Skip to content

cloud-kernel
cloud-kernel

openanolis / cloud-kernel
接近 2 年 前同步成功

通知 170
Star 36
Fork 7
cloud-kernel
cloud-kernel
切换分支/标签
  1. 02 4月, 2017 3 次提交
  3. 23 3月, 2017 2 次提交
    • M
      bpf: Add tests for map-in-map · fb30d4b7
      由 Martin KaFai Lau 提交于 
      Test cases for array of maps and hash of maps.
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      fb30d4b7
    • A
      bpf: fix hashmap extra_elems logic · 8c290e60
      由 Alexei Starovoitov 提交于 
      In both kmalloc and prealloc mode the bpf_map_update_elem() is using
per-cpu extra_elems to do atomic update when the map is full.
There are two issues with it. The logic can be misused, since it allows
max_entries+num_cpus elements to be present in the map. And alloc_extra_elems()
at map creation time can fail percpu alloc for large map values with a warn:
WARNING: CPU: 3 PID: 2752 at ../mm/percpu.c:892 pcpu_alloc+0x119/0xa60
illegal size (32824) or align (8) for percpu allocation

The fixes for both of these issues are different for kmalloc and prealloc modes.
For prealloc mode allocate extra num_possible_cpus elements and store
their pointers into extra_elems array instead of actual elements.
Hence we can use these hidden(spare) elements not only when the map is full
but during bpf_map_update_elem() that replaces existing element too.
That also improves performance, since pcpu_freelist_pop/push is avoided.
Unfortunately this approach cannot be used for kmalloc mode which needs
to kfree elements after rcu grace period. Therefore switch it back to normal
kmalloc even when full and old element exists like it was prior to
commit 6c905981 ("bpf: pre-allocate hash map elements").

Add tests to check for over max_entries and large map values.
Reported-by: NDave Jones <davej@codemonkey.org.uk>
Fixes: 6c905981 ("bpf: pre-allocate hash map elements")
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Acked-by: NMartin KaFai Lau <kafai@fb.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      8c290e60
  5. 22 3月, 2017 1 次提交
  7. 13 3月, 2017 1 次提交
  9. 10 3月, 2017 1 次提交
  11. 09 3月, 2017 1 次提交
    • C
      selftests/powerpc: Replace stxvx and lxvx with stxvd2x/lxvd2x · 605df8d6
      由 Cyril Bur 提交于 
      On POWER8 (ISA 2.07) lxvx and stxvx are defined to be extended mnemonics
of lxvd2x and stxvd2x. For POWER9 (ISA 3.0) the HW architects in their
infinite wisdom made lxvx and stxvx instructions in their own right.

POWER9 aware GCC will use the POWER9 instruction for lxvx and stxvx
causing these selftests to fail on POWER8. Further compounding the
issue, because of the way -mvsx works it will cause the power9
instructions to be used regardless of -mcpu=power8 to GCC or -mpower8 to
AS.

The safest way to address the problem for now is to not use the extended
mnemonic. We don't care how the CPU loads the values from memory since
the tests only performs register comparisons, so using stdvd2x/lxvd2x
does not impact the test.
Signed-off-by: NCyril Bur <cyrilbur@gmail.com>
Acked-by: Balbir Singh<bsingharora@gmail.com>
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
      605df8d6
  13. 03 3月, 2017 1 次提交
    • S
      selftest/powerpc: Fix false failures for skipped tests · a6d8a215
      由 Sachin Sant 提交于 
      Tests under alignment subdirectory are skipped when executed on previous
generation hardware, but harness still marks them as failed.

  test: test_copy_unaligned
  tags: git_version:unknown
  [SKIP] Test skipped on line 26
  skip: test_copy_unaligned
  selftests: copy_unaligned [FAIL]

The MAGIC_SKIP_RETURN_VALUE value assigned to rc variable is retained till
the program exit which causes the test to be marked as failed.

This patch resets the value before returning to the main() routine.
With this patch the test o/p is as follows:

  test: test_copy_unaligned
  tags: git_version:unknown
  [SKIP] Test skipped on line 26
  skip: test_copy_unaligned
  selftests: copy_unaligned [PASS]
Signed-off-by: NSachin Sant <sachinp@linux.vnet.ibm.com>
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
      a6d8a215
  15. 02 3月, 2017 2 次提交
    • S
      selftests: lib.mk Fix individual test builds · e53aff45
      由 Shuah Khan 提交于 
      In commit a8ba798b ("selftests: enable O and KBUILD_OUTPUT"), added
support to generate compile targets in a user specified directory. OUTPUT
variable controls the location which is undefined when tests are built in
the test directory or with "make -C tools/testing/selftests/x86".

make -C tools/testing/selftests/x86/
make: Entering directory '/lkml/linux_4.11/tools/testing/selftests/x86'
Makefile:44: warning: overriding recipe for target 'clean'
../lib.mk:51: warning: ignoring old recipe for target 'clean'
gcc -m64 -o /single_step_syscall_64 -O2 -g -std=gnu99 -pthread -Wall  single_step_syscall.c -lrt -ldl
/usr/bin/ld: cannot open output file /single_step_syscall_64: Permission denied
collect2: error: ld returned 1 exit status
Makefile:50: recipe for target '/single_step_syscall_64' failed
make: *** [/single_step_syscall_64] Error 1
make: Leaving directory '/lkml/linux_4.11/tools/testing/selftests/x86'

Same failure with "cd tools/testing/selftests/x86/;make" run.

Fix this with a change to lib.mk to define OUTPUT to be the pwd when
MAKELEVEL is 0. This covers both cases mentioned above.
Reported-by: NIngo Molnar <mingo@kernel.org>
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
      e53aff45
    • A
      selftests/x86: Add a basic selftest for ioperm · 0eb1d0fa
      由 Andy Lutomirski 提交于 
      This doesn't fully exercise the interaction between KVM and ioperm(),
but it does test basic functionality.
Signed-off-by: NAndy Lutomirski <luto@kernel.org>
Signed-off-by: NRadim Krčmář <rkrcmar@redhat.com>
      0eb1d0fa
  17. 01 3月, 2017 1 次提交
  19. 28 2月, 2017 1 次提交
  21. 25 2月, 2017 2 次提交
  23. 23 2月, 2017 7 次提交
  25. 19 2月, 2017 1 次提交
  27. 16 2月, 2017 1 次提交
  29. 14 2月, 2017 5 次提交
    • M
      selftests/powerpc: Fix remaining fallout from recent changes · 68bd42d9
      由 Michael Ellerman 提交于 
      In benchmarks we need to use $(TEST_GEN_PROGS) after we include lib.mk,
because lib.mk does the substitution to add $(OUTPUT).

In math the vmx and fpu names were typoed so they no longer matched
correctly, put back the 'v' and 'f'.

In tm we need to substitute $(OUTPUT) into SIGNAL_CONTEXT_CHK_TESTS so
that the rule matches.

In pmu there is an extraneous ':' on the end of $$BUILD_TARGET for the
clean and install rules, which breaks the logic in the child Makefiles.

Fixes: a8ba798b ("selftests: enable O and KBUILD_OUTPUT")
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
      68bd42d9
    • M
      selftests/powerpc: Fix the clean rule since recent changes · 2e8ec87d
      由 Michael Ellerman 提交于 
      The clean rule is broken for the powerpc tests:

  make[1]: Entering directory 'tools/testing/selftests/powerpc'
  Makefile:63: warning: overriding recipe for target 'clean'
  ../lib.mk:51: warning: ignoring old recipe for target 'clean'
  /bin/sh: 3: Syntax error: end of file unexpected (expecting "done")
  Makefile:63: recipe for target 'clean' failed

Fixes: a8ba798b ("selftests: enable O and KBUILD_OUTPUT")
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
      2e8ec87d
    • M
      selftests: Fix the .S and .S -> .o rules · 634ce97c
      由 Michael Ellerman 提交于 
      Both these rules incorrectly use $< (first prerequisite) rather than
$^ (all prerequisites), meaning they don't work if we're using more than
one .S file as input. Switch them to using $^.

They also don't include $(CPPFLAGS) and other variables used in the
default rules, which breaks targets that require those. Fix that by
using the builtin $(COMPILE.S) and $(LINK.S) rules.

Fixes: a8ba798b ("selftests: enable O and KBUILD_OUTPUT")
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
Tested by: Bamvor Jian Zhang <bamvor.zhangjian@linaro.org>
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
      634ce97c
    • M
      selftests: Fix the .c linking rule · 2047f1d8
      由 Michael Ellerman 提交于 
      Currently we can't build some tests, for example:

  $ make -C tools/testing/selftests/ TARGETS=vm
  ...
  gcc -Wall -I ../../../../usr/include   -lrt -lpthread ../../../../usr/include/linux/kernel.h userfaultfd.c -o tools/testing/selftests/vm/userfaultfd
  /tmp/ccmOkQSM.o: In function `stress':
  userfaultfd.c:(.text+0xc60): undefined reference to `pthread_create'
  userfaultfd.c:(.text+0xca5): undefined reference to `pthread_create'
  userfaultfd.c:(.text+0xcee): undefined reference to `pthread_create'
  userfaultfd.c:(.text+0xd30): undefined reference to `pthread_create'
  userfaultfd.c:(.text+0xd77): undefined reference to `pthread_join'
  userfaultfd.c:(.text+0xe7d): undefined reference to `pthread_join'
  userfaultfd.c:(.text+0xe9f): undefined reference to `pthread_cancel'
  userfaultfd.c:(.text+0xec6): undefined reference to `pthread_join'
  userfaultfd.c:(.text+0xf14): undefined reference to `pthread_join'
  /tmp/ccmOkQSM.o: In function `userfaultfd_stress':
  userfaultfd.c:(.text+0x13e2): undefined reference to `pthread_attr_setstacksize'
  collect2: error: ld returned 1 exit status

This is because the rule for linking .c files to binaries is incorrect.

The first bug is that it uses $< (first prerequisite) instead of $^ (all
preqrequisites), fix it by using ^$.

Secondly the ordering of the prerequisites vs $(LDLIBS) is wrong,
meaning on toolchains that use --as-needed we fail to link (as above).
Fix that by placing $(LDLIBS) *after* ^$.

Finally switch to using the default rule $(LINK.c), so that we get
$(CPPFLAGS) etc. included.

Fixes: a8ba798b ("selftests: enable O and KBUILD_OUTPUT")
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
Tested by: Bamvor Jian Zhang <bamvor.zhangjian@linaro.org>
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
      2047f1d8
    • M
      selftests: Fix selftests build to just build, not run tests · d83c3ba0
      由 Michael Ellerman 提交于 
      In commit 88baa78d ("selftests: remove duplicated all and clean
target"), the "all" target was removed from individual Makefiles and
added to lib.mk.

However the "all" target was added to lib.mk *after* the existing
"runtests" target. This means "runtests" becomes the first (default)
target for most of our Makefiles.

This has the effect of causing a plain "make" to build *and run* the
tests. Which is at best rude, but depending on which tests are run could
oops someone's build machine.

  $ make -C tools/testing/selftests/
  ...
  make[1]: Entering directory 'tools/testing/selftests/bpf'
  gcc -Wall -O2 -I../../../../usr/include   test_verifier.c -o tools/testing/selftests/bpf/test_verifier
  gcc -Wall -O2 -I../../../../usr/include   test_maps.c -o tools/testing/selftests/bpf/test_maps
  gcc -Wall -O2 -I../../../../usr/include   test_lru_map.c -o tools/testing/selftests/bpf/test_lru_map
  #0 add+sub+mul FAIL
  Failed to load prog 'Function not implemented'!
  #1 unreachable FAIL
  Unexpected error message!
  #2 unreachable2 FAIL
  ...

Fix it by moving the "all" target to the start of lib.mk, making it the
default target.

Fixes: 88baa78d ("selftests: remove duplicated all and clean target")
Signed-off-by: NMichael Ellerman <mpe@ellerman.id.au>
Tested by: Bamvor Jian Zhang <bamvor.zhangjian@linaro.org>
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
      d83c3ba0
  31. 11 2月, 2017 10 次提交