On Tue, Apr 28, 2020 at 11:35:15AM -0700, Andrii Nakryiko wrote: > On Tue, Apr 28, 2020 at 9:44 AM Alexei Starovoitov > <alexei.starovoitov@xxxxxxxxx> wrote: > > > > On Mon, Apr 27, 2020 at 09:46:24PM -0700, Andrii Nakryiko wrote: > > > Add another flavor of test_progs that is compiled and run with > > > AddressSanitizer and LeakSanitizer. This allows to find potential memory > > > correction bugs and memory leaks. Due to sometimes not trivial requirements on > > > the environment, this is (for now) done as a separate flavor, not by default. > > > Eventually I hope to enable it by default. > > > > > > To run ./test_progs-asan successfully, you need to have libasan installed in > > > the system, where version of the package depends on GCC version you have. > > > E.g., GCC8 needs libasan5, while GCC7 uses libasan4. > > > > > > For CentOS 7, to build everything successfully one would need to: > > > $ sudo yum install devtoolset-8-gcc devtoolset-libasan-devel > > > > > > For Arch Linux to run selftests, one would need to install gcc-libs package to > > > get libasan.so.5: > > > $ sudo pacman -S gcc-libs > > > > > > Cc: Julia Kartseva <hex@xxxxxx> > > > Signed-off-by: Andrii Nakryiko <andriin@xxxxxx> > > > > It needs a feature check. > > selftest shouldn't be forcing asan on everyone. > > Even after I did: > > sudo yum install devtoolset-8-libasan-devel > > it still failed to build: > > BINARY test_progs-asan > > /opt/rh/devtoolset-9/root/usr/libexec/gcc/x86_64-redhat-linux/9/ld: cannot find libasan_preinit.o: No such file or directory > > /opt/rh/devtoolset-9/root/usr/libexec/gcc/x86_64-redhat-linux/9/ld: cannot find -lasan > > > > Yeah, it worked for me initially because it still used GCC7 locally > and older version of libasan. > > On CentOS you have to run the following command to set up environment > (for current session only, though): > > $ scl enable devtoolset-8 bash > > What it does: > - adds /opt/rh/devtoolset-8/root/usr/bin to $PATH > - sets $LD_LIBRARY_PATH to > /opt/rh/devtoolset-8/root/usr/lib64:/opt/rh/devtoolset-8/root/usr/lib:/opt/rh/devtoolset-8/root/usr/lib64/dyninst:/opt/rh/devtoolset-8/root/usr/lib/dyninst:/opt/rh/devtoolset-8/root/usr/lib64:/opt/rh/devtoolset-8/root/usr/lib I don't want to do this, since I prefer gcc9 for my builds since it has better warnings. But yum cannot find devtoolset-9-libasan-devel it seems. > I'm going to add this to patch to ease some pain later. But yeah, I > think I have a better plan for ASAN builds. I'll add EXTRA_CFLAGS to > selftests Makefile, defaulted to nothing. Then for Travis CI (or > locally) one would do: > > $ make EXTRA_CFLAGS='-fsanitize-address' > > to build ASAN versions of all the same test runners (including > test_verifier, test_maps, etc). > > I think this will be better overall. > > > Also I really don't like that skeletons are now built three times for now good reason > > GEN-SKEL [test_progs-asan] test_stack_map.skel.h > > GEN-SKEL [test_progs-asan] test_core_reloc_nesting.skel.h > > default vs no_alu32 makes sense. They are different bpf.o files and different skeletons, > > but for asan there is no such need. > > I agree, luckily I don't really have to change anything with the above approach. > > > > > Please resubmit the rest of the patches, since asan isn't a prerequisite. > > I'll update this patch to just add EXTRA_CFLAGS, if you are ok with > this (and will leave instructions on installing libasan). yeah. EXTRA_CFLAGS approach should work. That will build both test_progs and test_progs-no_alu32 with libasan ? That would be ideal.