Skip to content

L
linux
Switch branch/tag
  1. 02 Mar, 2022 3 commits
  3. 01 Mar, 2022 2 commits
  5. 28 Feb, 2022 11 commits
  7. 25 Feb, 2022 1 commit
    • Yucong Sun's avatar
      bpf: Fix issue with bpf preload module taking over stdout/stdin of kernel. · 80bebebd
      Yucong Sun authored 
      In cb80ddc6 ("bpf: Convert bpf_preload.ko to use light skeleton.")
BPF preload was switched from user mode process to use in-kernel light
skeleton instead. However, in the kernel context, early in the boot
sequence, the first available FD can start from 0, instead of normally
3 for user mode process. So FDs 0 and 1 are then used for loaded BPF
programs and prevent init process from setting up stdin/stdout/stderr on
FD 0, 1, and 2 as expected.

Before the fix:

ls -lah /proc/1/fd/*

lrwx------1 root root 64 Feb 23 17:20 /proc/1/fd/0 -> /dev/null
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/1 -> /dev/null
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/2 -> /dev/console
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/6 -> /dev/console
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/7 -> /dev/console

After the fix:

ls -lah /proc/1/fd/*

lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/0 -> /dev/console
lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/1 -> /dev/console
lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/2 -> /dev/console

Fix by closing prog FDs after initialization. struct bpf_prog's
themselves are kept alive through direct kernel references taken with
bpf_link_get_from_fd().

Fixes: cb80ddc6 ("bpf: Convert bpf_preload.ko to use light skeleton.")
Signed-off-by: default avatarYucong Sun <fallentree@fb.com>
Signed-off-by: default avatarAndrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220225185923.2535519-1-fallentree@fb.com
      80bebebd
  9. 24 Feb, 2022 1 commit
  11. 23 Feb, 2022 3 commits
  13. 22 Feb, 2022 5 commits
  15. 20 Feb, 2022 3 commits
    • Andrii Nakryiko's avatar
      selftests/bpf: Fix btfgen tests · b03e1946
      Andrii Nakryiko authored 
      There turned out to be a few problems with btfgen selftests.

First, core_btfgen tests are failing in BPF CI due to the use of
full-featured bpftool, which has extra dependencies on libbfd, libcap,
etc, which are present in BPF CI's build environment, but those shared
libraries are missing in QEMU image in which test_progs is running.

To fix this problem, use minimal bootstrap version of bpftool instead.
It only depend on libelf and libz, same as libbpf, so doesn't add any
new requirements (and bootstrap bpftool still implementes entire
`bpftool gen` functionality, which is quite convenient).

Second problem is even more interesting. Both core_btfgen and core_reloc
reuse the same set of struct core_reloc_test_case array of test case
definitions. That in itself is not a problem, but btfgen test replaces
test_case->btf_src_file property with the path to temporary file into
which minimized BTF is output by bpftool. This interferes with original
core_reloc tests, depending on order of tests execution (core_btfgen is
run first in sequential mode and skrews up subsequent core_reloc run by
pointing to already deleted temporary file, instead of the original BTF
files) and whether those two runs share the same process (in parallel
mode the chances are high for them to run in two separate processes and
so not interfere with each other).

To prevent this interference, create and use local copy of a test
definition. Mark original array as constant to catch accidental
modifcations. Note that setup_type_id_case_success() and
setup_type_id_case_success() still modify common test_case->output
memory area, but it is ok as each setup function has to re-initialize it
completely anyways. In sequential mode it leads to deterministic and
correct initialization. In parallel mode they will either each have
their own process, or if core_reloc and core_btfgen happen to be run by
the same worker process, they will still do that sequentially within the
worker process. If they are sharded across multiple processes, they
don't really share anything anyways.

Also, rename core_btfgen into core_reloc_btfgen, as it is indeed just
a "flavor" of core_reloc test, not an independent set of tests. So make
it more obvious.

Last problem that needed solving was that location of bpftool differs
between test_progs and test_progs' flavors (e.g., test_progs-no_alu32).
To keep it simple, create a symlink to bpftool both inside
selftests/bpf/ directory and selftests/bpf/<flavor> subdirectory. That
way, from inside core_reloc test, location to bpftool is just "./bpftool".

v2->v3:
  - fix bpftool location relative the test_progs-no_alu32;
v1->v2:
  - fix corruption of core_reloc_test_case.

Fixes: 704c91e5 ("selftests/bpf: Test "bpftool gen min_core_btf")
Signed-off-by: default avatarAndrii Nakryiko <andrii@kernel.org>
Signed-off-by: default avatarAlexei Starovoitov <ast@kernel.org>
Acked-by: default avatarYucong Sun <sunyucong@gmail.com>
Link: https://lore.kernel.org/bpf/20220220042720.3336684-1-andrii@kernel.org
      b03e1946
    • Souptick Joarder (HPE)'s avatar
      bpf: Initialize ret to 0 inside btf_populate_kfunc_set() · d0b38229
      Souptick Joarder (HPE) authored 
      Kernel test robot reported below error ->

kernel/bpf/btf.c:6718 btf_populate_kfunc_set()
error: uninitialized symbol 'ret'.

Initialize ret to 0.

Fixes: dee872e1 ("bpf: Populate kfunc BTF ID sets in struct btf")
Reported-by: default avatarkernel test robot <lkp@intel.com>
Signed-off-by: default avatarSouptick Joarder (HPE) <jrdr.linux@gmail.com>
Signed-off-by: default avatarAlexei Starovoitov <ast@kernel.org>
Acked-by: default avatarKumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20220219163915.125770-1-jrdr.linux@gmail.com
      d0b38229
    • Yonghong Song's avatar
      selftests/bpf: Fix a clang deprecated-declarations compilation error · a33c0c79
      Yonghong Song authored 
      Build the kernel and selftest with clang compiler with LLVM=1,
  make -j LLVM=1
  make -C tools/testing/selftests/bpf -j LLVM=1

I hit the following selftests/bpf compilation error:
  In file included from test_cpp.cpp:3:
  /.../tools/testing/selftests/bpf/tools/include/bpf/libbpf.h:73:8:
    error: 'relaxed_core_relocs' is deprecated: libbpf v0.6+: field has no effect [-Werror,-Wdeprecated-declarations]
  struct bpf_object_open_opts {
         ^
  test_cpp.cpp:56:2: note: in implicit move constructor for 'bpf_object_open_opts' first required here
          LIBBPF_OPTS(bpf_object_open_opts, opts);
          ^
  /.../tools/testing/selftests/bpf/tools/include/bpf/libbpf_common.h:77:3: note: expanded from macro 'LIBBPF_OPTS'
                  (struct TYPE) {                                             \
                  ^
  /.../tools/testing/selftests/bpf/tools/include/bpf/libbpf.h:90:2: note: 'relaxed_core_relocs' has been explicitly marked deprecated here
          LIBBPF_DEPRECATED_SINCE(0, 6, "field has no effect")
          ^
  /.../tools/testing/selftests/bpf/tools/include/bpf/libbpf_common.h:24:4: note: expanded from macro 'LIBBPF_DEPRECATED_SINCE'
                  (LIBBPF_DEPRECATED("libbpf v" # major "." # minor "+: " msg))
                   ^
  /.../tools/testing/selftests/bpf/tools/include/bpf/libbpf_common.h:19:47: note: expanded from macro 'LIBBPF_DEPRECATED'
  #define LIBBPF_DEPRECATED(msg) __attribute__((deprecated(msg)))

There are two ways to fix the issue, one is to use GCC diagnostic ignore pragma, and the
other is to open code bpf_object_open_opts instead of using LIBBPF_OPTS.
Since in general LIBBPF_OPTS is preferred, the patch fixed the issue by
adding proper GCC diagnostic ignore pragmas.
Signed-off-by: default avatarYonghong Song <yhs@fb.com>
Signed-off-by: default avatarAndrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220217194005.2765348-1-yhs@fb.com
      a33c0c79
  17. 18 Feb, 2022 11 commits
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7