Automatic merge of rsync://rsync.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6.git

arch/ia64/Kconfig

@@ -329,7 +329,7 @@ menu "Power management and ACPI"
 
 config PM
 	bool "Power Management support"
-	depends on IA64_GENERIC || IA64_DIG || IA64_HP_ZX1 || IA64_HP_ZX1_SWIOTLB
+	depends on !IA64_HP_SIM
 	default y
 	help
 	  "Power Management" means that parts of your computer are shut

arch/ia64/configs/sn2_defconfig

@@ -574,6 +574,8 @@ CONFIG_SERIAL_NONSTANDARD=y
 # CONFIG_N_HDLC is not set
 # CONFIG_STALDRV is not set
 CONFIG_SGI_SNSC=y
+CONFIG_SGI_TIOCX=y
+CONFIG_SGI_MBCS=m
 
 #
 # Serial drivers

arch/ia64/hp/common/sba_iommu.c

 /*
 **  IA64 System Bus Adapter (SBA) I/O MMU manager
 **
-**	(c) Copyright 2002-2004 Alex Williamson
+**	(c) Copyright 2002-2005 Alex Williamson
 **	(c) Copyright 2002-2003 Grant Grundler
-**	(c) Copyright 2002-2004 Hewlett-Packard Company
+**	(c) Copyright 2002-2005 Hewlett-Packard Company
 **
 **	Portions (c) 2000 Grant Grundler (from parisc I/O MMU code)
 **	Portions (c) 1999 Dave S. Miller (from sparc64 I/O MMU code)
@@ -459,21 +459,32 @@ get_iovp_order (unsigned long size)
  * sba_search_bitmap - find free space in IO PDIR resource bitmap
  * @ioc: IO MMU structure which owns the pdir we are interested in.
  * @bits_wanted: number of entries we need.
+ * @use_hint: use res_hint to indicate where to start looking
  *
  * Find consecutive free bits in resource bitmap.
  * Each bit represents one entry in the IO Pdir.
  * Cool perf optimization: search for log2(size) bits at a time.
  */
 static SBA_INLINE unsigned long
-sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted)
+sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted, int use_hint)
 {
-	unsigned long *res_ptr = ioc->res_hint;
+	unsigned long *res_ptr;
 	unsigned long *res_end = (unsigned long *) &(ioc->res_map[ioc->res_size]);
-	unsigned long pide = ~0UL;
+	unsigned long flags, pide = ~0UL;
 
 	ASSERT(((unsigned long) ioc->res_hint & (sizeof(unsigned long) - 1UL)) == 0);
 	ASSERT(res_ptr < res_end);
 
+	spin_lock_irqsave(&ioc->res_lock, flags);
+
+	/* Allow caller to force a search through the entire resource space */
+	if (likely(use_hint)) {
+		res_ptr = ioc->res_hint;
+	} else {
+		res_ptr = (ulong *)ioc->res_map;
+		ioc->res_bitshift = 0;
+	}
+
 	/*
 	 * N.B.  REO/Grande defect AR2305 can cause TLB fetch timeouts
 	 * if a TLB entry is purged while in use.  sba_mark_invalid()
@@ -570,10 +581,12 @@ sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted)
 	prefetch(ioc->res_map);
 	ioc->res_hint = (unsigned long *) ioc->res_map;
 	ioc->res_bitshift = 0;
+	spin_unlock_irqrestore(&ioc->res_lock, flags);
 	return (pide);
 
 found_it:
 	ioc->res_hint = res_ptr;
+	spin_unlock_irqrestore(&ioc->res_lock, flags);
 	return (pide);
 }
 
@@ -594,36 +607,36 @@ sba_alloc_range(struct ioc *ioc, size_t size)
 	unsigned long itc_start;
 #endif
 	unsigned long pide;
-	unsigned long flags;
 
 	ASSERT(pages_needed);
 	ASSERT(0 == (size & ~iovp_mask));
 
-	spin_lock_irqsave(&ioc->res_lock, flags);
-
 #ifdef PDIR_SEARCH_TIMING
 	itc_start = ia64_get_itc();
 #endif
 	/*
 	** "seek and ye shall find"...praying never hurts either...
 	*/
-	pide = sba_search_bitmap(ioc, pages_needed);
+	pide = sba_search_bitmap(ioc, pages_needed, 1);
 	if (unlikely(pide >= (ioc->res_size << 3))) {
-		pide = sba_search_bitmap(ioc, pages_needed);
+		pide = sba_search_bitmap(ioc, pages_needed, 0);
 		if (unlikely(pide >= (ioc->res_size << 3))) {
 #if DELAYED_RESOURCE_CNT > 0
+			unsigned long flags;
+
 			/*
 			** With delayed resource freeing, we can give this one more shot.  We're
 			** getting close to being in trouble here, so do what we can to make this
 			** one count.
 			*/
-			spin_lock(&ioc->saved_lock);
+			spin_lock_irqsave(&ioc->saved_lock, flags);
 			if (ioc->saved_cnt > 0) {
 				struct sba_dma_pair *d;
 				int cnt = ioc->saved_cnt;
 
-				d = &(ioc->saved[ioc->saved_cnt]);
+				d = &(ioc->saved[ioc->saved_cnt - 1]);
 
+				spin_lock(&ioc->res_lock);
 				while (cnt--) {
 					sba_mark_invalid(ioc, d->iova, d->size);
 					sba_free_range(ioc, d->iova, d->size);
@@ -631,10 +644,11 @@ sba_alloc_range(struct ioc *ioc, size_t size)
 				}
 				ioc->saved_cnt = 0;
 				READ_REG(ioc->ioc_hpa+IOC_PCOM);	/* flush purges */
+				spin_unlock(&ioc->res_lock);
 			}
-			spin_unlock(&ioc->saved_lock);
+			spin_unlock_irqrestore(&ioc->saved_lock, flags);
 
-			pide = sba_search_bitmap(ioc, pages_needed);
+			pide = sba_search_bitmap(ioc, pages_needed, 0);
 			if (unlikely(pide >= (ioc->res_size << 3)))
 				panic(__FILE__ ": I/O MMU @ %p is out of mapping resources\n",
 				      ioc->ioc_hpa);
@@ -664,8 +678,6 @@ sba_alloc_range(struct ioc *ioc, size_t size)
 		(uint) ((unsigned long) ioc->res_hint - (unsigned long) ioc->res_map),
 		ioc->res_bitshift );
 
-	spin_unlock_irqrestore(&ioc->res_lock, flags);
-
 	return (pide);
 }
 
@@ -950,6 +962,30 @@ sba_map_single(struct device *dev, void *addr, size_t size, int dir)
 	return SBA_IOVA(ioc, iovp, offset);
 }
 
+#ifdef ENABLE_MARK_CLEAN
+static SBA_INLINE void
+sba_mark_clean(struct ioc *ioc, dma_addr_t iova, size_t size)
+{
+	u32	iovp = (u32) SBA_IOVP(ioc,iova);
+	int	off = PDIR_INDEX(iovp);
+	void	*addr;
+
+	if (size <= iovp_size) {
+		addr = phys_to_virt(ioc->pdir_base[off] &
+		                    ~0xE000000000000FFFULL);
+		mark_clean(addr, size);
+	} else {
+		do {
+			addr = phys_to_virt(ioc->pdir_base[off] &
+			                    ~0xE000000000000FFFULL);
+			mark_clean(addr, min(size, iovp_size));
+			off++;
+			size -= iovp_size;
+		} while (size > 0);
+	}
+}
+#endif
+
 /**
  * sba_unmap_single - unmap one IOVA and free resources
  * @dev: instance of PCI owned by the driver that's asking.
@@ -995,6 +1031,10 @@ void sba_unmap_single(struct device *dev, dma_addr_t iova, size_t size, int dir)
 	size += offset;
 	size = ROUNDUP(size, iovp_size);
 
+#ifdef ENABLE_MARK_CLEAN
+	if (dir == DMA_FROM_DEVICE)
+		sba_mark_clean(ioc, iova, size);
+#endif
 
 #if DELAYED_RESOURCE_CNT > 0
 	spin_lock_irqsave(&ioc->saved_lock, flags);
@@ -1021,30 +1061,6 @@ void sba_unmap_single(struct device *dev, dma_addr_t iova, size_t size, int dir)
 	READ_REG(ioc->ioc_hpa+IOC_PCOM);	/* flush purges */
 	spin_unlock_irqrestore(&ioc->res_lock, flags);
 #endif /* DELAYED_RESOURCE_CNT == 0 */
-#ifdef ENABLE_MARK_CLEAN
-	if (dir == DMA_FROM_DEVICE) {
-		u32 iovp = (u32) SBA_IOVP(ioc,iova);
-		int off = PDIR_INDEX(iovp);
-		void *addr;
-
-		if (size <= iovp_size) {
-			addr = phys_to_virt(ioc->pdir_base[off] &
-					    ~0xE000000000000FFFULL);
-			mark_clean(addr, size);
-		} else {
-			size_t byte_cnt = size;
-
-			do {
-				addr = phys_to_virt(ioc->pdir_base[off] &
-				                    ~0xE000000000000FFFULL);
-				mark_clean(addr, min(byte_cnt, iovp_size));
-				off++;
-				byte_cnt -= iovp_size;
-
-			   } while (byte_cnt > 0);
-		}
-	}
-#endif
 }
 
 

arch/ia64/kernel/entry.S

@@ -728,12 +728,8 @@ ENTRY(ia64_leave_syscall)
 	mov f8=f0		// clear f8
 	;;
 	ld8 r30=[r2],16		// M0|1 load cr.ifs
-	mov.m ar.ssd=r0		// M2 clear ar.ssd
-	cmp.eq p9,p0=r0,r0	// set p9 to indicate that we should restore cr.ifs
-	;;
 	ld8 r25=[r3],16		// M0|1 load ar.unat
-	mov.m ar.csd=r0		// M2 clear ar.csd
-	mov r22=r0		// clear r22
+	cmp.eq p9,p0=r0,r0	// set p9 to indicate that we should restore cr.ifs
 	;;
 	ld8 r26=[r2],PT(B0)-PT(AR_PFS)	// M0|1 load ar.pfs
 (pKStk)	mov r22=psr		// M2 read PSR now that interrupts are disabled
@@ -756,11 +752,15 @@ ENTRY(ia64_leave_syscall)
 	mov f7=f0		// clear f7
 	;;
 	ld8.fill r12=[r2]	// restore r12 (sp)
+	mov.m ar.ssd=r0		// M2 clear ar.ssd
+	mov r22=r0		// clear r22
+
 	ld8.fill r15=[r3]	// restore r15
+(pUStk) st1 [r14]=r17
 	addl r3=THIS_CPU(ia64_phys_stacked_size_p8),r0
 	;;
-(pUStk)	ld4 r3=[r3]		// r3 = cpu_data->phys_stacked_size_p8
-(pUStk) st1 [r14]=r17
+(pUStk)	ld4 r17=[r3]		// r17 = cpu_data->phys_stacked_size_p8
+	mov.m ar.csd=r0		// M2 clear ar.csd
 	mov b6=r18		// I0  restore b6
 	;;
 	mov r14=r0		// clear r14

arch/ia64/kernel/irq_ia64.c

@@ -63,20 +63,30 @@ EXPORT_SYMBOL(isa_irq_to_vector_map);
 static unsigned long ia64_vector_mask[BITS_TO_LONGS(IA64_NUM_DEVICE_VECTORS)];
 
 int
-assign_irq_vector (int irq)
+assign_irq_vector_nopanic (int irq)
 {
 	int pos, vector;
  again:
 	pos = find_first_zero_bit(ia64_vector_mask, IA64_NUM_DEVICE_VECTORS);
 	vector = IA64_FIRST_DEVICE_VECTOR + pos;
 	if (vector > IA64_LAST_DEVICE_VECTOR)
-		/* XXX could look for sharable vectors instead of panic'ing... */
-		panic("assign_irq_vector: out of interrupt vectors!");
+		return -1;
 	if (test_and_set_bit(pos, ia64_vector_mask))
 		goto again;
 	return vector;
 }
 
+int
+assign_irq_vector (int irq)
+{
+	int vector = assign_irq_vector_nopanic(irq);
+
+	if (vector < 0)
+		panic("assign_irq_vector: out of interrupt vectors!");
+
+	return vector;
+}
+
 void
 free_irq_vector (int vector)
 {

arch/ia64/kernel/perfmon.c

@@ -479,14 +479,6 @@ typedef struct {
 
 #define PFM_CMD_ARG_MANY	-1 /* cannot be zero */
 
-typedef struct {
-	int	debug;		/* turn on/off debugging via syslog */
-	int	debug_ovfl;	/* turn on/off debug printk in overflow handler */
-	int	fastctxsw;	/* turn on/off fast (unsecure) ctxsw */
-	int	expert_mode;	/* turn on/off value checking */
-	int 	debug_pfm_read;
-} pfm_sysctl_t;
-
 typedef struct {
 	unsigned long pfm_spurious_ovfl_intr_count;	/* keep track of spurious ovfl interrupts */
 	unsigned long pfm_replay_ovfl_intr_count;	/* keep track of replayed ovfl interrupts */
@@ -514,8 +506,8 @@ static LIST_HEAD(pfm_buffer_fmt_list);
 static pmu_config_t		*pmu_conf;
 
 /* sysctl() controls */
-static pfm_sysctl_t pfm_sysctl;
-int pfm_debug_var;
+pfm_sysctl_t pfm_sysctl;
+EXPORT_SYMBOL(pfm_sysctl);
 
 static ctl_table pfm_ctl_table[]={
 	{1, "debug", &pfm_sysctl.debug, sizeof(int), 0666, NULL, &proc_dointvec, NULL,},
@@ -1576,7 +1568,7 @@ pfm_read(struct file *filp, char __user *buf, size_t size, loff_t *ppos)
 		goto abort_locked;
 	}
 
-	DPRINT(("[%d] fd=%d type=%d\n", current->pid, msg->pfm_gen_msg.msg_ctx_fd, msg->pfm_gen_msg.msg_type));
+	DPRINT(("fd=%d type=%d\n", msg->pfm_gen_msg.msg_ctx_fd, msg->pfm_gen_msg.msg_type));
 
 	ret = -EFAULT;
   	if(copy_to_user(buf, msg, sizeof(pfm_msg_t)) == 0) ret = sizeof(pfm_msg_t);
@@ -3695,8 +3687,6 @@ pfm_debug(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 
 	pfm_sysctl.debug = m == 0 ? 0 : 1;
 
-	pfm_debug_var = pfm_sysctl.debug;
-
 	printk(KERN_INFO "perfmon debugging %s (timing reset)\n", pfm_sysctl.debug ? "on" : "off");
 
 	if (m == 0) {
@@ -4996,13 +4986,21 @@ pfm_context_force_terminate(pfm_context_t *ctx, struct pt_regs *regs)
 }
 
 static int pfm_ovfl_notify_user(pfm_context_t *ctx, unsigned long ovfl_pmds);
-
+ /*
+  * pfm_handle_work() can be called with interrupts enabled
+  * (TIF_NEED_RESCHED) or disabled. The down_interruptible
+  * call may sleep, therefore we must re-enable interrupts
+  * to avoid deadlocks. It is safe to do so because this function
+  * is called ONLY when returning to user level (PUStk=1), in which case
+  * there is no risk of kernel stack overflow due to deep
+  * interrupt nesting.
+  */
 void
 pfm_handle_work(void)
 {
 	pfm_context_t *ctx;
 	struct pt_regs *regs;
-	unsigned long flags;
+	unsigned long flags, dummy_flags;
 	unsigned long ovfl_regs;
 	unsigned int reason;
 	int ret;
@@ -5039,18 +5037,15 @@ pfm_handle_work(void)
 	//if (CTX_OVFL_NOBLOCK(ctx)) goto skip_blocking;
 	if (reason == PFM_TRAP_REASON_RESET) goto skip_blocking;
 
+	/*
+	 * restore interrupt mask to what it was on entry.
+	 * Could be enabled/diasbled.
+	 */
 	UNPROTECT_CTX(ctx, flags);
 
-	 /*
-	  * pfm_handle_work() is currently called with interrupts disabled.
-	  * The down_interruptible call may sleep, therefore we
-	  * must re-enable interrupts to avoid deadlocks. It is
-	  * safe to do so because this function is called ONLY
-	  * when returning to user level (PUStk=1), in which case
-	  * there is no risk of kernel stack overflow due to deep
-	  * interrupt nesting.
-	  */
-	BUG_ON(flags & IA64_PSR_I);
+	/*
+	 * force interrupt enable because of down_interruptible()
+	 */
 	local_irq_enable();
 
 	DPRINT(("before block sleeping\n"));
@@ -5064,12 +5059,12 @@ pfm_handle_work(void)
 	DPRINT(("after block sleeping ret=%d\n", ret));
 
 	/*
-	 * disable interrupts to restore state we had upon entering
-	 * this function
+	 * lock context and mask interrupts again
+	 * We save flags into a dummy because we may have
+	 * altered interrupts mask compared to entry in this
+	 * function.
 	 */
-	local_irq_disable();
-
-	PROTECT_CTX(ctx, flags);
+	PROTECT_CTX(ctx, dummy_flags);
 
 	/*
 	 * we need to read the ovfl_regs only after wake-up
@@ -5095,7 +5090,9 @@ pfm_handle_work(void)
 	ctx->ctx_ovfl_regs[0] = 0UL;
 
 nothing_to_do:
-
+	/*
+	 * restore flags as they were upon entry
+	 */
 	UNPROTECT_CTX(ctx, flags);
 }
 

arch/ia64/kernel/perfmon_default_smpl.c

@@ -20,24 +20,17 @@ MODULE_AUTHOR("Stephane Eranian <eranian@hpl.hp.com>");
 MODULE_DESCRIPTION("perfmon default sampling format");
 MODULE_LICENSE("GPL");
 
-MODULE_PARM(debug, "i");
-MODULE_PARM_DESC(debug, "debug");
-
-MODULE_PARM(debug_ovfl, "i");
-MODULE_PARM_DESC(debug_ovfl, "debug ovfl");
-
-
 #define DEFAULT_DEBUG 1
 
 #ifdef DEFAULT_DEBUG
 #define DPRINT(a) \
 	do { \
-		if (unlikely(debug >0)) { printk("%s.%d: CPU%d ", __FUNCTION__, __LINE__, smp_processor_id()); printk a; } \
+		if (unlikely(pfm_sysctl.debug >0)) { printk("%s.%d: CPU%d ", __FUNCTION__, __LINE__, smp_processor_id()); printk a; } \
 	} while (0)
 
 #define DPRINT_ovfl(a) \
 	do { \
-		if (unlikely(debug_ovfl >0)) { printk("%s.%d: CPU%d ", __FUNCTION__, __LINE__, smp_processor_id()); printk a; } \
+		if (unlikely(pfm_sysctl.debug > 0 && pfm_sysctl.debug_ovfl >0)) { printk("%s.%d: CPU%d ", __FUNCTION__, __LINE__, smp_processor_id()); printk a; } \
 	} while (0)
 
 #else
@@ -45,8 +38,6 @@ MODULE_PARM_DESC(debug_ovfl, "debug ovfl");
 #define DPRINT_ovfl(a)
 #endif
 
-static int debug, debug_ovfl;
-
 static int
 default_validate(struct task_struct *task, unsigned int flags, int cpu, void *data)
 {

arch/ia64/kernel/setup.c

@@ -4,10 +4,15 @@
  * Copyright (C) 1998-2001, 2003-2004 Hewlett-Packard Co
  *	David Mosberger-Tang <davidm@hpl.hp.com>
  *	Stephane Eranian <eranian@hpl.hp.com>
- * Copyright (C) 2000, Rohit Seth <rohit.seth@intel.com>
+ * Copyright (C) 2000, 2004 Intel Corp
+ * 	Rohit Seth <rohit.seth@intel.com>
+ * 	Suresh Siddha <suresh.b.siddha@intel.com>
+ * 	Gordon Jin <gordon.jin@intel.com>
  * Copyright (C) 1999 VA Linux Systems
  * Copyright (C) 1999 Walt Drummond <drummond@valinux.com>
  *
+ * 12/26/04 S.Siddha, G.Jin, R.Seth
+ *			Add multi-threading and multi-core detection
  * 11/12/01 D.Mosberger Convert get_cpuinfo() to seq_file based show_cpuinfo().
  * 04/04/00 D.Mosberger renamed cpu_initialized to cpu_online_map
  * 03/31/00 R.Seth	cpu_initialized and current->processor fixes
@@ -296,6 +301,34 @@ mark_bsp_online (void)
 #endif
 }
 
+#ifdef CONFIG_SMP
+static void
+check_for_logical_procs (void)
+{
+	pal_logical_to_physical_t info;
+	s64 status;
+
+	status = ia64_pal_logical_to_phys(0, &info);
+	if (status == -1) {
+		printk(KERN_INFO "No logical to physical processor mapping "
+		       "available\n");
+		return;
+	}
+	if (status) {
+		printk(KERN_ERR "ia64_pal_logical_to_phys failed with %ld\n",
+		       status);
+		return;
+	}
+	/*
+	 * Total number of siblings that BSP has.  Though not all of them 
+	 * may have booted successfully. The correct number of siblings 
+	 * booted is in info.overview_num_log.
+	 */
+	smp_num_siblings = info.overview_tpc;
+	smp_num_cpucores = info.overview_cpp;
+}
+#endif
+
 void __init
 setup_arch (char **cmdline_p)
 {
@@ -356,6 +389,19 @@ setup_arch (char **cmdline_p)
 
 #ifdef CONFIG_SMP
 	cpu_physical_id(0) = hard_smp_processor_id();
+
+	cpu_set(0, cpu_sibling_map[0]);
+	cpu_set(0, cpu_core_map[0]);
+
+	check_for_logical_procs();
+	if (smp_num_cpucores > 1)
+		printk(KERN_INFO
+		       "cpu package is Multi-Core capable: number of cores=%d\n",
+		       smp_num_cpucores);
+	if (smp_num_siblings > 1)
+		printk(KERN_INFO
+		       "cpu package is Multi-Threading capable: number of siblings=%d\n",
+		       smp_num_siblings);
 #endif
 
 	cpu_init();	/* initialize the bootstrap CPU */
@@ -459,12 +505,23 @@ show_cpuinfo (struct seq_file *m, void *v)
 		   "cpu regs   : %u\n"
 		   "cpu MHz    : %lu.%06lu\n"
 		   "itc MHz    : %lu.%06lu\n"
-		   "BogoMIPS   : %lu.%02lu\n\n",
+		   "BogoMIPS   : %lu.%02lu\n",
 		   cpunum, c->vendor, family, c->model, c->revision, c->archrev,
 		   features, c->ppn, c->number,
 		   c->proc_freq / 1000000, c->proc_freq % 1000000,
 		   c->itc_freq / 1000000, c->itc_freq % 1000000,
 		   lpj*HZ/500000, (lpj*HZ/5000) % 100);
+#ifdef CONFIG_SMP
+	seq_printf(m, "siblings   : %u\n", c->num_log);
+	if (c->threads_per_core > 1 || c->cores_per_socket > 1)
+		seq_printf(m,
+		   	   "physical id: %u\n"
+		   	   "core id    : %u\n"
+		   	   "thread id  : %u\n",
+		   	   c->socket_id, c->core_id, c->thread_id);
+#endif
+	seq_printf(m,"\n");
+
 	return 0;
 }
 
@@ -533,6 +590,14 @@ identify_cpu (struct cpuinfo_ia64 *c)
 	memcpy(c->vendor, cpuid.field.vendor, 16);
 #ifdef CONFIG_SMP
 	c->cpu = smp_processor_id();
+
+	/* below default values will be overwritten  by identify_siblings() 
+	 * for Multi-Threading/Multi-Core capable cpu's
+	 */
+	c->threads_per_core = c->cores_per_socket = c->num_log = 1;
+	c->socket_id = -1;
+
+	identify_siblings(c);
 #endif
 	c->ppn = cpuid.field.ppn;
 	c->number = cpuid.field.number;

arch/ia64/kernel/smpboot.c

 /*
  * SMP boot-related support
  *
- * Copyright (C) 1998-2003 Hewlett-Packard Co
+ * Copyright (C) 1998-2003, 2005 Hewlett-Packard Co
  *	David Mosberger-Tang <davidm@hpl.hp.com>
+ * Copyright (C) 2001, 2004-2005 Intel Corp
+ * 	Rohit Seth <rohit.seth@intel.com>
+ * 	Suresh Siddha <suresh.b.siddha@intel.com>
+ * 	Gordon Jin <gordon.jin@intel.com>
+ *	Ashok Raj  <ashok.raj@intel.com>
  *
  * 01/05/16 Rohit Seth <rohit.seth@intel.com>	Moved SMP booting functions from smp.c to here.
  * 01/04/27 David Mosberger <davidm@hpl.hp.com>	Added ITC synching code.
@@ -10,6 +15,11 @@
  *						smp_boot_cpus()/smp_commence() is replaced by
  *						smp_prepare_cpus()/__cpu_up()/smp_cpus_done().
  * 04/06/21 Ashok Raj		<ashok.raj@intel.com> Added CPU Hotplug Support
+ * 04/12/26 Jin Gordon <gordon.jin@intel.com>
+ * 04/12/26 Rohit Seth <rohit.seth@intel.com>
+ *						Add multi-threading and multi-core detection
+ * 05/01/30 Suresh Siddha <suresh.b.siddha@intel.com>
+ *						Setup cpu_sibling_map and cpu_core_map
  */
 #include <linux/config.h>
 
@@ -122,6 +132,11 @@ EXPORT_SYMBOL(cpu_online_map);
 cpumask_t cpu_possible_map;
 EXPORT_SYMBOL(cpu_possible_map);
 
+cpumask_t cpu_core_map[NR_CPUS] __cacheline_aligned;
+cpumask_t cpu_sibling_map[NR_CPUS] __cacheline_aligned;
+int smp_num_siblings = 1;
+int smp_num_cpucores = 1;
+
 /* which logical CPU number maps to which CPU (physical APIC ID) */
 volatile int ia64_cpu_to_sapicid[NR_CPUS];
 EXPORT_SYMBOL(ia64_cpu_to_sapicid);
@@ -156,7 +171,8 @@ sync_master (void *arg)
 	local_irq_save(flags);
 	{
 		for (i = 0; i < NUM_ROUNDS*NUM_ITERS; ++i) {
-			while (!go[MASTER]);
+			while (!go[MASTER])
+				cpu_relax();
 			go[MASTER] = 0;
 			go[SLAVE] = ia64_get_itc();
 		}
@@ -179,7 +195,8 @@ get_delta (long *rt, long *master)
 	for (i = 0; i < NUM_ITERS; ++i) {
 		t0 = ia64_get_itc();
 		go[MASTER] = 1;
-		while (!(tm = go[SLAVE]));
+		while (!(tm = go[SLAVE]))
+			cpu_relax();
 		go[SLAVE] = 0;
 		t1 = ia64_get_itc();
 
@@ -258,7 +275,8 @@ ia64_sync_itc (unsigned int master)
 		return;
 	}
 
-	while (go[MASTER]);	/* wait for master to be ready */
+	while (go[MASTER])
+		cpu_relax();	/* wait for master to be ready */
 
 	spin_lock_irqsave(&itc_sync_lock, flags);
 	{
@@ -595,7 +613,68 @@ void __devinit smp_prepare_boot_cpu(void)
 	cpu_set(smp_processor_id(), cpu_callin_map);
 }
 
+/*
+ * mt_info[] is a temporary store for all info returned by
+ * PAL_LOGICAL_TO_PHYSICAL, to be copied into cpuinfo_ia64 when the
+ * specific cpu comes.
+ */
+static struct {
+	__u32   socket_id;
+	__u16   core_id;
+	__u16   thread_id;
+	__u16   proc_fixed_addr;
+	__u8    valid;
+}mt_info[NR_CPUS] __devinit;
+
 #ifdef CONFIG_HOTPLUG_CPU
+static inline void
+remove_from_mtinfo(int cpu)
+{
+	int i;
+
+	for_each_cpu(i)
+		if (mt_info[i].valid &&  mt_info[i].socket_id ==
+		    				cpu_data(cpu)->socket_id)
+			mt_info[i].valid = 0;
+}
+
+static inline void
+clear_cpu_sibling_map(int cpu)
+{
+	int i;
+
+	for_each_cpu_mask(i, cpu_sibling_map[cpu])
+		cpu_clear(cpu, cpu_sibling_map[i]);
+	for_each_cpu_mask(i, cpu_core_map[cpu])
+		cpu_clear(cpu, cpu_core_map[i]);
+
+	cpu_sibling_map[cpu] = cpu_core_map[cpu] = CPU_MASK_NONE;
+}
+
+static void
+remove_siblinginfo(int cpu)
+{
+	int last = 0;
+
+	if (cpu_data(cpu)->threads_per_core == 1 &&
+	    cpu_data(cpu)->cores_per_socket == 1) {
+		cpu_clear(cpu, cpu_core_map[cpu]);
+		cpu_clear(cpu, cpu_sibling_map[cpu]);
+		return;
+	}
+
+	last = (cpus_weight(cpu_core_map[cpu]) == 1 ? 1 : 0);
+
+	/* remove it from all sibling map's */
+	clear_cpu_sibling_map(cpu);
+
+	/* if this cpu is the last in the core group, remove all its info 
+	 * from mt_info structure
+	 */
+	if (last)
+		remove_from_mtinfo(cpu);
+}
+
 extern void fixup_irqs(void);
 /* must be called with cpucontrol mutex held */
 int __cpu_disable(void)
@@ -608,6 +687,7 @@ int __cpu_disable(void)
 	if (cpu == 0)
 		return -EBUSY;
 
+	remove_siblinginfo(cpu);
 	fixup_irqs();
 	local_flush_tlb_all();
 	cpu_clear(cpu, cpu_callin_map);
@@ -660,6 +740,23 @@ smp_cpus_done (unsigned int dummy)
 	       (int)num_online_cpus(), bogosum/(500000/HZ), (bogosum/(5000/HZ))%100);
 }
 
+static inline void __devinit
+set_cpu_sibling_map(int cpu)
+{
+	int i;
+
+	for_each_online_cpu(i) {
+		if ((cpu_data(cpu)->socket_id == cpu_data(i)->socket_id)) {
+			cpu_set(i, cpu_core_map[cpu]);
+			cpu_set(cpu, cpu_core_map[i]);
+			if (cpu_data(cpu)->core_id == cpu_data(i)->core_id) {
+				cpu_set(i, cpu_sibling_map[cpu]);
+				cpu_set(cpu, cpu_sibling_map[i]);
+			}
+		}
+	}
+}
+
 int __devinit
 __cpu_up (unsigned int cpu)
 {
@@ -682,6 +779,15 @@ __cpu_up (unsigned int cpu)
 	if (ret < 0)
 		return ret;
 
+	if (cpu_data(cpu)->threads_per_core == 1 &&
+	    cpu_data(cpu)->cores_per_socket == 1) {
+		cpu_set(cpu, cpu_sibling_map[cpu]);
+		cpu_set(cpu, cpu_core_map[cpu]);
+		return 0;
+	}
+
+	set_cpu_sibling_map(cpu);
+
 	return 0;
 }
 
@@ -709,3 +815,106 @@ init_smp_config(void)
 		       ia64_sal_strerror(sal_ret));
 }
 
+static inline int __devinit
+check_for_mtinfo_index(void)
+{
+	int i;
+	
+	for_each_cpu(i)
+		if (!mt_info[i].valid)
+			return i;
+
+	return -1;
+}
+
+/*
+ * Search the mt_info to find out if this socket's cid/tid information is
+ * cached or not. If the socket exists, fill in the core_id and thread_id 
+ * in cpuinfo
+ */
+static int __devinit
+check_for_new_socket(__u16 logical_address, struct cpuinfo_ia64 *c)
+{
+	int i;
+	__u32 sid = c->socket_id;
+
+	for_each_cpu(i) {
+		if (mt_info[i].valid && mt_info[i].proc_fixed_addr == logical_address
+		    && mt_info[i].socket_id == sid) {
+			c->core_id = mt_info[i].core_id;
+			c->thread_id = mt_info[i].thread_id;
+			return 1; /* not a new socket */
+		}
+	}
+	return 0;
+}
+
+/*
+ * identify_siblings(cpu) gets called from identify_cpu. This populates the 
+ * information related to logical execution units in per_cpu_data structure.
+ */
+void __devinit
+identify_siblings(struct cpuinfo_ia64 *c)
+{
+	s64 status;
+	u16 pltid;
+	u64 proc_fixed_addr;
+	int count, i;
+	pal_logical_to_physical_t info;
+
+	if (smp_num_cpucores == 1 && smp_num_siblings == 1)
+		return;
+
+	if ((status = ia64_pal_logical_to_phys(0, &info)) != PAL_STATUS_SUCCESS) {
+		printk(KERN_ERR "ia64_pal_logical_to_phys failed with %ld\n",
+		       status);
+		return;
+	}
+	if ((status = ia64_sal_physical_id_info(&pltid)) != PAL_STATUS_SUCCESS) {
+		printk(KERN_ERR "ia64_sal_pltid failed with %ld\n", status);
+		return;
+	}
+	if ((status = ia64_pal_fixed_addr(&proc_fixed_addr)) != PAL_STATUS_SUCCESS) {
+		printk(KERN_ERR "ia64_pal_fixed_addr failed with %ld\n", status);
+		return;
+	}
+
+	c->socket_id =  (pltid << 8) | info.overview_ppid;
+	c->cores_per_socket = info.overview_cpp;
+	c->threads_per_core = info.overview_tpc;
+	count = c->num_log = info.overview_num_log;
+
+	/* If the thread and core id information is already cached, then
+	 * we will simply update cpu_info and return. Otherwise, we will
+	 * do the PAL calls and cache core and thread id's of all the siblings.
+	 */
+	if (check_for_new_socket(proc_fixed_addr, c))
+		return;
+
+	for (i = 0; i < count; i++) {
+		int index;
+
+		if (i && (status = ia64_pal_logical_to_phys(i, &info))
+			  != PAL_STATUS_SUCCESS) {
+                	printk(KERN_ERR "ia64_pal_logical_to_phys failed"
+					" with %ld\n", status);
+                	return;
+		}
+		if (info.log2_la == proc_fixed_addr) {
+			c->core_id = info.log1_cid;
+			c->thread_id = info.log1_tid;
+		}
+
+		index = check_for_mtinfo_index();
+		/* We will not do the mt_info caching optimization in this case.
+		 */
+		if (index < 0)
+			continue;
+
+		mt_info[index].valid = 1;
+		mt_info[index].socket_id = c->socket_id;
+		mt_info[index].core_id = info.log1_cid;
+		mt_info[index].thread_id = info.log1_tid;
+		mt_info[index].proc_fixed_addr = info.log2_la;
+	}
+}

arch/ia64/kernel/unwind.c

@@ -1943,23 +1943,30 @@ EXPORT_SYMBOL(unw_unwind);
 int
 unw_unwind_to_user (struct unw_frame_info *info)
 {
-	unsigned long ip, sp;
+	unsigned long ip, sp, pr = 0;
 
 	while (unw_unwind(info) >= 0) {
-		if (unw_get_rp(info, &ip) < 0) {
-			unw_get_ip(info, &ip);
-			UNW_DPRINT(0, "unwind.%s: failed to read return pointer (ip=0x%lx)\n",
-				   __FUNCTION__, ip);
-			return -1;
-		}
 		unw_get_sp(info, &sp);
-		if (sp >= (unsigned long)info->task + IA64_STK_OFFSET)
+		if ((long)((unsigned long)info->task + IA64_STK_OFFSET - sp)
+		    < IA64_PT_REGS_SIZE) {
+			UNW_DPRINT(0, "unwind.%s: ran off the top of the kernel stack\n",
+				   __FUNCTION__);
 			break;
-		if (ip < FIXADDR_USER_END)
+		}
+		if (unw_is_intr_frame(info) &&
+		    (pr & (1UL << PRED_USER_STACK)))
 			return 0;
+		if (unw_get_pr (info, &pr) < 0) {
+			unw_get_rp(info, &ip);
+			UNW_DPRINT(0, "unwind.%s: failed to read "
+				   "predicate register (ip=0x%lx)\n",
+				__FUNCTION__, ip);
+			return -1;
+		}
 	}
 	unw_get_ip(info, &ip);
-	UNW_DPRINT(0, "unwind.%s: failed to unwind to user-level (ip=0x%lx)\n", __FUNCTION__, ip);
+	UNW_DPRINT(0, "unwind.%s: failed to unwind to user-level (ip=0x%lx)\n",
+		   __FUNCTION__, ip);
 	return -1;
 }
 EXPORT_SYMBOL(unw_unwind_to_user);

arch/ia64/lib/memcpy_mck.S

@@ -300,7 +300,7 @@ EK(.ex_handler,	(p[D])	st8 [dst1] = t15, 4*8)
 	add	src_pre_mem=0,src0	// prefetch src pointer
 	add	dst_pre_mem=0,dst0	// prefetch dest pointer
 	and	src0=-8,src0		// 1st src pointer
-(p7)	mov	ar.lc = r21
+(p7)	mov	ar.lc = cnt
 (p8)	mov	ar.lc = r0
 	;;
 	TEXT_ALIGN(32)

arch/ia64/mm/contig.c

@@ -61,7 +61,8 @@ show_mem (void)
 	printk("%d reserved pages\n", reserved);
 	printk("%d pages shared\n", shared);
 	printk("%d pages swap cached\n", cached);
-	printk("%ld pages in page table cache\n", pgtable_cache_size);
+	printk("%ld pages in page table cache\n",
+		pgtable_quicklist_total_size());
 }
 
 /* physical address where the bootmem map is located */

arch/ia64/mm/discontig.c

@@ -582,7 +582,8 @@ void show_mem(void)
 	printk("%d reserved pages\n", total_reserved);
 	printk("%d pages shared\n", total_shared);
 	printk("%d pages swap cached\n", total_cached);
-	printk("Total of %ld pages in page table cache\n", pgtable_cache_size);
+	printk("Total of %ld pages in page table cache\n",
+		pgtable_quicklist_total_size());
 	printk("%d free buffer pages\n", nr_free_buffer_pages());
 }
 

arch/ia64/mm/fault.c

@@ -209,10 +209,13 @@ ia64_do_page_fault (unsigned long address, unsigned long isr, struct pt_regs *re
 	}
 
   no_context:
-	if (isr & IA64_ISR_SP) {
+	if ((isr & IA64_ISR_SP)
+	    || ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
+	{
 		/*
-		 * This fault was due to a speculative load set the "ed" bit in the psr to
-		 * ensure forward progress (target register will get a NaT).
+		 * This fault was due to a speculative load or lfetch.fault, set the "ed"
+		 * bit in the psr to ensure forward progress.  (Target register will get a
+		 * NaT for ld.s, lfetch will be canceled.)
 		 */
 		ia64_psr(regs)->ed = 1;
 		return;

arch/ia64/mm/init.c

@@ -39,6 +39,9 @@
 
 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 
+DEFINE_PER_CPU(unsigned long *, __pgtable_quicklist);
+DEFINE_PER_CPU(long, __pgtable_quicklist_size);
+
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
@@ -50,27 +53,53 @@ struct page *vmem_map;
 EXPORT_SYMBOL(vmem_map);
 #endif
 
-static int pgt_cache_water[2] = { 25, 50 };
-
-struct page *zero_page_memmap_ptr;		/* map entry for zero page */
+struct page *zero_page_memmap_ptr;	/* map entry for zero page */
 EXPORT_SYMBOL(zero_page_memmap_ptr);
 
+#define MIN_PGT_PAGES			25UL
+#define MAX_PGT_FREES_PER_PASS		16L
+#define PGT_FRACTION_OF_NODE_MEM	16
+
+static inline long
+max_pgt_pages(void)
+{
+	u64 node_free_pages, max_pgt_pages;
+
+#ifndef	CONFIG_NUMA
+	node_free_pages = nr_free_pages();
+#else
+	node_free_pages = nr_free_pages_pgdat(NODE_DATA(numa_node_id()));
+#endif
+	max_pgt_pages = node_free_pages / PGT_FRACTION_OF_NODE_MEM;
+	max_pgt_pages = max(max_pgt_pages, MIN_PGT_PAGES);
+	return max_pgt_pages;
+}
+
+static inline long
+min_pages_to_free(void)
+{
+	long pages_to_free;
+
+	pages_to_free = pgtable_quicklist_size - max_pgt_pages();
+	pages_to_free = min(pages_to_free, MAX_PGT_FREES_PER_PASS);
+	return pages_to_free;
+}
+
 void
-check_pgt_cache (void)
+check_pgt_cache(void)
 {
-	int low, high;
+	long pages_to_free;
 
-	low = pgt_cache_water[0];
-	high = pgt_cache_water[1];
+	if (unlikely(pgtable_quicklist_size <= MIN_PGT_PAGES))
+		return;
 
 	preempt_disable();
-	if (pgtable_cache_size > (u64) high) {
-		do {
-			if (pgd_quicklist)
-				free_page((unsigned long)pgd_alloc_one_fast(NULL));
-			if (pmd_quicklist)
-				free_page((unsigned long)pmd_alloc_one_fast(NULL, 0));
-		} while (pgtable_cache_size > (u64) low);
+	while (unlikely((pages_to_free = min_pages_to_free()) > 0)) {
+		while (pages_to_free--) {
+			free_page((unsigned long)pgtable_quicklist_alloc());
+		}
+		preempt_enable();
+		preempt_disable();
 	}
 	preempt_enable();
 }
@@ -523,11 +552,14 @@ void
 mem_init (void)
 {
 	long reserved_pages, codesize, datasize, initsize;
-	unsigned long num_pgt_pages;
 	pg_data_t *pgdat;
 	int i;
 	static struct kcore_list kcore_mem, kcore_vmem, kcore_kernel;
 
+	BUG_ON(PTRS_PER_PGD * sizeof(pgd_t) != PAGE_SIZE);
+	BUG_ON(PTRS_PER_PMD * sizeof(pmd_t) != PAGE_SIZE);
+	BUG_ON(PTRS_PER_PTE * sizeof(pte_t) != PAGE_SIZE);
+
 #ifdef CONFIG_PCI
 	/*
 	 * This needs to be called _after_ the command line has been parsed but _before_
@@ -564,18 +596,6 @@ mem_init (void)
 	       num_physpages << (PAGE_SHIFT - 10), codesize >> 10,
 	       reserved_pages << (PAGE_SHIFT - 10), datasize >> 10, initsize >> 10);
 
-	/*
-	 * Allow for enough (cached) page table pages so that we can map the entire memory
-	 * at least once.  Each task also needs a couple of page tables pages, so add in a
-	 * fudge factor for that (don't use "threads-max" here; that would be wrong!).
-	 * Don't allow the cache to be more than 10% of total memory, though.
-	 */
-#	define NUM_TASKS	500	/* typical number of tasks */
-	num_pgt_pages = nr_free_pages() / PTRS_PER_PGD + NUM_TASKS;
-	if (num_pgt_pages > nr_free_pages() / 10)
-		num_pgt_pages = nr_free_pages() / 10;
-	if (num_pgt_pages > (u64) pgt_cache_water[1])
-		pgt_cache_water[1] = num_pgt_pages;
 
 	/*
 	 * For fsyscall entrpoints with no light-weight handler, use the ordinary

arch/ia64/sn/include/pci/pcibr_provider.h

@@ -123,9 +123,11 @@ pcibr_lock(struct pcibus_info *pcibus_info)
 }
 #define pcibr_unlock(pcibus_info, flag)  spin_unlock_irqrestore(&pcibus_info->pbi_lock, flag)
 
+extern int  pcibr_init_provider(void);
 extern void *pcibr_bus_fixup(struct pcibus_bussoft *);
-extern uint64_t pcibr_dma_map(struct pcidev_info *, unsigned long, size_t, unsigned int);
-extern void pcibr_dma_unmap(struct pcidev_info *, dma_addr_t, int);
+extern dma_addr_t pcibr_dma_map(struct pci_dev *, unsigned long, size_t);
+extern dma_addr_t pcibr_dma_map_consistent(struct pci_dev *, unsigned long, size_t);
+extern void pcibr_dma_unmap(struct pci_dev *, dma_addr_t, int);
 
 /*
  * prototypes for the bridge asic register access routines in pcibr_reg.c

arch/ia64/sn/kernel/Makefile

@@ -10,3 +10,4 @@
 obj-y				+= setup.o bte.o bte_error.o irq.o mca.o idle.o \
 				   huberror.o io_init.o iomv.o klconflib.o sn2/
 obj-$(CONFIG_IA64_GENERIC)      += machvec.o
+obj-$(CONFIG_SGI_TIOCX)		+= tiocx.o

arch/ia64/sn/kernel/bte.c

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  */
 
 #include <linux/config.h>
@@ -170,10 +170,6 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 	/* Initialize the notification to a known value. */
 	*bte->most_rcnt_na = BTE_WORD_BUSY;
 
-	/* Set the status reg busy bit and transfer length */
-	BTE_PRINTKV(("IBLS = 0x%lx\n", IBLS_BUSY | transfer_size));
-	BTE_LNSTAT_STORE(bte, IBLS_BUSY | transfer_size);
-
 	/* Set the source and destination registers */
 	BTE_PRINTKV(("IBSA = 0x%lx)\n", (TO_PHYS(src))));
 	BTE_SRC_STORE(bte, TO_PHYS(src));
@@ -188,7 +184,7 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 
 	/* Initiate the transfer */
 	BTE_PRINTK(("IBCT = 0x%lx)\n", BTE_VALID_MODE(mode)));
-	BTE_CTRL_STORE(bte, BTE_VALID_MODE(mode));
+	BTE_START_TRANSFER(bte, transfer_size, BTE_VALID_MODE(mode));
 
 	itc_end = ia64_get_itc() + (40000000 * local_cpu_data->cyc_per_usec);
 
@@ -429,10 +425,16 @@ void bte_init_node(nodepda_t * mynodepda, cnodeid_t cnode)
 	mynodepda->bte_recovery_timer.data = (unsigned long)mynodepda;
 
 	for (i = 0; i < BTES_PER_NODE; i++) {
+		u64 *base_addr;
+
 		/* Which link status register should we use? */
-		unsigned long link_status = (i == 0 ? IIO_IBLS0 : IIO_IBLS1);
-		mynodepda->bte_if[i].bte_base_addr = (u64 *)
-		    REMOTE_HUB_ADDR(cnodeid_to_nasid(cnode), link_status);
+		base_addr = (u64 *)
+		    REMOTE_HUB_ADDR(cnodeid_to_nasid(cnode), BTE_BASE_ADDR(i));
+		mynodepda->bte_if[i].bte_base_addr = base_addr;
+		mynodepda->bte_if[i].bte_source_addr = BTE_SOURCE_ADDR(base_addr);
+		mynodepda->bte_if[i].bte_destination_addr = BTE_DEST_ADDR(base_addr);
+		mynodepda->bte_if[i].bte_control_addr = BTE_CTRL_ADDR(base_addr);
+		mynodepda->bte_if[i].bte_notify_addr = BTE_NOTIF_ADDR(base_addr);
 
 		/*
 		 * Initialize the notification and spinlock

arch/ia64/sn/kernel/bte_error.c

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  */
 
 #include <linux/types.h>
@@ -33,48 +33,28 @@ void bte_error_handler(unsigned long);
  * Wait until all BTE related CRBs are completed
  * and then reset the interfaces.
  */
-void bte_error_handler(unsigned long _nodepda)
+void shub1_bte_error_handler(unsigned long _nodepda)
 {
 	struct nodepda_s *err_nodepda = (struct nodepda_s *)_nodepda;
-	spinlock_t *recovery_lock = &err_nodepda->bte_recovery_lock;
 	struct timer_list *recovery_timer = &err_nodepda->bte_recovery_timer;
 	nasid_t nasid;
 	int i;
 	int valid_crbs;
-	unsigned long irq_flags;
-	volatile u64 *notify;
-	bte_result_t bh_error;
 	ii_imem_u_t imem;	/* II IMEM Register */
 	ii_icrb0_d_u_t icrbd;	/* II CRB Register D */
 	ii_ibcr_u_t ibcr;
 	ii_icmr_u_t icmr;
 	ii_ieclr_u_t ieclr;
 
-	BTE_PRINTK(("bte_error_handler(%p) - %d\n", err_nodepda,
+	BTE_PRINTK(("shub1_bte_error_handler(%p) - %d\n", err_nodepda,
 		    smp_processor_id()));
 
-	spin_lock_irqsave(recovery_lock, irq_flags);
-
 	if ((err_nodepda->bte_if[0].bh_error == BTE_SUCCESS) &&
 	    (err_nodepda->bte_if[1].bh_error == BTE_SUCCESS)) {
 		BTE_PRINTK(("eh:%p:%d Nothing to do.\n", err_nodepda,
 			    smp_processor_id()));
-		spin_unlock_irqrestore(recovery_lock, irq_flags);
 		return;
 	}
-	/*
-	 * Lock all interfaces on this node to prevent new transfers
-	 * from being queued.
-	 */
-	for (i = 0; i < BTES_PER_NODE; i++) {
-		if (err_nodepda->bte_if[i].cleanup_active) {
-			continue;
-		}
-		spin_lock(&err_nodepda->bte_if[i].spinlock);
-		BTE_PRINTK(("eh:%p:%d locked %d\n", err_nodepda,
-			    smp_processor_id(), i));
-		err_nodepda->bte_if[i].cleanup_active = 1;
-	}
 
 	/* Determine information about our hub */
 	nasid = cnodeid_to_nasid(err_nodepda->bte_if[0].bte_cnode);
@@ -101,7 +81,6 @@ void bte_error_handler(unsigned long _nodepda)
 		mod_timer(recovery_timer, HZ * 5);
 		BTE_PRINTK(("eh:%p:%d Marked Giving up\n", err_nodepda,
 			    smp_processor_id()));
-		spin_unlock_irqrestore(recovery_lock, irq_flags);
 		return;
 	}
 	if (icmr.ii_icmr_fld_s.i_crb_vld != 0) {
@@ -120,8 +99,6 @@ void bte_error_handler(unsigned long _nodepda)
 				BTE_PRINTK(("eh:%p:%d Valid %d, Giving up\n",
 					    err_nodepda, smp_processor_id(),
 					    i));
-				spin_unlock_irqrestore(recovery_lock,
-						       irq_flags);
 				return;
 			}
 		}
@@ -146,6 +123,51 @@ void bte_error_handler(unsigned long _nodepda)
 	ibcr.ii_ibcr_fld_s.i_soft_reset = 1;
 	REMOTE_HUB_S(nasid, IIO_IBCR, ibcr.ii_ibcr_regval);
 
+	del_timer(recovery_timer);
+}
+
+/*
+ * Wait until all BTE related CRBs are completed
+ * and then reset the interfaces.
+ */
+void bte_error_handler(unsigned long _nodepda)
+{
+	struct nodepda_s *err_nodepda = (struct nodepda_s *)_nodepda;
+	spinlock_t *recovery_lock = &err_nodepda->bte_recovery_lock;
+	int i;
+	nasid_t nasid;
+	unsigned long irq_flags;
+	volatile u64 *notify;
+	bte_result_t bh_error;
+
+	BTE_PRINTK(("bte_error_handler(%p) - %d\n", err_nodepda,
+		    smp_processor_id()));
+
+	spin_lock_irqsave(recovery_lock, irq_flags);
+
+	/*
+	 * Lock all interfaces on this node to prevent new transfers
+	 * from being queued.
+	 */
+	for (i = 0; i < BTES_PER_NODE; i++) {
+		if (err_nodepda->bte_if[i].cleanup_active) {
+			continue;
+		}
+		spin_lock(&err_nodepda->bte_if[i].spinlock);
+		BTE_PRINTK(("eh:%p:%d locked %d\n", err_nodepda,
+			    smp_processor_id(), i));
+		err_nodepda->bte_if[i].cleanup_active = 1;
+	}
+
+	if (is_shub1()) {
+		shub1_bte_error_handler(_nodepda);
+	} else {
+		nasid = cnodeid_to_nasid(err_nodepda->bte_if[0].bte_cnode);
+
+		if (ia64_sn_bte_recovery(nasid))
+			panic("bte_error_handler(): Fatal BTE Error");
+	}
+
 	for (i = 0; i < BTES_PER_NODE; i++) {
 		bh_error = err_nodepda->bte_if[i].bh_error;
 		if (bh_error != BTE_SUCCESS) {
@@ -165,8 +187,6 @@ void bte_error_handler(unsigned long _nodepda)
 		spin_unlock(&err_nodepda->bte_if[i].spinlock);
 	}
 
-	del_timer(recovery_timer);
-
 	spin_unlock_irqrestore(recovery_lock, irq_flags);
 }
 

arch/ia64/sn/kernel/huberror.c

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 1992 - 1997, 2000,2002-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 1992 - 1997, 2000,2002-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #include <linux/types.h>
@@ -38,8 +38,11 @@ static irqreturn_t hub_eint_handler(int irq, void *arg, struct pt_regs *ep)
 	if ((int)ret_stuff.v0)
 		panic("hubii_eint_handler(): Fatal TIO Error");
 
-	if (!(nasid & 1)) /* Not a TIO, handle CRB errors */
-		(void)hubiio_crb_error_handler(hubdev_info);
+	if (is_shub1()) {
+		if (!(nasid & 1)) /* Not a TIO, handle CRB errors */
+			(void)hubiio_crb_error_handler(hubdev_info);
+	} else 
+		bte_error_handler((unsigned long)NODEPDA(nasid_to_cnodeid(nasid)));
 
 	return IRQ_HANDLED;
 }

arch/ia64/sn/kernel/io_init.c

@@ -11,14 +11,15 @@
 #include <asm/sn/types.h>
 #include <asm/sn/sn_sal.h>
 #include <asm/sn/addrs.h>
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include "pci/pcibr_provider.h"
 #include "xtalk/xwidgetdev.h"
 #include <asm/sn/geo.h>
 #include "xtalk/hubdev.h"
 #include <asm/sn/io.h>
 #include <asm/sn/simulator.h>
+#include <asm/sn/tioca_provider.h>
 
 char master_baseio_wid;
 nasid_t master_nasid = INVALID_NASID;	/* Partition Master */
@@ -34,6 +35,37 @@ struct brick {
 
 int sn_ioif_inited = 0;		/* SN I/O infrastructure initialized? */
 
+struct sn_pcibus_provider *sn_pci_provider[PCIIO_ASIC_MAX_TYPES];	/* indexed by asic type */
+
+/*
+ * Hooks and struct for unsupported pci providers
+ */
+
+static dma_addr_t
+sn_default_pci_map(struct pci_dev *pdev, unsigned long paddr, size_t size)
+{
+	return 0;
+}
+
+static void
+sn_default_pci_unmap(struct pci_dev *pdev, dma_addr_t addr, int direction)
+{
+	return;
+}
+
+static void *
+sn_default_pci_bus_fixup(struct pcibus_bussoft *soft)
+{
+	return NULL;
+}
+
+static struct sn_pcibus_provider sn_pci_default_provider = {
+	.dma_map = sn_default_pci_map,
+	.dma_map_consistent = sn_default_pci_map,
+	.dma_unmap = sn_default_pci_unmap,
+	.bus_fixup = sn_default_pci_bus_fixup,
+};
+
 /*
  * Retrieve the DMA Flush List given nasid.  This list is needed 
  * to implement the WAR - Flush DMA data on PIO Reads.
@@ -201,6 +233,7 @@ static void sn_pci_fixup_slot(struct pci_dev *dev)
 	struct sn_irq_info *sn_irq_info;
 	struct pci_dev *host_pci_dev;
 	int status = 0;
+	struct pcibus_bussoft *bs;
 
 	dev->sysdata = kmalloc(sizeof(struct pcidev_info), GFP_KERNEL);
 	if (SN_PCIDEV_INFO(dev) <= 0)
@@ -241,6 +274,7 @@ static void sn_pci_fixup_slot(struct pci_dev *dev)
 	}
 
 	/* set up host bus linkages */
+	bs = SN_PCIBUS_BUSSOFT(dev->bus);
 	host_pci_dev =
 	    pci_find_slot(SN_PCIDEV_INFO(dev)->pdi_slot_host_handle >> 32,
 			  SN_PCIDEV_INFO(dev)->
@@ -248,10 +282,16 @@ static void sn_pci_fixup_slot(struct pci_dev *dev)
 	SN_PCIDEV_INFO(dev)->pdi_host_pcidev_info =
 	    SN_PCIDEV_INFO(host_pci_dev);
 	SN_PCIDEV_INFO(dev)->pdi_linux_pcidev = dev;
-	SN_PCIDEV_INFO(dev)->pdi_pcibus_info = SN_PCIBUS_BUSSOFT(dev->bus);
+	SN_PCIDEV_INFO(dev)->pdi_pcibus_info = bs;
+
+	if (bs && bs->bs_asic_type < PCIIO_ASIC_MAX_TYPES) {
+		SN_PCIDEV_BUSPROVIDER(dev) = sn_pci_provider[bs->bs_asic_type];
+	} else {
+		SN_PCIDEV_BUSPROVIDER(dev) = &sn_pci_default_provider;
+	}
 
 	/* Only set up IRQ stuff if this device has a host bus context */
-	if (SN_PCIDEV_BUSSOFT(dev) && sn_irq_info->irq_irq) {
+	if (bs && sn_irq_info->irq_irq) {
 		SN_PCIDEV_INFO(dev)->pdi_sn_irq_info = sn_irq_info;
 		dev->irq = SN_PCIDEV_INFO(dev)->pdi_sn_irq_info->irq_irq;
 		sn_irq_fixup(dev, sn_irq_info);
@@ -271,6 +311,7 @@ static void sn_pci_controller_fixup(int segment, int busnum)
 	struct pcibus_bussoft *prom_bussoft_ptr;
 	struct hubdev_info *hubdev_info;
 	void *provider_soft;
+	struct sn_pcibus_provider *provider;
 
 	status =
 	    sal_get_pcibus_info((u64) segment, (u64) busnum,
@@ -291,16 +332,22 @@ static void sn_pci_controller_fixup(int segment, int busnum)
 	/*
 	 * Per-provider fixup.  Copies the contents from prom to local
 	 * area and links SN_PCIBUS_BUSSOFT().
-	 *
-	 * Note:  Provider is responsible for ensuring that prom_bussoft_ptr
-	 * represents an asic-type that it can handle.
 	 */
 
-	if (prom_bussoft_ptr->bs_asic_type == PCIIO_ASIC_TYPE_PPB) {
-		return;		/* no further fixup necessary */
+	if (prom_bussoft_ptr->bs_asic_type >= PCIIO_ASIC_MAX_TYPES) {
+		return;		/* unsupported asic type */
+	}
+
+	provider = sn_pci_provider[prom_bussoft_ptr->bs_asic_type];
+	if (provider == NULL) {
+		return;		/* no provider registerd for this asic */
+	}
+
+	provider_soft = NULL;
+	if (provider->bus_fixup) {
+		provider_soft = (*provider->bus_fixup) (prom_bussoft_ptr);
 	}
 
-	provider_soft = pcibr_bus_fixup(prom_bussoft_ptr);
 	if (provider_soft == NULL) {
 		return;		/* fixup failed or not applicable */
 	}
@@ -338,6 +385,17 @@ static int __init sn_pci_init(void)
 	if (!ia64_platform_is("sn2") || IS_RUNNING_ON_SIMULATOR())
 		return 0;
 
+	/*
+	 * prime sn_pci_provider[].  Individial provider init routines will
+	 * override their respective default entries.
+	 */
+
+	for (i = 0; i < PCIIO_ASIC_MAX_TYPES; i++)
+		sn_pci_provider[i] = &sn_pci_default_provider;
+
+	pcibr_init_provider();
+	tioca_init_provider();
+
 	/*
 	 * This is needed to avoid bounce limit checks in the blk layer
 	 */

arch/ia64/sn/kernel/irq.c

@@ -13,8 +13,8 @@
 #include <asm/sn/addrs.h>
 #include <asm/sn/arch.h>
 #include "xtalk/xwidgetdev.h"
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include "pci/pcibr_provider.h"
 #include <asm/sn/shub_mmr.h>
 #include <asm/sn/sn_sal.h>
@@ -82,20 +82,9 @@ static void sn_ack_irq(unsigned int irq)
 	nasid = get_nasid();
 	event_occurred =
 	    HUB_L((uint64_t *) GLOBAL_MMR_ADDR(nasid, SH_EVENT_OCCURRED));
-	if (event_occurred & SH_EVENT_OCCURRED_UART_INT_MASK) {
-		mask |= (1 << SH_EVENT_OCCURRED_UART_INT_SHFT);
-	}
-	if (event_occurred & SH_EVENT_OCCURRED_IPI_INT_MASK) {
-		mask |= (1 << SH_EVENT_OCCURRED_IPI_INT_SHFT);
-	}
-	if (event_occurred & SH_EVENT_OCCURRED_II_INT0_MASK) {
-		mask |= (1 << SH_EVENT_OCCURRED_II_INT0_SHFT);
-	}
-	if (event_occurred & SH_EVENT_OCCURRED_II_INT1_MASK) {
-		mask |= (1 << SH_EVENT_OCCURRED_II_INT1_SHFT);
-	}
+	mask = event_occurred & SH_ALL_INT_MASK;
 	HUB_S((uint64_t *) GLOBAL_MMR_ADDR(nasid, SH_EVENT_OCCURRED_ALIAS),
-	      mask);
+		 mask);
 	__set_bit(irq, (volatile void *)pda->sn_in_service_ivecs);
 
 	move_irq(irq);

arch/ia64/sn/kernel/setup.c

@@ -29,6 +29,7 @@
 #include <linux/sched.h>
 #include <linux/root_dev.h>
 #include <linux/nodemask.h>
+#include <linux/pm.h>
 
 #include <asm/io.h>
 #include <asm/sal.h>
@@ -353,6 +354,14 @@ void __init sn_setup(char **cmdline_p)
 	screen_info = sn_screen_info;
 
 	sn_timer_init();
+
+	/*
+	 * set pm_power_off to a SAL call to allow
+	 * sn machines to power off. The SAL call can be replaced
+	 * by an ACPI interface call when ACPI is fully implemented
+	 * for sn.
+	 */
+	pm_power_off = ia64_sn_power_down;
 }
 
 /**

arch/ia64/sn/kernel/sn2/sn_hwperf.c

@@ -28,6 +28,7 @@
 #include <linux/vmalloc.h>
 #include <linux/seq_file.h>
 #include <linux/miscdevice.h>
+#include <linux/utsname.h>
 #include <linux/cpumask.h>
 #include <linux/smp_lock.h>
 #include <linux/nodemask.h>
@@ -43,6 +44,7 @@
 #include <asm/sn/module.h>
 #include <asm/sn/geo.h>
 #include <asm/sn/sn2/sn_hwperf.h>
+#include <asm/sn/addrs.h>
 
 static void *sn_hwperf_salheap = NULL;
 static int sn_hwperf_obj_cnt = 0;
@@ -81,26 +83,45 @@ static int sn_hwperf_enum_objects(int *nobj, struct sn_hwperf_object_info **ret)
 	return e;
 }
 
+static int sn_hwperf_location_to_bpos(char *location,
+	int *rack, int *bay, int *slot, int *slab)
+{
+	char type;
+
+	/* first scan for an old style geoid string */
+	if (sscanf(location, "%03d%c%02d#%d",
+		rack, &type, bay, slab) == 4)
+		*slot = 0; 
+	else /* scan for a new bladed geoid string */
+	if (sscanf(location, "%03d%c%02d^%02d#%d",
+		rack, &type, bay, slot, slab) != 5)
+		return -1; 
+	/* success */
+	return 0;
+}
+
 static int sn_hwperf_geoid_to_cnode(char *location)
 {
 	int cnode;
 	geoid_t geoid;
 	moduleid_t module_id;
-	char type;
-	int rack, slot, slab;
-	int this_rack, this_slot, this_slab;
+	int rack, bay, slot, slab;
+	int this_rack, this_bay, this_slot, this_slab;
 
-	if (sscanf(location, "%03d%c%02d#%d", &rack, &type, &slot, &slab) != 4)
+	if (sn_hwperf_location_to_bpos(location, &rack, &bay, &slot, &slab))
 		return -1;
 
 	for (cnode = 0; cnode < numionodes; cnode++) {
 		geoid = cnodeid_get_geoid(cnode);
 		module_id = geo_module(geoid);
 		this_rack = MODULE_GET_RACK(module_id);
-		this_slot = MODULE_GET_BPOS(module_id);
+		this_bay = MODULE_GET_BPOS(module_id);
+		this_slot = geo_slot(geoid);
 		this_slab = geo_slab(geoid);
-		if (rack == this_rack && slot == this_slot && slab == this_slab)
+		if (rack == this_rack && bay == this_bay &&
+			slot == this_slot && slab == this_slab) {
 			break;
+		}
 	}
 
 	return cnode < numionodes ? cnode : -1;
@@ -153,11 +174,36 @@ static const char *sn_hwperf_get_slabname(struct sn_hwperf_object_info *obj,
 	return slabname;
 }
 
+static void print_pci_topology(struct seq_file *s,
+	struct sn_hwperf_object_info *obj, int *ordinal,
+	u64 rack, u64 bay, u64 slot, u64 slab)
+{
+	char *p1;
+	char *p2;
+	char *pg;
+
+	if (!(pg = (char *)get_zeroed_page(GFP_KERNEL)))
+		return; /* ignore */
+	if (ia64_sn_ioif_get_pci_topology(rack, bay, slot, slab,
+		__pa(pg), PAGE_SIZE) == SN_HWPERF_OP_OK) {
+		for (p1=pg; *p1 && p1 < pg + PAGE_SIZE;) {
+			if (!(p2 = strchr(p1, '\n')))
+				break;
+			*p2 = '\0';
+			seq_printf(s, "pcibus %d %s-%s\n",
+				*ordinal, obj->location, p1);
+			(*ordinal)++;
+			p1 = p2 + 1;
+		}
+	}
+	free_page((unsigned long)pg);
+}
+
 static int sn_topology_show(struct seq_file *s, void *d)
 {
 	int sz;
 	int pt;
-	int e;
+	int e = 0;
 	int i;
 	int j;
 	const char *slabname;
@@ -169,11 +215,44 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	struct sn_hwperf_object_info *p;
 	struct sn_hwperf_object_info *obj = d;	/* this object */
 	struct sn_hwperf_object_info *objs = s->private; /* all objects */
+	int rack, bay, slot, slab;
+	u8 shubtype;
+	u8 system_size;
+	u8 sharing_size;
+	u8 partid;
+	u8 coher;
+	u8 nasid_shift;
+	u8 region_size;
+	u16 nasid_mask;
+	int nasid_msb;
+	int pci_bus_ordinal = 0;
 
 	if (obj == objs) {
-		seq_printf(s, "# sn_topology version 1\n");
+		seq_printf(s, "# sn_topology version 2\n");
 		seq_printf(s, "# objtype ordinal location partition"
 			" [attribute value [, ...]]\n");
+
+		if (ia64_sn_get_sn_info(0,
+			&shubtype, &nasid_mask, &nasid_shift, &system_size,
+			&sharing_size, &partid, &coher, &region_size))
+			BUG();
+		for (nasid_msb=63; nasid_msb > 0; nasid_msb--) {
+			if (((u64)nasid_mask << nasid_shift) & (1ULL << nasid_msb))
+				break;
+		}
+		seq_printf(s, "partition %u %s local "
+			"shubtype %s, "
+			"nasid_mask 0x%016lx, "
+			"nasid_bits %d:%d, "
+			"system_size %d, "
+			"sharing_size %d, "
+			"coherency_domain %d, "
+			"region_size %d\n",
+
+			partid, system_utsname.nodename,
+			shubtype ? "shub2" : "shub1", 
+			(u64)nasid_mask << nasid_shift, nasid_msb, nasid_shift,
+			system_size, sharing_size, coher, region_size);
 	}
 
 	if (SN_HWPERF_FOREIGN(obj)) {
@@ -181,7 +260,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 		return 0;
 	}
 
-	for (i = 0; obj->name[i]; i++) {
+	for (i = 0; i < SN_HWPERF_MAXSTRING && obj->name[i]; i++) {
 		if (obj->name[i] == ' ')
 			obj->name[i] = '_';
 	}
@@ -221,6 +300,17 @@ static int sn_topology_show(struct seq_file *s, void *d)
 				seq_putc(s, '\n');
 			}
 		}
+
+		/*
+		 * PCI busses attached to this node, if any
+		 */
+		if (sn_hwperf_location_to_bpos(obj->location,
+			&rack, &bay, &slot, &slab)) {
+			/* export pci bus info */
+			print_pci_topology(s, obj, &pci_bus_ordinal,
+				rack, bay, slot, slab);
+
+		}
 	}
 
 	if (obj->ports) {
@@ -397,6 +487,9 @@ static int sn_hwperf_map_err(int hwperf_err)
 		break;
 
 	case SN_HWPERF_OP_BUSY:
+		e = -EBUSY;
+		break;
+
 	case SN_HWPERF_OP_RECONFIGURE:
 		e = -EAGAIN;
 		break;
@@ -549,6 +642,7 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 		r = sn_hwperf_op_cpu(&op_info);
 		if (r) {
 			r = sn_hwperf_map_err(r);
+			a.v0 = v0;
 			goto error;
 		}
 		break;

arch/ia64/sn/pci/Makefile

@@ -7,4 +7,4 @@
 #
 # Makefile for the sn pci general routines.
 
-obj-y := pci_dma.o pcibr/ 
+obj-y := pci_dma.o tioca_provider.o pcibr/ 

arch/ia64/sn/pci/pci_dma.c

@@ -12,9 +12,8 @@
 #include <linux/module.h>
 #include <asm/dma.h>
 #include <asm/sn/sn_sal.h>
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
-#include "pci/pcibr_provider.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 
 #define SG_ENT_VIRT_ADDRESS(sg)	(page_address((sg)->page) + (sg)->offset)
 #define SG_ENT_PHYS_ADDRESS(SG)	virt_to_phys(SG_ENT_VIRT_ADDRESS(SG))
@@ -79,7 +78,8 @@ void *sn_dma_alloc_coherent(struct device *dev, size_t size,
 {
 	void *cpuaddr;
 	unsigned long phys_addr;
-	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(to_pci_dev(dev));
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev);
 
 	BUG_ON(dev->bus != &pci_bus_type);
 
@@ -102,8 +102,7 @@ void *sn_dma_alloc_coherent(struct device *dev, size_t size,
 	 * resources.
 	 */
 
-	*dma_handle = pcibr_dma_map(pcidev_info, phys_addr, size,
-				    SN_PCIDMA_CONSISTENT);
+	*dma_handle = provider->dma_map_consistent(pdev, phys_addr, size);
 	if (!*dma_handle) {
 		printk(KERN_ERR "%s: out of ATEs\n", __FUNCTION__);
 		free_pages((unsigned long)cpuaddr, get_order(size));
@@ -127,11 +126,12 @@ EXPORT_SYMBOL(sn_dma_alloc_coherent);
 void sn_dma_free_coherent(struct device *dev, size_t size, void *cpu_addr,
 			  dma_addr_t dma_handle)
 {
-	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(to_pci_dev(dev));
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev);
 
 	BUG_ON(dev->bus != &pci_bus_type);
 
-	pcibr_dma_unmap(pcidev_info, dma_handle, 0);
+	provider->dma_unmap(pdev, dma_handle, 0);
 	free_pages((unsigned long)cpu_addr, get_order(size));
 }
 EXPORT_SYMBOL(sn_dma_free_coherent);
@@ -159,12 +159,13 @@ dma_addr_t sn_dma_map_single(struct device *dev, void *cpu_addr, size_t size,
 {
 	dma_addr_t dma_addr;
 	unsigned long phys_addr;
-	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(to_pci_dev(dev));
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev);
 
 	BUG_ON(dev->bus != &pci_bus_type);
 
 	phys_addr = __pa(cpu_addr);
-	dma_addr = pcibr_dma_map(pcidev_info, phys_addr, size, 0);
+	dma_addr = provider->dma_map(pdev, phys_addr, size);
 	if (!dma_addr) {
 		printk(KERN_ERR "%s: out of ATEs\n", __FUNCTION__);
 		return 0;
@@ -187,10 +188,12 @@ EXPORT_SYMBOL(sn_dma_map_single);
 void sn_dma_unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size,
 			 int direction)
 {
-	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(to_pci_dev(dev));
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev);
 
 	BUG_ON(dev->bus != &pci_bus_type);
-	pcibr_dma_unmap(pcidev_info, dma_addr, direction);
+
+	provider->dma_unmap(pdev, dma_addr, direction);
 }
 EXPORT_SYMBOL(sn_dma_unmap_single);
 
@@ -207,12 +210,13 @@ void sn_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
 		     int nhwentries, int direction)
 {
 	int i;
-	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(to_pci_dev(dev));
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev);
 
 	BUG_ON(dev->bus != &pci_bus_type);
 
 	for (i = 0; i < nhwentries; i++, sg++) {
-		pcibr_dma_unmap(pcidev_info, sg->dma_address, direction);
+		provider->dma_unmap(pdev, sg->dma_address, direction);
 		sg->dma_address = (dma_addr_t) NULL;
 		sg->dma_length = 0;
 	}
@@ -233,7 +237,8 @@ int sn_dma_map_sg(struct device *dev, struct scatterlist *sg, int nhwentries,
 {
 	unsigned long phys_addr;
 	struct scatterlist *saved_sg = sg;
-	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(to_pci_dev(dev));
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct sn_pcibus_provider *provider = SN_PCIDEV_BUSPROVIDER(pdev);
 	int i;
 
 	BUG_ON(dev->bus != &pci_bus_type);
@@ -243,8 +248,8 @@ int sn_dma_map_sg(struct device *dev, struct scatterlist *sg, int nhwentries,
 	 */
 	for (i = 0; i < nhwentries; i++, sg++) {
 		phys_addr = SG_ENT_PHYS_ADDRESS(sg);
-		sg->dma_address = pcibr_dma_map(pcidev_info, phys_addr,
-						sg->length, 0);
+		sg->dma_address = provider->dma_map(pdev,
+						    phys_addr, sg->length);
 
 		if (!sg->dma_address) {
 			printk(KERN_ERR "%s: out of ATEs\n", __FUNCTION__);

arch/ia64/sn/pci/pcibr/pcibr_ate.c

@@ -8,8 +8,8 @@
 
 #include <linux/types.h>
 #include <asm/sn/sn_sal.h>
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include "pci/pcibr_provider.h"
 
 int pcibr_invalidate_ate = 0;	/* by default don't invalidate ATE on free */

arch/ia64/sn/pci/pcibr/pcibr_dma.c

@@ -12,8 +12,8 @@
 #include <asm/sn/geo.h>
 #include "xtalk/xwidgetdev.h"
 #include "xtalk/hubdev.h"
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include "pci/tiocp.h"
 #include "pci/pic.h"
 #include "pci/pcibr_provider.h"
@@ -40,7 +40,7 @@ extern int sn_ioif_inited;
  *      we do not have to allocate entries in the PMU.
  */
 
-static uint64_t
+static dma_addr_t
 pcibr_dmamap_ate32(struct pcidev_info *info,
 		   uint64_t paddr, size_t req_size, uint64_t flags)
 {
@@ -109,7 +109,7 @@ pcibr_dmamap_ate32(struct pcidev_info *info,
 	return pci_addr;
 }
 
-static uint64_t
+static dma_addr_t
 pcibr_dmatrans_direct64(struct pcidev_info * info, uint64_t paddr,
 			uint64_t dma_attributes)
 {
@@ -141,7 +141,7 @@ pcibr_dmatrans_direct64(struct pcidev_info * info, uint64_t paddr,
 
 }
 
-static uint64_t
+static dma_addr_t
 pcibr_dmatrans_direct32(struct pcidev_info * info,
 			uint64_t paddr, size_t req_size, uint64_t flags)
 {
@@ -180,11 +180,11 @@ pcibr_dmatrans_direct32(struct pcidev_info * info,
  * DMA mappings for Direct 64 and 32 do not have any DMA maps.
  */
 void
-pcibr_dma_unmap(struct pcidev_info *pcidev_info, dma_addr_t dma_handle,
-		int direction)
+pcibr_dma_unmap(struct pci_dev *hwdev, dma_addr_t dma_handle, int direction)
 {
-	struct pcibus_info *pcibus_info = (struct pcibus_info *)pcidev_info->
-	    pdi_pcibus_info;
+	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(hwdev);
+	struct pcibus_info *pcibus_info =
+	    (struct pcibus_info *)pcidev_info->pdi_pcibus_info;
 
 	if (IS_PCI32_MAPPED(dma_handle)) {
 		int ate_index;
@@ -316,64 +316,63 @@ void sn_dma_flush(uint64_t addr)
 }
 
 /*
- * Wrapper DMA interface.  Called from pci_dma.c routines.
+ * DMA interfaces.  Called from pci_dma.c routines.
  */
 
-uint64_t
-pcibr_dma_map(struct pcidev_info * pcidev_info, unsigned long phys_addr,
-	      size_t size, unsigned int flags)
+dma_addr_t
+pcibr_dma_map(struct pci_dev * hwdev, unsigned long phys_addr, size_t size)
 {
 	dma_addr_t dma_handle;
-	struct pci_dev *pcidev = pcidev_info->pdi_linux_pcidev;
-
-	if (flags & SN_PCIDMA_CONSISTENT) {
-		/* sn_pci_alloc_consistent interfaces */
-		if (pcidev->dev.coherent_dma_mask == ~0UL) {
-			dma_handle =
-			    pcibr_dmatrans_direct64(pcidev_info, phys_addr,
-						    PCI64_ATTR_BAR);
-		} else {
-			dma_handle =
-			    (dma_addr_t) pcibr_dmamap_ate32(pcidev_info,
-							    phys_addr, size,
-							    PCI32_ATE_BAR);
-		}
-	} else {
-		/* map_sg/map_single interfaces */
+	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(hwdev);
 
-		/* SN cannot support DMA addresses smaller than 32 bits. */
-		if (pcidev->dma_mask < 0x7fffffff) {
-			return 0;
-		}
+	/* SN cannot support DMA addresses smaller than 32 bits. */
+	if (hwdev->dma_mask < 0x7fffffff) {
+		return 0;
+	}
 
-		if (pcidev->dma_mask == ~0UL) {
+	if (hwdev->dma_mask == ~0UL) {
+		/*
+		 * Handle the most common case: 64 bit cards.  This
+		 * call should always succeed.
+		 */
+
+		dma_handle = pcibr_dmatrans_direct64(pcidev_info, phys_addr,
+						     PCI64_ATTR_PREF);
+	} else {
+		/* Handle 32-63 bit cards via direct mapping */
+		dma_handle = pcibr_dmatrans_direct32(pcidev_info, phys_addr,
+						     size, 0);
+		if (!dma_handle) {
 			/*
-			 * Handle the most common case: 64 bit cards.  This
-			 * call should always succeed.
+			 * It is a 32 bit card and we cannot do direct mapping,
+			 * so we use an ATE.
 			 */
 
-			dma_handle =
-			    pcibr_dmatrans_direct64(pcidev_info, phys_addr,
-						    PCI64_ATTR_PREF);
-		} else {
-			/* Handle 32-63 bit cards via direct mapping */
-			dma_handle =
-			    pcibr_dmatrans_direct32(pcidev_info, phys_addr,
-						    size, 0);
-			if (!dma_handle) {
-				/*
-				 * It is a 32 bit card and we cannot do direct mapping,
-				 * so we use an ATE.
-				 */
-
-				dma_handle =
-				    pcibr_dmamap_ate32(pcidev_info, phys_addr,
-						       size, PCI32_ATE_PREF);
-			}
+			dma_handle = pcibr_dmamap_ate32(pcidev_info, phys_addr,
+							size, PCI32_ATE_PREF);
 		}
 	}
 
 	return dma_handle;
 }
 
+dma_addr_t
+pcibr_dma_map_consistent(struct pci_dev * hwdev, unsigned long phys_addr,
+			 size_t size)
+{
+	dma_addr_t dma_handle;
+	struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(hwdev);
+
+	if (hwdev->dev.coherent_dma_mask == ~0UL) {
+		dma_handle = pcibr_dmatrans_direct64(pcidev_info, phys_addr,
+					    PCI64_ATTR_BAR);
+	} else {
+		dma_handle = (dma_addr_t) pcibr_dmamap_ate32(pcidev_info,
+						    phys_addr, size,
+						    PCI32_ATE_BAR);
+	}
+
+	return dma_handle;
+}
+
 EXPORT_SYMBOL(sn_dma_flush);

arch/ia64/sn/pci/pcibr/pcibr_provider.c

@@ -13,8 +13,8 @@
 #include "xtalk/xwidgetdev.h"
 #include <asm/sn/geo.h>
 #include "xtalk/hubdev.h"
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include "pci/pcibr_provider.h"
 #include <asm/sn/addrs.h>
 
@@ -168,3 +168,23 @@ void pcibr_change_devices_irq(struct sn_irq_info *sn_irq_info)
 		pcibr_force_interrupt(sn_irq_info);
 	}
 }
+
+/*
+ * Provider entries for PIC/CP
+ */
+
+struct sn_pcibus_provider pcibr_provider = {
+	.dma_map = pcibr_dma_map,
+	.dma_map_consistent = pcibr_dma_map_consistent,
+	.dma_unmap = pcibr_dma_unmap,
+	.bus_fixup = pcibr_bus_fixup,
+};
+
+int
+pcibr_init_provider(void)
+{
+	sn_pci_provider[PCIIO_ASIC_TYPE_PIC] = &pcibr_provider;
+	sn_pci_provider[PCIIO_ASIC_TYPE_TIOCP] = &pcibr_provider;
+
+	return 0;
+}

arch/ia64/sn/pci/pcibr/pcibr_reg.c

@@ -8,8 +8,8 @@
 
 #include <linux/types.h>
 #include <linux/interrupt.h>
-#include "pci/pcibus_provider_defs.h"
-#include "pci/pcidev.h"
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include "pci/tiocp.h"
 #include "pci/pic.h"
 #include "pci/pcibr_provider.h"

drivers/char/Kconfig

@@ -399,6 +399,20 @@ config SGI_SNSC
 	  controller communication from user space (you want this!),
 	  say Y.  Otherwise, say N.
 
+config SGI_TIOCX
+       bool "SGI TIO CX driver support"
+       depends on (IA64_SGI_SN2 || IA64_GENERIC)
+       help
+         If you have an SGI Altix and you have fpga devices attached
+         to your TIO, say Y here, otherwise say N.
+
+config SGI_MBCS
+       tristate "SGI FPGA Core Services driver support"
+       depends on (IA64_SGI_SN2 || IA64_GENERIC)
+       help
+         If you have an SGI Altix with an attached SABrick
+         say Y or M here, otherwise say N.
+
 source "drivers/serial/Kconfig"
 
 config UNIX98_PTYS

drivers/char/Makefile

@@ -42,11 +42,12 @@ obj-$(CONFIG_SX)		+= sx.o generic_serial.o
 obj-$(CONFIG_RIO)		+= rio/ generic_serial.o
 obj-$(CONFIG_HVC_CONSOLE)	+= hvc_console.o hvsi.o
 obj-$(CONFIG_RAW_DRIVER)	+= raw.o
-obj-$(CONFIG_SGI_SNSC)		+= snsc.o
+obj-$(CONFIG_SGI_SNSC)		+= snsc.o snsc_event.o
 obj-$(CONFIG_MMTIMER)		+= mmtimer.o
 obj-$(CONFIG_VIOCONS) += viocons.o
 obj-$(CONFIG_VIOTAPE)		+= viotape.o
 obj-$(CONFIG_HVCS)		+= hvcs.o
+obj-$(CONFIG_SGI_MBCS)		+= mbcs.o
 
 obj-$(CONFIG_PRINTER) += lp.o
 obj-$(CONFIG_TIPAR) += tipar.o

drivers/char/snsc.c

@@ -374,6 +374,7 @@ scdrv_init(void)
 	void *salbuf;
 	struct class_simple *snsc_class;
 	dev_t first_dev, dev;
+	nasid_t event_nasid = ia64_sn_get_console_nasid();
 
 	if (alloc_chrdev_region(&first_dev, 0, numionodes,
 				SYSCTL_BASENAME) < 0) {
@@ -441,6 +442,13 @@ scdrv_init(void)
 			ia64_sn_irtr_intr_enable(scd->scd_nasid,
 						 0 /*ignored */ ,
 						 SAL_IROUTER_INTR_RECV);
+
+                        /* on the console nasid, prepare to receive
+                         * system controller environmental events
+                         */
+                        if(scd->scd_nasid == event_nasid) {
+                                scdrv_event_init(scd);
+                        }
 	}
 	return 0;
 }

drivers/char/snsc.h

@@ -47,4 +47,44 @@ struct sysctl_data_s {
 	nasid_t scd_nasid;	/* Node on which subchannels are opened. */
 };
 
+
+/* argument types */
+#define IR_ARG_INT              0x00    /* 4-byte integer (big-endian)  */
+#define IR_ARG_ASCII            0x01    /* null-terminated ASCII string */
+#define IR_ARG_UNKNOWN          0x80    /* unknown data type.  The low
+                                         * 7 bits will contain the data
+                                         * length.                      */
+#define IR_ARG_UNKNOWN_LENGTH_MASK	0x7f
+
+
+/* system controller event codes */
+#define EV_CLASS_MASK		0xf000ul
+#define EV_SEVERITY_MASK	0x0f00ul
+#define EV_COMPONENT_MASK	0x00fful
+
+#define EV_CLASS_POWER		0x1000ul
+#define EV_CLASS_FAN		0x2000ul
+#define EV_CLASS_TEMP		0x3000ul
+#define EV_CLASS_ENV		0x4000ul
+#define EV_CLASS_TEST_FAULT	0x5000ul
+#define EV_CLASS_TEST_WARNING	0x6000ul
+#define EV_CLASS_PWRD_NOTIFY	0x8000ul
+
+#define EV_SEVERITY_POWER_STABLE	0x0000ul
+#define EV_SEVERITY_POWER_LOW_WARNING	0x0100ul
+#define EV_SEVERITY_POWER_HIGH_WARNING	0x0200ul
+#define EV_SEVERITY_POWER_HIGH_FAULT	0x0300ul
+#define EV_SEVERITY_POWER_LOW_FAULT	0x0400ul
+
+#define EV_SEVERITY_FAN_STABLE		0x0000ul
+#define EV_SEVERITY_FAN_WARNING		0x0100ul
+#define EV_SEVERITY_FAN_FAULT		0x0200ul
+
+#define EV_SEVERITY_TEMP_STABLE		0x0000ul
+#define EV_SEVERITY_TEMP_ADVISORY	0x0100ul
+#define EV_SEVERITY_TEMP_CRITICAL	0x0200ul
+#define EV_SEVERITY_TEMP_FAULT		0x0300ul
+
+void scdrv_event_init(struct sysctl_data_s *);
+
 #endif /* _SN_SYSCTL_H_ */

drivers/char/snsc_event.c

+/*
+ * SN Platform system controller communication support
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2004 Silicon Graphics, Inc. All rights reserved.
+ */
+
+/*
+ * System controller event handler
+ *
+ * These routines deal with environmental events arriving from the
+ * system controllers.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/sched.h>
+#include <linux/byteorder/generic.h>
+#include <asm/sn/sn_sal.h>
+#include "snsc.h"
+
+static struct subch_data_s *event_sd;
+
+void scdrv_event(unsigned long);
+DECLARE_TASKLET(sn_sysctl_event, scdrv_event, 0);
+
+/*
+ * scdrv_event_interrupt
+ *
+ * Pull incoming environmental events off the physical link to the
+ * system controller and put them in a temporary holding area in SAL.
+ * Schedule scdrv_event() to move them along to their ultimate
+ * destination.
+ */
+static irqreturn_t
+scdrv_event_interrupt(int irq, void *subch_data, struct pt_regs *regs)
+{
+	struct subch_data_s *sd = subch_data;
+	unsigned long flags;
+	int status;
+
+	spin_lock_irqsave(&sd->sd_rlock, flags);
+	status = ia64_sn_irtr_intr(sd->sd_nasid, sd->sd_subch);
+
+	if ((status > 0) && (status & SAL_IROUTER_INTR_RECV)) {
+		tasklet_schedule(&sn_sysctl_event);
+	}
+	spin_unlock_irqrestore(&sd->sd_rlock, flags);
+	return IRQ_HANDLED;
+}
+
+
+/*
+ * scdrv_parse_event
+ *
+ * Break an event (as read from SAL) into useful pieces so we can decide
+ * what to do with it.
+ */
+static int
+scdrv_parse_event(char *event, int *src, int *code, int *esp_code, char *desc)
+{
+	char *desc_end;
+
+	/* record event source address */
+	*src = be32_to_cpup((__be32 *)event);
+	event += 4; 			/* move on to event code */
+
+	/* record the system controller's event code */
+	*code = be32_to_cpup((__be32 *)event);
+	event += 4;			/* move on to event arguments */
+
+	/* how many arguments are in the packet? */
+	if (*event++ != 2) {
+		/* if not 2, give up */
+		return -1;
+	}
+
+	/* parse out the ESP code */
+	if (*event++ != IR_ARG_INT) {
+		/* not an integer argument, so give up */
+		return -1;
+	}
+	*esp_code = be32_to_cpup((__be32 *)event);
+	event += 4;
+
+	/* parse out the event description */
+	if (*event++ != IR_ARG_ASCII) {
+		/* not an ASCII string, so give up */
+		return -1;
+	}
+	event[CHUNKSIZE-1] = '\0';	/* ensure this string ends! */
+	event += 2; 			/* skip leading CR/LF */
+	desc_end = desc + sprintf(desc, "%s", event);
+
+	/* strip trailing CR/LF (if any) */
+	for (desc_end--;
+	     (desc_end != desc) && ((*desc_end == 0xd) || (*desc_end == 0xa));
+	     desc_end--) {
+		*desc_end = '\0';
+	}
+
+	return 0;
+}
+
+
+/*
+ * scdrv_event_severity
+ *
+ * Figure out how urgent a message we should write to the console/syslog
+ * via printk.
+ */
+static char *
+scdrv_event_severity(int code)
+{
+	int ev_class = (code & EV_CLASS_MASK);
+	int ev_severity = (code & EV_SEVERITY_MASK);
+	char *pk_severity = KERN_NOTICE;
+
+	switch (ev_class) {
+	case EV_CLASS_POWER:
+		switch (ev_severity) {
+		case EV_SEVERITY_POWER_LOW_WARNING:
+		case EV_SEVERITY_POWER_HIGH_WARNING:
+			pk_severity = KERN_WARNING;
+			break;
+		case EV_SEVERITY_POWER_HIGH_FAULT:
+		case EV_SEVERITY_POWER_LOW_FAULT:
+			pk_severity = KERN_ALERT;
+			break;
+		}
+		break;
+	case EV_CLASS_FAN:
+		switch (ev_severity) {
+		case EV_SEVERITY_FAN_WARNING:
+			pk_severity = KERN_WARNING;
+			break;
+		case EV_SEVERITY_FAN_FAULT:
+			pk_severity = KERN_CRIT;
+			break;
+		}
+		break;
+	case EV_CLASS_TEMP:
+		switch (ev_severity) {
+		case EV_SEVERITY_TEMP_ADVISORY:
+			pk_severity = KERN_WARNING;
+			break;
+		case EV_SEVERITY_TEMP_CRITICAL:
+			pk_severity = KERN_CRIT;
+			break;
+		case EV_SEVERITY_TEMP_FAULT:
+			pk_severity = KERN_ALERT;
+			break;
+		}
+		break;
+	case EV_CLASS_ENV:
+		pk_severity = KERN_ALERT;
+		break;
+	case EV_CLASS_TEST_FAULT:
+		pk_severity = KERN_ALERT;
+		break;
+	case EV_CLASS_TEST_WARNING:
+		pk_severity = KERN_WARNING;
+		break;
+	case EV_CLASS_PWRD_NOTIFY:
+		pk_severity = KERN_ALERT;
+		break;
+	}
+
+	return pk_severity;
+}
+
+
+/*
+ * scdrv_dispatch_event
+ *
+ * Do the right thing with an incoming event.  That's often nothing
+ * more than printing it to the system log.  For power-down notifications
+ * we start a graceful shutdown.
+ */
+static void
+scdrv_dispatch_event(char *event, int len)
+{
+	int code, esp_code, src;
+	char desc[CHUNKSIZE];
+	char *severity;
+
+	if (scdrv_parse_event(event, &src, &code, &esp_code, desc) < 0) {
+		/* ignore uninterpretible event */
+		return;
+	}
+
+	/* how urgent is the message? */
+	severity = scdrv_event_severity(code);
+
+	if ((code & EV_CLASS_MASK) == EV_CLASS_PWRD_NOTIFY) {
+		struct task_struct *p;
+
+		/* give a SIGPWR signal to init proc */
+
+		/* first find init's task */
+		read_lock(&tasklist_lock);
+		for_each_process(p) {
+			if (p->pid == 1)
+				break;
+		}
+		if (p) { /* we found init's task */
+			printk(KERN_EMERG "Power off indication received. Initiating power fail sequence...\n");
+			force_sig(SIGPWR, p);
+		} else { /* failed to find init's task - just give message(s) */
+			printk(KERN_WARNING "Failed to find init proc to handle power off!\n");
+			printk("%s|$(0x%x)%s\n", severity, esp_code, desc);
+		}
+		read_unlock(&tasklist_lock);
+	} else {
+		/* print to system log */
+		printk("%s|$(0x%x)%s\n", severity, esp_code, desc);
+	}
+}
+
+
+/*
+ * scdrv_event
+ *
+ * Called as a tasklet when an event arrives from the L1.  Read the event
+ * from where it's temporarily stored in SAL and call scdrv_dispatch_event()
+ * to send it on its way.  Keep trying to read events until SAL indicates
+ * that there are no more immediately available.
+ */
+void
+scdrv_event(unsigned long dummy)
+{
+	int status;
+	int len;
+	unsigned long flags;
+	struct subch_data_s *sd = event_sd;
+
+	/* anything to read? */
+	len = CHUNKSIZE;
+	spin_lock_irqsave(&sd->sd_rlock, flags);
+	status = ia64_sn_irtr_recv(sd->sd_nasid, sd->sd_subch,
+				   sd->sd_rb, &len);
+
+	while (!(status < 0)) {
+		spin_unlock_irqrestore(&sd->sd_rlock, flags);
+		scdrv_dispatch_event(sd->sd_rb, len);
+		len = CHUNKSIZE;
+		spin_lock_irqsave(&sd->sd_rlock, flags);
+		status = ia64_sn_irtr_recv(sd->sd_nasid, sd->sd_subch,
+					   sd->sd_rb, &len);
+	}
+	spin_unlock_irqrestore(&sd->sd_rlock, flags);
+}
+
+
+/*
+ * scdrv_event_init
+ *
+ * Sets up a system controller subchannel to begin receiving event
+ * messages. This is sort of a specialized version of scdrv_open()
+ * in drivers/char/sn_sysctl.c.
+ */
+void
+scdrv_event_init(struct sysctl_data_s *scd)
+{
+	int rv;
+
+	event_sd = kmalloc(sizeof (struct subch_data_s), GFP_KERNEL);
+	if (event_sd == NULL) {
+		printk(KERN_WARNING "%s: couldn't allocate subchannel info"
+		       " for event monitoring\n", __FUNCTION__);
+		return;
+	}
+
+	/* initialize subch_data_s fields */
+	memset(event_sd, 0, sizeof (struct subch_data_s));
+	event_sd->sd_nasid = scd->scd_nasid;
+	spin_lock_init(&event_sd->sd_rlock);
+
+	/* ask the system controllers to send events to this node */
+	event_sd->sd_subch = ia64_sn_sysctl_event_init(scd->scd_nasid);
+
+	if (event_sd->sd_subch < 0) {
+		kfree(event_sd);
+		printk(KERN_WARNING "%s: couldn't open event subchannel\n",
+		       __FUNCTION__);
+		return;
+	}
+
+	/* hook event subchannel up to the system controller interrupt */
+	rv = request_irq(SGI_UART_VECTOR, scdrv_event_interrupt,
+			 SA_SHIRQ | SA_INTERRUPT,
+			 "system controller events", event_sd);
+	if (rv) {
+		printk(KERN_WARNING "%s: irq request failed (%d)\n",
+		       __FUNCTION__, rv);
+		ia64_sn_irtr_close(event_sd->sd_nasid, event_sd->sd_subch);
+		kfree(event_sd);
+		return;
+	}
+}
+
+

include/asm-ia64/hw_irq.h

@@ -81,6 +81,7 @@ extern __u8 isa_irq_to_vector_map[16];
 
 extern struct hw_interrupt_type irq_type_ia64_lsapic;	/* CPU-internal interrupt controller */
 
+extern int assign_irq_vector_nopanic (int irq); /* allocate a free vector without panic */
 extern int assign_irq_vector (int irq);	/* allocate a free vector */
 extern void free_irq_vector (int vector);
 extern void ia64_send_ipi (int cpu, int vector, int delivery_mode, int redirect);

include/asm-ia64/perfmon.h

@@ -254,6 +254,18 @@ extern int pfm_mod_write_dbrs(struct task_struct *task, void *req, unsigned int
 #define PFM_CPUINFO_DCR_PP	0x2	/* if set the system wide session has started */
 #define PFM_CPUINFO_EXCL_IDLE	0x4	/* the system wide session excludes the idle task */
 
+/*
+ * sysctl control structure. visible to sampling formats
+ */
+typedef struct {
+	int	debug;		/* turn on/off debugging via syslog */
+	int	debug_ovfl;	/* turn on/off debug printk in overflow handler */
+	int	fastctxsw;	/* turn on/off fast (unsecure) ctxsw */
+	int	expert_mode;	/* turn on/off value checking */
+} pfm_sysctl_t;
+extern pfm_sysctl_t pfm_sysctl;
+
+
 #endif /* __KERNEL__ */
 
 #endif /* _ASM_IA64_PERFMON_H */