It's not clear what's so horrible about emitting a function call to handle a run-time sized bitmap. Moreover, gcc also emits a function call for a compile-time-constant-but-huge nbits, so the comment isn't even accurate. Link: http://lkml.kernel.org/r/20180818131623.8755-6-linux@rasmusvillemoes.dk Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Yury Norov <ynorov@caviumnetworks.com> Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk> Cc: Sudeep Holla <sudeep.holla@arm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Most other bitmap API, including the OOL version __bitmap_shift_right,
take unsigned nbits. This was accidentally left out from 2fbad29917c98.
Link: http://lkml.kernel.org/r/20180818131623.8755-5-linux@rasmusvillemoes.dk
Fixes: 2fbad29917c98 ("lib: bitmap: change bitmap_shift_right to take unsigned parameters")
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reported-by: Yury Norov <ynorov@caviumnetworks.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In the _zero, _fill and _copy functions, the small_const_nbits branch is
redundant. If nbits is small and const, gcc knows full well that
BITS_TO_LONGS(nbits) is 1, so len is also a compile-time constant
(sizeof(long)), and calling memset or memcpy with a length argument of
sizeof(long) makes gcc generate the expected code anyway:
#include <string.h>
void a(unsigned long *x) { memset(x, 0, 8); }
void b(unsigned long *x) { memset(x, 0xff, 8); }
void c(unsigned long *x, const unsigned long *y) { memcpy(x, y, 8); }
turns into
0000000000000000 <a>:
0: 48 c7 07 00 00 00 00 movq $0x0,(%rdi)
7: c3 retq
8: 0f 1f 84 00 00 00 00 nopl 0x0(%rax,%rax,1)
f: 00
0000000000000010 <b>:
10: 48 c7 07 ff ff ff ff movq $0xffffffffffffffff,(%rdi)
17: c3 retq
18: 0f 1f 84 00 00 00 00 nopl 0x0(%rax,%rax,1)
1f: 00
0000000000000020 <c>:
20: 48 8b 06 mov (%rsi),%rax
23: 48 89 07 mov %rax,(%rdi)
26: c3 retq
Link: http://lkml.kernel.org/r/20180818131623.8755-4-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Yury Norov <ynorov@caviumnetworks.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The static inlines in bitmap.h do not handle a compile-time constant nbits==0 correctly (they dereference the passed src or dst pointers, despite only 0 words being valid to access). I had the 0-day buildbot chew on a patch [1] that would cause build failures for such cases without complaining, suggesting that we don't have any such users currently, at least for the 70 .config/arch combinations that was built. Should any turn up, make sure they use the out-of-line versions, which do handle nbits==0 correctly. This is of course not the most efficient, but it's much less churn than teaching all the static inlines an "if (zero_const_nbits())", and since we don't have any current instances, this doesn't affect existing code at all. [1] lkml.kernel.org/r/20180815085539.27485-1-linux@rasmusvillemoes.dk Link: http://lkml.kernel.org/r/20180818131623.8755-3-linux@rasmusvillemoes.dk Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Yury Norov <ynorov@caviumnetworks.com> Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk> Cc: Sudeep Holla <sudeep.holla@arm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A lot of code become ugly because of open coding allocations for bitmaps. Introduce three helpers to allow users be more clear of intention and keep their code neat. Note, due to multiple circular dependencies we may not provide the helpers as inliners. For now we keep them exported and, perhaps, at some point in the future we will sort out header inclusion and inheritance. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Commit 2a98dc028f91 ("include/linux/bitmap.h: turn bitmap_set and
bitmap_clear into memset when possible") introduced an optimization to
bitmap_{set,clear}() which uses memset() when the start and length are
constants aligned to a byte.
This is wrong on big-endian systems; our bitmaps are arrays of unsigned
long, so bit n is not at byte n / 8 in memory. This was caught by the
Btrfs selftests, but the bitmap selftests also fail when run on a
big-endian machine.
We can still use memset if the start and length are aligned to an
unsigned long, so do that on big-endian. The same problem applies to
the memcmp in bitmap_equal(), so fix it there, too.
Fixes: 2a98dc028f91 ("include/linux/bitmap.h: turn bitmap_set and bitmap_clear into memset when possible")
Fixes: 2c6deb01525a ("bitmap: use memcmp optimisation in more situations")
Cc: stable@kernel.org
Reported-by: "Erhard F." <erhard_f@mailbox.org>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We've measured that we spend ~0.6% of sys cpu time in cpumask_next_and().
It's essentially a joined iteration in search for a non-zero bit, which is
currently implemented as a lookup join (find a nonzero bit on the lhs,
lookup the rhs to see if it's set there).
Implement a direct join (find a nonzero bit on the incrementally built
join). Also add generic bitmap benchmarks in the new `test_find_bit`
module for new function (see `find_next_and_bit` in [2] and [3] below).
For cpumask_next_and, direct benchmarking shows that it's 1.17x to 14x
faster with a geometric mean of 2.1 on 32 CPUs [1]. No impact on memory
usage. Note that on Arm, the new pure-C implementation still outperforms
the old one that uses a mix of C and asm (`find_next_bit`) [3].
[1] Approximate benchmark code:
```
unsigned long src1p[nr_cpumask_longs] = {pattern1};
unsigned long src2p[nr_cpumask_longs] = {pattern2};
for (/*a bunch of repetitions*/) {
for (int n = -1; n <= nr_cpu_ids; ++n) {
asm volatile("" : "+rm"(src1p)); // prevent any optimization
asm volatile("" : "+rm"(src2p));
unsigned long result = cpumask_next_and(n, src1p, src2p);
asm volatile("" : "+rm"(result));
}
}
```
Results:
pattern1 pattern2 time_before/time_after
0x0000ffff 0x0000ffff 1.65
0x0000ffff 0x00005555 2.24
0x0000ffff 0x00001111 2.94
0x0000ffff 0x00000000 14.0
0x00005555 0x0000ffff 1.67
0x00005555 0x00005555 1.71
0x00005555 0x00001111 1.90
0x00005555 0x00000000 6.58
0x00001111 0x0000ffff 1.46
0x00001111 0x00005555 1.49
0x00001111 0x00001111 1.45
0x00001111 0x00000000 3.10
0x00000000 0x0000ffff 1.18
0x00000000 0x00005555 1.18
0x00000000 0x00001111 1.17
0x00000000 0x00000000 1.25
-----------------------------
geo.mean 2.06
[2] test_find_next_bit, X86 (skylake)
[ 3913.477422] Start testing find_bit() with random-filled bitmap
[ 3913.477847] find_next_bit: 160868 cycles, 16484 iterations
[ 3913.477933] find_next_zero_bit: 169542 cycles, 16285 iterations
[ 3913.478036] find_last_bit: 201638 cycles, 16483 iterations
[ 3913.480214] find_first_bit: 4353244 cycles, 16484 iterations
[ 3913.480216] Start testing find_next_and_bit() with random-filled
bitmap
[ 3913.481074] find_next_and_bit: 89604 cycles, 8216 iterations
[ 3913.481075] Start testing find_bit() with sparse bitmap
[ 3913.481078] find_next_bit: 2536 cycles, 66 iterations
[ 3913.481252] find_next_zero_bit: 344404 cycles, 32703 iterations
[ 3913.481255] find_last_bit: 2006 cycles, 66 iterations
[ 3913.481265] find_first_bit: 17488 cycles, 66 iterations
[ 3913.481266] Start testing find_next_and_bit() with sparse bitmap
[ 3913.481272] find_next_and_bit: 764 cycles, 1 iterations
[3] test_find_next_bit, arm (v7 odroid XU3).
[ 267.206928] Start testing find_bit() with random-filled bitmap
[ 267.214752] find_next_bit: 4474 cycles, 16419 iterations
[ 267.221850] find_next_zero_bit: 5976 cycles, 16350 iterations
[ 267.229294] find_last_bit: 4209 cycles, 16419 iterations
[ 267.279131] find_first_bit: 1032991 cycles, 16420 iterations
[ 267.286265] Start testing find_next_and_bit() with random-filled
bitmap
[ 267.302386] find_next_and_bit: 2290 cycles, 8140 iterations
[ 267.309422] Start testing find_bit() with sparse bitmap
[ 267.316054] find_next_bit: 191 cycles, 66 iterations
[ 267.322726] find_next_zero_bit: 8758 cycles, 32703 iterations
[ 267.329803] find_last_bit: 84 cycles, 66 iterations
[ 267.336169] find_first_bit: 4118 cycles, 66 iterations
[ 267.342627] Start testing find_next_and_bit() with sparse bitmap
[ 267.356919] find_next_and_bit: 91 cycles, 1 iterations
[courbet@google.com: v6]
Link: http://lkml.kernel.org/r/20171129095715.23430-1-courbet@google.com
[geert@linux-m68k.org: m68k/bitops: always include <asm-generic/bitops/find.h>]
Link: http://lkml.kernel.org/r/1512556816-28627-1-git-send-email-geert@linux-m68k.org
Link: http://lkml.kernel.org/r/20171128131334.23491-1-courbet@google.com
Signed-off-by: Clement Courbet <courbet@google.com>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Yury Norov <ynorov@caviumnetworks.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Behaviour of bitmap_fill() differs from bitmap_zero() in a way how bits behind bitmap are handed. bitmap_zero() clears entire bitmap by unsigned long boundary, while bitmap_fill() mimics bitmap_set(). Here we change bitmap_fill() behaviour to be consistent with bitmap_zero() and add a note to documentation. The change might reveal some bugs in the code where unused bits are handled differently and in such cases bitmap_set() has to be used. Link: http://lkml.kernel.org/r/20180109172430.87452-4-andriy.shevchenko@linux.intel.com Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Suggested-by: Rasmus Villemoes <linux@rasmusvillemoes.dk> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Yury Norov <ynorov@caviumnetworks.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
with bitmap_{from,to}_arr32 over the kernel. Additionally to it:
* __check_eq_bitmap() now takes single nbits argument.
* __check_eq_u32_array is not used in new test but may be used in
future. So I don't remove it here, but annotate as __used.
Tested on arm64 and 32-bit BE mips.
[arnd@arndb.de: perf: arm_dsu_pmu: convert to bitmap_from_arr32]
Link: http://lkml.kernel.org/r/20180201172508.5739-2-ynorov@caviumnetworks.com
[ynorov@caviumnetworks.com: fix net/core/ethtool.c]
Link: http://lkml.kernel.org/r/20180205071747.4ekxtsbgxkj5b2fz@yury-thinkpad
Link: http://lkml.kernel.org/r/20171228150019.27953-2-ynorov@caviumnetworks.com
Signed-off-by: Yury Norov <ynorov@caviumnetworks.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: David Decotigny <decot@googlers.com>,
Cc: David S. Miller <davem@davemloft.net>,
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patchset replaces bitmap_{to,from}_u32array with more simple and
standard looking copy-like functions.
bitmap_from_u32array() takes 4 arguments (bitmap_to_u32array is similar):
- unsigned long *bitmap, which is destination;
- unsigned int nbits, the length of destination bitmap, in bits;
- const u32 *buf, the source; and
- unsigned int nwords, the length of source buffer in ints.
In description to the function it is detailed like:
* copy min(nbits, 32*nwords) bits from @buf to @bitmap, remaining
* bits between nword and nbits in @bitmap (if any) are cleared.
Having two size arguments looks unneeded and potentially dangerous.
It is unneeded because normally user of copy-like function should take
care of the size of destination and make it big enough to fit source
data.
And it is dangerous because function may hide possible error if user
doesn't provide big enough bitmap, and data becomes silently dropped.
That's why all copy-like functions have 1 argument for size of copying
data, and I don't see any reason to make bitmap_from_u32array()
different.
One exception that comes in mind is strncpy() which also provides size
of destination in arguments, but it's strongly argued by the possibility
of taking broken strings in source. This is not the case of
bitmap_{from,to}_u32array().
There is no many real users of bitmap_{from,to}_u32array(), and they all
very clearly provide size of destination matched with the size of
source, so additional functionality is not used in fact. Like this:
bitmap_from_u32array(to->link_modes.supported,
__ETHTOOL_LINK_MODE_MASK_NBITS,
link_usettings.link_modes.supported,
__ETHTOOL_LINK_MODE_MASK_NU32);
Where:
#define __ETHTOOL_LINK_MODE_MASK_NU32 \
DIV_ROUND_UP(__ETHTOOL_LINK_MODE_MASK_NBITS, 32)
In this patch, bitmap_copy_safe and bitmap_{from,to}_arr32 are introduced.
'Safe' in bitmap_copy_safe() stands for clearing unused bits in bitmap
beyond last bit till the end of last word. It is useful for hardening
API when bitmap is assumed to be exposed to userspace.
bitmap_{from,to}_arr32 functions are replacements for
bitmap_{from,to}_u32array. They don't take unneeded nwords argument, and
so simpler in implementation and understanding.
This patch suggests optimization for 32-bit systems - aliasing
bitmap_{from,to}_arr32 to bitmap_copy_safe.
Other possible optimization is aliasing 64-bit LE bitmap_{from,to}_arr32 to
more generic function(s). But I didn't end up with the function that would
be helpful by itself, and can be used to alias 64-bit LE
bitmap_{from,to}_arr32, like bitmap_copy_safe() does. So I preferred to
leave things as is.
The following patch switches kernel to new API and introduces test for it.
Discussion is here: https://lkml.org/lkml/2017/11/15/592
[ynorov@caviumnetworks.com: rename bitmap_copy_safe to bitmap_copy_clear_tail]
Link: http://lkml.kernel.org/r/20180201172508.5739-3-ynorov@caviumnetworks.com
Link: http://lkml.kernel.org/r/20171228150019.27953-1-ynorov@caviumnetworks.com
Signed-off-by: Yury Norov <ynorov@caviumnetworks.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: David Decotigny <decot@googlers.com>,
Cc: David S. Miller <davem@davemloft.net>,
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>