| Message ID | 20200923155436.2117661-1-andriin@fb.com |
|---|---|
| Headers | show |
| Series | libbpf: BTF writer APIs | expand |
Andrii Nakryiko wrote: > This patch set introduces a new set of BTF APIs to libbpf that allow to > conveniently produce BTF types and strings. Internals of struct btf were > changed such that it can transparently and automatically switch to writable > mode, which allows appending BTF types and strings. This will allow for libbpf > itself to do more intrusive modifications of program's BTF (by rewriting it, > at least as of right now), which is necessary for the upcoming libbpf static > linking. But they are complete and generic, so can be adopted by anyone who > has a need to produce BTF type information. I had this floating around on my todo list thanks a lot for doing it. I can remove a couple more silly hacks I have floating around now! > > One such example outside of libbpf is pahole, which was actually converted to > these APIs (locally, pending landing of these changes in libbpf) completely > and shows reduction in amount of custom pahole code necessary and brings nice > savings in memory usage (about 370MB reduction at peak for my kernel > configuration) and even BTF deduplication times (one second reduction, > 23.7s -> 22.7s). Memory savings are due to avoiding pahole's own copy of > "uncompressed" raw BTF data. Time reduction comes from faster string > search and deduplication by relying on hashmap instead of BST used by pahole's > own code. Consequently, these APIs are already tested on real-world > complicated kernel BTF, but there is also pretty extensive selftest doing > extra validations. > > Selftests in patch #9 add a set of generic ASSERT_{EQ,STREQ,ERR,OK} macros > that are useful for writing shorter and less repretitive selftests. I decided > to keep them local to that selftest for now, but if they prove to be useful in > more contexts we should move them to test_progs.h. And few more (e.g., > inequality tests) macros are probably necessary to have a more complete set. > > Cc: Arnaldo Carvalho de Melo <acme@redhat.com> For the series, I have a few nits I'll put in the patches, mostly spelling errors and a couple questions, otherwise this is awesome thanks. Acked-by: John Fastabend <john.fastabend@gmail.com> > > Andrii Nakryiko (9): > libbpf: refactor internals of BTF type index > libbpf: remove assumption of single contiguous memory for BTF data > libbpf: generalize common logic for managing dynamically-sized arrays > libbpf: extract generic string hashing function for reuse > libbpf: allow modification of BTF and add btf__add_str API > libbpf: add btf__new_empty() to create an empty BTF object > libbpf: add BTF writing APIs > libbpf: add btf__str_by_offset() as a more generic variant of > name_by_offset > selftests/bpf: test BTF writing APIs > > tools/lib/bpf/bpf.c | 2 +- > tools/lib/bpf/bpf.h | 2 +- > tools/lib/bpf/btf.c | 1311 +++++++++++++++-- > tools/lib/bpf/btf.h | 41 + > tools/lib/bpf/btf_dump.c | 9 +- > tools/lib/bpf/hashmap.h | 12 + > tools/lib/bpf/libbpf.map | 22 + > tools/lib/bpf/libbpf_internal.h | 3 + > .../selftests/bpf/prog_tests/btf_write.c | 271 ++++ > 9 files changed, 1553 insertions(+), 120 deletions(-) > create mode 100644 tools/testing/selftests/bpf/prog_tests/btf_write.c > > -- > 2.24.1 >
This patch set introduces a new set of BTF APIs to libbpf that allow to conveniently produce BTF types and strings. Internals of struct btf were changed such that it can transparently and automatically switch to writable mode, which allows appending BTF types and strings. This will allow for libbpf itself to do more intrusive modifications of program's BTF (by rewriting it, at least as of right now), which is necessary for the upcoming libbpf static linking. But they are complete and generic, so can be adopted by anyone who has a need to produce BTF type information. One such example outside of libbpf is pahole, which was actually converted to these APIs (locally, pending landing of these changes in libbpf) completely and shows reduction in amount of custom pahole code necessary and brings nice savings in memory usage (about 370MB reduction at peak for my kernel configuration) and even BTF deduplication times (one second reduction, 23.7s -> 22.7s). Memory savings are due to avoiding pahole's own copy of "uncompressed" raw BTF data. Time reduction comes from faster string search and deduplication by relying on hashmap instead of BST used by pahole's own code. Consequently, these APIs are already tested on real-world complicated kernel BTF, but there is also pretty extensive selftest doing extra validations. Selftests in patch #9 add a set of generic ASSERT_{EQ,STREQ,ERR,OK} macros that are useful for writing shorter and less repretitive selftests. I decided to keep them local to that selftest for now, but if they prove to be useful in more contexts we should move them to test_progs.h. And few more (e.g., inequality tests) macros are probably necessary to have a more complete set. Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Andrii Nakryiko (9): libbpf: refactor internals of BTF type index libbpf: remove assumption of single contiguous memory for BTF data libbpf: generalize common logic for managing dynamically-sized arrays libbpf: extract generic string hashing function for reuse libbpf: allow modification of BTF and add btf__add_str API libbpf: add btf__new_empty() to create an empty BTF object libbpf: add BTF writing APIs libbpf: add btf__str_by_offset() as a more generic variant of name_by_offset selftests/bpf: test BTF writing APIs tools/lib/bpf/bpf.c | 2 +- tools/lib/bpf/bpf.h | 2 +- tools/lib/bpf/btf.c | 1311 +++++++++++++++-- tools/lib/bpf/btf.h | 41 + tools/lib/bpf/btf_dump.c | 9 +- tools/lib/bpf/hashmap.h | 12 + tools/lib/bpf/libbpf.map | 22 + tools/lib/bpf/libbpf_internal.h | 3 + .../selftests/bpf/prog_tests/btf_write.c | 271 ++++ 9 files changed, 1553 insertions(+), 120 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/btf_write.c