Re: [PATCH dwarves v2 2/4] btf_encoder: Manually lay out updated ELF sections

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi.

On Thu, 4 Feb 2021 at 04:13, Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote:
>
> On Mon, Feb 1, 2021 at 9:26 AM Giuliano Procida <gprocida@xxxxxxxxxx> wrote:
> >
> > pahole -J needs to do the following to an ELF file:
> >
> > * add or update the ".BTF" section
> > * maybe update the section name string table
> > * update the Section Header Table (SHT)
> >
> > libelf either takes full control of layout or requires the user to
> > specify offset, size and alignment of all new and updated sections and
> > headers.
> >
> > To avoid libelf moving program segments in particular, we position the
>
> It's not clear to me what's wrong with libelf handling all the layout.
> Even if libelf will move program segments around, what's the harm?
> Does it break anything if we just let libelf do this?
>

It doesn't hurt the userspace case I care about. I've no idea what it
means in terms of vmlinux.

However, I wrote that text before I discovered that pahole -J isn't
actually used to modify kernel images.

One thing I haven't tried is to try to make .BTF loadable but leave
placement to libelf.

> > ".BTF" and section name string table (typically named ".shstrtab")
> > sections after all others. The SHT always lives at the end of the file.
> >
> > Note that the last section in an ELF file is normally the section name
> > string table and any ".BTF" section will normally be second last.
> > However, if these sections appear earlier, then we'll waste some space
> > in the ELF file when we rewrite them.
> >
> > Signed-off-by: Giuliano Procida <gprocida@xxxxxxxxxx>
> > ---
> >  libbtf.c | 64 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
> >  1 file changed, 62 insertions(+), 2 deletions(-)
> >
>
> [...]



[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux