On Fri, Sep 6, 2024 at 12:56 PM Eduard Zingerman <eddyz87@xxxxxxxxx> wrote: > > On Fri, 2024-09-06 at 14:24 +0100, Mykyta Yatsenko wrote: > > From: Mykyta Yatsenko <yatsenko@xxxxxxxx> > > > > Existing algorithm for BTF C dump sorting uses only types and names of > > the structs and unions for ordering. As dump contains structs with the > > same names but different contents, relative to each other ordering of > > those structs will be accidental. > > This patch addresses this problem by introducing a new sorting field > > that contains hash of the struct/union field names and types to > > disambiguate comparison of the non-unique named structs. > > > > Signed-off-by: Mykyta Yatsenko <yatsenko@xxxxxxxx> > > --- > > Note, this is still not fully stable, e.g.: > > $ for i in $(seq 1 10); \ > do touch ./kernel/bpf/verifier.c && \ > ccache-kernel-make.sh -j23 && \ > ./tools/bpf/bpftool/bpftool btf dump file vmlinux format c > ~/work/tmp/vmlinux.h.$i; \ > done > ... > $ md5sum ~/work/tmp/vmlinux.h.* | sort -k1 > 76c9b22274c4aa6253ffaafa33ceffd3 /home/eddy/work/tmp/vmlinux.h.2 > 76c9b22274c4aa6253ffaafa33ceffd3 /home/eddy/work/tmp/vmlinux.h.4 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.1 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.10 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.3 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.5 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.6 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.7 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.8 > a1c90a62e6cca59869a9cdffbaa3c4de /home/eddy/work/tmp/vmlinux.h.9 Still, quite stable, compared to what it is right now. I think the second part is more consistent ordering between CUs inside the pahole to keep it even more stable. Applied to bpf-next, thanks! > > [...] >
