FRV: Use the correct preemption primitives in kmap_atomic() and co [try #2]
PATCH 1/1 FRV: Use the correct preemption primitives in kmap_atomic() and co [try #2]
From: David Howells <[email protected]>
Use inc/dec_preempt_count() rather than preempt_enable/disable() and manually add in the compiler barriers that were provided by the latter. This makes FRV consistent with other archs.
Furthermore, the compiler barrier effects are now there unconditionally - at least as far as preemption is concerned - because we don't want the compiler moving memory accesses out of the section of code in which the mapping is in force - in effect the kmap_atomic() must imply a LOCK-class barrier and the kunmap_atomic() must imply an UNLOCK-class barrier to the compiler.
---
include/asm-frv/highmem.h | 27 ++++++++++++++------------- 1 files changed, 14 insertions(+), 13 deletions(-)
include/asm-frv/highmem.h +14 -13
@@ -82,11 +82,11 @@ ({ \ dampr = paddr | xAMPRx_L | xAMPRx_M | xAMPRx_S | xAMPRx_SS_16Kb | xAMPRx_V; \ \ if (type != __KM_CACHE) \ - asm volatile("movgs %0,dampr"#ampr :: "r"(dampr)); \ + asm volatile("movgs %0,dampr"#ampr :: "r"(dampr) : "memory"); \ else \ asm volatile("movgs %0,iampr"#ampr"\n" \ "movgs %0,dampr"#ampr"\n" \ - :: "r"(dampr) \ + :: "r"(dampr) : "memory" \ ); \ \ asm("movsg damlr"#ampr",%0" : "=r"(damlr)); \
@@ -104,7 +104,7 @@ ({ \ asmvolatile ("movgs %0,tplr \n" \"movgs %1,tppr \n" \"tlbpr %0,gr0,#2,#1" \ - : :"r" (damlr),"r" (dampr)); \ + : :"r" (damlr),"r" (dampr) :"memory" ); \ \/*printk("TLB: SECN sl=%d L=%08lx P=%08lx\n", slot, damlr, dampr);*/ \ \
@@ -115,7 +115,7 @@ static inline void *kmap_atomic(struct p {unsigned long paddr; -preempt_disable (); +inc_preempt_count (); paddr =page_to_phys (page);switch (type) {
@@ -138,16 +138,16 @@ static inline void *kmap_atomic(struct p } } -#define __kunmap_atomic_primary (type, ampr) \ -do { \ - asm volatile("movgs gr0,dampr"#ampr"\n"); \ - if (type == __KM_CACHE) \ - asm volatile("movgs gr0,iampr"#ampr"\n"); \ +#define __kunmap_atomic_primary (type, ampr) \ +do { \ + asm volatile("movgs gr0,dampr"#ampr"\n" ::: "memory"); \ + if (type == __KM_CACHE) \ + asm volatile("movgs gr0,iampr"#ampr"\n" ::: "memory"); \ } while(0) -#define __kunmap_atomic_secondary (slot, vaddr) \ -do { \ - asm volatile("tlbpr %0,gr0,#4,#1" : : "r"(vaddr)); \ +#define __kunmap_atomic_secondary (slot, vaddr) \ +do { \ + asm volatile("tlbpr %0,gr0,#4,#1" : : "r"(vaddr) : "memory"); \ } while(0)static inline void kunmap_atomic (void *kvaddr,enum km_type type)
@@ -170,7 +170,8 @@ static inline void kunmap_atomic(void *kdefault : BUG(); } -preempt_enable (); +dec_preempt_count (); +preempt_check_resched (); }#endif /* !__ASSEMBLY__ */
Quoted 6 lines · David Howells
From: David Howells <[email protected]>
Use inc/dec_preempt_count() rather than preempt_enable/disable() and manually add in the compiler barriers that were provided by the latter. This makes FRV consistent with other archs.
Quoted 7 lines
Furthermore, the compiler barrier effects are now there unconditionally - at least as far as preemption is concerned - because we don't want the compiler moving memory accesses out of the section of code in which the mapping is in force - in effect the
kmap_atomic()must imply a LOCK-class barrier and thekunmap_atomic()must imply an UNLOCK-class barrier to the compiler.Signed-off-by David Howells <[email protected]>
Quoted 76 lines
---
include/asm-frv/highmem.h | 27 ++++++++++++++------------- 1 files changed, 14 insertions(+), 13 deletions(-)include/asm-frv/highmem.h +14
-13@@ -82,11 +82,11 @@ ({ \ dampr = paddr | xAMPRx_L | xAMPRx_M | xAMPRx_S | xAMPRx_SS_16Kb | xAMPRx_V; \ \ if (type != __KM_CACHE) \ - asm volatile("movgs %0,dampr"#ampr :: "r"(dampr)); \ + asm volatile("movgs %0,dampr"#ampr :: "r"(dampr) : "memory"); \ else \ asm volatile("movgs %0,iampr"#ampr"\n" \ "movgs %0,dampr"#ampr"\n" \ - :: "r"(dampr) \ + :: "r"(dampr) : "memory" \ ); \ \ asm("movsg damlr"#ampr",%0" : "=r"(damlr)); \@@ -104,7 +104,7 @@ ({ \ asmvolatile ("movgs %0,tplr \n" \"movgs %1,tppr \n" \"tlbpr %0,gr0,#2,#1" \ - : :"r" (damlr),"r" (dampr)); \ + : :"r" (damlr),"r" (dampr) :"memory" ); \ \/*printk("TLB: SECN sl=%d L=%08lx P=%08lx\n", slot, damlr, dampr);*/ \ \@@ -115,7 +115,7 @@ static inline void *kmap_atomic(struct p {unsigned long paddr; -preempt_disable (); +inc_preempt_count (); paddr =page_to_phys (page);switch (type) {@@ -138,16 +138,16 @@ static inline void *kmap_atomic(struct p } } -#define __kunmap_atomic_primary (type, ampr) \ -do { \ - asm volatile("movgs gr0,dampr"#ampr"\n"); \ - if (type == __KM_CACHE) \ - asm volatile("movgs gr0,iampr"#ampr"\n"); \ +#define __kunmap_atomic_primary (type, ampr) \ +do { \ + asm volatile("movgs gr0,dampr"#ampr"\n" ::: "memory"); \ + if (type == __KM_CACHE) \ + asm volatile("movgs gr0,iampr"#ampr"\n" ::: "memory"); \ } while(0) -#define __kunmap_atomic_secondary (slot, vaddr) \ -do { \ - asm volatile("tlbpr %0,gr0,#4,#1" : : "r"(vaddr)); \ +#define __kunmap_atomic_secondary (slot, vaddr) \ +do { \ + asm volatile("tlbpr %0,gr0,#4,#1" : : "r"(vaddr) : "memory"); \ } while(0)static inline void kunmap_atomic (void *kvaddr,enum km_type type)@@ -170,7 +170,8 @@ static inline void kunmap_atomic(void *kdefault : BUG(); } -preempt_enable (); +dec_preempt_count (); +preempt_check_resched (); }#endif /* !__ASSEMBLY__ */