* [PATCH v2 1/6] bitmap: fix opencoded bitmap_allocate_region()
2023-08-11 0:57 [PATCH v2 0/6] bitmap: cleanup bitmap_*_region() implementation Yury Norov
@ 2023-08-11 0:57 ` Yury Norov
2023-08-11 0:57 ` [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set() Yury Norov
` (4 subsequent siblings)
5 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 0:57 UTC (permalink / raw)
To: linux-kernel, Andy Shevchenko, Rasmus Villemoes; +Cc: Yury Norov
bitmap_find_region() opencodes bitmap_allocate_region(). Fix it.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
---
lib/bitmap.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/lib/bitmap.c b/lib/bitmap.c
index ddb31015e38a..3a589016f5e0 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -1315,10 +1315,8 @@ int bitmap_find_free_region(unsigned long *bitmap, unsigned int bits, int order)
unsigned int pos, end; /* scans bitmap by regions of size order */
for (pos = 0 ; (end = pos + (1U << order)) <= bits; pos = end) {
- if (!__reg_op(bitmap, pos, order, REG_OP_ISFREE))
- continue;
- __reg_op(bitmap, pos, order, REG_OP_ALLOC);
- return pos;
+ if (!bitmap_allocate_region(bitmap, pos, order))
+ return pos;
}
return -ENOMEM;
}
--
2.39.2
^ permalink raw reply related [flat|nested] 15+ messages in thread* [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 0:57 [PATCH v2 0/6] bitmap: cleanup bitmap_*_region() implementation Yury Norov
2023-08-11 0:57 ` [PATCH v2 1/6] bitmap: fix opencoded bitmap_allocate_region() Yury Norov
@ 2023-08-11 0:57 ` Yury Norov
2023-08-11 6:21 ` Rasmus Villemoes
2023-08-11 9:25 ` Andy Shevchenko
2023-08-11 0:57 ` [PATCH v2 3/6] bitmap: replace _reg_op(REG_OP_RELEASE) with bitmap_clear() Yury Norov
` (3 subsequent siblings)
5 siblings, 2 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 0:57 UTC (permalink / raw)
To: linux-kernel, Andy Shevchenko, Rasmus Villemoes; +Cc: Yury Norov
_reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
---
lib/bitmap.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/lib/bitmap.c b/lib/bitmap.c
index 3a589016f5e0..c9afe704fe4b 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -1352,9 +1352,12 @@ EXPORT_SYMBOL(bitmap_release_region);
*/
int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
{
+ unsigned int nbits = pos + BIT(order);
+
if (!__reg_op(bitmap, pos, order, REG_OP_ISFREE))
return -EBUSY;
- return __reg_op(bitmap, pos, order, REG_OP_ALLOC);
+ bitmap_set(bitmap, pos, nbits);
+ return 0;
}
EXPORT_SYMBOL(bitmap_allocate_region);
--
2.39.2
^ permalink raw reply related [flat|nested] 15+ messages in thread* Re: [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 0:57 ` [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set() Yury Norov
@ 2023-08-11 6:21 ` Rasmus Villemoes
2023-08-11 12:56 ` Yury Norov
2023-08-11 9:25 ` Andy Shevchenko
1 sibling, 1 reply; 15+ messages in thread
From: Rasmus Villemoes @ 2023-08-11 6:21 UTC (permalink / raw)
To: Yury Norov, linux-kernel, Andy Shevchenko
On 11/08/2023 02.57, Yury Norov wrote:
> _reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
>
> Signed-off-by: Yury Norov <yury.norov@gmail.com>
> ---
> lib/bitmap.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/lib/bitmap.c b/lib/bitmap.c
> index 3a589016f5e0..c9afe704fe4b 100644
> --- a/lib/bitmap.c
> +++ b/lib/bitmap.c
> @@ -1352,9 +1352,12 @@ EXPORT_SYMBOL(bitmap_release_region);
> */
> int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
> {
> + unsigned int nbits = pos + BIT(order);
> +
That really doesn't sound right. Have you added self-tests for these
functions first and then used those to catch regressions?
Rasmus
^ permalink raw reply [flat|nested] 15+ messages in thread* Re: [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 6:21 ` Rasmus Villemoes
@ 2023-08-11 12:56 ` Yury Norov
2023-08-11 13:31 ` Rasmus Villemoes
0 siblings, 1 reply; 15+ messages in thread
From: Yury Norov @ 2023-08-11 12:56 UTC (permalink / raw)
To: Rasmus Villemoes; +Cc: linux-kernel, Andy Shevchenko
On Fri, Aug 11, 2023 at 08:21:33AM +0200, Rasmus Villemoes wrote:
> On 11/08/2023 02.57, Yury Norov wrote:
> > _reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
> >
> > Signed-off-by: Yury Norov <yury.norov@gmail.com>
> > ---
> > lib/bitmap.c | 5 ++++-
> > 1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/lib/bitmap.c b/lib/bitmap.c
> > index 3a589016f5e0..c9afe704fe4b 100644
> > --- a/lib/bitmap.c
> > +++ b/lib/bitmap.c
> > @@ -1352,9 +1352,12 @@ EXPORT_SYMBOL(bitmap_release_region);
> > */
> > int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
> > {
> > + unsigned int nbits = pos + BIT(order);
> > +
>
> That really doesn't sound right. Have you added self-tests for these
> functions first and then used those to catch regressions?
When bitmap_allocate_region() is broken, almost every arch build fails
to boot. Can you explain what exactly looks wrong to you?
Thanks,
Yury
^ permalink raw reply [flat|nested] 15+ messages in thread* Re: [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 12:56 ` Yury Norov
@ 2023-08-11 13:31 ` Rasmus Villemoes
2023-08-14 0:39 ` Yury Norov
0 siblings, 1 reply; 15+ messages in thread
From: Rasmus Villemoes @ 2023-08-11 13:31 UTC (permalink / raw)
To: Yury Norov; +Cc: linux-kernel, Andy Shevchenko
On 11/08/2023 14.56, Yury Norov wrote:
> On Fri, Aug 11, 2023 at 08:21:33AM +0200, Rasmus Villemoes wrote:
>> On 11/08/2023 02.57, Yury Norov wrote:
>>> _reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
>>>
>>> Signed-off-by: Yury Norov <yury.norov@gmail.com>
>>> ---
>>> lib/bitmap.c | 5 ++++-
>>> 1 file changed, 4 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/lib/bitmap.c b/lib/bitmap.c
>>> index 3a589016f5e0..c9afe704fe4b 100644
>>> --- a/lib/bitmap.c
>>> +++ b/lib/bitmap.c
>>> @@ -1352,9 +1352,12 @@ EXPORT_SYMBOL(bitmap_release_region);
>>> */
>>> int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
>>> {
>>> + unsigned int nbits = pos + BIT(order);
>>> +
>>
>> That really doesn't sound right. Have you added self-tests for these
>> functions first and then used those to catch regressions?
>
> When bitmap_allocate_region() is broken, almost every arch build fails
> to boot. Can you explain what exactly looks wrong to you?
The number of bits we are about to set should not be [position in bitmap
to start from] + [2^order]. The second half of that patch was
- return __reg_op(bitmap, pos, order, REG_OP_ALLOC);
+ bitmap_set(bitmap, pos, nbits);
+ return 0;
so instead of setting 1<<nbits starting at pos, you're now setting
pos+(1<<nbits) starting at pos. How is that correct?
Rasmus
^ permalink raw reply [flat|nested] 15+ messages in thread* Re: [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 13:31 ` Rasmus Villemoes
@ 2023-08-14 0:39 ` Yury Norov
0 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-14 0:39 UTC (permalink / raw)
To: Rasmus Villemoes; +Cc: linux-kernel, Andy Shevchenko
On Fri, Aug 11, 2023 at 03:31:01PM +0200, Rasmus Villemoes wrote:
> On 11/08/2023 14.56, Yury Norov wrote:
> > On Fri, Aug 11, 2023 at 08:21:33AM +0200, Rasmus Villemoes wrote:
> >> On 11/08/2023 02.57, Yury Norov wrote:
> >>> _reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
> >>>
> >>> Signed-off-by: Yury Norov <yury.norov@gmail.com>
> >>> ---
> >>> lib/bitmap.c | 5 ++++-
> >>> 1 file changed, 4 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/lib/bitmap.c b/lib/bitmap.c
> >>> index 3a589016f5e0..c9afe704fe4b 100644
> >>> --- a/lib/bitmap.c
> >>> +++ b/lib/bitmap.c
> >>> @@ -1352,9 +1352,12 @@ EXPORT_SYMBOL(bitmap_release_region);
> >>> */
> >>> int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
> >>> {
> >>> + unsigned int nbits = pos + BIT(order);
> >>> +
> >>
> >> That really doesn't sound right. Have you added self-tests for these
> >> functions first and then used those to catch regressions?
> >
> > When bitmap_allocate_region() is broken, almost every arch build fails
> > to boot. Can you explain what exactly looks wrong to you?
>
> The number of bits we are about to set should not be [position in bitmap
> to start from] + [2^order]. The second half of that patch was
>
> - return __reg_op(bitmap, pos, order, REG_OP_ALLOC);
> + bitmap_set(bitmap, pos, nbits);
> + return 0;
>
> so instead of setting 1<<nbits starting at pos, you're now setting
> pos+(1<<nbits) starting at pos. How is that correct?
You're right. Bad on me. :/
I'll send v3 with all the discussed cleared shortly.
Thanks,
YUry
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 0:57 ` [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set() Yury Norov
2023-08-11 6:21 ` Rasmus Villemoes
@ 2023-08-11 9:25 ` Andy Shevchenko
2023-08-11 13:01 ` Yury Norov
1 sibling, 1 reply; 15+ messages in thread
From: Andy Shevchenko @ 2023-08-11 9:25 UTC (permalink / raw)
To: Yury Norov; +Cc: linux-kernel, Rasmus Villemoes
On Thu, Aug 10, 2023 at 05:57:28PM -0700, Yury Norov wrote:
> _reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
...
> int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
> {
> + unsigned int nbits = pos + BIT(order);
Shouldn't this be unsigned long?
As the prototype of the used later unsigned long find_next_bit().
> if (!__reg_op(bitmap, pos, order, REG_OP_ISFREE))
> return -EBUSY;
> - return __reg_op(bitmap, pos, order, REG_OP_ALLOC);
> + bitmap_set(bitmap, pos, nbits);
> + return 0;
> }
--
With Best Regards,
Andy Shevchenko
^ permalink raw reply [flat|nested] 15+ messages in thread* Re: [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set()
2023-08-11 9:25 ` Andy Shevchenko
@ 2023-08-11 13:01 ` Yury Norov
0 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 13:01 UTC (permalink / raw)
To: Andy Shevchenko; +Cc: linux-kernel, Rasmus Villemoes
On Fri, Aug 11, 2023 at 12:25:02PM +0300, Andy Shevchenko wrote:
> On Thu, Aug 10, 2023 at 05:57:28PM -0700, Yury Norov wrote:
> > _reg_op(REG_OP_ALLOC) duplicates bitmap_set(). Fix it.
>
> ...
>
> > int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
> > {
> > + unsigned int nbits = pos + BIT(order);
>
> Shouldn't this be unsigned long?
> As the prototype of the used later unsigned long find_next_bit().
Linux doesn't support bitmaps longer than 2^32 bits. I really doubt
such huge bitmaps would ever exist, but if that will happen, too many
bitmap functions will have to be revisited.
For the reference, check the code of __reg_op(), which I remove in the
series - all indexes and offsets are int's there.
Thanks,
Yury
> > if (!__reg_op(bitmap, pos, order, REG_OP_ISFREE))
> > return -EBUSY;
> > - return __reg_op(bitmap, pos, order, REG_OP_ALLOC);
> > + bitmap_set(bitmap, pos, nbits);
> > + return 0;
> > }
>
> --
> With Best Regards,
> Andy Shevchenko
>
^ permalink raw reply [flat|nested] 15+ messages in thread
* [PATCH v2 3/6] bitmap: replace _reg_op(REG_OP_RELEASE) with bitmap_clear()
2023-08-11 0:57 [PATCH v2 0/6] bitmap: cleanup bitmap_*_region() implementation Yury Norov
2023-08-11 0:57 ` [PATCH v2 1/6] bitmap: fix opencoded bitmap_allocate_region() Yury Norov
2023-08-11 0:57 ` [PATCH v2 2/6] bitmap: replace _reg_op(REG_OP_ALLOC) with bitmap_set() Yury Norov
@ 2023-08-11 0:57 ` Yury Norov
2023-08-11 0:57 ` [PATCH v2 4/6] bitmap: replace _reg_op(REG_OP_ISFREE) with find_next_bit() Yury Norov
` (2 subsequent siblings)
5 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 0:57 UTC (permalink / raw)
To: linux-kernel, Andy Shevchenko, Rasmus Villemoes; +Cc: Yury Norov
_reg_op(REG_OP_RELEASE) duplicates bitmap_clear(). Fix it.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
---
lib/bitmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/bitmap.c b/lib/bitmap.c
index c9afe704fe4b..0ffafc41cd56 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -1335,7 +1335,7 @@ EXPORT_SYMBOL(bitmap_find_free_region);
*/
void bitmap_release_region(unsigned long *bitmap, unsigned int pos, int order)
{
- __reg_op(bitmap, pos, order, REG_OP_RELEASE);
+ bitmap_clear(bitmap, pos, pos + BIT(order));
}
EXPORT_SYMBOL(bitmap_release_region);
--
2.39.2
^ permalink raw reply related [flat|nested] 15+ messages in thread* [PATCH v2 4/6] bitmap: replace _reg_op(REG_OP_ISFREE) with find_next_bit()
2023-08-11 0:57 [PATCH v2 0/6] bitmap: cleanup bitmap_*_region() implementation Yury Norov
` (2 preceding siblings ...)
2023-08-11 0:57 ` [PATCH v2 3/6] bitmap: replace _reg_op(REG_OP_RELEASE) with bitmap_clear() Yury Norov
@ 2023-08-11 0:57 ` Yury Norov
2023-08-11 0:57 ` [PATCH v2 5/6] bitmap: drop _reg_op() function Yury Norov
2023-08-11 0:57 ` [PATCH v2 6/6] bitmap: move bitmap_*_region functions to bitmap.h Yury Norov
5 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 0:57 UTC (permalink / raw)
To: linux-kernel, Andy Shevchenko, Rasmus Villemoes; +Cc: Yury Norov
_reg_op(REG_OP_ISFREE) can be trivially replaced with find_next_bit().
Doing that opens room for potential small_const_nbits()
optimization.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
---
lib/bitmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/bitmap.c b/lib/bitmap.c
index 0ffafc41cd56..d4560acbe077 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -1354,7 +1354,7 @@ int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
{
unsigned int nbits = pos + BIT(order);
- if (!__reg_op(bitmap, pos, order, REG_OP_ISFREE))
+ if (find_next_bit(bitmap, pos, nbits) < nbits)
return -EBUSY;
bitmap_set(bitmap, pos, nbits);
return 0;
--
2.39.2
^ permalink raw reply related [flat|nested] 15+ messages in thread* [PATCH v2 5/6] bitmap: drop _reg_op() function
2023-08-11 0:57 [PATCH v2 0/6] bitmap: cleanup bitmap_*_region() implementation Yury Norov
` (3 preceding siblings ...)
2023-08-11 0:57 ` [PATCH v2 4/6] bitmap: replace _reg_op(REG_OP_ISFREE) with find_next_bit() Yury Norov
@ 2023-08-11 0:57 ` Yury Norov
2023-08-11 0:57 ` [PATCH v2 6/6] bitmap: move bitmap_*_region functions to bitmap.h Yury Norov
5 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 0:57 UTC (permalink / raw)
To: linux-kernel, Andy Shevchenko, Rasmus Villemoes; +Cc: Yury Norov
Now that all _reg_op() users are switched alternative functions, _reg_op()
machinery is not needed anymore.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
---
lib/bitmap.c | 76 ----------------------------------------------------
1 file changed, 76 deletions(-)
diff --git a/lib/bitmap.c b/lib/bitmap.c
index d4560acbe077..ae619523c3ec 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -1220,82 +1220,6 @@ void bitmap_fold(unsigned long *dst, const unsigned long *orig,
}
#endif /* CONFIG_NUMA */
-/*
- * Common code for bitmap_*_region() routines.
- * bitmap: array of unsigned longs corresponding to the bitmap
- * pos: the beginning of the region
- * order: region size (log base 2 of number of bits)
- * reg_op: operation(s) to perform on that region of bitmap
- *
- * Can set, verify and/or release a region of bits in a bitmap,
- * depending on which combination of REG_OP_* flag bits is set.
- *
- * A region of a bitmap is a sequence of bits in the bitmap, of
- * some size '1 << order' (a power of two), aligned to that same
- * '1 << order' power of two.
- *
- * Returns 1 if REG_OP_ISFREE succeeds (region is all zero bits).
- * Returns 0 in all other cases and reg_ops.
- */
-
-enum {
- REG_OP_ISFREE, /* true if region is all zero bits */
- REG_OP_ALLOC, /* set all bits in region */
- REG_OP_RELEASE, /* clear all bits in region */
-};
-
-static int __reg_op(unsigned long *bitmap, unsigned int pos, int order, int reg_op)
-{
- int nbits_reg; /* number of bits in region */
- int index; /* index first long of region in bitmap */
- int offset; /* bit offset region in bitmap[index] */
- int nlongs_reg; /* num longs spanned by region in bitmap */
- int nbitsinlong; /* num bits of region in each spanned long */
- unsigned long mask; /* bitmask for one long of region */
- int i; /* scans bitmap by longs */
- int ret = 0; /* return value */
-
- /*
- * Either nlongs_reg == 1 (for small orders that fit in one long)
- * or (offset == 0 && mask == ~0UL) (for larger multiword orders.)
- */
- nbits_reg = 1 << order;
- index = pos / BITS_PER_LONG;
- offset = pos - (index * BITS_PER_LONG);
- nlongs_reg = BITS_TO_LONGS(nbits_reg);
- nbitsinlong = min(nbits_reg, BITS_PER_LONG);
-
- /*
- * Can't do "mask = (1UL << nbitsinlong) - 1", as that
- * overflows if nbitsinlong == BITS_PER_LONG.
- */
- mask = (1UL << (nbitsinlong - 1));
- mask += mask - 1;
- mask <<= offset;
-
- switch (reg_op) {
- case REG_OP_ISFREE:
- for (i = 0; i < nlongs_reg; i++) {
- if (bitmap[index + i] & mask)
- goto done;
- }
- ret = 1; /* all bits in region free (zero) */
- break;
-
- case REG_OP_ALLOC:
- for (i = 0; i < nlongs_reg; i++)
- bitmap[index + i] |= mask;
- break;
-
- case REG_OP_RELEASE:
- for (i = 0; i < nlongs_reg; i++)
- bitmap[index + i] &= ~mask;
- break;
- }
-done:
- return ret;
-}
-
/**
* bitmap_find_free_region - find a contiguous aligned mem region
* @bitmap: array of unsigned longs corresponding to the bitmap
--
2.39.2
^ permalink raw reply related [flat|nested] 15+ messages in thread* [PATCH v2 6/6] bitmap: move bitmap_*_region functions to bitmap.h
2023-08-11 0:57 [PATCH v2 0/6] bitmap: cleanup bitmap_*_region() implementation Yury Norov
` (4 preceding siblings ...)
2023-08-11 0:57 ` [PATCH v2 5/6] bitmap: drop _reg_op() function Yury Norov
@ 2023-08-11 0:57 ` Yury Norov
2023-08-11 9:27 ` Andy Shevchenko
5 siblings, 1 reply; 15+ messages in thread
From: Yury Norov @ 2023-08-11 0:57 UTC (permalink / raw)
To: linux-kernel, Andy Shevchenko, Rasmus Villemoes; +Cc: Yury Norov
Now that bitmap_*_region() functions are implemented as thin wrappers
around others, it's worth to move them to the header, as it opens room
for compile-time optimizations.
Signed-off-by: Yury Norov <yury.norov@gmail.com>
---
include/linux/bitmap.h | 65 ++++++++++++++++++++++++++++++++++++++++--
lib/bitmap.c | 65 ------------------------------------------
2 files changed, 62 insertions(+), 68 deletions(-)
diff --git a/include/linux/bitmap.h b/include/linux/bitmap.h
index 03644237e1ef..aa33fc290619 100644
--- a/include/linux/bitmap.h
+++ b/include/linux/bitmap.h
@@ -216,9 +216,6 @@ void bitmap_onto(unsigned long *dst, const unsigned long *orig,
const unsigned long *relmap, unsigned int bits);
void bitmap_fold(unsigned long *dst, const unsigned long *orig,
unsigned int sz, unsigned int nbits);
-int bitmap_find_free_region(unsigned long *bitmap, unsigned int bits, int order);
-void bitmap_release_region(unsigned long *bitmap, unsigned int pos, int order);
-int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order);
#ifdef __BIG_ENDIAN
void bitmap_copy_le(unsigned long *dst, const unsigned long *src, unsigned int nbits);
@@ -599,6 +596,68 @@ static inline void bitmap_set_value8(unsigned long *map, unsigned long value,
map[index] |= value << offset;
}
+/**
+ * bitmap_allocate_region - allocate bitmap region
+ * @bitmap: array of unsigned longs corresponding to the bitmap
+ * @pos: beginning of bit region to allocate
+ * @order: region size (log base 2 of number of bits) to allocate
+ *
+ * Allocate (set bits in) a specified region of a bitmap.
+ *
+ * Return 0 on success, or %-EBUSY if specified region wasn't
+ * free (not all bits were zero).
+ */
+static inline int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
+{
+ unsigned int nbits = pos + BIT(order);
+
+ if (find_next_bit(bitmap, pos, nbits) < nbits)
+ return -EBUSY;
+ bitmap_set(bitmap, pos, nbits);
+ return 0;
+}
+
+/**
+ * bitmap_find_free_region - find a contiguous aligned mem region
+ * @bitmap: array of unsigned longs corresponding to the bitmap
+ * @bits: number of bits in the bitmap
+ * @order: region size (log base 2 of number of bits) to find
+ *
+ * Find a region of free (zero) bits in a @bitmap of @bits bits and
+ * allocate them (set them to one). Only consider regions of length
+ * a power (@order) of two, aligned to that power of two, which
+ * makes the search algorithm much faster.
+ *
+ * Return the bit offset in bitmap of the allocated region,
+ * or -errno on failure.
+ */
+static inline int bitmap_find_free_region(unsigned long *bitmap, unsigned int bits, int order)
+{
+ unsigned int pos, end; /* scans bitmap by regions of size order */
+
+ for (pos = 0 ; (end = pos + BIT(order)) <= bits; pos = end)
+ if (!bitmap_allocate_region(bitmap, pos, order))
+ return pos;
+
+ return -ENOMEM;
+}
+
+/**
+ * bitmap_release_region - release allocated bitmap region
+ * @bitmap: array of unsigned longs corresponding to the bitmap
+ * @pos: beginning of bit region to release
+ * @order: region size (log base 2 of number of bits) to release
+ *
+ * This is the complement to __bitmap_find_free_region() and releases
+ * the found region (by clearing it in the bitmap).
+ *
+ * No return value.
+ */
+static inline void bitmap_release_region(unsigned long *bitmap, unsigned int pos, int order)
+{
+ bitmap_clear(bitmap, pos, pos + BIT(order));
+}
+
#endif /* __ASSEMBLY__ */
#endif /* __LINUX_BITMAP_H */
diff --git a/lib/bitmap.c b/lib/bitmap.c
index ae619523c3ec..1c5d1f5d2071 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -1220,71 +1220,6 @@ void bitmap_fold(unsigned long *dst, const unsigned long *orig,
}
#endif /* CONFIG_NUMA */
-/**
- * bitmap_find_free_region - find a contiguous aligned mem region
- * @bitmap: array of unsigned longs corresponding to the bitmap
- * @bits: number of bits in the bitmap
- * @order: region size (log base 2 of number of bits) to find
- *
- * Find a region of free (zero) bits in a @bitmap of @bits bits and
- * allocate them (set them to one). Only consider regions of length
- * a power (@order) of two, aligned to that power of two, which
- * makes the search algorithm much faster.
- *
- * Return the bit offset in bitmap of the allocated region,
- * or -errno on failure.
- */
-int bitmap_find_free_region(unsigned long *bitmap, unsigned int bits, int order)
-{
- unsigned int pos, end; /* scans bitmap by regions of size order */
-
- for (pos = 0 ; (end = pos + (1U << order)) <= bits; pos = end) {
- if (!bitmap_allocate_region(bitmap, pos, order))
- return pos;
- }
- return -ENOMEM;
-}
-EXPORT_SYMBOL(bitmap_find_free_region);
-
-/**
- * bitmap_release_region - release allocated bitmap region
- * @bitmap: array of unsigned longs corresponding to the bitmap
- * @pos: beginning of bit region to release
- * @order: region size (log base 2 of number of bits) to release
- *
- * This is the complement to __bitmap_find_free_region() and releases
- * the found region (by clearing it in the bitmap).
- *
- * No return value.
- */
-void bitmap_release_region(unsigned long *bitmap, unsigned int pos, int order)
-{
- bitmap_clear(bitmap, pos, pos + BIT(order));
-}
-EXPORT_SYMBOL(bitmap_release_region);
-
-/**
- * bitmap_allocate_region - allocate bitmap region
- * @bitmap: array of unsigned longs corresponding to the bitmap
- * @pos: beginning of bit region to allocate
- * @order: region size (log base 2 of number of bits) to allocate
- *
- * Allocate (set bits in) a specified region of a bitmap.
- *
- * Return 0 on success, or %-EBUSY if specified region wasn't
- * free (not all bits were zero).
- */
-int bitmap_allocate_region(unsigned long *bitmap, unsigned int pos, int order)
-{
- unsigned int nbits = pos + BIT(order);
-
- if (find_next_bit(bitmap, pos, nbits) < nbits)
- return -EBUSY;
- bitmap_set(bitmap, pos, nbits);
- return 0;
-}
-EXPORT_SYMBOL(bitmap_allocate_region);
-
/**
* bitmap_copy_le - copy a bitmap, putting the bits into little-endian order.
* @dst: destination buffer
--
2.39.2
^ permalink raw reply related [flat|nested] 15+ messages in thread* Re: [PATCH v2 6/6] bitmap: move bitmap_*_region functions to bitmap.h
2023-08-11 0:57 ` [PATCH v2 6/6] bitmap: move bitmap_*_region functions to bitmap.h Yury Norov
@ 2023-08-11 9:27 ` Andy Shevchenko
2023-08-11 13:05 ` Yury Norov
0 siblings, 1 reply; 15+ messages in thread
From: Andy Shevchenko @ 2023-08-11 9:27 UTC (permalink / raw)
To: Yury Norov; +Cc: linux-kernel, Rasmus Villemoes
On Thu, Aug 10, 2023 at 05:57:32PM -0700, Yury Norov wrote:
> Now that bitmap_*_region() functions are implemented as thin wrappers
> around others, it's worth to move them to the header, as it opens room
> for compile-time optimizations.
...
> + * Return 0 on success, or %-EBUSY if specified region wasn't
> + * free (not all bits were zero).
Run
scripts/kernel-doc -v -none -Wall
against this
...
> + * Return the bit offset in bitmap of the allocated region,
> + * or -errno on failure.
Ditto.
...
> +static inline int bitmap_find_free_region(unsigned long *bitmap, unsigned int bits, int order)
> +{
> + unsigned int pos, end; /* scans bitmap by regions of size order */
> +
> + for (pos = 0 ; (end = pos + BIT(order)) <= bits; pos = end)
Extra space.
> + if (!bitmap_allocate_region(bitmap, pos, order))
> + return pos;
> +
> + return -ENOMEM;
> +}
...
> +/**
> + * bitmap_release_region - release allocated bitmap region
> + * @bitmap: array of unsigned longs corresponding to the bitmap
> + * @pos: beginning of bit region to release
> + * @order: region size (log base 2 of number of bits) to release
> + *
> + * This is the complement to __bitmap_find_free_region() and releases
> + * the found region (by clearing it in the bitmap).
> + * No return value.
Useless line.
> + */
...
Seems like the original code has all these, perhaps update in a separate patch?
--
With Best Regards,
Andy Shevchenko
^ permalink raw reply [flat|nested] 15+ messages in thread* Re: [PATCH v2 6/6] bitmap: move bitmap_*_region functions to bitmap.h
2023-08-11 9:27 ` Andy Shevchenko
@ 2023-08-11 13:05 ` Yury Norov
0 siblings, 0 replies; 15+ messages in thread
From: Yury Norov @ 2023-08-11 13:05 UTC (permalink / raw)
To: Andy Shevchenko; +Cc: linux-kernel, Rasmus Villemoes
On Fri, Aug 11, 2023 at 12:27:51PM +0300, Andy Shevchenko wrote:
> On Thu, Aug 10, 2023 at 05:57:32PM -0700, Yury Norov wrote:
> > Now that bitmap_*_region() functions are implemented as thin wrappers
> > around others, it's worth to move them to the header, as it opens room
> > for compile-time optimizations.
>
> ...
>
> > + * Return 0 on success, or %-EBUSY if specified region wasn't
> > + * free (not all bits were zero).
>
> Run
>
> scripts/kernel-doc -v -none -Wall
> ...
>
> Seems like the original code has all these, perhaps update in a separate patch?
Yes. This patch is named _move_, which means - I don't touch the code.
I already replaced (1U << order) to BIT(order), and I'm not happy with
that. We need to do that in a separate patch.
Thanks,
Yury
^ permalink raw reply [flat|nested] 15+ messages in thread