Re: [PATCH v10 03/14] xen/bitops: implement fls{l}() in common logic

2024-05-23 Thread Julien Grall 

Hi,

On 22/05/2024 09:15, Jan Beulich wrote:

On 22.05.2024 09:37, Oleksii K. wrote:

On Tue, 2024-05-21 at 13:18 +0200, Jan Beulich wrote:

On 17.05.2024 15:54, Oleksii Kurochko wrote:

To avoid the compilation error below, it is needed to update to
places
in common/page_alloc.c where flsl() is used as now flsl() returns
unsigned int:

./include/xen/kernel.h:18:21: error: comparison of distinct pointer
types lacks a cast [-Werror]
    18 | (void) (&_x == &_y);    \
   | ^~
     common/page_alloc.c:1843:34: note: in expansion of macro 'min'
  1843 | unsigned int inc_order = min(MAX_ORDER, flsl(e
- s) - 1);

generic_fls{l} was used instead of __builtin_clz{l}(x) as if x is
0,
the result in undefined.

The prototype of the per-architecture fls{l}() functions was
changed to
return 'unsigned int' to align with the generic implementation of
these
functions and avoid introducing signed/unsigned mismatches.

Signed-off-by: Oleksii Kurochko 
---
  The patch is almost independent from Andrew's patch series
  (
https://lore.kernel.org/xen-devel/20240313172716.2325427-1-andrew.coop...@citrix.com/T/#t
)
  except test_fls() function which IMO can be merged as a separate
patch after Andrew's patch
  will be fully ready.


If there wasn't this dependency (I don't think it's "almost
independent"),
I'd be offering R-b with again one nit below.


Aren't all changes, except those in xen/common/bitops.c, independent? I
could move these changes in xen/common/bitops.c to a separate commit. I
think it is safe to commit them ( an introduction of common logic for
fls{l}() and tests ) separately since the CI tests have passed.


Technically they might be, but contextually there are further conflicts.
Just try "patch --dry-run" on top of a plain staging tree. You really
need to settle, perhaps consulting Andrew, whether you want to go on top
of his change, or ahead of it. I'm not willing to approve a patch that's
presented one way but then is (kind of) expected to go in the other way.


I agree with what Jan wrote. I don't have any strong opinion on which 
order they should be merged. But, if your series is intended to be 
merged before Andrew's one then please rebase to vanilla staging.


I looked at the rest of the patch and it LGTM.

Cheers,

--
Julien Grall

Re: [PATCH v10 03/14] xen/bitops: implement fls{l}() in common logic

2024-05-22 Thread Jan Beulich 
On 22.05.2024 09:37, Oleksii K. wrote:
> On Tue, 2024-05-21 at 13:18 +0200, Jan Beulich wrote:
>> On 17.05.2024 15:54, Oleksii Kurochko wrote:
>>> To avoid the compilation error below, it is needed to update to
>>> places
>>> in common/page_alloc.c where flsl() is used as now flsl() returns
>>> unsigned int:
>>>
>>> ./include/xen/kernel.h:18:21: error: comparison of distinct pointer
>>> types lacks a cast [-Werror]
>>>    18 | (void) (&_x == &_y);    \
>>>   | ^~
>>>     common/page_alloc.c:1843:34: note: in expansion of macro 'min'
>>>  1843 | unsigned int inc_order = min(MAX_ORDER, flsl(e
>>> - s) - 1);
>>>
>>> generic_fls{l} was used instead of __builtin_clz{l}(x) as if x is
>>> 0,
>>> the result in undefined.
>>>
>>> The prototype of the per-architecture fls{l}() functions was
>>> changed to
>>> return 'unsigned int' to align with the generic implementation of
>>> these
>>> functions and avoid introducing signed/unsigned mismatches.
>>>
>>> Signed-off-by: Oleksii Kurochko 
>>> ---
>>>  The patch is almost independent from Andrew's patch series
>>>  (
>>> https://lore.kernel.org/xen-devel/20240313172716.2325427-1-andrew.coop...@citrix.com/T/#t
>>> )
>>>  except test_fls() function which IMO can be merged as a separate
>>> patch after Andrew's patch
>>>  will be fully ready.
>>
>> If there wasn't this dependency (I don't think it's "almost
>> independent"),
>> I'd be offering R-b with again one nit below.
> 
> Aren't all changes, except those in xen/common/bitops.c, independent? I
> could move these changes in xen/common/bitops.c to a separate commit. I
> think it is safe to commit them ( an introduction of common logic for
> fls{l}() and tests ) separately since the CI tests have passed.

Technically they might be, but contextually there are further conflicts.
Just try "patch --dry-run" on top of a plain staging tree. You really
need to settle, perhaps consulting Andrew, whether you want to go on top
of his change, or ahead of it. I'm not willing to approve a patch that's
presented one way but then is (kind of) expected to go in the other way.

Jan

Re: [PATCH v10 03/14] xen/bitops: implement fls{l}() in common logic

2024-05-22 Thread Oleksii K. 
On Tue, 2024-05-21 at 13:18 +0200, Jan Beulich wrote:
> On 17.05.2024 15:54, Oleksii Kurochko wrote:
> > To avoid the compilation error below, it is needed to update to
> > places
> > in common/page_alloc.c where flsl() is used as now flsl() returns
> > unsigned int:
> > 
> > ./include/xen/kernel.h:18:21: error: comparison of distinct pointer
> > types lacks a cast [-Werror]
> >    18 | (void) (&_x == &_y);    \
> >   | ^~
> >     common/page_alloc.c:1843:34: note: in expansion of macro 'min'
> >  1843 | unsigned int inc_order = min(MAX_ORDER, flsl(e
> > - s) - 1);
> > 
> > generic_fls{l} was used instead of __builtin_clz{l}(x) as if x is
> > 0,
> > the result in undefined.
> > 
> > The prototype of the per-architecture fls{l}() functions was
> > changed to
> > return 'unsigned int' to align with the generic implementation of
> > these
> > functions and avoid introducing signed/unsigned mismatches.
> > 
> > Signed-off-by: Oleksii Kurochko 
> > ---
> >  The patch is almost independent from Andrew's patch series
> >  (
> > https://lore.kernel.org/xen-devel/20240313172716.2325427-1-andrew.coop...@citrix.com/T/#t
> > )
> >  except test_fls() function which IMO can be merged as a separate
> > patch after Andrew's patch
> >  will be fully ready.
> 
> If there wasn't this dependency (I don't think it's "almost
> independent"),
> I'd be offering R-b with again one nit below.

Aren't all changes, except those in xen/common/bitops.c, independent? I
could move these changes in xen/common/bitops.c to a separate commit. I
think it is safe to commit them ( an introduction of common logic for
fls{l}() and tests ) separately since the CI tests have passed.

~ Oleksii

> 
> > --- a/xen/arch/x86/include/asm/bitops.h
> > +++ b/xen/arch/x86/include/asm/bitops.h
> > @@ -425,20 +425,21 @@ static always_inline unsigned int
> > arch_ffsl(unsigned long x)
> >   *
> >   * This is defined the same way as ffs.
> >   */
> > -static inline int flsl(unsigned long x)
> > +static always_inline unsigned int arch_flsl(unsigned long x)
> >  {
> > -    long r;
> > +    unsigned long r;
> >  
> >  asm ( "bsr %1,%0\n\t"
> >    "jnz 1f\n\t"
> >    "mov $-1,%0\n"
> >    "1:" : "=r" (r) : "rm" (x));
> > -    return (int)r+1;
> > +    return (unsigned int)r+1;
> 
> Since you now touch this, you'd better tidy it at the same time:
> 
>     return r + 1;
> 
> (i.e. style and no need for a cast).
> 
> Jan

Re: [PATCH v10 03/14] xen/bitops: implement fls{l}() in common logic

2024-05-21 Thread Jan Beulich 
On 17.05.2024 15:54, Oleksii Kurochko wrote:
> To avoid the compilation error below, it is needed to update to places
> in common/page_alloc.c where flsl() is used as now flsl() returns unsigned 
> int:
> 
> ./include/xen/kernel.h:18:21: error: comparison of distinct pointer types 
> lacks a cast [-Werror]
>18 | (void) (&_x == &_y);\
>   | ^~
> common/page_alloc.c:1843:34: note: in expansion of macro 'min'
>  1843 | unsigned int inc_order = min(MAX_ORDER, flsl(e - s) - 1);
> 
> generic_fls{l} was used instead of __builtin_clz{l}(x) as if x is 0,
> the result in undefined.
> 
> The prototype of the per-architecture fls{l}() functions was changed to
> return 'unsigned int' to align with the generic implementation of these
> functions and avoid introducing signed/unsigned mismatches.
> 
> Signed-off-by: Oleksii Kurochko 
> ---
>  The patch is almost independent from Andrew's patch series
>  ( 
> https://lore.kernel.org/xen-devel/20240313172716.2325427-1-andrew.coop...@citrix.com/T/#t)
>  except test_fls() function which IMO can be merged as a separate patch after 
> Andrew's patch
>  will be fully ready.

If there wasn't this dependency (I don't think it's "almost independent"),
I'd be offering R-b with again one nit below.

> --- a/xen/arch/x86/include/asm/bitops.h
> +++ b/xen/arch/x86/include/asm/bitops.h
> @@ -425,20 +425,21 @@ static always_inline unsigned int arch_ffsl(unsigned 
> long x)
>   *
>   * This is defined the same way as ffs.
>   */
> -static inline int flsl(unsigned long x)
> +static always_inline unsigned int arch_flsl(unsigned long x)
>  {
> -long r;
> +unsigned long r;
>  
>  asm ( "bsr %1,%0\n\t"
>"jnz 1f\n\t"
>"mov $-1,%0\n"
>"1:" : "=r" (r) : "rm" (x));
> -return (int)r+1;
> +return (unsigned int)r+1;

Since you now touch this, you'd better tidy it at the same time:

return r + 1;

(i.e. style and no need for a cast).

Jan

[PATCH v10 03/14] xen/bitops: implement fls{l}() in common logic

2024-05-17 Thread Oleksii Kurochko 
To avoid the compilation error below, it is needed to update to places
in common/page_alloc.c where flsl() is used as now flsl() returns unsigned int:

./include/xen/kernel.h:18:21: error: comparison of distinct pointer types lacks 
a cast [-Werror]
   18 | (void) (&_x == &_y);\
  | ^~
common/page_alloc.c:1843:34: note: in expansion of macro 'min'
 1843 | unsigned int inc_order = min(MAX_ORDER, flsl(e - s) - 1);

generic_fls{l} was used instead of __builtin_clz{l}(x) as if x is 0,
the result in undefined.

The prototype of the per-architecture fls{l}() functions was changed to
return 'unsigned int' to align with the generic implementation of these
functions and avoid introducing signed/unsigned mismatches.

Signed-off-by: Oleksii Kurochko 
---
 The patch is almost independent from Andrew's patch series
 ( 
https://lore.kernel.org/xen-devel/20240313172716.2325427-1-andrew.coop...@citrix.com/T/#t)
 except test_fls() function which IMO can be merged as a separate patch after 
Andrew's patch
 will be fully ready.
---
Changes in V10:
 - update return type of arch_flsl() across arcitectures to 'unsigned int' to 
be aligned
   with return type of generic flsl() in xen/bitops.h.
 - switch inline to always_inline for arch_flsl() across architectures to be in 
sync
   with other similar changes.
 - define arch_flsl as arch_fls not just only fls.
 - update the commit message ( add information that per-arch fls{l)() protypes 
were
   changed ).
---
Changes in V9:
 - update return type of fls and flsl() to unsigned int to be aligned with other
   bit ops.
 - update places where return value of fls() and flsl() is compared with int.
 - update the commit message.
---
Changes in V8:
 - do proper rebase: back definition of fls{l} to the current patch.
 - drop the changes which removed ffz() in PPC. it should be done not
   in this patch.
 - add a message after Signed-off.
---
Changes in V7:
 - Code style fixes
---
Changes in V6:
 - new patch for the patch series.
---
 xen/arch/arm/include/asm/arm32/bitops.h |  2 +-
 xen/arch/arm/include/asm/arm64/bitops.h |  6 ++
 xen/arch/arm/include/asm/bitops.h   |  9 +++--
 xen/arch/ppc/include/asm/bitops.h   |  3 ---
 xen/arch/x86/include/asm/bitops.h   | 12 +++-
 xen/common/bitops.c | 22 ++
 xen/common/page_alloc.c |  4 ++--
 xen/include/xen/bitops.h| 24 
 8 files changed, 61 insertions(+), 21 deletions(-)

diff --git a/xen/arch/arm/include/asm/arm32/bitops.h 
b/xen/arch/arm/include/asm/arm32/bitops.h
index d0309d47c1..9ee96f568b 100644
--- a/xen/arch/arm/include/asm/arm32/bitops.h
+++ b/xen/arch/arm/include/asm/arm32/bitops.h
@@ -1,7 +1,7 @@
 #ifndef _ARM_ARM32_BITOPS_H
 #define _ARM_ARM32_BITOPS_H
 
-#define flsl fls
+#define arch_flsl arch_fls
 
 /*
  * Little endian assembly bitops.  nr = 0 -> byte 0 bit 0.
diff --git a/xen/arch/arm/include/asm/arm64/bitops.h 
b/xen/arch/arm/include/asm/arm64/bitops.h
index 906d84e5f2..d942077392 100644
--- a/xen/arch/arm/include/asm/arm64/bitops.h
+++ b/xen/arch/arm/include/asm/arm64/bitops.h
@@ -1,17 +1,15 @@
 #ifndef _ARM_ARM64_BITOPS_H
 #define _ARM_ARM64_BITOPS_H
 
-static inline int flsl(unsigned long x)
+static always_inline unsigned int arch_flsl(unsigned long x)
 {
 uint64_t ret;
 
-if (__builtin_constant_p(x))
-   return generic_flsl(x);
-
 asm("clz\t%0, %1" : "=r" (ret) : "r" (x));
 
 return BITS_PER_LONG - ret;
 }
+#define arch_flsl arch_flsl
 
 /* Based on linux/include/asm-generic/bitops/find.h */
 
diff --git a/xen/arch/arm/include/asm/bitops.h 
b/xen/arch/arm/include/asm/bitops.h
index 8e16335e76..f428cf8338 100644
--- a/xen/arch/arm/include/asm/bitops.h
+++ b/xen/arch/arm/include/asm/bitops.h
@@ -78,17 +78,14 @@ bool clear_mask16_timeout(uint16_t mask, volatile void *p,
  * the clz instruction for much better code efficiency.
  */
 
-static inline int fls(unsigned int x)
+static always_inline unsigned int arch_fls(unsigned int x)
 {
-int ret;
-
-if (__builtin_constant_p(x))
-   return generic_fls(x);
+unsigned int ret;
 
 asm("clz\t%"__OP32"0, %"__OP32"1" : "=r" (ret) : "r" (x));
 return 32 - ret;
 }
-
+#define arch_fls arch_fls
 
 #define arch_ffs(x) ({ unsigned int __t = (x); fls(ISOLATE_LSB(__t)); })
 #define arch_ffsl(x) ({ unsigned long __t = (x); flsl(ISOLATE_LSB(__t)); })
diff --git a/xen/arch/ppc/include/asm/bitops.h 
b/xen/arch/ppc/include/asm/bitops.h
index 049aa62b89..46154fc957 100644
--- a/xen/arch/ppc/include/asm/bitops.h
+++ b/xen/arch/ppc/include/asm/bitops.h
@@ -119,9 +119,6 @@ static inline int test_and_set_bit(unsigned int nr, 
volatile void *addr)
 (volatile unsigned int *)addr + BITOP_WORD(nr)) != 0;
 }
 
-#define flsl(x) generic_flsl(x)
-#define fls(x) generic_fls(x)
-
 /**
  * hweightN - returns the hamming weight