forge/linux-pinenote
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
linux-pinenote/arch/sh/include/asm/cmpxchg-grb.h

73 lines
2.1 KiB
C
Raw Normal View History

sh: GUSA atomic rollback support. This implements kernel-level atomic rollback built on top of gUSA, as an alternative non-IRQ based atomicity method. This is generally a faster method for platforms that are lacking the LL/SC pairs that SH-4A and later use, and is only supportable on legacy cores. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:12:36 +09:00
#ifndef __ASM_SH_CMPXCHG_GRB_H
#define __ASM_SH_CMPXCHG_GRB_H
static inline unsigned long xchg_u32(volatile u32 *m, unsigned long val)
{
unsigned long retval;
__asm__ __volatile__ (
" .align 2 \n\t"
" mova 1f, r0 \n\t" /* r0 = end point */
" nop \n\t"
" mov r15, r1 \n\t" /* r1 = saved sp */
" mov #-4, r15 \n\t" /* LOGIN */
" mov.l @%1, %0 \n\t" /* load old value */
" mov.l %2, @%1 \n\t" /* store new value */
"1: mov r1, r15 \n\t" /* LOGOUT */
: "=&r" (retval),
sh: Fix up xchg/cmpxchg corruption with gUSA RB. gUSA special cases r15 for part of its login/out sequence, meaning that any parameters need to be explicitly prohibited from accidentally being assigned that particular register, and the compiler ultimately needs to use a temporary instead. Certain configurations have begun generating code paths that do indeed get allocated r15, resulting in immediate corruption of the exchanged value. This was observed in (amongst others) exit_mm() code generation where the xchg_u32 call was immediately corrupting a structure address. As this is a general gUSA restriction, the rest of the users likewise need to be updated to ensure sensible constraints. References: https://bugzilla.stlinux.com/show_bug.cgi?id=11229 Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com> Reviewed-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2011-06-08 15:22:39 +09:00
"+r" (m),
"+r" (val) /* inhibit r15 overloading */
:
sh: GUSA atomic rollback support. This implements kernel-level atomic rollback built on top of gUSA, as an alternative non-IRQ based atomicity method. This is generally a faster method for platforms that are lacking the LL/SC pairs that SH-4A and later use, and is only supportable on legacy cores. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:12:36 +09:00
: "memory", "r0", "r1");
return retval;
}
static inline unsigned long xchg_u8(volatile u8 *m, unsigned long val)
{
unsigned long retval;
__asm__ __volatile__ (
" .align 2 \n\t"
" mova 1f, r0 \n\t" /* r0 = end point */
" mov r15, r1 \n\t" /* r1 = saved sp */
" mov #-6, r15 \n\t" /* LOGIN */
" mov.b @%1, %0 \n\t" /* load old value */
" extu.b %0, %0 \n\t" /* extend as unsigned */
" mov.b %2, @%1 \n\t" /* store new value */
"1: mov r1, r15 \n\t" /* LOGOUT */
: "=&r" (retval),
sh: Fix up xchg/cmpxchg corruption with gUSA RB. gUSA special cases r15 for part of its login/out sequence, meaning that any parameters need to be explicitly prohibited from accidentally being assigned that particular register, and the compiler ultimately needs to use a temporary instead. Certain configurations have begun generating code paths that do indeed get allocated r15, resulting in immediate corruption of the exchanged value. This was observed in (amongst others) exit_mm() code generation where the xchg_u32 call was immediately corrupting a structure address. As this is a general gUSA restriction, the rest of the users likewise need to be updated to ensure sensible constraints. References: https://bugzilla.stlinux.com/show_bug.cgi?id=11229 Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com> Reviewed-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2011-06-08 15:22:39 +09:00
"+r" (m),
"+r" (val) /* inhibit r15 overloading */
:
sh: GUSA atomic rollback support. This implements kernel-level atomic rollback built on top of gUSA, as an alternative non-IRQ based atomicity method. This is generally a faster method for platforms that are lacking the LL/SC pairs that SH-4A and later use, and is only supportable on legacy cores. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:12:36 +09:00
: "memory" , "r0", "r1");
return retval;
}
static inline unsigned long __cmpxchg_u32(volatile int *m, unsigned long old,
unsigned long new)
{
unsigned long retval;
__asm__ __volatile__ (
" .align 2 \n\t"
" mova 1f, r0 \n\t" /* r0 = end point */
" nop \n\t"
" mov r15, r1 \n\t" /* r1 = saved sp */
" mov #-8, r15 \n\t" /* LOGIN */
sh: Fix up xchg/cmpxchg corruption with gUSA RB. gUSA special cases r15 for part of its login/out sequence, meaning that any parameters need to be explicitly prohibited from accidentally being assigned that particular register, and the compiler ultimately needs to use a temporary instead. Certain configurations have begun generating code paths that do indeed get allocated r15, resulting in immediate corruption of the exchanged value. This was observed in (amongst others) exit_mm() code generation where the xchg_u32 call was immediately corrupting a structure address. As this is a general gUSA restriction, the rest of the users likewise need to be updated to ensure sensible constraints. References: https://bugzilla.stlinux.com/show_bug.cgi?id=11229 Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com> Reviewed-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2011-06-08 15:22:39 +09:00
" mov.l @%3, %0 \n\t" /* load old value */
" cmp/eq %0, %1 \n\t"
sh: GUSA atomic rollback support. This implements kernel-level atomic rollback built on top of gUSA, as an alternative non-IRQ based atomicity method. This is generally a faster method for platforms that are lacking the LL/SC pairs that SH-4A and later use, and is only supportable on legacy cores. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:12:36 +09:00
" bf 1f \n\t" /* if not equal */
sh: Fix up xchg/cmpxchg corruption with gUSA RB. gUSA special cases r15 for part of its login/out sequence, meaning that any parameters need to be explicitly prohibited from accidentally being assigned that particular register, and the compiler ultimately needs to use a temporary instead. Certain configurations have begun generating code paths that do indeed get allocated r15, resulting in immediate corruption of the exchanged value. This was observed in (amongst others) exit_mm() code generation where the xchg_u32 call was immediately corrupting a structure address. As this is a general gUSA restriction, the rest of the users likewise need to be updated to ensure sensible constraints. References: https://bugzilla.stlinux.com/show_bug.cgi?id=11229 Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com> Reviewed-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2011-06-08 15:22:39 +09:00
" mov.l %2, @%3 \n\t" /* store new value */
sh: GUSA atomic rollback support. This implements kernel-level atomic rollback built on top of gUSA, as an alternative non-IRQ based atomicity method. This is generally a faster method for platforms that are lacking the LL/SC pairs that SH-4A and later use, and is only supportable on legacy cores. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:12:36 +09:00
"1: mov r1, r15 \n\t" /* LOGOUT */
sh: Fix up xchg/cmpxchg corruption with gUSA RB. gUSA special cases r15 for part of its login/out sequence, meaning that any parameters need to be explicitly prohibited from accidentally being assigned that particular register, and the compiler ultimately needs to use a temporary instead. Certain configurations have begun generating code paths that do indeed get allocated r15, resulting in immediate corruption of the exchanged value. This was observed in (amongst others) exit_mm() code generation where the xchg_u32 call was immediately corrupting a structure address. As this is a general gUSA restriction, the rest of the users likewise need to be updated to ensure sensible constraints. References: https://bugzilla.stlinux.com/show_bug.cgi?id=11229 Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com> Reviewed-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2011-06-08 15:22:39 +09:00
: "=&r" (retval),
"+r" (old), "+r" (new) /* old or new can be r15 */
: "r" (m)
sh: GUSA atomic rollback support. This implements kernel-level atomic rollback built on top of gUSA, as an alternative non-IRQ based atomicity method. This is generally a faster method for platforms that are lacking the LL/SC pairs that SH-4A and later use, and is only supportable on legacy cores. Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
2007-11-30 16:12:36 +09:00
: "memory" , "r0", "r1", "t");
return retval;
}
#endif /* __ASM_SH_CMPXCHG_GRB_H */
Reference in a new issue Copy permalink