Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Philippe Gerum wrote: > Gilles Chanteperdrix wrote: >> Jan Kiszka wrote: >>> +#include >>> + >>> +#if BITS_PER_LONG != 32 >>> +#error Upgrade to kernel 2.6! >>> +#endif >> I am not well versed with the 2.4 support. But are we sure we support no >> 64 bits architecture with 2.4 ? What about ppc64 ? >> > > No other 2.4 port will be merged, so this should be ok. ppc64 is 2.6 only for > us. Fine, so I merged my patch. Jan -- Siemens AG, Corporate Technology, CT SE 2 Corporate Competence Center Embedded Linux ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Gilles Chanteperdrix wrote: > Jan Kiszka wrote: >> +#include >> + >> +#if BITS_PER_LONG != 32 >> +#error Upgrade to kernel 2.6! >> +#endif > > I am not well versed with the 2.4 support. But are we sure we support no > 64 bits architecture with 2.4 ? What about ppc64 ? > No other 2.4 port will be merged, so this should be ok. ppc64 is 2.6 only for us. -- Philippe. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Jan Kiszka wrote: > Gilles Chanteperdrix wrote: >> Jan Kiszka wrote: >>> +#include >>> + >>> +#if BITS_PER_LONG != 32 >>> +#error Upgrade to kernel 2.6! >>> +#endif >> I am not well versed with the 2.4 support. But are we sure we support no >> 64 bits architecture with 2.4 ? What about ppc64 ? > > AFAIK, only ppc32 is supported on kernel 2.4 now, and I don't think > ppc64 ever was. Just i386 and ppc32 should bother us /wrt wrapping. It is OK for me then. -- Gilles. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Jan Kiszka wrote:
> Gilles Chanteperdrix wrote:
>> Jan Kiszka wrote:
>>> Gilles Chanteperdrix wrote:
Jan Kiszka wrote:
> ...and also automatically fixes the missing LOCK prefix for
> pthread_mutex_* services on x86_32 SMP.
This looks to me as a half-way unification. Can we not totally get rid
of atomic_32.h and atomic_64.h ? I mean since we are using unsigned long
as atomic_t on both platforms, there should not be much difference
(except maybe the inline asm).
>>> I could merge all atomic_32/64.h hunks into atomic.h if that this
>>> preferred, but I cannot help getting rid of the atomic_t vs. atomic64_t
>>> differences, thus the sub-arch specific part cannot be reduced as far as
>>> I see it ATM.
>> We could use atomic_long_t on the two arches.
>
> OK, but then it becomes wrapping business (2.4...) - on the long term a
> vanishing issue, granted. Will look into this.
Here we go, an atomic_long_t based unification patch. Cannot test 2.4
kernel built, unfortunately, but the rest is fine (note: requires
bootstrap run).
Jan
---
include/asm-generic/wrappers.h | 82 ++-
include/asm-x86/Makefile.am|2
include/asm-x86/atomic.h | 142 -
include/asm-x86/atomic_32.h| 101 -
include/asm-x86/atomic_64.h| 102 -
5 files changed, 203 insertions(+), 226 deletions(-)
Index: b/include/asm-x86/atomic.h
===
--- a/include/asm-x86/atomic.h
+++ b/include/asm-x86/atomic.h
@@ -1,5 +1,141 @@
-#ifdef __i386__
-#include "atomic_32.h"
+/*
+ * Copyright (C) 2007 Philippe Gerum <[EMAIL PROTECTED]>.
+ *
+ * Xenomai is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published
+ * by the Free Software Foundation; either version 2 of the License,
+ * or (at your option) any later version.
+ *
+ * Xenomai is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with Xenomai; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+ * 02111-1307, USA.
+ */
+
+#ifndef _XENO_ASM_X86_ATOMIC_H
+#define _XENO_ASM_X86_ATOMIC_H
+
+#include
+
+typedef unsigned long atomic_flags_t;
+
+#ifdef __KERNEL__
+
+#include
+#include
+#include
+
+#define xnarch_atomic_set(pcounter,i) atomic_long_set(pcounter,i)
+#define xnarch_atomic_get(pcounter)atomic_long_read(pcounter)
+#define xnarch_atomic_inc(pcounter)atomic_long_inc(pcounter)
+#define xnarch_atomic_dec(pcounter)atomic_long_dec(pcounter)
+#define xnarch_atomic_inc_and_test(pcounter) \
+ atomic_long_inc_and_test(pcounter)
+#define xnarch_atomic_dec_and_test(pcounter) \
+ atomic_long_dec_and_test(pcounter)
+#define xnarch_atomic_cmpxchg(pcounter,old,new) \
+ atomic_long_cmpxchg((pcounter),(old),(new))
+
+typedef atomic_long_t atomic_counter_t;
+typedef atomic_long_t xnarch_atomic_t;
+
+#define xnarch_atomic_set_mask(pflags,mask) \
+ atomic_set_mask((mask),(unsigned *)(pflags))
+#define xnarch_atomic_clear_mask(pflags,mask) \
+ atomic_clear_mask((mask),(unsigned *)(pflags))
+#define xnarch_atomic_xchg(ptr,x) xchg(ptr,x)
+
+#define xnarch_memory_barrier()smp_mb()
+
+#else /* !__KERNEL__ */
+
+#include
+
+#ifdef CONFIG_SMP
+#define LOCK_PREFIX "lock ; "
#else
-#include "atomic_64.h"
+#define LOCK_PREFIX ""
#endif
+
+typedef struct { unsigned long counter; } xnarch_atomic_t;
+
+#define xnarch_atomic_get(v) ((v)->counter)
+
+#define xnarch_atomic_set(v,i) (((v)->counter) = (i))
+
+#define xnarch_write_memory_barrier() xnarch_memory_barrier()
+
+#ifdef __i386__
+
+struct __xeno_xchg_dummy { unsigned long a[100]; };
+#define __xeno_xg(x) ((struct __xeno_xchg_dummy *)(x))
+
+static inline unsigned long xnarch_atomic_xchg (volatile void *ptr,
+ unsigned long x)
+{
+ __asm__ __volatile__("xchgl %0,%1"
+:"=r" (x)
+:"m" (*__xeno_xg(ptr)), "0" (x)
+:"memory");
+ return x;
+}
+
+static inline unsigned long
+xnarch_atomic_cmpxchg(xnarch_atomic_t *v, unsigned long old, unsigned long
newval)
+{
+ volatile void *ptr = &v->counter;
+ unsigned long prev;
+
+ __asm__ __volatile__(LOCK_PREFIX "cmpxchgl %1,%2"
+: "=a"(prev)
+: "r"(newval), "m"(*__xeno_xg(ptr)), "0"(old)
+: "memory");
+ return prev;
+}
+
+#define xnarch_memory_barrier()__asm__ __volatile__("": :
:"memory")
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Gilles Chanteperdrix wrote: > Jan Kiszka wrote: >> +#include >> + >> +#if BITS_PER_LONG != 32 >> +#error Upgrade to kernel 2.6! >> +#endif > > I am not well versed with the 2.4 support. But are we sure we support no > 64 bits architecture with 2.4 ? What about ppc64 ? AFAIK, only ppc32 is supported on kernel 2.4 now, and I don't think ppc64 ever was. Just i386 and ppc32 should bother us /wrt wrapping. Jan -- Siemens AG, Corporate Technology, CT SE 2 Corporate Competence Center Embedded Linux ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Jan Kiszka wrote: > +#include > + > +#if BITS_PER_LONG != 32 > +#error Upgrade to kernel 2.6! > +#endif I am not well versed with the 2.4 support. But are we sure we support no 64 bits architecture with 2.4 ? What about ppc64 ? -- Gilles. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Gilles Chanteperdrix wrote: > Jan Kiszka wrote: >> Gilles Chanteperdrix wrote: >>> Jan Kiszka wrote: ...and also automatically fixes the missing LOCK prefix for pthread_mutex_* services on x86_32 SMP. >>> This looks to me as a half-way unification. Can we not totally get rid >>> of atomic_32.h and atomic_64.h ? I mean since we are using unsigned long >>> as atomic_t on both platforms, there should not be much difference >>> (except maybe the inline asm). >>> >> I could merge all atomic_32/64.h hunks into atomic.h if that this >> preferred, but I cannot help getting rid of the atomic_t vs. atomic64_t >> differences, thus the sub-arch specific part cannot be reduced as far as >> I see it ATM. > > We could use atomic_long_t on the two arches. OK, but then it becomes wrapping business (2.4...) - on the long term a vanishing issue, granted. Will look into this. Jan signature.asc Description: OpenPGP digital signature ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Jan Kiszka wrote: > Gilles Chanteperdrix wrote: >> Jan Kiszka wrote: >>> ...and also automatically fixes the missing LOCK prefix for >>> pthread_mutex_* services on x86_32 SMP. >> This looks to me as a half-way unification. Can we not totally get rid >> of atomic_32.h and atomic_64.h ? I mean since we are using unsigned long >> as atomic_t on both platforms, there should not be much difference >> (except maybe the inline asm). >> > > I could merge all atomic_32/64.h hunks into atomic.h if that this > preferred, but I cannot help getting rid of the atomic_t vs. atomic64_t > differences, thus the sub-arch specific part cannot be reduced as far as > I see it ATM. We could use atomic_long_t on the two arches. -- Gilles. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
>> ...and also automatically fixes the missing LOCK prefix for
>> pthread_mutex_* services on x86_32 SMP.
>
> This looks to me as a half-way unification. Can we not totally get rid
> of atomic_32.h and atomic_64.h ? I mean since we are using unsigned long
> as atomic_t on both platforms, there should not be much difference
> (except maybe the inline asm).
>
I could merge all atomic_32/64.h hunks into atomic.h if that this
preferred, but I cannot help getting rid of the atomic_t vs. atomic64_t
differences, thus the sub-arch specific part cannot be reduced as far as
I see it ATM.
However, yesterday's version contained a regression /wrt 32 bit (missing
atomic_counter_t and xnarch_atomic_t type definitions), this one is
better:
---
include/asm-x86/atomic.h| 64
include/asm-x86/atomic_32.h | 31 -
include/asm-x86/atomic_64.h | 33 --
3 files changed, 65 insertions(+), 63 deletions(-)
Index: b/include/asm-x86/atomic.h
===
--- a/include/asm-x86/atomic.h
+++ b/include/asm-x86/atomic.h
@@ -1,5 +1,69 @@
+/*
+ * Copyright (C) 2007 Philippe Gerum <[EMAIL PROTECTED]>.
+ *
+ * Xenomai is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published
+ * by the Free Software Foundation; either version 2 of the License,
+ * or (at your option) any later version.
+ *
+ * Xenomai is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with Xenomai; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+ * 02111-1307, USA.
+ */
+
+#ifndef _XENO_ASM_X86_ATOMIC_H
+#define _XENO_ASM_X86_ATOMIC_H
+
+#include
+
+typedef unsigned long atomic_flags_t;
+
+#ifdef __KERNEL__
+
+#include
+#include
+#include
+
+#define xnarch_atomic_set_mask(pflags,mask) \
+ atomic_set_mask((mask),(unsigned *)(pflags))
+#define xnarch_atomic_clear_mask(pflags,mask) \
+ atomic_clear_mask((mask),(unsigned *)(pflags))
+#define xnarch_atomic_xchg(ptr,x) xchg(ptr,x)
+
+#define xnarch_memory_barrier() smp_mb()
+
+#else /* !__KERNEL__ */
+
+#include
+
+#ifdef CONFIG_SMP
+#define LOCK_PREFIX "lock ; "
+#else
+#define LOCK_PREFIX ""
+#endif
+
+typedef struct { unsigned long counter; } xnarch_atomic_t;
+
+#define xnarch_atomic_get(v) ((v)->counter)
+
+#define xnarch_atomic_set(v,i) (((v)->counter) = (i))
+
+#define xnarch_write_memory_barrier() xnarch_memory_barrier()
+
+#endif /* __KERNEL__ */
+
#ifdef __i386__
#include "atomic_32.h"
#else
#include "atomic_64.h"
#endif
+
+#include
+
+#endif /* !_XENO_ASM_X86_ATOMIC_64_H */
Index: b/include/asm-x86/atomic_32.h
===
--- a/include/asm-x86/atomic_32.h
+++ b/include/asm-x86/atomic_32.h
@@ -19,48 +19,26 @@
#ifndef _XENO_ASM_X86_ATOMIC_32_H
#define _XENO_ASM_X86_ATOMIC_32_H
-#define _XENO_ASM_X86_ATOMIC_H
#ifdef __KERNEL__
-#include
-#include
-#include
-
#define xnarch_atomic_set(pcounter,i) atomic_set(pcounter,i)
#define xnarch_atomic_get(pcounter)atomic_read(pcounter)
#define xnarch_atomic_inc(pcounter)atomic_inc(pcounter)
#define xnarch_atomic_dec(pcounter)atomic_dec(pcounter)
#define xnarch_atomic_inc_and_test(pcounter) atomic_inc_and_test(pcounter)
#define xnarch_atomic_dec_and_test(pcounter) atomic_dec_and_test(pcounter)
-#define xnarch_atomic_set_mask(pflags,mask)atomic_set_mask(mask,pflags)
-#define xnarch_atomic_clear_mask(pflags,mask) atomic_clear_mask(mask,pflags)
-#define xnarch_atomic_xchg(ptr,x) xchg(ptr,x)
#define xnarch_atomic_cmpxchg(pcounter,old,new) \
atomic_cmpxchg((pcounter),(old),(new))
-#define xnarch_memory_barrier() smp_mb()
-
typedef atomic_t atomic_counter_t;
typedef atomic_t xnarch_atomic_t;
#else /* !__KERNEL__ */
-#ifdef CONFIG_SMP
-#define LOCK_PREFIX "lock ; "
-#else
-#define LOCK_PREFIX ""
-#endif
-
-typedef struct { int counter; } xnarch_atomic_t;
-
struct __xeno_xchg_dummy { unsigned long a[100]; };
#define __xeno_xg(x) ((struct __xeno_xchg_dummy *)(x))
-#define xnarch_atomic_get(v) ((v)->counter)
-
-#define xnarch_atomic_set(v,i) (((v)->counter) = (i))
-
static inline unsigned long xnarch_atomic_xchg (volatile void *ptr,
unsigned long x)
{
@@ -84,17 +62,10 @@ xnarch_atomic_cmpxchg(xnarch_atomic_t *v
return prev;
}
-#define xnarch_memory_barrier() __asm__ __volatile__("": : :"memory")
-
+#define xnarch_memory_barrier
Re: [Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
Jan Kiszka wrote: > ...and also automatically fixes the missing LOCK prefix for > pthread_mutex_* services on x86_32 SMP. This looks to me as a half-way unification. Can we not totally get rid of atomic_32.h and atomic_64.h ? I mean since we are using unsigned long as atomic_t on both platforms, there should not be much difference (except maybe the inline asm). -- Gilles. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
[Xenomai-core] [PATCH 2/2] Unify asm-x86/atomic.h
...and also automatically fixes the missing LOCK prefix for
pthread_mutex_* services on x86_32 SMP.
---
include/asm-x86/atomic.h| 64
include/asm-x86/atomic_32.h | 31 -
include/asm-x86/atomic_64.h | 33 --
3 files changed, 64 insertions(+), 64 deletions(-)
Index: b/include/asm-x86/atomic.h
===
--- a/include/asm-x86/atomic.h
+++ b/include/asm-x86/atomic.h
@@ -1,5 +1,69 @@
+/*
+ * Copyright (C) 2007 Philippe Gerum <[EMAIL PROTECTED]>.
+ *
+ * Xenomai is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published
+ * by the Free Software Foundation; either version 2 of the License,
+ * or (at your option) any later version.
+ *
+ * Xenomai is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with Xenomai; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+ * 02111-1307, USA.
+ */
+
+#ifndef _XENO_ASM_X86_ATOMIC_H
+#define _XENO_ASM_X86_ATOMIC_H
+
+#include
+
+typedef unsigned long atomic_flags_t;
+
+#ifdef __KERNEL__
+
+#include
+#include
+#include
+
+#define xnarch_atomic_set_mask(pflags,mask) \
+ atomic_set_mask((mask),(unsigned *)(pflags))
+#define xnarch_atomic_clear_mask(pflags,mask) \
+ atomic_clear_mask((mask),(unsigned *)(pflags))
+#define xnarch_atomic_xchg(ptr,x) xchg(ptr,x)
+
+#define xnarch_memory_barrier() smp_mb()
+
+#else /* !__KERNEL__ */
+
+#include
+
+#ifdef CONFIG_SMP
+#define LOCK_PREFIX "lock ; "
+#else
+#define LOCK_PREFIX ""
+#endif
+
+typedef struct { unsigned long counter; } xnarch_atomic_t;
+
+#define xnarch_atomic_get(v) ((v)->counter)
+
+#define xnarch_atomic_set(v,i) (((v)->counter) = (i))
+
+#define xnarch_write_memory_barrier() xnarch_memory_barrier()
+
+#endif /* __KERNEL__ */
+
#ifdef __i386__
#include "atomic_32.h"
#else
#include "atomic_64.h"
#endif
+
+#include
+
+#endif /* !_XENO_ASM_X86_ATOMIC_64_H */
Index: b/include/asm-x86/atomic_32.h
===
--- a/include/asm-x86/atomic_32.h
+++ b/include/asm-x86/atomic_32.h
@@ -19,48 +19,23 @@
#ifndef _XENO_ASM_X86_ATOMIC_32_H
#define _XENO_ASM_X86_ATOMIC_32_H
-#define _XENO_ASM_X86_ATOMIC_H
#ifdef __KERNEL__
-#include
-#include
-#include
-
#define xnarch_atomic_set(pcounter,i) atomic_set(pcounter,i)
#define xnarch_atomic_get(pcounter)atomic_read(pcounter)
#define xnarch_atomic_inc(pcounter)atomic_inc(pcounter)
#define xnarch_atomic_dec(pcounter)atomic_dec(pcounter)
#define xnarch_atomic_inc_and_test(pcounter) atomic_inc_and_test(pcounter)
#define xnarch_atomic_dec_and_test(pcounter) atomic_dec_and_test(pcounter)
-#define xnarch_atomic_set_mask(pflags,mask)atomic_set_mask(mask,pflags)
-#define xnarch_atomic_clear_mask(pflags,mask) atomic_clear_mask(mask,pflags)
-#define xnarch_atomic_xchg(ptr,x) xchg(ptr,x)
#define xnarch_atomic_cmpxchg(pcounter,old,new) \
atomic_cmpxchg((pcounter),(old),(new))
-#define xnarch_memory_barrier() smp_mb()
-
-typedef atomic_t atomic_counter_t;
-typedef atomic_t xnarch_atomic_t;
-
#else /* !__KERNEL__ */
-#ifdef CONFIG_SMP
-#define LOCK_PREFIX "lock ; "
-#else
-#define LOCK_PREFIX ""
-#endif
-
-typedef struct { int counter; } xnarch_atomic_t;
-
struct __xeno_xchg_dummy { unsigned long a[100]; };
#define __xeno_xg(x) ((struct __xeno_xchg_dummy *)(x))
-#define xnarch_atomic_get(v) ((v)->counter)
-
-#define xnarch_atomic_set(v,i) (((v)->counter) = (i))
-
static inline unsigned long xnarch_atomic_xchg (volatile void *ptr,
unsigned long x)
{
@@ -88,13 +63,7 @@ xnarch_atomic_cmpxchg(xnarch_atomic_t *v
#define xnarch_read_memory_barrier() \
__asm__ __volatile__ (LOCK_PREFIX "addl $0,0(%%esp)": : :"memory")
-#define xnarch_write_memory_barrier() xnarch_memory_barrier()
#endif /* !__KERNEL__ */
-typedef unsigned long atomic_flags_t;
-
-#include
-#include
-
#endif /* !_XENO_ASM_X86_ATOMIC_32_H */
Index: b/include/asm-x86/atomic_64.h
===
--- a/include/asm-x86/atomic_64.h
+++ b/include/asm-x86/atomic_64.h
@@ -19,55 +19,25 @@
#ifndef _XENO_ASM_X86_ATOMIC_64_H
#define _XENO_ASM_X86_ATOMIC_64_H
-#define _XENO_ASM_X86_ATOMIC_H
-
-#include
-
-typedef unsigned long atomic_flags_t;
#ifdef __KERNEL__
-#include
-#include
-#include
-
#define xnarch_atomic_set(pcounter,i) atomic64_set(pcounter,i)
#define xnarch_at
