Re: [PATCH v2 bpf-next] kbuild,bpf: switch to using --btf_features for pahole v1.26 and later

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

 



On Sat, May 11, 2024 at 3:01 AM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
>
> On Sat, May 11, 2024 at 6:45 AM Andrii Nakryiko
> <andrii.nakryiko@xxxxxxxxx> wrote:
> >
> > On Thu, May 9, 2024 at 11:30 PM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
> > >
> > > On Fri, May 10, 2024 at 7:01 AM Andrii Nakryiko
> > > <andrii.nakryiko@xxxxxxxxx> wrote:
> > > >
> > > > On Thu, May 9, 2024 at 1:20 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote:
> > > > >
> > > > > On 07/05/2024 17:48, Andrii Nakryiko wrote:
> > > > > > On Tue, May 7, 2024 at 6:55 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote:
> > > > > >>
> > > > > >> The btf_features list can be used for pahole v1.26 and later -
> > > > > >> it is useful because if a feature is not yet implemented it will
> > > > > >> not exit with a failure message.  This will allow us to add feature
> > > > > >> requests to the pahole options without having to check pahole versions
> > > > > >> in future; if the version of pahole supports the feature it will be
> > > > > >> added.
> > > > > >>
> > > > > >> Signed-off-by: Alan Maguire <alan.maguire@xxxxxxxxxx>
> > > > > >> Tested-by: Eduard Zingerman <eddyz87@xxxxxxxxx>
> > > > > >> ---
> > > > > >>  scripts/Makefile.btf | 15 +++++++++++++--
> > > > > >>  1 file changed, 13 insertions(+), 2 deletions(-)
> > > > > >>
> > > > > >> diff --git a/scripts/Makefile.btf b/scripts/Makefile.btf
> > > > > >> index 82377e470aed..2d6e5ed9081e 100644
> > > > > >> --- a/scripts/Makefile.btf
> > > > > >> +++ b/scripts/Makefile.btf
> > > > > >> @@ -3,6 +3,8 @@
> > > > > >>  pahole-ver := $(CONFIG_PAHOLE_VERSION)
> > > > > >>  pahole-flags-y :=
> > > > > >>
> > > > > >> +ifeq ($(call test-le, $(pahole-ver), 125),y)
> > > > > >> +
> > > > > >>  # pahole 1.18 through 1.21 can't handle zero-sized per-CPU vars
> > > > > >>  ifeq ($(call test-le, $(pahole-ver), 121),y)
> > > > > >>  pahole-flags-$(call test-ge, $(pahole-ver), 118)       += --skip_encoding_btf_vars
> > > > > >> @@ -12,8 +14,17 @@ pahole-flags-$(call test-ge, $(pahole-ver), 121)     += --btf_gen_floats
> > > > > >>
> > > > > >>  pahole-flags-$(call test-ge, $(pahole-ver), 122)       += -j
> > > > > >>
> > > > > >> -pahole-flags-$(CONFIG_PAHOLE_HAS_LANG_EXCLUDE)         += --lang_exclude=rust
> > > > > >> +ifeq ($(pahole-ver), 125)
> > > > > >
> > > > > > it's a bit of a scope creep, but isn't it strange that we don't have
> > > > > > test-eq and have to work-around that with more verbose constructs?
> > > > >
> > > > > Looking at the history, I _think_ the concern that motivated the numeric
> > > > > comparison constructs was the shell process fork required for numeric
> > > > > comparisons. In the equality case, ifeq would work for both strings and
> > > > > numeric values. Adding a test-eq (in a similar form to test-ge) would
> > > > > require a fallback to shell expansion for older Make without intcmp, and
> > > > > that would be slower than using ifeq, if less verbose.
> > > > >
> > > > > > Let's do a good service to the community and add test-eq (and maybe
> > > > > > test-ne while at it, don't know, up to Masahiro)?
> > > > > >
> > > > >
> > > > > Sure, I'm happy to do this if kbuild folks agree. I've cc'ed them; I
> > > > > neglected to do this in the original patch, apologies about that.
> > > > >
> > > >
> > > > Ok, let's see if Masahiro would like this improvement or not. For now
> > > > this patch gets us into a nicer place where there are legacy parts and
> > > > a better --btf_features setup completely separate, so I applied the
> > > > patch as is to bpf-next. If we decide to do test-eq, we can improve
> > > > this further separately. Thanks!
> > >
> > >
> > > That is a noise change.
> > > You did not need to modify the line in the first place.
> > >
> >
> > Not sure which specific line you are referring to. But the idea here
> > is that starting from pahole 1.26 we want to stop to set those
> > "legacy" arguments (like --skip_encoding_btf_vars, --btf_gen_floats)
> > and *only* use more usable and forward-compatible --btf_features.
> >
> > >
> > > The previous
> > >
> > >   pahole-flags-$(call test-ge, $(pahole-ver), 125)
> > >
> > > works as-is.
>
>
> You did not not need to change
>
>   pahole-flags-$(call test-ge, $(pahole-ver), 125) += ...
>
>
> to
>
>
>   ifeq ($(pahole-ver), 125)
>   pahole-flags-y += ...
>   endif
>
>
>
> Please note it exists in
>
>   ifeq ($(call test-le, $(pahole-ver), 125),y)
>      ...
>   else
>
>
>
>
>
> if (pahole_ver <= 125) {
>       do_something();
>       if (pahole_ver >= 125)
>              do_other();
> }
>
>
>   and
>
>
> if (pahole_ver <= 125) {
>       do_something();
>       if (pahole_ver == 125)
>             do_other();
> }
>
>
> are equivalent, don't they?
>
>
>
> The former is more intuitive because pahole 1.25+ supports
> --skip_encoding_btf_inconsistent_proto --btf_gen_optimized

The point here is to not specify these "legacy" arguments starting
from pahole v1.26, and the patch makes it more obvious, IMO.

But I don't mind adding (test-ge,125) check back, Alan, please send a
follow up patch.

>
>
>
> I attached a simpler and more correct patch.

It's not more correct, because this patch didn't break anything. It's
just as correct.

>
>
>
>
>
>
>
>
>
>
> --
> Best Regards
> Masahiro Yamada





[Index of Archives]     [Linux&nblp;USB Development]     [Linux Media]     [Video for Linux]     [Linux Audio Users]     [Yosemite Secrets]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux