Import 2.3.11pre4

arch/alpha/kernel/Makefile

@@ -105,7 +105,7 @@ O_OBJS   += sys_takara.o
 endif
 
 # Device support
-ifdef CONFIG_ALPHA_MIATA
+ifneq ($(CONFIG_ALPHA_MIATA)$(CONFIG_ALPHA_DP264),)
 O_OBJS   += es1888.o
 endif
 ifneq ($(CONFIG_ALPHA_SX164)$(CONFIG_ALPHA_MIATA)$(CONFIG_ALPHA_DP264),)

arch/alpha/kernel/alpha_ksyms.c

@@ -48,8 +48,6 @@ extern void __divqu (void);
 extern void __remqu (void);
 
 EXPORT_SYMBOL(alpha_mv);
-EXPORT_SYMBOL(local_bh_count);
-EXPORT_SYMBOL(local_irq_count);
 EXPORT_SYMBOL(enable_irq);
 EXPORT_SYMBOL(disable_irq);
 EXPORT_SYMBOL(disable_irq_nosync);
@@ -178,6 +176,9 @@ EXPORT_SYMBOL(debug_spin_trylock);
 EXPORT_SYMBOL(write_lock);
 EXPORT_SYMBOL(read_lock);
 #endif
+#else /* __SMP__ */
+EXPORT_SYMBOL(__local_bh_count);
+EXPORT_SYMBOL(__local_irq_count);
 #endif /* __SMP__ */
 
 /*

arch/alpha/kernel/core_pyxis.c

@@ -425,19 +425,24 @@ pyxis_native_window_setup(void)
 {
 	/*
 	 * Set up the PCI->physical memory translation windows.
-	 * For now, windows 1,2 and 3 are disabled.  In the future, we may
+	 * For now, windows 2 and 3 are disabled.  In the future, we may
 	 * want to use them to do scatter/gather DMA. 
 	 *
-	 * Window 0 goes at 1 GB and is 1 GB large.
+	 * Window 0 goes at 2 GB and is 1 GB large.
+	 * Window 1 goes at 3 GB and is 1 GB large.
 	 */
 
-	*(vuip)PYXIS_W0_BASE = 1U | (PYXIS_DMA_WIN_BASE_DEFAULT & 0xfff00000U);
+	*(vuip)PYXIS_W0_BASE = PYXIS_DMA_WIN_BASE_DEFAULT | 1UL;
 	*(vuip)PYXIS_W0_MASK = (PYXIS_DMA_WIN_SIZE_DEFAULT - 1) & 0xfff00000U;
 	*(vuip)PYXIS_T0_BASE = 0;
 
-	*(vuip)PYXIS_W1_BASE = 0x0 ;
-	*(vuip)PYXIS_W2_BASE = 0x0 ;
-	*(vuip)PYXIS_W3_BASE = 0x0 ;
+	*(vuip)PYXIS_W1_BASE = (PYXIS_DMA_WIN_BASE_DEFAULT +
+				PYXIS_DMA_WIN_SIZE_DEFAULT) | 1U;
+	*(vuip)PYXIS_W1_MASK = (PYXIS_DMA_WIN_SIZE_DEFAULT - 1) & 0xfff00000U;
+	*(vuip)PYXIS_T1_BASE = PYXIS_DMA_WIN_SIZE_DEFAULT;
+
+	*(vuip)PYXIS_W2_BASE = 0x0;
+	*(vuip)PYXIS_W3_BASE = 0x0;
 	mb();
 }
 

arch/alpha/kernel/es1888.c

@@ -32,6 +32,7 @@ es1888_init(void)
 		continue;
 	inb(0x022a);			/* pause */
 	outb(0xc6, 0x022c);		/* enable extended mode */
+	inb(0x022a);			/* pause, also forces the write */
 	while (inb(0x022c) & 0x80)	/* wait for bit 7 to deassert */
 		continue;
 	outb(0xb1, 0x022c);		/* setup for write to Interrupt CR */
@@ -44,4 +45,5 @@ es1888_init(void)
 	while (inb(0x022c) & 0x80)	/* wait for bit 7 to deassert */
 		continue;
 	outb(0x18, 0x022c);		/* set DMA channel 1 */
+	inb(0x022c);			/* force the write */
 }

arch/alpha/kernel/irq.c

@@ -35,9 +35,12 @@
 #define vulp	volatile unsigned long *
 #define vuip	volatile unsigned int *
 
-unsigned int local_irq_count[NR_CPUS];
-unsigned int local_bh_count[NR_CPUS];
-unsigned long hardirq_no[NR_CPUS];
+/* Only uniprocessor needs this IRQ/BH locking depth, on SMP it lives
+   in the per-cpu structure for cache reasons.  */
+#ifndef __SMP__
+int __local_irq_count;
+int __local_bh_count;
+#endif
 
 #if NR_IRQS > 64
 #  error Unable to handle more than 64 irq levels.
@@ -384,6 +387,8 @@ static void *previous_irqholder = NULL;
 
 static void show(char * str, void *where);
 
+#define SYNC_OTHER_CPUS(x)	udelay((x)+1);
+
 static inline void
 wait_on_irq(int cpu, void *where)
 {
@@ -397,8 +402,8 @@ wait_on_irq(int cpu, void *where)
 		 * already executing in one..
 		 */
 		if (!atomic_read(&global_irq_count)) {
-			if (local_bh_count[cpu] ||
-			    !atomic_read(&global_bh_count))
+			if (local_bh_count(cpu)
+			    || !atomic_read(&global_bh_count))
 				break;
 		}
 
@@ -412,19 +417,15 @@ wait_on_irq(int cpu, void *where)
 				count = MAXCOUNT;
 			}
 			__sti();
-#if 0
-			SYNC_OTHER_CORES(cpu);
-#else
-			udelay(cpu+1);
-#endif
+			SYNC_OTHER_CPUS(cpu);
 			__cli();
 
 			if (atomic_read(&global_irq_count))
 				continue;
-			if (global_irq_lock.lock)
+			if (spin_is_locked(&global_irq_lock))
 				continue;
-			if (!local_bh_count[cpu] &&
-			    atomic_read(&global_bh_count))
+			if (!local_bh_count(cpu)
+			    && atomic_read(&global_bh_count))
 				continue;
 			if (spin_trylock(&global_irq_lock))
 				break;
@@ -469,14 +470,14 @@ get_irqlock(int cpu, void* where)
 void
 __global_cli(void)
 {
-	int cpu;
+	int cpu = smp_processor_id();
 	void *where = __builtin_return_address(0);
 
 	/*
 	 * Maximize ipl.  If ipl was previously 0 and if this thread
 	 * is not in an irq, then take global_irq_lock.
 	 */
-	if ((swpipl(7) == 0) && !local_irq_count[cpu = smp_processor_id()])
+	if (swpipl(7) == 0 && !local_irq_count(cpu))
 		get_irqlock(cpu, where);
 }
 
@@ -485,9 +486,8 @@ __global_sti(void)
 {
         int cpu = smp_processor_id();
 
-        if (!local_irq_count[cpu]) {
+        if (!local_irq_count(cpu))
 		release_irqlock(cpu);
-	}
 	__sti();
 }
 
@@ -512,7 +512,7 @@ __global_save_flags(void)
         retval = 2 + local_enabled;
 
         /* Check for global flags if we're not in an interrupt.  */
-        if (!local_irq_count[cpu]) {
+        if (!local_irq_count(cpu)) {
                 if (local_enabled)
                         retval = 1;
                 if (global_irq_holder == cpu)
@@ -550,7 +550,7 @@ __global_restore_flags(unsigned long flags)
 #define STUCK							\
   if (!--stuck) {						\
     printk("irq_enter stuck (irq=%d, cpu=%d, global=%d)\n",	\
-	   irq, cpu,global_irq_holder);				\
+	   irq, cpu, global_irq_holder);			\
     stuck = INIT_STUCK;						\
   }
 
@@ -566,11 +566,11 @@ irq_enter(int cpu, int irq)
 
 	hardirq_enter(cpu, irq);
 	barrier();
-	while (global_irq_lock.lock) {
+	while (spin_is_locked(&global_irq_lock)) {
 		if (cpu == global_irq_holder) {
-			int globl_locked = global_irq_lock.lock;
+			int globl_locked = spin_is_locked(&global_irq_lock);
 			int globl_icount = atomic_read(&global_irq_count);
-			int local_count = local_irq_count[cpu];
+			int local_count = local_irq_count(cpu);
 
 			/* It is very important that we load the state
 			   variables before we do the first call to
@@ -609,19 +609,16 @@ show(char * str, void *where)
 #endif
         int cpu = smp_processor_id();
 
-	int global_count = atomic_read(&global_irq_count);
-        int local_count0 = local_irq_count[0];
-        int local_count1 = local_irq_count[1];
-        long hardirq_no0 = hardirq_no[0];
-        long hardirq_no1 = hardirq_no[1];
-
         printk("\n%s, CPU %d: %p\n", str, cpu, where);
-        printk("irq:  %d [%d(0x%016lx) %d(0x%016lx)]\n", global_count,
-               local_count0, hardirq_no0, local_count1, hardirq_no1);
+        printk("irq:  %d [%d %d]\n",
+	       atomic_read(&global_irq_count),
+               cpu_data[0].irq_count,
+	       cpu_data[1].irq_count);
 
         printk("bh:   %d [%d %d]\n",
-	       atomic_read(&global_bh_count), local_bh_count[0],
-	       local_bh_count[1]);
+	       atomic_read(&global_bh_count),
+	       cpu_data[0].bh_count,
+	       cpu_data[1].bh_count);
 #if 0
         stack = (unsigned long *) &str;
         for (i = 40; i ; i--) {
@@ -644,6 +641,7 @@ wait_on_bh(void)
                         count = ~0;
                 }
                 /* nothing .. wait for the other bh's to go away */
+		barrier();
         } while (atomic_read(&global_bh_count) != 0);
 }
 
@@ -658,12 +656,8 @@ wait_on_bh(void)
 void
 synchronize_bh(void)
 {
-	if (atomic_read(&global_bh_count)) {
-		int cpu = smp_processor_id();
-                if (!local_irq_count[cpu] && !local_bh_count[cpu]) {
-			wait_on_bh();
-		}
-        }
+	if (atomic_read(&global_bh_count) && !in_interrupt())
+		wait_on_bh();
 }
 
 /*
@@ -680,6 +674,8 @@ synchronize_bh(void)
 void
 synchronize_irq(void)
 {
+#if 0
+	/* Joe's version.  */
 	int cpu = smp_processor_id();
 	int local_count;
 	int global_count;
@@ -688,7 +684,7 @@ synchronize_irq(void)
 
 	mb();
 	do {
-		local_count = local_irq_count[cpu];
+		local_count = local_irq_count(cpu);
 		global_count = atomic_read(&global_irq_count);
 		if (DEBUG_SYNCHRONIZE_IRQ && (--countdown == 0)) {
 			printk("%d:%d/%d\n", cpu, local_count, global_count);
@@ -696,12 +692,19 @@ synchronize_irq(void)
 			break;
 		}
 	} while (global_count != local_count);
+#else
+	/* Jay's version.  */
+	if (atomic_read(&global_irq_count)) {
+		cli();
+		sti();
+	}
+#endif
 }
 
 #else /* !__SMP__ */
 
-#define irq_enter(cpu, irq)	(++local_irq_count[cpu])
-#define irq_exit(cpu, irq)	(--local_irq_count[cpu])
+#define irq_enter(cpu, irq)	(++local_irq_count(cpu))
+#define irq_exit(cpu, irq)	(--local_irq_count(cpu))
 
 #endif /* __SMP__ */
 
@@ -868,31 +871,23 @@ do_entInt(unsigned long type, unsigned long vector, unsigned long la_ptr,
 	  unsigned long a3, unsigned long a4, unsigned long a5,
 	  struct pt_regs regs)
 {
-	unsigned long flags;
-
 	switch (type) {
 	case 0:
 #ifdef __SMP__
-		__save_and_cli(flags);
 		handle_ipi(&regs);
-		__restore_flags(flags);
 		return;
 #else
 		printk("Interprocessor interrupt? You must be kidding\n");
 #endif
 		break;
 	case 1:
-		__save_and_cli(flags);
 		handle_irq(RTC_IRQ, -1, &regs);
-		__restore_flags(flags);
 		return;
 	case 2:
 		alpha_mv.machine_check(vector, la_ptr, &regs);
 		return;
 	case 3:
-		__save_and_cli(flags);
 		alpha_mv.device_interrupt(vector, &regs);
-		__restore_flags(flags);
 		return;
 	case 4:
 		perf_irq(vector, &regs);

arch/alpha/kernel/irq.h

@@ -31,3 +31,19 @@ extern void handle_irq(int irq, int ack, struct pt_regs * regs);
 #define TIMER_IRQ  RTC_IRQ		 /* timer is the rtc */
 #endif
 
+extern char _stext;
+static inline void alpha_do_profile (unsigned long pc)
+{
+	if (prof_buffer && current->pid) {
+		pc -= (unsigned long) &_stext;
+		pc >>= prof_shift;
+		/*
+		 * Don't ignore out-of-bounds PC values silently,
+		 * put them into the last histogram slot, so if
+		 * present, they will show up as a sharp peak.
+		 */
+		if (pc > prof_len - 1)
+			pc = prof_len - 1;
+		atomic_inc((atomic_t *)&prof_buffer[pc]);
+	}
+}

arch/alpha/kernel/osf_sys.c

@@ -927,6 +927,7 @@ asmlinkage unsigned long osf_getsysinfo(unsigned long op, void *buffer,
 			return -EINVAL;
 		cpu = (struct percpu_struct*)
 		  ((char*)hwrpb + hwrpb->processor_offset);
+		w = cpu->type;
 		if (put_user(w, (unsigned long *)buffer))
 			return -EFAULT;
 		return 1;

arch/alpha/kernel/process.c

@@ -76,7 +76,7 @@ sys_sethae(unsigned long hae, unsigned long a1, unsigned long a2,
 }
 
 #ifdef __SMP__
-void
+int
 cpu_idle(void *unused)
 {
 	/* An endless idle loop with no priority at all.  */
@@ -329,7 +329,6 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 	p->tss.ksp = (unsigned long) childstack;
 	p->tss.pal_flags = 1;	/* set FEN, clear everything else */
 	p->tss.flags = current->tss.flags;
-	p->tss.mm_context = p->tss.asn = 0;
 
 	return 0;
 }

arch/alpha/kernel/proto.h

@@ -193,7 +193,7 @@ extern void entDbg(void);
 
 /* process.c */
 extern void generic_kill_arch (int mode, char *reboot_cmd);
-extern void cpu_idle(void *) __attribute__((noreturn));
+extern int cpu_idle(void *) __attribute__((noreturn));
 
 /* ptrace.c */
 extern int ptrace_set_bpt (struct task_struct *child);

arch/alpha/kernel/ptrace.c

@@ -231,7 +231,6 @@ sys_ptrace(long request, long pid, long addr, long data,
 	   int a4, int a5, struct pt_regs regs)
 {
 	struct task_struct *child;
-	unsigned long tmp;
 	long ret;
 
 	lock_kernel();

arch/alpha/kernel/smp.c

@@ -97,6 +97,8 @@ smp_store_cpu_info(int cpuid)
 	cpu_data[cpuid].loops_per_sec = loops_per_sec;
 	cpu_data[cpuid].last_asn
 	  = (cpuid << WIDTH_HARDWARE_ASN) + ASN_FIRST_VERSION;
+	cpu_data[cpuid].irq_count = 0;
+	cpu_data[cpuid].bh_count = 0;
 }
 
 /*
@@ -107,12 +109,6 @@ smp_setup_percpu_timer(int cpuid)
 {
 	cpu_data[cpuid].prof_counter = 1;
 	cpu_data[cpuid].prof_multiplier = 1;
-
-#ifdef NOT_YET_PROFILING
-	load_profile_irq(mid_xlate[cpu], lvl14_resolution);
-	if (cpu == smp_boot_cpuid)
-		enable_pil_irq(14);
-#endif
 }
 
 /*
@@ -586,14 +582,12 @@ void
 smp_percpu_timer_interrupt(struct pt_regs *regs)
 {
 	int cpu = smp_processor_id();
-	int user = user_mode(regs);
+	unsigned long user = user_mode(regs);
 	struct cpuinfo_alpha *data = &cpu_data[cpu];
 
-#ifdef NOT_YET_PROFILING
-	clear_profile_irq(mid_xlate[cpu]);
+	/* Record kernel PC.  */
 	if (!user)
 		alpha_do_profile(regs->pc);
-#endif
 
 	if (!--data->prof_counter) {
 		/* We need to make like a normal interrupt -- otherwise
@@ -630,28 +624,7 @@ smp_percpu_timer_interrupt(struct pt_regs *regs)
 int __init
 setup_profiling_timer(unsigned int multiplier)
 {
-#ifdef NOT_YET_PROFILING
-	int i;
-	unsigned long flags;
-
-	/* Prevent level14 ticker IRQ flooding. */
-	if((!multiplier) || (lvl14_resolution / multiplier) < 500)
-	        return -EINVAL;
-
-	save_and_cli(flags);
-	for (i = 0; i < NR_CPUS; i++) {
-	        if (cpu_present_mask & (1L << i)) {
-	                load_profile_irq(mid_xlate[i],
-					 lvl14_resolution / multiplier);
-	                prof_multiplier[i] = multiplier;
-	        }
-	}
-	restore_flags(flags);
-
-	return 0;
-#else
 	return -EINVAL;
-#endif
 }
 
 
@@ -893,9 +866,11 @@ ipi_flush_tlb_mm(void *x)
 void
 flush_tlb_mm(struct mm_struct *mm)
 {
-	if (mm == current->mm)
+	if (mm == current->mm) {
 		flush_tlb_current(mm);
-	else
+		if (atomic_read(&mm->count) == 1)
+			return;
+	} else
 		flush_tlb_other(mm);
 
 	if (smp_call_function(ipi_flush_tlb_mm, mm, 1, 1)) {
@@ -923,15 +898,17 @@ flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 	struct flush_tlb_page_struct data;
 	struct mm_struct *mm = vma->vm_mm;
 
+	if (mm == current->mm) {
+		flush_tlb_current_page(mm, vma, addr);
+		if (atomic_read(&mm->count) == 1)
+			return;
+	} else
+		flush_tlb_other(mm);
+
 	data.vma = vma;
 	data.mm = mm;
 	data.addr = addr;
 
-	if (mm == current->mm)
-		flush_tlb_current_page(mm, vma, addr);
-	else
-		flush_tlb_other(mm);
-	
 	if (smp_call_function(ipi_flush_tlb_page, &data, 1, 1)) {
 		printk(KERN_CRIT "flush_tlb_page: timed out\n");
 	}

arch/alpha/kernel/sys_dp264.c

@@ -393,8 +393,8 @@ monet_pci_fixup(void)
 {
 	layout_all_busses(DEFAULT_IO_BASE, DEFAULT_MEM_BASE);
 	common_pci_fixup(monet_map_irq, monet_swizzle);
-	/* es1888_init(); */ /* later? */
 	SMC669_Init(1);
+	es1888_init();
 }
 
 static void __init

arch/alpha/kernel/time.c

@@ -94,6 +94,10 @@ void timer_interrupt(int irq, void *dev, struct pt_regs * regs)
 	smp_percpu_timer_interrupt(regs);
 	if (smp_processor_id() != smp_boot_cpuid)
 		return;
+#else
+	/* Not SMP, do kernel PC profiling here.  */
+	if (!user_mode(regs))
+		alpha_do_profile(regs->pc);
 #endif
 
 	write_lock(&xtime_lock);

arch/alpha/math-emu/ieee-math.c

@@ -704,20 +704,21 @@ ieee_CVTQS (int f, unsigned long a, unsigned long *b)
  *       FPCR_INV if invalid operation occurred, etc.
  */
 unsigned long
-ieee_CVTQT (int f, unsigned long a, unsigned long *b)
+ieee_CVTQT (int f, long a, unsigned long *b)
 {
 	EXTENDED op_b;
 
-	op_b.s    = 0;
-	op_b.f[0] = a;
-	op_b.f[1] = 0;
-	if (sign(a) < 0) {
-		op_b.s = 1;
-		op_b.f[0] = -a;
+	if (a != 0) {
+		op_b.s = (a < 0 ? 1 : 0);
+		op_b.f[0] = (a < 0 ? -a : a);
+		op_b.f[1] = 0;
+		op_b.e = 55;
+		normalize(&op_b);
+		return round_t_ieee(f, &op_b, b);
+	} else {
+		*b = 0;
+		return 0;
 	}
-	op_b.e = 55;
-	normalize(&op_b);
-	return round_t_ieee(f, &op_b, b);
 }
 
 

arch/alpha/math-emu/ieee-math.h

@@ -20,7 +20,7 @@
 extern unsigned long ieee_CVTST (int rm, unsigned long a, unsigned long *b);
 extern unsigned long ieee_CVTTS (int rm, unsigned long a, unsigned long *b);
 extern unsigned long ieee_CVTQS (int rm, unsigned long a, unsigned long *b);
-extern unsigned long ieee_CVTQT (int rm, unsigned long a, unsigned long *b);
+extern unsigned long ieee_CVTQT (int rm, long a, unsigned long *b);
 extern unsigned long ieee_CVTTQ (int rm, unsigned long a, unsigned long *b);
 
 extern unsigned long ieee_CMPTEQ (unsigned long a, unsigned long b,

arch/alpha/mm/fault.c

@@ -41,7 +41,7 @@ void
 get_new_mmu_context(struct task_struct *p, struct mm_struct *mm)
 {
 	unsigned long new = __get_new_mmu_context(p, mm);
-	p->tss.mm_context = new;
+	mm->context = new;
 	p->tss.asn = new & HARDWARE_ASN_MASK;
 }
 

arch/alpha/mm/init.c

@@ -173,7 +173,7 @@ show_mem(void)
 
 extern unsigned long free_area_init(unsigned long, unsigned long);
 
-static inline struct thread_struct *
+static inline unsigned long
 load_PCB(struct thread_struct * pcb)
 {
 	register unsigned long sp __asm__("$30");
@@ -192,7 +192,7 @@ paging_init(unsigned long start_mem, unsigned long end_mem)
 	unsigned long newptbr;
 	struct memclust_struct * cluster;
 	struct memdesc_struct * memdesc;
-	struct thread_struct *original_pcb_ptr;
+	unsigned long original_pcb_ptr;
 
 	/* initialize mem_map[] */
 	start_mem = free_area_init(start_mem, end_mem);
@@ -246,11 +246,11 @@ paging_init(unsigned long start_mem, unsigned long end_mem)
 	   since KSEG values also happen to work, folks get confused.
 	   Check this here.  */
 
-	if ((unsigned long)original_pcb_ptr < PAGE_OFFSET) {
-		original_pcb_ptr = (struct thread_struct *)
-		  phys_to_virt((unsigned long) original_pcb_ptr);
+	if (original_pcb_ptr < PAGE_OFFSET) {
+		original_pcb_ptr = (unsigned long)
+			phys_to_virt(original_pcb_ptr);
 	}
-	original_pcb = *original_pcb_ptr;
+	original_pcb = *(struct thread_struct *) original_pcb_ptr;
 
 	return start_mem;
 }

arch/i386/kernel/ldt.c

@@ -2,7 +2,7 @@
  * linux/kernel/ldt.c
  *
  * Copyright (C) 1992 Krishna Balasubramanian and Linus Torvalds
- * Copyright (C) 1998 Ingo Molnar
+ * Copyright (C) 1999 Ingo Molnar <mingo@redhat.com>
  */
 
 #include <linux/errno.h>

arch/i386/kernel/process.c

@@ -539,7 +539,7 @@ int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
 		:"=&a" (retval), "=&S" (d0)
 		:"0" (__NR_clone), "i" (__NR_exit),
 		 "r" (arg), "r" (fn),
-		 "b" (flags | CLONE_VM)
+		 "b" (flags | CLONE_VM | CLONE_TLB)
 		: "memory");
 	return retval;
 }
@@ -762,7 +762,7 @@ void __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 		unsigned long new_cr3 = next->cr3;
 
 		tss->cr3 = new_cr3;
-		if (new_cr3 != prev->cr3) 
+		if (new_cr3 != prev->cr3)
 			asm volatile("movl %0,%%cr3": :"r" (new_cr3));
 	}
 

arch/ppc/Makefile

@@ -65,8 +65,7 @@ endif
 checks:
 	@$(MAKE) -C arch/$(ARCH)/kernel checks
 
-BOOT_TARGETS = netboot znetboot zImage floppy install \
-	vmlinux.coff znetboot.initrd zImage.initrd vmlinux.coff.initrd
+BOOT_TARGETS = zImage znetboot.initrd zImage.initrd
 
 ifdef CONFIG_MBX
 $(BOOT_TARGETS): $(CHECKS) vmlinux
@@ -77,6 +76,24 @@ $(BOOT_TARGETS): $(CHECKS) vmlinux
 	@$(MAKECOFFBOOT) $@
 	@$(MAKEBOOT) $@
 	@$(MAKECHRPBOOT) $@
+
+znetboot: $(CHECKS) vmlinux
+ifdef CONFIG_SMP
+ifdef CONFIG_PPC64
+	cp -f vmlinux /tftpboot/vmlinux.smp.64
+else
+	cp -f vmlinux /tftpboot/vmlinux.smp
+endif
+else
+ifdef CONFIG_PPC64
+	cp -f vmlinux /tftpboot/vmlinux.64
+else
+	cp -f vmlinux /tftpboot/vmlinux
+endif
+endif
+	@$(MAKECOFFBOOT) $@
+	@$(MAKEBOOT) $@
+	@$(MAKECHRPBOOT) $@
 endif
 
 pmac_config:

arch/ppc/amiga/amiints.c

@@ -108,7 +108,7 @@ __initfunc(void amiga_init_IRQ(void))
 	custom.intreq = 0x7fff;
 
 #ifdef CONFIG_APUS
-	/* Clear any inter-CPU interrupt requests. Circumvents bug in
+	/* Clear any inter-CPU interupt requests. Circumvents bug in
            Blizzard IPL emulation HW (or so it appears). */
 	APUS_WRITE(APUS_INT_LVL, INTLVL_SETRESET | INTLVL_MASK);
 

arch/ppc/amiga/config.c

@@ -30,6 +30,7 @@ void (*kbd_reset_setup) (char *, int) __initdata = 0;
 #include <linux/kd.h>
 #include <linux/tty.h>
 #include <linux/console.h>
+#include <linux/init.h>
 
 #include <asm/bootinfo.h>
 #include <asm/setup.h>

arch/ppc/chrpboot/main.c

@@ -71,7 +71,7 @@ chrpboot(int a1, int a2, void *prom)
     sa = *(unsigned long *)PROG_START+PROG_START;
     printf("start address = 0x%x\n\r", sa);
 
-    (*(void (*)())sa)(a1, a2, prom, 0, 0);
+    (*(void (*)())sa)(0, 0, prom, a1, a2);
 
     printf("returned?\n\r");
 

arch/ppc/config.in

-# $Id: config.in,v 1.94 1999/06/25 11:00:07 davem Exp $
+# $Id: config.in,v 1.95 1999/07/03 08:57:06 davem Exp $
 # For a description of the syntax of this configuration file,
 # see the Configure script.
 #

arch/ppc/kernel/chrp_pci.c

@@ -97,7 +97,7 @@ int gg2_pcibios_write_config_dword(unsigned char bus, unsigned char dev_fn,
 #define python_config_data(bus) ((0xfef00000+0xf8010)-(bus*0x100000))
 #define PYTHON_CFA(b, d, o)	(0x80 | ((b<<6) << 8) | ((d) << 16) \
 				 | (((o) & ~3) << 24))
-unsigned int python_busnr = 1;
+unsigned int python_busnr = 0;
 
 int python_pcibios_read_config_byte(unsigned char bus, unsigned char dev_fn,
 				    unsigned char offset, unsigned char *val)
@@ -347,7 +347,7 @@ chrp_setup_pci_ptrs(void)
 			} else if ( !strncmp("IBM,7043-260",
 			   get_property(find_path_device("/"), "name", NULL),12) )
 			{
-				pci_dram_offset = 0x80000000;
+				pci_dram_offset = 0x0;
 				isa_mem_base = 0xc0000000;
 				isa_io_base = 0xf8000000;
 			}

arch/ppc/kernel/chrp_setup.c

@@ -71,6 +71,8 @@ void chrp_calibrate_decr(void);
 void chrp_time_init(void);
 
 void chrp_setup_pci_ptrs(void);
+extern void chrp_progress(char *, unsigned short);
+void chrp_event_scan(void);
 
 extern int pckbd_setkeycode(unsigned int scancode, unsigned int keycode);
 extern int pckbd_getkeycode(unsigned int scancode);
@@ -589,10 +591,10 @@ __initfunc(void
 	chrp_setup_pci_ptrs();
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* take care of initrd if we have one */
-	if ( r3 )
+	if ( r6 )
 	{
-		initrd_start = r3 + KERNELBASE;
-		initrd_end = r3 + r4 + KERNELBASE;
+		initrd_start = r6 + KERNELBASE;
+		initrd_end = r6 + r7 + KERNELBASE;
 	}
 #endif /* CONFIG_BLK_DEV_INITRD */
 
@@ -658,6 +660,8 @@ __initfunc(void
 	ppc_md.ppc_kbd_sysrq_xlate	 = pckbd_sysrq_xlate;
 	SYSRQ_KEY = 0x54;
 #endif
+	if ( rtas_data )
+		ppc_md.progress = chrp_progress;
 #endif
 #endif
 
@@ -678,16 +682,33 @@ __initfunc(void
 	 * Print the banner, then scroll down so boot progress
 	 * can be printed.  -- Cort 
 	 */
-	chrp_progress("Linux/PPC "UTS_RELEASE"\n");
+	if ( ppc_md.progress ) ppc_md.progress("Linux/PPC "UTS_RELEASE"\n", 0x0);
 }
 
-void chrp_progress(char *s)
+void chrp_progress(char *s, unsigned short hex)
 {
 	extern unsigned int rtas_data;
+	unsigned long width;
+	struct device_node *root;
+	char *os = s;
 	
+	if ( (root = find_path_device("/rtas")) )
+		width = *(unsigned long *)get_property(root, "ibm,display-line-length", NULL);
+	else
+		width = 0x10;
+
 	if ( (_machine != _MACH_chrp) || !rtas_data )
 		return;
-	call_rtas( "display-character", 1, 1, NULL, '\r' );
-	while ( *s )
-		call_rtas( "display-character", 1, 1, NULL, *s++ );
+	if ( call_rtas( "display-character", 1, 1, NULL, '\r' ) )
+	{
+		/* assume no display-character RTAS method - use hex display */
+		return;
+	}
+	while ( *os )
+		call_rtas( "display-character", 1, 1, NULL, *os++ );
+	/* scan back for the last newline or carriage return */
+	for ( os-- ; (*os != '\n') && (*os != '\r') && (os > s) ; os--, width-- )
+		/* nothing */ ;
+	/*while ( width-- )*/
+	call_rtas( "display-character", 1, 1, NULL, ' ' );
 }

arch/ppc/kernel/head.S

 /*
  *  arch/ppc/kernel/head.S
  *
- *  $Id: head.S,v 1.133 1999/05/20 05:13:08 cort Exp $
+ *  $Id: head.S,v 1.134 1999/06/30 05:05:52 paulus Exp $
  *
  *  PowerPC version 
  *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
@@ -300,7 +300,7 @@ __secondary_start:
 	oris	r21,r11,(KERNELBASE+0x20000000)@h
 	mtspr	DBAT2L,r18		/* N.B. 6xx (not 601) have valid */
 	mtspr	DBAT2U,r21		/* bit in upper BAT register */
-	mtspr	IBAT2L,r28
+	mtspr	IBAT2L,r18
 	mtspr	IBAT2U,r21
 #endif /* CONFIG_PPC64 */
 #endif
@@ -1978,7 +1978,13 @@ DoSyscall:
 	lwz	r6,GPR4(r1)
 	lwz	r7,GPR5(r1)
 	lwz	r8,GPR6(r1)
-	mr	r9,r2
+	lwz	r9,GPR7(r1)
+	bl	printk
+	lis	r3,77f@ha
+	addi	r3,r3,77f@l
+	lwz	r4,GPR8(r1)
+	lwz	r5,GPR9(r1)
+	mr	r6,r2
 	bl	printk
 	lwz	r0,GPR0(r1)
 	lwz	r3,GPR3(r1)
@@ -2081,7 +2087,8 @@ syscall_ret_2:
 66:	li	r3,ENOSYS
 	b	52b
 #ifdef SHOW_SYSCALLS
-7:	.string	"syscall %d(%x, %x, %x, %x), current=%p\n"
+7:	.string	"syscall %d(%x, %x, %x, %x, %x, "
+77:	.string	"%x, %x), current=%p\n"
 79:	.string	" -> %x\n"
 	.align	2
 #endif

arch/ppc/kernel/irq.c

@@ -50,6 +50,7 @@
 #include <asm/io.h>
 #include <asm/pgtable.h>
 #include <asm/irq.h>
+#include <asm/bitops.h>
 #include <asm/gg2.h>
 #include <asm/cache.h>
 #include <asm/prom.h>

arch/ppc/kernel/openpic.c

@@ -182,6 +182,8 @@ __initfunc(void openpic_init(int main_pic))
     if (!OpenPIC)
 	panic("No OpenPIC found");
 
+    if ( ppc_md.progress ) ppc_md.progress("openpic enter",0x122);
+
     t = openpic_read(&OpenPIC->Global.Feature_Reporting0);
     switch (t & OPENPIC_FEATURE_VERSION_MASK) {
 	case 1:
@@ -201,6 +203,7 @@ __initfunc(void openpic_init(int main_pic))
 		     OPENPIC_FEATURE_LAST_PROCESSOR_SHIFT) + 1;
     NumSources = ((t & OPENPIC_FEATURE_LAST_SOURCE_MASK) >>
 		  OPENPIC_FEATURE_LAST_SOURCE_SHIFT) + 1;
+
     printk("OpenPIC Version %s (%d CPUs and %d IRQ sources) at %p\n", version,
 	   NumProcessors, NumSources, OpenPIC);
     timerfreq = openpic_read(&OpenPIC->Global.Timer_Frequency);
@@ -212,6 +215,8 @@ __initfunc(void openpic_init(int main_pic))
 
     if ( main_pic )
     {
+	    if ( ppc_md.progress ) ppc_md.progress("openpic main",0x3ff);
+      
 	    /* Initialize timer interrupts */
 	    for (i = 0; i < OPENPIC_NUM_TIMERS; i++) {
 		    /* Disabled, Priority 0 */
@@ -226,9 +231,12 @@ __initfunc(void openpic_init(int main_pic))
 		    openpic_initipi(i, 0, OPENPIC_VEC_IPI+i);
 	    }
 	    
+	    if ( ppc_md.progress ) ppc_md.progress("openpic initirq",0x3bb);
 	    /* Initialize external interrupts */
 	    /* SIOint (8259 cascade) is special */
 	    openpic_initirq(0, 8, OPENPIC_VEC_SOURCE, 1, 1);
+	    
+	    if ( ppc_md.progress ) ppc_md.progress("openpic map",0x3cc);
 	    /* Processor 0 */
 	    openpic_mapirq(0, 1<<0);
 	    for (i = 1; i < NumSources; i++) {
@@ -248,6 +256,7 @@ __initfunc(void openpic_init(int main_pic))
 	    openpic_set_priority(0, 0);
 	    openpic_disable_8259_pass_through();
     }
+    if ( ppc_md.progress ) ppc_md.progress("openpic exit",0x222);
 }
 
 

arch/ppc/kernel/pci.c

@@ -9,6 +9,7 @@
 #include <linux/string.h>
 #include <linux/init.h>
 #include <linux/config.h>
+#include <linux/pci.h>
 #include <linux/openpic.h>
 
 #include <asm/processor.h>

arch/ppc/kernel/pmac_setup.c

@@ -58,6 +58,7 @@
 #include <asm/ide.h>
 #include <asm/machdep.h>
 #include <asm/keyboard.h>
+#include <asm/dma.h>
 
 #include "time.h"
 #include "local_irq.h"
@@ -204,11 +205,12 @@ kdev_t sd_find_target(void *host, int tgt)
 {
     Scsi_Disk *dp;
     int i;
-
+#ifdef CONFIG_BLK_DEV_SD
     for (dp = rscsi_disks, i = 0; i < sd_template.dev_max; ++i, ++dp)
         if (dp->device != NULL && dp->device->host == host
             && dp->device->id == tgt)
             return MKDEV_SD(i);
+#endif /* CONFIG_BLK_DEV_SD */
     return 0;
 }
 #endif
@@ -517,13 +519,13 @@ pmac_halt(void)
 void
 pmac_ide_insw(ide_ioreg_t port, void *buf, int ns)
 {
-	_insw_ns(port+_IO_BASE, buf, ns);
+	_insw_ns((unsigned short *)(port+_IO_BASE), buf, ns);
 }
 
 void
 pmac_ide_outsw(ide_ioreg_t port, void *buf, int ns)
 {
-	_outsw_ns(port+_IO_BASE, buf, ns);
+	_outsw_ns((unsigned short *)(port+_IO_BASE), buf, ns);
 }
 
 int

arch/ppc/kernel/ppc-stub.c

@@ -107,6 +107,7 @@
 
 #include <asm/system.h>
 #include <asm/signal.h>
+#include <asm/system.h>
 #include <asm/kgdb.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>

arch/ppc/kernel/prep_setup.c

@@ -40,6 +40,7 @@
 #include <asm/residual.h>
 #include <asm/io.h>
 #include <asm/pgtable.h>
+#include <linux/ide.h>
 #include <asm/ide.h>
 #include <asm/cache.h>
 #include <asm/dma.h>

arch/ppc/kernel/process.c

 /*
- * $Id: process.c,v 1.86 1999/06/17 21:53:46 cort Exp $
+ * $Id: process.c,v 1.87 1999/07/03 08:57:07 davem Exp $
  *
  *  linux/arch/ppc/kernel/process.c
  *

arch/ppc/kernel/prom.c

 /*
- * $Id: prom.c,v 1.61 1999/06/17 06:05:52 paulus Exp $
+ * $Id: prom.c,v 1.62 1999/07/02 19:59:31 cort Exp $
  *
  * Procedures for interfacing to the Open Firmware PROM on
  * Power Macintosh computers.
@@ -502,8 +502,8 @@ prom_init(int r3, int r4, prom_entry pp)
 			return;
 		
 		/* copy the holding pattern code to someplace safe (8M) */
-		memcpy( (void *)(8<<20), RELOC(__secondary_hold), 0x10000 );
-		for (i = 8<<20; i < ((8<<20)+0x10000); i += 32)
+		memcpy( (void *)(8<<20), RELOC(__secondary_hold), 0x100 );
+		for (i = 8<<20; i < ((8<<20)+0x100); i += 32)
 		{
 			asm volatile("dcbf 0,%0" : : "r" (i) : "memory");
 			asm volatile("icbi 0,%0" : : "r" (i) : "memory");

arch/ppc/kernel/ptrace.c

@@ -81,6 +81,7 @@ clear_single_step(struct task_struct *task)
 	regs->msr &= ~MSR_SE;
 }
 
+#if 0
 /*
  * This routine gets a long from any process space by following the page
  * tables. NOTE! You should check that the long isn't on a page boundary,
@@ -283,11 +284,13 @@ static int write_long(struct task_struct * tsk, unsigned long addr,
 		put_long(tsk, vma,addr,data);
 	return 0;
 }
+#endif
 
 asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 {
 	struct task_struct *child;
 	int ret = -EPERM;
+	unsigned long flags;
 
 	lock_kernel();
 	if (request == PTRACE_TRACEME) {
@@ -302,7 +305,10 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 	if (pid == 1)		/* you may not mess with init */
 		goto out;
 	ret = -ESRCH;
-	if (!(child = find_task_by_pid(pid)))
+	read_lock(&tasklist_lock);
+	child = find_task_by_pid(pid);
+	read_unlock(&tasklist_lock);	/* FIXME!!! */
+	if ( !child )
 		goto out;
 	ret = -EPERM;
 	if (request == PTRACE_ATTACH) {
@@ -322,11 +328,15 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 		if (child->flags & PF_PTRACED)
 			goto out;
 		child->flags |= PF_PTRACED;
+		
+		write_lock_irqsave(&tasklist_lock, flags);
 		if (child->p_pptr != current) {
 			REMOVE_LINKS(child);
 			child->p_pptr = current;
 			SET_LINKS(child);
 		}
+		write_unlock_irqrestore(&tasklist_lock, flags);
+
 		send_sig(SIGSTOP, child, 1);
 		ret = 0;
 		goto out;
@@ -342,22 +352,19 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 		goto out;
 
 	switch (request) {
-	/* If I and D space are separate, these will need to be fixed. */
+	/* when I and D space are separate, these will need to be fixed. */
 		case PTRACE_PEEKTEXT: /* read word at location addr. */ 
 		case PTRACE_PEEKDATA: {
 			unsigned long tmp;
+			int copied;
 
-			down(&child->mm->mmap_sem);
-			ret = read_long(child, addr, &tmp);
-			up(&child->mm->mmap_sem);
-			if (ret < 0)
+			copied = access_process_vm(child, addr, &tmp, sizeof(tmp), 0);
+			ret = -EIO;
+			if (copied != sizeof(tmp))
 				goto out;
-			ret = verify_area(VERIFY_WRITE, (void *) data, sizeof(long));
-			if (!ret)
-				put_user(tmp, (unsigned long *) data);
+			ret = put_user(tmp,(unsigned long *) data);
 			goto out;
 		}
-
 	/* read the word at location addr in the USER area. */
 		case PTRACE_PEEKUSR: {
 			unsigned long tmp;
@@ -391,11 +398,11 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
       /* If I and D space are separate, this will have to be fixed. */
 		case PTRACE_POKETEXT: /* write the word at location addr. */
 		case PTRACE_POKEDATA:
-			down(&child->mm->mmap_sem);
-			ret = write_long(child,addr,data);
-			up(&child->mm->mmap_sem);
+			ret = 0;
+			if (access_process_vm(child, addr, &data, sizeof(data), 1) == sizeof(data))
+				goto out;
+			ret = -EIO;
 			goto out;
-
 		case PTRACE_POKEUSR: /* write the word at location addr in the USER area */
 			ret = -EIO;
 			if ((addr & 3) || addr < 0 || addr >= ((PT_FPR0 + 64) << 2))
@@ -459,9 +466,9 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 				goto out;
 			child->flags &= ~PF_TRACESYS;
 			set_single_step(child);
-			wake_up_process(child);
 			child->exit_code = data;
 			/* give it a chance to run. */
+			wake_up_process(child);
 			ret = 0;
 			goto out;
 		}
@@ -473,9 +480,11 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 			child->flags &= ~(PF_PTRACED|PF_TRACESYS);
 			wake_up_process(child);
 			child->exit_code = data;
+			write_lock_irqsave(&tasklist_lock, flags);
 			REMOVE_LINKS(child);
 			child->p_pptr = child->p_opptr;
 			SET_LINKS(child);
+			write_unlock_irqrestore(&tasklist_lock, flags);
 			/* make sure the single step bit is not set. */
 			clear_single_step(child);
 			ret = 0;
@@ -493,7 +502,6 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 
 asmlinkage void syscall_trace(void)
 {
-	lock_kernel();
 	if ((current->flags & (PF_PTRACED|PF_TRACESYS))
 			!= (PF_PTRACED|PF_TRACESYS))
 		goto out;
@@ -511,5 +519,4 @@ asmlinkage void syscall_trace(void)
 		current->exit_code = 0;
 	}
 out:
-	unlock_kernel();
 }

arch/ppc/kernel/setup.c

 /*
- * $Id: setup.c,v 1.136 1999/06/18 07:11:35 cort Exp $
+ * $Id: setup.c,v 1.138 1999/07/11 16:32:21 cort Exp $
  * Common prep/pmac/chrp boot and setup code.
  */
 
@@ -32,6 +32,7 @@
 #endif
 #include <asm/bootx.h>
 #include <asm/machdep.h>
+#include <asm/ide.h>
 
 extern void pmac_init(unsigned long r3,
                       unsigned long r4,
@@ -331,6 +332,7 @@ identify_machine(unsigned long r3, unsigned long r4, unsigned long r5,
 #ifdef __SMP__
 	if ( first_cpu_booted ) return 0;
 #endif /* __SMP__ */
+	if ( ppc_md.progress ) ppc_md.progress("id mach(): start", 0x100);
 	
 #ifndef CONFIG_MACH_SPECIFIC
 	/* boot loader will tell us if we're APUS */
@@ -477,13 +479,13 @@ identify_machine(unsigned long r3, unsigned long r4, unsigned long r5,
 	default:
 		printk("Unknown machine type in identify_machine!\n");
 	}
-
 	/* Check for nobats option (used in mapin_ram). */
 	if (strstr(cmd_line, "nobats")) {
 		extern int __map_without_bats;
 		__map_without_bats = 1;
 	}
-	
+
+	if ( ppc_md.progress ) ppc_md.progress("id mach(): done", 0x200);
 	return 0;
 }
 
@@ -539,6 +541,8 @@ __initfunc(void setup_arch(char **cmdline_p,
 	*memory_end_p = (unsigned long) end_of_DRAM;
 
 	ppc_md.setup_arch(memory_start_p, memory_end_p);
+	/* clear the progress line */
+	if ( ppc_md.progress ) ppc_md.progress(" ", 0xffff);
 }
 
 void ppc_generic_ide_fix_driveid(struct hd_driveid *id)

arch/ppc/kernel/smp.c

 /*
- * $Id: smp.c,v 1.54 1999/06/24 17:13:34 cort Exp $
+ * $Id: smp.c,v 1.55 1999/07/03 08:57:09 davem Exp $
  *
  * Smp support for ppc.
  *

arch/ppc/mm/init.c

 /*
- *  $Id: init.c,v 1.170 1999/06/29 12:33:51 davem Exp $
+ *  $Id: init.c,v 1.171 1999/07/08 23:20:14 cort Exp $
  *
  *  PowerPC version 
  *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
@@ -50,6 +50,7 @@
 #include <asm/mbx.h>
 #include <asm/smp.h>
 #include <asm/bootx.h>
+#include <asm/machdep.h>
 /* APUS includes */
 #include <asm/setup.h>
 #include <asm/amigahw.h>
@@ -961,7 +962,6 @@ static void *MMU_get_page(void)
 	} else {
 		p = find_mem_piece(PAGE_SIZE, PAGE_SIZE);
 	}
-	/*memset(p, 0, PAGE_SIZE);*/
 	__clear_user(p, PAGE_SIZE);
 	return p;
 }
@@ -1027,7 +1027,7 @@ __initfunc(void MMU_init(void))
 #ifdef __SMP__
 	if ( first_cpu_booted ) return;
 #endif /* __SMP__ */
-	
+	if ( ppc_md.progress ) ppc_md.progress("MMU:enter", 0x111);
 #ifndef CONFIG_8xx
 	if (have_of)
 		end_of_DRAM = pmac_find_end_of_memory();
@@ -1038,10 +1038,12 @@ __initfunc(void MMU_init(void))
 	else /* prep */
 		end_of_DRAM = prep_find_end_of_memory();
 
+	if ( ppc_md.progress ) ppc_md.progress("MMU:hash init", 0x300);
         hash_init();
         _SDR1 = __pa(Hash) | (Hash_mask >> 10);
 	ioremap_base = 0xf8000000;
 
+	if ( ppc_md.progress ) ppc_md.progress("MMU:mapin", 0x301);
 	/* Map in all of RAM starting at KERNELBASE */
 	mapin_ram();
 
@@ -1050,6 +1052,7 @@ __initfunc(void MMU_init(void))
 	 * the io areas.  RAM was mapped by mapin_ram().
 	 * -- Cort
 	 */
+	if ( ppc_md.progress ) ppc_md.progress("MMU:setbat", 0x302);
 	switch (_machine) {
 	case _MACH_prep:
 		setbat(0, 0x80000000, 0x80000000, 0x10000000, IO_PAGE);
@@ -1102,6 +1105,7 @@ __initfunc(void MMU_init(void))
         ioremap(0x80000000, 0x4000);
         ioremap(0x81000000, 0x4000);
 #endif /* CONFIG_8xx */
+	if ( ppc_md.progress ) ppc_md.progress("MMU:exit", 0x211);
 }
 
 /*
@@ -1310,7 +1314,7 @@ __initfunc(unsigned long *pmac_find_end_of_memory(void))
 	int i;
 	
 	/* max amount of RAM we allow -- Cort */
-#define RAM_LIMIT (768<<20)
+#define RAM_LIMIT (256<<20)
 
 	memory_node = find_devices("memory");
 	if (memory_node == NULL) {
@@ -1509,6 +1513,7 @@ __initfunc(static void hash_init(void))
 	extern unsigned int hash_page_patch_A[], hash_page_patch_B[],
 		hash_page_patch_C[], hash_page[];
 
+	if ( ppc_md.progress ) ppc_md.progress("hash:enter", 0x105);
 	/*
 	 * Allow 64k of hash table for every 16MB of memory,
 	 * up to a maximum of 2MB.
@@ -1542,6 +1547,7 @@ __initfunc(static void hash_init(void))
  	}
 #endif /* NO_RELOAD_HTAB */
 	
+	if ( ppc_md.progress ) ppc_md.progress("hash:find piece", 0x322);
 	/* Find some memory for the hash table. */
 	if ( Hash_size )
 		Hash = find_mem_piece(Hash_size, Hash_size);
@@ -1557,11 +1563,11 @@ __initfunc(static void hash_init(void))
 #else
 #define b(x) (x)
 #endif
-		/*memset(Hash, 0, Hash_size);*/
-		__clear_user(Hash, Hash_size);
 		
 		Hash_end = (PTE *) ((unsigned long)Hash + Hash_size);
-
+		__clear_user(Hash, Hash_size);
+		
+		if ( ppc_md.progress ) ppc_md.progress("hash:patch", 0x345);
 		/*
 		 * Patch up the instructions in head.S:hash_page
 		 */
@@ -1601,5 +1607,6 @@ __initfunc(static void hash_init(void))
 		flush_icache_range((unsigned long) b(hash_page),
 				   (unsigned long) b(hash_page + 1));
 	}
+	if ( ppc_md.progress ) ppc_md.progress("hash:done", 0x205);
 }
 #endif /* ndef CONFIG_8xx */

drivers/block/loop.c

@@ -375,7 +375,7 @@ static int loop_set_fd(struct loop_device *lo, kdev_t dev, unsigned int arg)
 		   a file structure */
 		lo->lo_backing_file = NULL;
 	} else if (S_ISREG(inode->i_mode)) {
-		if (!inode->i_op->bmap) { 
+		if (!inode->i_op->get_block) { 
 			printk(KERN_ERR "loop: device has no block access/not implemented\n");
 			goto out_putf;
 		}

drivers/net/pcnet32.c

@@ -13,7 +13,7 @@
  * 	This driver is for PCnet32 and PCnetPCI based ethercards
  */
 
-static const char *version = "pcnet32.c:v1.21 31.3.99 tsbogend@alpha.franken.de\n";
+static const char *version = "pcnet32.c:v1.23 6.7.1999 tsbogend@alpha.franken.de\n";
 
 #include <linux/config.h>
 #include <linux/module.h>
@@ -149,6 +149,13 @@ static int full_duplex[MAX_UNITS] = {0, };
  *         rewritten PCI card detection
  *         added dwio mode to get driver working on some PPC machines
  * v1.21:  added mii selection and mii ioctl
+ * v1.22:  changed pci scanning code to make PPC people happy
+ *         fixed switching to 32bit mode in pcnet32_open() (thanks
+ *         to Michael Richard <mcr@solidum.com> for noticing this one)
+ *	   added sub vendor/device id matching (thanks again to 
+ *	   Michael Richard <mcr@solidum.com>)
+ *         added chip id for 79c973/975 (thanks to Zach Brown <zab@zabbo.net>)
+ * v1.23   fixed small bug, when manual selecting MII speed/duplex
  */
 
 
@@ -185,6 +192,16 @@ static int full_duplex[MAX_UNITS] = {0, };
 
 #define PCNET32_TOTAL_SIZE 0x20
 
+/* some PCI ids */
+#ifndef PCI_DEVICE_ID_AMD_LANCE
+#define PCI_VENDOR_ID_AMD	      0x1022
+#define PCI_DEVICE_ID_AMD_LANCE	      0x2000
+#endif
+#ifndef PCI_DEVICE_ID_AMD_PCNETHOME
+#define PCI_DEVICE_ID_AMD_PCNETHOME   0x2001
+#endif
+
+
 #define CRC_POLYNOMIAL_LE 0xedb88320UL  /* Ethernet CRC, little endian */
 
 /* The PCNET32 Rx and Tx ring descriptors. */
@@ -272,14 +289,19 @@ enum pci_flags_bit {
 
 struct pcnet32_pci_id_info {
     const char *name;
-    u16 vendor_id, device_id, device_id_mask, flags;
+    u16 vendor_id, device_id, svid, sdid, flags;
     int io_size;
     int (*probe1) (struct device *, unsigned long, unsigned char, int, int);
 };
 
 static struct pcnet32_pci_id_info pcnet32_tbl[] = {
     { "AMD PCnetPCI series",
-	0x1022, 0x2000, 0xfffe, PCI_USES_IO|PCI_USES_MASTER, PCNET32_TOTAL_SIZE,
+	PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_LANCE, 0, 0,
+	PCI_USES_IO|PCI_USES_MASTER, PCNET32_TOTAL_SIZE,
+	pcnet32_probe1},
+    { "AMD PCnetHome series",
+	PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_PCNETHOME, 0, 0,
+	PCI_USES_IO|PCI_USES_MASTER, PCNET32_TOTAL_SIZE,
 	pcnet32_probe1},
     {0,}
 };
@@ -418,30 +440,27 @@ int __init pcnet32_probe (struct device *dev)
     
 #if defined(CONFIG_PCI)
     if (pci_present()) {
-	struct pci_dev *pdev;
-	unsigned char pci_bus, pci_device_fn;
-	int pci_index;
+	struct pci_dev *pdev = NULL;
 	
 	printk("pcnet32.c: PCI bios is present, checking for devices...\n");
-	for (pci_index = 0; pci_index < 0xff; pci_index++) {
-	    u16 vendor, device, pci_command;
+	while ((pdev = pci_find_class (PCI_CLASS_NETWORK_ETHERNET<<8, pdev))) {
+	    u16 pci_command;
 	    int chip_idx;
+	    u16 sdid,svid;
 
-	    if (pcibios_find_class (PCI_CLASS_NETWORK_ETHERNET << 8,
-				    pci_index, &pci_bus, &pci_device_fn) != PCIBIOS_SUCCESSFUL)
-		break;
-	    
-	    pcibios_read_config_word(pci_bus, pci_device_fn, PCI_VENDOR_ID, &vendor);
-	    pcibios_read_config_word(pci_bus, pci_device_fn, PCI_DEVICE_ID, &device);
-
+	    pci_read_config_word(pdev, PCI_SUBSYSTEM_VENDOR_ID, &sdid);
+	    pci_read_config_word(pdev, PCI_SUBSYSTEM_ID, &svid);
 	    for (chip_idx = 0; pcnet32_tbl[chip_idx].vendor_id; chip_idx++)
-		if (vendor == pcnet32_tbl[chip_idx].vendor_id &&
-		    (device & pcnet32_tbl[chip_idx].device_id_mask) == pcnet32_tbl[chip_idx].device_id)
+		if ((pdev->vendor == pcnet32_tbl[chip_idx].vendor_id) &&
+		    (pdev->device == pcnet32_tbl[chip_idx].device_id) &&
+		    (pcnet32_tbl[chip_idx].svid == 0 || 
+		     (svid == pcnet32_tbl[chip_idx].svid)) &&
+		    (pcnet32_tbl[chip_idx].sdid == 0 || 
+		     (sdid == pcnet32_tbl[chip_idx].sdid)))
 		    break;
 	    if (pcnet32_tbl[chip_idx].vendor_id == 0)
 		continue;
 	    
-	    pdev = pci_find_slot(pci_bus, pci_device_fn);
 	    ioaddr = pdev->base_address[0] & PCI_BASE_ADDRESS_IO_MASK;
 #if defined(ADDR_64BITS) && defined(__alpha__)
 	    ioaddr |= ((long)pdev->base_address[1]) << 32;
@@ -541,6 +560,10 @@ pcnet32_probe1(struct device *dev, unsigned long ioaddr, unsigned char irq_line,
 	chipname = "PCnet/FAST+ 79C972";
 	fdx = 1; mii = 1;
 	break;
+     case 0x2625:
+	chipname = "PCnet/FAST III 79C973";
+	fdx = 1; mii = 1;
+	break;
      case 0x2626:
 	chipname = "PCnet/Home 79C978";
 	fdx = 1;
@@ -561,6 +584,9 @@ pcnet32_probe1(struct device *dev, unsigned long ioaddr, unsigned char irq_line,
 	    printk("pcnet32: pcnet32 media reset to %#x.\n",  media);
 	a->write_bcr (ioaddr, 49, media);
 	break;
+     case 0x2627:
+	chipname = "PCnet/FAST III 79C975";
+	fdx = 1; mii = 1;
      default:
 	printk("pcnet32: PCnet version %#x, no PCnet32 chip.\n",chip_version);
 	return ENODEV;
@@ -693,7 +719,7 @@ pcnet32_open(struct device *dev)
     lp->a.reset (ioaddr);
 
     /* switch pcnet32 to 32bit mode */
-    lp->a.write_csr (ioaddr, 20, 2);
+    lp->a.write_bcr (ioaddr, 20, 2);
 
     if (pcnet32_debug > 1)
 	printk("%s: pcnet32_open() irq %d tx/rx rings %#x/%#x init %#x.\n",
@@ -725,7 +751,7 @@ pcnet32_open(struct device *dev)
 	val |= 0x10;
     lp->a.write_csr (ioaddr, 124, val);
     
-    if (lp->mii & (lp->options & PORT_ASEL)) {
+    if (lp->mii & !(lp->options & PORT_ASEL)) {
 	val = lp->a.read_bcr (ioaddr, 32) & ~0x38; /* disable Auto Negotiation, set 10Mpbs, HD */
 	if (lp->options & PORT_FD)
 	    val |= 0x10;
@@ -952,7 +978,7 @@ pcnet32_interrupt(int irq, void *dev_id, struct pt_regs * regs)
     struct device *dev = (struct device *)dev_id;
     struct pcnet32_private *lp;
     unsigned long ioaddr;
-    u16 csr0;
+    u16 csr0,rap;
     int boguscnt =  max_interrupt_work;
     int must_restart;
 
@@ -968,6 +994,7 @@ pcnet32_interrupt(int irq, void *dev_id, struct pt_regs * regs)
 
     dev->interrupt = 1;
 
+    rap = lp->a.read_rap(ioaddr);
     while ((csr0 = lp->a.read_csr (ioaddr, 0)) & 0x8600 && --boguscnt >= 0) {
 	/* Acknowledge all of the current interrupt sources ASAP. */
 	lp->a.write_csr (ioaddr, 0, csr0 & ~0x004f);
@@ -1069,6 +1096,7 @@ pcnet32_interrupt(int irq, void *dev_id, struct pt_regs * regs)
 
     /* Clear any other interrupt, and set interrupt enable. */
     lp->a.write_csr (ioaddr, 0, 0x7940);
+    lp->a.write_rap(ioaddr,rap);
 
     if (pcnet32_debug > 4)
 	printk("%s: exiting interrupt, csr0=%#4.4x.\n",

fs/buffer.c

@@ -757,9 +757,9 @@ static void end_buffer_io_async(struct buffer_head * bh, int uptodate)
 
 	free = test_and_clear_bit(PG_free_after, &page->flags);
 
-	if (page->owner != -1)
+	if (page->owner != (void *)-1)
 		PAGE_BUG(page);
-	page->owner = (int)current;
+	page->owner = current;
 	UnlockPage(page);
 
 	if (free)
@@ -1195,7 +1195,7 @@ static int create_page_buffers(int rw, struct page *page, kdev_t dev, int b[], i
 
 	if (!PageLocked(page))
 		BUG();
-	if (page->owner != (int)current)
+	if (page->owner != current)
 		PAGE_BUG(page);
 	/*
 	 * Allocate async buffer heads pointing to this page, just for I/O.
@@ -1557,7 +1557,7 @@ int brw_page(int rw, struct page *page, kdev_t dev, int b[], int size, int bmap)
 	}
 	if (!page->buffers)
 		BUG();
-	page->owner = -1;
+	page->owner = (void *)-1;
 
 	head = page->buffers;
 	bh = head;
@@ -1606,7 +1606,7 @@ int brw_page(int rw, struct page *page, kdev_t dev, int b[], int size, int bmap)
 	} else {
 		if (!nr && rw == READ) {
 			SetPageUptodate(page);
-			page->owner = (int)current;
+			page->owner = current;
 			UnlockPage(page);
 		}
 		if (nr && (rw == WRITE))
@@ -1640,7 +1640,7 @@ int block_read_full_page(struct file * file, struct page * page)
 
 	blocks = PAGE_SIZE >> inode->i_sb->s_blocksize_bits;
 	iblock = page->offset >> inode->i_sb->s_blocksize_bits;
-	page->owner = -1;
+	page->owner = (void *)-1;
 	head = page->buffers;
 	bh = head;
 	nr = 0;
@@ -1675,7 +1675,7 @@ int block_read_full_page(struct file * file, struct page * page)
 		 * uptodate as well.
 		 */
 		SetPageUptodate(page);
-		page->owner = (int)current;
+		page->owner = current;
 		UnlockPage(page);
 	}
 	return 0;
@@ -1969,8 +1969,25 @@ asmlinkage int sys_bdflush(int func, long data)
 		goto out;
 
 	if (func == 1) {
-		 error = sync_old_buffers();
-		 goto out;
+		struct mm_struct *user_mm;
+		/*
+		 * bdflush will spend all of it's time in kernel-space,
+		 * without touching user-space, so we can switch it into
+		 * 'lazy TLB mode' to reduce the cost of context-switches
+		 * to and from bdflush.
+		 */
+		user_mm = current->mm;
+		mmget(user_mm);
+		current->flags |= PF_LAZY_TLB;
+
+		error = sync_old_buffers();
+
+		current->flags &= ~PF_LAZY_TLB;
+		SET_PAGE_DIR(current, user_mm->pgd);
+		mmput(current->mm);
+		current->mm = user_mm;
+
+		goto out;
 	}
 
 	/* Basically func 1 means read param 1, 2 means write param 1, etc */

fs/fcntl.c

@@ -31,6 +31,7 @@ static inline int dupfd(struct file *file, unsigned int arg)
 	return error;
 
 out_putf:
+	write_unlock(&files->file_lock);
 	fput(file);
 	goto out;
 }

fs/file_table.c

@@ -182,7 +182,7 @@ int fs_may_remount_ro(struct super_block *sb)
 
 		/* Writable file? */
 		if (S_ISREG(inode->i_mode) && (file->f_mode & FMODE_WRITE))
-			return 0;
+			goto too_bad;
 	}
 	file_list_unlock();
 	return 1; /* Tis' cool bro. */

include/asm-alpha/core_pyxis.h

@@ -74,8 +74,8 @@
 #define PYXIS_MEM_R2_MASK 0x07ffffff  /* SPARSE Mem region 2 mask is 27 bits */
 #define PYXIS_MEM_R3_MASK 0x03ffffff  /* SPARSE Mem region 3 mask is 26 bits */
 
-#define PYXIS_DMA_WIN_BASE_DEFAULT	(1024*1024*1024)
-#define PYXIS_DMA_WIN_SIZE_DEFAULT	(1024*1024*1024)
+#define PYXIS_DMA_WIN_BASE_DEFAULT	(2UL*1024*1024*1024)
+#define PYXIS_DMA_WIN_SIZE_DEFAULT	(1UL*1024*1024*1024)
 
 #if defined(CONFIG_ALPHA_GENERIC) || defined(CONFIG_ALPHA_SRM_SETUP)
 #define PYXIS_DMA_WIN_BASE		alpha_mv.dma_win_base

include/asm-alpha/hardirq.h

@@ -5,8 +5,12 @@
 
 #include <linux/tasks.h>
 
-extern unsigned int local_irq_count[NR_CPUS];
-extern unsigned long hardirq_no[NR_CPUS];
+#ifndef __SMP__
+extern int __local_irq_count;
+#define local_irq_count(cpu)  ((void)(cpu), __local_irq_count)
+#else
+#define local_irq_count(cpu)  (cpu_data[cpu].irq_count)
+#endif
 
 /*
  * Are we in an interrupt context? Either doing bottom half
@@ -16,16 +20,16 @@ extern unsigned long hardirq_no[NR_CPUS];
 #define in_interrupt()						\
 ({								\
 	int __cpu = smp_processor_id();				\
-	(local_irq_count[__cpu] + local_bh_count[__cpu]) != 0;	\
+	(local_irq_count(__cpu) + local_bh_count(__cpu)) != 0;	\
 })
 
 #ifndef __SMP__
 
-#define hardirq_trylock(cpu)	(local_irq_count[cpu] == 0)
+#define hardirq_trylock(cpu)	(local_irq_count(cpu) == 0)
 #define hardirq_endlock(cpu)	((void) 0)
 
-#define hardirq_enter(cpu, irq)	(local_irq_count[cpu]++)
-#define hardirq_exit(cpu, irq)	(local_irq_count[cpu]--)
+#define hardirq_enter(cpu, irq)	(local_irq_count(cpu)++)
+#define hardirq_exit(cpu, irq)	(local_irq_count(cpu)--)
 
 #define synchronize_irq()	barrier()
 
@@ -50,21 +54,20 @@ static inline void release_irqlock(int cpu)
 
 static inline void hardirq_enter(int cpu, int irq)
 {
-	++local_irq_count[cpu];
+	++local_irq_count(cpu);
         atomic_inc(&global_irq_count);
-	hardirq_no[cpu] |= 1L << irq;		/* debugging only */
 }
 
 static inline void hardirq_exit(int cpu, int irq)
 {
-	hardirq_no[cpu] &= ~(1L << irq);	/* debugging only */
 	atomic_dec(&global_irq_count);
-        --local_irq_count[cpu];
+        --local_irq_count(cpu);
 }
 
 static inline int hardirq_trylock(int cpu)
 {
-	return !atomic_read(&global_irq_count) && !global_irq_lock.lock;
+	return (!atomic_read(&global_irq_count)
+		&& !spin_is_locked(&global_irq_lock));
 }
 
 #define hardirq_endlock(cpu)  ((void)0)

include/asm-alpha/mmu_context.h

@@ -66,7 +66,11 @@ extern unsigned long last_asn;
 #endif /* __SMP__ */
 
 #define WIDTH_HARDWARE_ASN	8
+#ifdef __SMP__
 #define WIDTH_THIS_PROCESSOR	5
+#else
+#define WIDTH_THIS_PROCESSOR	0
+#endif
 #define ASN_FIRST_VERSION (1UL << (WIDTH_THIS_PROCESSOR + WIDTH_HARDWARE_ASN))
 #define HARDWARE_ASN_MASK ((1UL << WIDTH_HARDWARE_ASN) - 1)
 
@@ -95,12 +99,11 @@ __get_new_mmu_context(struct task_struct *p, struct mm_struct *mm)
 	unsigned long asn = cpu_last_asn(smp_processor_id());
 	unsigned long next = asn + 1;
 
-	if ((next ^ asn) & ~MAX_ASN) {
+	if ((asn & HARDWARE_ASN_MASK) >= MAX_ASN) {
 		tbiap();
 		next = (asn & ~HARDWARE_ASN_MASK) + ASN_FIRST_VERSION;
 	}
 	cpu_last_asn(smp_processor_id()) = next;
-	mm->context = next;                      /* full version + asn */
 	return next;
 }
 
@@ -110,6 +113,12 @@ ev4_get_mmu_context(struct task_struct *p)
 	/* As described, ASN's are broken.  But we can optimize for
 	   switching between threads -- if the mm is unchanged from
 	   current we needn't flush.  */
+	/* ??? May not be needed because EV4 PALcode recognizes that
+	   ASN's are broken and does a tbiap itself on swpctx, under
+	   the "Must set ASN or flush" rule.  At least this is true
+	   for a 1992 SRM, reports Joseph Martin (jmartin@hlo.dec.com).
+	   I'm going to leave this here anyway, just to Be Sure.  -- r~  */
+
 	if (current->mm != p->mm)
 		tbiap();
 }
@@ -119,17 +128,23 @@ ev5_get_mmu_context(struct task_struct *p)
 {
 	/* Check if our ASN is of an older version, or on a different CPU,
 	   and thus invalid.  */
+	/* ??? If we have two threads on different cpus, we'll continually
+	   fight over the context.  Find a way to record a per-mm, per-cpu
+	   value for the asn.  */
 
-	long asn = cpu_last_asn(smp_processor_id());
+	unsigned long asn = cpu_last_asn(smp_processor_id());
 	struct mm_struct *mm = p->mm;
-	long mmc = mm->context;
+	unsigned long mmc = mm->context;
 	
-	if ((p->tss.mm_context ^ asn) & ~HARDWARE_ASN_MASK) {
-		if ((mmc ^ asn) & ~HARDWARE_ASN_MASK)
-			mmc = __get_new_mmu_context(p, mm);
-		p->tss.mm_context = mmc;
-		p->tss.asn = mmc & HARDWARE_ASN_MASK;
+	if ((mmc ^ asn) & ~HARDWARE_ASN_MASK) {
+		mmc = __get_new_mmu_context(p, mm);
+		mm->context = mmc;
 	}
+
+	/* Always update the PCB ASN.  Another thread may have allocated
+	   a new mm->context (via flush_tlb_mm) without the ASN serial
+	   number wrapping.  We have no way to detect when this is needed.  */
+	p->tss.asn = mmc & HARDWARE_ASN_MASK;
 }
 
 #ifdef CONFIG_ALPHA_GENERIC

include/asm-alpha/param.h

 #ifndef _ASM_ALPHA_PARAM_H
 #define _ASM_ALPHA_PARAM_H
 
+/* ??? Gross.  I don't want to parameterize this, and supposedly the
+   hardware ignores reprogramming.  We also need userland buy-in to the 
+   change in HZ, since this is visible in the wait4 resources etc.  */
+
 #ifndef HZ
-# define HZ	1024
+# ifndef CONFIG_ALPHA_RAWHIDE
+#  define HZ	1024
+# else
+#  define HZ	1200
+# endif
 #endif
 
 #define EXEC_PAGESIZE	8192

include/asm-alpha/pgtable.h

@@ -49,7 +49,6 @@ ev4_flush_tlb_other(struct mm_struct *mm)
 __EXTERN_INLINE void
 ev5_flush_tlb_current(struct mm_struct *mm)
 {
-	mm->context = 0;
 	get_new_mmu_context(current, mm);
 	reload_context(current);
 }

include/asm-alpha/processor.h

@@ -61,15 +61,6 @@ struct thread_struct {
 	 */
 	unsigned long flags;
 
-	/* The full version of the ASN including serial number.
-
-	   Two threads running on two different processors must of necessity
-	   have different serial numbers.  Having this duplicated from
-	   mm->context allows them to be slightly out of sync preventing 
-	   the asn from incrementing each and every time the two threads
-	   are scheduled.  */
-	unsigned long mm_context;
-
 	/* Perform syscall argument validation (get/set_fs). */
 	mm_segment_t fs;
 
@@ -86,7 +77,7 @@ struct thread_struct {
 	0, 0, 0, \
 	0, 0, 0, \
 	0, 0, 0, \
-	0, 0, \
+	0, \
 	KERNEL_DS \
 }
 

include/asm-alpha/ptrace.h

@@ -67,7 +67,7 @@ struct switch_stack {
 };
 
 #ifdef __KERNEL__
-#define user_mode(regs) ((regs)->ps & 8)
+#define user_mode(regs) (((regs)->ps & 8) != 0)
 #define instruction_pointer(regs) ((regs)->pc)
 extern void show_regs(struct pt_regs *);
 #endif

include/asm-alpha/smp.h

@@ -4,7 +4,6 @@
 #ifdef __SMP__
 
 #include <linux/tasks.h>
-#include <asm/init.h>
 #include <asm/pal.h>
 
 struct cpuinfo_alpha {
@@ -16,7 +15,8 @@ struct cpuinfo_alpha {
 	unsigned long ipi_count;
 	unsigned long prof_multiplier;
 	unsigned long prof_counter;
-} __cacheline_aligned;
+	int irq_count, bh_count;
+} __attribute__((aligned(64)));
 
 extern struct cpuinfo_alpha cpu_data[NR_CPUS];
 

include/asm-alpha/softirq.h

@@ -5,28 +5,33 @@
 #include <asm/atomic.h>
 #include <asm/hardirq.h>
 
-extern unsigned int local_bh_count[NR_CPUS];
+#ifndef __SMP__
+extern int __local_bh_count;
+#define local_bh_count(cpu)	((void)(cpu), __local_bh_count)
+#else
+#define local_bh_count(cpu)	(cpu_data[cpu].bh_count)
+#endif
 
 extern inline void cpu_bh_disable(int cpu)
 {
-	local_bh_count[cpu]++;
+	local_bh_count(cpu)++;
 	mb();
 }
 
 extern inline void cpu_bh_enable(int cpu)
 {
 	mb();
-	local_bh_count[cpu]--;
+	local_bh_count(cpu)--;
 }
 
 extern inline int cpu_bh_trylock(int cpu)
 {
-	return local_bh_count[cpu] ? 0 : (local_bh_count[cpu] = 1);
+	return local_bh_count(cpu) ? 0 : (local_bh_count(cpu) = 1);
 }
 
 extern inline void cpu_bh_endlock(int cpu)
 {
-	local_bh_count[cpu] = 0;
+	local_bh_count(cpu) = 0;
 }
 
 #define local_bh_enable()	cpu_bh_enable(smp_processor_id())

include/asm-alpha/spinlock.h

@@ -65,11 +65,12 @@
   #define SPIN_LOCK_UNLOCKED (spinlock_t) { 0 }
 #endif
 
-#define spin_lock_init(lock)			((void) 0)
-#define spin_lock(lock)				((void) 0)
-#define spin_trylock(lock)			(1)
-#define spin_unlock_wait(lock)			((void) 0)
-#define spin_unlock(lock)			((void) 0)
+#define spin_lock_init(lock)			((void)(lock))
+#define spin_lock(lock)				((void)(lock))
+#define spin_trylock(lock)			((void)(lock), 1)
+#define spin_unlock_wait(lock)			((void)(lock))
+#define spin_unlock(lock)			((void)(lock))
+#define spin_is_locked(lock)			((void)(lock), 0)
 
 /*
  * Read-write spinlocks, allowing multiple readers
@@ -91,10 +92,10 @@
   #define RW_LOCK_UNLOCKED (rwlock_t) { 0 }
 #endif
 
-#define read_lock(lock)				((void) 0)
-#define read_unlock(lock)			((void) 0)
-#define write_lock(lock)			((void) 0)
-#define write_unlock(lock)			((void) 0)
+#define read_lock(lock)				((void)(lock))
+#define read_unlock(lock)			((void)(lock))
+#define write_lock(lock)			((void)(lock))
+#define write_unlock(lock)			((void)(lock))
 
 #else /* __SMP__ */
 
@@ -131,8 +132,8 @@ typedef struct {
 #define spin_lock_init(x)	((x)->lock = 0)
 #endif
 
-#define spin_unlock_wait(x) \
-	({ do { barrier(); } while(((volatile spinlock_t *)x)->lock); })
+#define spin_is_locked(x)	((x)->lock != 0)
+#define spin_unlock_wait(x)	({ do { barrier(); } while ((x)->lock); })
 
 typedef struct { unsigned long a[100]; } __dummy_lock_t;
 #define __dummy_lock(lock) (*(__dummy_lock_t *)(lock))

include/asm-i386/mmu_context.h

@@ -5,9 +5,23 @@
 
 /*
  * get a new mmu context.. x86's don't know much about contexts,
- * but we have to reload the new LDT in exec(). 
+ * but we have to reload the new LDT in exec().
+ *
+ * We implement lazy MMU context-switching on x86 to optimize context
+ * switches done to/from kernel threads. Kernel threads 'inherit' the
+ * previous MM, so Linux doesnt have to flush the TLB. In most cases
+ * we switch back to the same process so we preserve the TLB cache.
+ * This all means that kernel threads have about as much overhead as
+ * a function call ...
  */
-#define get_mmu_context(tsk)	do { } while(0)
+#define get_mmu_context(prev, next) \
+	do { if (next->flags & PF_LAZY_TLB) \
+		{ mmget(prev->mm); next->mm = prev->mm; \
+			next->thread.cr3 = prev->thread.cr3; } } while(0)
+
+#define put_mmu_context(prev, next) \
+	do { if (prev->flags & PF_LAZY_TLB) \
+			{ mmput(prev->mm); } } while(0)
 
 #define init_new_context(mm)	do { } while(0)
 /*

include/asm-i386/pgtable.h

@@ -307,6 +307,9 @@ extern pte_t * __bad_pagetable(void);
 do { \
 	unsigned long __pgdir = __pa(pgdir); \
 	(tsk)->thread.cr3 = __pgdir; \
+	/* do not inherit lazy-TLB after exec() */ \
+	if ((pgdir != swapper_pg_dir) && ((tsk)->flags & PF_LAZY_TLB)) \
+		(tsk)->flags &= ~PF_LAZY_TLB; \
 	if ((tsk) == current) \
 		__asm__ __volatile__("movl %0,%%cr3": :"r" (__pgdir)); \
 } while (0)

include/asm-ppc/machdep.h

@@ -39,6 +39,8 @@ struct machdep_calls {
 	unsigned long	heartbeat_reset;
 	unsigned long	heartbeat_count;
 
+  	void		(*progress)(char *, unsigned short);
+
 	unsigned char 	(*nvram_read_val)(int addr);
 	void		(*nvram_write_val)(int addr, unsigned char val);
 

include/asm-ppc/system.h

@@ -75,8 +75,6 @@ extern void enable_kernel_fp(void);
 extern void cvt_fd(float *from, double *to, unsigned long *fpscr);
 extern void cvt_df(double *from, float *to, unsigned long *fpscr);
 extern int call_rtas(const char *, int, int, unsigned long *, ...);
-extern void chrp_progress(char *);
-void chrp_event_scan(void);
 
 struct device_node;
 extern void note_scsi_host(struct device_node *, void *);

include/linux/mm.h

@@ -128,7 +128,7 @@ typedef struct page {
 	wait_queue_head_t wait;
 	struct page **pprev_hash;
 	struct buffer_head * buffers;
-	int owner; /* temporary debugging check */
+	void *owner; /* temporary debugging check */
 } mem_map_t;
 
 #define get_page(p) do { atomic_inc(&(p)->count); \
@@ -167,11 +167,11 @@ typedef struct page {
 	do { int _ret = test_and_set_bit(PG_locked, &(page)->flags); \
 	if (_ret) PAGE_BUG(page); \
 	if (page->owner) PAGE_BUG(page); \
-	page->owner = (int)current; } while (0)
+	page->owner = current; } while (0)
 #define TryLockPage(page)	({ int _ret = test_and_set_bit(PG_locked, &(page)->flags); \
-				if (!_ret) page->owner = (int)current; _ret; })
+				if (!_ret) page->owner = current; _ret; })
 #define UnlockPage(page)	do { \
-					if (page->owner != (int)current) { \
+					if (page->owner != current) { \
 BUG(); } page->owner = 0; \
 if (!test_and_clear_bit(PG_locked, &(page)->flags)) { \
 			PAGE_BUG(page); } wake_up(&page->wait); } while (0)

include/linux/sched.h

@@ -35,6 +35,7 @@ extern unsigned long event;
 #define CLONE_PID	0x00001000	/* set if pid shared */
 #define CLONE_PTRACE	0x00002000	/* set if we want to let tracing continue on the child too */
 #define CLONE_VFORK	0x00004000	/* set if the parent wants the child to wake it up on mm_release */
+#define CLONE_TLB	0x00008000	/* system thread does lazy TLB flushing (kernel-internal only!) */
 
 /*
  * These are the constant used to fake the fixed-point load-average
@@ -333,6 +334,7 @@ struct task_struct {
 #define PF_SIGNALED	0x00000400	/* killed by a signal */
 #define PF_MEMALLOC	0x00000800	/* Allocating memory */
 #define PF_VFORK	0x00001000	/* Wake up parent in mm_release */
+#define PF_LAZY_TLB	0x00002000	/* thread does lazy TLB switching */
 
 #define PF_USEDFPU	0x00100000	/* task used FPU this quantum (SMP) */
 #define PF_DTRACE	0x00200000	/* delayed trace (used on m68k, i386) */

kernel/fork.c

@@ -536,6 +536,8 @@ static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
 		new_flags &= ~(PF_PTRACED|PF_TRACESYS);
 	if (clone_flags & CLONE_VFORK)
 		new_flags |= PF_VFORK;
+	if ((clone_flags & CLONE_TLB) && capable(CAP_SYS_ADMIN))
+		new_flags |= PF_LAZY_TLB;
 	p->flags = new_flags;
 }
 

kernel/sched.c

@@ -772,8 +772,18 @@ asmlinkage void schedule(void)
 #endif /* __SMP__ */
 
 	kstat.context_swtch++;
-	get_mmu_context(next);
+	/*
+	 * there are 3 processes which are affected by a context switch:
+	 *
+	 * prev == .... ==> (last => next)
+	 *
+	 * It's the 'much more previous' 'prev' that is on next's stack,
+	 * but prev is set to (the just run) 'last' process by switch_to().
+	 * This might sound slightly confusing but makes tons of sense.
+	 */
+	get_mmu_context(prev, next);
 	switch_to(prev, next, prev);
+	put_mmu_context(prev, next);
 	__schedule_tail(prev);
 
 same_process:
@@ -1921,6 +1931,10 @@ static void show_task(struct task_struct * p)
 		printk("%5d ", p->p_cptr->pid);
 	else
 		printk("      ");
+	if (p->flags & PF_LAZY_TLB)
+		printk(" (L-TLB) ");
+	else
+		printk(" (NOTLB) ");
 	if (p->p_ysptr)
 		printk("%7d", p->p_ysptr->pid);
 	else
@@ -2010,5 +2024,11 @@ void __init sched_init(void)
 	init_bh(TIMER_BH, timer_bh);
 	init_bh(TQUEUE_BH, tqueue_bh);
 	init_bh(IMMEDIATE_BH, immediate_bh);
+
+	/*
+	 * The boot idle thread does lazy MMU switching as well:
+	 */
+	mmget(&init_mm);
+	current->flags |= PF_LAZY_TLB;
 }
 

mm/filemap.c

@@ -463,7 +463,7 @@ static inline void __add_to_page_cache(struct page * page,
 
 	flags = page->flags & ~((1 << PG_uptodate) | (1 << PG_error));
 	page->flags = flags |  ((1 << PG_locked) | (1 << PG_referenced));
-	page->owner = (int)current;	/* REMOVEME */
+	page->owner = current;	/* REMOVEME */
 	get_page(page);
 	page->offset = offset;
 	add_page_to_inode_queue(inode, page);
@@ -1855,7 +1855,7 @@ generic_file_write(struct file *file, const char *buf,
 		if (!PageLocked(page)) {
 			PAGE_BUG(page);
 		} else {
-			if (page->owner != (int)current) {
+			if (page->owner != current) {
 				PAGE_BUG(page);
 			}
 		}

mm/mremap.c

@@ -118,7 +118,7 @@ static int move_page_tables(struct mm_struct * mm,
 	flush_cache_range(mm, new_addr, new_addr + len);
 	while ((offset += PAGE_SIZE) < len)
 		move_one_page(mm, new_addr + offset, old_addr + offset);
-	zap_page_range(mm, new_addr, new_addr + len);
+	zap_page_range(mm, new_addr, len);
 	flush_tlb_range(mm, new_addr, new_addr + len);
 	return -1;
 }