Merge bk://ppc.bkbits.net/for-linus-ppc64

into ppc970.osdl.org:/home/torvalds/v2.6/linux

arch/ppc64/kernel/Makefile

@@ -7,7 +7,7 @@ extra-y		:= head.o vmlinux.lds
 
 obj-y               :=	setup.o entry.o traps.o irq.o idle.o dma.o \
 			time.o process.o signal.o syscalls.o misc.o ptrace.o \
-			align.o semaphore.o bitops.o stab.o pacaData.o \
+			align.o semaphore.o bitops.o pacaData.o \
 			udbg.o binfmt_elf32.o sys_ppc32.o ioctl32.o \
 			ptrace32.o signal32.o rtc.o init_task.o \
 			lmb.o cputable.o cpu_setup_power4.o idle_power4.o \

arch/ppc64/mm/Makefile

@@ -4,6 +4,7 @@
 
 EXTRA_CFLAGS += -mno-minimal-toc
 
-obj-y := fault.o init.o imalloc.o hash_utils.o hash_low.o tlb.o slb_low.o slb.o
+obj-y := fault.o init.o imalloc.o hash_utils.o hash_low.o tlb.o \
+	slb_low.o slb.o stab.o
 obj-$(CONFIG_DISCONTIGMEM) += numa.o
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o

arch/ppc64/kernel/stab.c → arch/ppc64/mm/stab.c

@@ -20,74 +20,44 @@
 #include <asm/naca.h>
 #include <asm/cputable.h>
 
-static int make_ste(unsigned long stab, unsigned long esid,
-		    unsigned long vsid);
-
-void slb_initialize(void);
-
-/*
- * Build an entry for the base kernel segment and put it into
- * the segment table or SLB.  All other segment table or SLB
- * entries are faulted in.
- */
-void stab_initialize(unsigned long stab)
-{
-	unsigned long vsid = get_kernel_vsid(KERNELBASE);
-
-	if (cur_cpu_spec->cpu_features & CPU_FTR_SLB) {
-		slb_initialize();
-	} else {
-		asm volatile("isync; slbia; isync":::"memory");
-		make_ste(stab, GET_ESID(KERNELBASE), vsid);
-
-		/* Order update */
-		asm volatile("sync":::"memory"); 
-	}
-}
-
 /* Both the segment table and SLB code uses the following cache */
 #define NR_STAB_CACHE_ENTRIES 8
 DEFINE_PER_CPU(long, stab_cache_ptr);
 DEFINE_PER_CPU(long, stab_cache[NR_STAB_CACHE_ENTRIES]);
 
-/*
- * Segment table stuff
- */
-
 /*
  * Create a segment table entry for the given esid/vsid pair.
  */
 static int make_ste(unsigned long stab, unsigned long esid, unsigned long vsid)
 {
+	unsigned long esid_data, vsid_data;
 	unsigned long entry, group, old_esid, castout_entry, i;
 	unsigned int global_entry;
-	STE *ste, *castout_ste;
-	unsigned long kernel_segment = (REGION_ID(esid << SID_SHIFT) != 
-					USER_REGION_ID);
+	struct stab_entry *ste, *castout_ste;
+	unsigned long kernel_segment = (esid << SID_SHIFT) >= KERNELBASE;
+
+	vsid_data = vsid << STE_VSID_SHIFT;
+	esid_data = esid << SID_SHIFT | STE_ESID_KP | STE_ESID_V;
+	if (! kernel_segment)
+		esid_data |= STE_ESID_KS;
 
 	/* Search the primary group first. */
 	global_entry = (esid & 0x1f) << 3;
-	ste = (STE *)(stab | ((esid & 0x1f) << 7)); 
+	ste = (struct stab_entry *)(stab | ((esid & 0x1f) << 7));
 
 	/* Find an empty entry, if one exists. */
 	for (group = 0; group < 2; group++) {
 		for (entry = 0; entry < 8; entry++, ste++) {
-			if (!(ste->dw0.dw0.v)) {
-				ste->dw0.dword0 = 0;
-				ste->dw1.dword1 = 0;
-				ste->dw1.dw1.vsid = vsid;
-				ste->dw0.dw0.esid = esid;
-				ste->dw0.dw0.kp = 1;
-				if (!kernel_segment)
-					ste->dw0.dw0.ks = 1;
+			if (!(ste->esid_data & STE_ESID_V)) {
+				ste->vsid_data = vsid_data;
 				asm volatile("eieio":::"memory");
-				ste->dw0.dw0.v = 1;
+				ste->esid_data = esid_data;
 				return (global_entry | entry);
 			}
 		}
 		/* Now search the secondary group. */
 		global_entry = ((~esid) & 0x1f) << 3;
-		ste = (STE *)(stab | (((~esid) & 0x1f) << 7)); 
+		ste = (struct stab_entry *)(stab | (((~esid) & 0x1f) << 7));
 	}
 
 	/*
@@ -98,16 +68,16 @@ static int make_ste(unsigned long stab, unsigned long esid, unsigned long vsid)
 	for (i = 0; i < 16; i++) {
 		if (castout_entry < 8) {
 			global_entry = (esid & 0x1f) << 3;
-			ste = (STE *)(stab | ((esid & 0x1f) << 7)); 
+			ste = (struct stab_entry *)(stab | ((esid & 0x1f) << 7));
 			castout_ste = ste + castout_entry;
 		} else {
 			global_entry = ((~esid) & 0x1f) << 3;
-			ste = (STE *)(stab | (((~esid) & 0x1f) << 7)); 
+			ste = (struct stab_entry *)(stab | (((~esid) & 0x1f) << 7));
 			castout_ste = ste + (castout_entry - 8);
 		}
 
 		/* Dont cast out the first kernel segment */
-		if (castout_ste->dw0.dw0.esid != GET_ESID(KERNELBASE))
+		if ((castout_ste->esid_data & ESID_MASK) != KERNELBASE)
 			break;
 
 		castout_entry = (castout_entry + 1) & 0xf;
@@ -120,19 +90,15 @@ static int make_ste(unsigned long stab, unsigned long esid, unsigned long vsid)
 	/* Force previous translations to complete. DRENG */
 	asm volatile("isync" : : : "memory");
 
-	castout_ste->dw0.dw0.v = 0;
+	old_esid = castout_ste->esid_data >> SID_SHIFT;
+	castout_ste->esid_data = 0;		/* Invalidate old entry */
+
 	asm volatile("sync" : : : "memory");    /* Order update */
 
-	castout_ste->dw0.dword0 = 0;
-	castout_ste->dw1.dword1 = 0;
-	castout_ste->dw1.dw1.vsid = vsid;
-	old_esid = castout_ste->dw0.dw0.esid;
-	castout_ste->dw0.dw0.esid = esid;
-	castout_ste->dw0.dw0.kp = 1;
-	if (!kernel_segment)
-		castout_ste->dw0.dw0.ks = 1;
+	castout_ste->vsid_data = vsid_data;
 	asm volatile("eieio" : : : "memory");   /* Order update */
-	castout_ste->dw0.dw0.v  = 1;
+	castout_ste->esid_data = esid_data;
+
 	asm volatile("slbie  %0" : : "r" (old_esid << SID_SHIFT));
 	/* Ensure completion of slbie */
 	asm volatile("sync" : : : "memory");
@@ -140,109 +106,64 @@ static int make_ste(unsigned long stab, unsigned long esid, unsigned long vsid)
 	return (global_entry | (castout_entry & 0x7));
 }
 
-static inline void __ste_allocate(unsigned long esid, unsigned long vsid)
-{
-	unsigned char stab_entry;
-	unsigned long offset;
-	int region_id = REGION_ID(esid << SID_SHIFT);
-
-	stab_entry = make_ste(get_paca()->stab_addr, esid, vsid);
-
-	if (region_id != USER_REGION_ID)
-		return;
-
-	offset = __get_cpu_var(stab_cache_ptr);
-	if (offset < NR_STAB_CACHE_ENTRIES)
-		__get_cpu_var(stab_cache[offset++]) = stab_entry;
-	else
-		offset = NR_STAB_CACHE_ENTRIES+1;
-	__get_cpu_var(stab_cache_ptr) = offset;
-}
-
 /*
- * Allocate a segment table entry for the given ea.
+ * Allocate a segment table entry for the given ea and mm
  */
-int ste_allocate(unsigned long ea)
+static int __ste_allocate(unsigned long ea, struct mm_struct *mm)
 {
-	unsigned long vsid, esid;
-	mm_context_t context;
+	unsigned long vsid;
+	unsigned char stab_entry;
+	unsigned long offset;
 
 	/* Check for invalid effective addresses. */
 	if (!IS_VALID_EA(ea))
 		return 1;
 
 	/* Kernel or user address? */
-	if (REGION_ID(ea) >= KERNEL_REGION_ID) {
+	if (ea >= KERNELBASE) {
 		vsid = get_kernel_vsid(ea);
-		context = KERNEL_CONTEXT(ea);
 	} else {
-		if (!current->mm)
+		if (! mm)
 			return 1;
 
-		context = current->mm->context;
-		vsid = get_vsid(context.id, ea);
+		vsid = get_vsid(mm->context.id, ea);
 	}
 
-	esid = GET_ESID(ea);
-	__ste_allocate(esid, vsid);
+	stab_entry = make_ste(get_paca()->stab_addr, GET_ESID(ea), vsid);
+
+	if (ea < KERNELBASE) {
+		offset = __get_cpu_var(stab_cache_ptr);
+		if (offset < NR_STAB_CACHE_ENTRIES)
+			__get_cpu_var(stab_cache[offset++]) = stab_entry;
+		else
+			offset = NR_STAB_CACHE_ENTRIES+1;
+		__get_cpu_var(stab_cache_ptr) = offset;
+
 		/* Order update */
 		asm volatile("sync":::"memory");
+	}
 
 	return 0;
 }
 
+int ste_allocate(unsigned long ea)
+{
+	return __ste_allocate(ea, current->mm);
+}
+
 /*
- * preload some userspace segments into the segment table.
+ * Do the segment table work for a context switch: flush all user
+ * entries from the table, then preload some probably useful entries
+ * for the new task
  */
-static void preload_stab(struct task_struct *tsk, struct mm_struct *mm)
+void switch_stab(struct task_struct *tsk, struct mm_struct *mm)
 {
+	struct stab_entry *stab = (struct stab_entry *) get_paca()->stab_addr;
+	struct stab_entry *ste;
+	unsigned long offset = __get_cpu_var(stab_cache_ptr);
 	unsigned long pc = KSTK_EIP(tsk);
 	unsigned long stack = KSTK_ESP(tsk);
 	unsigned long unmapped_base;
-	unsigned long pc_esid = GET_ESID(pc);
-	unsigned long stack_esid = GET_ESID(stack);
-	unsigned long unmapped_base_esid;
-	unsigned long vsid;
-
-	if (test_tsk_thread_flag(tsk, TIF_32BIT))
-		unmapped_base = TASK_UNMAPPED_BASE_USER32;
-	else
-		unmapped_base = TASK_UNMAPPED_BASE_USER64;
-
-	unmapped_base_esid = GET_ESID(unmapped_base);
-
-	if (!IS_VALID_EA(pc) || (REGION_ID(pc) >= KERNEL_REGION_ID))
-		return;
-	vsid = get_vsid(mm->context.id, pc);
-	__ste_allocate(pc_esid, vsid);
-
-	if (pc_esid == stack_esid)
-		return;
-
-	if (!IS_VALID_EA(stack) || (REGION_ID(stack) >= KERNEL_REGION_ID))
-		return;
-	vsid = get_vsid(mm->context.id, stack);
-	__ste_allocate(stack_esid, vsid);
-
-	if (pc_esid == unmapped_base_esid || stack_esid == unmapped_base_esid)
-		return;
-
-	if (!IS_VALID_EA(unmapped_base) ||
-	    (REGION_ID(unmapped_base) >= KERNEL_REGION_ID))
-		return;
-	vsid = get_vsid(mm->context.id, unmapped_base);
-	__ste_allocate(unmapped_base_esid, vsid);
-
-	/* Order update */
-	asm volatile("sync" : : : "memory");
-}
-
-/* Flush all user entries from the segment table of the current processor. */
-void flush_stab(struct task_struct *tsk, struct mm_struct *mm)
-{
-	STE *stab = (STE *) get_paca()->stab_addr;
-	STE *ste;
-	unsigned long offset = __get_cpu_var(stab_cache_ptr);
 
 	/* Force previous translations to complete. DRENG */
 	asm volatile("isync" : : : "memory");
@@ -252,7 +173,7 @@ void flush_stab(struct task_struct *tsk, struct mm_struct *mm)
 
 		for (i = 0; i < offset; i++) {
 			ste = stab + __get_cpu_var(stab_cache[i]);
-			ste->dw0.dw0.v = 0;
+			ste->esid_data = 0; /* invalidate entry */
 		}
 	} else {
 		unsigned long entry;
@@ -263,12 +184,12 @@ void flush_stab(struct task_struct *tsk, struct mm_struct *mm)
 		/* Never flush the first entry. */
 		ste += 1;
 		for (entry = 1;
-		     entry < (PAGE_SIZE / sizeof(STE));
+		     entry < (PAGE_SIZE / sizeof(struct stab_entry));
 		     entry++, ste++) {
 			unsigned long ea;
-			ea = ste->dw0.dw0.esid << SID_SHIFT;
+			ea = ste->esid_data & ESID_MASK;
 			if (ea < KERNELBASE) {
-				ste->dw0.dw0.v = 0;
+				ste->esid_data = 0;
 			}
 		}
 	}
@@ -277,5 +198,47 @@ void flush_stab(struct task_struct *tsk, struct mm_struct *mm)
 
 	__get_cpu_var(stab_cache_ptr) = 0;
 
-	preload_stab(tsk, mm);
+	/* Now preload some entries for the new task */
+	if (test_tsk_thread_flag(tsk, TIF_32BIT))
+		unmapped_base = TASK_UNMAPPED_BASE_USER32;
+	else
+		unmapped_base = TASK_UNMAPPED_BASE_USER64;
+
+	__ste_allocate(pc, mm);
+
+	if (GET_ESID(pc) == GET_ESID(stack))
+		return;
+
+	__ste_allocate(stack, mm);
+
+	if ((GET_ESID(pc) == GET_ESID(unmapped_base))
+	    || (GET_ESID(stack) == GET_ESID(unmapped_base)))
+		return;
+
+	__ste_allocate(unmapped_base, mm);
+
+	/* Order update */
+	asm volatile("sync" : : : "memory");
+}
+
+extern void slb_initialize(void);
+
+/*
+ * Build an entry for the base kernel segment and put it into
+ * the segment table or SLB.  All other segment table or SLB
+ * entries are faulted in.
+ */
+void stab_initialize(unsigned long stab)
+{
+	unsigned long vsid = get_kernel_vsid(KERNELBASE);
+
+	if (cur_cpu_spec->cpu_features & CPU_FTR_SLB) {
+		slb_initialize();
+	} else {
+		asm volatile("isync; slbia; isync":::"memory");
+		make_ste(stab, GET_ESID(KERNELBASE), vsid);
+
+		/* Order update */
+		asm volatile("sync":::"memory");
+	}
 }

include/asm-ppc64/mmu.h

@@ -37,33 +37,17 @@ typedef struct {
 		mm_context_t ctx = { .id = REGION_ID(ea), KERNEL_LOW_HPAGES}; \
 		ctx; })
 
-typedef struct {
-	unsigned long esid: 36; /* Effective segment ID */
-	unsigned long resv0:20; /* Reserved */
-	unsigned long v:     1; /* Entry valid (v=1) or invalid */
-	unsigned long resv1: 1; /* Reserved */
-	unsigned long ks:    1; /* Supervisor (privileged) state storage key */
-	unsigned long kp:    1; /* Problem state storage key */
-	unsigned long n:     1; /* No-execute if n=1 */
-	unsigned long resv2: 3; /* padding to a 64b boundary */
-} ste_dword0;
+#define STE_ESID_V	0x80
+#define STE_ESID_KS	0x20
+#define STE_ESID_KP	0x10
+#define STE_ESID_N	0x08
 
-typedef struct {
-	unsigned long vsid: 52; /* Virtual segment ID */
-	unsigned long resv0:12; /* Padding to a 64b boundary */
-} ste_dword1;
+#define STE_VSID_SHIFT	12
 
-typedef struct _STE {
-	union {
-		unsigned long dword0;
-		ste_dword0    dw0;
-	} dw0;
-
-	union {
-		unsigned long dword1;
-		ste_dword1    dw1;
-	} dw1;
-} STE;
+struct stab_entry {
+	unsigned long esid_data;
+	unsigned long vsid_data;
+};
 
 /* Hardware Page Table Entry */
 

include/asm-ppc64/mmu_context.h

@@ -135,7 +135,7 @@ destroy_context(struct mm_struct *mm)
 	spin_unlock_irqrestore(&mmu_context_queue.lock, flags);
 }
 
-extern void flush_stab(struct task_struct *tsk, struct mm_struct *mm);
+extern void switch_stab(struct task_struct *tsk, struct mm_struct *mm);
 extern void switch_slb(struct task_struct *tsk, struct mm_struct *mm);
 
 /*
@@ -163,7 +163,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	if (cur_cpu_spec->cpu_features & CPU_FTR_SLB)
 		switch_slb(tsk, next);
 	else
-		flush_stab(tsk, next);
+		switch_stab(tsk, next);
 }
 
 #define deactivate_mm(tsk,mm)	do { } while (0)