On Wed, Apr 18, 2018 at 09:47:45PM -0700, Alexei Starovoitov wrote:
> On Thu, Apr 19, 2018 at 09:34:05AM +0800, Leo Yan wrote:
> > The code defines macro 'PAGE_OFFSET' and uses it to decide if the
> > address is in kernel space or not. But different architecture has
> > different 'PAGE_OFFSET' so this program cannot be used for all
> > platforms.
> >
> > This commit changes to check returned pointer from ksym_search() to
> > judge if the address falls into kernel space or not, and removes
> > macro 'PAGE_OFFSET' as it isn't used anymore. As result, this program
> > has no architecture dependency.
> >
> > Signed-off-by: Leo Yan <leo....@linaro.org>
> > ---
> > samples/bpf/sampleip_user.c | 8 +++-----
> > 1 file changed, 3 insertions(+), 5 deletions(-)
> >
> > diff --git a/samples/bpf/sampleip_user.c b/samples/bpf/sampleip_user.c
> > index 4ed690b..0eea1b3 100644
> > --- a/samples/bpf/sampleip_user.c
> > +++ b/samples/bpf/sampleip_user.c
> > @@ -26,7 +26,6 @@
> > #define DEFAULT_FREQ 99
> > #define DEFAULT_SECS 5
> > #define MAX_IPS 8192
> > -#define PAGE_OFFSET 0xffff880000000000
> >
> > static int nr_cpus;
> >
> > @@ -107,14 +106,13 @@ static void print_ip_map(int fd)
> > /* sort and print */
> > qsort(counts, max, sizeof(struct ipcount), count_cmp);
> > for (i = 0; i < max; i++) {
> > - if (counts[i].ip > PAGE_OFFSET) {
> > - sym = ksym_search(counts[i].ip);
>
> yes. it is x64 specific, since it's a sample code,
> but simply removing it is not a fix.
> It makes this sampleip code behaving incorrectly.
> To do such 'cleanup of ksym' please refactor it in the true generic way,
> so these ksym* helpers can work on all archs and put this new
> functionality into selftests.
Just want to check, are you suggesting to create a standalone
testing for kallsyms (like a folder tools/testing/selftests/ksym) or
do you mean to place the generic code into folder
tools/testing/selftests/bpf/?
> If you can convert these examples into proper self-checking tests
> that run out of selftests that would be awesome.
Thank you for suggestions, Alexei.
> Thanks!
>
