Merge master.kernel.org:/pub/scm/linux/kernel/git/willy/parisc-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/willy/parisc-2.6: (41 commits)
  [PARISC] Kill wall_jiffies use
  [PARISC] Honour "panic_on_oops" sysctl
  [PARISC] Fix fs/binfmt_som.c
  [PARISC] Export clear_user_page to modules
  [PARISC] Make DMA routines more stubby
  [PARISC] Define pci_get_legacy_ide_irq
  [PARISC] Fix CONFIG_DEBUG_SPINLOCK
  [PARISC] Fix HPUX compat compile with current GCC
  [PARISC] Fix iounmap compile warning
  [PARISC] Add support for Quicksilver AGPGART
  [PARISC] Move LBA and SBA register defines to the common ropes.h
  [PARISC] Create shared <asm/ropes.h> header
  [PARISC] Stash the lba_device in its struct device drvdata
  [PARISC] Generalize IS_ASTRO et al to take a parisc_device like
  [PARISC] Pretty print the name of the lba type on kernel boot
  [PARISC] Remove some obsolete comments and I checked that Reo is similar to Ike
  [PARISC] Add hardware found in the rp8400
  [PARISC] Allow nested interrupts
  [PARISC] Further updates to timer_interrupt()
  [PARISC] remove halftick and copy clocktick to local var (gcc can optimize usage)
  ...

arch/parisc/Kconfig

@@ -127,7 +127,7 @@ config PA11
 
 config PREFETCH
 	def_bool y
-	depends on PA8X00
+	depends on PA8X00 || PA7200
 
 config 64BIT
 	bool "64-bit kernel"

arch/parisc/hpux/fs.c

@@ -96,7 +96,7 @@ static int filldir(void * __buf, const char * name, int namlen, loff_t offset,
 	put_user(namlen, &dirent->d_namlen);
 	copy_to_user(dirent->d_name, name, namlen);
 	put_user(0, dirent->d_name + namlen);
-	((char *) dirent) += reclen;
+	dirent = (void __user *)dirent + reclen;
 	buf->current_dir = dirent;
 	buf->count -= reclen;
 	return 0;

arch/parisc/kernel/binfmt_elf32.c

@@ -87,7 +87,7 @@ struct elf_prpsinfo32
  */
 
 #define SET_PERSONALITY(ex, ibcs2) \
-	current->personality = PER_LINUX32; \
+	set_thread_flag(TIF_32BIT); \
 	current->thread.map_base = DEFAULT_MAP_BASE32; \
 	current->thread.task_size = DEFAULT_TASK_SIZE32 \
 
@@ -102,25 +102,3 @@ cputime_to_compat_timeval(const cputime_t cputime, struct compat_timeval *value)
 }
 
 #include "../../../fs/binfmt_elf.c"
-
-/* Set up a separate execution domain for ELF32 binaries running
- * on an ELF64 kernel */
-
-static struct exec_domain parisc32_exec_domain = { 
-	.name = "Linux/ELF32",
-	.pers_low = PER_LINUX32,
-	.pers_high = PER_LINUX32,
-};      
-
-static int __init parisc32_exec_init(void)
-{
-	/* steal the identity signal mappings from the default domain */
-	parisc32_exec_domain.signal_map = default_exec_domain.signal_map;
-	parisc32_exec_domain.signal_invmap = default_exec_domain.signal_invmap;
-
-	register_exec_domain(&parisc32_exec_domain);
-
-	return 0;
-}
-
-__initcall(parisc32_exec_init);

arch/parisc/kernel/cache.c

@@ -35,15 +35,12 @@ int icache_stride __read_mostly;
 EXPORT_SYMBOL(dcache_stride);
 
 
-#if defined(CONFIG_SMP)
 /* On some machines (e.g. ones with the Merced bus), there can be
  * only a single PxTLB broadcast at a time; this must be guaranteed
  * by software.  We put a spinlock around all TLB flushes  to
  * ensure this.
  */
 DEFINE_SPINLOCK(pa_tlb_lock);
-EXPORT_SYMBOL(pa_tlb_lock);
-#endif
 
 struct pdc_cache_info cache_info __read_mostly;
 #ifndef CONFIG_PA20
@@ -91,7 +88,8 @@ update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t pte)
 
 		flush_kernel_dcache_page(page);
 		clear_bit(PG_dcache_dirty, &page->flags);
-	}
+	} else if (parisc_requires_coherency())
+		flush_kernel_dcache_page(page);
 }
 
 void
@@ -370,3 +368,45 @@ void parisc_setup_cache_timing(void)
 
 	printk(KERN_INFO "Setting cache flush threshold to %x (%d CPUs online)\n", parisc_cache_flush_threshold, num_online_cpus());
 }
+
+extern void purge_kernel_dcache_page(unsigned long);
+extern void clear_user_page_asm(void *page, unsigned long vaddr);
+
+void clear_user_page(void *page, unsigned long vaddr, struct page *pg)
+{
+	purge_kernel_dcache_page((unsigned long)page);
+	purge_tlb_start();
+	pdtlb_kernel(page);
+	purge_tlb_end();
+	clear_user_page_asm(page, vaddr);
+}
+EXPORT_SYMBOL(clear_user_page);
+
+void flush_kernel_dcache_page_addr(void *addr)
+{
+	flush_kernel_dcache_page_asm(addr);
+	purge_tlb_start();
+	pdtlb_kernel(addr);
+	purge_tlb_end();
+}
+EXPORT_SYMBOL(flush_kernel_dcache_page_addr);
+
+void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
+		    struct page *pg)
+{
+	/* no coherency needed (all in kmap/kunmap) */
+	copy_user_page_asm(vto, vfrom);
+	if (!parisc_requires_coherency())
+		flush_kernel_dcache_page_asm(vto);
+}
+EXPORT_SYMBOL(copy_user_page);
+
+#ifdef CONFIG_PA8X00
+
+void kunmap_parisc(void *addr)
+{
+	if (parisc_requires_coherency())
+		flush_kernel_dcache_page_addr(addr);
+}
+EXPORT_SYMBOL(kunmap_parisc);
+#endif

arch/parisc/kernel/entry.S

@@ -30,6 +30,7 @@
 
 
 #include <asm/psw.h>
+#include <asm/cache.h>		/* for L1_CACHE_SHIFT */
 #include <asm/assembly.h>	/* for LDREG/STREG defines */
 #include <asm/pgtable.h>
 #include <asm/signal.h>
@@ -478,11 +479,7 @@
 	bb,>=,n		\pmd,_PxD_PRESENT_BIT,\fault
 	DEP		%r0,31,PxD_FLAG_SHIFT,\pmd /* clear flags */
 	copy		\pmd,%r9
-#ifdef CONFIG_64BIT
-	shld		%r9,PxD_VALUE_SHIFT,\pmd
-#else
-	shlw		%r9,PxD_VALUE_SHIFT,\pmd
-#endif
+	SHLREG		%r9,PxD_VALUE_SHIFT,\pmd
 	EXTR		\va,31-PAGE_SHIFT,ASM_BITS_PER_PTE,\index
 	DEP		%r0,31,PAGE_SHIFT,\pmd  /* clear offset */
 	shladd		\index,BITS_PER_PTE_ENTRY,\pmd,\pmd
@@ -970,11 +967,7 @@ intr_return:
 	/* shift left ____cacheline_aligned (aka L1_CACHE_BYTES) amount
 	** irq_stat[] is defined using ____cacheline_aligned.
 	*/
-#ifdef CONFIG_64BIT
-	shld	%r1, 6, %r20
-#else
-	shlw	%r1, 5, %r20
-#endif
+	SHLREG	%r1,L1_CACHE_SHIFT,%r20
 	add     %r19,%r20,%r19	/* now have &irq_stat[smp_processor_id()] */
 #endif /* CONFIG_SMP */
 
@@ -1076,7 +1069,7 @@ intr_do_preempt:
 	BL	preempt_schedule_irq, %r2
 	nop
 
-	b	intr_restore		/* ssm PSW_SM_I done by intr_restore */
+	b,n	intr_restore		/* ssm PSW_SM_I done by intr_restore */
 #endif /* CONFIG_PREEMPT */
 
 	.import do_signal,code
@@ -2115,11 +2108,7 @@ syscall_check_bh:
 	ldw     TI_CPU-THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r26 /* cpu # */
 
 	/* shift left ____cacheline_aligned (aka L1_CACHE_BYTES) bits */
-#ifdef CONFIG_64BIT
-	shld	%r26, 6, %r20
-#else
-	shlw	%r26, 5, %r20
-#endif
+	SHLREG	%r26,L1_CACHE_SHIFT,%r20
 	add     %r19,%r20,%r19	/* now have &irq_stat[smp_processor_id()] */
 #endif /* CONFIG_SMP */
 

arch/parisc/kernel/hardware.c

@@ -231,6 +231,7 @@ static struct hp_hardware hp_hardware_list[] __initdata = {
 	{HPHW_NPROC,0x5E6,0x4,0x91,"Keystone/Matterhorn W2 650"},
 	{HPHW_NPROC,0x5E7,0x4,0x91,"Caribe W2 800"},
 	{HPHW_NPROC,0x5E8,0x4,0x91,"Pikes Peak W2"},
+	{HPHW_NPROC,0x5EB,0x4,0x91,"Perf/Leone 875 W2+"},
 	{HPHW_NPROC,0x5FF,0x4,0x91,"Hitachi W"},
 	{HPHW_NPROC,0x600,0x4,0x81,"Gecko (712/60)"},
 	{HPHW_NPROC,0x601,0x4,0x81,"Gecko 80 (712/80)"},
@@ -584,8 +585,10 @@ static struct hp_hardware hp_hardware_list[] __initdata = {
 	{HPHW_CONSOLE, 0x01A, 0x0001F, 0x00, "Jason/Anole 64 Null Console"}, 
 	{HPHW_CONSOLE, 0x01B, 0x0001F, 0x00, "Jason/Anole 100 Null Console"}, 
 	{HPHW_FABRIC, 0x004, 0x000AA, 0x80, "Halfdome DNA Central Agent"}, 
+	{HPHW_FABRIC, 0x005, 0x000AA, 0x80, "Keystone DNA Central Agent"},
 	{HPHW_FABRIC, 0x007, 0x000AA, 0x80, "Caribe DNA Central Agent"}, 
 	{HPHW_FABRIC, 0x004, 0x000AB, 0x00, "Halfdome TOGO Fabric Crossbar"}, 
+	{HPHW_FABRIC, 0x005, 0x000AB, 0x00, "Keystone TOGO Fabric Crossbar"},
 	{HPHW_FABRIC, 0x004, 0x000AC, 0x00, "Halfdome Sakura Fabric Router"}, 
 	{HPHW_FIO, 0x025, 0x0002E, 0x80, "Armyknife Optional X.25"}, 
 	{HPHW_FIO, 0x004, 0x0004F, 0x0, "8-Port X.25 EISA-ACC (AMSO)"}, 

arch/parisc/kernel/irq.c

@@ -45,6 +45,17 @@ extern irqreturn_t ipi_interrupt(int, void *, struct pt_regs *);
 */
 static volatile unsigned long cpu_eiem = 0;
 
+/*
+** ack bitmap ... habitually set to 1, but reset to zero
+** between ->ack() and ->end() of the interrupt to prevent
+** re-interruption of a processing interrupt.
+*/
+static volatile unsigned long global_ack_eiem = ~0UL;
+/*
+** Local bitmap, same as above but for per-cpu interrupts
+*/
+static DEFINE_PER_CPU(unsigned long, local_ack_eiem) = ~0UL;
+
 static void cpu_disable_irq(unsigned int irq)
 {
 	unsigned long eirr_bit = EIEM_MASK(irq);
@@ -62,13 +73,6 @@ static void cpu_enable_irq(unsigned int irq)
 
 	cpu_eiem |= eirr_bit;
 
-	/* FIXME: while our interrupts aren't nested, we cannot reset
-	 * the eiem mask if we're already in an interrupt.  Once we
-	 * implement nested interrupts, this can go away
-	 */
-	if (!in_interrupt())
-		set_eiem(cpu_eiem);
-
 	/* This is just a simple NOP IPI.  But what it does is cause
 	 * all the other CPUs to do a set_eiem(cpu_eiem) at the end
 	 * of the interrupt handler */
@@ -84,13 +88,45 @@ static unsigned int cpu_startup_irq(unsigned int irq)
 void no_ack_irq(unsigned int irq) { }
 void no_end_irq(unsigned int irq) { }
 
+void cpu_ack_irq(unsigned int irq)
+{
+	unsigned long mask = EIEM_MASK(irq);
+	int cpu = smp_processor_id();
+
+	/* Clear in EIEM so we can no longer process */
+	if (CHECK_IRQ_PER_CPU(irq_desc[irq].status))
+		per_cpu(local_ack_eiem, cpu) &= ~mask;
+	else
+		global_ack_eiem &= ~mask;
+
+	/* disable the interrupt */
+	set_eiem(cpu_eiem & global_ack_eiem & per_cpu(local_ack_eiem, cpu));
+	/* and now ack it */
+	mtctl(mask, 23);
+}
+
+void cpu_end_irq(unsigned int irq)
+{
+	unsigned long mask = EIEM_MASK(irq);
+	int cpu = smp_processor_id();
+
+	/* set it in the eiems---it's no longer in process */
+	if (CHECK_IRQ_PER_CPU(irq_desc[irq].status))
+		per_cpu(local_ack_eiem, cpu) |= mask;
+	else
+		global_ack_eiem |= mask;
+
+	/* enable the interrupt */
+	set_eiem(cpu_eiem & global_ack_eiem & per_cpu(local_ack_eiem, cpu));
+}
+
 #ifdef CONFIG_SMP
 int cpu_check_affinity(unsigned int irq, cpumask_t *dest)
 {
 	int cpu_dest;
 
 	/* timer and ipi have to always be received on all CPUs */
-	if (irq == TIMER_IRQ || irq == IPI_IRQ) {
+	if (CHECK_IRQ_PER_CPU(irq)) {
 		/* Bad linux design decision.  The mask has already
 		 * been set; we must reset it */
 		irq_desc[irq].affinity = CPU_MASK_ALL;
@@ -119,8 +155,8 @@ static struct hw_interrupt_type cpu_interrupt_type = {
 	.shutdown	= cpu_disable_irq,
 	.enable		= cpu_enable_irq,
 	.disable	= cpu_disable_irq,
-	.ack		= no_ack_irq,
-	.end		= no_end_irq,
+	.ack		= cpu_ack_irq,
+	.end		= cpu_end_irq,
 #ifdef CONFIG_SMP
 	.set_affinity	= cpu_set_affinity_irq,
 #endif
@@ -209,7 +245,7 @@ int show_interrupts(struct seq_file *p, void *v)
 ** Then use that to get the Transaction address and data.
 */
 
-int cpu_claim_irq(unsigned int irq, struct hw_interrupt_type *type, void *data)
+int cpu_claim_irq(unsigned int irq, struct irq_chip *type, void *data)
 {
 	if (irq_desc[irq].action)
 		return -EBUSY;
@@ -298,82 +334,69 @@ unsigned int txn_alloc_data(unsigned int virt_irq)
 	return virt_irq - CPU_IRQ_BASE;
 }
 
+static inline int eirr_to_irq(unsigned long eirr)
+{
+#ifdef CONFIG_64BIT
+	int bit = fls64(eirr);
+#else
+	int bit = fls(eirr);
+#endif
+	return (BITS_PER_LONG - bit) + TIMER_IRQ;
+}
+
 /* ONLY called from entry.S:intr_extint() */
 void do_cpu_irq_mask(struct pt_regs *regs)
 {
 	unsigned long eirr_val;
-
-	irq_enter();
-
-	/*
-	 * Don't allow TIMER or IPI nested interrupts.
-	 * Allowing any single interrupt to nest can lead to that CPU
-	 * handling interrupts with all enabled interrupts unmasked.
-	 */
-	set_eiem(0UL);
-
-	/* 1) only process IRQs that are enabled/unmasked (cpu_eiem)
-	 * 2) We loop here on EIRR contents in order to avoid
-	 *    nested interrupts or having to take another interrupt
-	 *    when we could have just handled it right away.
-	 */
-	for (;;) {
-		unsigned long bit = (1UL << (BITS_PER_LONG - 1));
-		unsigned int irq;
-		eirr_val = mfctl(23) & cpu_eiem;
-		if (!eirr_val)
-			break;
-
-		mtctl(eirr_val, 23); /* reset bits we are going to process */
-
-		/* Work our way from MSb to LSb...same order we alloc EIRs */
-		for (irq = TIMER_IRQ; eirr_val && bit; bit>>=1, irq++) {
+	int irq, cpu = smp_processor_id();
 #ifdef CONFIG_SMP
-			cpumask_t dest = irq_desc[irq].affinity;
+	cpumask_t dest;
 #endif
-			if (!(bit & eirr_val))
-				continue;
 
-			/* clear bit in mask - can exit loop sooner */
-			eirr_val &= ~bit;
+	local_irq_disable();
+	irq_enter();
 
-#ifdef CONFIG_SMP
-			/* FIXME: because generic set affinity mucks
-			 * with the affinity before sending it to us
-			 * we can get the situation where the affinity is
-			 * wrong for our CPU type interrupts */
-			if (irq != TIMER_IRQ && irq != IPI_IRQ &&
-			    !cpu_isset(smp_processor_id(), dest)) {
-				int cpu = first_cpu(dest);
-
-				printk(KERN_DEBUG "redirecting irq %d from CPU %d to %d\n",
-				       irq, smp_processor_id(), cpu);
-				gsc_writel(irq + CPU_IRQ_BASE,
-					   cpu_data[cpu].hpa);
-				continue;
-			}
-#endif
+	eirr_val = mfctl(23) & cpu_eiem & global_ack_eiem &
+		per_cpu(local_ack_eiem, cpu);
+	if (!eirr_val)
+		goto set_out;
+	irq = eirr_to_irq(eirr_val);
 
-			__do_IRQ(irq, regs);
-		}
+#ifdef CONFIG_SMP
+	dest = irq_desc[irq].affinity;
+	if (CHECK_IRQ_PER_CPU(irq_desc[irq].status) &&
+	    !cpu_isset(smp_processor_id(), dest)) {
+		int cpu = first_cpu(dest);
+
+		printk(KERN_DEBUG "redirecting irq %d from CPU %d to %d\n",
+		       irq, smp_processor_id(), cpu);
+		gsc_writel(irq + CPU_IRQ_BASE,
+			   cpu_data[cpu].hpa);
+		goto set_out;
 	}
+#endif
+	__do_IRQ(irq, regs);
 
-	set_eiem(cpu_eiem);	/* restore original mask */
+ out:
 	irq_exit();
-}
+	return;
 
+ set_out:
+	set_eiem(cpu_eiem & global_ack_eiem & per_cpu(local_ack_eiem, cpu));
+	goto out;
+}
 
 static struct irqaction timer_action = {
 	.handler = timer_interrupt,
 	.name = "timer",
-	.flags = IRQF_DISABLED,
+	.flags = IRQF_DISABLED | IRQF_TIMER | IRQF_PERCPU,
 };
 
 #ifdef CONFIG_SMP
 static struct irqaction ipi_action = {
 	.handler = ipi_interrupt,
 	.name = "IPI",
-	.flags = IRQF_DISABLED,
+	.flags = IRQF_DISABLED | IRQF_PERCPU,
 };
 #endif
 

arch/parisc/kernel/processor.c

@@ -143,8 +143,9 @@ static int __init processor_probe(struct parisc_device *dev)
 	p = &cpu_data[cpuid];
 	boot_cpu_data.cpu_count++;
 
-	/* initialize counters */
-	memset(p, 0, sizeof(struct cpuinfo_parisc));
+	/* initialize counters - CPU 0 gets it_value set in time_init() */
+	if (cpuid)
+		memset(p, 0, sizeof(struct cpuinfo_parisc));
 
 	p->loops_per_jiffy = loops_per_jiffy;
 	p->dev = dev;		/* Save IODC data in case we need it */

arch/parisc/kernel/signal.c

@@ -26,7 +26,6 @@
 #include <linux/stddef.h>
 #include <linux/compat.h>
 #include <linux/elf.h>
-#include <linux/personality.h>
 #include <asm/ucontext.h>
 #include <asm/rt_sigframe.h>
 #include <asm/uaccess.h>
@@ -433,13 +432,13 @@ setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
 	if (in_syscall) {
 		regs->gr[31] = haddr;
 #ifdef __LP64__
-		if (personality(current->personality) == PER_LINUX)
+		if (!test_thread_flag(TIF_32BIT))
 			sigframe_size |= 1;
 #endif
 	} else {
 		unsigned long psw = USER_PSW;
 #ifdef __LP64__
-		if (personality(current->personality) == PER_LINUX)
+		if (!test_thread_flag(TIF_32BIT))
 			psw |= PSW_W;
 #endif
 

arch/parisc/kernel/smp.c

@@ -262,6 +262,9 @@ ipi_interrupt(int irq, void *dev_id, struct pt_regs *regs)
 					this_cpu, which);
 				return IRQ_NONE;
 			} /* Switch */
+		/* let in any pending interrupts */
+		local_irq_enable();
+		local_irq_disable();
 		} /* while (ops) */
 	}
 	return IRQ_HANDLED;
@@ -430,8 +433,9 @@ smp_do_timer(struct pt_regs *regs)
 static void __init
 smp_cpu_init(int cpunum)
 {
-	extern int init_per_cpu(int);  /* arch/parisc/kernel/setup.c */
+	extern int init_per_cpu(int);  /* arch/parisc/kernel/processor.c */
 	extern void init_IRQ(void);    /* arch/parisc/kernel/irq.c */
+	extern void start_cpu_itimer(void); /* arch/parisc/kernel/time.c */
 
 	/* Set modes and Enable floating point coprocessor */
 	(void) init_per_cpu(cpunum);
@@ -457,6 +461,7 @@ smp_cpu_init(int cpunum)
 	enter_lazy_tlb(&init_mm, current);
 
 	init_IRQ();   /* make sure no IRQ's are enabled or pending */
+	start_cpu_itimer();
 }
 
 

arch/parisc/kernel/sys_parisc.c

@@ -31,6 +31,8 @@
 #include <linux/shm.h>
 #include <linux/smp_lock.h>
 #include <linux/syscalls.h>
+#include <linux/utsname.h>
+#include <linux/personality.h>
 
 int sys_pipe(int __user *fildes)
 {
@@ -248,3 +250,46 @@ asmlinkage int sys_free_hugepages(unsigned long addr)
 {
 	return -EINVAL;
 }
+
+long parisc_personality(unsigned long personality)
+{
+	long err;
+
+	if (personality(current->personality) == PER_LINUX32
+	    && personality == PER_LINUX)
+		personality = PER_LINUX32;
+
+	err = sys_personality(personality);
+	if (err == PER_LINUX32)
+		err = PER_LINUX;
+
+	return err;
+}
+
+static inline int override_machine(char __user *mach) {
+#ifdef CONFIG_COMPAT
+	if (personality(current->personality) == PER_LINUX32) {
+		if (__put_user(0, mach + 6) ||
+		    __put_user(0, mach + 7))
+			return -EFAULT;
+	}
+
+	return 0;
+#else /*!CONFIG_COMPAT*/
+	return 0;
+#endif /*CONFIG_COMPAT*/
+}
+
+long parisc_newuname(struct new_utsname __user *utsname)
+{
+	int err = 0;
+
+	down_read(&uts_sem);
+	if (copy_to_user(utsname, &system_utsname, sizeof(*utsname)))
+		err = -EFAULT;
+	up_read(&uts_sem);
+
+	err = override_machine(utsname->machine);
+
+	return (long)err;
+}

arch/parisc/kernel/syscall_table.S

@@ -132,7 +132,7 @@
 	ENTRY_SAME(socketpair)
 	ENTRY_SAME(setpgid)
 	ENTRY_SAME(send)
-	ENTRY_SAME(newuname)
+	ENTRY_OURS(newuname)
 	ENTRY_SAME(umask)		/* 60 */
 	ENTRY_SAME(chroot)
 	ENTRY_SAME(ustat)
@@ -221,7 +221,7 @@
 	ENTRY_SAME(fchdir)
 	ENTRY_SAME(bdflush)
 	ENTRY_SAME(sysfs)		/* 135 */
-	ENTRY_SAME(personality)
+	ENTRY_OURS(personality)
 	ENTRY_SAME(ni_syscall)	/* for afs_syscall */
 	ENTRY_SAME(setfsuid)
 	ENTRY_SAME(setfsgid)

arch/parisc/kernel/time.c

@@ -32,8 +32,7 @@
 
 #include <linux/timex.h>
 
-static long clocktick __read_mostly;	/* timer cycles per tick */
-static long halftick __read_mostly;
+static unsigned long clocktick __read_mostly;	/* timer cycles per tick */
 
 #ifdef CONFIG_SMP
 extern void smp_do_timer(struct pt_regs *regs);
@@ -41,46 +40,106 @@ extern void smp_do_timer(struct pt_regs *regs);
 
 irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
 {
-	long now;
-	long next_tick;
-	int nticks;
-	int cpu = smp_processor_id();
+	unsigned long now;
+	unsigned long next_tick;
+	unsigned long cycles_elapsed;
+	unsigned long cycles_remainder;
+	unsigned int cpu = smp_processor_id();
+
+	/* gcc can optimize for "read-only" case with a local clocktick */
+	unsigned long cpt = clocktick;
 
 	profile_tick(CPU_PROFILING, regs);
 
-	now = mfctl(16);
-	/* initialize next_tick to time at last clocktick */
+	/* Initialize next_tick to the expected tick time. */
 	next_tick = cpu_data[cpu].it_value;
 
-	/* since time passes between the interrupt and the mfctl()
-	 * above, it is never true that last_tick + clocktick == now.  If we
-	 * never miss a clocktick, we could set next_tick = last_tick + clocktick
-	 * but maybe we'll miss ticks, hence the loop.
-	 *
-	 * Variables are *signed*.
+	/* Get current interval timer.
+	 * CR16 reads as 64 bits in CPU wide mode.
+	 * CR16 reads as 32 bits in CPU narrow mode.
 	 */
+	now = mfctl(16);
+
+	cycles_elapsed = now - next_tick;
 
-	nticks = 0;
-	while((next_tick - now) < halftick) {
-		next_tick += clocktick;
-		nticks++;
+	if ((cycles_elapsed >> 5) < cpt) {
+		/* use "cheap" math (add/subtract) instead
+		 * of the more expensive div/mul method
+		 */
+		cycles_remainder = cycles_elapsed;
+		while (cycles_remainder > cpt) {
+			cycles_remainder -= cpt;
+		}
+	} else {
+		cycles_remainder = cycles_elapsed % cpt;
 	}
-	mtctl(next_tick, 16);
+
+	/* Can we differentiate between "early CR16" (aka Scenario 1) and
+	 * "long delay" (aka Scenario 3)? I don't think so.
+	 *
+	 * We expected timer_interrupt to be delivered at least a few hundred
+	 * cycles after the IT fires. But it's arbitrary how much time passes
+	 * before we call it "late". I've picked one second.
+	 */
+/* aproximate HZ with shifts. Intended math is "(elapsed/clocktick) > HZ" */
+#if HZ == 1000
+	if (cycles_elapsed > (cpt << 10) )
+#elif HZ == 250
+	if (cycles_elapsed > (cpt << 8) )
+#elif HZ == 100
+	if (cycles_elapsed > (cpt << 7) )
+#else
+#warn WTF is HZ set to anyway?
+	if (cycles_elapsed > (HZ * cpt) )
+#endif
+	{
+		/* Scenario 3: very long delay?  bad in any case */
+		printk (KERN_CRIT "timer_interrupt(CPU %d): delayed!"
+			" cycles %lX rem %lX "
+			" next/now %lX/%lX\n",
+			cpu,
+			cycles_elapsed, cycles_remainder,
+			next_tick, now );
+	}
+
+	/* convert from "division remainder" to "remainder of clock tick" */
+	cycles_remainder = cpt - cycles_remainder;
+
+	/* Determine when (in CR16 cycles) next IT interrupt will fire.
+	 * We want IT to fire modulo clocktick even if we miss/skip some.
+	 * But those interrupts don't in fact get delivered that regularly.
+	 */
+	next_tick = now + cycles_remainder;
+
 	cpu_data[cpu].it_value = next_tick;
 
-	while (nticks--) {
+	/* Skip one clocktick on purpose if we are likely to miss next_tick.
+	 * We want to avoid the new next_tick being less than CR16.
+	 * If that happened, itimer wouldn't fire until CR16 wrapped.
+	 * We'll catch the tick we missed on the tick after that.
+	 */
+	if (!(cycles_remainder >> 13))
+		next_tick += cpt;
+
+	/* Program the IT when to deliver the next interrupt. */
+        /* Only bottom 32-bits of next_tick are written to cr16.  */
+	mtctl(next_tick, 16);
+
+
+	/* Done mucking with unreliable delivery of interrupts.
+	 * Go do system house keeping.
+	 */
 #ifdef CONFIG_SMP
-		smp_do_timer(regs);
+	smp_do_timer(regs);
 #else
-		update_process_times(user_mode(regs));
+	update_process_times(user_mode(regs));
 #endif
-		if (cpu == 0) {
-			write_seqlock(&xtime_lock);
-			do_timer(1);
-			write_sequnlock(&xtime_lock);
-		}
+	if (cpu == 0) {
+		write_seqlock(&xtime_lock);
+		do_timer(regs);
+		write_sequnlock(&xtime_lock);
 	}
-    
+
 	/* check soft power switch status */
 	if (cpu == 0 && !atomic_read(&power_tasklet.count))
 		tasklet_schedule(&power_tasklet);
@@ -106,14 +165,12 @@ unsigned long profile_pc(struct pt_regs *regs)
 EXPORT_SYMBOL(profile_pc);
 
 
-/*** converted from ia64 ***/
 /*
  * Return the number of micro-seconds that elapsed since the last
  * update to wall time (aka xtime).  The xtime_lock
  * must be at least read-locked when calling this routine.
  */
-static inline unsigned long
-gettimeoffset (void)
+static inline unsigned long gettimeoffset (void)
 {
 #ifndef CONFIG_SMP
 	/*
@@ -121,21 +178,44 @@ gettimeoffset (void)
 	 *    Once parisc-linux learns the cr16 difference between processors,
 	 *    this could be made to work.
 	 */
-	long last_tick;
-	long elapsed_cycles;
-
-	/* it_value is the intended time of the next tick */
-	last_tick = cpu_data[smp_processor_id()].it_value;
-
-	/* Subtract one tick and account for possible difference between
-	 * when we expected the tick and when it actually arrived.
-	 * (aka wall vs real)
-	 */
-	last_tick -= clocktick * (jiffies - wall_jiffies + 1);
-	elapsed_cycles = mfctl(16) - last_tick;
+	unsigned long now;
+	unsigned long prev_tick;
+	unsigned long next_tick;
+	unsigned long elapsed_cycles;
+	unsigned long usec;
+	unsigned long cpuid = smp_processor_id();
+	unsigned long cpt = clocktick;
+
+	next_tick = cpu_data[cpuid].it_value;
+	now = mfctl(16);	/* Read the hardware interval timer.  */
+
+	prev_tick = next_tick - cpt;
+
+	/* Assume Scenario 1: "now" is later than prev_tick.  */
+	elapsed_cycles = now - prev_tick;
+
+/* aproximate HZ with shifts. Intended math is "(elapsed/clocktick) > HZ" */
+#if HZ == 1000
+	if (elapsed_cycles > (cpt << 10) )
+#elif HZ == 250
+	if (elapsed_cycles > (cpt << 8) )
+#elif HZ == 100
+	if (elapsed_cycles > (cpt << 7) )
+#else
+#warn WTF is HZ set to anyway?
+	if (elapsed_cycles > (HZ * cpt) )
+#endif
+	{
+		/* Scenario 3: clock ticks are missing. */
+		printk (KERN_CRIT "gettimeoffset(CPU %ld): missing %ld ticks!"
+			" cycles %lX prev/now/next %lX/%lX/%lX  clock %lX\n",
+			cpuid, elapsed_cycles / cpt,
+			elapsed_cycles, prev_tick, now, next_tick, cpt);
+	}
 
-	/* the precision of this math could be improved */
-	return elapsed_cycles / (PAGE0->mem_10msec / 10000);
+	/* FIXME: Can we improve the precision? Not with PAGE0. */
+	usec = (elapsed_cycles * 10000) / PAGE0->mem_10msec;
+	return usec;
 #else
 	return 0;
 #endif
@@ -146,6 +226,7 @@ do_gettimeofday (struct timeval *tv)
 {
 	unsigned long flags, seq, usec, sec;
 
+	/* Hold xtime_lock and adjust timeval.  */
 	do {
 		seq = read_seqbegin_irqsave(&xtime_lock, flags);
 		usec = gettimeoffset();
@@ -153,25 +234,13 @@ do_gettimeofday (struct timeval *tv)
 		usec += (xtime.tv_nsec / 1000);
 	} while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
 
-	if (unlikely(usec > LONG_MAX)) {
-		/* This can happen if the gettimeoffset adjustment is
-		 * negative and xtime.tv_nsec is smaller than the
-		 * adjustment */
-		printk(KERN_ERR "do_gettimeofday() spurious xtime.tv_nsec of %ld\n", usec);
-		usec += USEC_PER_SEC;
-		--sec;
-		/* This should never happen, it means the negative
-		 * time adjustment was more than a second, so there's
-		 * something seriously wrong */
-		BUG_ON(usec > LONG_MAX);
-	}
-
-
+	/* Move adjusted usec's into sec's.  */
 	while (usec >= USEC_PER_SEC) {
 		usec -= USEC_PER_SEC;
 		++sec;
 	}
 
+	/* Return adjusted result.  */
 	tv->tv_sec = sec;
 	tv->tv_usec = usec;
 }
@@ -223,22 +292,23 @@ unsigned long long sched_clock(void)
 }
 
 
+void __init start_cpu_itimer(void)
+{
+	unsigned int cpu = smp_processor_id();
+	unsigned long next_tick = mfctl(16) + clocktick;
+
+	mtctl(next_tick, 16);		/* kick off Interval Timer (CR16) */
+
+	cpu_data[cpu].it_value = next_tick;
+}
+
 void __init time_init(void)
 {
-	unsigned long next_tick;
 	static struct pdc_tod tod_data;
 
 	clocktick = (100 * PAGE0->mem_10msec) / HZ;
-	halftick = clocktick / 2;
 
-	/* Setup clock interrupt timing */
-
-	next_tick = mfctl(16);
-	next_tick += clocktick;
-	cpu_data[smp_processor_id()].it_value = next_tick;
-
-	/* kick off Itimer (CR16) */
-	mtctl(next_tick, 16);
+	start_cpu_itimer();	/* get CPU 0 started */
 
 	if(pdc_tod_read(&tod_data) == 0) {
 		write_seqlock_irq(&xtime_lock);

arch/parisc/kernel/traps.c

@@ -16,6 +16,7 @@
 #include <linux/errno.h>
 #include <linux/ptrace.h>
 #include <linux/timer.h>
+#include <linux/delay.h>
 #include <linux/mm.h>
 #include <linux/module.h>
 #include <linux/smp.h>
@@ -245,6 +246,15 @@ void die_if_kernel(char *str, struct pt_regs *regs, long err)
 		current->comm, current->pid, str, err);
 	show_regs(regs);
 
+	if (in_interrupt())
+		panic("Fatal exception in interrupt");
+
+	if (panic_on_oops) {
+		printk(KERN_EMERG "Fatal exception: panic in 5 seconds\n");
+		ssleep(5);
+		panic("Fatal exception");
+	}
+
 	/* Wot's wrong wif bein' racy? */
 	if (current->thread.flags & PARISC_KERNEL_DEATH) {
 		printk(KERN_CRIT "%s() recursion detected.\n", __FUNCTION__);

arch/parisc/mm/init.c

@@ -31,10 +31,7 @@
 
 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 
-extern char _text;	/* start of kernel code, defined by linker */
 extern int  data_start;
-extern char _end;	/* end of BSS, defined by linker */
-extern char __init_begin, __init_end;
 
 #ifdef CONFIG_DISCONTIGMEM
 struct node_map_data node_data[MAX_NUMNODES] __read_mostly;
@@ -319,8 +316,8 @@ static void __init setup_bootmem(void)
 
 	reserve_bootmem_node(NODE_DATA(0), 0UL,
 			(unsigned long)(PAGE0->mem_free + PDC_CONSOLE_IO_IODC_SIZE));
-	reserve_bootmem_node(NODE_DATA(0),__pa((unsigned long)&_text),
-			(unsigned long)(&_end - &_text));
+	reserve_bootmem_node(NODE_DATA(0), __pa((unsigned long)_text),
+			(unsigned long)(_end - _text));
 	reserve_bootmem_node(NODE_DATA(0), (bootmap_start_pfn << PAGE_SHIFT),
 			((bootmap_pfn - bootmap_start_pfn) << PAGE_SHIFT));
 
@@ -355,8 +352,8 @@ static void __init setup_bootmem(void)
 #endif
 
 	data_resource.start =  virt_to_phys(&data_start);
-	data_resource.end = virt_to_phys(&_end)-1;
-	code_resource.start = virt_to_phys(&_text);
+	data_resource.end = virt_to_phys(_end) - 1;
+	code_resource.start = virt_to_phys(_text);
 	code_resource.end = virt_to_phys(&data_start)-1;
 
 	/* We don't know which region the kernel will be in, so try
@@ -385,12 +382,12 @@ void free_initmem(void)
 	 */
 	local_irq_disable();
 
-	memset(&__init_begin, 0x00, 
-		(unsigned long)&__init_end - (unsigned long)&__init_begin);
+	memset(__init_begin, 0x00,
+		(unsigned long)__init_end - (unsigned long)__init_begin);
 
 	flush_data_cache();
 	asm volatile("sync" : : );
-	flush_icache_range((unsigned long)&__init_begin, (unsigned long)&__init_end);
+	flush_icache_range((unsigned long)__init_begin, (unsigned long)__init_end);
 	asm volatile("sync" : : );
 
 	local_irq_enable();
@@ -398,8 +395,8 @@ void free_initmem(void)
 	
 	/* align __init_begin and __init_end to page size,
 	   ignoring linker script where we might have tried to save RAM */
-	init_begin = PAGE_ALIGN((unsigned long)(&__init_begin));
-	init_end   = PAGE_ALIGN((unsigned long)(&__init_end));
+	init_begin = PAGE_ALIGN((unsigned long)(__init_begin));
+	init_end   = PAGE_ALIGN((unsigned long)(__init_end));
 	for (addr = init_begin; addr < init_end; addr += PAGE_SIZE) {
 		ClearPageReserved(virt_to_page(addr));
 		init_page_count(virt_to_page(addr));
@@ -578,7 +575,7 @@ static void __init map_pages(unsigned long start_vaddr, unsigned long start_padd
 	extern const unsigned long fault_vector_20;
 	extern void * const linux_gateway_page;
 
-	ro_start = __pa((unsigned long)&_text);
+	ro_start = __pa((unsigned long)_text);
 	ro_end   = __pa((unsigned long)&data_start);
 	fv_addr  = __pa((unsigned long)&fault_vector_20) & PAGE_MASK;
 	gw_addr  = __pa((unsigned long)&linux_gateway_page) & PAGE_MASK;

arch/parisc/mm/ioremap.c

@@ -188,7 +188,7 @@ void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned l
 }
 EXPORT_SYMBOL(__ioremap);
 
-void iounmap(void __iomem *addr)
+void iounmap(const volatile void __iomem *addr)
 {
 	if (addr > high_memory)
 		return vfree((void *) (PAGE_MASK & (unsigned long __force) addr));

drivers/char/agp/Kconfig

 config AGP
 	tristate "/dev/agpgart (AGP Support)"
-	depends on ALPHA || IA64 || PPC || X86
+	depends on ALPHA || IA64 || PARISC || PPC || X86
 	depends on PCI
 	---help---
 	  AGP (Accelerated Graphics Port) is a bus system mainly used to
@@ -122,6 +122,14 @@ config AGP_HP_ZX1
 	  This option gives you AGP GART support for the HP ZX1 chipset
 	  for IA64 processors.
 
+config AGP_PARISC
+	tristate "HP Quicksilver AGP support"
+	depends on AGP && PARISC && 64BIT
+	help
+	  This option gives you AGP GART support for the HP Quicksilver
+	  AGP bus adapter on HP PA-RISC machines (Ok, just on the C8000
+	  workstation...)
+
 config AGP_ALPHA_CORE
 	tristate "Alpha AGP support"
 	depends on AGP && (ALPHA_GENERIC || ALPHA_TITAN || ALPHA_MARVEL)

drivers/char/agp/Makefile

@@ -8,6 +8,7 @@ obj-$(CONFIG_AGP_AMD64)		+= amd64-agp.o
 obj-$(CONFIG_AGP_ALPHA_CORE)	+= alpha-agp.o
 obj-$(CONFIG_AGP_EFFICEON)	+= efficeon-agp.o
 obj-$(CONFIG_AGP_HP_ZX1)	+= hp-agp.o
+obj-$(CONFIG_AGP_PARISC)	+= parisc-agp.o
 obj-$(CONFIG_AGP_I460)		+= i460-agp.o
 obj-$(CONFIG_AGP_INTEL)		+= intel-agp.o
 obj-$(CONFIG_AGP_NVIDIA)	+= nvidia-agp.o

drivers/parisc/iosapic.c

@@ -146,7 +146,7 @@
 #include <asm/superio.h>
 #endif
 
-#include <asm/iosapic.h>
+#include <asm/ropes.h>
 #include "./iosapic_private.h"
 
 #define MODULE_NAME "iosapic"
@@ -692,6 +692,7 @@ static void iosapic_end_irq(unsigned int irq)
 	DBG(KERN_DEBUG "end_irq(%d): eoi(%p, 0x%x)\n", irq,
 			vi->eoi_addr, vi->eoi_data);
 	iosapic_eoi(vi->eoi_addr, vi->eoi_data);
+	cpu_end_irq(irq);
 }
 
 static unsigned int iosapic_startup_irq(unsigned int irq)
@@ -728,7 +729,7 @@ static struct hw_interrupt_type iosapic_interrupt_type = {
 	.shutdown =	iosapic_disable_irq,
 	.enable =	iosapic_enable_irq,
 	.disable =	iosapic_disable_irq,
-	.ack =		no_ack_irq,
+	.ack =		cpu_ack_irq,
 	.end =		iosapic_end_irq,
 #ifdef CONFIG_SMP
 	.set_affinity =	iosapic_set_affinity_irq,

drivers/parisc/lba_pci.c

@@ -46,9 +46,9 @@
 #include <asm/page.h>
 #include <asm/system.h>
 
+#include <asm/ropes.h>
 #include <asm/hardware.h>	/* for register_parisc_driver() stuff */
 #include <asm/parisc-device.h>
-#include <asm/iosapic.h>	/* for iosapic_register() */
 #include <asm/io.h>		/* read/write stuff */
 
 #undef DEBUG_LBA	/* general stuff */
@@ -100,113 +100,10 @@
 
 #define MODULE_NAME "LBA"
 
-#define LBA_FUNC_ID	0x0000	/* function id */
-#define LBA_FCLASS	0x0008	/* function class, bist, header, rev... */
-#define LBA_CAPABLE	0x0030	/* capabilities register */
-
-#define LBA_PCI_CFG_ADDR	0x0040	/* poke CFG address here */
-#define LBA_PCI_CFG_DATA	0x0048	/* read or write data here */
-
-#define LBA_PMC_MTLT	0x0050	/* Firmware sets this - read only. */
-#define LBA_FW_SCRATCH	0x0058	/* Firmware writes the PCI bus number here. */
-#define LBA_ERROR_ADDR	0x0070	/* On error, address gets logged here */
-
-#define LBA_ARB_MASK	0x0080	/* bit 0 enable arbitration. PAT/PDC enables */
-#define LBA_ARB_PRI	0x0088	/* firmware sets this. */
-#define LBA_ARB_MODE	0x0090	/* firmware sets this. */
-#define LBA_ARB_MTLT	0x0098	/* firmware sets this. */
-
-#define LBA_MOD_ID	0x0100	/* Module ID. PDC_PAT_CELL reports 4 */
-
-#define LBA_STAT_CTL	0x0108	/* Status & Control */
-#define   LBA_BUS_RESET		0x01	/*  Deassert PCI Bus Reset Signal */
-#define   CLEAR_ERRLOG		0x10	/*  "Clear Error Log" cmd */
-#define   CLEAR_ERRLOG_ENABLE	0x20	/*  "Clear Error Log" Enable */
-#define   HF_ENABLE	0x40	/*    enable HF mode (default is -1 mode) */
-
-#define LBA_LMMIO_BASE	0x0200	/* < 4GB I/O address range */
-#define LBA_LMMIO_MASK	0x0208
-
-#define LBA_GMMIO_BASE	0x0210	/* > 4GB I/O address range */
-#define LBA_GMMIO_MASK	0x0218
-
-#define LBA_WLMMIO_BASE	0x0220	/* All < 4GB ranges under the same *SBA* */
-#define LBA_WLMMIO_MASK	0x0228
-
-#define LBA_WGMMIO_BASE	0x0230	/* All > 4GB ranges under the same *SBA* */
-#define LBA_WGMMIO_MASK	0x0238
-
-#define LBA_IOS_BASE	0x0240	/* I/O port space for this LBA */
-#define LBA_IOS_MASK	0x0248
-
-#define LBA_ELMMIO_BASE	0x0250	/* Extra LMMIO range */
-#define LBA_ELMMIO_MASK	0x0258
-
-#define LBA_EIOS_BASE	0x0260	/* Extra I/O port space */
-#define LBA_EIOS_MASK	0x0268
-
-#define LBA_GLOBAL_MASK	0x0270	/* Mercury only: Global Address Mask */
-#define LBA_DMA_CTL	0x0278	/* firmware sets this */
-
-#define LBA_IBASE	0x0300	/* SBA DMA support */
-#define LBA_IMASK	0x0308
-
-/* FIXME: ignore DMA Hint stuff until we can measure performance */
-#define LBA_HINT_CFG	0x0310
-#define LBA_HINT_BASE	0x0380	/* 14 registers at every 8 bytes. */
-
-#define LBA_BUS_MODE	0x0620
-
-/* ERROR regs are needed for config cycle kluges */
-#define LBA_ERROR_CONFIG 0x0680
-#define     LBA_SMART_MODE 0x20
-#define LBA_ERROR_STATUS 0x0688
-#define LBA_ROPE_CTL     0x06A0
-
-#define LBA_IOSAPIC_BASE	0x800 /* Offset of IRQ logic */
-
 /* non-postable I/O port space, densely packed */
 #define LBA_PORT_BASE	(PCI_F_EXTEND | 0xfee00000UL)
 static void __iomem *astro_iop_base __read_mostly;
 
-#define ELROY_HVERS	0x782
-#define MERCURY_HVERS	0x783
-#define QUICKSILVER_HVERS	0x784
-
-static inline int IS_ELROY(struct parisc_device *d)
-{
-	return (d->id.hversion == ELROY_HVERS);
-}
-
-static inline int IS_MERCURY(struct parisc_device *d)
-{
-	return (d->id.hversion == MERCURY_HVERS);
-}
-
-static inline int IS_QUICKSILVER(struct parisc_device *d)
-{
-	return (d->id.hversion == QUICKSILVER_HVERS);
-}
-
-
-/*
-** lba_device: Per instance Elroy data structure
-*/
-struct lba_device {
-	struct pci_hba_data hba;
-
-	spinlock_t	lba_lock;
-	void		*iosapic_obj;
-
-#ifdef CONFIG_64BIT
-	void __iomem *	iop_base;    /* PA_VIEW - for IO port accessor funcs */
-#endif
-
-	int		flags;       /* state/functionality enabled */
-	int		hw_rev;      /* HW revision of chip */
-};
-
-
 static u32 lba_t32;
 
 /* lba flags */
@@ -1542,8 +1439,8 @@ lba_driver_probe(struct parisc_device *dev)
 		default: version = "TR4+";
 		}
 
-		printk(KERN_INFO "%s version %s (0x%x) found at 0x%lx\n",
-			MODULE_NAME, version, func_class & 0xf, dev->hpa.start);
+		printk(KERN_INFO "Elroy version %s (0x%x) found at 0x%lx\n",
+		       version, func_class & 0xf, dev->hpa.start);
 
 		if (func_class < 2) {
 			printk(KERN_WARNING "Can't support LBA older than "
@@ -1563,14 +1460,18 @@ lba_driver_probe(struct parisc_device *dev)
 		}
 
 	} else if (IS_MERCURY(dev) || IS_QUICKSILVER(dev)) {
+		int major, minor;
+
 		func_class &= 0xff;
-		version = kmalloc(6, GFP_KERNEL);
-		snprintf(version, 6, "TR%d.%d",(func_class >> 4),(func_class & 0xf));
+		major = func_class >> 4, minor = func_class & 0xf;
+
 		/* We could use one printk for both Elroy and Mercury,
                  * but for the mask for func_class.
                  */ 
-		printk(KERN_INFO "%s version %s (0x%x) found at 0x%lx\n",
-		       MODULE_NAME, version, func_class & 0xff, dev->hpa.start);
+		printk(KERN_INFO "%s version TR%d.%d (0x%x) found at 0x%lx\n",
+		       IS_MERCURY(dev) ? "Mercury" : "Quicksilver", major,
+		       minor, func_class, dev->hpa.start);
+
 		cfg_ops = &mercury_cfg_ops;
 	} else {
 		printk(KERN_ERR "Unknown LBA found at 0x%lx\n", dev->hpa.start);
@@ -1600,6 +1501,7 @@ lba_driver_probe(struct parisc_device *dev)
 	lba_dev->hba.dev = dev;
 	lba_dev->iosapic_obj = tmp_obj;  /* save interrupt handle */
 	lba_dev->hba.iommu = sba_get_iommu(dev);  /* get iommu data */
+	parisc_set_drvdata(dev, lba_dev);
 
 	/* ------------ Second : initialize common stuff ---------- */
 	pci_bios = &lba_bios_ops;

drivers/serial/8250_gsc.c

@@ -22,7 +22,6 @@
 #include <asm/hardware.h>
 #include <asm/parisc-device.h>
 #include <asm/io.h>
-#include <asm/serial.h> /* for LASI_BASE_BAUD */
 
 #include "8250.h"
 
@@ -54,7 +53,8 @@ serial_init_chip(struct parisc_device *dev)
 
 	memset(&port, 0, sizeof(port));
 	port.iotype	= UPIO_MEM;
-	port.uartclk	= LASI_BASE_BAUD * 16;
+	/* 7.272727MHz on Lasi.  Assumed the same for Dino, Wax and Timi. */
+	port.uartclk	= 7272727;
 	port.mapbase	= address;
 	port.membase	= ioremap_nocache(address, 16);
 	port.irq	= dev->irq;

drivers/serial/Kconfig

@@ -556,10 +556,11 @@ config SERIAL_MUX
 	default y
 	---help---
 	  Saying Y here will enable the hardware MUX serial driver for
-	  the Nova and K class systems.  The hardware MUX is not 8250/16550 
-	  compatible therefore the /dev/ttyB0 device is shared between the 
-	  Serial MUX and the PDC software console.  The following steps 
-	  need to be completed to use the Serial MUX:
+	  the Nova, K class systems and D class with a 'remote control card'.
+	  The hardware MUX is not 8250/16550 compatible therefore the
+	  /dev/ttyB0 device is shared between the Serial MUX and the PDC
+	  software console. The following steps need to be completed to use
+	  the Serial MUX:
 
 	    1. create the device entry (mknod /dev/ttyB0 c 11 0)
 	    2. Edit the /etc/inittab to start a getty listening on /dev/ttyB0

fs/binfmt_som.c

@@ -29,6 +29,7 @@
 #include <linux/personality.h>
 #include <linux/init.h>
 
+#include <asm/a.out.h>
 #include <asm/uaccess.h>
 #include <asm/pgtable.h>
 
@@ -194,6 +195,7 @@ load_som_binary(struct linux_binprm * bprm, struct pt_regs * regs)
 	unsigned long som_entry;
 	struct som_hdr *som_ex;
 	struct som_exec_auxhdr *hpuxhdr;
+	struct files_struct *files;
 
 	/* Get the exec-header */
 	som_ex = (struct som_hdr *) bprm->buf;
@@ -208,15 +210,27 @@ load_som_binary(struct linux_binprm * bprm, struct pt_regs * regs)
 	size = som_ex->aux_header_size;
 	if (size > SOM_PAGESIZE)
 		goto out;
-	hpuxhdr = (struct som_exec_auxhdr *) kmalloc(size, GFP_KERNEL);
+	hpuxhdr = kmalloc(size, GFP_KERNEL);
 	if (!hpuxhdr)
 		goto out;
 
 	retval = kernel_read(bprm->file, som_ex->aux_header_location,
 			(char *) hpuxhdr, size);
+	if (retval != size) {
+		if (retval >= 0)
+			retval = -EIO;
+		goto out_free;
+	}
+
+	files = current->files; /* Refcounted so ok */
+	retval = unshare_files();
 	if (retval < 0)
 		goto out_free;
-#error "Fix security hole before enabling me"
+	if (files == current->files) {
+		put_files_struct(files);
+		files = NULL;
+	}
+
 	retval = get_unused_fd();
 	if (retval < 0)
 		goto out_free;

include/asm-parisc/agp.h

+#ifndef _ASM_PARISC_AGP_H
+#define _ASM_PARISC_AGP_H
+
+/*
+ * PARISC specific AGP definitions.
+ * Copyright (c) 2006 Kyle McMartin <kyle@parisc-linux.org>
+ *
+ */
+
+#define map_page_into_agp(page)		/* nothing */
+#define unmap_page_from_agp(page)	/* nothing */
+#define flush_agp_mappings()		/* nothing */
+#define flush_agp_cache()		mb()
+
+/* Convert a physical address to an address suitable for the GART. */
+#define phys_to_gart(x) (x)
+#define gart_to_phys(x) (x)
+
+/* GATT allocation. Returns/accepts GATT kernel virtual address. */
+#define alloc_gatt_pages(order)		\
+	((char *)__get_free_pages(GFP_KERNEL, (order)))
+#define free_gatt_pages(table, order)	\
+	free_pages((unsigned long)(table), (order))
+
+#endif /* _ASM_PARISC_AGP_H */

include/asm-parisc/assembly.h

@@ -29,7 +29,8 @@
 #define LDREGX  ldd,s
 #define LDREGM	ldd,mb
 #define STREGM	std,ma
-#define SHRREG  shrd
+#define SHRREG	shrd
+#define SHLREG	shld
 #define RP_OFFSET	16
 #define FRAME_SIZE	128
 #define CALLEE_REG_FRAME_SIZE	144
@@ -39,7 +40,8 @@
 #define LDREGX  ldwx,s
 #define LDREGM	ldwm
 #define STREGM	stwm
-#define SHRREG  shr
+#define SHRREG	shr
+#define SHLREG	shlw
 #define RP_OFFSET	20
 #define FRAME_SIZE	64
 #define CALLEE_REG_FRAME_SIZE	128

include/asm-parisc/cacheflush.h

@@ -191,16 +191,38 @@ flush_anon_page(struct page *page, unsigned long vmaddr)
 }
 #define ARCH_HAS_FLUSH_ANON_PAGE
 
-static inline void
-flush_kernel_dcache_page(struct page *page)
+#define ARCH_HAS_FLUSH_KERNEL_DCACHE_PAGE
+void flush_kernel_dcache_page_addr(void *addr);
+static inline void flush_kernel_dcache_page(struct page *page)
 {
-	flush_kernel_dcache_page_asm(page_address(page));
+	flush_kernel_dcache_page_addr(page_address(page));
 }
-#define ARCH_HAS_FLUSH_KERNEL_DCACHE_PAGE
 
 #ifdef CONFIG_DEBUG_RODATA
 void mark_rodata_ro(void);
 #endif
 
+#ifdef CONFIG_PA8X00
+/* Only pa8800, pa8900 needs this */
+#define ARCH_HAS_KMAP
+
+void kunmap_parisc(void *addr);
+
+static inline void *kmap(struct page *page)
+{
+	might_sleep();
+	return page_address(page);
+}
+
+#define kunmap(page)			kunmap_parisc(page_address(page))
+
+#define kmap_atomic(page, idx)		page_address(page)
+
+#define kunmap_atomic(addr, idx)	kunmap_parisc(addr)
+
+#define kmap_atomic_pfn(pfn, idx)	page_address(pfn_to_page(pfn))
+#define kmap_atomic_to_page(ptr)	virt_to_page(ptr)
+#endif
+
 #endif /* _PARISC_CACHEFLUSH_H */
 

include/asm-parisc/compat.h

@@ -5,7 +5,7 @@
  */
 #include <linux/types.h>
 #include <linux/sched.h>
-#include <linux/personality.h>
+#include <linux/thread_info.h>
 
 #define COMPAT_USER_HZ 100
 
@@ -152,7 +152,7 @@ static __inline__ void __user *compat_alloc_user_space(long len)
 
 static inline int __is_compat_task(struct task_struct *t)
 {
-	return personality(t->personality) == PER_LINUX32;
+	return test_ti_thread_flag(t->thread_info, TIF_32BIT);
 }
 
 static inline int is_compat_task(void)

include/asm-parisc/dma.h

@@ -72,18 +72,13 @@
 #define DMA2_MASK_ALL_REG       0xDE    /* all-channels mask (w) */
 #define DMA2_EXT_MODE_REG	(0x400 | DMA2_MODE_REG)
 
-extern spinlock_t dma_spin_lock;
-
 static __inline__ unsigned long claim_dma_lock(void)
 {
-	unsigned long flags;
-	spin_lock_irqsave(&dma_spin_lock, flags);
-	return flags;
+	return 0;
 }
 
 static __inline__ void release_dma_lock(unsigned long flags)
 {
-	spin_unlock_irqrestore(&dma_spin_lock, flags);
 }
 
 

include/asm-parisc/futex.h

-#ifndef _ASM_FUTEX_H
-#define _ASM_FUTEX_H
+#ifndef _ASM_PARISC_FUTEX_H
+#define _ASM_PARISC_FUTEX_H
 
-#include <asm-generic/futex.h>
+#ifdef __KERNEL__
 
+#include <linux/futex.h>
+#include <asm/errno.h>
+#include <asm/uaccess.h>
+
+static inline int
+futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+{
+	int op = (encoded_op >> 28) & 7;
+	int cmp = (encoded_op >> 24) & 15;
+	int oparg = (encoded_op << 8) >> 20;
+	int cmparg = (encoded_op << 20) >> 20;
+	int oldval = 0, ret;
+	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
+		oparg = 1 << oparg;
+
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+		return -EFAULT;
+
+	inc_preempt_count();
+
+	switch (op) {
+	case FUTEX_OP_SET:
+	case FUTEX_OP_ADD:
+	case FUTEX_OP_OR:
+	case FUTEX_OP_ANDN:
+	case FUTEX_OP_XOR:
+	default:
+		ret = -ENOSYS;
+	}
+
+	dec_preempt_count();
+
+	if (!ret) {
+		switch (cmp) {
+		case FUTEX_OP_CMP_EQ: ret = (oldval == cmparg); break;
+		case FUTEX_OP_CMP_NE: ret = (oldval != cmparg); break;
+		case FUTEX_OP_CMP_LT: ret = (oldval < cmparg); break;
+		case FUTEX_OP_CMP_GE: ret = (oldval >= cmparg); break;
+		case FUTEX_OP_CMP_LE: ret = (oldval <= cmparg); break;
+		case FUTEX_OP_CMP_GT: ret = (oldval > cmparg); break;
+		default: ret = -ENOSYS;
+		}
+	}
+	return ret;
+}
+
+/* Non-atomic version */
+static inline int
+futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+{
+	int err = 0;
+	int uval;
+
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+		return -EFAULT;
+
+	err = get_user(uval, uaddr);
+	if (err) return -EFAULT;
+	if (uval == oldval)
+		err = put_user(newval, uaddr);
+	if (err) return -EFAULT;
+	return uval;
+}
+
+#endif
 #endif

include/asm-parisc/io.h

@@ -134,7 +134,7 @@ extern inline void __iomem * ioremap(unsigned long offset, unsigned long size)
 }
 #define ioremap_nocache(off, sz)	ioremap((off), (sz))
 
-extern void iounmap(void __iomem *addr);
+extern void iounmap(const volatile void __iomem *addr);
 
 static inline unsigned char __raw_readb(const volatile void __iomem *addr)
 {

include/asm-parisc/iosapic.h

-/*
-** This file is private to iosapic driver.
-** If stuff needs to be used by another driver, move it to a common file.
-**
-** WARNING: fields most data structures here are ordered to make sure
-**          they pack nicely for 64-bit compilation. (ie sizeof(long) == 8)
-*/
-
-
-/*
-** I/O SAPIC init function
-** Caller knows where an I/O SAPIC is. LBA has an integrated I/O SAPIC.
-** Call setup as part of per instance initialization.
-** (ie *not* init_module() function unless only one is present.)
-** fixup_irq is to initialize PCI IRQ line support and
-** virtualize pcidev->irq value. To be called by pci_fixup_bus().
-*/
-extern void *iosapic_register(unsigned long hpa);
-extern int iosapic_fixup_irq(void *obj, struct pci_dev *pcidev);
-
-
-#ifdef __IA64__
-/*
-** PA: PIB (Processor Interrupt Block) is handled by Runway bus adapter.
-**     and is hardcoded to 0xfeeNNNN0 where NNNN is id_eid field.
-**
-** IA64: PIB is handled by "Local SAPIC" (integrated in the processor).
-*/
-struct local_sapic_info {
-	struct local_sapic_info *lsi_next;      /* point to next CPU info */
-	int                     *lsi_cpu_id;    /* point to logical CPU id */
-	unsigned long           *lsi_id_eid;    /* point to IA-64 CPU id */
-	int                     *lsi_status;    /* point to CPU status   */
-	void                    *lsi_private;   /* point to special info */
-};
-
-/*
-** "root" data structure which ties everything together.
-** Should always be able to start with sapic_root and locate
-** the desired information.
-*/
-struct sapic_info {
-	struct sapic_info	*si_next;	/* info is per cell */
-	int                     si_cellid;      /* cell id */
-	unsigned int            si_status;       /* status  */
-	char                    *si_pib_base;   /* intr blk base address */
-	local_sapic_info_t      *si_local_info;
-	io_sapic_info_t         *si_io_info;
-	extint_info_t           *si_extint_info;/* External Intr info      */
-};
-
-#endif /* IA64 */
-

include/asm-parisc/irq.h

@@ -31,7 +31,7 @@ static __inline__ int irq_canonicalize(int irq)
 	return (irq == 2) ? 9 : irq;
 }
 
-struct hw_interrupt_type;
+struct irq_chip;
 
 /*
  * Some useful "we don't have to do anything here" handlers.  Should
@@ -39,6 +39,8 @@ struct hw_interrupt_type;
  */
 void no_ack_irq(unsigned int irq);
 void no_end_irq(unsigned int irq);
+void cpu_ack_irq(unsigned int irq);
+void cpu_end_irq(unsigned int irq);
 
 extern int txn_alloc_irq(unsigned int nbits);
 extern int txn_claim_irq(int);
@@ -46,7 +48,7 @@ extern unsigned int txn_alloc_data(unsigned int);
 extern unsigned long txn_alloc_addr(unsigned int);
 extern unsigned long txn_affinity_addr(unsigned int irq, int cpu);
 
-extern int cpu_claim_irq(unsigned int irq, struct hw_interrupt_type *, void *);
+extern int cpu_claim_irq(unsigned int irq, struct irq_chip *, void *);
 extern int cpu_check_affinity(unsigned int irq, cpumask_t *dest);
 
 /* soft power switch support (power.c) */

include/asm-parisc/mckinley.h

+#ifndef ASM_PARISC_MCKINLEY_H
+#define ASM_PARISC_MCKINLEY_H
+#ifdef __KERNEL__
+
+/* declared in arch/parisc/kernel/setup.c */
+extern struct proc_dir_entry * proc_mckinley_root;
+
+#endif /*__KERNEL__*/
+#endif /*ASM_PARISC_MCKINLEY_H*/

include/asm-parisc/page.h

@@ -26,24 +26,10 @@
 
 struct page;
 
-extern void purge_kernel_dcache_page(unsigned long);
-extern void copy_user_page_asm(void *to, void *from);
-extern void clear_user_page_asm(void *page, unsigned long vaddr);
-
-static inline void
-copy_user_page(void *vto, void *vfrom, unsigned long vaddr, struct page *pg)
-{
-	copy_user_page_asm(vto, vfrom);
-	flush_kernel_dcache_page_asm(vto);
-	/* XXX: ppc flushes icache too, should we? */
-}
-
-static inline void
-clear_user_page(void *page, unsigned long vaddr, struct page *pg)
-{
-	purge_kernel_dcache_page((unsigned long)page);
-	clear_user_page_asm(page, vaddr);
-}
+void copy_user_page_asm(void *to, void *from);
+void copy_user_page(void *vto, void *vfrom, unsigned long vaddr,
+			   struct page *pg);
+void clear_user_page(void *page, unsigned long vaddr, struct page *pg);
 
 /*
  * These are used to make use of C type-checking..

include/asm-parisc/param.h

@@ -2,13 +2,9 @@
 #define _ASMPARISC_PARAM_H
 
 #ifdef __KERNEL__
-# ifdef CONFIG_PA20
-#  define HZ		1000		/* Faster machines */
-# else
-#  define HZ		100		/* Internal kernel timer frequency */
-# endif
-# define USER_HZ	100		/* .. some user interfaces are in "ticks" */
-# define CLOCKS_PER_SEC	(USER_HZ)	/* like times() */
+#define HZ		CONFIG_HZ
+#define USER_HZ		100		/* some user API use "ticks" */
+#define CLOCKS_PER_SEC	(USER_HZ)	/* like times() */
 #endif
 
 #ifndef HZ

include/asm-parisc/parisc-device.h

+#ifndef _ASM_PARISC_PARISC_DEVICE_H_
+#define _ASM_PARISC_PARISC_DEVICE_H_
+
 #include <linux/device.h>
 
 struct parisc_device {
@@ -57,3 +60,5 @@ parisc_get_drvdata(struct parisc_device *d)
 }
 
 extern struct bus_type parisc_bus_type;
+
+#endif /*_ASM_PARISC_PARISC_DEVICE_H_*/

include/asm-parisc/pci.h

@@ -293,4 +293,9 @@ static inline void pcibios_penalize_isa_irq(int irq, int active)
 	/* We don't need to penalize isa irq's */
 }
 
+static inline int pci_get_legacy_ide_irq(struct pci_dev *dev, int channel)
+{
+	return channel ? 15 : 14;
+}
+
 #endif /* __ASM_PARISC_PCI_H */

include/asm-parisc/prefetch.h

+/*
+ * include/asm-parisc/prefetch.h
+ *
+ * PA 2.0 defines data prefetch instructions on page 6-11 of the Kane book.
+ * In addition, many implementations do hardware prefetching of both
+ * instructions and data.
+ *
+ * PA7300LC (page 14-4 of the ERS) also implements prefetching by a load
+ * to gr0 but not in a way that Linux can use.  If the load would cause an
+ * interruption (eg due to prefetching 0), it is suppressed on PA2.0
+ * processors, but not on 7300LC.
+ *
+ */
+
+#ifndef __ASM_PARISC_PREFETCH_H
+#define __ASM_PARISC_PREFETCH_H
+
+#ifndef __ASSEMBLY__
+#ifdef CONFIG_PREFETCH
+
+#define ARCH_HAS_PREFETCH
+extern inline void prefetch(const void *addr)
+{
+	__asm__("ldw 0(%0), %%r0" : : "r" (addr));
+}
+
+/* LDD is a PA2.0 addition. */
+#ifdef CONFIG_PA20
+#define ARCH_HAS_PREFETCHW
+extern inline void prefetchw(const void *addr)
+{
+	__asm__("ldd 0(%0), %%r0" : : "r" (addr));
+}
+#endif /* CONFIG_PA20 */
+
+#endif /* CONFIG_PREFETCH */
+#endif /* __ASSEMBLY__ */
+
+#endif /* __ASM_PARISC_PROCESSOR_H */

include/asm-parisc/processor.h

@@ -9,6 +9,8 @@
 #define __ASM_PARISC_PROCESSOR_H
 
 #ifndef __ASSEMBLY__
+#include <asm/prefetch.h>	/* lockdep.h needs <linux/prefetch.h> */
+
 #include <linux/threads.h>
 #include <linux/spinlock_types.h>
 
@@ -276,7 +278,7 @@ on downward growing arches, it looks like this:
  */
 
 #ifdef __LP64__
-#define USER_WIDE_MODE	(personality(current->personality) == PER_LINUX)
+#define USER_WIDE_MODE	(!test_thread_flag(TIF_32BIT))
 #else
 #define USER_WIDE_MODE	0
 #endif
@@ -328,33 +330,20 @@ extern unsigned long get_wchan(struct task_struct *p);
 #define KSTK_EIP(tsk)	((tsk)->thread.regs.iaoq[0])
 #define KSTK_ESP(tsk)	((tsk)->thread.regs.gr[30])
 
+#define cpu_relax()	barrier()
 
-/*
- * PA 2.0 defines data prefetch instructions on page 6-11 of the Kane book.
- * In addition, many implementations do hardware prefetching of both
- * instructions and data.
- *
- * PA7300LC (page 14-4 of the ERS) also implements prefetching by a load
- * to gr0 but not in a way that Linux can use.  If the load would cause an
- * interruption (eg due to prefetching 0), it is suppressed on PA2.0
- * processors, but not on 7300LC.
- */
-#ifdef  CONFIG_PREFETCH
-#define ARCH_HAS_PREFETCH
-#define ARCH_HAS_PREFETCHW
-
-extern inline void prefetch(const void *addr)
-{
-	__asm__("ldw 0(%0), %%r0" : : "r" (addr));
-}
-
-extern inline void prefetchw(const void *addr)
+/* Used as a macro to identify the combined VIPT/PIPT cached
+ * CPUs which require a guarantee of coherency (no inequivalent
+ * aliases with different data, whether clean or not) to operate */
+static inline int parisc_requires_coherency(void)
 {
-	__asm__("ldd 0(%0), %%r0" : : "r" (addr));
-}
+#ifdef CONFIG_PA8X00
+	/* FIXME: also pa8900 - when we see one */
+	return boot_cpu_data.cpu_type == mako;
+#else
+	return 0;
 #endif
-
-#define cpu_relax()	barrier()
+}
 
 #endif /* __ASSEMBLY__ */
 

include/asm-parisc/serial.h

@@ -3,20 +3,8 @@
  */
 
 /*
- * This assumes you have a 7.272727 MHz clock for your UART.
- * The documentation implies a 40Mhz clock, and elsewhere a 7Mhz clock
- * Clarified: 7.2727MHz on LASI. Not yet clarified for DINO
+ * This is used for 16550-compatible UARTs
  */
+#define BASE_BAUD ( 1843200 / 16 )
 
-#define LASI_BASE_BAUD ( 7272727 / 16 )
-#define BASE_BAUD  LASI_BASE_BAUD
-
-/*
- * We don't use the ISA probing code, so these entries are just to reserve
- * space.  Some example (maximal) configurations:
- * - 712 w/ additional Lasi & RJ16 ports: 4
- * - J5k w/ PCI serial cards: 2 + 4 * card ~= 34
- * A500 w/ PCI serial cards: 5 + 4 * card ~= 17
- */
- 
 #define SERIAL_PORT_DFNS

include/asm-parisc/spinlock.h

@@ -56,50 +56,79 @@ static inline int __raw_spin_trylock(raw_spinlock_t *x)
 }
 
 /*
- * Read-write spinlocks, allowing multiple readers
- * but only one writer.
+ * Read-write spinlocks, allowing multiple readers but only one writer.
+ * Linux rwlocks are unfair to writers; they can be starved for an indefinite
+ * time by readers.  With care, they can also be taken in interrupt context.
+ *
+ * In the PA-RISC implementation, we have a spinlock and a counter.
+ * Readers use the lock to serialise their access to the counter (which
+ * records how many readers currently hold the lock).
+ * Writers hold the spinlock, preventing any readers or other writers from
+ * grabbing the rwlock.
  */
 
-#define __raw_read_trylock(lock) generic__raw_read_trylock(lock)
-
-/* read_lock, read_unlock are pretty straightforward.  Of course it somehow
- * sucks we end up saving/restoring flags twice for read_lock_irqsave aso. */
-
+/* Note that we have to ensure interrupts are disabled in case we're
+ * interrupted by some other code that wants to grab the same read lock */
 static  __inline__ void __raw_read_lock(raw_rwlock_t *rw)
 {
-	__raw_spin_lock(&rw->lock);
-
+	unsigned long flags;
+	local_irq_save(flags);
+	__raw_spin_lock_flags(&rw->lock, flags);
 	rw->counter++;
-
 	__raw_spin_unlock(&rw->lock);
+	local_irq_restore(flags);
 }
 
+/* Note that we have to ensure interrupts are disabled in case we're
+ * interrupted by some other code that wants to grab the same read lock */
 static  __inline__ void __raw_read_unlock(raw_rwlock_t *rw)
 {
-	__raw_spin_lock(&rw->lock);
-
+	unsigned long flags;
+	local_irq_save(flags);
+	__raw_spin_lock_flags(&rw->lock, flags);
 	rw->counter--;
-
 	__raw_spin_unlock(&rw->lock);
+	local_irq_restore(flags);
 }
 
-/* write_lock is less trivial.  We optimistically grab the lock and check
- * if we surprised any readers.  If so we release the lock and wait till
- * they're all gone before trying again
- *
- * Also note that we don't use the _irqsave / _irqrestore suffixes here.
- * If we're called with interrupts enabled and we've got readers (or other
- * writers) in interrupt handlers someone fucked up and we'd dead-lock
- * sooner or later anyway.   prumpf */
+/* Note that we have to ensure interrupts are disabled in case we're
+ * interrupted by some other code that wants to grab the same read lock */
+static __inline__ int __raw_read_trylock(raw_rwlock_t *rw)
+{
+	unsigned long flags;
+ retry:
+	local_irq_save(flags);
+	if (__raw_spin_trylock(&rw->lock)) {
+		rw->counter++;
+		__raw_spin_unlock(&rw->lock);
+		local_irq_restore(flags);
+		return 1;
+	}
 
-static  __inline__ void __raw_write_lock(raw_rwlock_t *rw)
+	local_irq_restore(flags);
+	/* If write-locked, we fail to acquire the lock */
+	if (rw->counter < 0)
+		return 0;
+
+	/* Wait until we have a realistic chance at the lock */
+	while (__raw_spin_is_locked(&rw->lock) && rw->counter >= 0)
+		cpu_relax();
+
+	goto retry;
+}
+
+/* Note that we have to ensure interrupts are disabled in case we're
+ * interrupted by some other code that wants to read_trylock() this lock */
+static __inline__ void __raw_write_lock(raw_rwlock_t *rw)
 {
+	unsigned long flags;
 retry:
-	__raw_spin_lock(&rw->lock);
+	local_irq_save(flags);
+	__raw_spin_lock_flags(&rw->lock, flags);
 
-	if(rw->counter != 0) {
-		/* this basically never happens */
+	if (rw->counter != 0) {
 		__raw_spin_unlock(&rw->lock);
+		local_irq_restore(flags);
 
 		while (rw->counter != 0)
 			cpu_relax();
@@ -107,31 +136,37 @@ static  __inline__ void __raw_write_lock(raw_rwlock_t *rw)
 		goto retry;
 	}
 
-	/* got it.  now leave without unlocking */
-	rw->counter = -1; /* remember we are locked */
+	rw->counter = -1; /* mark as write-locked */
+	mb();
+	local_irq_restore(flags);
 }
 
-/* write_unlock is absolutely trivial - we don't have to wait for anything */
-
-static  __inline__ void __raw_write_unlock(raw_rwlock_t *rw)
+static __inline__ void __raw_write_unlock(raw_rwlock_t *rw)
 {
 	rw->counter = 0;
 	__raw_spin_unlock(&rw->lock);
 }
 
-static  __inline__ int __raw_write_trylock(raw_rwlock_t *rw)
+/* Note that we have to ensure interrupts are disabled in case we're
+ * interrupted by some other code that wants to read_trylock() this lock */
+static __inline__ int __raw_write_trylock(raw_rwlock_t *rw)
 {
-	__raw_spin_lock(&rw->lock);
-	if (rw->counter != 0) {
-		/* this basically never happens */
-		__raw_spin_unlock(&rw->lock);
-
-		return 0;
+	unsigned long flags;
+	int result = 0;
+
+	local_irq_save(flags);
+	if (__raw_spin_trylock(&rw->lock)) {
+		if (rw->counter == 0) {
+			rw->counter = -1;
+			result = 1;
+		} else {
+			/* Read-locked.  Oh well. */
+			__raw_spin_unlock(&rw->lock);
+		}
 	}
+	local_irq_restore(flags);
 
-	/* got it.  now leave without unlocking */
-	rw->counter = -1; /* remember we are locked */
-	return 1;
+	return result;
 }
 
 /*

include/linux/debug_locks.h

@@ -43,6 +43,8 @@ extern int debug_locks_off(void);
 # define locking_selftest()	do { } while (0)
 #endif
 
+struct task_struct;
+
 #ifdef CONFIG_LOCKDEP
 extern void debug_show_all_locks(void);
 extern void debug_show_held_locks(struct task_struct *task);