On Wed, Jul 20, 2022 at 12:48 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote:
>
> On 19/07/2022 18:51, Alexei Starovoitov wrote:
> > On Tue, Jul 19, 2022 at 10:19 AM Martin KaFai Lau <kafai@xxxxxx> wrote:
> >>
> >> On Tue, Jul 19, 2022 at 12:53:01PM +0300, Dan Carpenter wrote:
> >>> The return from strcmp() is inverted so the it returns true instead
> >>> of false and vise versa.
> >>>
> >>> Fixes: a1c9d61b19cb ("libbpf: Improve library identification for uprobe binary path resolution")
> >>> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> >>> ---
> >>> Spotted during review. *cmp() functions should always have a comparison
> >>> to zero.
> >>> if (strcmp(a, b) < 0) { <-- means a < b
> >>> if (strcmp(a, b) >= 0) { <-- means a >= b
> >>> if (strcmp(a, b) != 0) { <-- means a != b
> >>> etc.
> >>>
> >>> tools/lib/bpf/libbpf_internal.h | 6 +++---
> >>> 1 file changed, 3 insertions(+), 3 deletions(-)
> >>>
> >>> diff --git a/tools/lib/bpf/libbpf_internal.h b/tools/lib/bpf/libbpf_internal.h
> >>> index 9cd7829cbe41..008485296a29 100644
> >>> --- a/tools/lib/bpf/libbpf_internal.h
> >>> +++ b/tools/lib/bpf/libbpf_internal.h
> >>> @@ -108,9 +108,9 @@ static inline bool str_has_sfx(const char *str, const char *sfx)
> >>> size_t str_len = strlen(str);
> >>> size_t sfx_len = strlen(sfx);
> >>>
> >>> - if (sfx_len <= str_len)
> >>> - return strcmp(str + str_len - sfx_len, sfx);
> >>> - return false;
> >>> + if (sfx_len > str_len)
> >>> + return false;
> >>> + return strcmp(str + str_len - sfx_len, sfx) == 0;
> >> Please tag the subject with "bpf" next time.
> >>
> >> Acked-by: Martin KaFai Lau <kafai@xxxxxx>
> >
> > Alan,
> >
> > If it was so broken how did it work earlier?
> > Do we have a test for this?
> >
>
> We have tests for automatic path determination, yes, but those
> tests specify libc.so.6, so are testing the strstr(file, ".so.")
> predicate below:
>
> if (str_has_sfx(file, ".so") || strstr(file, ".so.")) {
>
> Problem is, on many systems, libc.so is a GNU ld script rather
> than an actual library:
>
> cat /usr/lib64/libc.so
> /* GNU ld script
> Use the shared library, but some functions are only in
> the static library, so try that secondarily. */
> OUTPUT_FORMAT(elf64-x86-64)
> GROUP ( /lib64/libc.so.6 /usr/lib64/libc_nonshared.a AS_NEEDED ( /lib64/ld-linux-x86-64.so.2 ) )
>
> ...so we can't rely on system library .so files actually containing
> the library to instrument.
>
> Maybe we can do something with liburandom_read.so now we have it
> there; I was looking at extending the auto-path determination
> to usdt too, so we could add a new test to cover this then I think.
Library path resolution should already work for USDTs (note that I
reuse resolve_full_path() in bpf_program__attach_usdt()), but having
explicit tests would be great. It might be simplest to temporarily
override LD_LIBRARY_PATH with a path to liburandom_read.so? So please
consider sending a patch with tests, thanks!
>
> Alan
