Old binutils get confused about .buildid overlapping (in VA space) with
earlier section. That confusion results in weird errors down the road,
like this one:
objcopy: xen.efi: Data Directory size (1c) exceeds space left in section (8)
While the bug is fixed in later binutils version, force alignment of the
buildid to avoid overlapping and make it work with older versions too.
This can be reverted once toolchain base is raised at or above binutils
2.36.
Details at:
https://lore.kernel.org/all/3TMd7J2u5gCA8ouIG_Xfcw7s5JKMG06XsDIesEB3Fi9htUJ43Lfl057wXohlpCHcszqoCmicpIlneEDO26ZqT8QfC2Y39VxBuqD3nS1j5Q4=@trmm.net/
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Fixes: eee5909e9d1e ("x86/EFI: use less crude a way of generating the build ID")
---
While the actual bug apparently is in binutils, the commit mentioned in
Fixes: tag introduced part that triggers that bug.
---
xen/arch/x86/xen.lds.S | 2 ++
1 file changed, 2 insertions(+)
diff --git a/xen/arch/x86/xen.lds.S b/xen/arch/x86/xen.lds.S
index 5d02f83a409c..967f9167f757 100644
--- a/xen/arch/x86/xen.lds.S
+++ b/xen/arch/x86/xen.lds.S
@@ -171,6 +171,8 @@ SECTIONS
__note_gnu_build_id_end = .;
} PHDR(note) PHDR(text)
#elif defined(BUILD_ID_EFI)
+ /* Workaround bug in binutils < 2.38 */
+ . = ALIGN(32);
DECL_SECTION(.buildid) {
__note_gnu_build_id_start = .;
*(.buildid)
--
2.51.0
On 19/11/2025 7:12 pm, Marek Marczykowski-Górecki wrote:
> Old binutils get confused about .buildid overlapping (in VA space) with
> earlier section. That confusion results in weird errors down the road,
> like this one:
>
> objcopy: xen.efi: Data Directory size (1c) exceeds space left in section (8)
>
> While the bug is fixed in later binutils version, force alignment of the
> buildid to avoid overlapping and make it work with older versions too.
> This can be reverted once toolchain base is raised at or above binutils
> 2.36.
>
> Details at:
> https://lore.kernel.org/all/3TMd7J2u5gCA8ouIG_Xfcw7s5JKMG06XsDIesEB3Fi9htUJ43Lfl057wXohlpCHcszqoCmicpIlneEDO26ZqT8QfC2Y39VxBuqD3nS1j5Q4=@trmm.net/
>
> Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
> Fixes: eee5909e9d1e ("x86/EFI: use less crude a way of generating the build ID")
We typically put these tags in the other order. I can fix on commit.
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>, although...
> ---
> While the actual bug apparently is in binutils, the commit mentioned in
> Fixes: tag introduced part that triggers that bug.
> ---
> xen/arch/x86/xen.lds.S | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/xen/arch/x86/xen.lds.S b/xen/arch/x86/xen.lds.S
> index 5d02f83a409c..967f9167f757 100644
> --- a/xen/arch/x86/xen.lds.S
> +++ b/xen/arch/x86/xen.lds.S
> @@ -171,6 +171,8 @@ SECTIONS
> __note_gnu_build_id_end = .;
> } PHDR(note) PHDR(text)
> #elif defined(BUILD_ID_EFI)
> + /* Workaround bug in binutils < 2.38 */
... this says 2.38, but the commit message says 2.36. Which one is it?
~Andrew
> + . = ALIGN(32);
> DECL_SECTION(.buildid) {
> __note_gnu_build_id_start = .;
> *(.buildid)
On Thu, Nov 20, 2025 at 12:43:57AM +0000, Andrew Cooper wrote:
> On 19/11/2025 7:12 pm, Marek Marczykowski-Górecki wrote:
> > Old binutils get confused about .buildid overlapping (in VA space) with
> > earlier section. That confusion results in weird errors down the road,
> > like this one:
> >
> > objcopy: xen.efi: Data Directory size (1c) exceeds space left in section (8)
> >
> > While the bug is fixed in later binutils version, force alignment of the
> > buildid to avoid overlapping and make it work with older versions too.
> > This can be reverted once toolchain base is raised at or above binutils
> > 2.36.
> >
> > Details at:
> > https://lore.kernel.org/all/3TMd7J2u5gCA8ouIG_Xfcw7s5JKMG06XsDIesEB3Fi9htUJ43Lfl057wXohlpCHcszqoCmicpIlneEDO26ZqT8QfC2Y39VxBuqD3nS1j5Q4=@trmm.net/
> >
> > Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
> > Fixes: eee5909e9d1e ("x86/EFI: use less crude a way of generating the build ID")
>
> We typically put these tags in the other order. I can fix on commit.
>
> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>, although...
>
> > ---
> > While the actual bug apparently is in binutils, the commit mentioned in
> > Fixes: tag introduced part that triggers that bug.
> > ---
> > xen/arch/x86/xen.lds.S | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/xen/arch/x86/xen.lds.S b/xen/arch/x86/xen.lds.S
> > index 5d02f83a409c..967f9167f757 100644
> > --- a/xen/arch/x86/xen.lds.S
> > +++ b/xen/arch/x86/xen.lds.S
> > @@ -171,6 +171,8 @@ SECTIONS
> > __note_gnu_build_id_end = .;
> > } PHDR(note) PHDR(text)
> > #elif defined(BUILD_ID_EFI)
> > + /* Workaround bug in binutils < 2.38 */
>
> ... this says 2.38, but the commit message says 2.36. Which one is it?
2.36
> ~Andrew
>
> > + . = ALIGN(32);
> > DECL_SECTION(.buildid) {
> > __note_gnu_build_id_start = .;
> > *(.buildid)
>
--
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
On 20/11/2025 1:33 am, Marek Marczykowski-Górecki wrote:
> On Thu, Nov 20, 2025 at 12:43:57AM +0000, Andrew Cooper wrote:
>> On 19/11/2025 7:12 pm, Marek Marczykowski-Górecki wrote:
>>> Old binutils get confused about .buildid overlapping (in VA space) with
>>> earlier section. That confusion results in weird errors down the road,
>>> like this one:
>>>
>>> objcopy: xen.efi: Data Directory size (1c) exceeds space left in section (8)
>>>
>>> While the bug is fixed in later binutils version, force alignment of the
>>> buildid to avoid overlapping and make it work with older versions too.
>>> This can be reverted once toolchain base is raised at or above binutils
>>> 2.36.
>>>
>>> Details at:
>>> https://lore.kernel.org/all/3TMd7J2u5gCA8ouIG_Xfcw7s5JKMG06XsDIesEB3Fi9htUJ43Lfl057wXohlpCHcszqoCmicpIlneEDO26ZqT8QfC2Y39VxBuqD3nS1j5Q4=@trmm.net/
>>>
>>> Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
>>> Fixes: eee5909e9d1e ("x86/EFI: use less crude a way of generating the build ID")
>> We typically put these tags in the other order. I can fix on commit.
>>
>> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>, although...
>>
>>> ---
>>> While the actual bug apparently is in binutils, the commit mentioned in
>>> Fixes: tag introduced part that triggers that bug.
>>> ---
>>> xen/arch/x86/xen.lds.S | 2 ++
>>> 1 file changed, 2 insertions(+)
>>>
>>> diff --git a/xen/arch/x86/xen.lds.S b/xen/arch/x86/xen.lds.S
>>> index 5d02f83a409c..967f9167f757 100644
>>> --- a/xen/arch/x86/xen.lds.S
>>> +++ b/xen/arch/x86/xen.lds.S
>>> @@ -171,6 +171,8 @@ SECTIONS
>>> __note_gnu_build_id_end = .;
>>> } PHDR(note) PHDR(text)
>>> #elif defined(BUILD_ID_EFI)
>>> + /* Workaround bug in binutils < 2.38 */
>> ... this says 2.38, but the commit message says 2.36. Which one is it?
> 2.36
Thanks. I'll adjust on commit.
~Andrew
© 2016 - 2025 Red Hat, Inc.