Merge bk://bk.arm.linux.org.uk/linux-2.6-rmk

into ppc970.osdl.org:/home/torvalds/v2.5/linux

arch/ia64/Kconfig

@@ -288,39 +288,6 @@ config FORCE_MAX_ZONEORDER
 	int
 	default "18"
 
-choice
-	prompt "Huge TLB page size"
-	depends on HUGETLB_PAGE
-	default HUGETLB_PAGE_SIZE_16MB
-
-config HUGETLB_PAGE_SIZE_4GB
-	depends on MCKINLEY
-	bool "4GB"
-
-config HUGETLB_PAGE_SIZE_1GB
-	depends on MCKINLEY
-	bool "1GB"
-
-config HUGETLB_PAGE_SIZE_256MB
-	bool "256MB"
-
-config HUGETLB_PAGE_SIZE_64MB
-	bool "64MB"
-
-config HUGETLB_PAGE_SIZE_16MB
-	bool "16MB"
-
-config HUGETLB_PAGE_SIZE_4MB
-	bool "4MB"
-
-config HUGETLB_PAGE_SIZE_1MB
-	bool "1MB"
-
-config HUGETLB_PAGE_SIZE_256KB
-	bool "256KB"
-
-endchoice
-
 config IA64_PAL_IDLE
 	bool "Use PAL_HALT_LIGHT in idle loop"
 	help

arch/ia64/kernel/head.S

@@ -816,6 +816,19 @@ GLOBAL_ENTRY(ia64_delay_loop)
 	br.ret.sptk.many rp
 END(ia64_delay_loop)
 
+GLOBAL_ENTRY(ia64_invoke_kernel_thread_helper)
+	.prologue
+	.save rp, r0				// this is the end of the call-chain
+	.body
+	alloc r2 = ar.pfs, 0, 0, 2, 0
+	mov out0 = r9
+	mov out1 = r11;;
+	br.call.sptk.many rp = kernel_thread_helper;;
+	mov out0 = r8
+	br.call.sptk.many rp = sys_exit;;
+1:	br.sptk.few 1b				// not reached
+END(ia64_invoke_kernel_thread_helper)
+
 #ifdef CONFIG_IA64_BRL_EMU
 
 /*

arch/ia64/kernel/iosapic.c

@@ -103,6 +103,7 @@ static spinlock_t iosapic_lock = SPIN_LOCK_UNLOCKED;
 
 static struct iosapic_intr_info {
 	char		*addr;		/* base address of IOSAPIC */
+	u32		low32;		/* current value of low word of Redirection table entry */
 	unsigned int	gsi_base;	/* first GSI assigned to this IOSAPIC */
 	char		rte_index;	/* IOSAPIC RTE index (-1 => not an IOSAPIC interrupt) */
 	unsigned char	dmode	: 3;	/* delivery mode (see iosapic.h) */
@@ -213,6 +214,7 @@ set_rte (unsigned int vector, unsigned int dest)
 	writel(high32, addr + IOSAPIC_WINDOW);
 	writel(IOSAPIC_RTE_LOW(rte_index), addr + IOSAPIC_REG_SELECT);
 	writel(low32, addr + IOSAPIC_WINDOW);
+	iosapic_intr_info[vector].low32 = low32;
 }
 
 static void
@@ -239,9 +241,10 @@ mask_irq (unsigned int irq)
 	spin_lock_irqsave(&iosapic_lock, flags);
 	{
 		writel(IOSAPIC_RTE_LOW(rte_index), addr + IOSAPIC_REG_SELECT);
-		low32 = readl(addr + IOSAPIC_WINDOW);
 
-		low32 |= (1 << IOSAPIC_MASK_SHIFT);    /* set only the mask bit */
+		/* set only the mask bit */
+		low32 = iosapic_intr_info[vec].low32 |= IOSAPIC_MASK;
+
 		writel(low32, addr + IOSAPIC_WINDOW);
 	}
 	spin_unlock_irqrestore(&iosapic_lock, flags);
@@ -264,9 +267,7 @@ unmask_irq (unsigned int irq)
 	spin_lock_irqsave(&iosapic_lock, flags);
 	{
 		writel(IOSAPIC_RTE_LOW(rte_index), addr + IOSAPIC_REG_SELECT);
-		low32 = readl(addr + IOSAPIC_WINDOW);
-
-		low32 &= ~(1 << IOSAPIC_MASK_SHIFT);    /* clear only the mask bit */
+		low32 = iosapic_intr_info[vec].low32 &= ~IOSAPIC_MASK;
 		writel(low32, addr + IOSAPIC_WINDOW);
 	}
 	spin_unlock_irqrestore(&iosapic_lock, flags);
@@ -307,9 +308,7 @@ iosapic_set_affinity (unsigned int irq, cpumask_t mask)
 	{
 		/* get current delivery mode by reading the low32 */
 		writel(IOSAPIC_RTE_LOW(rte_index), addr + IOSAPIC_REG_SELECT);
-		low32 = readl(addr + IOSAPIC_WINDOW);
-
-		low32 &= ~(7 << IOSAPIC_DELIVERY_SHIFT);
+		low32 = iosapic_intr_info[vec].low32 & ~(7 << IOSAPIC_DELIVERY_SHIFT);
 		if (redir)
 		        /* change delivery mode to lowest priority */
 			low32 |= (IOSAPIC_LOWEST_PRIORITY << IOSAPIC_DELIVERY_SHIFT);
@@ -317,6 +316,7 @@ iosapic_set_affinity (unsigned int irq, cpumask_t mask)
 		        /* change delivery mode to fixed */
 			low32 |= (IOSAPIC_FIXED << IOSAPIC_DELIVERY_SHIFT);
 
+		iosapic_intr_info[vec].low32 = low32;
 		writel(IOSAPIC_RTE_HIGH(rte_index), addr + IOSAPIC_REG_SELECT);
 		writel(high32, addr + IOSAPIC_WINDOW);
 		writel(IOSAPIC_RTE_LOW(rte_index), addr + IOSAPIC_REG_SELECT);

arch/ia64/kernel/irq.c

@@ -455,7 +455,6 @@ unsigned int do_IRQ(unsigned long irq, struct pt_regs *regs)
 	unsigned int status;
 	int cpu;
 
-	irq_enter();
 	cpu = smp_processor_id(); /* for CONFIG_PREEMPT, this must come after irq_enter()! */
 
 	kstat_cpu(cpu).irqs[irq]++;
@@ -525,7 +524,6 @@ unsigned int do_IRQ(unsigned long irq, struct pt_regs *regs)
 		desc->handler->end(irq);
 		spin_unlock(&desc->lock);
 	}
-	irq_exit();
 	return 1;
 }
 

arch/ia64/kernel/irq_ia64.c

@@ -120,6 +120,7 @@ ia64_handle_irq (ia64_vector vector, struct pt_regs *regs)
 	 * 16 (without this, it would be ~240, which could easily lead
 	 * to kernel stack overflows).
 	 */
+	irq_enter();
 	saved_tpr = ia64_getreg(_IA64_REG_CR_TPR);
 	ia64_srlz_d();
 	while (vector != IA64_SPURIOUS_INT_VECTOR) {
@@ -143,8 +144,7 @@ ia64_handle_irq (ia64_vector vector, struct pt_regs *regs)
 	 * handler needs to be able to wait for further keyboard interrupts, which can't
 	 * come through until ia64_eoi() has been done.
 	 */
-	if (local_softirq_pending())
-		do_softirq();
+	irq_exit();
 }
 
 #ifdef CONFIG_SMP

arch/ia64/kernel/ivt.S

@@ -118,10 +118,11 @@ ENTRY(vhpt_miss)
 #ifdef CONFIG_HUGETLB_PAGE
 	extr.u r26=r25,2,6
 	;;
-	cmp.eq p8,p0=HPAGE_SHIFT,r26
+	cmp.ne p8,p0=r18,r26
+	sub r27=r26,r18
 	;;
 (p8)	dep r25=r18,r25,2,6
-(p8)	shr r22=r22,HPAGE_SHIFT-PAGE_SHIFT
+(p8)	shr r22=r22,r27
 #endif
 	;;
 	cmp.eq p6,p7=5,r17			// is IFA pointing into to region 5?

arch/ia64/kernel/perfmon.c

@@ -82,7 +82,7 @@
 #define PFM_REG_IMPL		0x1 /* register implemented */
 #define PFM_REG_END		0x2 /* end marker */
 #define PFM_REG_MONITOR		(0x1<<4|PFM_REG_IMPL) /* a PMC with a pmc.pm field only */
-#define PFM_REG_COUNTING	(0x2<<4|PFM_REG_MONITOR|PFM_REG_IMPL) /* a monitor + pmc.oi+ PMD used as a counter */
+#define PFM_REG_COUNTING	(0x2<<4|PFM_REG_MONITOR) /* a monitor + pmc.oi+ PMD used as a counter */
 #define PFM_REG_CONTROL		(0x4<<4|PFM_REG_IMPL) /* PMU control register */
 #define	PFM_REG_CONFIG		(0x8<<4|PFM_REG_IMPL) /* configuration register */
 #define PFM_REG_BUFFER	 	(0xc<<4|PFM_REG_IMPL) /* PMD used as buffer */
@@ -109,14 +109,15 @@
 #define PMD_PMD_DEP(i)	   pmu_conf.pmd_desc[i].dep_pmd[0]
 #define PMC_PMD_DEP(i)	   pmu_conf.pmc_desc[i].dep_pmd[0]
 
-/* k assumed unsigned (up to 64 registers) */
-#define IBR_IS_IMPL(k)	  (k< IA64_NUM_DBG_REGS)
-#define DBR_IS_IMPL(k)	  (k< IA64_NUM_DBG_REGS)
+#define PFM_NUM_IBRS	  IA64_NUM_DBG_REGS
+#define PFM_NUM_DBRS	  IA64_NUM_DBG_REGS
 
 #define CTX_OVFL_NOBLOCK(c)	((c)->ctx_fl_block == 0)
 #define CTX_HAS_SMPL(c)		((c)->ctx_fl_is_sampling)
 #define PFM_CTX_TASK(h)		(h)->ctx_task
 
+#define PMU_PMC_OI		5 /* position of pmc.oi bit */
+
 /* XXX: does not support more than 64 PMDs */
 #define CTX_USED_PMD(ctx, mask) (ctx)->ctx_used_pmds[0] |= (mask)
 #define CTX_IS_USED_PMD(ctx, c) (((ctx)->ctx_used_pmds[0] & (1UL << (c))) != 0UL)
@@ -218,6 +219,8 @@
 /*
  * debugging
  */
+#define PFM_DEBUGGING 1
+#ifdef PFM_DEBUGGING
 #define DPRINT(a) \
 	do { \
 		if (unlikely(pfm_sysctl.debug >0)) { printk("%s.%d: CPU%d [%d] ", __FUNCTION__, __LINE__, smp_processor_id(), current->pid); printk a; } \
@@ -227,18 +230,7 @@
 	do { \
 		if (unlikely(pfm_sysctl.debug > 0 && pfm_sysctl.debug_ovfl >0)) { printk("%s.%d: CPU%d [%d] ", __FUNCTION__, __LINE__, smp_processor_id(), current->pid); printk a; } \
 	} while (0)
-/*
- * Architected PMC structure
- */
-typedef struct {
-	unsigned long pmc_plm:4;	/* privilege level mask */
-	unsigned long pmc_ev:1;		/* external visibility */
-	unsigned long pmc_oi:1;		/* overflow interrupt */
-	unsigned long pmc_pm:1;		/* privileged monitor */
-	unsigned long pmc_ig1:1;	/* reserved */
-	unsigned long pmc_es:8;		/* event select */
-	unsigned long pmc_ig2:48;	/* reserved */
-} pfm_monitor_t;
+#endif
 
 /*
  * 64-bit software counter structure
@@ -469,20 +461,13 @@ typedef struct {
 #define PFM_CMD_STOP		0x08	/* command does not work on zombie context */
 
 
-#define PFM_CMD_IDX(cmd)	(cmd)
-#define PFM_CMD_IS_VALID(cmd)	((PFM_CMD_IDX(cmd) >= 0) && (PFM_CMD_IDX(cmd) < PFM_CMD_COUNT) \
-				  && pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_func != NULL)
-
-#define PFM_CMD_NAME(cmd)	pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_name
-#define PFM_CMD_READ_ARG(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_flags & PFM_CMD_ARG_READ)
-#define PFM_CMD_RW_ARG(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_flags & PFM_CMD_ARG_RW)
-#define PFM_CMD_USE_FD(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_flags & PFM_CMD_FD)
-#define PFM_CMD_STOPPED(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_flags & PFM_CMD_STOP)
+#define PFM_CMD_NAME(cmd)	pfm_cmd_tab[(cmd)].cmd_name
+#define PFM_CMD_READ_ARG(cmd)	(pfm_cmd_tab[(cmd)].cmd_flags & PFM_CMD_ARG_READ)
+#define PFM_CMD_RW_ARG(cmd)	(pfm_cmd_tab[(cmd)].cmd_flags & PFM_CMD_ARG_RW)
+#define PFM_CMD_USE_FD(cmd)	(pfm_cmd_tab[(cmd)].cmd_flags & PFM_CMD_FD)
+#define PFM_CMD_STOPPED(cmd)	(pfm_cmd_tab[(cmd)].cmd_flags & PFM_CMD_STOP)
 
 #define PFM_CMD_ARG_MANY	-1 /* cannot be zero */
-#define PFM_CMD_NARG(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_narg)
-#define PFM_CMD_ARG_SIZE(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_argsize)
-#define PFM_CMD_GETSIZE(cmd)	(pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_getsize)
 
 typedef struct {
 	int	debug;		/* turn on/off debugging via syslog */
@@ -2834,10 +2819,11 @@ static int
 pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
 	struct thread_struct *thread = NULL;
+	struct task_struct *task;
 	pfarg_reg_t *req = (pfarg_reg_t *)arg;
-	unsigned long value;
-	unsigned long smpl_pmds, reset_pmds;
-	unsigned int cnum, reg_flags, flags;
+	unsigned long value, pmc_pm;
+	unsigned long smpl_pmds, reset_pmds, impl_pmds;
+	unsigned int cnum, reg_flags, flags, pmc_type;
 	int i, can_access_pmu = 0, is_loaded, is_system;
 	int is_monitor, is_counting, state;
 	int ret = -EINVAL;
@@ -2846,12 +2832,13 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	state     = ctx->ctx_state;
 	is_loaded = state == PFM_CTX_LOADED ? 1 : 0;
 	is_system = ctx->ctx_fl_system;
+	task      = ctx->ctx_task;
+	impl_pmds = pmu_conf.impl_pmds[0];
 
 	if (state == PFM_CTX_TERMINATED || state == PFM_CTX_ZOMBIE) return -EINVAL;
 
-
 	if (is_loaded) {
-		thread = &ctx->ctx_task->thread;
+		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
@@ -2861,7 +2848,7 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			DPRINT(("[%d] should be running on CPU%d\n", current->pid, ctx->ctx_cpu));
 			return -EBUSY;
 		}
-		can_access_pmu = GET_PMU_OWNER() == ctx->ctx_task || is_system ? 1 : 0;
+		can_access_pmu = GET_PMU_OWNER() == task || is_system ? 1 : 0;
 	}
 
 	for (i = 0; i < count; i++, req++) {
@@ -2873,16 +2860,24 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		reset_pmds = req->reg_reset_pmds[0];
 		flags      = 0;
 
-		is_counting = PMC_IS_COUNTING(cnum);
-		is_monitor  = PMC_IS_MONITOR(cnum);
+
+		if (cnum >= PMU_MAX_PMCS) {
+			DPRINT(("pmc%u is invalid\n", cnum));
+			goto error;
+		}
+
+		pmc_type   = pmu_conf.pmc_desc[cnum].type;
+		pmc_pm     = (value >> pmu_conf.pmc_desc[cnum].pm_pos) & 0x1;
+		is_counting = (pmc_type & PFM_REG_COUNTING) == PFM_REG_COUNTING ? 1 : 0;
+		is_monitor  = (pmc_type & PFM_REG_MONITOR) == PFM_REG_MONITOR ? 1 : 0;
 
 		/*
 		 * we reject all non implemented PMC as well
 		 * as attempts to modify PMC[0-3] which are used
 		 * as status registers by the PMU
 		 */
-		if (PMC_IS_IMPL(cnum) == 0 || PMC_IS_CONTROL(cnum)) {
-			DPRINT(("pmc%u is unimplemented or invalid\n", cnum));
+		if ((pmc_type & PFM_REG_IMPL) == 0 || (pmc_type & PFM_REG_CONTROL) == PFM_REG_CONTROL) {
+			DPRINT(("pmc%u is unimplemented or no-access pmc_type=%x\n", cnum, pmc_type));
 			goto error;
 		}
 		/*
@@ -2890,21 +2885,20 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		 * 	- system-wide session: PMCx.pm=1 (privileged monitor)
 		 * 	- per-task           : PMCx.pm=0 (user monitor)
 		 */
-		if ((is_monitor || is_counting) && value != PMC_DFL_VAL(cnum) && PFM_CHECK_PMC_PM(ctx, cnum, value)) {
-			DPRINT(("pmc%u pmc_pm=%ld fl_system=%d\n",
+		if (is_monitor && value != PMC_DFL_VAL(cnum) && is_system ^ pmc_pm) {
+			DPRINT(("pmc%u pmc_pm=%lu is_system=%d\n",
 				cnum,
-				PMC_PM(cnum, value),
-				ctx->ctx_fl_system));
+				pmc_pm,
+				is_system));
 			goto error;
 		}
 
 		if (is_counting) {
-			pfm_monitor_t *p = (pfm_monitor_t *)&value;
 			/*
 		 	 * enforce generation of overflow interrupt. Necessary on all
 		 	 * CPUs.
 		 	 */
-			p->pmc_oi = 1;
+			value |= 1 << PMU_PMC_OI;
 
 			if (reg_flags & PFM_REGFL_OVFL_NOTIFY) {
 				flags |= PFM_REGFL_OVFL_NOTIFY;
@@ -2913,13 +2907,13 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			if (reg_flags & PFM_REGFL_RANDOM) flags |= PFM_REGFL_RANDOM;
 
 			/* verify validity of smpl_pmds */
-			if ((smpl_pmds & pmu_conf.impl_pmds[0]) != smpl_pmds) {
+			if ((smpl_pmds & impl_pmds) != smpl_pmds) {
 				DPRINT(("invalid smpl_pmds 0x%lx for pmc%u\n", smpl_pmds, cnum));
 				goto error;
 			}
 
 			/* verify validity of reset_pmds */
-			if ((reset_pmds & pmu_conf.impl_pmds[0]) != reset_pmds) {
+			if ((reset_pmds & impl_pmds) != reset_pmds) {
 				DPRINT(("invalid reset_pmds 0x%lx for pmc%u\n", reset_pmds, cnum));
 				goto error;
 			}
@@ -2935,7 +2929,7 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		 * execute write checker, if any
 		 */
 		if (PMC_WR_FUNC(cnum)) {
-			ret = PMC_WR_FUNC(cnum)(ctx->ctx_task, ctx, cnum, &value, regs);
+			ret = PMC_WR_FUNC(cnum)(task, ctx, cnum, &value, regs);
 			if (ret) goto error;
 			ret = -EINVAL;
 		}
@@ -2997,7 +2991,7 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		 *
 		 * The value in ctx_pmcs[] can only be changed in pfm_write_pmcs().
 		 *
-		 * The value in t->pmc[] may be modified on overflow, i.e.,  when
+		 * The value in thread->pmcs[] may be modified on overflow, i.e.,  when
 		 * monitoring needs to be stopped.
 		 */
 		if (is_monitor) CTX_USED_MONITOR(ctx, 1UL << cnum);
@@ -3056,11 +3050,6 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	return 0;
 error:
 	PFM_REG_RETFLAG_SET(req->reg_flags, PFM_REG_RETFL_EINVAL);
-
-	req->reg_flags = PFM_REG_RETFL_EINVAL;
-
-	DPRINT(("pmc[%u]=0x%lx error %d\n", cnum, value, ret));
-
 	return ret;
 }
 
@@ -3068,6 +3057,7 @@ static int
 pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
 	struct thread_struct *thread = NULL;
+	struct task_struct *task;
 	pfarg_reg_t *req = (pfarg_reg_t *)arg;
 	unsigned long value, hw_value, ovfl_mask;
 	unsigned int cnum;
@@ -3080,25 +3070,26 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	is_loaded = state == PFM_CTX_LOADED ? 1 : 0;
 	is_system = ctx->ctx_fl_system;
 	ovfl_mask = pmu_conf.ovfl_val;
+	task      = ctx->ctx_task;
 
-	if (state == PFM_CTX_TERMINATED || state == PFM_CTX_ZOMBIE) return -EINVAL;
+	if (unlikely(state == PFM_CTX_TERMINATED || state == PFM_CTX_ZOMBIE)) return -EINVAL;
 
 	/*
 	 * on both UP and SMP, we can only write to the PMC when the task is
 	 * the owner of the local PMU.
 	 */
-	if (is_loaded) {
-		thread = &ctx->ctx_task->thread;
+	if (likely(is_loaded)) {
+		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
 		 * It does not have to be the owner (ctx_task) of the context per se.
 		 */
-		if (is_system && ctx->ctx_cpu != smp_processor_id()) {
+		if (unlikely(is_system && ctx->ctx_cpu != smp_processor_id())) {
 			DPRINT(("[%d] should be running on CPU%d\n", current->pid, ctx->ctx_cpu));
 			return -EBUSY;
 		}
-		can_access_pmu = GET_PMU_OWNER() == ctx->ctx_task || is_system ? 1 : 0;
+		can_access_pmu = GET_PMU_OWNER() == task || is_system ? 1 : 0;
 	}
 
 	for (i = 0; i < count; i++, req++) {
@@ -3118,7 +3109,7 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		if (PMD_WR_FUNC(cnum)) {
 			unsigned long v = value;
 
-			ret = PMD_WR_FUNC(cnum)(ctx->ctx_task, ctx, cnum, &v, regs);
+			ret = PMD_WR_FUNC(cnum)(task, ctx, cnum, &v, regs);
 			if (ret) goto abort_mission;
 
 			value = v;
@@ -3243,16 +3234,6 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	 * for now, we have only one possibility for error
 	 */
 	PFM_REG_RETFLAG_SET(req->reg_flags, PFM_REG_RETFL_EINVAL);
-
-	/*
-	 * we change the return value to EFAULT in case we cannot write register return code.
-	 * The caller first must correct this error, then a resubmission of the request will
-	 * eventually yield the EINVAL.
-	 */
-	req->reg_flags = PFM_REG_RETFL_EINVAL;
-
-	DPRINT(("pmd[%u]=0x%lx ret %d\n", cnum, value, ret));
-
 	return ret;
 }
 
@@ -3269,11 +3250,12 @@ static int
 pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
 	struct thread_struct *thread = NULL;
-	unsigned long val = 0UL, lval, ovfl_mask;
+	struct task_struct *task;
+	unsigned long val = 0UL, lval, ovfl_mask, sval;
 	pfarg_reg_t *req = (pfarg_reg_t *)arg;
 	unsigned int cnum, reg_flags = 0;
 	int i, can_access_pmu = 0, state;
-	int is_loaded, is_system;
+	int is_loaded, is_system, is_counting;
 	int ret = -EINVAL;
 
 	/*
@@ -3285,32 +3267,33 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	is_loaded = state == PFM_CTX_LOADED ? 1 : 0;
 	is_system = ctx->ctx_fl_system;
 	ovfl_mask = pmu_conf.ovfl_val;
+	task      = ctx->ctx_task;
 
 	if (state == PFM_CTX_ZOMBIE) return -EINVAL;
 
-	if (is_loaded) {
-		thread = &ctx->ctx_task->thread;
+	if (likely(is_loaded)) {
+		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
 		 * It does not have to be the owner (ctx_task) of the context per se.
 		 */
-		if (is_system && ctx->ctx_cpu != smp_processor_id()) {
+		if (unlikely(is_system && ctx->ctx_cpu != smp_processor_id())) {
 			DPRINT(("[%d] should be running on CPU%d\n", current->pid, ctx->ctx_cpu));
 			return -EBUSY;
 		}
 		/*
 		 * this can be true when not self-monitoring only in UP
 		 */
-		can_access_pmu = GET_PMU_OWNER() == ctx->ctx_task || is_system ? 1 : 0;
+		can_access_pmu = GET_PMU_OWNER() == task || is_system ? 1 : 0;
 
 		if (can_access_pmu) ia64_srlz_d();
 	}
 
-	DPRINT(("enter loaded=%d access_pmu=%d ctx_state=%d\n",
+	DPRINT(("loaded=%d access_pmu=%d ctx_state=%d\n",
 		is_loaded,
 		can_access_pmu,
-		ctx->ctx_state));
+		state));
 
 	/*
 	 * on both UP and SMP, we can only read the PMD from the hardware register when
@@ -3319,11 +3302,10 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 
 	for (i = 0; i < count; i++, req++) {
 
-		lval        = 0UL;
 		cnum        = req->reg_num;
 		reg_flags   = req->reg_flags;
 
-		if (!PMD_IS_IMPL(cnum)) goto error;
+		if (unlikely(!PMD_IS_IMPL(cnum))) goto error;
 		/*
 		 * we can only read the register that we use. That includes
 		 * the one we explicitely initialize AND the one we want included
@@ -3332,7 +3314,11 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		 * Having this restriction allows optimization in the ctxsw routine
 		 * without compromising security (leaks)
 		 */
-		if (!CTX_IS_USED_PMD(ctx, cnum)) goto error;
+		if (unlikely(!CTX_IS_USED_PMD(ctx, cnum))) goto error;
+
+		sval        = ctx->ctx_pmds[cnum].val;
+		lval        = ctx->ctx_pmds[cnum].lval;
+		is_counting = PMD_IS_COUNTING(cnum);
 
 		/*
 		 * If the task is not the current one, then we check if the
@@ -3347,23 +3333,21 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			 * if context is zombie, then task does not exist anymore.
 			 * In this case, we use the full value saved in the context (pfm_flush_regs()).
 			 */
-			val = state == PFM_CTX_LOADED ? thread->pmds[cnum] : 0UL;
+			val = is_loaded ? thread->pmds[cnum] : 0UL;
 		}
 
-		if (PMD_IS_COUNTING(cnum)) {
+		if (is_counting) {
 			/*
 			 * XXX: need to check for overflow when loaded
 			 */
 			val &= ovfl_mask;
-			val += ctx->ctx_pmds[cnum].val;
-
-			lval = ctx->ctx_pmds[cnum].lval;
+			val += sval;
 		}
 
 		/*
 		 * execute read checker, if any
 		 */
-		if (PMD_RD_FUNC(cnum)) {
+		if (unlikely(PMD_RD_FUNC(cnum))) {
 			unsigned long v = val;
 			ret = PMD_RD_FUNC(cnum)(ctx->ctx_task, ctx, cnum, &v, regs);
 			if (ret) goto error;
@@ -3373,12 +3357,7 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 
 		PFM_REG_RETFLAG_SET(reg_flags, 0);
 
-		DPRINT(("pmd[%u]=0x%lx loaded=%d access_pmu=%d ctx_state=%d\n",
-			cnum,
-			val,
-			is_loaded,
-			can_access_pmu,
-			ctx->ctx_state));
+		DPRINT(("pmd[%u]=0x%lx\n", cnum, val));
 
 		/*
 		 * update register return value, abort all if problem during copy.
@@ -3393,12 +3372,7 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	return 0;
 
 error:
-	PFM_REG_RETFLAG_SET(reg_flags, PFM_REG_RETFL_EINVAL);
-
-	req->reg_flags = PFM_REG_RETFL_EINVAL;
-
-	DPRINT(("error pmd[%u]=0x%lx\n", cnum, val));
-
+	PFM_REG_RETFLAG_SET(req->reg_flags, PFM_REG_RETFL_EINVAL);
 	return ret;
 }
 
@@ -3628,7 +3602,7 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			prefetch(ctx->ctx_smpl_hdr);
 
 			rst_ctrl.bits.mask_monitoring = 0;
-			rst_ctrl.bits.reset_ovfl_pmds = 1;
+			rst_ctrl.bits.reset_ovfl_pmds = 0;
 
 			if (state == PFM_CTX_LOADED)
 				ret = pfm_buf_fmt_restart_active(fmt, task, &rst_ctrl, ctx->ctx_smpl_hdr, regs);
@@ -3748,6 +3722,7 @@ static int
 pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
 	struct thread_struct *thread = NULL;
+	struct task_struct *task;
 	pfarg_dbreg_t *req = (pfarg_dbreg_t *)arg;
 	unsigned long flags;
 	dbreg_t dbreg;
@@ -3762,6 +3737,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 	state     = ctx->ctx_state;
 	is_loaded = state == PFM_CTX_LOADED ? 1 : 0;
 	is_system = ctx->ctx_fl_system;
+	task      = ctx->ctx_task;
 
 	if (state == PFM_CTX_TERMINATED || state == PFM_CTX_ZOMBIE) return -EINVAL;
 
@@ -3770,17 +3746,17 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 	 * the owner of the local PMU.
 	 */
 	if (is_loaded) {
-		thread = &ctx->ctx_task->thread;
+		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
 		 * It does not have to be the owner (ctx_task) of the context per se.
 		 */
-		if (is_system && ctx->ctx_cpu != smp_processor_id()) {
+		if (unlikely(is_system && ctx->ctx_cpu != smp_processor_id())) {
 			DPRINT(("[%d] should be running on CPU%d\n", current->pid, ctx->ctx_cpu));
 			return -EBUSY;
 		}
-		can_access_pmu = GET_PMU_OWNER() == ctx->ctx_task || is_system ? 1 : 0;
+		can_access_pmu = GET_PMU_OWNER() == task || is_system ? 1 : 0;
 	}
 
 	/*
@@ -3796,7 +3772,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 	 * don't bother if we are loaded and task is being debugged
 	 */
 	if (is_loaded && (thread->flags & IA64_THREAD_DBG_VALID) != 0) {
-		DPRINT(("debug registers already in use for [%d]\n", ctx->ctx_task->pid));
+		DPRINT(("debug registers already in use for [%d]\n", task->pid));
 		return -EBUSY;
 	}
 
@@ -3837,7 +3813,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 	 * is shared by all processes running on it
  	 */
 	if (first_time && can_access_pmu) {
-		DPRINT(("[%d] clearing ibrs, dbrs\n", ctx->ctx_task->pid));
+		DPRINT(("[%d] clearing ibrs, dbrs\n", task->pid));
 		for (i=0; i < pmu_conf.num_ibrs; i++) {
 			ia64_set_ibr(i, 0UL);
 			ia64_srlz_i();
@@ -3860,7 +3836,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 
 		ret = -EINVAL;
 
-		if ((mode == PFM_CODE_RR && !IBR_IS_IMPL(rnum)) || ((mode == PFM_DATA_RR) && !DBR_IS_IMPL(rnum))) {
+		if ((mode == PFM_CODE_RR && rnum >= PFM_NUM_IBRS) || ((mode == PFM_DATA_RR) && rnum >= PFM_NUM_DBRS)) {
 			DPRINT(("invalid register %u val=0x%lx mode=%d i=%d count=%d\n",
 				  rnum, dbreg.val, mode, i, count));
 
@@ -4434,6 +4410,7 @@ pfm_context_unload(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg
 	struct task_struct *task = PFM_CTX_TASK(ctx);
 	struct pt_regs *tregs;
 	int state, is_system;
+	int ret;
 
 	DPRINT(("ctx_state=%d task [%d]\n", ctx->ctx_state, task ? task->pid : -1));
 
@@ -4451,7 +4428,8 @@ pfm_context_unload(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg
 	/*
 	 * clear psr and dcr bits
 	 */
-	pfm_stop(ctx, NULL, 0, regs);
+	ret = pfm_stop(ctx, NULL, 0, regs);
+	if (ret) return ret;
 
 	ctx->ctx_state = state = PFM_CTX_UNLOADED;
 
@@ -4760,37 +4738,45 @@ sys_perfmonctl (int fd, int cmd, void *arg, int count, long arg5, long arg6, lon
 	void *args_k = NULL;
 	long ret; /* will expand int return types */
 	size_t base_sz, sz, xtra_sz = 0;
-	int narg, completed_args = 0, call_made = 0;
+	int narg, completed_args = 0, call_made = 0, cmd_flags;
+	int (*func)(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs);
+	int (*getsize)(void *arg, size_t *sz);
 #define PFM_MAX_ARGSIZE	4096
 
 	/*
-	 * reject any call if perfmon was disabled at initialization time
-	 mask*/
-	if (PFM_IS_DISABLED()) return -ENOSYS;
+	 * reject any call if perfmon was disabled at initialization
+	 */
+	if (unlikely(PFM_IS_DISABLED())) return -ENOSYS;
 
-	if (unlikely(PFM_CMD_IS_VALID(cmd) == 0)) {
+	if (unlikely(cmd < 0 || cmd >= PFM_CMD_COUNT)) {
 		DPRINT(("[%d] invalid cmd=%d\n", current->pid, cmd));
 		return -EINVAL;
 	}
 
-	DPRINT(("cmd=%s idx=%d valid=%d narg=0x%x argsz=%lu count=%d\n",
+	func      = pfm_cmd_tab[cmd].cmd_func;
+	narg      = pfm_cmd_tab[cmd].cmd_narg;
+	base_sz   = pfm_cmd_tab[cmd].cmd_argsize;
+	getsize   = pfm_cmd_tab[cmd].cmd_getsize;
+	cmd_flags = pfm_cmd_tab[cmd].cmd_flags;
+
+	if (unlikely(func == NULL)) {
+		DPRINT(("[%d] invalid cmd=%d\n", current->pid, cmd));
+		return -EINVAL;
+	}
+
+	DPRINT(("cmd=%s idx=%d narg=0x%x argsz=%lu count=%d\n",
 		PFM_CMD_NAME(cmd),
-		PFM_CMD_IDX(cmd),
-		PFM_CMD_IS_VALID(cmd),
-		PFM_CMD_NARG(cmd),
-		PFM_CMD_ARG_SIZE(cmd), 
+		cmd,
+		narg,
+		base_sz,
 		count));
 
 	/*
 	 * check if number of arguments matches what the command expects
 	 */
-	narg = PFM_CMD_NARG(cmd);
-	if ((narg == PFM_CMD_ARG_MANY && count <= 0) || (narg > 0 && narg != count))
+	if (unlikely((narg == PFM_CMD_ARG_MANY && count <= 0) || (narg > 0 && narg != count)))
 		return -EINVAL;
 
-	/* get single argument size */
-	base_sz = PFM_CMD_ARG_SIZE(cmd);
-
 restart_args:
 	sz = xtra_sz + base_sz*count;
 	/*
@@ -4804,7 +4790,7 @@ sys_perfmonctl (int fd, int cmd, void *arg, int count, long arg5, long arg6, lon
 	/*
 	 * allocate default-sized argument buffer
 	 */
-	if (count && args_k == NULL) {
+	if (likely(count && args_k == NULL)) {
 		args_k = kmalloc(PFM_MAX_ARGSIZE, GFP_KERNEL);
 		if (args_k == NULL) return -ENOMEM;
 	}
@@ -4824,11 +4810,11 @@ sys_perfmonctl (int fd, int cmd, void *arg, int count, long arg5, long arg6, lon
 	/*
 	 * check if command supports extra parameters
 	 */
-	if (completed_args == 0 && PFM_CMD_GETSIZE(cmd)) {
+	if (completed_args == 0 && getsize) {
 		/*
 		 * get extra parameters size (based on main argument)
 		 */
-		ret = PFM_CMD_GETSIZE(cmd)(args_k, &xtra_sz);
+		ret = (*getsize)(args_k, &xtra_sz);
 		if (ret) goto error_args;
 
 		completed_args = 1;
@@ -4836,45 +4822,45 @@ sys_perfmonctl (int fd, int cmd, void *arg, int count, long arg5, long arg6, lon
 		DPRINT(("[%d] restart_args sz=%lu xtra_sz=%lu\n", current->pid, sz, xtra_sz));
 
 		/* retry if necessary */
-		if (xtra_sz) goto restart_args;
+		if (likely(xtra_sz)) goto restart_args;
 	}
 
-	if (PFM_CMD_USE_FD(cmd))  {
-
-		ret = -EBADF;
+	if (unlikely((cmd_flags & PFM_CMD_FD) == 0)) goto skip_fd;
 
-		file = fget(fd);
-		if (file == NULL) {
-			DPRINT(("[%d] invalid fd %d\n", current->pid, fd));
-			goto error_args;
-		}
-		if (PFM_IS_FILE(file) == 0) {
-			DPRINT(("[%d] fd %d not related to perfmon\n", current->pid, fd));
-			goto error_args;
-		}
+	ret = -EBADF;
 
+	file = fget(fd);
+	if (unlikely(file == NULL)) {
+		DPRINT(("[%d] invalid fd %d\n", current->pid, fd));
+		goto error_args;
+	}
+	if (unlikely(PFM_IS_FILE(file) == 0)) {
+		DPRINT(("[%d] fd %d not related to perfmon\n", current->pid, fd));
+		goto error_args;
+	}
 
-		ctx = (pfm_context_t *)file->private_data;
-		if (ctx == NULL) {
-			DPRINT(("[%d] no context for fd %d\n", current->pid, fd));
-			goto error_args;
-		}
+	ctx = (pfm_context_t *)file->private_data;
+	if (unlikely(ctx == NULL)) {
+		DPRINT(("[%d] no context for fd %d\n", current->pid, fd));
+		goto error_args;
+	}
+	prefetch(&ctx->ctx_state);
 
-		PROTECT_CTX(ctx, flags);
+	PROTECT_CTX(ctx, flags);
 
-		/*
-		 * check task is stopped
-		 */
-		ret = pfm_check_task_state(ctx, cmd, flags);
-		if (ret) goto abort_locked;
-	}
+	/*
+	 * check task is stopped
+	 */
+	ret = pfm_check_task_state(ctx, cmd, flags);
+	if (unlikely(ret)) goto abort_locked;
 
-	ret = (*pfm_cmd_tab[PFM_CMD_IDX(cmd)].cmd_func)(ctx, args_k, count, regs);
+skip_fd:
+	ret = (*func)(ctx, args_k, count, regs);
 
 	call_made = 1;
 
 abort_locked:
-	if (ctx) {
+	if (likely(ctx)) {
 		DPRINT(("[%d] context unlocked\n", current->pid));
 		UNPROTECT_CTX(ctx, flags);
 		fput(file);
@@ -4907,7 +4893,7 @@ pfm_resume_after_ovfl(pfm_context_t *ctx, unsigned long ovfl_regs, struct pt_reg
 	if (CTX_HAS_SMPL(ctx)) {
 
 		rst_ctrl.bits.mask_monitoring = 0;
-		rst_ctrl.bits.reset_ovfl_pmds = 1;
+		rst_ctrl.bits.reset_ovfl_pmds = 0;
 
 		if (state == PFM_CTX_LOADED)
 			ret = pfm_buf_fmt_restart_active(fmt, current, &rst_ctrl, ctx->ctx_smpl_hdr, regs);
@@ -5096,7 +5082,7 @@ pfm_ovfl_notify_user(pfm_context_t *ctx, unsigned long ovfl_pmds)
 		msg->pfm_ovfl_msg.msg_ovfl_pmds[1] = 0UL;
 		msg->pfm_ovfl_msg.msg_ovfl_pmds[2] = 0UL;
 		msg->pfm_ovfl_msg.msg_ovfl_pmds[3] = 0UL;
-		msg->pfm_ovfl_msg.msg_tstamp       = ia64_get_itc(); /* relevant on UP only */
+		msg->pfm_ovfl_msg.msg_tstamp       = 0UL;
 	}
 
 	DPRINT(("ovfl msg: msg=%p no_msg=%d fd=%d pid=%d ovfl_pmds=0x%lx\n",
@@ -5119,10 +5105,12 @@ pfm_end_notify_user(pfm_context_t *ctx)
 		printk(KERN_ERR "perfmon: pfm_end_notify_user no more notification msgs\n");
 		return -1;
 	}
+	/* no leak */
+	memset(msg, 0, sizeof(*msg));
 
 	msg->pfm_end_msg.msg_type    = PFM_MSG_END;
 	msg->pfm_end_msg.msg_ctx_fd  = ctx->ctx_fd;
-	msg->pfm_ovfl_msg.msg_tstamp = ia64_get_itc(); /* relevant on UP only */
+	msg->pfm_ovfl_msg.msg_tstamp = 0UL;
 
 	DPRINT(("end msg: msg=%p no_msg=%d ctx_fd=%d pid=%d\n",
 		msg,
@@ -5141,8 +5129,8 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 {
 	pfm_ovfl_arg_t ovfl_arg;
 	unsigned long mask;
-	unsigned long old_val, ovfl_val;
-	unsigned long ovfl_notify = 0UL, ovfl_pmds = 0UL, smpl_pmds = 0UL;
+	unsigned long old_val, ovfl_val, new_val;
+	unsigned long ovfl_notify = 0UL, ovfl_pmds = 0UL, smpl_pmds = 0UL, reset_pmds;
 	unsigned long tstamp;
 	pfm_ovfl_ctrl_t	ovfl_ctrl;
 	unsigned int i, has_smpl;
@@ -5155,21 +5143,19 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 	 */
 	if (unlikely((pmc0 & 0x1) == 0)) goto sanity_check;
 
-	tstamp = ia64_get_itc();
-
+	tstamp   = ia64_get_itc();
 	mask     = pmc0 >> PMU_FIRST_COUNTER;
 	ovfl_val = pmu_conf.ovfl_val;
+	has_smpl = CTX_HAS_SMPL(ctx);
 
 	DPRINT_ovfl(("pmc0=0x%lx pid=%d iip=0x%lx, %s "
-		     "used_pmds=0x%lx reload_pmcs=0x%lx\n",
+		     "used_pmds=0x%lx\n",
 			pmc0,
 			task ? task->pid: -1,
 			(regs ? regs->cr_iip : 0),
 			CTX_OVFL_NOBLOCK(ctx) ? "nonblocking" : "blocking",
-			ctx->ctx_used_pmds[0],
-			ctx->ctx_reload_pmcs[0]));
+			ctx->ctx_used_pmds[0]));
 
-	has_smpl = CTX_HAS_SMPL(ctx);
 
 	/*
 	 * first we update the virtual counters
@@ -5180,29 +5166,31 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 		/* skip pmd which did not overflow */
 		if ((mask & 0x1) == 0) continue;
 
-		DPRINT_ovfl(("pmd[%d] overflowed hw_pmd=0x%lx ctx_pmd=0x%lx\n",
-			i, ia64_get_pmd(i), ctx->ctx_pmds[i].val));
-
 		/*
 		 * Note that the pmd is not necessarily 0 at this point as qualified events
 		 * may have happened before the PMU was frozen. The residual count is not
 		 * taken into consideration here but will be with any read of the pmd via
 		 * pfm_read_pmds().
 		 */
-		old_val               = ctx->ctx_pmds[i].val;
-		ctx->ctx_pmds[i].val += 1 + ovfl_val;
+		old_val              = new_val = ctx->ctx_pmds[i].val;
+		new_val             += 1 + ovfl_val;
+		ctx->ctx_pmds[i].val = new_val;
 
 		/*
 		 * check for overflow condition
 		 */
-		if (likely(old_val > ctx->ctx_pmds[i].val)) {
+		if (likely(old_val > new_val)) {
 			ovfl_pmds |= 1UL << i;
 			if (PMC_OVFL_NOTIFY(ctx, i)) ovfl_notify |= 1UL << i;
 		}
 
-		DPRINT_ovfl(("ctx_pmd[%d].val=0x%lx old_val=0x%lx pmd=0x%lx ovfl_pmds=0x%lx ovfl_notify=0x%lx smpl_pmds=0x%lx\n",
-			i, ctx->ctx_pmds[i].val, old_val,
-			ia64_get_pmd(i) & ovfl_val, ovfl_pmds, ovfl_notify, smpl_pmds));
+		DPRINT_ovfl(("ctx_pmd[%d].val=0x%lx old_val=0x%lx pmd=0x%lx ovfl_pmds=0x%lx ovfl_notify=0x%lx\n",
+			i,
+			new_val,
+			old_val,
+			ia64_get_pmd(i) & ovfl_val,
+			ovfl_pmds,
+			ovfl_notify));
 	}
 
 	/*
@@ -5214,6 +5202,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 	 * reset all control bits
 	 */
 	ovfl_ctrl.val = 0;
+	reset_pmds    = 0UL;
 
 	/*
 	 * if a sampling format module exists, then we "cache" the overflow by 
@@ -5225,7 +5214,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 		int j, k, ret = 0;
 		int this_cpu = smp_processor_id();
 
-		pmd_mask = ovfl_pmds >> PMU_FIRST_COUNTER;
+		pmd_mask   = ovfl_pmds >> PMU_FIRST_COUNTER;
 
 		prefetch(ctx->ctx_smpl_hdr);
 
@@ -5275,7 +5264,10 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 			ovfl_ctrl.bits.notify_user     |= ovfl_arg.ovfl_ctrl.bits.notify_user;
 			ovfl_ctrl.bits.block_task      |= ovfl_arg.ovfl_ctrl.bits.block_task;
 			ovfl_ctrl.bits.mask_monitoring |= ovfl_arg.ovfl_ctrl.bits.mask_monitoring;
-			ovfl_ctrl.bits.reset_ovfl_pmds |= ovfl_arg.ovfl_ctrl.bits.reset_ovfl_pmds; /* yes or no */
+			/*
+			 * build the bitmask of pmds to reset now
+			 */
+			if (ovfl_arg.ovfl_ctrl.bits.reset_ovfl_pmds) reset_pmds |= mask;
 
 			pfm_stats[this_cpu].pfm_smpl_handler_cycles += end_cycles - start_cycles;
 		}
@@ -5287,6 +5279,10 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 				current->pid,
 				pmd_mask<<PMU_FIRST_COUNTER));
 		}
+		/*
+		 * remove the pmds we reset now from the set of pmds to reset in pfm_restart()
+		 */
+		ovfl_pmds &= ~reset_pmds;
 	} else {
 		/*
 		 * when no sampling module is used, then the default
@@ -5296,14 +5292,21 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 		ovfl_ctrl.bits.block_task      = ovfl_notify ? 1 : 0;
 		ovfl_ctrl.bits.mask_monitoring = ovfl_notify ? 1 : 0; /* XXX: change for saturation */
 		ovfl_ctrl.bits.reset_ovfl_pmds = ovfl_notify ? 0 : 1;
+		/*
+		 * if needed, we reset all overflowed pmds
+		 */
+		if (ovfl_notify == 0) reset_pmds = ovfl_pmds;
 	}
 
+	DPRINT(("current [%d] ovfl_pmds=0x%lx reset_pmds=0x%lx\n",
+		current->pid,
+		ovfl_pmds,
+		reset_pmds));
 	/*
-	 * if we (still) have some overflowed PMD but no notification is requested
-	 * then we use the short reset period.
+	 * reset the requested PMD registers using the short reset values
 	 */
-	if (ovfl_ctrl.bits.reset_ovfl_pmds) {
-		unsigned long bm = ovfl_pmds;
+	if (reset_pmds) {
+		unsigned long bm = reset_pmds;
 		pfm_reset_regs(ctx, &bm, PFM_PMD_SHORT_RESET);
 	}
 

arch/ia64/kernel/perfmon_default_smpl.c

@@ -178,6 +178,7 @@ default_handler(struct task_struct *task, void *buf, pfm_ovfl_arg_t *arg, struct
 	ent->tstamp    = stamp;
 	ent->cpu       = smp_processor_id();
 	ent->set       = arg->active_set;
+	ent->tgid      = current->tgid;
 
 	/*
 	 * selectively store PMDs in increasing index number

arch/ia64/kernel/process.c

@@ -259,10 +259,12 @@ ia64_load_extra (struct task_struct *task)
  *
  * We get here through the following  call chain:
  *
- *	<clone syscall>
- *	sys_clone
- *	do_fork
- *	copy_thread
+ *	from user-level:	from kernel:
+ *
+ *	<clone syscall>	        <some kernel call frames>
+ *	sys_clone		   :
+ *	do_fork			do_fork
+ *	copy_thread		copy_thread
  *
  * This means that the stack layout is as follows:
  *
@@ -276,9 +278,6 @@ ia64_load_extra (struct task_struct *task)
  *	|                     | <-- sp (lowest addr)
  *	+---------------------+
  *
- * Note: if we get called through kernel_thread() then the memory above "(highest addr)"
- * is valid kernel stack memory that needs to be copied as well.
- *
  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
@@ -291,9 +290,9 @@ copy_thread (int nr, unsigned long clone_flags,
 	     unsigned long user_stack_base, unsigned long user_stack_size,
 	     struct task_struct *p, struct pt_regs *regs)
 {
-	unsigned long rbs, child_rbs, rbs_size, stack_offset, stack_top, stack_used;
-	struct switch_stack *child_stack, *stack;
 	extern char ia64_ret_from_clone, ia32_ret_from_clone;
+	struct switch_stack *child_stack, *stack;
+	unsigned long rbs, child_rbs, rbs_size;
 	struct pt_regs *child_ptregs;
 	int retval = 0;
 
@@ -306,16 +305,13 @@ copy_thread (int nr, unsigned long clone_flags,
 		return 0;
 #endif
 
-	stack_top = (unsigned long) current + IA64_STK_OFFSET;
 	stack = ((struct switch_stack *) regs) - 1;
-	stack_used = stack_top - (unsigned long) stack;
-	stack_offset = IA64_STK_OFFSET - stack_used;
 
-	child_stack = (struct switch_stack *) ((unsigned long) p + stack_offset);
-	child_ptregs = (struct pt_regs *) (child_stack + 1);
+	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
+	child_stack = (struct switch_stack *) child_ptregs - 1;
 
 	/* copy parent's switch_stack & pt_regs to child: */
-	memcpy(child_stack, stack, stack_used);
+	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 
 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
@@ -324,7 +320,7 @@ copy_thread (int nr, unsigned long clone_flags,
 	/* copy the parent's register backing store to the child: */
 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 
-	if (user_mode(child_ptregs)) {
+	if (likely(user_mode(child_ptregs))) {
 		if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
 		if (user_stack_base) {
@@ -341,14 +337,14 @@ copy_thread (int nr, unsigned long clone_flags,
 		 * been taken care of by the caller of sys_clone()
 		 * already.
 		 */
-		child_ptregs->r12 = (unsigned long) (child_ptregs + 1); /* kernel sp */
+		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
 	}
+	child_stack->ar_bspstore = child_rbs + rbs_size;
 	if (IS_IA32_PROCESS(regs))
 		child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 	else
 		child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
-	child_stack->ar_bspstore = child_rbs + rbs_size;
 
 	/* copy parts of thread_struct: */
 	p->thread.ksp = (unsigned long) child_stack - 16;
@@ -358,8 +354,8 @@ copy_thread (int nr, unsigned long clone_flags,
 	 * therefore we must specify them explicitly here and not include them in
 	 * IA64_PSR_BITS_TO_CLEAR.
 	 */
-	child_ptregs->cr_ipsr =  ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
-			      & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
+	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
+				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 
 	/*
 	 * NOTE: The calling convention considers all floating point
@@ -578,27 +574,43 @@ ia64_set_personality (struct elf64_hdr *elf_ex, int ibcs2_interpreter)
 pid_t
 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 {
-	struct task_struct *parent = current;
-	int result; 
-	pid_t tid;
+	extern void ia64_invoke_kernel_thread_helper (void);
+	unsigned long *helper_fptr = (unsigned long *) &ia64_invoke_kernel_thread_helper;
+	struct {
+		struct switch_stack sw;
+		struct pt_regs pt;
+	} regs;
+
+	memset(&regs, 0, sizeof(regs));
+	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
+	regs.pt.r1 = helper_fptr[1];		/* set GP */
+	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
+	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
+	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
+	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
+	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
+	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
+	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
+
+	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
+}
+EXPORT_SYMBOL(kernel_thread);
 
-	tid = clone(flags | CLONE_VM | CLONE_UNTRACED, 0);
-	if (parent != current) {
+/* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
+int
+kernel_thread_helper (int (*fn)(void *), void *arg)
+{
 #ifdef CONFIG_IA32_SUPPORT
-		if (IS_IA32_PROCESS(ia64_task_regs(current))) {
-			/* A kernel thread is always a 64-bit process. */
-			current->thread.map_base  = DEFAULT_MAP_BASE;
-			current->thread.task_size = DEFAULT_TASK_SIZE;
-			ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
-			ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
-		}
-#endif
-		result = (*fn)(arg);
-		_exit(result);
+	if (IS_IA32_PROCESS(ia64_task_regs(current))) {
+		/* A kernel thread is always a 64-bit process. */
+		current->thread.map_base  = DEFAULT_MAP_BASE;
+		current->thread.task_size = DEFAULT_TASK_SIZE;
+		ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
+		ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 	}
-	return tid;
+#endif
+	return (*fn)(arg);
 }
-EXPORT_SYMBOL(kernel_thread);
 
 /*
  * Flush thread state.  This is called when a thread does an execve().

arch/ia64/mm/hugetlbpage.c

 /*
  * IA-64 Huge TLB Page Support for Kernel.
  *
- * Copyright (C) 2002, Rohit Seth <rohit.seth@intel.com>
+ * Copyright (C) 2002-2004 Rohit Seth <rohit.seth@intel.com>
+ * Copyright (C) 2003-2004 Ken Chen <kenneth.w.chen@intel.com>
+ *
+ * Sep, 2003: add numa support
+ * Feb, 2004: dynamic hugetlb page size via boot parameter
  */
 
 #include <linux/config.h>
@@ -18,11 +22,10 @@
 #include <asm/tlb.h>
 #include <asm/tlbflush.h>
 
-#define TASK_HPAGE_BASE (REGION_HPAGE << REGION_SHIFT)
-
 static long	htlbpagemem;
 int		htlbpage_max;
 static long	htlbzone_pages;
+unsigned int	hpage_shift=HPAGE_SHIFT_DEFAULT;
 
 static struct list_head hugepage_freelists[MAX_NUMNODES];
 static spinlock_t htlbpage_lock = SPIN_LOCK_UNLOCKED;
@@ -407,7 +410,7 @@ unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, u
 		return -EINVAL;
 	/* This code assumes that REGION_HPAGE != 0. */
 	if ((REGION_NUMBER(addr) != REGION_HPAGE) || (addr & (HPAGE_SIZE - 1)))
-		addr = TASK_HPAGE_BASE;
+		addr = HPAGE_REGION_BASE;
 	else
 		addr = ALIGN(addr, HPAGE_SIZE);
 	for (vmm = find_vma(current->mm, addr); ; vmm = vmm->vm_next) {
@@ -520,6 +523,35 @@ static int __init hugetlb_setup(char *s)
 }
 __setup("hugepages=", hugetlb_setup);
 
+static int __init hugetlb_setup_sz(char *str)
+{
+	u64 tr_pages;
+	unsigned long long size;
+
+	if (ia64_pal_vm_page_size(&tr_pages, NULL) != 0)
+		/*
+		 * shouldn't happen, but just in case.
+		 */
+		tr_pages = 0x15557000UL;
+
+	size = memparse(str, &str);
+	if (*str || (size & (size-1)) || !(tr_pages & size) ||
+		size <= PAGE_SIZE ||
+		size >= (1UL << PAGE_SHIFT << MAX_ORDER)) {
+		printk(KERN_WARNING "Invalid huge page size specified\n");
+		return 1;
+	}
+
+	hpage_shift = __ffs(size);
+	/*
+	 * boot cpu already executed ia64_mmu_init, and has HPAGE_SHIFT_DEFAULT
+	 * override here with new page shift.
+	 */
+	ia64_set_rr(HPAGE_REGION_BASE, hpage_shift << 2);
+	return 1;
+}
+__setup("hugepagesz=", hugetlb_setup_sz);
+
 static int __init hugetlb_init(void)
 {
 	int i;
@@ -540,7 +572,7 @@ static int __init hugetlb_init(void)
 	printk("Total HugeTLB memory allocated, %ld\n", htlbpagemem);
 	return 0;
 }
-module_init(hugetlb_init);
+__initcall(hugetlb_init);
 
 int hugetlb_report_meminfo(char *buf)
 {

arch/ia64/mm/init.c

@@ -342,6 +342,10 @@ ia64_mmu_init (void *my_cpu_data)
 
 	ia64_tlb_init();
 
+#ifdef	CONFIG_HUGETLB_PAGE
+	ia64_set_rr(HPAGE_REGION_BASE, HPAGE_SHIFT << 2);
+#endif
+
 #ifdef	CONFIG_IA64_MCA
 	cpu = smp_processor_id();
 

arch/ppc64/kernel/prom.c

@@ -546,7 +546,7 @@ prom_initialize_lmb(unsigned long mem)
 			opt++;
 		if (!strncmp(opt, RELOC("off"), 3))
 			nodart = 1;
-		else if (!strncmp(opt, RELOC("on"), 2))
+		else if (!strncmp(opt, RELOC("force"), 5))
 			RELOC(dart_force_on) = 1;
 	}
 #else

arch/ppc64/kernel/vio.c

@@ -29,6 +29,8 @@
 
 #define DBGENTER() pr_debug("%s entered\n", __FUNCTION__)
 
+extern struct subsystem devices_subsys; /* needed for vio_find_name() */
+
 struct iommu_table *vio_build_iommu_table(struct vio_dev *dev);
 
 static int vio_num_address_cells;
@@ -157,8 +159,7 @@ static int __init vio_bus_init(void)
 
 	node_vroot = find_devices("vdevice");
 	if ((node_vroot == NULL) || (node_vroot->child == NULL)) {
-		printk(KERN_INFO "VIO: missing or empty /vdevice node; no virtual IO"
-			" devices present.\n");
+		/* this machine doesn't do virtual IO, and that's ok */
 		return 0;
 	}
 
@@ -260,7 +261,7 @@ struct vio_dev * __devinit vio_register_device(struct device_node *of_node)
 	/* init generic 'struct device' fields: */
 	viodev->dev.parent = &vio_bus_device->dev;
 	viodev->dev.bus = &vio_bus_type;
-	snprintf(viodev->dev.bus_id, BUS_ID_SIZE, "%lx", viodev->unit_address);
+	snprintf(viodev->dev.bus_id, BUS_ID_SIZE, "%x", viodev->unit_address);
 	viodev->dev.release = vio_dev_release;
 
 	/* register with generic device framework */
@@ -299,6 +300,42 @@ const void * vio_get_attribute(struct vio_dev *vdev, void* which, int* length)
 }
 EXPORT_SYMBOL(vio_get_attribute);
 
+/* vio_find_name() - internal because only vio.c knows how we formatted the
+ * kobject name
+ * XXX once vio_bus_type.devices is actually used as a kset in
+ * drivers/base/bus.c, this function should be removed in favor of
+ * "device_find(kobj_name, &vio_bus_type)"
+ */
+static struct vio_dev *vio_find_name(const char *kobj_name)
+{
+	struct kobject *found;
+
+	found = kset_find_obj(&devices_subsys.kset, kobj_name);
+	if (!found)
+		return NULL;
+
+	return to_vio_dev(container_of(found, struct device, kobj));
+}
+
+/**
+ * vio_find_node - find an already-registered vio_dev
+ * @vnode: device_node of the virtual device we're looking for
+ */
+struct vio_dev *vio_find_node(struct device_node *vnode)
+{
+	uint32_t *unit_address;
+	char kobj_name[BUS_ID_SIZE];
+
+	/* construct the kobject name from the device node */
+	unit_address = (uint32_t *)get_property(vnode, "reg", NULL);
+	if (!unit_address)
+		return NULL;
+	snprintf(kobj_name, BUS_ID_SIZE, "%x", *unit_address);
+
+	return vio_find_name(kobj_name);
+}
+EXPORT_SYMBOL(vio_find_node);
+
 /**
  * vio_build_iommu_table: - gets the dma information from OF and builds the TCE tree.
  * @dev: the virtual device.

arch/x86_64/kernel/x8664_ksyms.c

@@ -196,6 +196,7 @@ EXPORT_SYMBOL(die_chain);
 
 #ifdef CONFIG_SMP
 EXPORT_SYMBOL(cpu_sibling_map);
+EXPORT_SYMBOL(smp_num_siblings);
 #endif
 
 extern void do_softirq_thunk(void);

drivers/net/ibmveth.c

@@ -66,16 +66,16 @@
   printk(KERN_INFO "%s: " fmt, __FILE__, ## args)
 
 #define ibmveth_error_printk(fmt, args...) \
-  printk(KERN_ERR "(%s:%3.3d ua:%lx) ERROR: " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
+  printk(KERN_ERR "(%s:%3.3d ua:%x) ERROR: " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
 
 #ifdef DEBUG
 #define ibmveth_debug_printk_no_adapter(fmt, args...) \
   printk(KERN_DEBUG "(%s:%3.3d): " fmt, __FILE__, __LINE__ , ## args)
 #define ibmveth_debug_printk(fmt, args...) \
-  printk(KERN_DEBUG "(%s:%3.3d ua:%lx): " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
+  printk(KERN_DEBUG "(%s:%3.3d ua:%x): " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
 #define ibmveth_assert(expr) \
   if(!(expr)) {                                   \
-    printk(KERN_DEBUG "assertion failed (%s:%3.3d ua:%lx): %s\n", __FILE__, __LINE__, adapter->vdev->unit_address, #expr); \
+    printk(KERN_DEBUG "assertion failed (%s:%3.3d ua:%x): %s\n", __FILE__, __LINE__, adapter->vdev->unit_address, #expr); \
     BUG(); \
   }
 #else
@@ -869,7 +869,7 @@ static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_
 	unsigned int *mcastFilterSize_p;
 
 
-	ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%lx\n", 
+	ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%x\n", 
 					dev->unit_address);
 
 	mac_addr_p = (unsigned int *) vio_get_attribute(dev, VETH_MAC_ADDR, 0);
@@ -1014,7 +1014,7 @@ static int ibmveth_seq_show(struct seq_file *seq, void *v)
 
 	seq_printf(seq, "%s %s\n\n", ibmveth_driver_string, ibmveth_driver_version);
 	
-	seq_printf(seq, "Unit Address:    0x%lx\n", adapter->vdev->unit_address);
+	seq_printf(seq, "Unit Address:    0x%x\n", adapter->vdev->unit_address);
 	seq_printf(seq, "LIOBN:           0x%lx\n", adapter->liobn);
 	seq_printf(seq, "Current MAC:     %02X:%02X:%02X:%02X:%02X:%02X\n",
 		   current_mac[0], current_mac[1], current_mac[2],

drivers/usb/class/usblp.c

@@ -603,7 +603,7 @@ static ssize_t usblp_write(struct file *file, const char __user *buffer, size_t
 {
 	DECLARE_WAITQUEUE(wait, current);
 	struct usblp *usblp = file->private_data;
-	int timeout, err = 0, transfer_length;
+	int timeout, err = 0, transfer_length = 0;
 	size_t writecount = 0;
 
 	while (writecount < count) {
@@ -654,6 +654,16 @@ static ssize_t usblp_write(struct file *file, const char __user *buffer, size_t
 			continue;
 		}
 
+		/* We must increment writecount here, and not at the
+		 * end of the loop. Otherwise, the final loop iteration may
+		 * be skipped, leading to incomplete printer output.
+		 */
+		writecount += transfer_length;
+		if (writecount == count) {
+			up(&usblp->sem);
+			break;
+		}
+
 		transfer_length=(count - writecount);
 		if (transfer_length > USBLP_BUF_SIZE)
 			transfer_length = USBLP_BUF_SIZE;
@@ -677,8 +687,6 @@ static ssize_t usblp_write(struct file *file, const char __user *buffer, size_t
 			break;
 		}
 		up (&usblp->sem);
-
-		writecount += transfer_length;
 	}
 
 	return count;

fs/compat_ioctl.c

@@ -1458,6 +1458,7 @@ static int cdrom_do_generic_command(unsigned int fd, unsigned int cmd, unsigned
 	struct cdrom_generic_command *cgc;
 	struct cdrom_generic_command32 *cgc32;
 	unsigned char dir;
+	int itmp;
 
 	cgc = compat_alloc_user_space(sizeof(*cgc));
 	cgc32 = compat_ptr(arg);
@@ -1469,12 +1470,16 @@ static int cdrom_do_generic_command(unsigned int fd, unsigned int cmd, unsigned
 	    __cgc_do_ptr((void **) &cgc->sense, &cgc32->sense))
 		return -EFAULT;
 
-	if (get_user(dir, &cgc->data_direction) ||
-	    put_user(dir, &cgc32->data_direction))
+	if (get_user(dir, &cgc32->data_direction) ||
+	    put_user(dir, &cgc->data_direction))
 		return -EFAULT;
 
-	if (copy_in_user(&cgc->quiet, &cgc32->quiet,
-			 2 * sizeof(int)))
+	if (get_user(itmp, &cgc32->quiet) ||
+	    put_user(itmp, &cgc->quiet))
+		return -EFAULT;
+
+	if (get_user(itmp, &cgc32->timeout) ||
+	    put_user(itmp, &cgc->timeout))
 		return -EFAULT;
 
 	if (__cgc_do_ptr(&cgc->reserved[0], &cgc32->reserved[0]))

fs/hpfs/buffer.c

@@ -62,56 +62,28 @@ void hpfs_unlock_inode(struct inode *i)
 
 void hpfs_lock_2inodes(struct inode *i1, struct inode *i2)
 {
-	struct hpfs_inode_info *hpfs_i1 = NULL, *hpfs_i2 = NULL;
-
-	if (!i1) {
-		if (i2) {
-			hpfs_i2 = hpfs_i(i2);
+	if (!i2 || i1 == i2) {
+		hpfs_lock_inode(i1);
+	} else if (!i1) {
+		hpfs_lock_inode(i2);
+	} else {
+		struct hpfs_inode_info *hpfs_i1 = hpfs_i(i1);
+		struct hpfs_inode_info *hpfs_i2 = hpfs_i(i2);
+		if (i1->i_ino < i2->i_ino) {
+			down(&hpfs_i1->i_sem);
+			down(&hpfs_i2->i_sem);
+		} else {
 			down(&hpfs_i2->i_sem);
-		}
-		return;
-	}
-	if (!i2) {
-		if (i1) {
-			hpfs_i1 = hpfs_i(i1);
 			down(&hpfs_i1->i_sem);
 		}
-		return;
 	}
-	if (i1->i_ino < i2->i_ino) {
-		down(&hpfs_i1->i_sem);
-		down(&hpfs_i2->i_sem);
-	} else if (i1->i_ino > i2->i_ino) {
-		down(&hpfs_i2->i_sem);
-		down(&hpfs_i1->i_sem);
-	} else down(&hpfs_i1->i_sem);
 }
 
 void hpfs_unlock_2inodes(struct inode *i1, struct inode *i2)
 {
-	struct hpfs_inode_info *hpfs_i1 = NULL, *hpfs_i2 = NULL;
-
-	if (!i1) {
-		if (i2) {
-			hpfs_i2 = hpfs_i(i2);
-			up(&hpfs_i2->i_sem);
-		}
-		return;
-	}
-	if (!i2) {
-		if (i1) {
-			hpfs_i1 = hpfs_i(i1);
-			up(&hpfs_i1->i_sem);
-		}
-		return;
-	}
-	if (i1->i_ino < i2->i_ino) {
-		up(&hpfs_i2->i_sem);
-		up(&hpfs_i1->i_sem);
-	} else if (i1->i_ino > i2->i_ino) {
-		up(&hpfs_i1->i_sem);
-		up(&hpfs_i2->i_sem);
-	} else up(&hpfs_i1->i_sem);
+	/* order of up() doesn't matter here */
+	hpfs_unlock_inode(i1);
+	hpfs_unlock_inode(i2);
 }
 
 void hpfs_lock_3inodes(struct inode *i1, struct inode *i2, struct inode *i3)

include/asm-ia64/iosapic.h

@@ -45,9 +45,9 @@
 /*
  * Mask bit
  */
+
 #define	IOSAPIC_MASK_SHIFT		16
-#define	IOSAPIC_UNMASK			0
-#define	IOSAPIC_MSAK			1
+#define	IOSAPIC_MASK			(1<<IOSAPIC_MASK_SHIFT)
 
 #ifndef __ASSEMBLY__
 

include/asm-ia64/mmu_context.h

@@ -140,8 +140,9 @@ reload_context (mm_context_t context)
 {
 	unsigned long rid;
 	unsigned long rid_incr = 0;
-	unsigned long rr0, rr1, rr2, rr3, rr4;
+	unsigned long rr0, rr1, rr2, rr3, rr4, old_rr4;
 
+	old_rr4 = ia64_get_rr(0x8000000000000000);
 	rid = context << 3;	/* make space for encoding the region number */
 	rid_incr = 1 << 8;
 
@@ -152,7 +153,7 @@ reload_context (mm_context_t context)
 	rr3 = rr0 + 3*rid_incr;
 	rr4 = rr0 + 4*rid_incr;
 #ifdef  CONFIG_HUGETLB_PAGE
-	rr4 = (rr4 & (~(0xfcUL))) | (HPAGE_SHIFT << 2);
+	rr4 = (rr4 & (~(0xfcUL))) | (old_rr4 & 0xfc);
 #endif
 
 	ia64_set_rr(0x0000000000000000, rr0);

include/asm-ia64/page.h

@@ -37,31 +37,14 @@
 #define RGN_MAP_LIMIT	((1UL << (4*PAGE_SHIFT - 12)) - PAGE_SIZE)	/* per region addr limit */
 
 #ifdef CONFIG_HUGETLB_PAGE
+# define REGION_HPAGE		(4UL)	/* note: this is hardcoded in reload_context()!*/
+# define REGION_SHIFT		61
+# define HPAGE_REGION_BASE	(REGION_HPAGE << REGION_SHIFT)
+# define HPAGE_SHIFT		hpage_shift
+# define HPAGE_SHIFT_DEFAULT	28	/* check ia64 SDM for architecture supported size */
+# define HPAGE_SIZE		(__IA64_UL_CONST(1) << HPAGE_SHIFT)
+# define HPAGE_MASK		(~(HPAGE_SIZE - 1))
 
-# if defined(CONFIG_HUGETLB_PAGE_SIZE_4GB)
-#  define HPAGE_SHIFT	32
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_1GB)
-#  define HPAGE_SHIFT	30
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_256MB)
-#  define HPAGE_SHIFT	28
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_64MB)
-#  define HPAGE_SHIFT	26
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_16MB)
-#  define HPAGE_SHIFT	24
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_4MB)
-#  define HPAGE_SHIFT	22
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_1MB)
-#  define HPAGE_SHIFT	20
-# elif defined(CONFIG_HUGETLB_PAGE_SIZE_256KB)
-#  define HPAGE_SHIFT	18
-# else
-#  error Unsupported IA-64 HugeTLB Page Size!
-# endif
-
-# define REGION_HPAGE	(4UL)	/* note: this is hardcoded in mmu_context.h:reload_context()!*/
-# define REGION_SHIFT	61
-# define HPAGE_SIZE	(__IA64_UL_CONST(1) << HPAGE_SHIFT)
-# define HPAGE_MASK	(~(HPAGE_SIZE - 1))
 # define HAVE_ARCH_HUGETLB_UNMAPPED_AREA
 # define ARCH_HAS_HUGEPAGE_ONLY_RANGE
 #endif /* CONFIG_HUGETLB_PAGE */
@@ -140,6 +123,7 @@ typedef union ia64_va {
 # define is_hugepage_only_range(addr, len)		\
 	 (REGION_NUMBER(addr) == REGION_HPAGE &&	\
 	  REGION_NUMBER((addr)+(len)) == REGION_HPAGE)
+extern unsigned int hpage_shift;
 #endif
 
 static __inline__ int

include/asm-ia64/perfmon_default_smpl.h

@@ -59,7 +59,7 @@ typedef struct {
  * last_reset_value member indicates the initial value of the overflowed PMD. 
  */
 typedef struct {
-        int             pid;                    /* active process at PMU interrupt point */
+        int             pid;                    /* thread id (for NPTL, this is gettid()) */
         unsigned char   reserved1[3];           /* reserved for future use */
         unsigned char   ovfl_pmd;               /* index of overflowed PMD */
 
@@ -69,7 +69,7 @@ typedef struct {
 
         unsigned short  cpu;                    /* cpu on which the overfow occured */
         unsigned short  set;                    /* event set active when overflow ocurred   */
-        unsigned int    reserved2;              /* for future use */
+        int    		tgid;              	/* thread group id (for NPTL, this is getpid()) */
 } pfm_default_smpl_entry_t;
 
 #define PFM_DEFAULT_MAX_PMDS		64 /* how many pmds supported by data structures (sizeof(unsigned long) */

include/asm-ia64/scatterlist.h

@@ -2,7 +2,7 @@
 #define _ASM_IA64_SCATTERLIST_H
 
 /*
- * Modified 1998-1999, 2001-2002
+ * Modified 1998-1999, 2001-2002, 2004
  *	David Mosberger-Tang <davidm@hpl.hp.com>, Hewlett-Packard Co
  */
 
@@ -15,6 +15,14 @@ struct scatterlist {
 	unsigned int dma_length;
 };
 
-#define ISA_DMA_THRESHOLD	(~0UL)
+/*
+ * It used to be that ISA_DMA_THRESHOLD had something to do with the
+ * DMA-limits of ISA-devices.  Nowadays, its only remaining use (apart
+ * from the aha1542.c driver, which isn't 64-bit clean anyhow) is to
+ * tell the block-layer (via BLK_BOUNCE_ISA) what the max. physical
+ * address of a page is that is allocated with GFP_DMA.  On IA-64,
+ * that's 4GB - 1.
+ */
+#define ISA_DMA_THRESHOLD	0xffffffff
 
 #endif /* _ASM_IA64_SCATTERLIST_H */

include/asm-ppc64/vio.h

@@ -94,8 +94,8 @@ static inline struct vio_driver *to_vio_driver(struct device_driver *drv)
 struct vio_dev {
 	struct device_node *archdata;   /* Open Firmware node */
 	void *driver_data;              /* data private to the driver */
-	unsigned long unit_address;	
 	struct iommu_table *iommu_table;     /* vio_map_* uses this */
+	uint32_t unit_address;	
 	unsigned int irq;
 
 	struct device dev;

include/net/tcp.h

@@ -263,7 +263,10 @@ static __inline__ int tw_del_dead_node(struct tcp_tw_bucket *tw)
 #define tw_for_each(tw, node, head) \
 	hlist_for_each_entry(tw, node, head, tw_node)
 
-#define tw_for_each_inmate(tw, node, safe, jail) \
+#define tw_for_each_inmate(tw, node, jail) \
+	hlist_for_each_entry(tw, node, jail, tw_death_node)
+
+#define tw_for_each_inmate_safe(tw, node, safe, jail) \
 	hlist_for_each_entry_safe(tw, node, safe, jail, tw_death_node)
 
 #define tcptw_sk(__sk)	((struct tcp_tw_bucket *)(__sk))

net/ipv4/tcp_minisocks.c

@@ -427,7 +427,7 @@ static u32 twkill_thread_slots;
 static int tcp_do_twkill_work(int slot, unsigned int quota)
 {
 	struct tcp_tw_bucket *tw;
-	struct hlist_node *node, *safe;
+	struct hlist_node *node;
 	unsigned int killed;
 	int ret;
 
@@ -439,8 +439,8 @@ static int tcp_do_twkill_work(int slot, unsigned int quota)
 	 */
 	killed = 0;
 	ret = 0;
-	tw_for_each_inmate(tw, node, safe,
-			   &tcp_tw_death_row[slot]) {
+rescan:
+	tw_for_each_inmate(tw, node, &tcp_tw_death_row[slot]) {
 		__tw_del_dead_node(tw);
 		spin_unlock(&tw_death_lock);
 		tcp_timewait_kill(tw);
@@ -451,6 +451,14 @@ static int tcp_do_twkill_work(int slot, unsigned int quota)
 			ret = 1;
 			break;
 		}
+
+		/* While we dropped tw_death_lock, another cpu may have
+		 * killed off the next TW bucket in the list, therefore
+		 * do a fresh re-read of the hlist head node with the
+		 * lock reacquired.  We still use the hlist traversal
+		 * macro in order to get the prefetches.
+		 */
+		goto rescan;
 	}
 
 	tcp_tw_count -= killed;
@@ -637,7 +645,7 @@ void tcp_twcal_tick(unsigned long dummy)
 			struct hlist_node *node, *safe;
 			struct tcp_tw_bucket *tw;
 
-			tw_for_each_inmate(tw, node, safe,
+			tw_for_each_inmate_safe(tw, node, safe,
 					   &tcp_twcal_row[slot]) {
 				__tw_del_dead_node(tw);
 				tcp_timewait_kill(tw);

sound/oss/Kconfig

@@ -1147,7 +1147,7 @@ config SOUND_RME96XX
 	help
 	  Say Y or M if you have a Hammerfall or Hammerfall light
 	  multichannel card from RME. If you want to acess advanced
-	  features of the card, read Documentation/sound/rme96xx.
+	  features of the card, read Documentation/sound/oss/rme96xx.
 
 config SOUND_AD1980
 	tristate "AD1980 front/back switch plugin"