Re: [PATCH 2/2] bitmap: Add test for bitmap_cut()
On Mon, 15 Jun 2020 14:43:53 +0300 Andy Shevchenko wrote: > On Mon, Jun 15, 2020 at 01:08:25PM +0200, Stefano Brivio wrote: > > > > [...] > > > > By the way, tests for 'parse', 'parse_user' and 'parselist' report > > issues: > > I believe this [1] will fix it. > > [1]: 81c4f4d924d5 ("lib: fix bitmap_parse() on 64-bit big endian archs") Yes, thanks, that works for me too. -- Stefano
Re: [PATCH 2/2] bitmap: Add test for bitmap_cut()
On Mon, Jun 15, 2020 at 12:46:16PM +0300, Andy Shevchenko wrote: > On Mon, Jun 15, 2020 at 12:41:55PM +0300, Andy Shevchenko wrote: > > On Sun, Jun 14, 2020 at 07:40:54PM +0200, Stefano Brivio wrote: > > > Inspired by an original patch from Yury Norov: introduce a test for > > > bitmap_cut() that also makes sure functionality is as described for > > > partially overlapping src and dst. > > > > Taking into account recent fixes for BE 64-bit, do we have test cases for a > > such? > > It might be enough to have only these, but perhaps s390 guys can help? > > Alexander, can you apply this patch (w/o the first one, which is suppose to > fix) and confirm that you have test case failure, followed by applying first > one and confirm a fix? This failure goes away when patch #1 is applied: test_bitmap: [lib/test_bitmap.c:666] bitmaps contents differ: expected "0-16,18-19,21,24,26-27,29", got "1,3-4,6,9,11-12,14,16,18-19,21,24,26-27,29" Thus, I confirm. [...] > -- > With Best Regards, > Andy Shevchenko > >
Re: [PATCH 2/2] bitmap: Add test for bitmap_cut()
On Mon, Jun 15, 2020 at 01:08:25PM +0200, Stefano Brivio wrote: > On Mon, 15 Jun 2020 12:46:16 +0300 > Andy Shevchenko wrote: > > > On Mon, Jun 15, 2020 at 12:41:55PM +0300, Andy Shevchenko wrote: > > > On Sun, Jun 14, 2020 at 07:40:54PM +0200, Stefano Brivio wrote: > > > > Inspired by an original patch from Yury Norov: introduce a test for > > > > bitmap_cut() that also makes sure functionality is as described for > > > > partially overlapping src and dst. > > > > > > Taking into account recent fixes for BE 64-bit, do we have test cases for > > > a such? > > > > It might be enough to have only these, but perhaps s390 guys can help? > > There's no behaviour difference due to endianness in this test itself -- > just word size was a topic, hence that BITS_PER_LONG usage with > redundant values (checked on i686). > > That is, if you have: > { 0xUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL }, > > then 1 as array subscript always denotes the second item (from the left) > there, it doesn't matter how and where different architectures store it. > > Indeed, if bitmap_cut() directly addressed single bytes within the > words, I would need to pay special attention there. The values I picked > for these tests are also meant to show any issue in that sense. > > > Alexander, can you apply this patch (w/o the first one, which is suppose to > > fix) and confirm that you have test case failure, followed by applying first > > one and confirm a fix? > > I did that already on s390x (of course, I thought :)), I can confirm > that. Without patch 1/2 the test also fails there: > > [ 20.917848] test_bitmap: [lib/test_bitmap.c:666] bitmaps contents differ: > expected "0-16,18-19,21,24,26-27,29", got > "1,3-4,6,9,11-12,14,16,18-19,21,24,26-27,29" Thanks! Reviewed-by: Andy Shevchenko > If Alexander wants to test this on a z14 or z15, sure, it won't harm. Sure. > By the way, tests for 'parse', 'parse_user' and 'parselist' report > issues: I believe this [1] will fix it. [1]: 81c4f4d924d5 ("lib: fix bitmap_parse() on 64-bit big endian archs") -- With Best Regards, Andy Shevchenko
Re: [PATCH 2/2] bitmap: Add test for bitmap_cut()
On Mon, 15 Jun 2020 12:46:16 +0300 Andy Shevchenko wrote: > On Mon, Jun 15, 2020 at 12:41:55PM +0300, Andy Shevchenko wrote: > > On Sun, Jun 14, 2020 at 07:40:54PM +0200, Stefano Brivio wrote: > > > Inspired by an original patch from Yury Norov: introduce a test for > > > bitmap_cut() that also makes sure functionality is as described for > > > partially overlapping src and dst. > > > > Taking into account recent fixes for BE 64-bit, do we have test cases for a > > such? > > It might be enough to have only these, but perhaps s390 guys can help? There's no behaviour difference due to endianness in this test itself -- just word size was a topic, hence that BITS_PER_LONG usage with redundant values (checked on i686). That is, if you have: { 0xUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL }, then 1 as array subscript always denotes the second item (from the left) there, it doesn't matter how and where different architectures store it. Indeed, if bitmap_cut() directly addressed single bytes within the words, I would need to pay special attention there. The values I picked for these tests are also meant to show any issue in that sense. > Alexander, can you apply this patch (w/o the first one, which is suppose to > fix) and confirm that you have test case failure, followed by applying first > one and confirm a fix? I did that already on s390x (of course, I thought :)), I can confirm that. Without patch 1/2 the test also fails there: [ 20.917848] test_bitmap: [lib/test_bitmap.c:666] bitmaps contents differ: expected "0-16,18-19,21,24,26-27,29", got "1,3-4,6,9,11-12,14,16,18-19,21,24,26-27,29" If Alexander wants to test this on a z14 or z15, sure, it won't harm. By the way, tests for 'parse', 'parse_user' and 'parselist' report issues: [ 20.390401] test_bitmap: loaded. [ 20.394839] test_bitmap: parse: 4: input is 1, result is 0x1, expected 0x1 [ 20.395011] test_bitmap: parse: 5: input is deadbeef, result is 0xdeadbeef, expected 0xdeadbeef [ 20.395059] test_bitmap: parse: 6: input is 1,0, result is 0x1, expected 0x1 [ 20.395099] test_bitmap: parse: 7: input is deadbeef, ,0,1, result is 0x1, expected 0xdeadbeef [ 20.396696] test_bitmap: parse: 8: input is deadbeef,1,0, result is 0x1, expected 0x1 [ 20.396735] test_bitmap: parse: 9: input is baadf00d,deadbeef,1,0, result is 0x1, expected 0x1 [ 20.396835] test_bitmap: parse: 10: input is badf00d,deadbeef,1,0, errno is -75, expected 0 [ 20.396878] test_bitmap: parse: 11: input is badf00d,deadbeef,1,0, errno is -75, expected 0 [ 20.396913] test_bitmap: parse: 12: input is badf00d,deadbeef,1,0 , errno is -75, expected 0 [ 20.396957] test_bitmap: parse: 13: input is , badf00d,deadbeef,1,0 , , errno is -75, expected 0 [ 20.396983] test_bitmap: parse: 14: input is , badf00d, ,, ,,deadbeef,1,0 , , errno is -75, expected 0 [ 20.397052] test_bitmap: parse: 16: input is 3,0, errno is 0, expected -75 [ 20.397712] test_bitmap: parse_user: 4: input is 1, result is 0x1, expected 0x1 [ 20.397832] test_bitmap: parse_user: 5: input is deadbeef, result is 0xdeadbeef, expected 0xdeadbeef [ 20.397928] test_bitmap: parse_user: 6: input is 1,0, result is 0x1, expected 0x1 [ 20.398022] test_bitmap: parse_user: 7: input is deadbeef, ,0,1, result is 0x1, expected 0xdeadbeef [ 20.398116] test_bitmap: parse_user: 8: input is deadbeef,1,0, result is 0x1, expected 0x1 [ 20.398209] test_bitmap: parse_user: 9: input is baadf00d,deadbeef,1,0, result is 0x1, expected 0x1 [ 20.398301] test_bitmap: parse_user: 10: input is badf00d,deadbeef,1,0, errno is -75, expected 0 [ 20.398393] test_bitmap: parse_user: 11: input is badf00d,deadbeef,1,0, errno is -75, expected 0 [ 20.398484] test_bitmap: parse_user: 12: input is badf00d,deadbeef,1,0 , errno is -75, expected 0 [ 20.398574] test_bitmap: parse_user: 13: input is , badf00d,deadbeef,1,0 , , errno is -75, expected 0 [ 20.398666] test_bitmap: parse_user: 14: input is , badf00d, ,, ,,deadbeef,1,0 , , errno is -75, expected 0 [ 20.398794] test_bitmap: parse_user: 16: input is 3,0, errno is 0, expected -75 [ 20.399906] test_bitmap: parselist: 14: input is '0-2047:128/256' OK, Time: 2641 [ 20.400914] test_bitmap: parselist_user: 14: input is '0-2047:128/256' OK, Time: 19961 [ 20.421406] test_bitmap: all 1679 tests passed and at a glance those *seem* to be bugs in the tests themselves, not in the actual functions they test. Sure, they should be fixed, but I can't take care of that right now. -- Stefano
Re: [PATCH 2/2] bitmap: Add test for bitmap_cut()
On Mon, Jun 15, 2020 at 12:41:55PM +0300, Andy Shevchenko wrote: > On Sun, Jun 14, 2020 at 07:40:54PM +0200, Stefano Brivio wrote: > > Inspired by an original patch from Yury Norov: introduce a test for > > bitmap_cut() that also makes sure functionality is as described for > > partially overlapping src and dst. > > Taking into account recent fixes for BE 64-bit, do we have test cases for a > such? It might be enough to have only these, but perhaps s390 guys can help? Alexander, can you apply this patch (w/o the first one, which is suppose to fix) and confirm that you have test case failure, followed by applying first one and confirm a fix? > > Signed-off-by: Stefano Brivio > > --- > > v2: > > - use expect_eq_bitmap() instead of open coding result check (Andy > > Shevchenko) > > - don't use uncommon Co-authored-by: tag (Andy Shevchenko), drop > > it altogether as Yury asked me to go ahead with this and I haven't > > heard back in a while. Patch is now rather different anyway > > - avoid stack overflow, buffer needs to be five unsigned longs and > > not four as 'in' is shifted by one, spotted by kernel test robot > > with CONFIG_STACKPROTECTOR_STRONG > > > > lib/test_bitmap.c | 58 +++ > > 1 file changed, 58 insertions(+) > > > > diff --git a/lib/test_bitmap.c b/lib/test_bitmap.c > > index 6b13150667f5..df903c53952b 100644 > > --- a/lib/test_bitmap.c > > +++ b/lib/test_bitmap.c > > @@ -610,6 +610,63 @@ static void __init test_for_each_set_clump8(void) > > expect_eq_clump8(start, CLUMP_EXP_NUMBITS, clump_exp, ); > > } > > > > +struct test_bitmap_cut { > > + unsigned int first; > > + unsigned int cut; > > + unsigned int nbits; > > + unsigned long in[4]; > > + unsigned long expected[4]; > > +}; > > + > > +static struct test_bitmap_cut test_cut[] = { > > + { 0, 0, 8, { 0x000aUL, }, { 0x000aUL, }, }, > > + { 0, 0, 32, { 0xdadadeadUL, }, { 0xdadadeadUL, }, }, > > + { 0, 3, 8, { 0x00aaUL, }, { 0x0015UL, }, }, > > + { 3, 3, 8, { 0x00aaUL, }, { 0x0012UL, }, }, > > + { 0, 1, 32, { 0xa5a5a5a5UL, }, { 0x52d2d2d2UL, }, }, > > + { 0, 8, 32, { 0xdeadc0deUL, }, { 0x00deadc0UL, }, }, > > + { 1, 1, 32, { 0x5a5a5a5aUL, }, { 0x2d2d2d2cUL, }, }, > > + { 0, 15, 32, { 0xa5a5a5a5UL, }, { 0x00014b4bUL, }, }, > > + { 0, 16, 32, { 0xa5a5a5a5UL, }, { 0xa5a5UL, }, }, > > + { 15, 15, 32, { 0xa5a5a5a5UL, }, { 0x000125a5UL, }, }, > > + { 15, 16, 32, { 0xa5a5a5a5UL, }, { 0xa5a5UL, }, }, > > + { 16, 15, 32, { 0xa5a5a5a5UL, }, { 0x0001a5a5UL, }, }, > > + > > + { BITS_PER_LONG, BITS_PER_LONG, BITS_PER_LONG, > > + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, > > + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, > > + }, > > + { 1, BITS_PER_LONG - 1, BITS_PER_LONG, > > + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, > > + { 0x0001UL, 0x0001UL, }, > > + }, > > + > > + { 0, BITS_PER_LONG * 2, BITS_PER_LONG * 2 + 1, > > + { 0xa5a5a5a5UL, 0x0001UL, 0x0001UL, 0x0001UL }, > > + { 0x0001UL, }, > > + }, > > + { 16, BITS_PER_LONG * 2 + 1, BITS_PER_LONG * 2 + 1 + 16, > > + { 0xUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL }, > > + { 0x2d2dUL, }, > > + }, > > +}; > > + > > +static void __init test_bitmap_cut(void) > > +{ > > + unsigned long b[5], *in = [1], *out = [0]; /* Partial overlap */ > > + int i; > > + > > + for (i = 0; i < ARRAY_SIZE(test_cut); i++) { > > + struct test_bitmap_cut *t = _cut[i]; > > + > > + memcpy(in, t->in, sizeof(t->in)); > > + > > + bitmap_cut(out, in, t->first, t->cut, t->nbits); > > + > > + expect_eq_bitmap(t->expected, out, t->nbits); > > + } > > +} > > + > > static void __init selftest(void) > > { > > test_zero_clear(); > > @@ -623,6 +680,7 @@ static void __init selftest(void) > > test_bitmap_parselist_user(); > > test_mem_optimisations(); > > test_for_each_set_clump8(); > > + test_bitmap_cut(); > > } > > > > KSTM_MODULE_LOADERS(test_bitmap); > > -- > > 2.27.0 > > > > -- > With Best Regards, > Andy Shevchenko > > -- With Best Regards, Andy Shevchenko
Re: [PATCH 2/2] bitmap: Add test for bitmap_cut()
On Sun, Jun 14, 2020 at 07:40:54PM +0200, Stefano Brivio wrote: > Inspired by an original patch from Yury Norov: introduce a test for > bitmap_cut() that also makes sure functionality is as described for > partially overlapping src and dst. Taking into account recent fixes for BE 64-bit, do we have test cases for a such? > Signed-off-by: Stefano Brivio > --- > v2: > - use expect_eq_bitmap() instead of open coding result check (Andy > Shevchenko) > - don't use uncommon Co-authored-by: tag (Andy Shevchenko), drop > it altogether as Yury asked me to go ahead with this and I haven't > heard back in a while. Patch is now rather different anyway > - avoid stack overflow, buffer needs to be five unsigned longs and > not four as 'in' is shifted by one, spotted by kernel test robot > with CONFIG_STACKPROTECTOR_STRONG > > lib/test_bitmap.c | 58 +++ > 1 file changed, 58 insertions(+) > > diff --git a/lib/test_bitmap.c b/lib/test_bitmap.c > index 6b13150667f5..df903c53952b 100644 > --- a/lib/test_bitmap.c > +++ b/lib/test_bitmap.c > @@ -610,6 +610,63 @@ static void __init test_for_each_set_clump8(void) > expect_eq_clump8(start, CLUMP_EXP_NUMBITS, clump_exp, ); > } > > +struct test_bitmap_cut { > + unsigned int first; > + unsigned int cut; > + unsigned int nbits; > + unsigned long in[4]; > + unsigned long expected[4]; > +}; > + > +static struct test_bitmap_cut test_cut[] = { > + { 0, 0, 8, { 0x000aUL, }, { 0x000aUL, }, }, > + { 0, 0, 32, { 0xdadadeadUL, }, { 0xdadadeadUL, }, }, > + { 0, 3, 8, { 0x00aaUL, }, { 0x0015UL, }, }, > + { 3, 3, 8, { 0x00aaUL, }, { 0x0012UL, }, }, > + { 0, 1, 32, { 0xa5a5a5a5UL, }, { 0x52d2d2d2UL, }, }, > + { 0, 8, 32, { 0xdeadc0deUL, }, { 0x00deadc0UL, }, }, > + { 1, 1, 32, { 0x5a5a5a5aUL, }, { 0x2d2d2d2cUL, }, }, > + { 0, 15, 32, { 0xa5a5a5a5UL, }, { 0x00014b4bUL, }, }, > + { 0, 16, 32, { 0xa5a5a5a5UL, }, { 0xa5a5UL, }, }, > + { 15, 15, 32, { 0xa5a5a5a5UL, }, { 0x000125a5UL, }, }, > + { 15, 16, 32, { 0xa5a5a5a5UL, }, { 0xa5a5UL, }, }, > + { 16, 15, 32, { 0xa5a5a5a5UL, }, { 0x0001a5a5UL, }, }, > + > + { BITS_PER_LONG, BITS_PER_LONG, BITS_PER_LONG, > + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, > + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, > + }, > + { 1, BITS_PER_LONG - 1, BITS_PER_LONG, > + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, > + { 0x0001UL, 0x0001UL, }, > + }, > + > + { 0, BITS_PER_LONG * 2, BITS_PER_LONG * 2 + 1, > + { 0xa5a5a5a5UL, 0x0001UL, 0x0001UL, 0x0001UL }, > + { 0x0001UL, }, > + }, > + { 16, BITS_PER_LONG * 2 + 1, BITS_PER_LONG * 2 + 1 + 16, > + { 0xUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL }, > + { 0x2d2dUL, }, > + }, > +}; > + > +static void __init test_bitmap_cut(void) > +{ > + unsigned long b[5], *in = [1], *out = [0]; /* Partial overlap */ > + int i; > + > + for (i = 0; i < ARRAY_SIZE(test_cut); i++) { > + struct test_bitmap_cut *t = _cut[i]; > + > + memcpy(in, t->in, sizeof(t->in)); > + > + bitmap_cut(out, in, t->first, t->cut, t->nbits); > + > + expect_eq_bitmap(t->expected, out, t->nbits); > + } > +} > + > static void __init selftest(void) > { > test_zero_clear(); > @@ -623,6 +680,7 @@ static void __init selftest(void) > test_bitmap_parselist_user(); > test_mem_optimisations(); > test_for_each_set_clump8(); > + test_bitmap_cut(); > } > > KSTM_MODULE_LOADERS(test_bitmap); > -- > 2.27.0 > -- With Best Regards, Andy Shevchenko
[PATCH 2/2] bitmap: Add test for bitmap_cut()
Inspired by an original patch from Yury Norov: introduce a test for bitmap_cut() that also makes sure functionality is as described for partially overlapping src and dst. Signed-off-by: Stefano Brivio --- v2: - use expect_eq_bitmap() instead of open coding result check (Andy Shevchenko) - don't use uncommon Co-authored-by: tag (Andy Shevchenko), drop it altogether as Yury asked me to go ahead with this and I haven't heard back in a while. Patch is now rather different anyway - avoid stack overflow, buffer needs to be five unsigned longs and not four as 'in' is shifted by one, spotted by kernel test robot with CONFIG_STACKPROTECTOR_STRONG lib/test_bitmap.c | 58 +++ 1 file changed, 58 insertions(+) diff --git a/lib/test_bitmap.c b/lib/test_bitmap.c index 6b13150667f5..df903c53952b 100644 --- a/lib/test_bitmap.c +++ b/lib/test_bitmap.c @@ -610,6 +610,63 @@ static void __init test_for_each_set_clump8(void) expect_eq_clump8(start, CLUMP_EXP_NUMBITS, clump_exp, ); } +struct test_bitmap_cut { + unsigned int first; + unsigned int cut; + unsigned int nbits; + unsigned long in[4]; + unsigned long expected[4]; +}; + +static struct test_bitmap_cut test_cut[] = { + { 0, 0, 8, { 0x000aUL, }, { 0x000aUL, }, }, + { 0, 0, 32, { 0xdadadeadUL, }, { 0xdadadeadUL, }, }, + { 0, 3, 8, { 0x00aaUL, }, { 0x0015UL, }, }, + { 3, 3, 8, { 0x00aaUL, }, { 0x0012UL, }, }, + { 0, 1, 32, { 0xa5a5a5a5UL, }, { 0x52d2d2d2UL, }, }, + { 0, 8, 32, { 0xdeadc0deUL, }, { 0x00deadc0UL, }, }, + { 1, 1, 32, { 0x5a5a5a5aUL, }, { 0x2d2d2d2cUL, }, }, + { 0, 15, 32, { 0xa5a5a5a5UL, }, { 0x00014b4bUL, }, }, + { 0, 16, 32, { 0xa5a5a5a5UL, }, { 0xa5a5UL, }, }, + { 15, 15, 32, { 0xa5a5a5a5UL, }, { 0x000125a5UL, }, }, + { 15, 16, 32, { 0xa5a5a5a5UL, }, { 0xa5a5UL, }, }, + { 16, 15, 32, { 0xa5a5a5a5UL, }, { 0x0001a5a5UL, }, }, + + { BITS_PER_LONG, BITS_PER_LONG, BITS_PER_LONG, + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, + }, + { 1, BITS_PER_LONG - 1, BITS_PER_LONG, + { 0xa5a5a5a5UL, 0xa5a5a5a5UL, }, + { 0x0001UL, 0x0001UL, }, + }, + + { 0, BITS_PER_LONG * 2, BITS_PER_LONG * 2 + 1, + { 0xa5a5a5a5UL, 0x0001UL, 0x0001UL, 0x0001UL }, + { 0x0001UL, }, + }, + { 16, BITS_PER_LONG * 2 + 1, BITS_PER_LONG * 2 + 1 + 16, + { 0xUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL, 0x5a5a5a5aUL }, + { 0x2d2dUL, }, + }, +}; + +static void __init test_bitmap_cut(void) +{ + unsigned long b[5], *in = [1], *out = [0]; /* Partial overlap */ + int i; + + for (i = 0; i < ARRAY_SIZE(test_cut); i++) { + struct test_bitmap_cut *t = _cut[i]; + + memcpy(in, t->in, sizeof(t->in)); + + bitmap_cut(out, in, t->first, t->cut, t->nbits); + + expect_eq_bitmap(t->expected, out, t->nbits); + } +} + static void __init selftest(void) { test_zero_clear(); @@ -623,6 +680,7 @@ static void __init selftest(void) test_bitmap_parselist_user(); test_mem_optimisations(); test_for_each_set_clump8(); + test_bitmap_cut(); } KSTM_MODULE_LOADERS(test_bitmap); -- 2.27.0