| Message ID | 20220731232649.4668-1-james.hilliard1@gmail.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 47ea7417b0744324424405fc1207e266053237a9 |
| Delegated to: | BPF |
| Headers | show |
| Series | libbpf: skip empty sections in bpf_object__init_global_data_maps | expand |
| Context | Check | Description |
|---|---|---|
| netdev/tree_selection | success | Not a local patch |
| bpf/vmtest-bpf-next-PR | success | PR summary |
| bpf/vmtest-bpf-next-VM_Test-3 | success | Logs for Kernel LATEST on z15 with gcc |
| bpf/vmtest-bpf-next-VM_Test-1 | success | Logs for Kernel LATEST on ubuntu-latest with gcc |
| bpf/vmtest-bpf-next-VM_Test-2 | success | Logs for Kernel LATEST on ubuntu-latest with llvm-16 |
On Sun, Jul 31, 2022 at 05:26:49PM -0600, James Hilliard wrote: > The GNU assembler generates an empty .bss section. This is a well > established behavior in GAS that happens in all supported targets. > > The LLVM assembler doesn't generate an empty .bss section. > > bpftool chokes on the empty .bss section. > > Additionally in bpf_object__elf_collect the sec_desc->data is not > initialized when a section is not recognized. In this case, this > happens with .comment. > > So we must check that sec_desc->data is initialized before checking > if the size is 0. oops David send same change but I asked him to move the check to bpf_object__elf_collect [1] .. but with your explanation this fix actualy looks fine to me jirka [1] https://lore.kernel.org/bpf/YuKaFiZ+ksB5f0Ye@krava/ > > Signed-off-by: James Hilliard <james.hilliard1@gmail.com> > Cc: Jose E. Marchesi <jose.marchesi@oracle.com> > --- > tools/lib/bpf/libbpf.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c > index 50d41815f431..77e3797cf75a 100644 > --- a/tools/lib/bpf/libbpf.c > +++ b/tools/lib/bpf/libbpf.c > @@ -1642,6 +1642,10 @@ static int bpf_object__init_global_data_maps(struct bpf_object *obj) > for (sec_idx = 1; sec_idx < obj->efile.sec_cnt; sec_idx++) { > sec_desc = &obj->efile.secs[sec_idx]; > > + /* Skip recognized sections with size 0. */ > + if (sec_desc->data && sec_desc->data->d_size == 0) > + continue; > + > switch (sec_desc->sec_type) { > case SEC_DATA: > sec_name = elf_sec_name(obj, elf_sec_by_idx(obj, sec_idx)); > -- > 2.34.1 >
On 8/1/22 13:24, Jiri Olsa wrote: > On Sun, Jul 31, 2022 at 05:26:49PM -0600, James Hilliard wrote: >> The GNU assembler generates an empty .bss section. This is a well >> established behavior in GAS that happens in all supported targets. >> >> The LLVM assembler doesn't generate an empty .bss section. >> >> bpftool chokes on the empty .bss section. >> >> Additionally in bpf_object__elf_collect the sec_desc->data is not >> initialized when a section is not recognized. In this case, this >> happens with .comment. >> >> So we must check that sec_desc->data is initialized before checking >> if the size is 0. > > oops David send same change but I asked him to move the check > to bpf_object__elf_collect [1] .. but with your explanation this > fix actualy looks fine to me FWIW, I only just got back to actually making that change. This patch has a much better explanation than the one I sent so +1 from me also David > > jirka > > > [1] https://lore.kernel.org/bpf/YuKaFiZ+ksB5f0Ye@krava/ > >> >> Signed-off-by: James Hilliard <james.hilliard1@gmail.com> >> Cc: Jose E. Marchesi <jose.marchesi@oracle.com> >> --- >> tools/lib/bpf/libbpf.c | 4 ++++ >> 1 file changed, 4 insertions(+) >> >> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c >> index 50d41815f431..77e3797cf75a 100644 >> --- a/tools/lib/bpf/libbpf.c >> +++ b/tools/lib/bpf/libbpf.c >> @@ -1642,6 +1642,10 @@ static int bpf_object__init_global_data_maps(struct bpf_object *obj) >> for (sec_idx = 1; sec_idx < obj->efile.sec_cnt; sec_idx++) { >> sec_desc = &obj->efile.secs[sec_idx]; >> >> + /* Skip recognized sections with size 0. */ >> + if (sec_desc->data && sec_desc->data->d_size == 0) >> + continue; >> + >> switch (sec_desc->sec_type) { >> case SEC_DATA: >> sec_name = elf_sec_name(obj, elf_sec_by_idx(obj, sec_idx)); >> -- >> 2.34.1 >>
On Mon, Aug 01, 2022 at 03:21:58PM -0700, David Faust wrote: > > > On 8/1/22 13:24, Jiri Olsa wrote: > > On Sun, Jul 31, 2022 at 05:26:49PM -0600, James Hilliard wrote: > >> The GNU assembler generates an empty .bss section. This is a well > >> established behavior in GAS that happens in all supported targets. > >> > >> The LLVM assembler doesn't generate an empty .bss section. > >> > >> bpftool chokes on the empty .bss section. > >> > >> Additionally in bpf_object__elf_collect the sec_desc->data is not > >> initialized when a section is not recognized. In this case, this > >> happens with .comment. > >> > >> So we must check that sec_desc->data is initialized before checking > >> if the size is 0. > > > > oops David send same change but I asked him to move the check > > to bpf_object__elf_collect [1] .. but with your explanation this > > fix actualy looks fine to me > > FWIW, I only just got back to actually making that change. This > patch has a much better explanation than the one I sent so +1 from > me also thanks, I'm acking this one then Acked-by: Jiri Olsa <jolsa@kernel.org> jirka > > David > > > > > jirka > > > > > > [1] https://lore.kernel.org/bpf/YuKaFiZ+ksB5f0Ye@krava/ > > > >> > >> Signed-off-by: James Hilliard <james.hilliard1@gmail.com> > >> Cc: Jose E. Marchesi <jose.marchesi@oracle.com> > >> --- > >> tools/lib/bpf/libbpf.c | 4 ++++ > >> 1 file changed, 4 insertions(+) > >> > >> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c > >> index 50d41815f431..77e3797cf75a 100644 > >> --- a/tools/lib/bpf/libbpf.c > >> +++ b/tools/lib/bpf/libbpf.c > >> @@ -1642,6 +1642,10 @@ static int bpf_object__init_global_data_maps(struct bpf_object *obj) > >> for (sec_idx = 1; sec_idx < obj->efile.sec_cnt; sec_idx++) { > >> sec_desc = &obj->efile.secs[sec_idx]; > >> > >> + /* Skip recognized sections with size 0. */ > >> + if (sec_desc->data && sec_desc->data->d_size == 0) > >> + continue; > >> + > >> switch (sec_desc->sec_type) { > >> case SEC_DATA: > >> sec_name = elf_sec_name(obj, elf_sec_by_idx(obj, sec_idx)); > >> -- > >> 2.34.1 > >>
Hello: This patch was applied to bpf/bpf-next.git (master) by Andrii Nakryiko <andrii@kernel.org>: On Sun, 31 Jul 2022 17:26:49 -0600 you wrote: > The GNU assembler generates an empty .bss section. This is a well > established behavior in GAS that happens in all supported targets. > > The LLVM assembler doesn't generate an empty .bss section. > > bpftool chokes on the empty .bss section. > > [...] Here is the summary with links: - libbpf: skip empty sections in bpf_object__init_global_data_maps https://git.kernel.org/bpf/bpf-next/c/47ea7417b074 You are awesome, thank you!
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c index 50d41815f431..77e3797cf75a 100644 --- a/tools/lib/bpf/libbpf.c +++ b/tools/lib/bpf/libbpf.c @@ -1642,6 +1642,10 @@ static int bpf_object__init_global_data_maps(struct bpf_object *obj) for (sec_idx = 1; sec_idx < obj->efile.sec_cnt; sec_idx++) { sec_desc = &obj->efile.secs[sec_idx]; + /* Skip recognized sections with size 0. */ + if (sec_desc->data && sec_desc->data->d_size == 0) + continue; + switch (sec_desc->sec_type) { case SEC_DATA: sec_name = elf_sec_name(obj, elf_sec_by_idx(obj, sec_idx));
The GNU assembler generates an empty .bss section. This is a well established behavior in GAS that happens in all supported targets. The LLVM assembler doesn't generate an empty .bss section. bpftool chokes on the empty .bss section. Additionally in bpf_object__elf_collect the sec_desc->data is not initialized when a section is not recognized. In this case, this happens with .comment. So we must check that sec_desc->data is initialized before checking if the size is 0. Signed-off-by: James Hilliard <james.hilliard1@gmail.com> Cc: Jose E. Marchesi <jose.marchesi@oracle.com> --- tools/lib/bpf/libbpf.c | 4 ++++ 1 file changed, 4 insertions(+)