feel like every time I say that
> > about the
> > linker I end up spending a month in there...
>
> You can find it here:
>
> https://github.com/AlexGhiti/riscv-linux/tree/int/alex/riscv_relocatable_v1
>
> Zong fixed the bug introduced by those 2 new relocations and everything
> works
> like a charm, so I'm not sure you have to dig in the linker :)
>

I'm not quite familiar with btf, so I have no idea why there are two
weak symbols be added in 8580ac9404f6 ("bpf: Process in-kernel BTF")
as well, According on relocation mechanism, maybe it is unnecessary to
handle weak undefined symbol at this time, because there is no
substantive help to relocate the absolute value 0. I just simply
ignore the non-relative relocation types to make processing can go
forward, and it works for me based on v5.5-rc5.

> Alex
>

Re: linux-next: build warning after merge of the bpf-next tree

[Alexei Starovoitov]

On Sun, Jan 12, 2020 at 8:33 PM Zong Li <zong.li@sifive.com> wrote:
>
> I'm not quite familiar with btf, so I have no idea why there are two
> weak symbols be added in 8580ac9404f6 ("bpf: Process in-kernel BTF")

I can explain what these weak symbols are for, but that won't change
the fact that compiler or linker are buggy. The weak symbols should work
in all cases and compiler should pick correct relocation.
In this case it sounds that compiler picked relative relocation and failed
to reach zero from that address.

Re: linux-next: build warning after merge of the bpf-next tree

[Alexandre Ghiti]

On 1/14/20 6:23 AM, Alexei Starovoitov wrote:
> On Sun, Jan 12, 2020 at 8:33 PM Zong Li<zong.li@sifive.com>  wrote:
>> I'm not quite familiar with btf, so I have no idea why there are two
>> weak symbols be added in 8580ac9404f6 ("bpf: Process in-kernel BTF")
> I can explain what these weak symbols are for, but that won't change
> the fact that compiler or linker are buggy. The weak symbols should work
> in all cases and compiler should pick correct relocation.
> In this case it sounds that compiler picked relative relocation and failed
> to reach zero from that address.

Sorry for the response delay: I now agree that there is nothing weird 
about those
relocations. All compiler/linker I took a look at (arm64, ppc64 and 
riscv64) correctly
emit an absolute relocation to the address 0 in case of a weak 
unresolved symbol,
so there's no buggy compiler/linker.

And regarding ppc warning, the kernel being compiled as -pie, the 
scripts looks
for absolute relocations which it considers as "bad", except for one 
that is known
to be weak and that is ignored: I have just sent a patch to fix this 
script so that weak
undefined symbol relocations are not considered as bad.

Thanks,

Alex