diff options
Diffstat (limited to 'arch/tile/include/asm')
75 files changed, 2360 insertions, 1702 deletions
diff --git a/arch/tile/include/asm/Kbuild b/arch/tile/include/asm/Kbuild index 0bb42642343..0aa5675e702 100644 --- a/arch/tile/include/asm/Kbuild +++ b/arch/tile/include/asm/Kbuild @@ -1,33 +1,32 @@ -include include/asm-generic/Kbuild.asm header-y += ../arch/ -header-y += ucontext.h -header-y += hardwall.h - generic-y += bug.h generic-y += bugs.h +generic-y += clkdev.h generic-y += cputime.h -generic-y += device.h generic-y += div64.h generic-y += emergency-restart.h generic-y += errno.h +generic-y += exec.h generic-y += fb.h generic-y += fcntl.h +generic-y += hash.h +generic-y += hw_irq.h generic-y += ioctl.h generic-y += ioctls.h -generic-y += ipc.h generic-y += ipcbuf.h generic-y += irq_regs.h -generic-y += kdebug.h generic-y += local.h -generic-y += module.h +generic-y += local64.h +generic-y += mcs_spinlock.h generic-y += msgbuf.h generic-y += mutex.h generic-y += param.h generic-y += parport.h generic-y += poll.h generic-y += posix_types.h +generic-y += preempt.h generic-y += resource.h generic-y += scatterlist.h generic-y += sembuf.h @@ -39,6 +38,6 @@ generic-y += sockios.h generic-y += statfs.h generic-y += termbits.h generic-y += termios.h +generic-y += trace_clock.h generic-y += types.h -generic-y += ucontext.h generic-y += xor.h diff --git a/arch/tile/include/asm/atomic.h b/arch/tile/include/asm/atomic.h index 921dbeb8a70..70979846076 100644 --- a/arch/tile/include/asm/atomic.h +++ b/arch/tile/include/asm/atomic.h @@ -17,10 +17,12 @@ #ifndef _ASM_TILE_ATOMIC_H #define _ASM_TILE_ATOMIC_H +#include <asm/cmpxchg.h> + #ifndef __ASSEMBLY__ #include <linux/compiler.h> -#include <asm/system.h> +#include <linux/types.h> #define ATOMIC_INIT(i) { (i) } @@ -112,6 +114,32 @@ static inline int atomic_read(const atomic_t *v) #define atomic_inc_and_test(v) (atomic_inc_return(v) == 0) /** + * atomic_xchg - atomically exchange contents of memory with a new value + * @v: pointer of type atomic_t + * @i: integer value to store in memory + * + * Atomically sets @v to @i and returns old @v + */ +static inline int atomic_xchg(atomic_t *v, int n) +{ + return xchg(&v->counter, n); +} + +/** + * atomic_cmpxchg - atomically exchange contents of memory if it matches + * @v: pointer of type atomic_t + * @o: old value that memory should have + * @n: new value to write to memory if it matches + * + * Atomically checks if @v holds @o and replaces it with @n if so. + * Returns the old value at @v. + */ +static inline int atomic_cmpxchg(atomic_t *v, int o, int n) +{ + return cmpxchg(&v->counter, o, n); +} + +/** * atomic_add_negative - add and test if negative * @v: pointer of type atomic_t * @i: integer value to add @@ -121,54 +149,6 @@ static inline int atomic_read(const atomic_t *v) */ #define atomic_add_negative(i, v) (atomic_add_return((i), (v)) < 0) -/* Nonexistent functions intended to cause link errors. */ -extern unsigned long __xchg_called_with_bad_pointer(void); -extern unsigned long __cmpxchg_called_with_bad_pointer(void); - -#define xchg(ptr, x) \ - ({ \ - typeof(*(ptr)) __x; \ - switch (sizeof(*(ptr))) { \ - case 4: \ - __x = (typeof(__x))(typeof(__x-__x))atomic_xchg( \ - (atomic_t *)(ptr), \ - (u32)(typeof((x)-(x)))(x)); \ - break; \ - case 8: \ - __x = (typeof(__x))(typeof(__x-__x))atomic64_xchg( \ - (atomic64_t *)(ptr), \ - (u64)(typeof((x)-(x)))(x)); \ - break; \ - default: \ - __xchg_called_with_bad_pointer(); \ - } \ - __x; \ - }) - -#define cmpxchg(ptr, o, n) \ - ({ \ - typeof(*(ptr)) __x; \ - switch (sizeof(*(ptr))) { \ - case 4: \ - __x = (typeof(__x))(typeof(__x-__x))atomic_cmpxchg( \ - (atomic_t *)(ptr), \ - (u32)(typeof((o)-(o)))(o), \ - (u32)(typeof((n)-(n)))(n)); \ - break; \ - case 8: \ - __x = (typeof(__x))(typeof(__x-__x))atomic64_cmpxchg( \ - (atomic64_t *)(ptr), \ - (u64)(typeof((o)-(o)))(o), \ - (u64)(typeof((n)-(n)))(n)); \ - break; \ - default: \ - __cmpxchg_called_with_bad_pointer(); \ - } \ - __x; \ - }) - -#define tas(ptr) (xchg((ptr), 1)) - #endif /* __ASSEMBLY__ */ #ifndef __tilegx__ @@ -177,4 +157,52 @@ extern unsigned long __cmpxchg_called_with_bad_pointer(void); #include <asm/atomic_64.h> #endif +#ifndef __ASSEMBLY__ + +/** + * atomic64_xchg - atomically exchange contents of memory with a new value + * @v: pointer of type atomic64_t + * @i: integer value to store in memory + * + * Atomically sets @v to @i and returns old @v + */ +static inline long long atomic64_xchg(atomic64_t *v, long long n) +{ + return xchg64(&v->counter, n); +} + +/** + * atomic64_cmpxchg - atomically exchange contents of memory if it matches + * @v: pointer of type atomic64_t + * @o: old value that memory should have + * @n: new value to write to memory if it matches + * + * Atomically checks if @v holds @o and replaces it with @n if so. + * Returns the old value at @v. + */ +static inline long long atomic64_cmpxchg(atomic64_t *v, long long o, + long long n) +{ + return cmpxchg64(&v->counter, o, n); +} + +static inline long long atomic64_dec_if_positive(atomic64_t *v) +{ + long long c, old, dec; + + c = atomic64_read(v); + for (;;) { + dec = c - 1; + if (unlikely(dec < 0)) + break; + old = atomic64_cmpxchg((v), c, dec); + if (likely(old == c)) + break; + c = old; + } + return dec; +} + +#endif /* __ASSEMBLY__ */ + #endif /* _ASM_TILE_ATOMIC_H */ diff --git a/arch/tile/include/asm/atomic_32.h b/arch/tile/include/asm/atomic_32.h index c03349e0ca9..1b109fad9ff 100644 --- a/arch/tile/include/asm/atomic_32.h +++ b/arch/tile/include/asm/atomic_32.h @@ -17,44 +17,11 @@ #ifndef _ASM_TILE_ATOMIC_32_H #define _ASM_TILE_ATOMIC_32_H +#include <asm/barrier.h> #include <arch/chip.h> #ifndef __ASSEMBLY__ -/* Tile-specific routines to support <linux/atomic.h>. */ -int _atomic_xchg(atomic_t *v, int n); -int _atomic_xchg_add(atomic_t *v, int i); -int _atomic_xchg_add_unless(atomic_t *v, int a, int u); -int _atomic_cmpxchg(atomic_t *v, int o, int n); - -/** - * atomic_xchg - atomically exchange contents of memory with a new value - * @v: pointer of type atomic_t - * @i: integer value to store in memory - * - * Atomically sets @v to @i and returns old @v - */ -static inline int atomic_xchg(atomic_t *v, int n) -{ - smp_mb(); /* barrier for proper semantics */ - return _atomic_xchg(v, n); -} - -/** - * atomic_cmpxchg - atomically exchange contents of memory if it matches - * @v: pointer of type atomic_t - * @o: old value that memory should have - * @n: new value to write to memory if it matches - * - * Atomically checks if @v holds @o and replaces it with @n if so. - * Returns the old value at @v. - */ -static inline int atomic_cmpxchg(atomic_t *v, int o, int n) -{ - smp_mb(); /* barrier for proper semantics */ - return _atomic_cmpxchg(v, o, n); -} - /** * atomic_add - add integer to atomic variable * @i: integer value to add @@ -64,7 +31,7 @@ static inline int atomic_cmpxchg(atomic_t *v, int o, int n) */ static inline void atomic_add(int i, atomic_t *v) { - _atomic_xchg_add(v, i); + _atomic_xchg_add(&v->counter, i); } /** @@ -77,7 +44,7 @@ static inline void atomic_add(int i, atomic_t *v) static inline int atomic_add_return(int i, atomic_t *v) { smp_mb(); /* barrier for proper semantics */ - return _atomic_xchg_add(v, i) + i; + return _atomic_xchg_add(&v->counter, i) + i; } /** @@ -92,7 +59,7 @@ static inline int atomic_add_return(int i, atomic_t *v) static inline int __atomic_add_unless(atomic_t *v, int a, int u) { smp_mb(); /* barrier for proper semantics */ - return _atomic_xchg_add_unless(v, a, u); + return _atomic_xchg_add_unless(&v->counter, a, u); } /** @@ -107,64 +74,31 @@ static inline int __atomic_add_unless(atomic_t *v, int a, int u) */ static inline void atomic_set(atomic_t *v, int n) { - _atomic_xchg(v, n); + _atomic_xchg(&v->counter, n); } /* A 64bit atomic type */ typedef struct { - u64 __aligned(8) counter; + long long counter; } atomic64_t; #define ATOMIC64_INIT(val) { (val) } -u64 _atomic64_xchg(atomic64_t *v, u64 n); -u64 _atomic64_xchg_add(atomic64_t *v, u64 i); -u64 _atomic64_xchg_add_unless(atomic64_t *v, u64 a, u64 u); -u64 _atomic64_cmpxchg(atomic64_t *v, u64 o, u64 n); - /** * atomic64_read - read atomic variable * @v: pointer of type atomic64_t * * Atomically reads the value of @v. */ -static inline u64 atomic64_read(const atomic64_t *v) +static inline long long atomic64_read(const atomic64_t *v) { /* * Requires an atomic op to read both 32-bit parts consistently. * Casting away const is safe since the atomic support routines * do not write to memory if the value has not been modified. */ - return _atomic64_xchg_add((atomic64_t *)v, 0); -} - -/** - * atomic64_xchg - atomically exchange contents of memory with a new value - * @v: pointer of type atomic64_t - * @i: integer value to store in memory - * - * Atomically sets @v to @i and returns old @v - */ -static inline u64 atomic64_xchg(atomic64_t *v, u64 n) -{ - smp_mb(); /* barrier for proper semantics */ - return _atomic64_xchg(v, n); -} - -/** - * atomic64_cmpxchg - atomically exchange contents of memory if it matches - * @v: pointer of type atomic64_t - * @o: old value that memory should have - * @n: new value to write to memory if it matches - * - * Atomically checks if @v holds @o and replaces it with @n if so. - * Returns the old value at @v. - */ -static inline u64 atomic64_cmpxchg(atomic64_t *v, u64 o, u64 n) -{ - smp_mb(); /* barrier for proper semantics */ - return _atomic64_cmpxchg(v, o, n); + return _atomic64_xchg_add((long long *)&v->counter, 0); } /** @@ -174,9 +108,9 @@ static inline u64 atomic64_cmpxchg(atomic64_t *v, u64 o, u64 n) * * Atomically adds @i to @v. */ -static inline void atomic64_add(u64 i, atomic64_t *v) +static inline void atomic64_add(long long i, atomic64_t *v) { - _atomic64_xchg_add(v, i); + _atomic64_xchg_add(&v->counter, i); } /** @@ -186,10 +120,10 @@ static inline void atomic64_add(u64 i, atomic64_t *v) * * Atomically adds @i to @v and returns @i + @v */ -static inline u64 atomic64_add_return(u64 i, atomic64_t *v) +static inline long long atomic64_add_return(long long i, atomic64_t *v) { smp_mb(); /* barrier for proper semantics */ - return _atomic64_xchg_add(v, i) + i; + return _atomic64_xchg_add(&v->counter, i) + i; } /** @@ -199,12 +133,13 @@ static inline u64 atomic64_add_return(u64 i, atomic64_t *v) * @u: ...unless v is equal to u. * * Atomically adds @a to @v, so long as @v was not already @u. - * Returns the old value of @v. + * Returns non-zero if @v was not @u, and zero otherwise. */ -static inline u64 atomic64_add_unless(atomic64_t *v, u64 a, u64 u) +static inline long long atomic64_add_unless(atomic64_t *v, long long a, + long long u) { smp_mb(); /* barrier for proper semantics */ - return _atomic64_xchg_add_unless(v, a, u) != u; + return _atomic64_xchg_add_unless(&v->counter, a, u) != u; } /** @@ -217,9 +152,9 @@ static inline u64 atomic64_add_unless(atomic64_t *v, u64 a, u64 u) * atomic64_set() can't be just a raw store, since it would be lost if it * fell between the load and store of one of the other atomic ops. */ -static inline void atomic64_set(atomic64_t *v, u64 n) +static inline void atomic64_set(atomic64_t *v, long long n) { - _atomic64_xchg(v, n); + _atomic64_xchg(&v->counter, n); } #define atomic64_add_negative(a, v) (atomic64_add_return((a), (v)) < 0) @@ -234,16 +169,6 @@ static inline void atomic64_set(atomic64_t *v, u64 n) #define atomic64_dec_and_test(v) (atomic64_dec_return((v)) == 0) #define atomic64_inc_not_zero(v) atomic64_add_unless((v), 1LL, 0LL) -/* - * We need to barrier before modifying the word, since the _atomic_xxx() - * routines just tns the lock and then read/modify/write of the word. - * But after the word is updated, the routine issues an "mf" before returning, - * and since it's a function call, we don't even need a compiler barrier. - */ -#define smp_mb__before_atomic_dec() smp_mb() -#define smp_mb__before_atomic_inc() smp_mb() -#define smp_mb__after_atomic_dec() do { } while (0) -#define smp_mb__after_atomic_inc() do { } while (0) #endif /* !__ASSEMBLY__ */ @@ -251,21 +176,6 @@ static inline void atomic64_set(atomic64_t *v, u64 n) * Internal definitions only beyond this point. */ -#define ATOMIC_LOCKS_FOUND_VIA_TABLE() \ - (!CHIP_HAS_CBOX_HOME_MAP() && defined(CONFIG_SMP)) - -#if ATOMIC_LOCKS_FOUND_VIA_TABLE() - -/* Number of entries in atomic_lock_ptr[]. */ -#define ATOMIC_HASH_L1_SHIFT 6 -#define ATOMIC_HASH_L1_SIZE (1 << ATOMIC_HASH_L1_SHIFT) - -/* Number of locks in each struct pointed to by atomic_lock_ptr[]. */ -#define ATOMIC_HASH_L2_SHIFT (CHIP_L2_LOG_LINE_SIZE() - 2) -#define ATOMIC_HASH_L2_SIZE (1 << ATOMIC_HASH_L2_SHIFT) - -#else /* ATOMIC_LOCKS_FOUND_VIA_TABLE() */ - /* * Number of atomic locks in atomic_locks[]. Must be a power of two. * There is no reason for more than PAGE_SIZE / 8 entries, since that @@ -280,8 +190,6 @@ static inline void atomic64_set(atomic64_t *v, u64 n) extern int atomic_locks[]; #endif -#endif /* ATOMIC_LOCKS_FOUND_VIA_TABLE() */ - /* * All the code that may fault while holding an atomic lock must * place the pointer to the lock in ATOMIC_LOCK_REG so the fault code @@ -302,7 +210,14 @@ void __init_atomic_per_cpu(void); void __atomic_fault_unlock(int *lock_ptr); #endif +/* Return a pointer to the lock for the given address. */ +int *__atomic_hashed_lock(volatile void *v); + /* Private helper routines in lib/atomic_asm_32.S */ +struct __get_user { + unsigned long val; + int err; +}; extern struct __get_user __atomic_cmpxchg(volatile int *p, int *lock, int o, int n); extern struct __get_user __atomic_xchg(volatile int *p, int *lock, int n); @@ -312,11 +227,16 @@ extern struct __get_user __atomic_xchg_add_unless(volatile int *p, extern struct __get_user __atomic_or(volatile int *p, int *lock, int n); extern struct __get_user __atomic_andn(volatile int *p, int *lock, int n); extern struct __get_user __atomic_xor(volatile int *p, int *lock, int n); -extern u64 __atomic64_cmpxchg(volatile u64 *p, int *lock, u64 o, u64 n); -extern u64 __atomic64_xchg(volatile u64 *p, int *lock, u64 n); -extern u64 __atomic64_xchg_add(volatile u64 *p, int *lock, u64 n); -extern u64 __atomic64_xchg_add_unless(volatile u64 *p, - int *lock, u64 o, u64 n); +extern long long __atomic64_cmpxchg(volatile long long *p, int *lock, + long long o, long long n); +extern long long __atomic64_xchg(volatile long long *p, int *lock, long long n); +extern long long __atomic64_xchg_add(volatile long long *p, int *lock, + long long n); +extern long long __atomic64_xchg_add_unless(volatile long long *p, + int *lock, long long o, long long n); + +/* Return failure from the atomic wrappers. */ +struct __get_user __atomic_bad_address(int __user *addr); #endif /* !__ASSEMBLY__ */ diff --git a/arch/tile/include/asm/atomic_64.h b/arch/tile/include/asm/atomic_64.h index 27fe667fddf..7b11c5fadd4 100644 --- a/arch/tile/include/asm/atomic_64.h +++ b/arch/tile/include/asm/atomic_64.h @@ -19,6 +19,7 @@ #ifndef __ASSEMBLY__ +#include <asm/barrier.h> #include <arch/spr_def.h> /* First, the 32-bit atomic ops that are "real" on our 64-bit platform. */ @@ -31,25 +32,6 @@ * on any routine which updates memory and returns a value. */ -static inline int atomic_cmpxchg(atomic_t *v, int o, int n) -{ - int val; - __insn_mtspr(SPR_CMPEXCH_VALUE, o); - smp_mb(); /* barrier for proper semantics */ - val = __insn_cmpexch4((void *)&v->counter, n); - smp_mb(); /* barrier for proper semantics */ - return val; -} - -static inline int atomic_xchg(atomic_t *v, int n) -{ - int val; - smp_mb(); /* barrier for proper semantics */ - val = __insn_exch4((void *)&v->counter, n); - smp_mb(); /* barrier for proper semantics */ - return val; -} - static inline void atomic_add(int i, atomic_t *v) { __insn_fetchadd4((void *)&v->counter, i); @@ -71,7 +53,7 @@ static inline int __atomic_add_unless(atomic_t *v, int a, int u) if (oldval == u) break; guess = oldval; - oldval = atomic_cmpxchg(v, guess, guess + a); + oldval = cmpxchg(&v->counter, guess, guess + a); } while (guess != oldval); return oldval; } @@ -83,25 +65,6 @@ static inline int __atomic_add_unless(atomic_t *v, int a, int u) #define atomic64_read(v) ((v)->counter) #define atomic64_set(v, i) ((v)->counter = (i)) -static inline long atomic64_cmpxchg(atomic64_t *v, long o, long n) -{ - long val; - smp_mb(); /* barrier for proper semantics */ - __insn_mtspr(SPR_CMPEXCH_VALUE, o); - val = __insn_cmpexch((void *)&v->counter, n); - smp_mb(); /* barrier for proper semantics */ - return val; -} - -static inline long atomic64_xchg(atomic64_t *v, long n) -{ - long val; - smp_mb(); /* barrier for proper semantics */ - val = __insn_exch((void *)&v->counter, n); - smp_mb(); /* barrier for proper semantics */ - return val; -} - static inline void atomic64_add(long i, atomic64_t *v) { __insn_fetchadd((void *)&v->counter, i); @@ -123,7 +86,7 @@ static inline long atomic64_add_unless(atomic64_t *v, long a, long u) if (oldval == u) break; guess = oldval; - oldval = atomic64_cmpxchg(v, guess, guess + a); + oldval = cmpxchg(&v->counter, guess, guess + a); } while (guess != oldval); return oldval != u; } @@ -142,12 +105,6 @@ static inline long atomic64_add_unless(atomic64_t *v, long a, long u) #define atomic64_inc_not_zero(v) atomic64_add_unless((v), 1, 0) -/* Atomic dec and inc don't implement barrier, so provide them if needed. */ -#define smp_mb__before_atomic_dec() smp_mb() -#define smp_mb__after_atomic_dec() smp_mb() -#define smp_mb__before_atomic_inc() smp_mb() -#define smp_mb__after_atomic_inc() smp_mb() - /* Define this to indicate that cmpxchg is an efficient operation. */ #define __HAVE_ARCH_CMPXCHG diff --git a/arch/tile/include/asm/barrier.h b/arch/tile/include/asm/barrier.h new file mode 100644 index 00000000000..96a42ae79f4 --- /dev/null +++ b/arch/tile/include/asm/barrier.h @@ -0,0 +1,92 @@ +/* + * Copyright 2010 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + */ + +#ifndef _ASM_TILE_BARRIER_H +#define _ASM_TILE_BARRIER_H + +#ifndef __ASSEMBLY__ + +#include <linux/types.h> +#include <arch/chip.h> +#include <arch/spr_def.h> +#include <asm/timex.h> + +#define __sync() __insn_mf() + +#include <hv/syscall_public.h> +/* + * Issue an uncacheable load to each memory controller, then + * wait until those loads have completed. + */ +static inline void __mb_incoherent(void) +{ + long clobber_r10; + asm volatile("swint2" + : "=R10" (clobber_r10) + : "R10" (HV_SYS_fence_incoherent) + : "r0", "r1", "r2", "r3", "r4", + "r5", "r6", "r7", "r8", "r9", + "r11", "r12", "r13", "r14", + "r15", "r16", "r17", "r18", "r19", + "r20", "r21", "r22", "r23", "r24", + "r25", "r26", "r27", "r28", "r29"); +} + +/* Fence to guarantee visibility of stores to incoherent memory. */ +static inline void +mb_incoherent(void) +{ + __insn_mf(); + + { +#if CHIP_HAS_TILE_WRITE_PENDING() + const unsigned long WRITE_TIMEOUT_CYCLES = 400; + unsigned long start = get_cycles_low(); + do { + if (__insn_mfspr(SPR_TILE_WRITE_PENDING) == 0) + return; + } while ((get_cycles_low() - start) < WRITE_TIMEOUT_CYCLES); +#endif /* CHIP_HAS_TILE_WRITE_PENDING() */ + (void) __mb_incoherent(); + } +} + +#define fast_wmb() __sync() +#define fast_rmb() __sync() +#define fast_mb() __sync() +#define fast_iob() mb_incoherent() + +#define wmb() fast_wmb() +#define rmb() fast_rmb() +#define mb() fast_mb() +#define iob() fast_iob() + +#ifndef __tilegx__ /* 32 bit */ +/* + * We need to barrier before modifying the word, since the _atomic_xxx() + * routines just tns the lock and then read/modify/write of the word. + * But after the word is updated, the routine issues an "mf" before returning, + * and since it's a function call, we don't even need a compiler barrier. + */ +#define smp_mb__before_atomic() smp_mb() +#define smp_mb__after_atomic() do { } while (0) +#else /* 64 bit */ +#define smp_mb__before_atomic() smp_mb() +#define smp_mb__after_atomic() smp_mb() +#endif + +#include <asm-generic/barrier.h> + +#endif /* !__ASSEMBLY__ */ +#endif /* _ASM_TILE_BARRIER_H */ diff --git a/arch/tile/include/asm/bitops.h b/arch/tile/include/asm/bitops.h index 16f1fa51fea..20caa346ac0 100644 --- a/arch/tile/include/asm/bitops.h +++ b/arch/tile/include/asm/bitops.h @@ -17,6 +17,7 @@ #define _ASM_TILE_BITOPS_H #include <linux/types.h> +#include <asm/barrier.h> #ifndef _LINUX_BITOPS_H #error only <linux/bitops.h> can be included directly @@ -29,17 +30,6 @@ #endif /** - * __ffs - find first set bit in word - * @word: The word to search - * - * Undefined if no set bit exists, so code should check against 0 first. - */ -static inline unsigned long __ffs(unsigned long word) -{ - return __builtin_ctzl(word); -} - -/** * ffz - find first zero bit in word * @word: The word to search * @@ -50,31 +40,9 @@ static inline unsigned long ffz(unsigned long word) return __builtin_ctzl(~word); } -/** - * __fls - find last set bit in word - * @word: The word to search - * - * Undefined if no set bit exists, so code should check against 0 first. - */ -static inline unsigned long __fls(unsigned long word) -{ - return (sizeof(word) * 8) - 1 - __builtin_clzl(word); -} - -/** - * ffs - find first set bit in word - * @x: the word to search - * - * This is defined the same way as the libc and compiler builtin ffs - * routines, therefore differs in spirit from the other bitops. - * - * ffs(value) returns 0 if value is 0 or the position of the first - * set bit if value is nonzero. The first (least significant) bit - * is at position 1. - */ -static inline int ffs(int x) +static inline int fls64(__u64 w) { - return __builtin_ffs(x); + return (sizeof(__u64) * 8) - __builtin_clzll(w); } /** @@ -90,12 +58,7 @@ static inline int ffs(int x) */ static inline int fls(int x) { - return (sizeof(int) * 8) - __builtin_clz(x); -} - -static inline int fls64(__u64 w) -{ - return (sizeof(__u64) * 8) - __builtin_clzll(w); + return fls64((unsigned int) x); } static inline unsigned int __arch_hweight32(unsigned int w) @@ -118,6 +81,9 @@ static inline unsigned long __arch_hweight64(__u64 w) return __builtin_popcountll(w); } +#include <asm-generic/bitops/builtin-__ffs.h> +#include <asm-generic/bitops/builtin-__fls.h> +#include <asm-generic/bitops/builtin-ffs.h> #include <asm-generic/bitops/const_hweight.h> #include <asm-generic/bitops/lock.h> #include <asm-generic/bitops/find.h> diff --git a/arch/tile/include/asm/bitops_32.h b/arch/tile/include/asm/bitops_32.h index 571b118bfd9..bbf7b666f21 100644 --- a/arch/tile/include/asm/bitops_32.h +++ b/arch/tile/include/asm/bitops_32.h @@ -16,8 +16,7 @@ #define _ASM_TILE_BITOPS_32_H #include <linux/compiler.h> -#include <linux/atomic.h> -#include <asm/system.h> +#include <asm/barrier.h> /* Tile-specific routines to support <asm/bitops.h>. */ unsigned long _atomic_or(volatile unsigned long *p, unsigned long mask); @@ -50,8 +49,8 @@ static inline void set_bit(unsigned nr, volatile unsigned long *addr) * restricted to acting on a single-word quantity. * * clear_bit() may not contain a memory barrier, so if it is used for - * locking purposes, you should call smp_mb__before_clear_bit() and/or - * smp_mb__after_clear_bit() to ensure changes are visible on other cpus. + * locking purposes, you should call smp_mb__before_atomic() and/or + * smp_mb__after_atomic() to ensure changes are visible on other cpus. */ static inline void clear_bit(unsigned nr, volatile unsigned long *addr) { @@ -122,10 +121,6 @@ static inline int test_and_change_bit(unsigned nr, return (_atomic_xor(addr, mask) & mask) != 0; } -/* See discussion at smp_mb__before_atomic_dec() in <asm/atomic_32.h>. */ -#define smp_mb__before_clear_bit() smp_mb() -#define smp_mb__after_clear_bit() do {} while (0) - #include <asm-generic/bitops/ext2-atomic.h> #endif /* _ASM_TILE_BITOPS_32_H */ diff --git a/arch/tile/include/asm/bitops_64.h b/arch/tile/include/asm/bitops_64.h index e9c8e381ee0..bb1a29221fc 100644 --- a/arch/tile/include/asm/bitops_64.h +++ b/arch/tile/include/asm/bitops_64.h @@ -16,8 +16,7 @@ #define _ASM_TILE_BITOPS_64_H #include <linux/compiler.h> -#include <linux/atomic.h> -#include <asm/system.h> +#include <asm/cmpxchg.h> /* See <asm/bitops.h> for API comments. */ @@ -33,20 +32,15 @@ static inline void clear_bit(unsigned nr, volatile unsigned long *addr) __insn_fetchand((void *)(addr + nr / BITS_PER_LONG), ~mask); } -#define smp_mb__before_clear_bit() smp_mb() -#define smp_mb__after_clear_bit() smp_mb() - - static inline void change_bit(unsigned nr, volatile unsigned long *addr) { - unsigned long old, mask = (1UL << (nr % BITS_PER_LONG)); - long guess, oldval; + unsigned long mask = (1UL << (nr % BITS_PER_LONG)); + unsigned long guess, oldval; addr += nr / BITS_PER_LONG; - old = *addr; + oldval = *addr; do { guess = oldval; - oldval = atomic64_cmpxchg((atomic64_t *)addr, - guess, guess ^ mask); + oldval = cmpxchg(addr, guess, guess ^ mask); } while (guess != oldval); } @@ -86,13 +80,12 @@ static inline int test_and_change_bit(unsigned nr, volatile unsigned long *addr) { unsigned long mask = (1UL << (nr % BITS_PER_LONG)); - long guess, oldval = *addr; + unsigned long guess, oldval; addr += nr / BITS_PER_LONG; oldval = *addr; do { guess = oldval; - oldval = atomic64_cmpxchg((atomic64_t *)addr, - guess, guess ^ mask); + oldval = cmpxchg(addr, guess, guess ^ mask); } while (guess != oldval); return (oldval & mask) != 0; } diff --git a/arch/tile/include/asm/bitsperlong.h b/arch/tile/include/asm/bitsperlong.h deleted file mode 100644 index 58c771f2af2..00000000000 --- a/arch/tile/include/asm/bitsperlong.h +++ /dev/null @@ -1,26 +0,0 @@ -/* - * Copyright 2010 Tilera Corporation. All Rights Reserved. - * - * This program 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, version 2. - * - * This program 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, GOOD TITLE or - * NON INFRINGEMENT. See the GNU General Public License for - * more details. - */ - -#ifndef _ASM_TILE_BITSPERLONG_H -#define _ASM_TILE_BITSPERLONG_H - -#ifdef __LP64__ -# define __BITS_PER_LONG 64 -#else -# define __BITS_PER_LONG 32 -#endif - -#include <asm-generic/bitsperlong.h> - -#endif /* _ASM_TILE_BITSPERLONG_H */ diff --git a/arch/tile/include/asm/byteorder.h b/arch/tile/include/asm/byteorder.h deleted file mode 100644 index 9558416d578..00000000000 --- a/arch/tile/include/asm/byteorder.h +++ /dev/null @@ -1 +0,0 @@ -#include <linux/byteorder/little_endian.h> diff --git a/arch/tile/include/asm/cache.h b/arch/tile/include/asm/cache.h index 392e5333dd8..6160761d5f6 100644 --- a/arch/tile/include/asm/cache.h +++ b/arch/tile/include/asm/cache.h @@ -27,11 +27,17 @@ #define L2_CACHE_ALIGN(x) (((x)+(L2_CACHE_BYTES-1)) & -L2_CACHE_BYTES) /* - * TILE-Gx is fully coherent so we don't need to define ARCH_DMA_MINALIGN. + * TILEPro I/O is not always coherent (networking typically uses coherent + * I/O, but PCI traffic does not) and setting ARCH_DMA_MINALIGN to the + * L2 cacheline size helps ensure that kernel heap allocations are aligned. + * TILE-Gx I/O is always coherent when used on hash-for-home pages. + * + * However, it's possible at runtime to request not to use hash-for-home + * for the kernel heap, in which case the kernel will use flush-and-inval + * to manage coherence. As a result, we use L2_CACHE_BYTES for the + * DMA minimum alignment to avoid false sharing in the kernel heap. */ -#ifndef __tilegx__ #define ARCH_DMA_MINALIGN L2_CACHE_BYTES -#endif /* use the cache line size for the L2, which is where it counts */ #define SMP_CACHE_BYTES_SHIFT L2_CACHE_SHIFT @@ -43,9 +49,16 @@ #define __read_mostly __attribute__((__section__(".data..read_mostly"))) /* - * Attribute for data that is kept read/write coherent until the end of - * initialization, then bumped to read/only incoherent for performance. + * Originally we used small TLB pages for kernel data and grouped some + * things together as "write once", enforcing the property at the end + * of initialization by making those pages read-only and non-coherent. + * This allowed better cache utilization since cache inclusion did not + * need to be maintained. However, to do this requires an extra TLB + * entry, which on balance is more of a performance hit than the + * non-coherence is a performance gain, so we now just make "read + * mostly" and "write once" be synonyms. We keep the attribute + * separate in case we change our minds at a future date. */ -#define __write_once __attribute__((__section__(".w1data"))) +#define __write_once __read_mostly #endif /* _ASM_TILE_CACHE_H */ diff --git a/arch/tile/include/asm/cacheflush.h b/arch/tile/include/asm/cacheflush.h index e925f4bb498..92ee4c8a4f7 100644 --- a/arch/tile/include/asm/cacheflush.h +++ b/arch/tile/include/asm/cacheflush.h @@ -20,7 +20,6 @@ /* Keep includes the same across arches. */ #include <linux/mm.h> #include <linux/cache.h> -#include <asm/system.h> #include <arch/icache.h> /* Caches are physically-indexed and so don't need special treatment */ @@ -76,23 +75,6 @@ static inline void copy_to_user_page(struct vm_area_struct *vma, #define copy_from_user_page(vma, page, vaddr, dst, src, len) \ memcpy((dst), (src), (len)) -/* - * Invalidate a VA range; pads to L2 cacheline boundaries. - * - * Note that on TILE64, __inv_buffer() actually flushes modified - * cache lines in addition to invalidating them, i.e., it's the - * same as __finv_buffer(). - */ -static inline void __inv_buffer(void *buffer, size_t size) -{ - char *next = (char *)((long)buffer & -L2_CACHE_BYTES); - char *finish = (char *)L2_CACHE_ALIGN((long)buffer + size); - while (next < finish) { - __insn_inv(next); - next += CHIP_INV_STRIDE(); - } -} - /* Flush a VA range; pads to L2 cacheline boundaries. */ static inline void __flush_buffer(void *buffer, size_t size) { @@ -116,13 +98,6 @@ static inline void __finv_buffer(void *buffer, size_t size) } -/* Invalidate a VA range and wait for it to be complete. */ -static inline void inv_buffer(void *buffer, size_t size) -{ - __inv_buffer(buffer, size); - mb(); -} - /* * Flush a locally-homecached VA range and wait for the evicted * cachelines to hit memory. @@ -143,6 +118,26 @@ static inline void finv_buffer_local(void *buffer, size_t size) mb_incoherent(); } +#ifdef __tilepro__ +/* Invalidate a VA range; pads to L2 cacheline boundaries. */ +static inline void __inv_buffer(void *buffer, size_t size) +{ + char *next = (char *)((long)buffer & -L2_CACHE_BYTES); + char *finish = (char *)L2_CACHE_ALIGN((long)buffer + size); + while (next < finish) { + __insn_inv(next); + next += CHIP_INV_STRIDE(); + } +} + +/* Invalidate a VA range and wait for it to be complete. */ +static inline void inv_buffer(void *buffer, size_t size) +{ + __inv_buffer(buffer, size); + mb(); +} +#endif + /* * Flush and invalidate a VA range that is homed remotely, waiting * until the memory controller holds the flushed values. If "hfh" is @@ -152,4 +147,14 @@ static inline void finv_buffer_local(void *buffer, size_t size) */ void finv_buffer_remote(void *buffer, size_t size, int hfh); +/* + * On SMP systems, when the scheduler does migration-cost autodetection, + * it needs a way to flush as much of the CPU's caches as possible: + * + * TODO: fill this in! + */ +static inline void sched_cacheflush(void) +{ +} + #endif /* _ASM_TILE_CACHEFLUSH_H */ diff --git a/arch/tile/include/asm/checksum.h b/arch/tile/include/asm/checksum.h index a120766c726..b21a2fdec9f 100644 --- a/arch/tile/include/asm/checksum.h +++ b/arch/tile/include/asm/checksum.h @@ -21,4 +21,22 @@ __wsum do_csum(const unsigned char *buff, int len); #define do_csum do_csum +/* + * Return the sum of all the 16-bit subwords in a long. + * This sums two subwords on a 32-bit machine, and four on 64 bits. + * The implementation does two vector adds to capture any overflow. + */ +static inline unsigned int csum_long(unsigned long x) +{ + unsigned long ret; +#ifdef __tilegx__ + ret = __insn_v2sadu(x, 0); + ret = __insn_v2sadu(ret, 0); +#else + ret = __insn_sadh_u(x, 0); + ret = __insn_sadh_u(ret, 0); +#endif + return ret; +} + #endif /* _ASM_TILE_CHECKSUM_H */ diff --git a/arch/tile/include/asm/cmpxchg.h b/arch/tile/include/asm/cmpxchg.h new file mode 100644 index 00000000000..0ccda3c425b --- /dev/null +++ b/arch/tile/include/asm/cmpxchg.h @@ -0,0 +1,134 @@ +/* + * cmpxchg.h -- forked from asm/atomic.h with this copyright: + * + * Copyright 2010 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + * + */ + +#ifndef _ASM_TILE_CMPXCHG_H +#define _ASM_TILE_CMPXCHG_H + +#ifndef __ASSEMBLY__ + +#include <asm/barrier.h> + +/* Nonexistent functions intended to cause compile errors. */ +extern void __xchg_called_with_bad_pointer(void) + __compiletime_error("Bad argument size for xchg"); +extern void __cmpxchg_called_with_bad_pointer(void) + __compiletime_error("Bad argument size for cmpxchg"); + +#ifndef __tilegx__ + +/* Note the _atomic_xxx() routines include a final mb(). */ +int _atomic_xchg(int *ptr, int n); +int _atomic_xchg_add(int *v, int i); +int _atomic_xchg_add_unless(int *v, int a, int u); +int _atomic_cmpxchg(int *ptr, int o, int n); +long long _atomic64_xchg(long long *v, long long n); +long long _atomic64_xchg_add(long long *v, long long i); +long long _atomic64_xchg_add_unless(long long *v, long long a, long long u); +long long _atomic64_cmpxchg(long long *v, long long o, long long n); + +#define xchg(ptr, n) \ + ({ \ + if (sizeof(*(ptr)) != 4) \ + __xchg_called_with_bad_pointer(); \ + smp_mb(); \ + (typeof(*(ptr)))_atomic_xchg((int *)(ptr), (int)(n)); \ + }) + +#define cmpxchg(ptr, o, n) \ + ({ \ + if (sizeof(*(ptr)) != 4) \ + __cmpxchg_called_with_bad_pointer(); \ + smp_mb(); \ + (typeof(*(ptr)))_atomic_cmpxchg((int *)ptr, (int)o, \ + (int)n); \ + }) + +#define xchg64(ptr, n) \ + ({ \ + if (sizeof(*(ptr)) != 8) \ + __xchg_called_with_bad_pointer(); \ + smp_mb(); \ + (typeof(*(ptr)))_atomic64_xchg((long long *)(ptr), \ + (long long)(n)); \ + }) + +#define cmpxchg64(ptr, o, n) \ + ({ \ + if (sizeof(*(ptr)) != 8) \ + __cmpxchg_called_with_bad_pointer(); \ + smp_mb(); \ + (typeof(*(ptr)))_atomic64_cmpxchg((long long *)ptr, \ + (long long)o, (long long)n); \ + }) + +#else + +#define xchg(ptr, n) \ + ({ \ + typeof(*(ptr)) __x; \ + smp_mb(); \ + switch (sizeof(*(ptr))) { \ + case 4: \ + __x = (typeof(__x))(unsigned long) \ + __insn_exch4((ptr), \ + (u32)(unsigned long)(n)); \ + break; \ + case 8: \ + __x = (typeof(__x)) \ + __insn_exch((ptr), (unsigned long)(n)); \ + break; \ + default: \ + __xchg_called_with_bad_pointer(); \ + break; \ + } \ + smp_mb(); \ + __x; \ + }) + +#define cmpxchg(ptr, o, n) \ + ({ \ + typeof(*(ptr)) __x; \ + __insn_mtspr(SPR_CMPEXCH_VALUE, (unsigned long)(o)); \ + smp_mb(); \ + switch (sizeof(*(ptr))) { \ + case 4: \ + __x = (typeof(__x))(unsigned long) \ + __insn_cmpexch4((ptr), \ + (u32)(unsigned long)(n)); \ + break; \ + case 8: \ + __x = (typeof(__x))__insn_cmpexch((ptr), \ + (long long)(n)); \ + break; \ + default: \ + __cmpxchg_called_with_bad_pointer(); \ + break; \ + } \ + smp_mb(); \ + __x; \ + }) + +#define xchg64 xchg +#define cmpxchg64 cmpxchg + +#endif + +#define tas(ptr) xchg((ptr), 1) + +#endif /* __ASSEMBLY__ */ + +#endif /* _ASM_TILE_CMPXCHG_H */ diff --git a/arch/tile/include/asm/compat.h b/arch/tile/include/asm/compat.h index bf95f55b82b..ffd4493efc7 100644 --- a/arch/tile/include/asm/compat.h +++ b/arch/tile/include/asm/compat.h @@ -44,7 +44,6 @@ typedef __kernel_uid32_t __compat_gid32_t; typedef __kernel_mode_t compat_mode_t; typedef __kernel_dev_t compat_dev_t; typedef __kernel_loff_t compat_loff_t; -typedef __kernel_nlink_t compat_nlink_t; typedef __kernel_ipc_pid_t compat_ipc_pid_t; typedef __kernel_daddr_t compat_daddr_t; typedef __kernel_fsid_t compat_fsid_t; @@ -111,6 +110,68 @@ struct compat_flock64 { typedef u32 compat_sigset_word; +typedef union compat_sigval { + compat_int_t sival_int; + compat_uptr_t sival_ptr; +} compat_sigval_t; + +#define COMPAT_SI_PAD_SIZE (128/sizeof(int) - 3) + +typedef struct compat_siginfo { + int si_signo; + int si_errno; + int si_code; + + union { + int _pad[COMPAT_SI_PAD_SIZE]; + + /* kill() */ + struct { + unsigned int _pid; /* sender's pid */ + unsigned int _uid; /* sender's uid */ + } _kill; + + /* POSIX.1b timers */ + struct { + compat_timer_t _tid; /* timer id */ + int _overrun; /* overrun count */ + compat_sigval_t _sigval; /* same as below */ + int _sys_private; /* not to be passed to user */ + int _overrun_incr; /* amount to add to overrun */ + } _timer; + + /* POSIX.1b signals */ + struct { + unsigned int _pid; /* sender's pid */ + unsigned int _uid; /* sender's uid */ + compat_sigval_t _sigval; + } _rt; + + /* SIGCHLD */ + struct { + unsigned int _pid; /* which child */ + unsigned int _uid; /* sender's uid */ + int _status; /* exit code */ + compat_clock_t _utime; + compat_clock_t _stime; + } _sigchld; + + /* SIGILL, SIGFPE, SIGSEGV, SIGBUS */ + struct { + unsigned int _addr; /* faulting insn/memory ref. */ +#ifdef __ARCH_SI_TRAPNO + int _trapno; /* TRAP # which caused the signal */ +#endif + } _sigfault; + + /* SIGPOLL */ + struct { + int _band; /* POLL_IN, POLL_OUT, POLL_MSG */ + int _fd; + } _sigpoll; + } _sifields; +} compat_siginfo_t; + #define COMPAT_OFF_T_MAX 0x7fffffff #define COMPAT_LOFF_T_MAX 0x7fffffffffffffffL @@ -211,57 +272,26 @@ extern int compat_setup_rt_frame(int sig, struct k_sigaction *ka, struct pt_regs *regs); /* Compat syscalls. */ -struct compat_sigaction; struct compat_siginfo; struct compat_sigaltstack; -long compat_sys_execve(const char __user *path, - compat_uptr_t __user *argv, - compat_uptr_t __user *envp, struct pt_regs *); -long compat_sys_rt_sigaction(int sig, struct compat_sigaction __user *act, - struct compat_sigaction __user *oact, - size_t sigsetsize); -long compat_sys_rt_sigqueueinfo(int pid, int sig, - struct compat_siginfo __user *uinfo); -long compat_sys_rt_sigreturn(struct pt_regs *); -long compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr, - struct compat_sigaltstack __user *uoss_ptr, - struct pt_regs *); +long compat_sys_rt_sigreturn(void); long compat_sys_truncate64(char __user *filename, u32 dummy, u32 low, u32 high); long compat_sys_ftruncate64(unsigned int fd, u32 dummy, u32 low, u32 high); long compat_sys_pread64(unsigned int fd, char __user *ubuf, size_t count, u32 dummy, u32 low, u32 high); long compat_sys_pwrite64(unsigned int fd, char __user *ubuf, size_t count, u32 dummy, u32 low, u32 high); -long compat_sys_lookup_dcookie(u32 low, u32 high, char __user *buf, size_t len); long compat_sys_sync_file_range2(int fd, unsigned int flags, u32 offset_lo, u32 offset_hi, u32 nbytes_lo, u32 nbytes_hi); long compat_sys_fallocate(int fd, int mode, u32 offset_lo, u32 offset_hi, u32 len_lo, u32 len_hi); -long compat_sys_sched_rr_get_interval(compat_pid_t pid, - struct compat_timespec __user *interval); - -/* Versions of compat functions that differ from generic Linux. */ -struct compat_msgbuf; -long tile_compat_sys_msgsnd(int msqid, - struct compat_msgbuf __user *msgp, - size_t msgsz, int msgflg); -long tile_compat_sys_msgrcv(int msqid, - struct compat_msgbuf __user *msgp, - size_t msgsz, long msgtyp, int msgflg); -long tile_compat_sys_ptrace(compat_long_t request, compat_long_t pid, - compat_long_t addr, compat_long_t data); - -/* Tilera Linux syscalls that don't have "compat" versions. */ -#define compat_sys_flush_cache sys_flush_cache - -/* These are the intvec_64.S trampolines. */ -long _compat_sys_execve(const char __user *path, - const compat_uptr_t __user *argv, - const compat_uptr_t __user *envp); -long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr, - struct compat_sigaltstack __user *uoss_ptr); +long compat_sys_llseek(unsigned int fd, unsigned int offset_high, + unsigned int offset_low, loff_t __user * result, + unsigned int origin); + +/* Assembly trampoline to avoid clobbering r0. */ long _compat_sys_rt_sigreturn(void); #endif /* _ASM_TILE_COMPAT_H */ diff --git a/arch/tile/include/asm/siginfo.h b/arch/tile/include/asm/device.h index 56d661bb010..6ab8bf146d4 100644 --- a/arch/tile/include/asm/siginfo.h +++ b/arch/tile/include/asm/device.h @@ -10,25 +10,27 @@ * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or * NON INFRINGEMENT. See the GNU General Public License for * more details. + * Arch specific extensions to struct device */ -#ifndef _ASM_TILE_SIGINFO_H -#define _ASM_TILE_SIGINFO_H +#ifndef _ASM_TILE_DEVICE_H +#define _ASM_TILE_DEVICE_H -#define __ARCH_SI_TRAPNO +struct dev_archdata { + /* DMA operations on that device */ + struct dma_map_ops *dma_ops; -#ifdef __LP64__ -# define __ARCH_SI_PREAMBLE_SIZE (4 * sizeof(int)) -#endif + /* Offset of the DMA address from the PA. */ + dma_addr_t dma_offset; -#include <asm-generic/siginfo.h> + /* + * Highest DMA address that can be generated by devices that + * have limited DMA capability, i.e. non 64-bit capable. + */ + dma_addr_t max_direct_dma_addr; +}; -/* - * Additional Tile-specific SIGILL si_codes - */ -#define ILL_DBLFLT (__SI_FAULT|9) /* double fault */ -#define ILL_HARDWALL (__SI_FAULT|10) /* user networks hardwall violation */ -#undef NSIGILL -#define NSIGILL 10 +struct pdev_archdata { +}; -#endif /* _ASM_TILE_SIGINFO_H */ +#endif /* _ASM_TILE_DEVICE_H */ diff --git a/arch/tile/include/asm/dma-mapping.h b/arch/tile/include/asm/dma-mapping.h index eaa06d175b3..1eae359d831 100644 --- a/arch/tile/include/asm/dma-mapping.h +++ b/arch/tile/include/asm/dma-mapping.h @@ -20,69 +20,94 @@ #include <linux/cache.h> #include <linux/io.h> -/* - * Note that on x86 and powerpc, there is a "struct dma_mapping_ops" - * that is used for all the DMA operations. For now, we don't have an - * equivalent on tile, because we only have a single way of doing DMA. - * (Tilera bug 7994 to use dma_mapping_ops.) - */ +#ifdef __tilegx__ +#define ARCH_HAS_DMA_GET_REQUIRED_MASK +#endif + +extern struct dma_map_ops *tile_dma_map_ops; +extern struct dma_map_ops *gx_pci_dma_map_ops; +extern struct dma_map_ops *gx_legacy_pci_dma_map_ops; +extern struct dma_map_ops *gx_hybrid_pci_dma_map_ops; + +static inline struct dma_map_ops *get_dma_ops(struct device *dev) +{ + if (dev && dev->archdata.dma_ops) + return dev->archdata.dma_ops; + else + return tile_dma_map_ops; +} + +static inline dma_addr_t get_dma_offset(struct device *dev) +{ + return dev->archdata.dma_offset; +} + +static inline void set_dma_offset(struct device *dev, dma_addr_t off) +{ + dev->archdata.dma_offset = off; +} + +static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr) +{ + return paddr; +} + +static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr) +{ + return daddr; +} + +static inline void dma_mark_clean(void *addr, size_t size) {} + +#include <asm-generic/dma-mapping-common.h> + +static inline void set_dma_ops(struct device *dev, struct dma_map_ops *ops) +{ + dev->archdata.dma_ops = ops; +} -#define dma_alloc_noncoherent(d, s, h, f) dma_alloc_coherent(d, s, h, f) -#define dma_free_noncoherent(d, s, v, h) dma_free_coherent(d, s, v, h) - -extern dma_addr_t dma_map_single(struct device *dev, void *ptr, size_t size, - enum dma_data_direction); -extern void dma_unmap_single(struct device *dev, dma_addr_t dma_addr, - size_t size, enum dma_data_direction); -extern int dma_map_sg(struct device *dev, struct scatterlist *sg, int nents, - enum dma_data_direction); -extern void dma_unmap_sg(struct device *dev, struct scatterlist *sg, - int nhwentries, enum dma_data_direction); -extern dma_addr_t dma_map_page(struct device *dev, struct page *page, - unsigned long offset, size_t size, - enum dma_data_direction); -extern void dma_unmap_page(struct device *dev, dma_addr_t dma_address, - size_t size, enum dma_data_direction); -extern void dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg, - int nelems, enum dma_data_direction); -extern void dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg, - int nelems, enum dma_data_direction); - - -void *dma_alloc_coherent(struct device *dev, size_t size, - dma_addr_t *dma_handle, gfp_t flag); - -void dma_free_coherent(struct device *dev, size_t size, - void *vaddr, dma_addr_t dma_handle); - -extern void dma_sync_single_for_cpu(struct device *, dma_addr_t, size_t, - enum dma_data_direction); -extern void dma_sync_single_for_device(struct device *, dma_addr_t, - size_t, enum dma_data_direction); -extern void dma_sync_single_range_for_cpu(struct device *, dma_addr_t, - unsigned long offset, size_t, - enum dma_data_direction); -extern void dma_sync_single_range_for_device(struct device *, dma_addr_t, - unsigned long offset, size_t, - enum dma_data_direction); -extern void dma_cache_sync(struct device *dev, void *vaddr, size_t, - enum dma_data_direction); +static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size) +{ + if (!dev->dma_mask) + return 0; + + return addr + size - 1 <= *dev->dma_mask; +} static inline int dma_mapping_error(struct device *dev, dma_addr_t dma_addr) { - return 0; + debug_dma_mapping_error(dev, dma_addr); + return get_dma_ops(dev)->mapping_error(dev, dma_addr); } static inline int dma_supported(struct device *dev, u64 mask) { - return 1; + return get_dma_ops(dev)->dma_supported(dev, mask); } static inline int dma_set_mask(struct device *dev, u64 mask) { + struct dma_map_ops *dma_ops = get_dma_ops(dev); + + /* + * For PCI devices with 64-bit DMA addressing capability, promote + * the dma_ops to hybrid, with the consistent memory DMA space limited + * to 32-bit. For 32-bit capable devices, limit the streaming DMA + * address range to max_direct_dma_addr. + */ + if (dma_ops == gx_pci_dma_map_ops || + dma_ops == gx_hybrid_pci_dma_map_ops || + dma_ops == gx_legacy_pci_dma_map_ops) { + if (mask == DMA_BIT_MASK(64) && + dma_ops == gx_legacy_pci_dma_map_ops) + set_dma_ops(dev, gx_hybrid_pci_dma_map_ops); + else if (mask > dev->archdata.max_direct_dma_addr) + mask = dev->archdata.max_direct_dma_addr; + } + if (!dev->dma_mask || !dma_supported(dev, mask)) return -EIO; @@ -91,4 +116,43 @@ dma_set_mask(struct device *dev, u64 mask) return 0; } +static inline void *dma_alloc_attrs(struct device *dev, size_t size, + dma_addr_t *dma_handle, gfp_t flag, + struct dma_attrs *attrs) +{ + struct dma_map_ops *dma_ops = get_dma_ops(dev); + void *cpu_addr; + + cpu_addr = dma_ops->alloc(dev, size, dma_handle, flag, attrs); + + debug_dma_alloc_coherent(dev, size, *dma_handle, cpu_addr); + + return cpu_addr; +} + +static inline void dma_free_attrs(struct device *dev, size_t size, + void *cpu_addr, dma_addr_t dma_handle, + struct dma_attrs *attrs) +{ + struct dma_map_ops *dma_ops = get_dma_ops(dev); + + debug_dma_free_coherent(dev, size, cpu_addr, dma_handle); + + dma_ops->free(dev, size, cpu_addr, dma_handle, attrs); +} + +#define dma_alloc_coherent(d, s, h, f) dma_alloc_attrs(d, s, h, f, NULL) +#define dma_alloc_noncoherent(d, s, h, f) dma_alloc_attrs(d, s, h, f, NULL) +#define dma_free_coherent(d, s, v, h) dma_free_attrs(d, s, v, h, NULL) +#define dma_free_noncoherent(d, s, v, h) dma_free_attrs(d, s, v, h, NULL) + +/* + * dma_alloc_noncoherent() is #defined to return coherent memory, + * so there's no need to do any flushing here. + */ +static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size, + enum dma_data_direction direction) +{ +} + #endif /* _ASM_TILE_DMA_MAPPING_H */ diff --git a/arch/tile/include/asm/elf.h b/arch/tile/include/asm/elf.h index 623a6bb741c..41d9878a968 100644 --- a/arch/tile/include/asm/elf.h +++ b/arch/tile/include/asm/elf.h @@ -30,7 +30,6 @@ typedef unsigned long elf_greg_t; #define ELF_NGREG (sizeof(struct pt_regs) / sizeof(elf_greg_t)) typedef elf_greg_t elf_gregset_t[ELF_NGREG]; -#define EM_TILE64 187 #define EM_TILEPRO 188 #define EM_TILEGX 191 @@ -44,7 +43,11 @@ typedef elf_fpreg_t elf_fpregset_t[ELF_NFPREG]; #else #define ELF_CLASS ELFCLASS32 #endif +#ifdef __BIG_ENDIAN__ +#define ELF_DATA ELFDATA2MSB +#else #define ELF_DATA ELFDATA2LSB +#endif /* * There seems to be a bug in how compat_binfmt_elf.c works: it @@ -59,6 +62,7 @@ enum { ELF_ARCH = CHIP_ELF_TYPE() }; */ #define elf_check_arch(x) \ ((x)->e_ident[EI_CLASS] == ELF_CLASS && \ + (x)->e_ident[EI_DATA] == ELF_DATA && \ (x)->e_machine == CHIP_ELF_TYPE()) /* The module loader only handles a few relocation types. */ @@ -127,6 +131,15 @@ extern int dump_task_regs(struct task_struct *, elf_gregset_t *); struct linux_binprm; extern int arch_setup_additional_pages(struct linux_binprm *bprm, int executable_stack); +#define ARCH_DLINFO \ +do { \ + NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_BASE); \ +} while (0) + +struct mm_struct; +extern unsigned long arch_randomize_brk(struct mm_struct *mm); +#define arch_randomize_brk arch_randomize_brk + #ifdef CONFIG_COMPAT #define COMPAT_ELF_PLATFORM "tilegx-m32" @@ -143,6 +156,7 @@ extern int arch_setup_additional_pages(struct linux_binprm *bprm, #define compat_start_thread(regs, ip, usp) do { \ regs->pc = ptr_to_compat_reg((void *)(ip)); \ regs->sp = ptr_to_compat_reg((void *)(usp)); \ + single_step_execve(); \ } while (0) /* @@ -151,12 +165,12 @@ extern int arch_setup_additional_pages(struct linux_binprm *bprm, #undef SET_PERSONALITY #define SET_PERSONALITY(ex) \ do { \ - current->personality = PER_LINUX; \ + set_personality(PER_LINUX | (current->personality & (~PER_MASK))); \ current_thread_info()->status &= ~TS_COMPAT; \ } while (0) #define COMPAT_SET_PERSONALITY(ex) \ do { \ - current->personality = PER_LINUX_32BIT; \ + set_personality(PER_LINUX | (current->personality & (~PER_MASK))); \ current_thread_info()->status |= TS_COMPAT; \ } while (0) @@ -164,4 +178,6 @@ do { \ #endif /* CONFIG_COMPAT */ +#define CORE_DUMP_USE_REGSET + #endif /* _ASM_TILE_ELF_H */ diff --git a/arch/tile/include/asm/fixmap.h b/arch/tile/include/asm/fixmap.h index c66f7933bea..ffe2637aeb3 100644 --- a/arch/tile/include/asm/fixmap.h +++ b/arch/tile/include/asm/fixmap.h @@ -25,9 +25,6 @@ #include <asm/kmap_types.h> #endif -#define __fix_to_virt(x) (FIXADDR_TOP - ((x) << PAGE_SHIFT)) -#define __virt_to_fix(x) ((FIXADDR_TOP - ((x)&PAGE_MASK)) >> PAGE_SHIFT) - /* * Here we define all the compile-time 'special' virtual * addresses. The point is to have a constant address at @@ -45,15 +42,23 @@ * * TLB entries of such buffers will not be flushed across * task switches. - * - * We don't bother with a FIX_HOLE since above the fixmaps - * is unmapped memory in any case. */ enum fixed_addresses { +#ifdef __tilegx__ + /* + * TILEPro has unmapped memory above so the hole isn't needed, + * and in any case the hole pushes us over a single 16MB pmd. + */ + FIX_HOLE, +#endif #ifdef CONFIG_HIGHMEM FIX_KMAP_BEGIN, /* reserved pte's for temporary kernel mappings */ FIX_KMAP_END = FIX_KMAP_BEGIN+(KM_TYPE_NR*NR_CPUS)-1, #endif +#ifdef __tilegx__ /* see homecache.c */ + FIX_HOMECACHE_BEGIN, + FIX_HOMECACHE_END = FIX_HOMECACHE_BEGIN+(NR_CPUS)-1, +#endif __end_of_permanent_fixed_addresses, /* @@ -70,48 +75,12 @@ enum fixed_addresses { #endif }; -extern void __set_fixmap(enum fixed_addresses idx, - unsigned long phys, pgprot_t flags); - -#define set_fixmap(idx, phys) \ - __set_fixmap(idx, phys, PAGE_KERNEL) -#define clear_fixmap(idx) \ - __set_fixmap(idx, 0, __pgprot(0)) - #define __FIXADDR_SIZE (__end_of_permanent_fixed_addresses << PAGE_SHIFT) #define __FIXADDR_BOOT_SIZE (__end_of_fixed_addresses << PAGE_SHIFT) #define FIXADDR_START (FIXADDR_TOP + PAGE_SIZE - __FIXADDR_SIZE) #define FIXADDR_BOOT_START (FIXADDR_TOP + PAGE_SIZE - __FIXADDR_BOOT_SIZE) -extern void __this_fixmap_does_not_exist(void); - -/* - * 'index to address' translation. If anyone tries to use the idx - * directly without tranlation, we catch the bug with a NULL-deference - * kernel oops. Illegal ranges of incoming indices are caught too. - */ -static __always_inline unsigned long fix_to_virt(const unsigned int idx) -{ - /* - * this branch gets completely eliminated after inlining, - * except when someone tries to use fixaddr indices in an - * illegal way. (such as mixing up address types or using - * out-of-range indices). - * - * If it doesn't get removed, the linker will complain - * loudly with a reasonably clear error message.. - */ - if (idx >= __end_of_fixed_addresses) - __this_fixmap_does_not_exist(); - - return __fix_to_virt(idx); -} - -static inline unsigned long virt_to_fix(const unsigned long vaddr) -{ - BUG_ON(vaddr >= FIXADDR_TOP || vaddr < FIXADDR_START); - return __virt_to_fix(vaddr); -} +#include <asm-generic/fixmap.h> #endif /* !__ASSEMBLY__ */ diff --git a/arch/tile/include/asm/ftrace.h b/arch/tile/include/asm/ftrace.h index 461459b06d9..13a9bb81a8a 100644 --- a/arch/tile/include/asm/ftrace.h +++ b/arch/tile/include/asm/ftrace.h @@ -15,6 +15,26 @@ #ifndef _ASM_TILE_FTRACE_H #define _ASM_TILE_FTRACE_H -/* empty */ +#ifdef CONFIG_FUNCTION_TRACER + +#define MCOUNT_ADDR ((unsigned long)(__mcount)) +#define MCOUNT_INSN_SIZE 8 /* sizeof mcount call */ + +#ifndef __ASSEMBLY__ +extern void __mcount(void); + +#ifdef CONFIG_DYNAMIC_FTRACE +static inline unsigned long ftrace_call_adjust(unsigned long addr) +{ + return addr; +} + +struct dyn_arch_ftrace { +}; +#endif /* CONFIG_DYNAMIC_FTRACE */ + +#endif /* __ASSEMBLY__ */ + +#endif /* CONFIG_FUNCTION_TRACER */ #endif /* _ASM_TILE_FTRACE_H */ diff --git a/arch/tile/include/asm/futex.h b/arch/tile/include/asm/futex.h index d03ec124a59..1a6ef1b69cb 100644 --- a/arch/tile/include/asm/futex.h +++ b/arch/tile/include/asm/futex.h @@ -28,29 +28,82 @@ #include <linux/futex.h> #include <linux/uaccess.h> #include <linux/errno.h> +#include <asm/atomic.h> -extern struct __get_user futex_set(u32 __user *v, int i); -extern struct __get_user futex_add(u32 __user *v, int n); -extern struct __get_user futex_or(u32 __user *v, int n); -extern struct __get_user futex_andn(u32 __user *v, int n); -extern struct __get_user futex_cmpxchg(u32 __user *v, int o, int n); +/* + * Support macros for futex operations. Do not use these macros directly. + * They assume "ret", "val", "oparg", and "uaddr" in the lexical context. + * __futex_cmpxchg() additionally assumes "oldval". + */ + +#ifdef __tilegx__ + +#define __futex_asm(OP) \ + asm("1: {" #OP " %1, %3, %4; movei %0, 0 }\n" \ + ".pushsection .fixup,\"ax\"\n" \ + "0: { movei %0, %5; j 9f }\n" \ + ".section __ex_table,\"a\"\n" \ + ".align 8\n" \ + ".quad 1b, 0b\n" \ + ".popsection\n" \ + "9:" \ + : "=r" (ret), "=r" (val), "+m" (*(uaddr)) \ + : "r" (uaddr), "r" (oparg), "i" (-EFAULT)) + +#define __futex_set() __futex_asm(exch4) +#define __futex_add() __futex_asm(fetchadd4) +#define __futex_or() __futex_asm(fetchor4) +#define __futex_andn() ({ oparg = ~oparg; __futex_asm(fetchand4); }) +#define __futex_cmpxchg() \ + ({ __insn_mtspr(SPR_CMPEXCH_VALUE, oldval); __futex_asm(cmpexch4); }) + +#define __futex_xor() \ + ({ \ + u32 oldval, n = oparg; \ + if ((ret = __get_user(oldval, uaddr)) == 0) { \ + do { \ + oparg = oldval ^ n; \ + __futex_cmpxchg(); \ + } while (ret == 0 && oldval != val); \ + } \ + }) + +/* No need to prefetch, since the atomic ops go to the home cache anyway. */ +#define __futex_prolog() -#ifndef __tilegx__ -extern struct __get_user futex_xor(u32 __user *v, int n); #else -static inline struct __get_user futex_xor(u32 __user *uaddr, int n) -{ - struct __get_user asm_ret = __get_user_4(uaddr); - if (!asm_ret.err) { - int oldval, newval; - do { - oldval = asm_ret.val; - newval = oldval ^ n; - asm_ret = futex_cmpxchg(uaddr, oldval, newval); - } while (asm_ret.err == 0 && oldval != asm_ret.val); + +#define __futex_call(FN) \ + { \ + struct __get_user gu = FN((u32 __force *)uaddr, lock, oparg); \ + val = gu.val; \ + ret = gu.err; \ } - return asm_ret; -} + +#define __futex_set() __futex_call(__atomic_xchg) +#define __futex_add() __futex_call(__atomic_xchg_add) +#define __futex_or() __futex_call(__atomic_or) +#define __futex_andn() __futex_call(__atomic_andn) +#define __futex_xor() __futex_call(__atomic_xor) + +#define __futex_cmpxchg() \ + { \ + struct __get_user gu = __atomic_cmpxchg((u32 __force *)uaddr, \ + lock, oldval, oparg); \ + val = gu.val; \ + ret = gu.err; \ + } + +/* + * Find the lock pointer for the atomic calls to use, and issue a + * prefetch to the user address to bring it into cache. Similar to + * __atomic_setup(), but we can't do a read into the L1 since it might + * fault; instead we do a prefetch into the L2. + */ +#define __futex_prolog() \ + int *lock; \ + __insn_prefetch(uaddr); \ + lock = __atomic_hashed_lock((int __force *)uaddr) #endif static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr) @@ -59,8 +112,12 @@ static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr) int cmp = (encoded_op >> 24) & 15; int oparg = (encoded_op << 8) >> 20; int cmparg = (encoded_op << 20) >> 20; - int ret; - struct __get_user asm_ret; + int uninitialized_var(val), ret; + + __futex_prolog(); + + /* The 32-bit futex code makes this assumption, so validate it here. */ + BUILD_BUG_ON(sizeof(atomic_t) != sizeof(int)); if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) oparg = 1 << oparg; @@ -71,46 +128,45 @@ static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr) pagefault_disable(); switch (op) { case FUTEX_OP_SET: - asm_ret = futex_set(uaddr, oparg); + __futex_set(); break; case FUTEX_OP_ADD: - asm_ret = futex_add(uaddr, oparg); + __futex_add(); break; case FUTEX_OP_OR: - asm_ret = futex_or(uaddr, oparg); + __futex_or(); break; case FUTEX_OP_ANDN: - asm_ret = futex_andn(uaddr, oparg); + __futex_andn(); break; case FUTEX_OP_XOR: - asm_ret = futex_xor(uaddr, oparg); + __futex_xor(); break; default: - asm_ret.err = -ENOSYS; + ret = -ENOSYS; + break; } pagefault_enable(); - ret = asm_ret.err; - if (!ret) { switch (cmp) { case FUTEX_OP_CMP_EQ: - ret = (asm_ret.val == cmparg); + ret = (val == cmparg); break; case FUTEX_OP_CMP_NE: - ret = (asm_ret.val != cmparg); + ret = (val != cmparg); break; case FUTEX_OP_CMP_LT: - ret = (asm_ret.val < cmparg); + ret = (val < cmparg); break; case FUTEX_OP_CMP_GE: - ret = (asm_ret.val >= cmparg); + ret = (val >= cmparg); break; case FUTEX_OP_CMP_LE: - ret = (asm_ret.val <= cmparg); + ret = (val <= cmparg); break; case FUTEX_OP_CMP_GT: - ret = (asm_ret.val > cmparg); + ret = (val > cmparg); break; default: ret = -ENOSYS; @@ -120,22 +176,20 @@ static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr) } static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr, - u32 oldval, u32 newval) + u32 oldval, u32 oparg) { - struct __get_user asm_ret; + int ret, val; + + __futex_prolog(); if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32))) return -EFAULT; - asm_ret = futex_cmpxchg(uaddr, oldval, newval); - *uval = asm_ret.val; - return asm_ret.err; -} + __futex_cmpxchg(); -#ifndef __tilegx__ -/* Return failure from the atomic wrappers. */ -struct __get_user __atomic_bad_address(int __user *addr); -#endif + *uval = val; + return ret; +} #endif /* !__ASSEMBLY__ */ diff --git a/arch/tile/include/asm/hardirq.h b/arch/tile/include/asm/hardirq.h index 822390f9a15..54110af2398 100644 --- a/arch/tile/include/asm/hardirq.h +++ b/arch/tile/include/asm/hardirq.h @@ -42,6 +42,4 @@ DECLARE_PER_CPU(irq_cpustat_t, irq_stat); #include <linux/irq_cpustat.h> /* Standard mappings for irq_cpustat_t above */ -#define HARDIRQ_BITS 8 - #endif /* _ASM_TILE_HARDIRQ_H */ diff --git a/arch/tile/include/asm/hardwall.h b/arch/tile/include/asm/hardwall.h index 2ac422848c7..2f572b6b7bc 100644 --- a/arch/tile/include/asm/hardwall.h +++ b/arch/tile/include/asm/hardwall.h @@ -11,45 +11,13 @@ * NON INFRINGEMENT. See the GNU General Public License for * more details. * - * Provide methods for the HARDWALL_FILE for accessing the UDN. + * Provide methods for access control of per-cpu resources like + * UDN, IDN, or IPI. */ - #ifndef _ASM_TILE_HARDWALL_H #define _ASM_TILE_HARDWALL_H -#include <linux/ioctl.h> - -#define HARDWALL_IOCTL_BASE 0xa2 - -/* - * The HARDWALL_CREATE() ioctl is a macro with a "size" argument. - * The resulting ioctl value is passed to the kernel in conjunction - * with a pointer to a little-endian bitmask of cpus, which must be - * physically in a rectangular configuration on the chip. - * The "size" is the number of bytes of cpu mask data. - */ -#define _HARDWALL_CREATE 1 -#define HARDWALL_CREATE(size) \ - _IOC(_IOC_READ, HARDWALL_IOCTL_BASE, _HARDWALL_CREATE, (size)) - -#define _HARDWALL_ACTIVATE 2 -#define HARDWALL_ACTIVATE \ - _IO(HARDWALL_IOCTL_BASE, _HARDWALL_ACTIVATE) - -#define _HARDWALL_DEACTIVATE 3 -#define HARDWALL_DEACTIVATE \ - _IO(HARDWALL_IOCTL_BASE, _HARDWALL_DEACTIVATE) - -#define _HARDWALL_GET_ID 4 -#define HARDWALL_GET_ID \ - _IO(HARDWALL_IOCTL_BASE, _HARDWALL_GET_ID) - -#ifndef __KERNEL__ - -/* This is the canonical name expected by userspace. */ -#define HARDWALL_FILE "/dev/hardwall" - -#else +#include <uapi/asm/hardwall.h> /* /proc hooks for hardwall. */ struct proc_dir_entry; @@ -59,7 +27,4 @@ int proc_pid_hardwall(struct task_struct *task, char *buffer); #else static inline void proc_tile_hardwall_init(struct proc_dir_entry *root) {} #endif - -#endif - #endif /* _ASM_TILE_HARDWALL_H */ diff --git a/arch/tile/include/asm/highmem.h b/arch/tile/include/asm/highmem.h index b2a6c5de79a..fc8429a31c8 100644 --- a/arch/tile/include/asm/highmem.h +++ b/arch/tile/include/asm/highmem.h @@ -59,7 +59,7 @@ void *kmap_fix_kpte(struct page *page, int finished); /* This macro is used only in map_new_virtual() to map "page". */ #define kmap_prot page_to_kpgprot(page) -void *__kmap_atomic(struct page *page); +void *kmap_atomic(struct page *page); void __kunmap_atomic(void *kvaddr); void *kmap_atomic_pfn(unsigned long pfn); void *kmap_atomic_prot_pfn(unsigned long pfn, pgprot_t prot); diff --git a/arch/tile/include/asm/homecache.h b/arch/tile/include/asm/homecache.h index a8243865d49..7ddd1b8d691 100644 --- a/arch/tile/include/asm/homecache.h +++ b/arch/tile/include/asm/homecache.h @@ -33,8 +33,7 @@ struct zone; /* * Is this page immutable (unwritable) and thus able to be cached more - * widely than would otherwise be possible? On tile64 this means we - * mark the PTE to cache locally; on tilepro it means we have "nc" set. + * widely than would otherwise be possible? This means we have "nc" set. */ #define PAGE_HOME_IMMUTABLE -2 @@ -44,16 +43,8 @@ struct zone; */ #define PAGE_HOME_INCOHERENT -3 -#if CHIP_HAS_CBOX_HOME_MAP() /* Home for the page is distributed via hash-for-home. */ #define PAGE_HOME_HASH -4 -#endif - -/* Homing is unknown or unspecified. Not valid for page_home(). */ -#define PAGE_HOME_UNKNOWN -5 - -/* Home on the current cpu. Not valid for page_home(). */ -#define PAGE_HOME_HERE -6 /* Support wrapper to use instead of explicit hv_flush_remote(). */ extern void flush_remote(unsigned long cache_pfn, unsigned long cache_length, @@ -79,10 +70,17 @@ extern void homecache_change_page_home(struct page *, int order, int home); /* * Flush a page out of whatever cache(s) it is in. * This is more than just finv, since it properly handles waiting - * for the data to reach memory on tilepro, but it can be quite - * heavyweight, particularly on hash-for-home memory. + * for the data to reach memory, but it can be quite + * heavyweight, particularly on incoherent or immutable memory. + */ +extern void homecache_finv_page(struct page *); + +/* + * Flush a page out of the specified home cache. + * Note that the specified home need not be the actual home of the page, + * as for example might be the case when coordinating with I/O devices. */ -extern void homecache_flush_cache(struct page *, int order); +extern void homecache_finv_map_page(struct page *, int home); /* * Allocate a page with the given GFP flags, home, and optionally @@ -104,10 +102,10 @@ extern struct page *homecache_alloc_pages_node(int nid, gfp_t gfp_mask, * routines use homecache_change_page_home() to reset the home * back to the default before returning the page to the allocator. */ +void __homecache_free_pages(struct page *, unsigned int order); void homecache_free_pages(unsigned long addr, unsigned int order); -#define homecache_free_page(page) \ - homecache_free_pages((page), 0) - +#define __homecache_free_page(page) __homecache_free_pages((page), 0) +#define homecache_free_page(page) homecache_free_pages((page), 0) /* diff --git a/arch/tile/include/asm/hugetlb.h b/arch/tile/include/asm/hugetlb.h index d396d180516..3257733003f 100644 --- a/arch/tile/include/asm/hugetlb.h +++ b/arch/tile/include/asm/hugetlb.h @@ -16,6 +16,7 @@ #define _ASM_TILE_HUGETLB_H #include <asm/page.h> +#include <asm-generic/hugetlb.h> static inline int is_hugepage_only_range(struct mm_struct *mm, @@ -106,4 +107,29 @@ static inline void arch_release_hugepage(struct page *page) { } +static inline void arch_clear_hugepage_flags(struct page *page) +{ +} + +#ifdef CONFIG_HUGETLB_SUPER_PAGES +static inline pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma, + struct page *page, int writable) +{ + size_t pagesize = huge_page_size(hstate_vma(vma)); + if (pagesize != PUD_SIZE && pagesize != PMD_SIZE) + entry = pte_mksuper(entry); + return entry; +} +#define arch_make_huge_pte arch_make_huge_pte + +/* Sizes to scale up page size for PTEs with HV_PTE_SUPER bit. */ +enum { + HUGE_SHIFT_PGDIR = 0, + HUGE_SHIFT_PMD = 1, + HUGE_SHIFT_PAGE = 2, + HUGE_SHIFT_ENTRIES +}; +extern int huge_shift[HUGE_SHIFT_ENTRIES]; +#endif + #endif /* _ASM_TILE_HUGETLB_H */ diff --git a/arch/tile/include/asm/io.h b/arch/tile/include/asm/io.h index d2152deb1f3..9fe434969fa 100644 --- a/arch/tile/include/asm/io.h +++ b/arch/tile/include/asm/io.h @@ -19,7 +19,8 @@ #include <linux/bug.h> #include <asm/page.h> -#define IO_SPACE_LIMIT 0xfffffffful +/* Maximum PCI I/O space address supported. */ +#define IO_SPACE_LIMIT 0xffffffff /* * Convert a physical pointer to a virtual kernel pointer for /dev/mem @@ -62,6 +63,92 @@ extern void iounmap(volatile void __iomem *addr); #define mm_ptov(addr) ((void *)phys_to_virt(addr)) #define mm_vtop(addr) ((unsigned long)virt_to_phys(addr)) +#if CHIP_HAS_MMIO() + +/* + * We use inline assembly to guarantee that the compiler does not + * split an access into multiple byte-sized accesses as it might + * sometimes do if a register data structure is marked "packed". + * Obviously on tile we can't tolerate such an access being + * actually unaligned, but we want to avoid the case where the + * compiler conservatively would generate multiple accesses even + * for an aligned read or write. + */ + +static inline u8 __raw_readb(const volatile void __iomem *addr) +{ + return *(const volatile u8 __force *)addr; +} + +static inline u16 __raw_readw(const volatile void __iomem *addr) +{ + u16 ret; + asm volatile("ld2u %0, %1" : "=r" (ret) : "r" (addr)); + barrier(); + return le16_to_cpu(ret); +} + +static inline u32 __raw_readl(const volatile void __iomem *addr) +{ + u32 ret; + /* Sign-extend to conform to u32 ABI sign-extension convention. */ + asm volatile("ld4s %0, %1" : "=r" (ret) : "r" (addr)); + barrier(); + return le32_to_cpu(ret); +} + +static inline u64 __raw_readq(const volatile void __iomem *addr) +{ + u64 ret; + asm volatile("ld %0, %1" : "=r" (ret) : "r" (addr)); + barrier(); + return le64_to_cpu(ret); +} + +static inline void __raw_writeb(u8 val, volatile void __iomem *addr) +{ + *(volatile u8 __force *)addr = val; +} + +static inline void __raw_writew(u16 val, volatile void __iomem *addr) +{ + asm volatile("st2 %0, %1" :: "r" (addr), "r" (cpu_to_le16(val))); +} + +static inline void __raw_writel(u32 val, volatile void __iomem *addr) +{ + asm volatile("st4 %0, %1" :: "r" (addr), "r" (cpu_to_le32(val))); +} + +static inline void __raw_writeq(u64 val, volatile void __iomem *addr) +{ + asm volatile("st %0, %1" :: "r" (addr), "r" (cpu_to_le64(val))); +} + +/* + * The on-chip I/O hardware on tilegx is configured with VA=PA for the + * kernel's PA range. The low-level APIs and field names use "va" and + * "void *" nomenclature, to be consistent with the general notion + * that the addresses in question are virtualizable, but in the kernel + * context we are actually manipulating PA values. (In other contexts, + * e.g. access from user space, we do in fact use real virtual addresses + * in the va fields.) To allow readers of the code to understand what's + * happening, we direct their attention to this comment by using the + * following two functions that just duplicate __va() and __pa(). + */ +typedef unsigned long tile_io_addr_t; +static inline tile_io_addr_t va_to_tile_io_addr(void *va) +{ + BUILD_BUG_ON(sizeof(phys_addr_t) != sizeof(tile_io_addr_t)); + return __pa(va); +} +static inline void *tile_io_addr_to_va(tile_io_addr_t tile_io_addr) +{ + return __va(tile_io_addr); +} + +#else /* CHIP_HAS_MMIO() */ + #ifdef CONFIG_PCI extern u8 _tile_readb(unsigned long addr); @@ -73,10 +160,19 @@ extern void _tile_writew(u16 val, unsigned long addr); extern void _tile_writel(u32 val, unsigned long addr); extern void _tile_writeq(u64 val, unsigned long addr); -#else +#define __raw_readb(addr) _tile_readb((unsigned long)addr) +#define __raw_readw(addr) _tile_readw((unsigned long)addr) +#define __raw_readl(addr) _tile_readl((unsigned long)addr) +#define __raw_readq(addr) _tile_readq((unsigned long)addr) +#define __raw_writeb(val, addr) _tile_writeb(val, (unsigned long)addr) +#define __raw_writew(val, addr) _tile_writew(val, (unsigned long)addr) +#define __raw_writel(val, addr) _tile_writel(val, (unsigned long)addr) +#define __raw_writeq(val, addr) _tile_writeq(val, (unsigned long)addr) + +#else /* CONFIG_PCI */ /* - * The Tile architecture does not support IOMEM unless PCI is enabled. + * The tilepro architecture does not support IOMEM unless PCI is enabled. * Unfortunately we can't yet simply not declare these methods, * since some generic code that compiles into the kernel, but * we never run, uses them unconditionally. @@ -88,65 +184,58 @@ static inline int iomem_panic(void) return 0; } -static inline u8 _tile_readb(unsigned long addr) +static inline u8 readb(unsigned long addr) { return iomem_panic(); } -static inline u16 _tile_readw(unsigned long addr) +static inline u16 _readw(unsigned long addr) { return iomem_panic(); } -static inline u32 _tile_readl(unsigned long addr) +static inline u32 readl(unsigned long addr) { return iomem_panic(); } -static inline u64 _tile_readq(unsigned long addr) +static inline u64 readq(unsigned long addr) { return iomem_panic(); } -static inline void _tile_writeb(u8 val, unsigned long addr) +static inline void writeb(u8 val, unsigned long addr) { iomem_panic(); } -static inline void _tile_writew(u16 val, unsigned long addr) +static inline void writew(u16 val, unsigned long addr) { iomem_panic(); } -static inline void _tile_writel(u32 val, unsigned long addr) +static inline void writel(u32 val, unsigned long addr) { iomem_panic(); } -static inline void _tile_writeq(u64 val, unsigned long addr) +static inline void writeq(u64 val, unsigned long addr) { iomem_panic(); } -#endif +#endif /* CONFIG_PCI */ -#define readb(addr) _tile_readb((unsigned long)addr) -#define readw(addr) _tile_readw((unsigned long)addr) -#define readl(addr) _tile_readl((unsigned long)addr) -#define readq(addr) _tile_readq((unsigned long)addr) -#define writeb(val, addr) _tile_writeb(val, (unsigned long)addr) -#define writew(val, addr) _tile_writew(val, (unsigned long)addr) -#define writel(val, addr) _tile_writel(val, (unsigned long)addr) -#define writeq(val, addr) _tile_writeq(val, (unsigned long)addr) - -#define __raw_readb readb -#define __raw_readw readw -#define __raw_readl readl -#define __raw_readq readq -#define __raw_writeb writeb -#define __raw_writew writew -#define __raw_writel writel -#define __raw_writeq writeq +#endif /* CHIP_HAS_MMIO() */ + +#define readb __raw_readb +#define readw __raw_readw +#define readl __raw_readl +#define readq __raw_readq +#define writeb __raw_writeb +#define writew __raw_writew +#define writel __raw_writel +#define writeq __raw_writeq #define readb_relaxed readb #define readw_relaxed readw @@ -162,9 +251,11 @@ static inline void _tile_writeq(u64 val, unsigned long addr) #define iowrite32 writel #define iowrite64 writeq -static inline void memset_io(void *dst, int val, size_t len) +#if CHIP_HAS_MMIO() || defined(CONFIG_PCI) + +static inline void memset_io(volatile void *dst, int val, size_t len) { - int x; + size_t x; BUG_ON((unsigned long)dst & 0x3); val = (val & 0xff) * 0x01010101; for (x = 0; x < len; x += 4) @@ -174,7 +265,7 @@ static inline void memset_io(void *dst, int val, size_t len) static inline void memcpy_fromio(void *dst, const volatile void __iomem *src, size_t len) { - int x; + size_t x; BUG_ON((unsigned long)src & 0x3); for (x = 0; x < len; x += 4) *(u32 *)(dst + x) = readl(src + x); @@ -183,14 +274,116 @@ static inline void memcpy_fromio(void *dst, const volatile void __iomem *src, static inline void memcpy_toio(volatile void __iomem *dst, const void *src, size_t len) { - int x; + size_t x; BUG_ON((unsigned long)dst & 0x3); for (x = 0; x < len; x += 4) writel(*(u32 *)(src + x), dst + x); } +#endif + +#if CHIP_HAS_MMIO() && defined(CONFIG_TILE_PCI_IO) + +static inline u8 inb(unsigned long addr) +{ + return readb((volatile void __iomem *) addr); +} + +static inline u16 inw(unsigned long addr) +{ + return readw((volatile void __iomem *) addr); +} + +static inline u32 inl(unsigned long addr) +{ + return readl((volatile void __iomem *) addr); +} + +static inline void outb(u8 b, unsigned long addr) +{ + writeb(b, (volatile void __iomem *) addr); +} + +static inline void outw(u16 b, unsigned long addr) +{ + writew(b, (volatile void __iomem *) addr); +} + +static inline void outl(u32 b, unsigned long addr) +{ + writel(b, (volatile void __iomem *) addr); +} + +static inline void insb(unsigned long addr, void *buffer, int count) +{ + if (count) { + u8 *buf = buffer; + do { + u8 x = inb(addr); + *buf++ = x; + } while (--count); + } +} + +static inline void insw(unsigned long addr, void *buffer, int count) +{ + if (count) { + u16 *buf = buffer; + do { + u16 x = inw(addr); + *buf++ = x; + } while (--count); + } +} + +static inline void insl(unsigned long addr, void *buffer, int count) +{ + if (count) { + u32 *buf = buffer; + do { + u32 x = inl(addr); + *buf++ = x; + } while (--count); + } +} + +static inline void outsb(unsigned long addr, const void *buffer, int count) +{ + if (count) { + const u8 *buf = buffer; + do { + outb(*buf++, addr); + } while (--count); + } +} + +static inline void outsw(unsigned long addr, const void *buffer, int count) +{ + if (count) { + const u16 *buf = buffer; + do { + outw(*buf++, addr); + } while (--count); + } +} + +static inline void outsl(unsigned long addr, const void *buffer, int count) +{ + if (count) { + const u32 *buf = buffer; + do { + outl(*buf++, addr); + } while (--count); + } +} + +extern void __iomem *ioport_map(unsigned long port, unsigned int len); +extern void ioport_unmap(void __iomem *addr); + +#else + /* - * The Tile architecture does not support IOPORT, even with PCI. + * The TilePro architecture does not support IOPORT, even with PCI. * Unfortunately we can't yet simply not declare these methods, * since some generic code that compiles into the kernel, but * we never run, uses them unconditionally. @@ -198,7 +391,12 @@ static inline void memcpy_toio(volatile void __iomem *dst, const void *src, static inline long ioport_panic(void) { +#ifdef __tilegx__ + panic("PCI IO space support is disabled. Configure the kernel with" + " CONFIG_TILE_PCI_IO to enable it"); +#else panic("inb/outb and friends do not exist on tile"); +#endif return 0; } @@ -243,13 +441,6 @@ static inline void outl(u32 b, unsigned long addr) ioport_panic(); } -#define inb_p(addr) inb(addr) -#define inw_p(addr) inw(addr) -#define inl_p(addr) inl(addr) -#define outb_p(x, addr) outb((x), (addr)) -#define outw_p(x, addr) outw((x), (addr)) -#define outl_p(x, addr) outl((x), (addr)) - static inline void insb(unsigned long addr, void *buffer, int count) { ioport_panic(); @@ -280,6 +471,15 @@ static inline void outsl(unsigned long addr, const void *buffer, int count) ioport_panic(); } +#endif /* CHIP_HAS_MMIO() && defined(CONFIG_TILE_PCI_IO) */ + +#define inb_p(addr) inb(addr) +#define inw_p(addr) inw(addr) +#define inl_p(addr) inl(addr) +#define outb_p(x, addr) outb((x), (addr)) +#define outw_p(x, addr) outw((x), (addr)) +#define outl_p(x, addr) outl((x), (addr)) + #define ioread16be(addr) be16_to_cpu(ioread16(addr)) #define ioread32be(addr) be32_to_cpu(ioread32(addr)) #define iowrite16be(v, addr) iowrite16(be16_to_cpu(v), (addr)) diff --git a/arch/tile/include/asm/irq.h b/arch/tile/include/asm/irq.h index f80f8ceabc6..1fe86911838 100644 --- a/arch/tile/include/asm/irq.h +++ b/arch/tile/include/asm/irq.h @@ -18,10 +18,12 @@ #include <linux/hardirq.h> /* The hypervisor interface provides 32 IRQs. */ -#define NR_IRQS 32 +#define NR_IRQS 32 /* IRQ numbers used for linux IPIs. */ -#define IRQ_RESCHEDULE 1 +#define IRQ_RESCHEDULE 0 +/* Interrupts for dynamic allocation start at 1. Let the core allocate irq0 */ +#define NR_IRQS_LEGACY 1 #define irq_canonicalize(irq) (irq) diff --git a/arch/tile/include/asm/irqflags.h b/arch/tile/include/asm/irqflags.h index 5db0ce54284..71af5747874 100644 --- a/arch/tile/include/asm/irqflags.h +++ b/arch/tile/include/asm/irqflags.h @@ -18,32 +18,20 @@ #include <arch/interrupts.h> #include <arch/chip.h> -#if !defined(__tilegx__) && defined(__ASSEMBLY__) - /* * The set of interrupts we want to allow when interrupts are nominally * disabled. The remainder are effectively "NMI" interrupts from * the point of view of the generic Linux code. Note that synchronous * interrupts (aka "non-queued") are not blocked by the mask in any case. */ -#if CHIP_HAS_AUX_PERF_COUNTERS() -#define LINUX_MASKABLE_INTERRUPTS_HI \ - (~(INT_MASK_HI(INT_PERF_COUNT) | INT_MASK_HI(INT_AUX_PERF_COUNT))) -#else -#define LINUX_MASKABLE_INTERRUPTS_HI \ - (~(INT_MASK_HI(INT_PERF_COUNT))) -#endif - -#else - -#if CHIP_HAS_AUX_PERF_COUNTERS() -#define LINUX_MASKABLE_INTERRUPTS \ - (~(INT_MASK(INT_PERF_COUNT) | INT_MASK(INT_AUX_PERF_COUNT))) -#else #define LINUX_MASKABLE_INTERRUPTS \ - (~(INT_MASK(INT_PERF_COUNT))) -#endif + (~((_AC(1,ULL) << INT_PERF_COUNT) | (_AC(1,ULL) << INT_AUX_PERF_COUNT))) +#if CHIP_HAS_SPLIT_INTR_MASK() +/* The same macro, but for the two 32-bit SPRs separately. */ +#define LINUX_MASKABLE_INTERRUPTS_LO (-1) +#define LINUX_MASKABLE_INTERRUPTS_HI \ + (~((1 << (INT_PERF_COUNT - 32)) | (1 << (INT_AUX_PERF_COUNT - 32)))) #endif #ifndef __ASSEMBLY__ @@ -52,7 +40,15 @@ #include <asm/percpu.h> #include <arch/spr_def.h> -/* Set and clear kernel interrupt masks. */ +/* + * Set and clear kernel interrupt masks. + * + * NOTE: __insn_mtspr() is a compiler builtin marked as a memory + * clobber. We rely on it being equivalent to a compiler barrier in + * this code since arch_local_irq_save() and friends must act as + * compiler barriers. This compiler semantic is baked into enough + * places that the compiler will maintain it going forward. + */ #if CHIP_HAS_SPLIT_INTR_MASK() #if INT_PERF_COUNT < 32 || INT_AUX_PERF_COUNT < 32 || INT_MEM_ERROR >= 32 # error Fix assumptions about which word various interrupts are in @@ -90,6 +86,14 @@ __insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_0, (unsigned long)(__m)); \ __insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_1, (unsigned long)(__m>>32)); \ } while (0) +#define interrupt_mask_save_mask() \ + (__insn_mfspr(SPR_INTERRUPT_MASK_SET_K_0) | \ + (((unsigned long long)__insn_mfspr(SPR_INTERRUPT_MASK_SET_K_1))<<32)) +#define interrupt_mask_restore_mask(mask) do { \ + unsigned long long __m = (mask); \ + __insn_mtspr(SPR_INTERRUPT_MASK_K_0, (unsigned long)(__m)); \ + __insn_mtspr(SPR_INTERRUPT_MASK_K_1, (unsigned long)(__m>>32)); \ +} while (0) #else #define interrupt_mask_set(n) \ __insn_mtspr(SPR_INTERRUPT_MASK_SET_K, (1UL << (n))) @@ -101,6 +105,10 @@ __insn_mtspr(SPR_INTERRUPT_MASK_SET_K, (mask)) #define interrupt_mask_reset_mask(mask) \ __insn_mtspr(SPR_INTERRUPT_MASK_RESET_K, (mask)) +#define interrupt_mask_save_mask() \ + __insn_mfspr(SPR_INTERRUPT_MASK_K) +#define interrupt_mask_restore_mask(mask) \ + __insn_mtspr(SPR_INTERRUPT_MASK_K, (mask)) #endif /* @@ -114,7 +122,13 @@ * to know our current state. */ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); -#define INITIAL_INTERRUPTS_ENABLED INT_MASK(INT_MEM_ERROR) +#define INITIAL_INTERRUPTS_ENABLED (1ULL << INT_MEM_ERROR) + +#ifdef CONFIG_DEBUG_PREEMPT +/* Due to inclusion issues, we can't rely on <linux/smp.h> here. */ +extern unsigned int debug_smp_processor_id(void); +# define smp_processor_id() debug_smp_processor_id() +#endif /* Disable interrupts. */ #define arch_local_irq_disable() \ @@ -122,11 +136,20 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); /* Disable all interrupts, including NMIs. */ #define arch_local_irq_disable_all() \ - interrupt_mask_set_mask(-1UL) + interrupt_mask_set_mask(-1ULL) + +/* + * Read the set of maskable interrupts. + * We avoid the preemption warning here via __this_cpu_ptr since even + * if irqs are already enabled, it's harmless to read the wrong cpu's + * enabled mask. + */ +#define arch_local_irqs_enabled() \ + (*__this_cpu_ptr(&interrupts_enabled_mask)) /* Re-enable all maskable interrupts. */ #define arch_local_irq_enable() \ - interrupt_mask_reset_mask(__get_cpu_var(interrupts_enabled_mask)) + interrupt_mask_reset_mask(arch_local_irqs_enabled()) /* Disable or enable interrupts based on flag argument. */ #define arch_local_irq_restore(disabled) do { \ @@ -153,7 +176,7 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); /* Prevent the given interrupt from being enabled next time we enable irqs. */ #define arch_local_irq_mask(interrupt) \ - (__get_cpu_var(interrupts_enabled_mask) &= ~INT_MASK(interrupt)) + this_cpu_and(interrupts_enabled_mask, ~(1ULL << (interrupt))) /* Prevent the given interrupt from being enabled immediately. */ #define arch_local_irq_mask_now(interrupt) do { \ @@ -163,7 +186,7 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); /* Allow the given interrupt to be enabled next time we enable irqs. */ #define arch_local_irq_unmask(interrupt) \ - (__get_cpu_var(interrupts_enabled_mask) |= INT_MASK(interrupt)) + this_cpu_or(interrupts_enabled_mask, (1ULL << (interrupt))) /* Allow the given interrupt to be enabled immediately, if !irqs_disabled. */ #define arch_local_irq_unmask_now(interrupt) do { \ @@ -179,7 +202,7 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); #ifdef __tilegx__ #if INT_MEM_ERROR != 0 -# error Fix IRQ_DISABLED() macro +# error Fix IRQS_DISABLED() macro #endif /* Return 0 or 1 to indicate whether interrupts are currently disabled. */ @@ -207,9 +230,10 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); mtspr SPR_INTERRUPT_MASK_SET_K, tmp /* Enable interrupts. */ -#define IRQ_ENABLE(tmp0, tmp1) \ +#define IRQ_ENABLE_LOAD(tmp0, tmp1) \ GET_INTERRUPTS_ENABLED_MASK_PTR(tmp0); \ - ld tmp0, tmp0; \ + ld tmp0, tmp0 +#define IRQ_ENABLE_APPLY(tmp0, tmp1) \ mtspr SPR_INTERRUPT_MASK_RESET_K, tmp0 #else /* !__tilegx__ */ @@ -237,7 +261,7 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); /* Disable interrupts. */ #define IRQ_DISABLE(tmp0, tmp1) \ { \ - movei tmp0, -1; \ + movei tmp0, LINUX_MASKABLE_INTERRUPTS_LO; \ moveli tmp1, lo16(LINUX_MASKABLE_INTERRUPTS_HI) \ }; \ { \ @@ -253,17 +277,22 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask); mtspr SPR_INTERRUPT_MASK_SET_K_1, tmp /* Enable interrupts. */ -#define IRQ_ENABLE(tmp0, tmp1) \ +#define IRQ_ENABLE_LOAD(tmp0, tmp1) \ GET_INTERRUPTS_ENABLED_MASK_PTR(tmp0); \ { \ lw tmp0, tmp0; \ addi tmp1, tmp0, 4 \ }; \ - lw tmp1, tmp1; \ + lw tmp1, tmp1 +#define IRQ_ENABLE_APPLY(tmp0, tmp1) \ mtspr SPR_INTERRUPT_MASK_RESET_K_0, tmp0; \ mtspr SPR_INTERRUPT_MASK_RESET_K_1, tmp1 #endif +#define IRQ_ENABLE(tmp0, tmp1) \ + IRQ_ENABLE_LOAD(tmp0, tmp1); \ + IRQ_ENABLE_APPLY(tmp0, tmp1) + /* * Do the CPU's IRQ-state tracing from assembly code. We call a * C function, but almost everywhere we do, we don't mind clobbering diff --git a/arch/tile/include/asm/hw_irq.h b/arch/tile/include/asm/kdebug.h index 4fac5fbf333..5bbbfa904c2 100644 --- a/arch/tile/include/asm/hw_irq.h +++ b/arch/tile/include/asm/kdebug.h @@ -1,5 +1,5 @@ /* - * Copyright 2010 Tilera Corporation. All Rights Reserved. + * Copyright 2012 Tilera Corporation. All Rights Reserved. * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License @@ -12,7 +12,17 @@ * more details. */ -#ifndef _ASM_TILE_HW_IRQ_H -#define _ASM_TILE_HW_IRQ_H +#ifndef _ASM_TILE_KDEBUG_H +#define _ASM_TILE_KDEBUG_H -#endif /* _ASM_TILE_HW_IRQ_H */ +#include <linux/notifier.h> + +enum die_val { + DIE_OOPS = 1, + DIE_BREAK, + DIE_SSTEPBP, + DIE_PAGE_FAULT, + DIE_COMPILED_BPT +}; + +#endif /* _ASM_TILE_KDEBUG_H */ diff --git a/arch/tile/include/asm/kexec.h b/arch/tile/include/asm/kexec.h index c11a6cc73bb..fc98ccfc98a 100644 --- a/arch/tile/include/asm/kexec.h +++ b/arch/tile/include/asm/kexec.h @@ -19,12 +19,24 @@ #include <asm/page.h> +#ifndef __tilegx__ /* Maximum physical address we can use pages from. */ #define KEXEC_SOURCE_MEMORY_LIMIT TASK_SIZE /* Maximum address we can reach in physical address mode. */ #define KEXEC_DESTINATION_MEMORY_LIMIT TASK_SIZE /* Maximum address we can use for the control code buffer. */ #define KEXEC_CONTROL_MEMORY_LIMIT TASK_SIZE +#else +/* We need to limit the memory below PGDIR_SIZE since + * we only setup page table for [0, PGDIR_SIZE) before final kexec. + */ +/* Maximum physical address we can use pages from. */ +#define KEXEC_SOURCE_MEMORY_LIMIT PGDIR_SIZE +/* Maximum address we can reach in physical address mode. */ +#define KEXEC_DESTINATION_MEMORY_LIMIT PGDIR_SIZE +/* Maximum address we can use for the control code buffer. */ +#define KEXEC_CONTROL_MEMORY_LIMIT PGDIR_SIZE +#endif #define KEXEC_CONTROL_PAGE_SIZE PAGE_SIZE diff --git a/arch/tile/include/asm/kgdb.h b/arch/tile/include/asm/kgdb.h new file mode 100644 index 00000000000..280c181cf0d --- /dev/null +++ b/arch/tile/include/asm/kgdb.h @@ -0,0 +1,71 @@ +/* + * Copyright 2013 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + * + * TILE-Gx KGDB support. + */ + +#ifndef __TILE_KGDB_H__ +#define __TILE_KGDB_H__ + +#include <linux/kdebug.h> +#include <arch/opcode.h> + +#define GDB_SIZEOF_REG sizeof(unsigned long) + +/* + * TILE-Gx gdb is expecting the following register layout: + * 56 GPRs(R0 - R52, TP, SP, LR), 8 special GPRs(networks and ZERO), + * plus the PC and the faultnum. + * + * Even though kernel not use the 8 special GPRs, they need to be present + * in the registers sent for correct processing in the host-side gdb. + * + */ +#define DBG_MAX_REG_NUM (56+8+2) +#define NUMREGBYTES (DBG_MAX_REG_NUM * GDB_SIZEOF_REG) + +/* + * BUFMAX defines the maximum number of characters in inbound/outbound + * buffers at least NUMREGBYTES*2 are needed for register packets, + * Longer buffer is needed to list all threads. + */ +#define BUFMAX 2048 + +#define BREAK_INSTR_SIZE TILEGX_BUNDLE_SIZE_IN_BYTES + +/* + * Require cache flush for set/clear a software breakpoint or write memory. + */ +#define CACHE_FLUSH_IS_SAFE 1 + +/* + * The compiled-in breakpoint instruction can be used to "break" into + * the debugger via magic system request key (sysrq-G). + */ +static tile_bundle_bits compiled_bpt = TILEGX_BPT_BUNDLE | DIE_COMPILED_BPT; + +enum tilegx_regnum { + TILEGX_PC_REGNUM = TREG_LAST_GPR + 9, + TILEGX_FAULTNUM_REGNUM, +}; + +/* + * Generate a breakpoint exception to "break" into the debugger. + */ +static inline void arch_kgdb_breakpoint(void) +{ + asm volatile (".quad %0\n\t" + ::""(compiled_bpt)); +} + +#endif /* __TILE_KGDB_H__ */ diff --git a/arch/tile/include/asm/kmap_types.h b/arch/tile/include/asm/kmap_types.h index 3d0f2024626..92b28e3e997 100644 --- a/arch/tile/include/asm/kmap_types.h +++ b/arch/tile/include/asm/kmap_types.h @@ -23,35 +23,6 @@ * adds 4MB of required address-space. For now we leave KM_TYPE_NR * set to depth 8. */ -enum km_type { - KM_TYPE_NR = 8 -}; - -/* - * We provide dummy definitions of all the stray values that used to be - * required for kmap_atomic() and no longer are. - */ -enum { - KM_BOUNCE_READ, - KM_SKB_SUNRPC_DATA, - KM_SKB_DATA_SOFTIRQ, - KM_USER0, - KM_USER1, - KM_BIO_SRC_IRQ, - KM_BIO_DST_IRQ, - KM_PTE0, - KM_PTE1, - KM_IRQ0, - KM_IRQ1, - KM_SOFTIRQ0, - KM_SOFTIRQ1, - KM_SYNC_ICACHE, - KM_SYNC_DCACHE, - KM_UML_USERCOPY, - KM_IRQ_PTE, - KM_NMI, - KM_NMI_PTE, - KM_KDB -}; +#define KM_TYPE_NR 8 #endif /* _ASM_TILE_KMAP_TYPES_H */ diff --git a/arch/tile/include/asm/kprobes.h b/arch/tile/include/asm/kprobes.h new file mode 100644 index 00000000000..d8f9a83943b --- /dev/null +++ b/arch/tile/include/asm/kprobes.h @@ -0,0 +1,79 @@ +/* + * arch/tile/include/asm/kprobes.h + * + * Copyright 2012 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + */ + +#ifndef _ASM_TILE_KPROBES_H +#define _ASM_TILE_KPROBES_H + +#include <linux/types.h> +#include <linux/ptrace.h> +#include <linux/percpu.h> + +#include <arch/opcode.h> + +#define __ARCH_WANT_KPROBES_INSN_SLOT +#define MAX_INSN_SIZE 2 + +#define kretprobe_blacklist_size 0 + +typedef tile_bundle_bits kprobe_opcode_t; + +#define flush_insn_slot(p) \ + flush_icache_range((unsigned long)p->addr, \ + (unsigned long)p->addr + \ + (MAX_INSN_SIZE * sizeof(kprobe_opcode_t))) + +struct kprobe; + +/* Architecture specific copy of original instruction. */ +struct arch_specific_insn { + kprobe_opcode_t *insn; +}; + +struct prev_kprobe { + struct kprobe *kp; + unsigned long status; + unsigned long saved_pc; +}; + +#define MAX_JPROBES_STACK_SIZE 128 +#define MAX_JPROBES_STACK_ADDR \ + (((unsigned long)current_thread_info()) + THREAD_SIZE - 32 \ + - sizeof(struct pt_regs)) + +#define MIN_JPROBES_STACK_SIZE(ADDR) \ + ((((ADDR) + MAX_JPROBES_STACK_SIZE) > MAX_JPROBES_STACK_ADDR) \ + ? MAX_JPROBES_STACK_ADDR - (ADDR) \ + : MAX_JPROBES_STACK_SIZE) + +/* per-cpu kprobe control block. */ +struct kprobe_ctlblk { + unsigned long kprobe_status; + unsigned long kprobe_saved_pc; + unsigned long jprobe_saved_sp; + struct prev_kprobe prev_kprobe; + struct pt_regs jprobe_saved_regs; + char jprobes_stack[MAX_JPROBES_STACK_SIZE]; +}; + +extern tile_bundle_bits breakpoint2_insn; +extern tile_bundle_bits breakpoint_insn; + +void arch_remove_kprobe(struct kprobe *); + +extern int kprobe_exceptions_notify(struct notifier_block *self, + unsigned long val, void *data); + +#endif /* _ASM_TILE_KPROBES_H */ diff --git a/arch/tile/include/asm/memprof.h b/arch/tile/include/asm/memprof.h deleted file mode 100644 index 359949be28c..00000000000 --- a/arch/tile/include/asm/memprof.h +++ /dev/null @@ -1,33 +0,0 @@ -/* - * Copyright 2010 Tilera Corporation. All Rights Reserved. - * - * This program 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, version 2. - * - * This program 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, GOOD TITLE or - * NON INFRINGEMENT. See the GNU General Public License for - * more details. - * - * The hypervisor's memory controller profiling infrastructure allows - * the programmer to find out what fraction of the available memory - * bandwidth is being consumed at each memory controller. The - * profiler provides start, stop, and clear operations to allows - * profiling over a specific time window, as well as an interface for - * reading the most recent profile values. - * - * This header declares IOCTL codes necessary to control memprof. - */ -#ifndef _ASM_TILE_MEMPROF_H -#define _ASM_TILE_MEMPROF_H - -#include <linux/ioctl.h> - -#define MEMPROF_IOCTL_TYPE 0xB4 -#define MEMPROF_IOCTL_START _IO(MEMPROF_IOCTL_TYPE, 0) -#define MEMPROF_IOCTL_STOP _IO(MEMPROF_IOCTL_TYPE, 1) -#define MEMPROF_IOCTL_CLEAR _IO(MEMPROF_IOCTL_TYPE, 2) - -#endif /* _ASM_TILE_MEMPROF_H */ diff --git a/arch/tile/include/asm/mman.h b/arch/tile/include/asm/mman.h deleted file mode 100644 index 81b8fc348d6..00000000000 --- a/arch/tile/include/asm/mman.h +++ /dev/null @@ -1,41 +0,0 @@ -/* - * Copyright 2010 Tilera Corporation. All Rights Reserved. - * - * This program 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, version 2. - * - * This program 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, GOOD TITLE or - * NON INFRINGEMENT. See the GNU General Public License for - * more details. - */ - -#ifndef _ASM_TILE_MMAN_H -#define _ASM_TILE_MMAN_H - -#include <asm-generic/mman-common.h> -#include <arch/chip.h> - -/* Standard Linux flags */ - -#define MAP_POPULATE 0x0040 /* populate (prefault) pagetables */ -#define MAP_NONBLOCK 0x0080 /* do not block on IO */ -#define MAP_GROWSDOWN 0x0100 /* stack-like segment */ -#define MAP_STACK MAP_GROWSDOWN /* provide convenience alias */ -#define MAP_LOCKED 0x0200 /* pages are locked */ -#define MAP_NORESERVE 0x0400 /* don't check for reservations */ -#define MAP_DENYWRITE 0x0800 /* ETXTBSY */ -#define MAP_EXECUTABLE 0x1000 /* mark it as an executable */ -#define MAP_HUGETLB 0x4000 /* create a huge page mapping */ - - -/* - * Flags for mlockall - */ -#define MCL_CURRENT 1 /* lock all current mappings */ -#define MCL_FUTURE 2 /* lock all future mappings */ - - -#endif /* _ASM_TILE_MMAN_H */ diff --git a/arch/tile/include/asm/mmu.h b/arch/tile/include/asm/mmu.h index 92f94c77b6e..0cab1182bde 100644 --- a/arch/tile/include/asm/mmu.h +++ b/arch/tile/include/asm/mmu.h @@ -21,7 +21,8 @@ struct mm_context { * Written under the mmap_sem semaphore; read without the * semaphore but atomically, but it is conservatively set. */ - unsigned int priority_cached; + unsigned long priority_cached; + unsigned long vdso_base; }; typedef struct mm_context mm_context_t; diff --git a/arch/tile/include/asm/mmu_context.h b/arch/tile/include/asm/mmu_context.h index 15fb2464112..4734215e2ad 100644 --- a/arch/tile/include/asm/mmu_context.h +++ b/arch/tile/include/asm/mmu_context.h @@ -30,18 +30,22 @@ init_new_context(struct task_struct *tsk, struct mm_struct *mm) return 0; } -/* Note that arch/tile/kernel/head.S also calls hv_install_context() */ +/* + * Note that arch/tile/kernel/head_NN.S and arch/tile/mm/migrate_NN.S + * also call hv_install_context(). + */ static inline void __install_page_table(pgd_t *pgdir, int asid, pgprot_t prot) { /* FIXME: DIRECTIO should not always be set. FIXME. */ - int rc = hv_install_context(__pa(pgdir), prot, asid, HV_CTX_DIRECTIO); + int rc = hv_install_context(__pa(pgdir), prot, asid, + HV_CTX_DIRECTIO | CTX_PAGE_FLAG); if (rc < 0) panic("hv_install_context failed: %d", rc); } static inline void install_page_table(pgd_t *pgdir, int asid) { - pte_t *ptep = virt_to_pte(NULL, (unsigned long)pgdir); + pte_t *ptep = virt_to_kpte((unsigned long)pgdir); __install_page_table(pgdir, asid, *ptep); } diff --git a/arch/tile/include/asm/mmzone.h b/arch/tile/include/asm/mmzone.h index 9d3dbce8f95..804f1098b6c 100644 --- a/arch/tile/include/asm/mmzone.h +++ b/arch/tile/include/asm/mmzone.h @@ -42,7 +42,7 @@ static inline int pfn_to_nid(unsigned long pfn) #define kern_addr_valid(kaddr) virt_addr_valid((void *)kaddr) -static inline int pfn_valid(int pfn) +static inline int pfn_valid(unsigned long pfn) { int nid = pfn_to_nid(pfn); diff --git a/arch/tile/include/asm/module.h b/arch/tile/include/asm/module.h new file mode 100644 index 00000000000..44ed07ccd3d --- /dev/null +++ b/arch/tile/include/asm/module.h @@ -0,0 +1,40 @@ +/* + * Copyright 2011 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + */ + +#ifndef _ASM_TILE_MODULE_H +#define _ASM_TILE_MODULE_H + +#include <arch/chip.h> + +#include <asm-generic/module.h> + +/* We can't use modules built with different page sizes. */ +#if defined(CONFIG_PAGE_SIZE_16KB) +# define MODULE_PGSZ " 16KB" +#elif defined(CONFIG_PAGE_SIZE_64KB) +# define MODULE_PGSZ " 64KB" +#else +# define MODULE_PGSZ "" +#endif + +/* We don't really support no-SMP so tag if someone tries. */ +#ifdef CONFIG_SMP +#define MODULE_NOSMP "" +#else +#define MODULE_NOSMP " nosmp" +#endif + +#define MODULE_ARCH_VERMAGIC CHIP_ARCH_NAME MODULE_PGSZ MODULE_NOSMP + +#endif /* _ASM_TILE_MODULE_H */ diff --git a/arch/tile/include/asm/page.h b/arch/tile/include/asm/page.h index db93518fac0..67276800861 100644 --- a/arch/tile/include/asm/page.h +++ b/arch/tile/include/asm/page.h @@ -20,8 +20,17 @@ #include <arch/chip.h> /* PAGE_SHIFT and HPAGE_SHIFT determine the page sizes. */ -#define PAGE_SHIFT HV_LOG2_PAGE_SIZE_SMALL -#define HPAGE_SHIFT HV_LOG2_PAGE_SIZE_LARGE +#if defined(CONFIG_PAGE_SIZE_16KB) +#define PAGE_SHIFT 14 +#define CTX_PAGE_FLAG HV_CTX_PG_SM_16K +#elif defined(CONFIG_PAGE_SIZE_64KB) +#define PAGE_SHIFT 16 +#define CTX_PAGE_FLAG HV_CTX_PG_SM_64K +#else +#define PAGE_SHIFT HV_LOG2_DEFAULT_PAGE_SIZE_SMALL +#define CTX_PAGE_FLAG 0 +#endif +#define HPAGE_SHIFT HV_LOG2_DEFAULT_PAGE_SIZE_LARGE #define PAGE_SIZE (_AC(1, UL) << PAGE_SHIFT) #define HPAGE_SIZE (_AC(1, UL) << HPAGE_SHIFT) @@ -30,6 +39,12 @@ #define HPAGE_MASK (~(HPAGE_SIZE - 1)) /* + * We do define AT_SYSINFO_EHDR to support vDSO, + * but don't use the gate mechanism. + */ +#define __HAVE_ARCH_GATE_AREA 1 + +/* * If the Kconfig doesn't specify, set a maximum zone order that * is enough so that we can create huge pages from small pages given * the respective sizes of the two page types. See <linux/mmzone.h>. @@ -78,8 +93,7 @@ typedef HV_PTE pgprot_t; /* * User L2 page tables are managed as one L2 page table per page, * because we use the page allocator for them. This keeps the allocation - * simple and makes it potentially useful to implement HIGHPTE at some point. - * However, it's also inefficient, since L2 page tables are much smaller + * simple, but it's also inefficient, since L2 page tables are much smaller * than pages (currently 2KB vs 64KB). So we should revisit this. */ typedef struct page *pgtable_t; @@ -128,14 +142,18 @@ static inline __attribute_const__ int get_order(unsigned long size) #define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT) -#define HUGE_MAX_HSTATE 2 +#define HUGE_MAX_HSTATE 6 #ifdef CONFIG_HUGETLB_PAGE #define HAVE_ARCH_HUGETLB_UNMAPPED_AREA #endif +/* Allow overriding how much VA or PA the kernel will use. */ +#define MAX_PA_WIDTH CHIP_PA_WIDTH() +#define MAX_VA_WIDTH CHIP_VA_WIDTH() + /* Each memory controller has PAs distinct in their high bits. */ -#define NR_PA_HIGHBIT_SHIFT (CHIP_PA_WIDTH() - CHIP_LOG_NUM_MSHIMS()) +#define NR_PA_HIGHBIT_SHIFT (MAX_PA_WIDTH - CHIP_LOG_NUM_MSHIMS()) #define NR_PA_HIGHBIT_VALUES (1 << CHIP_LOG_NUM_MSHIMS()) #define __pa_to_highbits(pa) ((phys_addr_t)(pa) >> NR_PA_HIGHBIT_SHIFT) #define __pfn_to_highbits(pfn) ((pfn) >> (NR_PA_HIGHBIT_SHIFT - PAGE_SHIFT)) @@ -146,7 +164,7 @@ static inline __attribute_const__ int get_order(unsigned long size) * We reserve the lower half of memory for user-space programs, and the * upper half for system code. We re-map all of physical memory in the * upper half, which takes a quarter of our VA space. Then we have - * the vmalloc regions. The supervisor code lives at 0xfffffff700000000, + * the vmalloc regions. The supervisor code lives at the highest address, * with the hypervisor above that. * * Loadable kernel modules are placed immediately after the static @@ -158,27 +176,18 @@ static inline __attribute_const__ int get_order(unsigned long size) * Similarly, for now we don't play any struct page mapping games. */ -#if CHIP_PA_WIDTH() + 2 > CHIP_VA_WIDTH() +#if MAX_PA_WIDTH + 2 > MAX_VA_WIDTH # error Too much PA to map with the VA available! #endif -#define HALF_VA_SPACE (_AC(1, UL) << (CHIP_VA_WIDTH() - 1)) - -#define MEM_LOW_END (HALF_VA_SPACE - 1) /* low half */ -#define MEM_HIGH_START (-HALF_VA_SPACE) /* high half */ -#define PAGE_OFFSET MEM_HIGH_START -#define _VMALLOC_START _AC(0xfffffff500000000, UL) /* 4 GB */ -#define HUGE_VMAP_BASE _AC(0xfffffff600000000, UL) /* 4 GB */ -#define MEM_SV_START _AC(0xfffffff700000000, UL) /* 256 MB */ -#define MEM_SV_INTRPT MEM_SV_START -#define MEM_MODULE_START _AC(0xfffffff710000000, UL) /* 256 MB */ -#define MEM_MODULE_END (MEM_MODULE_START + (256*1024*1024)) -#define MEM_HV_START _AC(0xfffffff800000000, UL) /* 32 GB */ - -/* Highest DTLB address we will use */ -#define KERNEL_HIGH_VADDR MEM_SV_START -/* Since we don't currently provide any fixmaps, we use an impossible VA. */ -#define FIXADDR_TOP MEM_HV_START +#define PAGE_OFFSET (-(_AC(1, UL) << (MAX_VA_WIDTH - 1))) +#define KERNEL_HIGH_VADDR _AC(0xfffffff800000000, UL) /* high 32GB */ +#define FIXADDR_BASE (KERNEL_HIGH_VADDR - 0x300000000) /* 4 GB */ +#define FIXADDR_TOP (KERNEL_HIGH_VADDR - 0x200000000) /* 4 GB */ +#define _VMALLOC_START FIXADDR_TOP +#define MEM_SV_START (KERNEL_HIGH_VADDR - 0x100000000) /* 256 MB */ +#define MEM_MODULE_START (MEM_SV_START + (256*1024*1024)) /* 256 MB */ +#define MEM_MODULE_END (MEM_MODULE_START + (256*1024*1024)) #else /* !__tilegx__ */ @@ -200,25 +209,18 @@ static inline __attribute_const__ int get_order(unsigned long size) * values, and after that, we show "typical" values, since the actual * addresses depend on kernel #defines. * - * MEM_HV_INTRPT 0xfe000000 - * MEM_SV_INTRPT (kernel code) 0xfd000000 + * MEM_HV_START 0xfe000000 + * MEM_SV_START (kernel code) 0xfd000000 * MEM_USER_INTRPT (user vector) 0xfc000000 - * FIX_KMAP_xxx 0xf8000000 (via NR_CPUS * KM_TYPE_NR) - * PKMAP_BASE 0xf7000000 (via LAST_PKMAP) - * HUGE_VMAP 0xf3000000 (via CONFIG_NR_HUGE_VMAPS) - * VMALLOC_START 0xf0000000 (via __VMALLOC_RESERVE) + * FIX_KMAP_xxx 0xfa000000 (via NR_CPUS * KM_TYPE_NR) + * PKMAP_BASE 0xf9000000 (via LAST_PKMAP) + * VMALLOC_START 0xf7000000 (via VMALLOC_RESERVE) * mapped LOWMEM 0xc0000000 */ #define MEM_USER_INTRPT _AC(0xfc000000, UL) -#if CONFIG_KERNEL_PL == 1 -#define MEM_SV_INTRPT _AC(0xfd000000, UL) -#define MEM_HV_INTRPT _AC(0xfe000000, UL) -#else -#define MEM_GUEST_INTRPT _AC(0xfd000000, UL) -#define MEM_SV_INTRPT _AC(0xfe000000, UL) -#define MEM_HV_INTRPT _AC(0xff000000, UL) -#endif +#define MEM_SV_START _AC(0xfd000000, UL) +#define MEM_HV_START _AC(0xfe000000, UL) #define INTRPT_SIZE 0x4000 @@ -239,7 +241,7 @@ static inline __attribute_const__ int get_order(unsigned long size) #endif /* __tilegx__ */ -#ifndef __ASSEMBLY__ +#if !defined(__ASSEMBLY__) && !defined(VDSO_BUILD) #ifdef CONFIG_HIGHMEM @@ -325,6 +327,7 @@ static inline int pfn_valid(unsigned long pfn) struct mm_struct; extern pte_t *virt_to_pte(struct mm_struct *mm, unsigned long addr); +extern pte_t *virt_to_kpte(unsigned long kaddr); #endif /* !__ASSEMBLY__ */ diff --git a/arch/tile/include/asm/pci.h b/arch/tile/include/asm/pci.h index 5d5a635530b..dfedd7ac729 100644 --- a/arch/tile/include/asm/pci.h +++ b/arch/tile/include/asm/pci.h @@ -15,9 +15,12 @@ #ifndef _ASM_TILE_PCI_H #define _ASM_TILE_PCI_H +#include <linux/dma-mapping.h> #include <linux/pci.h> #include <asm-generic/pci_iomap.h> +#ifndef __tilegx__ + /* * Structure of a PCI controller (host bridge) */ @@ -25,7 +28,6 @@ struct pci_controller { int index; /* PCI domain number */ struct pci_bus *root_bus; - int first_busno; int last_busno; int hv_cfg_fd[2]; /* config{0,1} fds for this PCIe controller */ @@ -41,20 +43,161 @@ struct pci_controller { }; /* + * This flag tells if the platform is TILEmpower that needs + * special configuration for the PLX switch chip. + */ +extern int tile_plx_gen1; + +static inline void pci_iounmap(struct pci_dev *dev, void __iomem *addr) {} + +#define TILE_NUM_PCIE 2 + +/* * The hypervisor maps the entirety of CPA-space as bus addresses, so * bus addresses are physical addresses. The networking and block * device layers use this boolean for bounce buffer decisions. */ #define PCI_DMA_BUS_IS_PHYS 1 -int __devinit tile_pci_init(void); -int __devinit pcibios_init(void); +/* generic pci stuff */ +#include <asm-generic/pci.h> -static inline void pci_iounmap(struct pci_dev *dev, void __iomem *addr) {} +#else -void __devinit pcibios_fixup_bus(struct pci_bus *bus); +#include <asm/page.h> +#include <gxio/trio.h> -#define TILE_NUM_PCIE 2 +/** + * We reserve the hugepage-size address range at the top of the 64-bit address + * space to serve as the PCI window, emulating the BAR0 space of an endpoint + * device. This window is used by the chip-to-chip applications running on + * the RC node. The reason for carving out this window is that Mem-Maps that + * back up this window will not overlap with those that map the real physical + * memory. + */ +#define PCIE_HOST_BAR0_SIZE HPAGE_SIZE +#define PCIE_HOST_BAR0_START HPAGE_MASK + +/** + * The first PAGE_SIZE of the above "BAR" window is mapped to the + * gxpci_host_regs structure. + */ +#define PCIE_HOST_REGS_SIZE PAGE_SIZE + +/* + * This is the PCI address where the Mem-Map interrupt regions start. + * We use the 2nd to the last huge page of the 64-bit address space. + * The last huge page is used for the rootcomplex "bar", for C2C purpose. + */ +#define MEM_MAP_INTR_REGIONS_BASE (HPAGE_MASK - HPAGE_SIZE) + +/* + * Each Mem-Map interrupt region occupies 4KB. + */ +#define MEM_MAP_INTR_REGION_SIZE (1 << TRIO_MAP_MEM_LIM__ADDR_SHIFT) + +/* + * Allocate the PCI BAR window right below 4GB. + */ +#define TILE_PCI_BAR_WINDOW_TOP (1ULL << 32) + +/* + * Allocate 1GB for the PCI BAR window. + */ +#define TILE_PCI_BAR_WINDOW_SIZE (1 << 30) + +/* + * This is the highest bus address targeting the host memory that + * can be generated by legacy PCI devices with 32-bit or less + * DMA capability, dictated by the BAR window size and location. + */ +#define TILE_PCI_MAX_DIRECT_DMA_ADDRESS \ + (TILE_PCI_BAR_WINDOW_TOP - TILE_PCI_BAR_WINDOW_SIZE - 1) + +/* + * We shift the PCI bus range for all the physical memory up by the whole PA + * range. The corresponding CPA of an incoming PCI request will be the PCI + * address minus TILE_PCI_MEM_MAP_BASE_OFFSET. This also implies + * that the 64-bit capable devices will be given DMA addresses as + * the CPA plus TILE_PCI_MEM_MAP_BASE_OFFSET. To support 32-bit + * devices, we create a separate map region that handles the low + * 4GB. + * + * This design lets us avoid the "PCI hole" problem where the host bridge + * won't pass DMA traffic with target addresses that happen to fall within the + * BAR space. This enables us to use all the physical memory for DMA, instead + * of wasting the same amount of physical memory as the BAR window size. + */ +#define TILE_PCI_MEM_MAP_BASE_OFFSET (1ULL << CHIP_PA_WIDTH()) + +/* + * Start of the PCI memory resource, which starts at the end of the + * maximum system physical RAM address. + */ +#define TILE_PCI_MEM_START (1ULL << CHIP_PA_WIDTH()) + +/* + * Structure of a PCI controller (host bridge) on Gx. + */ +struct pci_controller { + + /* Pointer back to the TRIO that this PCIe port is connected to. */ + gxio_trio_context_t *trio; + int mac; /* PCIe mac index on the TRIO shim */ + int trio_index; /* Index of TRIO shim that contains the MAC. */ + + int pio_mem_index; /* PIO region index for memory access */ + +#ifdef CONFIG_TILE_PCI_IO + int pio_io_index; /* PIO region index for I/O space access */ +#endif + + /* + * Mem-Map regions for all the memory controllers so that Linux can + * map all of its physical memory space to the PCI bus. + */ + int mem_maps[MAX_NUMNODES]; + + int index; /* PCI domain number */ + struct pci_bus *root_bus; + + /* PCI I/O space resource for this controller. */ + struct resource io_space; + char io_space_name[32]; + + /* PCI memory space resource for this controller. */ + struct resource mem_space; + char mem_space_name[32]; + + uint64_t mem_offset; /* cpu->bus memory mapping offset. */ + + int first_busno; + + struct pci_ops *ops; + + /* Table that maps the INTx numbers to Linux irq numbers. */ + int irq_intx_table[4]; +}; + +extern struct pci_controller pci_controllers[TILEGX_NUM_TRIO * TILEGX_TRIO_PCIES]; +extern gxio_trio_context_t trio_contexts[TILEGX_NUM_TRIO]; +extern int num_trio_shims; + +extern void pci_iounmap(struct pci_dev *dev, void __iomem *); + +/* + * The PCI address space does not equal the physical memory address + * space (we have an IOMMU). The IDE and SCSI device layers use this + * boolean for bounce buffer decisions. + */ +#define PCI_DMA_BUS_IS_PHYS 0 + +#endif /* __tilegx__ */ + +int __init tile_pci_init(void); +int __init pcibios_init(void); + +void pcibios_fixup_bus(struct pci_bus *bus); #define pci_domain_nr(bus) (((struct pci_controller *)(bus)->sysdata)->index) @@ -77,13 +220,8 @@ static inline int pcibios_assign_all_busses(void) } #define PCIBIOS_MIN_MEM 0 -#define PCIBIOS_MIN_IO 0 - -/* - * This flag tells if the platform is TILEmpower that needs - * special configuration for the PLX switch chip. - */ -extern int tile_plx_gen1; +/* Minimum PCI I/O address, starting at the page boundary. */ +#define PCIBIOS_MIN_IO PAGE_SIZE /* Use any cpu for PCI. */ #define cpumask_of_pcibus(bus) cpu_online_mask @@ -91,7 +229,4 @@ extern int tile_plx_gen1; /* implement the pci_ DMA API in terms of the generic device dma_ one */ #include <asm-generic/pci-dma-compat.h> -/* generic pci stuff */ -#include <asm-generic/pci.h> - #endif /* _ASM_TILE_PCI_H */ diff --git a/arch/tile/include/asm/percpu.h b/arch/tile/include/asm/percpu.h index 63294f5a8ef..4f7ae39fa20 100644 --- a/arch/tile/include/asm/percpu.h +++ b/arch/tile/include/asm/percpu.h @@ -15,9 +15,37 @@ #ifndef _ASM_TILE_PERCPU_H #define _ASM_TILE_PERCPU_H -register unsigned long __my_cpu_offset __asm__("tp"); -#define __my_cpu_offset __my_cpu_offset -#define set_my_cpu_offset(tp) (__my_cpu_offset = (tp)) +register unsigned long my_cpu_offset_reg asm("tp"); + +#ifdef CONFIG_PREEMPT +/* + * For full preemption, we can't just use the register variable + * directly, since we need barrier() to hazard against it, causing the + * compiler to reload anything computed from a previous "tp" value. + * But we also don't want to use volatile asm, since we'd like the + * compiler to be able to cache the value across multiple percpu reads. + * So we use a fake stack read as a hazard against barrier(). + * The 'U' constraint is like 'm' but disallows postincrement. + */ +static inline unsigned long __my_cpu_offset(void) +{ + unsigned long tp; + register unsigned long *sp asm("sp"); + asm("move %0, tp" : "=r" (tp) : "U" (*sp)); + return tp; +} +#define __my_cpu_offset __my_cpu_offset() +#else +/* + * We don't need to hazard against barrier() since "tp" doesn't ever + * change with PREEMPT_NONE, and with PREEMPT_VOLUNTARY it only + * changes at function call points, at which we are already re-reading + * the value of "tp" due to "my_cpu_offset_reg" being a global variable. + */ +#define __my_cpu_offset my_cpu_offset_reg +#endif + +#define set_my_cpu_offset(tp) (my_cpu_offset_reg = (tp)) #include <asm-generic/percpu.h> diff --git a/arch/tile/include/asm/auxvec.h b/arch/tile/include/asm/perf_event.h index 1d393edb064..59c5b164e5b 100644 --- a/arch/tile/include/asm/auxvec.h +++ b/arch/tile/include/asm/perf_event.h @@ -1,5 +1,5 @@ /* - * Copyright 2010 Tilera Corporation. All Rights Reserved. + * Copyright 2014 Tilera Corporation. All Rights Reserved. * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License @@ -12,9 +12,11 @@ * more details. */ -#ifndef _ASM_TILE_AUXVEC_H -#define _ASM_TILE_AUXVEC_H +#ifndef _ASM_TILE_PERF_EVENT_H +#define _ASM_TILE_PERF_EVENT_H -/* No extensions to auxvec */ +#include <linux/percpu.h> +DECLARE_PER_CPU(u64, perf_irqs); -#endif /* _ASM_TILE_AUXVEC_H */ +unsigned long handle_syscall_link_address(void); +#endif /* _ASM_TILE_PERF_EVENT_H */ diff --git a/arch/tile/include/asm/pgalloc.h b/arch/tile/include/asm/pgalloc.h index e919c0bdc22..1b902508b66 100644 --- a/arch/tile/include/asm/pgalloc.h +++ b/arch/tile/include/asm/pgalloc.h @@ -19,24 +19,24 @@ #include <linux/mm.h> #include <linux/mmzone.h> #include <asm/fixmap.h> +#include <asm/page.h> #include <hv/hypervisor.h> /* Bits for the size of the second-level page table. */ -#define L2_KERNEL_PGTABLE_SHIFT \ - (HV_LOG2_PAGE_SIZE_LARGE - HV_LOG2_PAGE_SIZE_SMALL + HV_LOG2_PTE_SIZE) +#define L2_KERNEL_PGTABLE_SHIFT _HV_LOG2_L2_SIZE(HPAGE_SHIFT, PAGE_SHIFT) + +/* How big is a kernel L2 page table? */ +#define L2_KERNEL_PGTABLE_SIZE (1UL << L2_KERNEL_PGTABLE_SHIFT) /* We currently allocate user L2 page tables by page (unlike kernel L2s). */ -#if L2_KERNEL_PGTABLE_SHIFT < HV_LOG2_PAGE_SIZE_SMALL -#define L2_USER_PGTABLE_SHIFT HV_LOG2_PAGE_SIZE_SMALL +#if L2_KERNEL_PGTABLE_SHIFT < PAGE_SHIFT +#define L2_USER_PGTABLE_SHIFT PAGE_SHIFT #else #define L2_USER_PGTABLE_SHIFT L2_KERNEL_PGTABLE_SHIFT #endif /* How many pages do we need, as an "order", for a user L2 page table? */ -#define L2_USER_PGTABLE_ORDER (L2_USER_PGTABLE_SHIFT - HV_LOG2_PAGE_SIZE_SMALL) - -/* How big is a kernel L2 page table? */ -#define L2_KERNEL_PGTABLE_SIZE (1 << L2_KERNEL_PGTABLE_SHIFT) +#define L2_USER_PGTABLE_ORDER (L2_USER_PGTABLE_SHIFT - PAGE_SHIFT) static inline void set_pmd(pmd_t *pmdp, pmd_t pmd) { @@ -50,14 +50,14 @@ static inline void set_pmd(pmd_t *pmdp, pmd_t pmd) static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd, pte_t *ptep) { - set_pmd(pmd, ptfn_pmd(__pa(ptep) >> HV_LOG2_PAGE_TABLE_ALIGN, + set_pmd(pmd, ptfn_pmd(HV_CPA_TO_PTFN(__pa(ptep)), __pgprot(_PAGE_PRESENT))); } static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd, pgtable_t page) { - set_pmd(pmd, ptfn_pmd(HV_PFN_TO_PTFN(page_to_pfn(page)), + set_pmd(pmd, ptfn_pmd(HV_CPA_TO_PTFN(PFN_PHYS(page_to_pfn(page))), __pgprot(_PAGE_PRESENT))); } @@ -68,8 +68,20 @@ static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd, extern pgd_t *pgd_alloc(struct mm_struct *mm); extern void pgd_free(struct mm_struct *mm, pgd_t *pgd); -extern pgtable_t pte_alloc_one(struct mm_struct *mm, unsigned long address); -extern void pte_free(struct mm_struct *mm, struct page *pte); +extern pgtable_t pgtable_alloc_one(struct mm_struct *mm, unsigned long address, + int order); +extern void pgtable_free(struct mm_struct *mm, struct page *pte, int order); + +static inline pgtable_t pte_alloc_one(struct mm_struct *mm, + unsigned long address) +{ + return pgtable_alloc_one(mm, address, L2_USER_PGTABLE_ORDER); +} + +static inline void pte_free(struct mm_struct *mm, struct page *pte) +{ + pgtable_free(mm, pte, L2_USER_PGTABLE_ORDER); +} #define pmd_pgtable(pmd) pmd_page(pmd) @@ -85,8 +97,13 @@ static inline void pte_free_kernel(struct mm_struct *mm, pte_t *pte) pte_free(mm, virt_to_page(pte)); } -extern void __pte_free_tlb(struct mmu_gather *tlb, struct page *pte, - unsigned long address); +extern void __pgtable_free_tlb(struct mmu_gather *tlb, struct page *pte, + unsigned long address, int order); +static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *pte, + unsigned long address) +{ + __pgtable_free_tlb(tlb, pte, address, L2_USER_PGTABLE_ORDER); +} #define check_pgt_cache() do { } while (0) @@ -104,19 +121,44 @@ void shatter_pmd(pmd_t *pmd); void shatter_huge_page(unsigned long addr); #ifdef __tilegx__ -/* We share a single page allocator for both L1 and L2 page tables. */ -#if HV_L1_SIZE != HV_L2_SIZE -# error Rework assumption that L1 and L2 page tables are same size. -#endif -#define L1_USER_PGTABLE_ORDER L2_USER_PGTABLE_ORDER + #define pud_populate(mm, pud, pmd) \ pmd_populate_kernel((mm), (pmd_t *)(pud), (pte_t *)(pmd)) -#define pmd_alloc_one(mm, addr) \ - ((pmd_t *)page_to_virt(pte_alloc_one((mm), (addr)))) -#define pmd_free(mm, pmdp) \ - pte_free((mm), virt_to_page(pmdp)) -#define __pmd_free_tlb(tlb, pmdp, address) \ - __pte_free_tlb((tlb), virt_to_page(pmdp), (address)) + +/* Bits for the size of the L1 (intermediate) page table. */ +#define L1_KERNEL_PGTABLE_SHIFT _HV_LOG2_L1_SIZE(HPAGE_SHIFT) + +/* How big is a kernel L2 page table? */ +#define L1_KERNEL_PGTABLE_SIZE (1UL << L1_KERNEL_PGTABLE_SHIFT) + +/* We currently allocate L1 page tables by page. */ +#if L1_KERNEL_PGTABLE_SHIFT < PAGE_SHIFT +#define L1_USER_PGTABLE_SHIFT PAGE_SHIFT +#else +#define L1_USER_PGTABLE_SHIFT L1_KERNEL_PGTABLE_SHIFT #endif +/* How many pages do we need, as an "order", for an L1 page table? */ +#define L1_USER_PGTABLE_ORDER (L1_USER_PGTABLE_SHIFT - PAGE_SHIFT) + +static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long address) +{ + struct page *p = pgtable_alloc_one(mm, address, L1_USER_PGTABLE_ORDER); + return (pmd_t *)page_to_virt(p); +} + +static inline void pmd_free(struct mm_struct *mm, pmd_t *pmdp) +{ + pgtable_free(mm, virt_to_page(pmdp), L1_USER_PGTABLE_ORDER); +} + +static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmdp, + unsigned long address) +{ + __pgtable_free_tlb(tlb, virt_to_page(pmdp), address, + L1_USER_PGTABLE_ORDER); +} + +#endif /* __tilegx__ */ + #endif /* _ASM_TILE_PGALLOC_H */ diff --git a/arch/tile/include/asm/pgtable.h b/arch/tile/include/asm/pgtable.h index 1a20b7ef8ea..33587f16c15 100644 --- a/arch/tile/include/asm/pgtable.h +++ b/arch/tile/include/asm/pgtable.h @@ -27,9 +27,10 @@ #include <linux/slab.h> #include <linux/list.h> #include <linux/spinlock.h> +#include <linux/pfn.h> #include <asm/processor.h> #include <asm/fixmap.h> -#include <asm/system.h> +#include <asm/page.h> struct mm_struct; struct vm_area_struct; @@ -70,6 +71,7 @@ extern void set_page_homes(void); #define _PAGE_PRESENT HV_PTE_PRESENT #define _PAGE_HUGE_PAGE HV_PTE_PAGE +#define _PAGE_SUPER_PAGE HV_PTE_SUPER #define _PAGE_READABLE HV_PTE_READABLE #define _PAGE_WRITABLE HV_PTE_WRITABLE #define _PAGE_EXECUTABLE HV_PTE_EXECUTABLE @@ -86,6 +88,7 @@ extern void set_page_homes(void); #define _PAGE_ALL (\ _PAGE_PRESENT | \ _PAGE_HUGE_PAGE | \ + _PAGE_SUPER_PAGE | \ _PAGE_READABLE | \ _PAGE_WRITABLE | \ _PAGE_EXECUTABLE | \ @@ -163,7 +166,7 @@ extern void set_page_homes(void); (pgprot_t) { ((oldprot).val & ~_PAGE_ALL) | (newprot).val } /* Just setting the PFN to zero suffices. */ -#define pte_pgprot(x) hv_pte_set_pfn((x), 0) +#define pte_pgprot(x) hv_pte_set_pa((x), 0) /* * For PTEs and PDEs, we must clear the Present bit first when @@ -188,6 +191,7 @@ static inline void __pte_clear(pte_t *ptep) * Undefined behaviour if not.. */ #define pte_present hv_pte_get_present +#define pte_mknotpresent hv_pte_clear_present #define pte_user hv_pte_get_user #define pte_read hv_pte_get_readable #define pte_dirty hv_pte_get_dirty @@ -195,6 +199,7 @@ static inline void __pte_clear(pte_t *ptep) #define pte_write hv_pte_get_writable #define pte_exec hv_pte_get_executable #define pte_huge hv_pte_get_page +#define pte_super hv_pte_get_super #define pte_rdprotect hv_pte_clear_readable #define pte_exprotect hv_pte_clear_executable #define pte_mkclean hv_pte_clear_dirty @@ -207,6 +212,7 @@ static inline void __pte_clear(pte_t *ptep) #define pte_mkyoung hv_pte_set_accessed #define pte_mkwrite hv_pte_set_writable #define pte_mkhuge hv_pte_set_page +#define pte_mksuper hv_pte_set_super #define pte_special(pte) 0 #define pte_mkspecial(pte) (pte) @@ -262,7 +268,7 @@ static inline int pte_none(pte_t pte) static inline unsigned long pte_pfn(pte_t pte) { - return hv_pte_get_pfn(pte); + return PFN_DOWN(hv_pte_get_pa(pte)); } /* Set or get the remote cache cpu in a pgprot with remote caching. */ @@ -271,7 +277,7 @@ extern int get_remote_cache_cpu(pgprot_t prot); static inline pte_t pfn_pte(unsigned long pfn, pgprot_t prot) { - return hv_pte_set_pfn(prot, pfn); + return hv_pte_set_pa(prot, PFN_PHYS(pfn)); } /* Support for priority mappings. */ @@ -313,7 +319,7 @@ extern void check_mm_caching(struct mm_struct *prev, struct mm_struct *next); */ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) { - return pfn_pte(hv_pte_get_pfn(pte), newprot); + return pfn_pte(pte_pfn(pte), newprot); } /* @@ -336,13 +342,8 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) */ #define pgd_offset_k(address) pgd_offset(&init_mm, address) -#if defined(CONFIG_HIGHPTE) -extern pte_t *pte_offset_map(pmd_t *, unsigned long address); -#define pte_unmap(pte) kunmap_atomic(pte) -#else #define pte_offset_map(dir, address) pte_offset_kernel(dir, address) #define pte_unmap(pte) do { } while (0) -#endif /* Clear a non-executable kernel PTE and flush it from the TLB. */ #define kpte_clear_flush(ptep, vaddr) \ @@ -361,9 +362,6 @@ do { \ #define kern_addr_valid(addr) (1) #endif /* CONFIG_FLATMEM */ -#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ - remap_pfn_range(vma, vaddr, pfn, size, prot) - extern void vmalloc_sync_all(void); #endif /* !__ASSEMBLY__ */ @@ -411,6 +409,46 @@ static inline unsigned long pmd_index(unsigned long address) return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1); } +#define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG +static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma, + unsigned long address, + pmd_t *pmdp) +{ + return ptep_test_and_clear_young(vma, address, pmdp_ptep(pmdp)); +} + +#define __HAVE_ARCH_PMDP_SET_WRPROTECT +static inline void pmdp_set_wrprotect(struct mm_struct *mm, + unsigned long address, pmd_t *pmdp) +{ + ptep_set_wrprotect(mm, address, pmdp_ptep(pmdp)); +} + + +#define __HAVE_ARCH_PMDP_GET_AND_CLEAR +static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm, + unsigned long address, + pmd_t *pmdp) +{ + return pte_pmd(ptep_get_and_clear(mm, address, pmdp_ptep(pmdp))); +} + +static inline void __set_pmd(pmd_t *pmdp, pmd_t pmdval) +{ + set_pte(pmdp_ptep(pmdp), pmd_pte(pmdval)); +} + +#define set_pmd_at(mm, addr, pmdp, pmdval) __set_pmd(pmdp, pmdval) + +/* Create a pmd from a PTFN. */ +static inline pmd_t ptfn_pmd(unsigned long ptfn, pgprot_t prot) +{ + return pte_pmd(hv_pte_set_ptfn(prot, ptfn)); +} + +/* Return the page-table frame number (ptfn) that a pmd_t points at. */ +#define pmd_ptfn(pmd) hv_pte_get_ptfn(pmd_pte(pmd)) + /* * A given kernel pmd_t maps to a specific virtual address (either a * kernel huge page or a kernel pte_t table). Since kernel pte_t @@ -431,7 +469,48 @@ static inline unsigned long pmd_page_vaddr(pmd_t pmd) * OK for pte_lockptr(), since we just end up with potentially one * lock being used for several pte_t arrays. */ -#define pmd_page(pmd) pfn_to_page(HV_PTFN_TO_PFN(pmd_ptfn(pmd))) +#define pmd_page(pmd) pfn_to_page(PFN_DOWN(HV_PTFN_TO_CPA(pmd_ptfn(pmd)))) + +static inline void pmd_clear(pmd_t *pmdp) +{ + __pte_clear(pmdp_ptep(pmdp)); +} + +#define pmd_mknotpresent(pmd) pte_pmd(pte_mknotpresent(pmd_pte(pmd))) +#define pmd_young(pmd) pte_young(pmd_pte(pmd)) +#define pmd_mkyoung(pmd) pte_pmd(pte_mkyoung(pmd_pte(pmd))) +#define pmd_mkold(pmd) pte_pmd(pte_mkold(pmd_pte(pmd))) +#define pmd_mkwrite(pmd) pte_pmd(pte_mkwrite(pmd_pte(pmd))) +#define pmd_write(pmd) pte_write(pmd_pte(pmd)) +#define pmd_wrprotect(pmd) pte_pmd(pte_wrprotect(pmd_pte(pmd))) +#define pmd_mkdirty(pmd) pte_pmd(pte_mkdirty(pmd_pte(pmd))) +#define pmd_huge_page(pmd) pte_huge(pmd_pte(pmd)) +#define pmd_mkhuge(pmd) pte_pmd(pte_mkhuge(pmd_pte(pmd))) +#define __HAVE_ARCH_PMD_WRITE + +#define pfn_pmd(pfn, pgprot) pte_pmd(pfn_pte((pfn), (pgprot))) +#define pmd_pfn(pmd) pte_pfn(pmd_pte(pmd)) +#define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot)) + +static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot) +{ + return pfn_pmd(pmd_pfn(pmd), newprot); +} + +#ifdef CONFIG_TRANSPARENT_HUGEPAGE +#define has_transparent_hugepage() 1 +#define pmd_trans_huge pmd_huge_page + +static inline pmd_t pmd_mksplitting(pmd_t pmd) +{ + return pte_pmd(hv_pte_set_client2(pmd_pte(pmd))); +} + +static inline int pmd_trans_splitting(pmd_t pmd) +{ + return hv_pte_get_client2(pmd_pte(pmd)); +} +#endif /* CONFIG_TRANSPARENT_HUGEPAGE */ /* * The pte page can be thought of an array like this: pte_t[PTRS_PER_PTE] @@ -449,17 +528,13 @@ static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address) return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address); } -static inline int pmd_huge_page(pmd_t pmd) -{ - return pmd_val(pmd) & _PAGE_HUGE_PAGE; -} - #include <asm-generic/pgtable.h> /* Support /proc/NN/pgtable API. */ struct seq_file; int arch_proc_pgtable_show(struct seq_file *m, struct mm_struct *mm, - unsigned long vaddr, pte_t *ptep, void **datap); + unsigned long vaddr, unsigned long pagesize, + pte_t *ptep, void **datap); #endif /* !__ASSEMBLY__ */ diff --git a/arch/tile/include/asm/pgtable_32.h b/arch/tile/include/asm/pgtable_32.h index 9f98529761f..d26a4227903 100644 --- a/arch/tile/include/asm/pgtable_32.h +++ b/arch/tile/include/asm/pgtable_32.h @@ -20,11 +20,12 @@ * The level-1 index is defined by the huge page size. A PGD is composed * of PTRS_PER_PGD pgd_t's and is the top level of the page table. */ -#define PGDIR_SHIFT HV_LOG2_PAGE_SIZE_LARGE -#define PGDIR_SIZE HV_PAGE_SIZE_LARGE +#define PGDIR_SHIFT HPAGE_SHIFT +#define PGDIR_SIZE HPAGE_SIZE #define PGDIR_MASK (~(PGDIR_SIZE-1)) -#define PTRS_PER_PGD (1 << (32 - PGDIR_SHIFT)) -#define SIZEOF_PGD (PTRS_PER_PGD * sizeof(pgd_t)) +#define PTRS_PER_PGD _HV_L1_ENTRIES(HPAGE_SHIFT) +#define PGD_INDEX(va) _HV_L1_INDEX(va, HPAGE_SHIFT) +#define SIZEOF_PGD _HV_L1_SIZE(HPAGE_SHIFT) /* * The level-2 index is defined by the difference between the huge @@ -33,8 +34,9 @@ * Note that the hypervisor docs use PTE for what we call pte_t, so * this nomenclature is somewhat confusing. */ -#define PTRS_PER_PTE (1 << (HV_LOG2_PAGE_SIZE_LARGE - HV_LOG2_PAGE_SIZE_SMALL)) -#define SIZEOF_PTE (PTRS_PER_PTE * sizeof(pte_t)) +#define PTRS_PER_PTE _HV_L2_ENTRIES(HPAGE_SHIFT, PAGE_SHIFT) +#define PTE_INDEX(va) _HV_L2_INDEX(va, HPAGE_SHIFT, PAGE_SHIFT) +#define SIZEOF_PTE _HV_L2_SIZE(HPAGE_SHIFT, PAGE_SHIFT) #ifndef __ASSEMBLY__ @@ -53,17 +55,9 @@ #define PKMAP_BASE ((FIXADDR_BOOT_START - PAGE_SIZE*LAST_PKMAP) & PGDIR_MASK) #ifdef CONFIG_HIGHMEM -# define __VMAPPING_END (PKMAP_BASE & ~(HPAGE_SIZE-1)) +# define _VMALLOC_END (PKMAP_BASE & ~(HPAGE_SIZE-1)) #else -# define __VMAPPING_END (FIXADDR_START & ~(HPAGE_SIZE-1)) -#endif - -#ifdef CONFIG_HUGEVMAP -#define HUGE_VMAP_END __VMAPPING_END -#define HUGE_VMAP_BASE (HUGE_VMAP_END - CONFIG_NR_HUGE_VMAPS * HPAGE_SIZE) -#define _VMALLOC_END HUGE_VMAP_BASE -#else -#define _VMALLOC_END __VMAPPING_END +# define _VMALLOC_END (FIXADDR_START & ~(HPAGE_SIZE-1)) #endif /* @@ -82,10 +76,12 @@ extern unsigned long VMALLOC_RESERVE /* = CONFIG_VMALLOC_RESERVE */; /* We have no pmd or pud since we are strictly a two-level page table */ #include <asm-generic/pgtable-nopmd.h> +static inline int pud_huge_page(pud_t pud) { return 0; } + /* We don't define any pgds for these addresses. */ static inline int pgd_addr_invalid(unsigned long addr) { - return addr >= MEM_HV_INTRPT; + return addr >= MEM_HV_START; } /* @@ -111,24 +107,14 @@ static inline pte_t ptep_get_and_clear(struct mm_struct *mm, return pte; } -static inline void __set_pmd(pmd_t *pmdp, pmd_t pmdval) -{ - set_pte(&pmdp->pud.pgd, pmdval.pud.pgd); -} - -/* Create a pmd from a PTFN. */ -static inline pmd_t ptfn_pmd(unsigned long ptfn, pgprot_t prot) -{ - return (pmd_t){ { hv_pte_set_ptfn(prot, ptfn) } }; -} - -/* Return the page-table frame number (ptfn) that a pmd_t points at. */ -#define pmd_ptfn(pmd) hv_pte_get_ptfn((pmd).pud.pgd) - -static inline void pmd_clear(pmd_t *pmdp) -{ - __pte_clear(&pmdp->pud.pgd); -} +/* + * pmds are wrappers around pgds, which are the same as ptes. + * It's often convenient to "cast" back and forth and use the pte methods, + * which are the methods supplied by the hypervisor. + */ +#define pmd_pte(pmd) ((pmd).pud.pgd) +#define pmdp_ptep(pmdp) (&(pmdp)->pud.pgd) +#define pte_pmd(pte) ((pmd_t){ { (pte) } }) #endif /* __ASSEMBLY__ */ diff --git a/arch/tile/include/asm/pgtable_64.h b/arch/tile/include/asm/pgtable_64.h index fd80328523b..2c8a9cd102d 100644 --- a/arch/tile/include/asm/pgtable_64.h +++ b/arch/tile/include/asm/pgtable_64.h @@ -21,17 +21,19 @@ #define PGDIR_SIZE HV_L1_SPAN #define PGDIR_MASK (~(PGDIR_SIZE-1)) #define PTRS_PER_PGD HV_L0_ENTRIES -#define SIZEOF_PGD (PTRS_PER_PGD * sizeof(pgd_t)) +#define PGD_INDEX(va) HV_L0_INDEX(va) +#define SIZEOF_PGD HV_L0_SIZE /* * The level-1 index is defined by the huge page size. A PMD is composed * of PTRS_PER_PMD pgd_t's and is the middle level of the page table. */ -#define PMD_SHIFT HV_LOG2_PAGE_SIZE_LARGE -#define PMD_SIZE HV_PAGE_SIZE_LARGE +#define PMD_SHIFT HPAGE_SHIFT +#define PMD_SIZE HPAGE_SIZE #define PMD_MASK (~(PMD_SIZE-1)) -#define PTRS_PER_PMD (1 << (PGDIR_SHIFT - PMD_SHIFT)) -#define SIZEOF_PMD (PTRS_PER_PMD * sizeof(pmd_t)) +#define PTRS_PER_PMD _HV_L1_ENTRIES(HPAGE_SHIFT) +#define PMD_INDEX(va) _HV_L1_INDEX(va, HPAGE_SHIFT) +#define SIZEOF_PMD _HV_L1_SIZE(HPAGE_SHIFT) /* * The level-2 index is defined by the difference between the huge @@ -40,25 +42,34 @@ * Note that the hypervisor docs use PTE for what we call pte_t, so * this nomenclature is somewhat confusing. */ -#define PTRS_PER_PTE (1 << (HV_LOG2_PAGE_SIZE_LARGE - HV_LOG2_PAGE_SIZE_SMALL)) -#define SIZEOF_PTE (PTRS_PER_PTE * sizeof(pte_t)) +#define PTRS_PER_PTE _HV_L2_ENTRIES(HPAGE_SHIFT, PAGE_SHIFT) +#define PTE_INDEX(va) _HV_L2_INDEX(va, HPAGE_SHIFT, PAGE_SHIFT) +#define SIZEOF_PTE _HV_L2_SIZE(HPAGE_SHIFT, PAGE_SHIFT) /* - * Align the vmalloc area to an L2 page table, and leave a guard page - * at the beginning and end. The vmalloc code also puts in an internal + * Align the vmalloc area to an L2 page table. Omit guard pages at + * the beginning and end for simplicity (particularly in the per-cpu + * memory allocation code). The vmalloc code puts in an internal * guard page between each allocation. */ -#define _VMALLOC_END HUGE_VMAP_BASE -#define VMALLOC_END (_VMALLOC_END - PAGE_SIZE) -#define VMALLOC_START (_VMALLOC_START + PAGE_SIZE) - -#define HUGE_VMAP_END (HUGE_VMAP_BASE + PGDIR_SIZE) +#define _VMALLOC_END MEM_SV_START +#define VMALLOC_END _VMALLOC_END +#define VMALLOC_START _VMALLOC_START #ifndef __ASSEMBLY__ /* We have no pud since we are a three-level page table. */ #include <asm-generic/pgtable-nopud.h> +/* + * pmds are the same as pgds and ptes, so converting is a no-op. + */ +#define pmd_pte(pmd) (pmd) +#define pmdp_ptep(pmdp) (pmdp) +#define pte_pmd(pte) (pte) + +#define pud_pte(pud) ((pud).pgd) + static inline int pud_none(pud_t pud) { return pud_val(pud) == 0; @@ -69,6 +80,11 @@ static inline int pud_present(pud_t pud) return pud_val(pud) & _PAGE_PRESENT; } +static inline int pud_huge_page(pud_t pud) +{ + return pud_val(pud) & _PAGE_HUGE_PAGE; +} + #define pmd_ERROR(e) \ pr_err("%s:%d: bad pmd 0x%016llx.\n", __FILE__, __LINE__, pmd_val(e)) @@ -85,6 +101,9 @@ static inline int pud_bad(pud_t pud) /* Return the page-table frame number (ptfn) that a pud_t points at. */ #define pud_ptfn(pud) hv_pte_get_ptfn((pud).pgd) +/* Return the page frame number (pfn) that a pud_t points at. */ +#define pud_pfn(pud) pte_pfn(pud_pte(pud)) + /* * A given kernel pud_t maps to a kernel pmd_t table at a specific * virtual address. Since kernel pmd_t tables can be aligned at @@ -98,7 +117,7 @@ static inline int pud_bad(pud_t pud) * A pud_t points to a pmd_t array. Since we can have multiple per * page, we don't have a one-to-one mapping of pud_t's to pages. */ -#define pud_page(pud) pfn_to_page(HV_PTFN_TO_PFN(pud_ptfn(pud))) +#define pud_page(pud) pfn_to_page(PFN_DOWN(HV_PTFN_TO_CPA(pud_ptfn(pud)))) static inline unsigned long pud_index(unsigned long address) { @@ -108,28 +127,6 @@ static inline unsigned long pud_index(unsigned long address) #define pmd_offset(pud, address) \ ((pmd_t *)pud_page_vaddr(*(pud)) + pmd_index(address)) -static inline void __set_pmd(pmd_t *pmdp, pmd_t pmdval) -{ - set_pte(pmdp, pmdval); -} - -/* Create a pmd from a PTFN and pgprot. */ -static inline pmd_t ptfn_pmd(unsigned long ptfn, pgprot_t prot) -{ - return hv_pte_set_ptfn(prot, ptfn); -} - -/* Return the page-table frame number (ptfn) that a pmd_t points at. */ -static inline unsigned long pmd_ptfn(pmd_t pmd) -{ - return hv_pte_get_ptfn(pmd); -} - -static inline void pmd_clear(pmd_t *pmdp) -{ - __pte_clear(pmdp); -} - /* Normalize an address to having the correct high bits set. */ #define pgd_addr_normalize pgd_addr_normalize static inline unsigned long pgd_addr_normalize(unsigned long addr) @@ -141,8 +138,7 @@ static inline unsigned long pgd_addr_normalize(unsigned long addr) /* We don't define any pgds for these addresses. */ static inline int pgd_addr_invalid(unsigned long addr) { - return addr >= MEM_HV_START || - (addr > MEM_LOW_END && addr < MEM_HIGH_START); + return addr >= KERNEL_HIGH_VADDR || addr != pgd_addr_normalize(addr); } /* diff --git a/arch/tile/include/asm/pmc.h b/arch/tile/include/asm/pmc.h new file mode 100644 index 00000000000..7ae3956d900 --- /dev/null +++ b/arch/tile/include/asm/pmc.h @@ -0,0 +1,64 @@ +/* + * Copyright 2014 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + */ + +#ifndef _ASM_TILE_PMC_H +#define _ASM_TILE_PMC_H + +#include <linux/ptrace.h> + +#define TILE_BASE_COUNTERS 2 + +/* Bitfields below are derived from SPR PERF_COUNT_CTL*/ +#ifndef __tilegx__ +/* PERF_COUNT_CTL on TILEPro */ +#define TILE_CTL_EXCL_USER (1 << 7) /* exclude user level */ +#define TILE_CTL_EXCL_KERNEL (1 << 8) /* exclude kernel level */ +#define TILE_CTL_EXCL_HV (1 << 9) /* exclude hypervisor level */ + +#define TILE_SEL_MASK 0x7f /* 7 bits for event SEL, + COUNT_0_SEL */ +#define TILE_PLM_MASK 0x780 /* 4 bits priv level msks, + COUNT_0_MASK*/ +#define TILE_EVENT_MASK (TILE_SEL_MASK | TILE_PLM_MASK) + +#else /* __tilegx__*/ +/* PERF_COUNT_CTL on TILEGx*/ +#define TILE_CTL_EXCL_USER (1 << 10) /* exclude user level */ +#define TILE_CTL_EXCL_KERNEL (1 << 11) /* exclude kernel level */ +#define TILE_CTL_EXCL_HV (1 << 12) /* exclude hypervisor level */ + +#define TILE_SEL_MASK 0x3f /* 6 bits for event SEL, + COUNT_0_SEL*/ +#define TILE_BOX_MASK 0x1c0 /* 3 bits box msks, + COUNT_0_BOX */ +#define TILE_PLM_MASK 0x3c00 /* 4 bits priv level msks, + COUNT_0_MASK */ +#define TILE_EVENT_MASK (TILE_SEL_MASK | TILE_BOX_MASK | TILE_PLM_MASK) +#endif /* __tilegx__*/ + +/* Takes register and fault number. Returns error to disable the interrupt. */ +typedef int (*perf_irq_t)(struct pt_regs *, int); + +int userspace_perf_handler(struct pt_regs *regs, int fault); + +perf_irq_t reserve_pmc_hardware(perf_irq_t new_perf_irq); +void release_pmc_hardware(void); + +unsigned long pmc_get_overflow(void); +void pmc_ack_overflow(unsigned long status); + +void unmask_pmc_interrupts(void); +void mask_pmc_interrupts(void); + +#endif /* _ASM_TILE_PMC_H */ diff --git a/arch/tile/include/asm/processor.h b/arch/tile/include/asm/processor.h index 34c1e01ffb5..42323636c45 100644 --- a/arch/tile/include/asm/processor.h +++ b/arch/tile/include/asm/processor.h @@ -15,6 +15,8 @@ #ifndef _ASM_TILE_PROCESSOR_H #define _ASM_TILE_PROCESSOR_H +#include <arch/chip.h> + #ifndef __ASSEMBLY__ /* @@ -25,7 +27,6 @@ #include <asm/ptrace.h> #include <asm/percpu.h> -#include <arch/chip.h> #include <arch/spr_def.h> struct task_struct; @@ -76,6 +77,17 @@ struct async_tlb { #ifdef CONFIG_HARDWALL struct hardwall_info; +struct hardwall_task { + /* Which hardwall is this task tied to? (or NULL if none) */ + struct hardwall_info *info; + /* Chains this task into the list at info->task_head. */ + struct list_head list; +}; +#ifdef __tilepro__ +#define HARDWALL_TYPES 1 /* udn */ +#else +#define HARDWALL_TYPES 3 /* udn, idn, and ipi */ +#endif #endif struct thread_struct { @@ -99,47 +111,38 @@ struct thread_struct { unsigned long long interrupt_mask; /* User interrupt-control 0 state */ unsigned long intctrl_0; -#if CHIP_HAS_PROC_STATUS_SPR() + /* Is this task currently doing a backtrace? */ + bool in_backtrace; /* Any other miscellaneous processor state bits */ unsigned long proc_status; -#endif #if !CHIP_HAS_FIXED_INTVEC_BASE() /* Interrupt base for PL0 interrupts */ unsigned long interrupt_vector_base; #endif -#if CHIP_HAS_TILE_RTF_HWM() /* Tile cache retry fifo high-water mark */ unsigned long tile_rtf_hwm; -#endif #if CHIP_HAS_DSTREAM_PF() /* Data stream prefetch control */ unsigned long dstream_pf; #endif #ifdef CONFIG_HARDWALL - /* Is this task tied to an activated hardwall? */ - struct hardwall_info *hardwall; - /* Chains this task into the list at hardwall->list. */ - struct list_head hardwall_list; + /* Hardwall information for various resources. */ + struct hardwall_task hardwall[HARDWALL_TYPES]; #endif #if CHIP_HAS_TILE_DMA() /* Async DMA TLB fault information */ struct async_tlb dma_async_tlb; #endif -#if CHIP_HAS_SN_PROC() - /* Was static network processor when we were switched out? */ - int sn_proc_running; - /* Async SNI TLB fault information */ - struct async_tlb sn_async_tlb; -#endif }; #endif /* !__ASSEMBLY__ */ /* * Start with "sp" this many bytes below the top of the kernel stack. - * This preserves the invariant that a called function may write to *sp. + * This allows us to be cache-aware when handling the initial save + * of the pt_regs value to the stack. */ -#define STACK_TOP_DELTA 8 +#define STACK_TOP_DELTA 64 /* * When entering the kernel via a fault, start with the top of the @@ -155,7 +158,7 @@ struct thread_struct { #ifndef __ASSEMBLY__ #ifdef __tilegx__ -#define TASK_SIZE_MAX (MEM_LOW_END + 1) +#define TASK_SIZE_MAX (_AC(1, UL) << (MAX_VA_WIDTH - 1)) #else #define TASK_SIZE_MAX PAGE_OFFSET #endif @@ -169,10 +172,10 @@ struct thread_struct { #define TASK_SIZE TASK_SIZE_MAX #endif -/* We provide a minimal "vdso" a la x86; just the sigreturn code for now. */ -#define VDSO_BASE (TASK_SIZE - PAGE_SIZE) +#define VDSO_BASE ((unsigned long)current->active_mm->context.vdso_base) +#define VDSO_SYM(x) (VDSO_BASE + (unsigned long)(x)) -#define STACK_TOP VDSO_BASE +#define STACK_TOP TASK_SIZE /* STACK_TOP_MAX is used temporarily in execve and should not check COMPAT. */ #define STACK_TOP_MAX TASK_SIZE_MAX @@ -202,6 +205,7 @@ static inline void start_thread(struct pt_regs *regs, { regs->pc = pc; regs->sp = usp; + single_step_execve(); } /* Free all resources held by a thread. */ @@ -210,35 +214,40 @@ static inline void release_thread(struct task_struct *dead_task) /* Nothing for now */ } -/* Prepare to copy thread state - unlazy all lazy status. */ -#define prepare_to_copy(tsk) do { } while (0) - -extern int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags); - extern int do_work_pending(struct pt_regs *regs, u32 flags); /* * Return saved (kernel) PC of a blocked thread. - * Only used in a printk() in kernel/sched.c, so don't work too hard. + * Only used in a printk() in kernel/sched/core.c, so don't work too hard. */ #define thread_saved_pc(t) ((t)->thread.pc) unsigned long get_wchan(struct task_struct *p); /* Return initial ksp value for given task. */ -#define task_ksp0(task) ((unsigned long)(task)->stack + THREAD_SIZE) +#define task_ksp0(task) \ + ((unsigned long)(task)->stack + THREAD_SIZE - STACK_TOP_DELTA) /* Return some info about the user process TASK. */ -#define KSTK_TOP(task) (task_ksp0(task) - STACK_TOP_DELTA) #define task_pt_regs(task) \ - ((struct pt_regs *)(task_ksp0(task) - KSTK_PTREGS_GAP) - 1) + ((struct pt_regs *)(task_ksp0(task) - KSTK_PTREGS_GAP) - 1) +#define current_pt_regs() \ + ((struct pt_regs *)((stack_pointer | (THREAD_SIZE - 1)) - \ + STACK_TOP_DELTA - (KSTK_PTREGS_GAP - 1)) - 1) #define task_sp(task) (task_pt_regs(task)->sp) #define task_pc(task) (task_pt_regs(task)->pc) /* Aliases for pc and sp (used in fs/proc/array.c) */ #define KSTK_EIP(task) task_pc(task) #define KSTK_ESP(task) task_sp(task) +/* Fine-grained unaligned JIT support */ +#define GET_UNALIGN_CTL(tsk, adr) get_unalign_ctl((tsk), (adr)) +#define SET_UNALIGN_CTL(tsk, val) set_unalign_ctl((tsk), (val)) + +extern int get_unalign_ctl(struct task_struct *tsk, unsigned long adr); +extern int set_unalign_ctl(struct task_struct *tsk, unsigned int val); + /* Standard format for printing registers and other word-size data. */ #ifdef __tilegx__ # define REGFMT "0x%016lx" @@ -267,7 +276,6 @@ extern char chip_model[64]; /* Data on which physical memory controller corresponds to which NUMA node. */ extern int node_controller[]; -#if CHIP_HAS_CBOX_HOME_MAP() /* Does the heap allocator return hash-for-home pages by default? */ extern int hash_default; @@ -277,11 +285,6 @@ extern int kstack_hash; /* Does MAP_ANONYMOUS return hash-for-home pages by default? */ #define uheap_hash hash_default -#else -#define hash_default 0 -#define kstack_hash 0 -#define uheap_hash 0 -#endif /* Are we using huge pages in the TLB for kernel data? */ extern int kdata_huge; @@ -329,7 +332,6 @@ extern int kdata_huge; /* * Provide symbolic constants for PLs. - * Note that assembly code assumes that USER_PL is zero. */ #define USER_PL 0 #if CONFIG_KERNEL_PL == 2 @@ -338,20 +340,38 @@ extern int kdata_huge; #define KERNEL_PL CONFIG_KERNEL_PL /* SYSTEM_SAVE_K_0 holds the current cpu number ORed with ksp0. */ -#define CPU_LOG_MASK_VALUE 12 -#define CPU_MASK_VALUE ((1 << CPU_LOG_MASK_VALUE) - 1) -#if CONFIG_NR_CPUS > CPU_MASK_VALUE -# error Too many cpus! +#ifdef __tilegx__ +#define CPU_SHIFT 48 +#if CHIP_VA_WIDTH() > CPU_SHIFT +# error Too many VA bits! #endif +#define MAX_CPU_ID ((1 << (64 - CPU_SHIFT)) - 1) #define raw_smp_processor_id() \ - ((int)__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & CPU_MASK_VALUE) + ((int)(__insn_mfspr(SPR_SYSTEM_SAVE_K_0) >> CPU_SHIFT)) #define get_current_ksp0() \ - (__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & ~CPU_MASK_VALUE) + ((unsigned long)(((long)__insn_mfspr(SPR_SYSTEM_SAVE_K_0) << \ + (64 - CPU_SHIFT)) >> (64 - CPU_SHIFT))) +#define next_current_ksp0(task) ({ \ + unsigned long __ksp0 = task_ksp0(task) & ((1UL << CPU_SHIFT) - 1); \ + unsigned long __cpu = (long)raw_smp_processor_id() << CPU_SHIFT; \ + __ksp0 | __cpu; \ +}) +#else +#define LOG2_NR_CPU_IDS 6 +#define MAX_CPU_ID ((1 << LOG2_NR_CPU_IDS) - 1) +#define raw_smp_processor_id() \ + ((int)__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & MAX_CPU_ID) +#define get_current_ksp0() \ + (__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & ~MAX_CPU_ID) #define next_current_ksp0(task) ({ \ unsigned long __ksp0 = task_ksp0(task); \ int __cpu = raw_smp_processor_id(); \ - BUG_ON(__ksp0 & CPU_MASK_VALUE); \ + BUG_ON(__ksp0 & MAX_CPU_ID); \ __ksp0 | __cpu; \ }) +#endif +#if CONFIG_NR_CPUS > (MAX_CPU_ID + 1) +# error Too many cpus! +#endif #endif /* _ASM_TILE_PROCESSOR_H */ diff --git a/arch/tile/include/asm/ptrace.h b/arch/tile/include/asm/ptrace.h index c6cddd7e8d5..b9620c077ab 100644 --- a/arch/tile/include/asm/ptrace.h +++ b/arch/tile/include/asm/ptrace.h @@ -11,87 +11,20 @@ * NON INFRINGEMENT. See the GNU General Public License for * more details. */ - #ifndef _ASM_TILE_PTRACE_H #define _ASM_TILE_PTRACE_H -#include <arch/chip.h> -#include <arch/abi.h> - -/* These must match struct pt_regs, below. */ -#if CHIP_WORD_SIZE() == 32 -#define PTREGS_OFFSET_REG(n) ((n)*4) -#else -#define PTREGS_OFFSET_REG(n) ((n)*8) -#endif -#define PTREGS_OFFSET_BASE 0 -#define PTREGS_OFFSET_TP PTREGS_OFFSET_REG(53) -#define PTREGS_OFFSET_SP PTREGS_OFFSET_REG(54) -#define PTREGS_OFFSET_LR PTREGS_OFFSET_REG(55) -#define PTREGS_NR_GPRS 56 -#define PTREGS_OFFSET_PC PTREGS_OFFSET_REG(56) -#define PTREGS_OFFSET_EX1 PTREGS_OFFSET_REG(57) -#define PTREGS_OFFSET_FAULTNUM PTREGS_OFFSET_REG(58) -#define PTREGS_OFFSET_ORIG_R0 PTREGS_OFFSET_REG(59) -#define PTREGS_OFFSET_FLAGS PTREGS_OFFSET_REG(60) -#if CHIP_HAS_CMPEXCH() -#define PTREGS_OFFSET_CMPEXCH PTREGS_OFFSET_REG(61) -#endif -#define PTREGS_SIZE PTREGS_OFFSET_REG(64) +#include <linux/compiler.h> #ifndef __ASSEMBLY__ - -#ifdef __KERNEL__ /* Benefit from consistent use of "long" on all chips. */ typedef unsigned long pt_reg_t; -#else -/* Provide appropriate length type to userspace regardless of -m32/-m64. */ -typedef uint_reg_t pt_reg_t; -#endif - -/* - * This struct defines the way the registers are stored on the stack during a - * system call or exception. "struct sigcontext" has the same shape. - */ -struct pt_regs { - /* Saved main processor registers; 56..63 are special. */ - /* tp, sp, and lr must immediately follow regs[] for aliasing. */ - pt_reg_t regs[53]; - pt_reg_t tp; /* aliases regs[TREG_TP] */ - pt_reg_t sp; /* aliases regs[TREG_SP] */ - pt_reg_t lr; /* aliases regs[TREG_LR] */ - - /* Saved special registers. */ - pt_reg_t pc; /* stored in EX_CONTEXT_K_0 */ - pt_reg_t ex1; /* stored in EX_CONTEXT_K_1 (PL and ICS bit) */ - pt_reg_t faultnum; /* fault number (INT_SWINT_1 for syscall) */ - pt_reg_t orig_r0; /* r0 at syscall entry, else zero */ - pt_reg_t flags; /* flags (see below) */ -#if !CHIP_HAS_CMPEXCH() - pt_reg_t pad[3]; -#else - pt_reg_t cmpexch; /* value of CMPEXCH_VALUE SPR at interrupt */ - pt_reg_t pad[2]; #endif -}; - -#endif /* __ASSEMBLY__ */ -#define PTRACE_GETREGS 12 -#define PTRACE_SETREGS 13 -#define PTRACE_GETFPREGS 14 -#define PTRACE_SETFPREGS 15 +#include <uapi/asm/ptrace.h> -/* Support TILE-specific ptrace options, with events starting at 16. */ -#define PTRACE_O_TRACEMIGRATE 0x00010000 -#define PTRACE_EVENT_MIGRATE 16 -#ifdef __KERNEL__ #define PTRACE_O_MASK_TILE (PTRACE_O_TRACEMIGRATE) -#define PT_TRACE_MIGRATE 0x00080000 -#define PT_TRACE_MASK_TILE (PT_TRACE_MIGRATE) -#endif - -#ifdef __KERNEL__ +#define PT_TRACE_MIGRATE PT_EVENT_FLAG(PTRACE_EVENT_MIGRATE) /* Flag bits in pt_regs.flags */ #define PT_FLAGS_DISABLE_IRQ 1 /* on return to kernel, disable irqs */ @@ -100,17 +33,20 @@ struct pt_regs { #ifndef __ASSEMBLY__ +#define regs_return_value(regs) ((regs)->regs[0]) #define instruction_pointer(regs) ((regs)->pc) #define profile_pc(regs) instruction_pointer(regs) +#define user_stack_pointer(regs) ((regs)->sp) /* Does the process account for user or for system time? */ -#define user_mode(regs) (EX1_PL((regs)->ex1) == USER_PL) +#define user_mode(regs) (EX1_PL((regs)->ex1) < KERNEL_PL) /* Fill in a struct pt_regs with the current kernel registers. */ struct pt_regs *get_pt_regs(struct pt_regs *); /* Trace the current syscall. */ -extern void do_syscall_trace(void); +extern int do_syscall_trace_enter(struct pt_regs *regs); +extern void do_syscall_trace_exit(struct pt_regs *regs); #define arch_has_single_step() (1) @@ -144,8 +80,7 @@ extern void single_step_execve(void); struct task_struct; -extern void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, - int error_code); +extern void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs); #ifdef __tilegx__ /* We need this since sigval_t has a user pointer in it, for GETSIGINFO etc. */ @@ -159,6 +94,4 @@ extern void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, #define SINGLESTEP_STATE_TARGET_LB 2 #define SINGLESTEP_STATE_TARGET_UB 7 -#endif /* !__KERNEL__ */ - #endif /* _ASM_TILE_PTRACE_H */ diff --git a/arch/tile/include/asm/sections.h b/arch/tile/include/asm/sections.h index d062d463fca..5d5d3b739a6 100644 --- a/arch/tile/include/asm/sections.h +++ b/arch/tile/include/asm/sections.h @@ -25,16 +25,22 @@ extern char _sinitdata[], _einitdata[]; /* Write-once data is writable only till the end of initialization. */ extern char __w1data_begin[], __w1data_end[]; +extern char vdso_start[], vdso_end[]; +#ifdef CONFIG_COMPAT +extern char vdso32_start[], vdso32_end[]; +#endif /* Not exactly sections, but PC comparison points in the code. */ extern char __rt_sigreturn[], __rt_sigreturn_end[]; -#ifndef __tilegx__ +#ifdef __tilegx__ +extern char __start_unalign_asm_code[], __end_unalign_asm_code[]; +#else extern char sys_cmpxchg[], __sys_cmpxchg_end[]; extern char __sys_cmpxchg_grab_lock[]; extern char __start_atomic_asm_code[], __end_atomic_asm_code[]; #endif -/* Handle the discontiguity between _sdata and _stext. */ +/* Handle the discontiguity between _sdata and _text. */ static inline int arch_is_kernel_data(unsigned long addr) { return addr >= (unsigned long)_sdata && diff --git a/arch/tile/include/asm/setup.h b/arch/tile/include/asm/setup.h index 7caf0f36b03..e98909033e5 100644 --- a/arch/tile/include/asm/setup.h +++ b/arch/tile/include/asm/setup.h @@ -11,26 +11,42 @@ * NON INFRINGEMENT. See the GNU General Public License for * more details. */ - #ifndef _ASM_TILE_SETUP_H #define _ASM_TILE_SETUP_H -#define COMMAND_LINE_SIZE 2048 - -#ifdef __KERNEL__ #include <linux/pfn.h> #include <linux/init.h> +#include <uapi/asm/setup.h> /* * Reserved space for vmalloc and iomap - defined in asm/page.h */ #define MAXMEM_PFN PFN_DOWN(MAXMEM) +int tile_console_write(const char *buf, int count); void early_panic(const char *fmt, ...); -void warn_early_printk(void); -void __init disable_early_printk(void); -#endif /* __KERNEL__ */ +/* Init-time routine to do tile-specific per-cpu setup. */ +void setup_cpu(int boot); + +/* User-level DMA management functions */ +void grant_dma_mpls(void); +void restrict_dma_mpls(void); + +#ifdef CONFIG_HARDWALL +/* User-level network management functions */ +void reset_network_state(void); +struct task_struct; +void hardwall_switch_tasks(struct task_struct *prev, struct task_struct *next); +void hardwall_deactivate_all(struct task_struct *task); +int hardwall_ipi_valid(int cpu); + +/* Hook hardwall code into changes in affinity. */ +#define arch_set_cpus_allowed(p, new_mask) do { \ + if (!cpumask_equal(&p->cpus_allowed, new_mask)) \ + hardwall_deactivate_all(p); \ +} while (0) +#endif #endif /* _ASM_TILE_SETUP_H */ diff --git a/arch/tile/include/asm/sigcontext.h b/arch/tile/include/asm/sigcontext.h deleted file mode 100644 index 6348e59d372..00000000000 --- a/arch/tile/include/asm/sigcontext.h +++ /dev/null @@ -1,37 +0,0 @@ -/* - * Copyright 2010 Tilera Corporation. All Rights Reserved. - * - * This program 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, version 2. - * - * This program 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, GOOD TITLE or - * NON INFRINGEMENT. See the GNU General Public License for - * more details. - */ - -#ifndef _ASM_TILE_SIGCONTEXT_H -#define _ASM_TILE_SIGCONTEXT_H - -/* Don't pollute the namespace since <signal.h> includes this file. */ -#define __need_int_reg_t -#include <arch/abi.h> - -/* - * struct sigcontext has the same shape as struct pt_regs, - * but is simplified since we know the fault is from userspace. - */ -struct sigcontext { - __uint_reg_t gregs[53]; /* General-purpose registers. */ - __uint_reg_t tp; /* Aliases gregs[TREG_TP]. */ - __uint_reg_t sp; /* Aliases gregs[TREG_SP]. */ - __uint_reg_t lr; /* Aliases gregs[TREG_LR]. */ - __uint_reg_t pc; /* Program counter. */ - __uint_reg_t ics; /* In Interrupt Critical Section? */ - __uint_reg_t faultnum; /* Fault number. */ - __uint_reg_t pad[5]; -}; - -#endif /* _ASM_TILE_SIGCONTEXT_H */ diff --git a/arch/tile/include/asm/signal.h b/arch/tile/include/asm/signal.h index 1e5e49aad54..10e183de96d 100644 --- a/arch/tile/include/asm/signal.h +++ b/arch/tile/include/asm/signal.h @@ -11,19 +11,11 @@ * NON INFRINGEMENT. See the GNU General Public License for * more details. */ - #ifndef _ASM_TILE_SIGNAL_H #define _ASM_TILE_SIGNAL_H -/* Do not notify a ptracer when this signal is handled. */ -#define SA_NOPTRACE 0x02000000u - -/* Used in earlier Tilera releases, so keeping for binary compatibility. */ -#define SA_RESTORER 0x04000000u - -#include <asm-generic/signal.h> +#include <uapi/asm/signal.h> -#if defined(__KERNEL__) #if !defined(__ASSEMBLY__) struct pt_regs; int restore_sigcontext(struct pt_regs *, struct sigcontext __user *); @@ -34,6 +26,4 @@ void signal_fault(const char *type, struct pt_regs *, void trace_unhandled_signal(const char *type, struct pt_regs *regs, unsigned long address, int signo); #endif -#endif - #endif /* _ASM_TILE_SIGNAL_H */ diff --git a/arch/tile/include/asm/smp.h b/arch/tile/include/asm/smp.h index 532124ae4b1..9a326b64f7a 100644 --- a/arch/tile/include/asm/smp.h +++ b/arch/tile/include/asm/smp.h @@ -43,10 +43,6 @@ void evaluate_message(int tag); /* Boot a secondary cpu */ void online_secondary(void); -/* Call a function on a specified set of CPUs (may include this one). */ -extern void on_each_cpu_mask(const struct cpumask *mask, - void (*func)(void *), void *info, bool wait); - /* Topology of the supervisor tile grid, and coordinates of boot processor */ extern HV_Topology smp_topology; @@ -91,9 +87,6 @@ void print_disabled_cpus(void); #else /* !CONFIG_SMP */ -#define on_each_cpu_mask(mask, func, info, wait) \ - do { if (cpumask_test_cpu(0, (mask))) func(info); } while (0) - #define smp_master_cpu 0 #define smp_height 1 #define smp_width 1 @@ -108,10 +101,8 @@ void print_disabled_cpus(void); extern struct cpumask cpu_lotar_map; #define cpu_is_valid_lotar(cpu) cpumask_test_cpu((cpu), &cpu_lotar_map) -#if CHIP_HAS_CBOX_HOME_MAP() /* Which processors are used for hash-for-home mapping */ extern struct cpumask hash_for_home_map; -#endif /* Which cpus can have their cache flushed by hv_flush_remote(). */ extern struct cpumask cpu_cacheable_map; diff --git a/arch/tile/include/asm/spinlock_32.h b/arch/tile/include/asm/spinlock_32.h index a5e4208d34f..c0a77b38d39 100644 --- a/arch/tile/include/asm/spinlock_32.h +++ b/arch/tile/include/asm/spinlock_32.h @@ -19,7 +19,6 @@ #include <linux/atomic.h> #include <asm/page.h> -#include <asm/system.h> #include <linux/compiler.h> /* diff --git a/arch/tile/include/asm/spinlock_64.h b/arch/tile/include/asm/spinlock_64.h index 72be5904e02..9a12b9c7e5d 100644 --- a/arch/tile/include/asm/spinlock_64.h +++ b/arch/tile/include/asm/spinlock_64.h @@ -27,7 +27,7 @@ * Return the "current" portion of a ticket lock value, * i.e. the number that currently owns the lock. */ -static inline int arch_spin_current(u32 val) +static inline u32 arch_spin_current(u32 val) { return val >> __ARCH_SPIN_CURRENT_SHIFT; } @@ -36,7 +36,7 @@ static inline int arch_spin_current(u32 val) * Return the "next" portion of a ticket lock value, * i.e. the number that the next task to try to acquire the lock will get. */ -static inline int arch_spin_next(u32 val) +static inline u32 arch_spin_next(u32 val) { return val & __ARCH_SPIN_NEXT_MASK; } @@ -137,7 +137,7 @@ static inline void arch_read_unlock(arch_rwlock_t *rw) static inline void arch_write_unlock(arch_rwlock_t *rw) { __insn_mf(); - rw->lock = 0; + __insn_exch4(&rw->lock, 0); /* Avoid waiting in the write buffer. */ } static inline int arch_read_trylock(arch_rwlock_t *rw) diff --git a/arch/tile/include/asm/stack.h b/arch/tile/include/asm/stack.h index 4d97a2db932..0e9d382a2d4 100644 --- a/arch/tile/include/asm/stack.h +++ b/arch/tile/include/asm/stack.h @@ -25,7 +25,6 @@ struct KBacktraceIterator { BacktraceIterator it; struct task_struct *task; /* task we are backtracing */ - pte_t *pgtable; /* page table for user space access */ int end; /* iteration complete. */ int new_context; /* new context is starting */ int profile; /* profiling, so stop on async intrpt */ diff --git a/arch/tile/include/asm/stat.h b/arch/tile/include/asm/stat.h deleted file mode 100644 index c0db34d56be..00000000000 --- a/arch/tile/include/asm/stat.h +++ /dev/null @@ -1,4 +0,0 @@ -#if defined(__KERNEL__) && defined(CONFIG_COMPAT) -#define __ARCH_WANT_STAT64 /* Used for compat_sys_stat64() etc. */ -#endif -#include <asm-generic/stat.h> diff --git a/arch/tile/include/asm/string.h b/arch/tile/include/asm/string.h index 7535cf1a30e..92b271bd9eb 100644 --- a/arch/tile/include/asm/string.h +++ b/arch/tile/include/asm/string.h @@ -21,8 +21,10 @@ #define __HAVE_ARCH_MEMMOVE #define __HAVE_ARCH_STRCHR #define __HAVE_ARCH_STRLEN +#define __HAVE_ARCH_STRNLEN extern __kernel_size_t strlen(const char *); +extern __kernel_size_t strnlen(const char *, __kernel_size_t); extern char *strchr(const char *s, int c); extern void *memchr(const void *s, int c, size_t n); extern void *memset(void *, int, __kernel_size_t); diff --git a/arch/tile/include/asm/swab.h b/arch/tile/include/asm/swab.h deleted file mode 100644 index 7c37b38f6c8..00000000000 --- a/arch/tile/include/asm/swab.h +++ /dev/null @@ -1,23 +0,0 @@ -/* - * Copyright 2010 Tilera Corporation. All Rights Reserved. - * - * This program 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, version 2. - * - * This program 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, GOOD TITLE or - * NON INFRINGEMENT. See the GNU General Public License for - * more details. - */ - -#ifndef _ASM_TILE_SWAB_H -#define _ASM_TILE_SWAB_H - -/* Tile gcc is always >= 4.3.0, so we use __builtin_bswap. */ -#define __arch_swab32(x) __builtin_bswap32(x) -#define __arch_swab64(x) __builtin_bswap64(x) -#define __arch_swab16(x) (__builtin_bswap32(x) >> 16) - -#endif /* _ASM_TILE_SWAB_H */ diff --git a/arch/tile/include/asm/switch_to.h b/arch/tile/include/asm/switch_to.h new file mode 100644 index 00000000000..b8f888cbe6b --- /dev/null +++ b/arch/tile/include/asm/switch_to.h @@ -0,0 +1,79 @@ +/* + * Copyright 2010 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + */ + +#ifndef _ASM_TILE_SWITCH_TO_H +#define _ASM_TILE_SWITCH_TO_H + +#include <arch/sim_def.h> + +/* + * switch_to(n) should switch tasks to task nr n, first + * checking that n isn't the current task, in which case it does nothing. + * The number of callee-saved registers saved on the kernel stack + * is defined here for use in copy_thread() and must agree with __switch_to(). + */ +#define CALLEE_SAVED_FIRST_REG 30 +#define CALLEE_SAVED_REGS_COUNT 24 /* r30 to r52, plus an empty to align */ + +#ifndef __ASSEMBLY__ + +struct task_struct; + +/* + * Pause the DMA engine and static network before task switching. + */ +#define prepare_arch_switch(next) _prepare_arch_switch(next) +void _prepare_arch_switch(struct task_struct *next); + +struct task_struct; +#define switch_to(prev, next, last) ((last) = _switch_to((prev), (next))) +extern struct task_struct *_switch_to(struct task_struct *prev, + struct task_struct *next); + +/* Helper function for _switch_to(). */ +extern struct task_struct *__switch_to(struct task_struct *prev, + struct task_struct *next, + unsigned long new_system_save_k_0); + +/* Address that switched-away from tasks are at. */ +extern unsigned long get_switch_to_pc(void); + +/* + * Kernel threads can check to see if they need to migrate their + * stack whenever they return from a context switch; for user + * threads, we defer until they are returning to user-space. + */ +#define finish_arch_switch(prev) do { \ + if (unlikely((prev)->state == TASK_DEAD)) \ + __insn_mtspr(SPR_SIM_CONTROL, SIM_CONTROL_OS_EXIT | \ + ((prev)->pid << _SIM_CONTROL_OPERATOR_BITS)); \ + __insn_mtspr(SPR_SIM_CONTROL, SIM_CONTROL_OS_SWITCH | \ + (current->pid << _SIM_CONTROL_OPERATOR_BITS)); \ + if (current->mm == NULL && !kstack_hash && \ + current_thread_info()->homecache_cpu != smp_processor_id()) \ + homecache_migrate_kthread(); \ +} while (0) + +/* Support function for forking a new task. */ +void ret_from_fork(void); + +/* Support function for forking a new kernel thread. */ +void ret_from_kernel_thread(void *fn, void *arg); + +/* Called from ret_from_xxx() when a new process starts up. */ +struct task_struct *sim_notify_fork(struct task_struct *prev); + +#endif /* !__ASSEMBLY__ */ + +#endif /* _ASM_TILE_SWITCH_TO_H */ diff --git a/arch/tile/include/asm/syscall.h b/arch/tile/include/asm/syscall.h index d35e0dcb67b..9644b88f133 100644 --- a/arch/tile/include/asm/syscall.h +++ b/arch/tile/include/asm/syscall.h @@ -22,6 +22,12 @@ #include <linux/err.h> #include <arch/abi.h> +/* The array of function pointers for syscalls. */ +extern void *sys_call_table[]; +#ifdef CONFIG_COMPAT +extern void *compat_sys_call_table[]; +#endif + /* * Only the low 32 bits of orig_r0 are meaningful, so we return int. * This importantly ignores the high bits on 64-bit, so comparisons diff --git a/arch/tile/include/asm/syscalls.h b/arch/tile/include/asm/syscalls.h index 3b5507c31ea..07b298450ef 100644 --- a/arch/tile/include/asm/syscalls.h +++ b/arch/tile/include/asm/syscalls.h @@ -24,12 +24,6 @@ #include <linux/types.h> #include <linux/compat.h> -/* The array of function pointers for syscalls. */ -extern void *sys_call_table[]; -#ifdef CONFIG_COMPAT -extern void *compat_sys_call_table[]; -#endif - /* * Note that by convention, any syscall which requires the current * register set takes an additional "struct pt_regs *" pointer; a @@ -43,15 +37,15 @@ long sys32_fadvise64(int fd, u32 offset_lo, u32 offset_hi, u32 len, int advice); int sys32_fadvise64_64(int fd, u32 offset_lo, u32 offset_hi, u32 len_lo, u32 len_hi, int advice); -long sys_flush_cache(void); +long sys_cacheflush(unsigned long addr, unsigned long len, + unsigned long flags); #ifndef __tilegx__ /* No mmap() in the 32-bit kernel. */ #define sys_mmap sys_mmap #endif #ifndef __tilegx__ /* mm/fault.c */ -long sys_cmpxchg_badaddr(unsigned long address, struct pt_regs *); -long _sys_cmpxchg_badaddr(unsigned long address); +long sys_cmpxchg_badaddr(unsigned long address); #endif #ifdef CONFIG_COMPAT @@ -62,14 +56,14 @@ long sys_truncate64(const char __user *path, loff_t length); long sys_ftruncate64(unsigned int fd, loff_t length); #endif +/* Provide versions of standard syscalls that use current_pt_regs(). */ +long sys_rt_sigreturn(void); +#define sys_rt_sigreturn sys_rt_sigreturn + /* These are the intvec*.S trampolines. */ -long _sys_sigaltstack(const stack_t __user *, stack_t __user *); long _sys_rt_sigreturn(void); long _sys_clone(unsigned long clone_flags, unsigned long newsp, void __user *parent_tid, void __user *child_tid); -long _sys_execve(const char __user *filename, - const char __user *const __user *argv, - const char __user *const __user *envp); #include <asm-generic/syscalls.h> diff --git a/arch/tile/include/asm/system.h b/arch/tile/include/asm/system.h deleted file mode 100644 index 23d1842f483..00000000000 --- a/arch/tile/include/asm/system.h +++ /dev/null @@ -1,261 +0,0 @@ -/* - * Copyright 2010 Tilera Corporation. All Rights Reserved. - * - * This program 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, version 2. - * - * This program 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, GOOD TITLE or - * NON INFRINGEMENT. See the GNU General Public License for - * more details. - */ - -#ifndef _ASM_TILE_SYSTEM_H -#define _ASM_TILE_SYSTEM_H - -#ifndef __ASSEMBLY__ - -#include <linux/types.h> -#include <linux/irqflags.h> - -/* NOTE: we can't include <linux/ptrace.h> due to #include dependencies. */ -#include <asm/ptrace.h> - -#include <arch/chip.h> -#include <arch/sim_def.h> -#include <arch/spr_def.h> - -/* - * read_barrier_depends - Flush all pending reads that subsequents reads - * depend on. - * - * No data-dependent reads from memory-like regions are ever reordered - * over this barrier. All reads preceding this primitive are guaranteed - * to access memory (but not necessarily other CPUs' caches) before any - * reads following this primitive that depend on the data return by - * any of the preceding reads. This primitive is much lighter weight than - * rmb() on most CPUs, and is never heavier weight than is - * rmb(). - * - * These ordering constraints are respected by both the local CPU - * and the compiler. - * - * Ordering is not guaranteed by anything other than these primitives, - * not even by data dependencies. See the documentation for - * memory_barrier() for examples and URLs to more information. - * - * For example, the following code would force ordering (the initial - * value of "a" is zero, "b" is one, and "p" is "&a"): - * - * <programlisting> - * CPU 0 CPU 1 - * - * b = 2; - * memory_barrier(); - * p = &b; q = p; - * read_barrier_depends(); - * d = *q; - * </programlisting> - * - * because the read of "*q" depends on the read of "p" and these - * two reads are separated by a read_barrier_depends(). However, - * the following code, with the same initial values for "a" and "b": - * - * <programlisting> - * CPU 0 CPU 1 - * - * a = 2; - * memory_barrier(); - * b = 3; y = b; - * read_barrier_depends(); - * x = a; - * </programlisting> - * - * does not enforce ordering, since there is no data dependency between - * the read of "a" and the read of "b". Therefore, on some CPUs, such - * as Alpha, "y" could be set to 3 and "x" to 0. Use rmb() - * in cases like this where there are no data dependencies. - */ - -#define read_barrier_depends() do { } while (0) - -#define __sync() __insn_mf() - -#if CHIP_HAS_SPLIT_CYCLE() -#define get_cycles_low() __insn_mfspr(SPR_CYCLE_LOW) -#else -#define get_cycles_low() __insn_mfspr(SPR_CYCLE) /* just get all 64 bits */ -#endif - -#if !CHIP_HAS_MF_WAITS_FOR_VICTIMS() -#include <hv/syscall_public.h> -/* - * Issue an uncacheable load to each memory controller, then - * wait until those loads have completed. - */ -static inline void __mb_incoherent(void) -{ - long clobber_r10; - asm volatile("swint2" - : "=R10" (clobber_r10) - : "R10" (HV_SYS_fence_incoherent) - : "r0", "r1", "r2", "r3", "r4", - "r5", "r6", "r7", "r8", "r9", - "r11", "r12", "r13", "r14", - "r15", "r16", "r17", "r18", "r19", - "r20", "r21", "r22", "r23", "r24", - "r25", "r26", "r27", "r28", "r29"); -} -#endif - -/* Fence to guarantee visibility of stores to incoherent memory. */ -static inline void -mb_incoherent(void) -{ - __insn_mf(); - -#if !CHIP_HAS_MF_WAITS_FOR_VICTIMS() - { -#if CHIP_HAS_TILE_WRITE_PENDING() - const unsigned long WRITE_TIMEOUT_CYCLES = 400; - unsigned long start = get_cycles_low(); - do { - if (__insn_mfspr(SPR_TILE_WRITE_PENDING) == 0) - return; - } while ((get_cycles_low() - start) < WRITE_TIMEOUT_CYCLES); -#endif /* CHIP_HAS_TILE_WRITE_PENDING() */ - (void) __mb_incoherent(); - } -#endif /* CHIP_HAS_MF_WAITS_FOR_VICTIMS() */ -} - -#define fast_wmb() __sync() -#define fast_rmb() __sync() -#define fast_mb() __sync() -#define fast_iob() mb_incoherent() - -#define wmb() fast_wmb() -#define rmb() fast_rmb() -#define mb() fast_mb() -#define iob() fast_iob() - -#ifdef CONFIG_SMP -#define smp_mb() mb() -#define smp_rmb() rmb() -#define smp_wmb() wmb() -#define smp_read_barrier_depends() read_barrier_depends() -#else -#define smp_mb() barrier() -#define smp_rmb() barrier() -#define smp_wmb() barrier() -#define smp_read_barrier_depends() do { } while (0) -#endif - -#define set_mb(var, value) \ - do { var = value; mb(); } while (0) - -/* - * Pause the DMA engine and static network before task switching. - */ -#define prepare_arch_switch(next) _prepare_arch_switch(next) -void _prepare_arch_switch(struct task_struct *next); - - -/* - * switch_to(n) should switch tasks to task nr n, first - * checking that n isn't the current task, in which case it does nothing. - * The number of callee-saved registers saved on the kernel stack - * is defined here for use in copy_thread() and must agree with __switch_to(). - */ -#endif /* !__ASSEMBLY__ */ -#define CALLEE_SAVED_FIRST_REG 30 -#define CALLEE_SAVED_REGS_COUNT 24 /* r30 to r52, plus an empty to align */ -#ifndef __ASSEMBLY__ -struct task_struct; -#define switch_to(prev, next, last) ((last) = _switch_to((prev), (next))) -extern struct task_struct *_switch_to(struct task_struct *prev, - struct task_struct *next); - -/* Helper function for _switch_to(). */ -extern struct task_struct *__switch_to(struct task_struct *prev, - struct task_struct *next, - unsigned long new_system_save_k_0); - -/* Address that switched-away from tasks are at. */ -extern unsigned long get_switch_to_pc(void); - -/* - * On SMP systems, when the scheduler does migration-cost autodetection, - * it needs a way to flush as much of the CPU's caches as possible: - * - * TODO: fill this in! - */ -static inline void sched_cacheflush(void) -{ -} - -#define arch_align_stack(x) (x) - -/* - * Is the kernel doing fixups of unaligned accesses? If <0, no kernel - * intervention occurs and SIGBUS is delivered with no data address - * info. If 0, the kernel single-steps the instruction to discover - * the data address to provide with the SIGBUS. If 1, the kernel does - * a fixup. - */ -extern int unaligned_fixup; - -/* Is the kernel printing on each unaligned fixup? */ -extern int unaligned_printk; - -/* Number of unaligned fixups performed */ -extern unsigned int unaligned_fixup_count; - -/* Init-time routine to do tile-specific per-cpu setup. */ -void setup_cpu(int boot); - -/* User-level DMA management functions */ -void grant_dma_mpls(void); -void restrict_dma_mpls(void); - -#ifdef CONFIG_HARDWALL -/* User-level network management functions */ -void reset_network_state(void); -void grant_network_mpls(void); -void restrict_network_mpls(void); -int hardwall_deactivate(struct task_struct *task); - -/* Hook hardwall code into changes in affinity. */ -#define arch_set_cpus_allowed(p, new_mask) do { \ - if (p->thread.hardwall && !cpumask_equal(&p->cpus_allowed, new_mask)) \ - hardwall_deactivate(p); \ -} while (0) -#endif - -/* - * Kernel threads can check to see if they need to migrate their - * stack whenever they return from a context switch; for user - * threads, we defer until they are returning to user-space. - */ -#define finish_arch_switch(prev) do { \ - if (unlikely((prev)->state == TASK_DEAD)) \ - __insn_mtspr(SPR_SIM_CONTROL, SIM_CONTROL_OS_EXIT | \ - ((prev)->pid << _SIM_CONTROL_OPERATOR_BITS)); \ - __insn_mtspr(SPR_SIM_CONTROL, SIM_CONTROL_OS_SWITCH | \ - (current->pid << _SIM_CONTROL_OPERATOR_BITS)); \ - if (current->mm == NULL && !kstack_hash && \ - current_thread_info()->homecache_cpu != smp_processor_id()) \ - homecache_migrate_kthread(); \ -} while (0) - -/* Support function for forking a new task. */ -void ret_from_fork(void); - -/* Called from ret_from_fork() when a new process starts up. */ -struct task_struct *sim_notify_fork(struct task_struct *prev); - -#endif /* !__ASSEMBLY__ */ - -#endif /* _ASM_TILE_SYSTEM_H */ diff --git a/arch/tile/include/asm/thread_info.h b/arch/tile/include/asm/thread_info.h index bc4f562bd45..48e4fd0f38e 100644 --- a/arch/tile/include/asm/thread_info.h +++ b/arch/tile/include/asm/thread_info.h @@ -39,6 +39,11 @@ struct thread_info { struct restart_block restart_block; struct single_step_state *step_state; /* single step state (if non-zero) */ + int align_ctl; /* controls unaligned access */ +#ifdef __tilegx__ + unsigned long unalign_jit_tmp[4]; /* temp r0..r3 storage */ + void __user *unalign_jit_base; /* unalign fixup JIT base */ +#endif }; /* @@ -56,6 +61,7 @@ struct thread_info { .fn = do_no_restart_syscall, \ }, \ .step_state = NULL, \ + .align_ctl = 0, \ } #define init_thread_info (init_thread_union.thread_info) @@ -77,40 +83,36 @@ struct thread_info { #ifndef __ASSEMBLY__ +void arch_release_thread_info(struct thread_info *info); + /* How to get the thread information struct from C. */ register unsigned long stack_pointer __asm__("sp"); #define current_thread_info() \ ((struct thread_info *)(stack_pointer & -THREAD_SIZE)) -#define __HAVE_ARCH_THREAD_INFO_ALLOCATOR -extern struct thread_info *alloc_thread_info_node(struct task_struct *task, int node); -extern void free_thread_info(struct thread_info *info); - /* Sit on a nap instruction until interrupted. */ extern void smp_nap(void); -/* Enable interrupts racelessly and nap forever: helper for cpu_idle(). */ +/* Enable interrupts racelessly and nap forever: helper for arch_cpu_idle(). */ extern void _cpu_idle(void); -/* Switch boot idle thread to a freshly-allocated stack and free old stack. */ -extern void cpu_idle_on_new_stack(struct thread_info *old_ti, - unsigned long new_sp, - unsigned long new_ss10); - #else /* __ASSEMBLY__ */ -/* how to get the thread information struct from ASM */ +/* + * How to get the thread information struct from assembly. + * Note that we use different macros since different architectures + * have different semantics in their "mm" instruction and we would + * like to guarantee that the macro expands to exactly one instruction. + */ #ifdef __tilegx__ -#define GET_THREAD_INFO(reg) move reg, sp; mm reg, zero, LOG2_THREAD_SIZE, 63 +#define EXTRACT_THREAD_INFO(reg) mm reg, zero, LOG2_THREAD_SIZE, 63 #else #define GET_THREAD_INFO(reg) mm reg, sp, zero, LOG2_THREAD_SIZE, 31 #endif #endif /* !__ASSEMBLY__ */ -#define PREEMPT_ACTIVE 0x10000000 - /* * Thread information flags that various assembly files may need to access. * Keep flags accessed frequently in low bits, particular since it makes @@ -126,6 +128,8 @@ extern void cpu_idle_on_new_stack(struct thread_info *old_ti, #define TIF_SECCOMP 6 /* secure computing */ #define TIF_MEMDIE 7 /* OOM killer at work */ #define TIF_NOTIFY_RESUME 8 /* callback before returning to user */ +#define TIF_SYSCALL_TRACEPOINT 9 /* syscall tracepoint instrumentation */ +#define TIF_POLLING_NRFLAG 10 /* idle is polling for TIF_NEED_RESCHED */ #define _TIF_SIGPENDING (1<<TIF_SIGPENDING) #define _TIF_NEED_RESCHED (1<<TIF_NEED_RESCHED) @@ -136,12 +140,20 @@ extern void cpu_idle_on_new_stack(struct thread_info *old_ti, #define _TIF_SECCOMP (1<<TIF_SECCOMP) #define _TIF_MEMDIE (1<<TIF_MEMDIE) #define _TIF_NOTIFY_RESUME (1<<TIF_NOTIFY_RESUME) +#define _TIF_SYSCALL_TRACEPOINT (1<<TIF_SYSCALL_TRACEPOINT) +#define _TIF_POLLING_NRFLAG (1<<TIF_POLLING_NRFLAG) /* Work to do on any return to user space. */ #define _TIF_ALLWORK_MASK \ (_TIF_SIGPENDING|_TIF_NEED_RESCHED|_TIF_SINGLESTEP|\ _TIF_ASYNC_TLB|_TIF_NOTIFY_RESUME) +/* Work to do at syscall entry. */ +#define _TIF_SYSCALL_ENTRY_WORK (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_TRACEPOINT) + +/* Work to do at syscall exit. */ +#define _TIF_SYSCALL_EXIT_WORK (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_TRACEPOINT) + /* * Thread-synchronous status. * @@ -152,18 +164,31 @@ extern void cpu_idle_on_new_stack(struct thread_info *old_ti, #ifdef __tilegx__ #define TS_COMPAT 0x0001 /* 32-bit compatibility mode */ #endif -#define TS_POLLING 0x0004 /* in idle loop but not sleeping */ #define TS_RESTORE_SIGMASK 0x0008 /* restore signal mask in do_signal */ -#define tsk_is_polling(t) (task_thread_info(t)->status & TS_POLLING) - #ifndef __ASSEMBLY__ #define HAVE_SET_RESTORE_SIGMASK 1 static inline void set_restore_sigmask(void) { struct thread_info *ti = current_thread_info(); ti->status |= TS_RESTORE_SIGMASK; - set_bit(TIF_SIGPENDING, &ti->flags); + WARN_ON(!test_bit(TIF_SIGPENDING, &ti->flags)); +} +static inline void clear_restore_sigmask(void) +{ + current_thread_info()->status &= ~TS_RESTORE_SIGMASK; +} +static inline bool test_restore_sigmask(void) +{ + return current_thread_info()->status & TS_RESTORE_SIGMASK; +} +static inline bool test_and_clear_restore_sigmask(void) +{ + struct thread_info *ti = current_thread_info(); + if (!(ti->status & TS_RESTORE_SIGMASK)) + return false; + ti->status &= ~TS_RESTORE_SIGMASK; + return true; } #endif /* !__ASSEMBLY__ */ diff --git a/arch/tile/include/asm/timex.h b/arch/tile/include/asm/timex.h index 29921f0b86d..dc987d53e2a 100644 --- a/arch/tile/include/asm/timex.h +++ b/arch/tile/include/asm/timex.h @@ -29,11 +29,13 @@ typedef unsigned long long cycles_t; #if CHIP_HAS_SPLIT_CYCLE() cycles_t get_cycles(void); +#define get_cycles_low() __insn_mfspr(SPR_CYCLE_LOW) #else static inline cycles_t get_cycles(void) { return __insn_mfspr(SPR_CYCLE); } +#define get_cycles_low() __insn_mfspr(SPR_CYCLE) /* just get all 64 bits */ #endif cycles_t get_clock_rate(void); diff --git a/arch/tile/include/asm/tlbflush.h b/arch/tile/include/asm/tlbflush.h index 96199d214fb..dcf91b25a1e 100644 --- a/arch/tile/include/asm/tlbflush.h +++ b/arch/tile/include/asm/tlbflush.h @@ -38,16 +38,11 @@ DECLARE_PER_CPU(int, current_asid); /* The hypervisor tells us what ASIDs are available to us. */ extern int min_asid, max_asid; -static inline unsigned long hv_page_size(const struct vm_area_struct *vma) -{ - return (vma->vm_flags & VM_HUGETLB) ? HPAGE_SIZE : PAGE_SIZE; -} - /* Pass as vma pointer for non-executable mapping, if no vma available. */ -#define FLUSH_NONEXEC ((const struct vm_area_struct *)-1UL) +#define FLUSH_NONEXEC ((struct vm_area_struct *)-1UL) /* Flush a single user page on this cpu. */ -static inline void local_flush_tlb_page(const struct vm_area_struct *vma, +static inline void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long addr, unsigned long page_size) { @@ -60,7 +55,7 @@ static inline void local_flush_tlb_page(const struct vm_area_struct *vma, } /* Flush range of user pages on this cpu. */ -static inline void local_flush_tlb_pages(const struct vm_area_struct *vma, +static inline void local_flush_tlb_pages(struct vm_area_struct *vma, unsigned long addr, unsigned long page_size, unsigned long len) @@ -117,10 +112,10 @@ extern void flush_tlb_all(void); extern void flush_tlb_kernel_range(unsigned long start, unsigned long end); extern void flush_tlb_current_task(void); extern void flush_tlb_mm(struct mm_struct *); -extern void flush_tlb_page(const struct vm_area_struct *, unsigned long); -extern void flush_tlb_page_mm(const struct vm_area_struct *, +extern void flush_tlb_page(struct vm_area_struct *, unsigned long); +extern void flush_tlb_page_mm(struct vm_area_struct *, struct mm_struct *, unsigned long); -extern void flush_tlb_range(const struct vm_area_struct *, +extern void flush_tlb_range(struct vm_area_struct *, unsigned long start, unsigned long end); #define flush_tlb() flush_tlb_current_task() diff --git a/arch/tile/include/asm/topology.h b/arch/tile/include/asm/topology.h index 6fdd0c86019..93831184423 100644 --- a/arch/tile/include/asm/topology.h +++ b/arch/tile/include/asm/topology.h @@ -44,66 +44,6 @@ static inline const struct cpumask *cpumask_of_node(int node) /* For now, use numa node -1 for global allocation. */ #define pcibus_to_node(bus) ((void)(bus), -1) -/* - * TILE architecture has many cores integrated in one processor, so we need - * setup bigger balance_interval for both CPU/NODE scheduling domains to - * reduce process scheduling costs. - */ - -/* sched_domains SD_CPU_INIT for TILE architecture */ -#define SD_CPU_INIT (struct sched_domain) { \ - .min_interval = 4, \ - .max_interval = 128, \ - .busy_factor = 64, \ - .imbalance_pct = 125, \ - .cache_nice_tries = 1, \ - .busy_idx = 2, \ - .idle_idx = 1, \ - .newidle_idx = 0, \ - .wake_idx = 0, \ - .forkexec_idx = 0, \ - \ - .flags = 1*SD_LOAD_BALANCE \ - | 1*SD_BALANCE_NEWIDLE \ - | 1*SD_BALANCE_EXEC \ - | 1*SD_BALANCE_FORK \ - | 0*SD_BALANCE_WAKE \ - | 0*SD_WAKE_AFFINE \ - | 0*SD_PREFER_LOCAL \ - | 0*SD_SHARE_CPUPOWER \ - | 0*SD_SHARE_PKG_RESOURCES \ - | 0*SD_SERIALIZE \ - , \ - .last_balance = jiffies, \ - .balance_interval = 32, \ -} - -/* sched_domains SD_NODE_INIT for TILE architecture */ -#define SD_NODE_INIT (struct sched_domain) { \ - .min_interval = 16, \ - .max_interval = 512, \ - .busy_factor = 32, \ - .imbalance_pct = 125, \ - .cache_nice_tries = 1, \ - .busy_idx = 3, \ - .idle_idx = 1, \ - .newidle_idx = 2, \ - .wake_idx = 1, \ - .flags = 1*SD_LOAD_BALANCE \ - | 1*SD_BALANCE_NEWIDLE \ - | 1*SD_BALANCE_EXEC \ - | 1*SD_BALANCE_FORK \ - | 0*SD_BALANCE_WAKE \ - | 0*SD_WAKE_AFFINE \ - | 0*SD_PREFER_LOCAL \ - | 0*SD_SHARE_CPUPOWER \ - | 0*SD_SHARE_PKG_RESOURCES \ - | 1*SD_SERIALIZE \ - , \ - .last_balance = jiffies, \ - .balance_interval = 128, \ -} - /* By definition, we create nodes based on online memory. */ #define node_has_online_mem(nid) 1 @@ -116,9 +56,6 @@ static inline const struct cpumask *cpumask_of_node(int node) #define topology_core_id(cpu) (cpu) #define topology_core_cpumask(cpu) ((void)(cpu), cpu_online_mask) #define topology_thread_cpumask(cpu) cpumask_of(cpu) - -/* indicates that pointers to the topology struct cpumask maps are valid */ -#define arch_provides_topology_pointers yes #endif #endif /* _ASM_TILE_TOPOLOGY_H */ diff --git a/arch/tile/include/asm/traps.h b/arch/tile/include/asm/traps.h index 5f20f920f93..4b99a1c3aab 100644 --- a/arch/tile/include/asm/traps.h +++ b/arch/tile/include/asm/traps.h @@ -15,12 +15,13 @@ #ifndef _ASM_TILE_TRAPS_H #define _ASM_TILE_TRAPS_H +#ifndef __ASSEMBLY__ #include <arch/chip.h> /* mm/fault.c */ void do_page_fault(struct pt_regs *, int fault_num, unsigned long address, unsigned long write); -#if CHIP_HAS_TILE_DMA() || CHIP_HAS_SN_PROC() +#if CHIP_HAS_TILE_DMA() void do_async_page_fault(struct pt_regs *); #endif @@ -64,7 +65,21 @@ void do_breakpoint(struct pt_regs *, int fault_num); #ifdef __tilegx__ +/* kernel/single_step.c */ void gx_singlestep_handle(struct pt_regs *, int fault_num); + +/* kernel/intvec_64.S */ +void fill_ra_stack(void); + +/* Handle unalign data fixup. */ +extern void do_unaligned(struct pt_regs *regs, int vecnum); +#endif + +#endif /* __ASSEMBLY__ */ + +#ifdef __tilegx__ +/* 128 byte JIT per unalign fixup. */ +#define UNALIGN_JIT_SHIFT 7 #endif -#endif /* _ASM_TILE_SYSCALLS_H */ +#endif /* _ASM_TILE_TRAPS_H */ diff --git a/arch/tile/include/asm/uaccess.h b/arch/tile/include/asm/uaccess.h index ef34d2caa5b..b6cde3209b9 100644 --- a/arch/tile/include/asm/uaccess.h +++ b/arch/tile/include/asm/uaccess.h @@ -114,45 +114,79 @@ struct exception_table_entry { extern int fixup_exception(struct pt_regs *regs); /* - * We return the __get_user_N function results in a structure, - * thus in r0 and r1. If "err" is zero, "val" is the result - * of the read; otherwise, "err" is -EFAULT. - * - * We rarely need 8-byte values on a 32-bit architecture, but - * we size the structure to accommodate. In practice, for the - * the smaller reads, we can zero the high word for free, and - * the caller will ignore it by virtue of casting anyway. + * Support macros for __get_user(). + * + * Implementation note: The "case 8" logic of casting to the type of + * the result of subtracting the value from itself is basically a way + * of keeping all integer types the same, but casting any pointers to + * ptrdiff_t, i.e. also an integer type. This way there are no + * questionable casts seen by the compiler on an ILP32 platform. + * + * Note that __get_user() and __put_user() assume proper alignment. */ -struct __get_user { - unsigned long long val; - int err; -}; -/* - * FIXME: we should express these as inline extended assembler, since - * they're fundamentally just a variable dereference and some - * supporting exception_table gunk. Note that (a la i386) we can - * extend the copy_to_user and copy_from_user routines to call into - * such extended assembler routines, though we will have to use a - * different return code in that case (1, 2, or 4, rather than -EFAULT). - */ -extern struct __get_user __get_user_1(const void __user *); -extern struct __get_user __get_user_2(const void __user *); -extern struct __get_user __get_user_4(const void __user *); -extern struct __get_user __get_user_8(const void __user *); -extern int __put_user_1(long, void __user *); -extern int __put_user_2(long, void __user *); -extern int __put_user_4(long, void __user *); -extern int __put_user_8(long long, void __user *); - -/* Unimplemented routines to cause linker failures */ -extern struct __get_user __get_user_bad(void); -extern int __put_user_bad(void); +#ifdef __LP64__ +#define _ASM_PTR ".quad" +#define _ASM_ALIGN ".align 8" +#else +#define _ASM_PTR ".long" +#define _ASM_ALIGN ".align 4" +#endif + +#define __get_user_asm(OP, x, ptr, ret) \ + asm volatile("1: {" #OP " %1, %2; movei %0, 0 }\n" \ + ".pushsection .fixup,\"ax\"\n" \ + "0: { movei %1, 0; movei %0, %3 }\n" \ + "j 9f\n" \ + ".section __ex_table,\"a\"\n" \ + _ASM_ALIGN "\n" \ + _ASM_PTR " 1b, 0b\n" \ + ".popsection\n" \ + "9:" \ + : "=r" (ret), "=r" (x) \ + : "r" (ptr), "i" (-EFAULT)) + +#ifdef __tilegx__ +#define __get_user_1(x, ptr, ret) __get_user_asm(ld1u, x, ptr, ret) +#define __get_user_2(x, ptr, ret) __get_user_asm(ld2u, x, ptr, ret) +#define __get_user_4(x, ptr, ret) __get_user_asm(ld4s, x, ptr, ret) +#define __get_user_8(x, ptr, ret) __get_user_asm(ld, x, ptr, ret) +#else +#define __get_user_1(x, ptr, ret) __get_user_asm(lb_u, x, ptr, ret) +#define __get_user_2(x, ptr, ret) __get_user_asm(lh_u, x, ptr, ret) +#define __get_user_4(x, ptr, ret) __get_user_asm(lw, x, ptr, ret) +#ifdef __LITTLE_ENDIAN +#define __lo32(a, b) a +#define __hi32(a, b) b +#else +#define __lo32(a, b) b +#define __hi32(a, b) a +#endif +#define __get_user_8(x, ptr, ret) \ + ({ \ + unsigned int __a, __b; \ + asm volatile("1: { lw %1, %3; addi %2, %3, 4 }\n" \ + "2: { lw %2, %2; movei %0, 0 }\n" \ + ".pushsection .fixup,\"ax\"\n" \ + "0: { movei %1, 0; movei %2, 0 }\n" \ + "{ movei %0, %4; j 9f }\n" \ + ".section __ex_table,\"a\"\n" \ + ".align 4\n" \ + ".word 1b, 0b\n" \ + ".word 2b, 0b\n" \ + ".popsection\n" \ + "9:" \ + : "=r" (ret), "=r" (__a), "=&r" (__b) \ + : "r" (ptr), "i" (-EFAULT)); \ + (x) = (__typeof(x))(__typeof((x)-(x))) \ + (((u64)__hi32(__a, __b) << 32) | \ + __lo32(__a, __b)); \ + }) +#endif + +extern int __get_user_bad(void) + __attribute__((warning("sizeof __get_user argument not 1, 2, 4 or 8"))); -/* - * Careful: we have to cast the result to the type of the pointer - * for sign reasons. - */ /** * __get_user: - Get a simple variable from user space, with less checking. * @x: Variable to store result. @@ -174,30 +208,64 @@ extern int __put_user_bad(void); * function. */ #define __get_user(x, ptr) \ -({ struct __get_user __ret; \ - __typeof__(*(ptr)) const __user *__gu_addr = (ptr); \ - __chk_user_ptr(__gu_addr); \ - switch (sizeof(*(__gu_addr))) { \ - case 1: \ - __ret = __get_user_1(__gu_addr); \ - break; \ - case 2: \ - __ret = __get_user_2(__gu_addr); \ - break; \ - case 4: \ - __ret = __get_user_4(__gu_addr); \ - break; \ - case 8: \ - __ret = __get_user_8(__gu_addr); \ - break; \ - default: \ - __ret = __get_user_bad(); \ - break; \ - } \ - (x) = (__typeof__(*__gu_addr)) (__typeof__(*__gu_addr - *__gu_addr)) \ - __ret.val; \ - __ret.err; \ -}) + ({ \ + int __ret; \ + __chk_user_ptr(ptr); \ + switch (sizeof(*(ptr))) { \ + case 1: __get_user_1(x, ptr, __ret); break; \ + case 2: __get_user_2(x, ptr, __ret); break; \ + case 4: __get_user_4(x, ptr, __ret); break; \ + case 8: __get_user_8(x, ptr, __ret); break; \ + default: __ret = __get_user_bad(); break; \ + } \ + __ret; \ + }) + +/* Support macros for __put_user(). */ + +#define __put_user_asm(OP, x, ptr, ret) \ + asm volatile("1: {" #OP " %1, %2; movei %0, 0 }\n" \ + ".pushsection .fixup,\"ax\"\n" \ + "0: { movei %0, %3; j 9f }\n" \ + ".section __ex_table,\"a\"\n" \ + _ASM_ALIGN "\n" \ + _ASM_PTR " 1b, 0b\n" \ + ".popsection\n" \ + "9:" \ + : "=r" (ret) \ + : "r" (ptr), "r" (x), "i" (-EFAULT)) + +#ifdef __tilegx__ +#define __put_user_1(x, ptr, ret) __put_user_asm(st1, x, ptr, ret) +#define __put_user_2(x, ptr, ret) __put_user_asm(st2, x, ptr, ret) +#define __put_user_4(x, ptr, ret) __put_user_asm(st4, x, ptr, ret) +#define __put_user_8(x, ptr, ret) __put_user_asm(st, x, ptr, ret) +#else +#define __put_user_1(x, ptr, ret) __put_user_asm(sb, x, ptr, ret) +#define __put_user_2(x, ptr, ret) __put_user_asm(sh, x, ptr, ret) +#define __put_user_4(x, ptr, ret) __put_user_asm(sw, x, ptr, ret) +#define __put_user_8(x, ptr, ret) \ + ({ \ + u64 __x = (__typeof((x)-(x)))(x); \ + int __lo = (int) __x, __hi = (int) (__x >> 32); \ + asm volatile("1: { sw %1, %2; addi %0, %1, 4 }\n" \ + "2: { sw %0, %3; movei %0, 0 }\n" \ + ".pushsection .fixup,\"ax\"\n" \ + "0: { movei %0, %4; j 9f }\n" \ + ".section __ex_table,\"a\"\n" \ + ".align 4\n" \ + ".word 1b, 0b\n" \ + ".word 2b, 0b\n" \ + ".popsection\n" \ + "9:" \ + : "=&r" (ret) \ + : "r" (ptr), "r" (__lo32(__lo, __hi)), \ + "r" (__hi32(__lo, __hi)), "i" (-EFAULT)); \ + }) +#endif + +extern int __put_user_bad(void) + __attribute__((warning("sizeof __put_user argument not 1, 2, 4 or 8"))); /** * __put_user: - Write a simple value into user space, with less checking. @@ -217,39 +285,19 @@ extern int __put_user_bad(void); * function. * * Returns zero on success, or -EFAULT on error. - * - * Implementation note: The "case 8" logic of casting to the type of - * the result of subtracting the value from itself is basically a way - * of keeping all integer types the same, but casting any pointers to - * ptrdiff_t, i.e. also an integer type. This way there are no - * questionable casts seen by the compiler on an ILP32 platform. */ #define __put_user(x, ptr) \ ({ \ - int __pu_err = 0; \ - __typeof__(*(ptr)) __user *__pu_addr = (ptr); \ - typeof(*__pu_addr) __pu_val = (x); \ - __chk_user_ptr(__pu_addr); \ - switch (sizeof(__pu_val)) { \ - case 1: \ - __pu_err = __put_user_1((long)__pu_val, __pu_addr); \ - break; \ - case 2: \ - __pu_err = __put_user_2((long)__pu_val, __pu_addr); \ - break; \ - case 4: \ - __pu_err = __put_user_4((long)__pu_val, __pu_addr); \ - break; \ - case 8: \ - __pu_err = \ - __put_user_8((__typeof__(__pu_val - __pu_val))__pu_val,\ - __pu_addr); \ - break; \ - default: \ - __pu_err = __put_user_bad(); \ - break; \ + int __ret; \ + __chk_user_ptr(ptr); \ + switch (sizeof(*(ptr))) { \ + case 1: __put_user_1(x, ptr, __ret); break; \ + case 2: __put_user_2(x, ptr, __ret); break; \ + case 4: __put_user_4(x, ptr, __ret); break; \ + case 8: __put_user_8(x, ptr, __ret); break; \ + default: __ret = __put_user_bad(); break; \ } \ - __pu_err; \ + __ret; \ }) /* @@ -353,7 +401,12 @@ _copy_from_user(void *to, const void __user *from, unsigned long n) return n; } -#ifdef CONFIG_DEBUG_COPY_FROM_USER +#ifdef CONFIG_DEBUG_STRICT_USER_COPY_CHECKS +/* + * There are still unprovable places in the generic code as of 2.6.34, so this + * option is not really compatible with -Werror, which is more useful in + * general. + */ extern void copy_from_user_overflow(void) __compiletime_warning("copy_from_user() size is not provably correct"); @@ -378,7 +431,7 @@ static inline unsigned long __must_check copy_from_user(void *to, /** * __copy_in_user() - copy data within user space, with less checking. * @to: Destination address, in user space. - * @from: Source address, in kernel space. + * @from: Source address, in user space. * @n: Number of bytes to copy. * * Context: User context only. This function may sleep. @@ -395,7 +448,7 @@ extern unsigned long __copy_in_user_inatomic( static inline unsigned long __must_check __copy_in_user(void __user *to, const void __user *from, unsigned long n) { - might_sleep(); + might_fault(); return __copy_in_user_inatomic(to, from, n); } @@ -520,37 +573,6 @@ static inline unsigned long __must_check flush_user( } /** - * inv_user: - Invalidate a block of memory in user space from cache. - * @mem: Destination address, in user space. - * @len: Number of bytes to invalidate. - * - * Returns number of bytes that could not be invalidated. - * On success, this will be zero. - * - * Note that on Tile64, the "inv" operation is in fact a - * "flush and invalidate", so cache write-backs will occur prior - * to the cache being marked invalid. - */ -extern unsigned long inv_user_asm(void __user *mem, unsigned long len); -static inline unsigned long __must_check __inv_user( - void __user *mem, unsigned long len) -{ - int retval; - - might_fault(); - retval = inv_user_asm(mem, len); - mb_incoherent(); - return retval; -} -static inline unsigned long __must_check inv_user( - void __user *mem, unsigned long len) -{ - if (access_ok(VERIFY_WRITE, mem, len)) - return __inv_user(mem, len); - return len; -} - -/** * finv_user: - Flush-inval a block of memory in user space from cache. * @mem: Destination address, in user space. * @len: Number of bytes to invalidate. diff --git a/arch/tile/include/asm/unaligned.h b/arch/tile/include/asm/unaligned.h index 137e2de5b10..5a58a0d1144 100644 --- a/arch/tile/include/asm/unaligned.h +++ b/arch/tile/include/asm/unaligned.h @@ -15,10 +15,29 @@ #ifndef _ASM_TILE_UNALIGNED_H #define _ASM_TILE_UNALIGNED_H -#include <linux/unaligned/le_struct.h> -#include <linux/unaligned/be_byteshift.h> -#include <linux/unaligned/generic.h> -#define get_unaligned __get_unaligned_le -#define put_unaligned __put_unaligned_le +/* + * We could implement faster get_unaligned_[be/le]64 using the ldna + * instruction on tilegx; however, we need to either copy all of the + * other generic functions to here (which is pretty ugly) or else + * modify both the generic code and other arch code to allow arch + * specific unaligned data access functions. Given these functions + * are not often called, we'll stick with the generic version. + */ +#include <asm-generic/unaligned.h> + +/* + * Is the kernel doing fixups of unaligned accesses? If <0, no kernel + * intervention occurs and SIGBUS is delivered with no data address + * info. If 0, the kernel single-steps the instruction to discover + * the data address to provide with the SIGBUS. If 1, the kernel does + * a fixup. + */ +extern int unaligned_fixup; + +/* Is the kernel printing on each unaligned fixup? */ +extern int unaligned_printk; + +/* Number of unaligned fixups performed */ +extern unsigned int unaligned_fixup_count; #endif /* _ASM_TILE_UNALIGNED_H */ diff --git a/arch/tile/include/asm/unistd.h b/arch/tile/include/asm/unistd.h index f70bf1c541f..940831fe9e9 100644 --- a/arch/tile/include/asm/unistd.h +++ b/arch/tile/include/asm/unistd.h @@ -11,37 +11,10 @@ * NON INFRINGEMENT. See the GNU General Public License for * more details. */ - -#if !defined(_ASM_TILE_UNISTD_H) || defined(__SYSCALL) -#define _ASM_TILE_UNISTD_H - -#if !defined(__LP64__) || defined(__SYSCALL_COMPAT) -/* Use the flavor of this syscall that matches the 32-bit API better. */ -#define __ARCH_WANT_SYNC_FILE_RANGE2 -#endif - -/* Use the standard ABI for syscalls. */ -#include <asm-generic/unistd.h> - -/* Additional Tilera-specific syscalls. */ -#define __NR_flush_cache (__NR_arch_specific_syscall + 1) -__SYSCALL(__NR_flush_cache, sys_flush_cache) - -#ifndef __tilegx__ -/* "Fast" syscalls provide atomic support for 32-bit chips. */ -#define __NR_FAST_cmpxchg -1 -#define __NR_FAST_atomic_update -2 -#define __NR_FAST_cmpxchg64 -3 -#define __NR_cmpxchg_badaddr (__NR_arch_specific_syscall + 0) -__SYSCALL(__NR_cmpxchg_badaddr, sys_cmpxchg_badaddr) -#endif - -#ifdef __KERNEL__ /* In compat mode, we use sys_llseek() for compat_sys_llseek(). */ #ifdef CONFIG_COMPAT #define __ARCH_WANT_SYS_LLSEEK #endif #define __ARCH_WANT_SYS_NEWFSTATAT -#endif - -#endif /* _ASM_TILE_UNISTD_H */ +#define __ARCH_WANT_SYS_CLONE +#include <uapi/asm/unistd.h> diff --git a/arch/tile/include/asm/vdso.h b/arch/tile/include/asm/vdso.h new file mode 100644 index 00000000000..9f6a78d665f --- /dev/null +++ b/arch/tile/include/asm/vdso.h @@ -0,0 +1,49 @@ +/* + * Copyright 2012 Tilera Corporation. All Rights Reserved. + * + * This program 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, version 2. + * + * This program 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, GOOD TITLE or + * NON INFRINGEMENT. See the GNU General Public License for + * more details. + */ + +#ifndef __TILE_VDSO_H__ +#define __TILE_VDSO_H__ + +#include <linux/types.h> + +/* + * Note about the vdso_data structure: + * + * NEVER USE THEM IN USERSPACE CODE DIRECTLY. The layout of the + * structure is supposed to be known only to the function in the vdso + * itself and may change without notice. + */ + +struct vdso_data { + __u64 tz_update_count; /* Timezone atomicity ctr */ + __u64 tb_update_count; /* Timebase atomicity ctr */ + __u64 xtime_tod_stamp; /* TOD clock for xtime */ + __u64 xtime_clock_sec; /* Kernel time second */ + __u64 xtime_clock_nsec; /* Kernel time nanosecond */ + __u64 wtom_clock_sec; /* Wall to monotonic clock second */ + __u64 wtom_clock_nsec; /* Wall to monotonic clock nanosecond */ + __u32 mult; /* Cycle to nanosecond multiplier */ + __u32 shift; /* Cycle to nanosecond divisor (power of two) */ + __u32 tz_minuteswest; /* Minutes west of Greenwich */ + __u32 tz_dsttime; /* Type of dst correction */ +}; + +extern struct vdso_data *vdso_data; + +/* __vdso_rt_sigreturn is defined with the addresses in the vdso page. */ +extern void __vdso_rt_sigreturn(void); + +extern int setup_vdso_pages(void); + +#endif /* __TILE_VDSO_H__ */ |