diff options
Diffstat (limited to 'arch/ia64/sn')
-rw-r--r-- | arch/ia64/sn/kernel/Makefile | 3 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/io_init.c | 104 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/irq.c | 4 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/mca.c | 5 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/pio_phys.S | 71 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/setup.c | 50 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/sn2/prominfo_proc.c | 25 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/sn2/sn2_smp.c | 58 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/sn2/sn_hwperf.c | 5 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/sn2/sn_proc_fs.c | 22 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/sn2/timer.c | 19 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/sn2/timer_interrupt.c | 7 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/tiocx.c | 6 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/xpc_channel.c | 8 | ||||
-rw-r--r-- | arch/ia64/sn/kernel/xpc_main.c | 20 | ||||
-rw-r--r-- | arch/ia64/sn/pci/pci_dma.c | 16 | ||||
-rw-r--r-- | arch/ia64/sn/pci/pcibr/pcibr_ate.c | 29 | ||||
-rw-r--r-- | arch/ia64/sn/pci/pcibr/pcibr_dma.c | 14 | ||||
-rw-r--r-- | arch/ia64/sn/pci/pcibr/pcibr_provider.c | 9 | ||||
-rw-r--r-- | arch/ia64/sn/pci/tioce_provider.c | 326 |
20 files changed, 592 insertions, 209 deletions
diff --git a/arch/ia64/sn/kernel/Makefile b/arch/ia64/sn/kernel/Makefile index 3e9b4eea741..ab9c48c8801 100644 --- a/arch/ia64/sn/kernel/Makefile +++ b/arch/ia64/sn/kernel/Makefile @@ -10,7 +10,8 @@ CPPFLAGS += -I$(srctree)/arch/ia64/sn/include obj-y += setup.o bte.o bte_error.o irq.o mca.o idle.o \ - huberror.o io_init.o iomv.o klconflib.o sn2/ + huberror.o io_init.o iomv.o klconflib.o pio_phys.o \ + sn2/ obj-$(CONFIG_IA64_GENERIC) += machvec.o obj-$(CONFIG_SGI_TIOCX) += tiocx.o obj-$(CONFIG_IA64_SGI_SN_XP) += xp.o diff --git a/arch/ia64/sn/kernel/io_init.c b/arch/ia64/sn/kernel/io_init.c index d7e4d79e16a..dfb3f290237 100644 --- a/arch/ia64/sn/kernel/io_init.c +++ b/arch/ia64/sn/kernel/io_init.c @@ -23,6 +23,10 @@ #include "xtalk/hubdev.h" #include "xtalk/xwidgetdev.h" + +extern void sn_init_cpei_timer(void); +extern void register_sn_procfs(void); + static struct list_head sn_sysdata_list; /* sysdata list struct */ @@ -40,12 +44,12 @@ struct brick { struct slab_info slab_info[MAX_SLABS + 1]; }; -int sn_ioif_inited = 0; /* SN I/O infrastructure initialized? */ +int sn_ioif_inited; /* SN I/O infrastructure initialized? */ struct sn_pcibus_provider *sn_pci_provider[PCIIO_ASIC_MAX_TYPES]; /* indexed by asic type */ -static int max_segment_number = 0; /* Default highest segment number */ -static int max_pcibus_number = 255; /* Default highest pci bus number */ +static int max_segment_number; /* Default highest segment number */ +static int max_pcibus_number = 255; /* Default highest pci bus number */ /* * Hooks and struct for unsupported pci providers @@ -84,7 +88,6 @@ static inline u64 sal_get_device_dmaflush_list(u64 nasid, u64 widget_num, u64 device_num, u64 address) { - struct ia64_sal_retval ret_stuff; ret_stuff.status = 0; ret_stuff.v0 = 0; @@ -94,7 +97,6 @@ sal_get_device_dmaflush_list(u64 nasid, u64 widget_num, u64 device_num, (u64) nasid, (u64) widget_num, (u64) device_num, (u64) address, 0, 0, 0); return ret_stuff.status; - } /* @@ -102,7 +104,6 @@ sal_get_device_dmaflush_list(u64 nasid, u64 widget_num, u64 device_num, */ static inline u64 sal_get_hubdev_info(u64 handle, u64 address) { - struct ia64_sal_retval ret_stuff; ret_stuff.status = 0; ret_stuff.v0 = 0; @@ -118,7 +119,6 @@ static inline u64 sal_get_hubdev_info(u64 handle, u64 address) */ static inline u64 sal_get_pcibus_info(u64 segment, u64 busnum, u64 address) { - struct ia64_sal_retval ret_stuff; ret_stuff.status = 0; ret_stuff.v0 = 0; @@ -215,7 +215,7 @@ static void __init sn_fixup_ionodes(void) struct hubdev_info *hubdev; u64 status; u64 nasid; - int i, widget, device; + int i, widget, device, size; /* * Get SGI Specific HUB chipset information. @@ -251,48 +251,37 @@ static void __init sn_fixup_ionodes(void) if (!hubdev->hdi_flush_nasid_list.widget_p) continue; + size = (HUB_WIDGET_ID_MAX + 1) * + sizeof(struct sn_flush_device_kernel *); hubdev->hdi_flush_nasid_list.widget_p = - kmalloc((HUB_WIDGET_ID_MAX + 1) * - sizeof(struct sn_flush_device_kernel *), - GFP_KERNEL); - memset(hubdev->hdi_flush_nasid_list.widget_p, 0x0, - (HUB_WIDGET_ID_MAX + 1) * - sizeof(struct sn_flush_device_kernel *)); + kzalloc(size, GFP_KERNEL); + if (!hubdev->hdi_flush_nasid_list.widget_p) + BUG(); for (widget = 0; widget <= HUB_WIDGET_ID_MAX; widget++) { - sn_flush_device_kernel = kmalloc(DEV_PER_WIDGET * - sizeof(struct - sn_flush_device_kernel), - GFP_KERNEL); + size = DEV_PER_WIDGET * + sizeof(struct sn_flush_device_kernel); + sn_flush_device_kernel = kzalloc(size, GFP_KERNEL); if (!sn_flush_device_kernel) BUG(); - memset(sn_flush_device_kernel, 0x0, - DEV_PER_WIDGET * - sizeof(struct sn_flush_device_kernel)); dev_entry = sn_flush_device_kernel; for (device = 0; device < DEV_PER_WIDGET; device++,dev_entry++) { - dev_entry->common = kmalloc(sizeof(struct - sn_flush_device_common), - GFP_KERNEL); + size = sizeof(struct sn_flush_device_common); + dev_entry->common = kzalloc(size, GFP_KERNEL); if (!dev_entry->common) BUG(); - memset(dev_entry->common, 0x0, sizeof(struct - sn_flush_device_common)); if (sn_prom_feature_available( PRF_DEVICE_FLUSH_LIST)) status = sal_get_device_dmaflush_list( - nasid, - widget, - device, - (u64)(dev_entry->common)); + nasid, widget, device, + (u64)(dev_entry->common)); else status = sn_device_fixup_war(nasid, - widget, - device, - dev_entry->common); + widget, device, + dev_entry->common); if (status != SALRET_OK) panic("SAL call failed: %s\n", ia64_sal_strerror(status)); @@ -383,13 +372,12 @@ void sn_pci_fixup_slot(struct pci_dev *dev) pci_dev_get(dev); /* for the sysdata pointer */ pcidev_info = kzalloc(sizeof(struct pcidev_info), GFP_KERNEL); - if (pcidev_info <= 0) + if (!pcidev_info) BUG(); /* Cannot afford to run out of memory */ - sn_irq_info = kmalloc(sizeof(struct sn_irq_info), GFP_KERNEL); - if (sn_irq_info <= 0) + sn_irq_info = kzalloc(sizeof(struct sn_irq_info), GFP_KERNEL); + if (!sn_irq_info) BUG(); /* Cannot afford to run out of memory */ - memset(sn_irq_info, 0, sizeof(struct sn_irq_info)); /* Call to retrieve pci device information needed by kernel. */ status = sal_get_pcidev_info((u64) segment, (u64) dev->bus->number, @@ -482,13 +470,13 @@ void sn_pci_fixup_slot(struct pci_dev *dev) */ void sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus) { - int status = 0; + int status; int nasid, cnode; struct pci_controller *controller; struct sn_pci_controller *sn_controller; struct pcibus_bussoft *prom_bussoft_ptr; struct hubdev_info *hubdev_info; - void *provider_soft = NULL; + void *provider_soft; struct sn_pcibus_provider *provider; status = sal_get_pcibus_info((u64) segment, (u64) busnum, @@ -535,6 +523,8 @@ void sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus) bus->sysdata = controller; if (provider->bus_fixup) provider_soft = (*provider->bus_fixup) (prom_bussoft_ptr, controller); + else + provider_soft = NULL; if (provider_soft == NULL) { /* fixup failed or not applicable */ @@ -617,15 +607,15 @@ void sn_bus_store_sysdata(struct pci_dev *dev) void sn_bus_free_sysdata(void) { struct sysdata_el *element; - struct list_head *list; + struct list_head *list, *safe; -sn_sysdata_free_start: - list_for_each(list, &sn_sysdata_list) { + list_for_each_safe(list, safe, &sn_sysdata_list) { element = list_entry(list, struct sysdata_el, entry); list_del(&element->entry); + list_del(&(((struct pcidev_info *) + (element->sysdata))->pdi_list)); kfree(element->sysdata); kfree(element); - goto sn_sysdata_free_start; } return; } @@ -638,13 +628,8 @@ sn_sysdata_free_start: static int __init sn_pci_init(void) { - int i = 0; - int j = 0; + int i, j; struct pci_dev *pci_dev = NULL; - extern void sn_init_cpei_timer(void); -#ifdef CONFIG_PROC_FS - extern void register_sn_procfs(void); -#endif if (!ia64_platform_is("sn2") || IS_RUNNING_ON_FAKE_PROM()) return 0; @@ -700,32 +685,29 @@ static int __init sn_pci_init(void) */ void hubdev_init_node(nodepda_t * npda, cnodeid_t node) { - struct hubdev_info *hubdev_info; + int size; + pg_data_t *pg; + + size = sizeof(struct hubdev_info); if (node >= num_online_nodes()) /* Headless/memless IO nodes */ - hubdev_info = - (struct hubdev_info *)alloc_bootmem_node(NODE_DATA(0), - sizeof(struct - hubdev_info)); + pg = NODE_DATA(0); else - hubdev_info = - (struct hubdev_info *)alloc_bootmem_node(NODE_DATA(node), - sizeof(struct - hubdev_info)); - npda->pdinfo = (void *)hubdev_info; + pg = NODE_DATA(node); + hubdev_info = (struct hubdev_info *)alloc_bootmem_node(pg, size); + + npda->pdinfo = (void *)hubdev_info; } geoid_t cnodeid_get_geoid(cnodeid_t cnode) { - struct hubdev_info *hubdev; hubdev = (struct hubdev_info *)(NODEPDA(cnode)->pdinfo); return hubdev->hdi_geoid; - } subsys_initcall(sn_pci_init); diff --git a/arch/ia64/sn/kernel/irq.c b/arch/ia64/sn/kernel/irq.c index 74d87d903d5..c373113d073 100644 --- a/arch/ia64/sn/kernel/irq.c +++ b/arch/ia64/sn/kernel/irq.c @@ -299,7 +299,9 @@ void sn_irq_unfixup(struct pci_dev *pci_dev) return; sn_irq_info = SN_PCIDEV_INFO(pci_dev)->pdi_sn_irq_info; - if (!sn_irq_info || !sn_irq_info->irq_irq) { + if (!sn_irq_info) + return; + if (!sn_irq_info->irq_irq) { kfree(sn_irq_info); return; } diff --git a/arch/ia64/sn/kernel/mca.c b/arch/ia64/sn/kernel/mca.c index 9ab684d1bb5..3db62f24596 100644 --- a/arch/ia64/sn/kernel/mca.c +++ b/arch/ia64/sn/kernel/mca.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) 2000-2006 Silicon Graphics, Inc. All Rights Reserved. */ #include <linux/types.h> @@ -137,7 +137,8 @@ int sn_salinfo_platform_oemdata(const u8 *sect_header, u8 **oemdata, u64 *oemdat static int __init sn_salinfo_init(void) { - salinfo_platform_oemdata = &sn_salinfo_platform_oemdata; + if (ia64_platform_is("sn2")) + salinfo_platform_oemdata = &sn_salinfo_platform_oemdata; return 0; } diff --git a/arch/ia64/sn/kernel/pio_phys.S b/arch/ia64/sn/kernel/pio_phys.S new file mode 100644 index 00000000000..3c7d48d6ecb --- /dev/null +++ b/arch/ia64/sn/kernel/pio_phys.S @@ -0,0 +1,71 @@ +/* + * 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) 2000-2005 Silicon Graphics, Inc. All rights reserved. + * + * This file contains macros used to access MMR registers via + * uncached physical addresses. + * pio_phys_read_mmr - read an MMR + * pio_phys_write_mmr - write an MMR + * pio_atomic_phys_write_mmrs - atomically write 1 or 2 MMRs with psr.ic=0 + * Second MMR will be skipped if address is NULL + * + * Addresses passed to these routines should be uncached physical addresses + * ie., 0x80000.... + */ + + + +#include <asm/asmmacro.h> +#include <asm/page.h> + +GLOBAL_ENTRY(pio_phys_read_mmr) + .prologue + .regstk 1,0,0,0 + .body + mov r2=psr + rsm psr.i | psr.dt + ;; + srlz.d + ld8.acq r8=[r32] + ;; + mov psr.l=r2;; + srlz.d + br.ret.sptk.many rp +END(pio_phys_read_mmr) + +GLOBAL_ENTRY(pio_phys_write_mmr) + .prologue + .regstk 2,0,0,0 + .body + mov r2=psr + rsm psr.i | psr.dt + ;; + srlz.d + st8.rel [r32]=r33 + ;; + mov psr.l=r2;; + srlz.d + br.ret.sptk.many rp +END(pio_phys_write_mmr) + +GLOBAL_ENTRY(pio_atomic_phys_write_mmrs) + .prologue + .regstk 4,0,0,0 + .body + mov r2=psr + cmp.ne p9,p0=r34,r0; + rsm psr.i | psr.dt | psr.ic + ;; + srlz.d + st8.rel [r32]=r33 +(p9) st8.rel [r34]=r35 + ;; + mov psr.l=r2;; + srlz.d + br.ret.sptk.many rp +END(pio_atomic_phys_write_mmrs) + + diff --git a/arch/ia64/sn/kernel/setup.c b/arch/ia64/sn/kernel/setup.c index aac1ba32c5e..8b6d5c84470 100644 --- a/arch/ia64/sn/kernel/setup.c +++ b/arch/ia64/sn/kernel/setup.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 1999,2001-2005 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 1999,2001-2006 Silicon Graphics, Inc. All rights reserved. */ #include <linux/config.h> @@ -67,6 +67,7 @@ extern unsigned long last_time_offset; extern void (*ia64_mark_idle) (int); extern void snidle(int); extern unsigned char acpi_kbd_controller_present; +extern unsigned long long (*ia64_printk_clock)(void); unsigned long sn_rtc_cycles_per_second; EXPORT_SYMBOL(sn_rtc_cycles_per_second); @@ -74,7 +75,7 @@ EXPORT_SYMBOL(sn_rtc_cycles_per_second); DEFINE_PER_CPU(struct sn_hub_info_s, __sn_hub_info); EXPORT_PER_CPU_SYMBOL(__sn_hub_info); -DEFINE_PER_CPU(short, __sn_cnodeid_to_nasid[MAX_NUMNODES]); +DEFINE_PER_CPU(short, __sn_cnodeid_to_nasid[MAX_COMPACT_NODES]); EXPORT_PER_CPU_SYMBOL(__sn_cnodeid_to_nasid); DEFINE_PER_CPU(struct nodepda_s *, __sn_nodepda); @@ -316,6 +317,7 @@ struct pcdp_vga_device { #define PCDP_PCI_TRANS_IOPORT 0x02 #define PCDP_PCI_TRANS_MMIO 0x01 +#if defined(CONFIG_VT) && defined(CONFIG_VGA_CONSOLE) static void sn_scan_pcdp(void) { @@ -357,6 +359,17 @@ sn_scan_pcdp(void) break; /* once we find the primary, we're done */ } } +#endif + +static unsigned long sn2_rtc_initial; + +static unsigned long long ia64_sn2_printk_clock(void) +{ + unsigned long rtc_now = rtc_time(); + + return (rtc_now - sn2_rtc_initial) * + (1000000000 / sn_rtc_cycles_per_second); +} /** * sn_setup - SN platform setup routine @@ -372,6 +385,7 @@ void __init sn_setup(char **cmdline_p) u32 version = sn_sal_rev(); extern void sn_cpu_init(void); + sn2_rtc_initial = rtc_time(); ia64_sn_plat_set_error_handling_features(); // obsolete ia64_sn_set_os_feature(OSF_MCA_SLV_TO_OS_INIT_SLV); ia64_sn_set_os_feature(OSF_FEAT_LOG_SBES); @@ -423,19 +437,6 @@ void __init sn_setup(char **cmdline_p) */ build_cnode_tables(); - /* - * Old PROMs do not provide an ACPI FADT. Disable legacy keyboard - * support here so we don't have to listen to failed keyboard probe - * messages. - */ - if (version <= 0x0209 && acpi_kbd_controller_present) { - printk(KERN_INFO "Disabling legacy keyboard support as prom " - "is too old and doesn't provide FADT\n"); - acpi_kbd_controller_present = 0; - } - - printk("SGI SAL version %x.%02x\n", version >> 8, version & 0x00FF); - status = ia64_sal_freq_base(SAL_FREQ_BASE_REALTIME_CLOCK, &ticks_per_sec, &drift); @@ -449,6 +450,21 @@ void __init sn_setup(char **cmdline_p) platform_intr_list[ACPI_INTERRUPT_CPEI] = IA64_CPE_VECTOR; + ia64_printk_clock = ia64_sn2_printk_clock; + + /* + * Old PROMs do not provide an ACPI FADT. Disable legacy keyboard + * support here so we don't have to listen to failed keyboard probe + * messages. + */ + if (version <= 0x0209 && acpi_kbd_controller_present) { + printk(KERN_INFO "Disabling legacy keyboard support as prom " + "is too old and doesn't provide FADT\n"); + acpi_kbd_controller_present = 0; + } + + printk("SGI SAL version %x.%02x\n", version >> 8, version & 0x00FF); + /* * we set the default root device to /dev/hda * to make simulation easy @@ -482,6 +498,7 @@ void __init sn_setup(char **cmdline_p) * for sn. */ pm_power_off = ia64_sn_power_down; + current->thread.flags |= IA64_THREAD_MIGRATION; } /** @@ -644,7 +661,8 @@ void __init sn_cpu_init(void) SH2_PIO_WRITE_STATUS_1, SH2_PIO_WRITE_STATUS_3}; u64 *pio; pio = is_shub1() ? pio1 : pio2; - pda->pio_write_status_addr = (volatile unsigned long *) LOCAL_MMR_ADDR(pio[slice]); + pda->pio_write_status_addr = + (volatile unsigned long *)GLOBAL_MMR_ADDR(nasid, pio[slice]); pda->pio_write_status_val = is_shub1() ? SH_PIO_WRITE_STATUS_PENDING_WRITE_COUNT_MASK : 0; } diff --git a/arch/ia64/sn/kernel/sn2/prominfo_proc.c b/arch/ia64/sn/kernel/sn2/prominfo_proc.c index 81c63b2f8ae..6ae276d5d50 100644 --- a/arch/ia64/sn/kernel/sn2/prominfo_proc.c +++ b/arch/ia64/sn/kernel/sn2/prominfo_proc.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 1999,2001-2004 Silicon Graphics, Inc. All Rights Reserved. + * Copyright (C) 1999,2001-2004, 2006 Silicon Graphics, Inc. All Rights Reserved. * * Module to export the system's Firmware Interface Tables, including * PROM revision numbers and banners, in /proc @@ -190,7 +190,7 @@ static int read_version_entry(char *page, char **start, off_t off, int count, int *eof, void *data) { - int len = 0; + int len; /* data holds the NASID of the node */ len = dump_version(page, (unsigned long)data); @@ -202,7 +202,7 @@ static int read_fit_entry(char *page, char **start, off_t off, int count, int *eof, void *data) { - int len = 0; + int len; /* data holds the NASID of the node */ len = dump_fit(page, (unsigned long)data); @@ -229,13 +229,16 @@ int __init prominfo_init(void) struct proc_dir_entry *p; cnodeid_t cnodeid; unsigned long nasid; + int size; char name[NODE_NAME_LEN]; if (!ia64_platform_is("sn2")) return 0; - proc_entries = kmalloc(num_online_nodes() * sizeof(struct proc_dir_entry *), - GFP_KERNEL); + size = num_online_nodes() * sizeof(struct proc_dir_entry *); + proc_entries = kzalloc(size, GFP_KERNEL); + if (!proc_entries) + return -ENOMEM; sgi_prominfo_entry = proc_mkdir("sgi_prominfo", NULL); @@ -244,14 +247,12 @@ int __init prominfo_init(void) sprintf(name, "node%d", cnodeid); *entp = proc_mkdir(name, sgi_prominfo_entry); nasid = cnodeid_to_nasid(cnodeid); - p = create_proc_read_entry( - "fit", 0, *entp, read_fit_entry, - (void *)nasid); + p = create_proc_read_entry("fit", 0, *entp, read_fit_entry, + (void *)nasid); if (p) p->owner = THIS_MODULE; - p = create_proc_read_entry( - "version", 0, *entp, read_version_entry, - (void *)nasid); + p = create_proc_read_entry("version", 0, *entp, + read_version_entry, (void *)nasid); if (p) p->owner = THIS_MODULE; entp++; @@ -263,7 +264,7 @@ int __init prominfo_init(void) void __exit prominfo_exit(void) { struct proc_dir_entry **entp; - unsigned cnodeid; + unsigned int cnodeid; char name[NODE_NAME_LEN]; entp = proc_entries; diff --git a/arch/ia64/sn/kernel/sn2/sn2_smp.c b/arch/ia64/sn/kernel/sn2/sn2_smp.c index f153a4c35c7..d9d306c79f2 100644 --- a/arch/ia64/sn/kernel/sn2/sn2_smp.c +++ b/arch/ia64/sn/kernel/sn2/sn2_smp.c @@ -46,8 +46,14 @@ DECLARE_PER_CPU(struct ptc_stats, ptcstats); static __cacheline_aligned DEFINE_SPINLOCK(sn2_global_ptc_lock); -void sn2_ptc_deadlock_recovery(short *, short, short, int, volatile unsigned long *, unsigned long, - volatile unsigned long *, unsigned long); +extern unsigned long +sn2_ptc_deadlock_recovery_core(volatile unsigned long *, unsigned long, + volatile unsigned long *, unsigned long, + volatile unsigned long *, unsigned long); +void +sn2_ptc_deadlock_recovery(short *, short, short, int, + volatile unsigned long *, unsigned long, + volatile unsigned long *, unsigned long); /* * Note: some is the following is captured here to make degugging easier @@ -59,16 +65,6 @@ void sn2_ptc_deadlock_recovery(short *, short, short, int, volatile unsigned lon #define reset_max_active_on_deadlock() 1 #define PTC_LOCK(sh1) ((sh1) ? &sn2_global_ptc_lock : &sn_nodepda->ptc_lock) -static inline void ptc_lock(int sh1, unsigned long *flagp) -{ - spin_lock_irqsave(PTC_LOCK(sh1), *flagp); -} - -static inline void ptc_unlock(int sh1, unsigned long flags) -{ - spin_unlock_irqrestore(PTC_LOCK(sh1), flags); -} - struct ptc_stats { unsigned long ptc_l; unsigned long change_rid; @@ -82,6 +78,8 @@ struct ptc_stats { unsigned long shub_ptc_flushes_not_my_mm; }; +#define sn2_ptctest 0 + static inline unsigned long wait_piowc(void) { volatile unsigned long *piows; @@ -95,6 +93,27 @@ static inline unsigned long wait_piowc(void) return (ws & SH_PIO_WRITE_STATUS_WRITE_DEADLOCK_MASK) != 0; } +/** + * sn_migrate - SN-specific task migration actions + * @task: Task being migrated to new CPU + * + * SN2 PIO writes from separate CPUs are not guaranteed to arrive in order. + * Context switching user threads which have memory-mapped MMIO may cause + * PIOs to issue from seperate CPUs, thus the PIO writes must be drained + * from the previous CPU's Shub before execution resumes on the new CPU. + */ +void sn_migrate(struct task_struct *task) +{ + pda_t *last_pda = pdacpu(task_thread_info(task)->last_cpu); + volatile unsigned long *adr = last_pda->pio_write_status_addr; + unsigned long val = last_pda->pio_write_status_val; + + /* Drain PIO writes from old CPU's Shub */ + while (unlikely((*adr & SH_PIO_WRITE_STATUS_PENDING_WRITE_COUNT_MASK) + != val)) + cpu_relax(); +} + void sn_tlb_migrate_finish(struct mm_struct *mm) { /* flush_tlb_mm is inefficient if more than 1 users of mm */ @@ -200,7 +219,7 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, max_active = max_active_pio(shub1); itc = ia64_get_itc(); - ptc_lock(shub1, &flags); + spin_lock_irqsave(PTC_LOCK(shub1), flags); itc2 = ia64_get_itc(); __get_cpu_var(ptcstats).lock_itc_clocks += itc2 - itc; @@ -258,7 +277,7 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, ia64_srlz_d(); } - ptc_unlock(shub1, flags); + spin_unlock_irqrestore(PTC_LOCK(shub1), flags); preempt_enable(); } @@ -270,11 +289,12 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start, * TLB flush transaction. The recovery sequence is somewhat tricky & is * coded in assembly language. */ -void sn2_ptc_deadlock_recovery(short *nasids, short ib, short ie, int mynasid, volatile unsigned long *ptc0, unsigned long data0, - volatile unsigned long *ptc1, unsigned long data1) + +void +sn2_ptc_deadlock_recovery(short *nasids, short ib, short ie, int mynasid, + volatile unsigned long *ptc0, unsigned long data0, + volatile unsigned long *ptc1, unsigned long data1) { - extern unsigned long sn2_ptc_deadlock_recovery_core(volatile unsigned long *, unsigned long, - volatile unsigned long *, unsigned long, volatile unsigned long *, unsigned long); short nasid, i; unsigned long *piows, zeroval, n; @@ -447,7 +467,7 @@ static struct proc_dir_entry *proc_sn2_ptc; static int __init sn2_ptc_init(void) { if (!ia64_platform_is("sn2")) - return -ENOSYS; + return 0; if (!(proc_sn2_ptc = create_proc_entry(PTC_BASENAME, 0444, NULL))) { printk(KERN_ERR "unable to create %s proc entry", PTC_BASENAME); diff --git a/arch/ia64/sn/kernel/sn2/sn_hwperf.c b/arch/ia64/sn/kernel/sn2/sn_hwperf.c index 19b54fbcd7e..70db21f3df2 100644 --- a/arch/ia64/sn/kernel/sn2/sn_hwperf.c +++ b/arch/ia64/sn/kernel/sn2/sn_hwperf.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2004-2005 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 2004-2006 Silicon Graphics, Inc. All rights reserved. * * SGI Altix topology and hardware performance monitoring API. * Mark Goodwin <markgw@sgi.com>. @@ -973,6 +973,9 @@ static int __devinit sn_hwperf_misc_register_init(void) { int e; + if (!ia64_platform_is("sn2")) + return 0; + sn_hwperf_init(); /* diff --git a/arch/ia64/sn/kernel/sn2/sn_proc_fs.c b/arch/ia64/sn/kernel/sn2/sn_proc_fs.c index a06719d752a..c686d9c12f7 100644 --- a/arch/ia64/sn/kernel/sn2/sn_proc_fs.c +++ b/arch/ia64/sn/kernel/sn2/sn_proc_fs.c @@ -6,11 +6,11 @@ * Copyright (C) 2000-2005 Silicon Graphics, Inc. All rights reserved. */ #include <linux/config.h> -#include <asm/uaccess.h> #ifdef CONFIG_PROC_FS #include <linux/proc_fs.h> #include <linux/seq_file.h> +#include <asm/uaccess.h> #include <asm/sn/sn_sal.h> static int partition_id_show(struct seq_file *s, void *p) @@ -90,10 +90,10 @@ static int coherence_id_open(struct inode *inode, struct file *file) return single_open(file, coherence_id_show, NULL); } -static struct proc_dir_entry *sn_procfs_create_entry( - const char *name, struct proc_dir_entry *parent, - int (*openfunc)(struct inode *, struct file *), - int (*releasefunc)(struct inode *, struct file *)) +static struct proc_dir_entry +*sn_procfs_create_entry(const char *name, struct proc_dir_entry *parent, + int (*openfunc)(struct inode *, struct file *), + int (*releasefunc)(struct inode *, struct file *)) { struct proc_dir_entry *e = create_proc_entry(name, 0444, parent); @@ -126,24 +126,24 @@ void register_sn_procfs(void) return; sn_procfs_create_entry("partition_id", sgi_proc_dir, - partition_id_open, single_release); + partition_id_open, single_release); sn_procfs_create_entry("system_serial_number", sgi_proc_dir, - system_serial_number_open, single_release); + system_serial_number_open, single_release); sn_procfs_create_entry("licenseID", sgi_proc_dir, - licenseID_open, single_release); + licenseID_open, single_release); e = sn_procfs_create_entry("sn_force_interrupt", sgi_proc_dir, - sn_force_interrupt_open, single_release); + sn_force_interrupt_open, single_release); if (e) e->proc_fops->write = sn_force_interrupt_write_proc; sn_procfs_create_entry("coherence_id", sgi_proc_dir, - coherence_id_open, single_release); + coherence_id_open, single_release); sn_procfs_create_entry("sn_topology", sgi_proc_dir, - sn_topology_open, sn_topology_release); + sn_topology_open, sn_topology_release); } #endif /* CONFIG_PROC_FS */ diff --git a/arch/ia64/sn/kernel/sn2/timer.c b/arch/ia64/sn/kernel/sn2/timer.c index deb9baf4d47..56a88b6df4b 100644 --- a/arch/ia64/sn/kernel/sn2/timer.c +++ b/arch/ia64/sn/kernel/sn2/timer.c @@ -14,6 +14,7 @@ #include <asm/hw_irq.h> #include <asm/system.h> +#include <asm/timex.h> #include <asm/sn/leds.h> #include <asm/sn/shub_mmr.h> @@ -28,9 +29,27 @@ static struct time_interpolator sn2_interpolator = { .source = TIME_SOURCE_MMIO64 }; +/* + * sn udelay uses the RTC instead of the ITC because the ITC is not + * synchronized across all CPUs, and the thread may migrate to another CPU + * if preemption is enabled. + */ +static void +ia64_sn_udelay (unsigned long usecs) +{ + unsigned long start = rtc_time(); + unsigned long end = start + + usecs * sn_rtc_cycles_per_second / 1000000; + + while (time_before((unsigned long)rtc_time(), end)) + cpu_relax(); +} + void __init sn_timer_init(void) { sn2_interpolator.frequency = sn_rtc_cycles_per_second; sn2_interpolator.addr = RTC_COUNTER_ADDR; register_time_interpolator(&sn2_interpolator); + + ia64_udelay = &ia64_sn_udelay; } diff --git a/arch/ia64/sn/kernel/sn2/timer_interrupt.c b/arch/ia64/sn/kernel/sn2/timer_interrupt.c index adf5db2e2af..fa7f6994591 100644 --- a/arch/ia64/sn/kernel/sn2/timer_interrupt.c +++ b/arch/ia64/sn/kernel/sn2/timer_interrupt.c @@ -1,7 +1,7 @@ /* * * - * Copyright (c) 2005 Silicon Graphics, Inc. All Rights Reserved. + * Copyright (c) 2005, 2006 Silicon Graphics, Inc. All Rights Reserved. * * This program is free software; you can redistribute it and/or modify it * under the terms of version 2 of the GNU General Public License @@ -22,11 +22,6 @@ * License along with this program; if not, write the Free Software * Foundation, Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA. * - * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy, - * Mountain View, CA 94043, or: - * - * http://www.sgi.com - * * For further information regarding this notice, see: * * http://oss.sgi.com/projects/GenInfo/NoticeExplan diff --git a/arch/ia64/sn/kernel/tiocx.c b/arch/ia64/sn/kernel/tiocx.c index d263d3e8fbb..99cb28e7429 100644 --- a/arch/ia64/sn/kernel/tiocx.c +++ b/arch/ia64/sn/kernel/tiocx.c @@ -284,12 +284,10 @@ struct sn_irq_info *tiocx_irq_alloc(nasid_t nasid, int widget, int irq, if ((nasid & 1) == 0) return NULL; - sn_irq_info = kmalloc(sn_irq_size, GFP_KERNEL); + sn_irq_info = kzalloc(sn_irq_size, GFP_KERNEL); if (sn_irq_info == NULL) return NULL; - memset(sn_irq_info, 0x0, sn_irq_size); - status = tiocx_intr_alloc(nasid, widget, __pa(sn_irq_info), irq, req_nasid, slice); if (status) { @@ -486,7 +484,7 @@ static int __init tiocx_init(void) int found_tiocx_device = 0; if (!ia64_platform_is("sn2")) - return -ENODEV; + return 0; bus_register(&tiocx_bus_type); diff --git a/arch/ia64/sn/kernel/xpc_channel.c b/arch/ia64/sn/kernel/xpc_channel.c index 36e5437a0fb..cdf6856ce08 100644 --- a/arch/ia64/sn/kernel/xpc_channel.c +++ b/arch/ia64/sn/kernel/xpc_channel.c @@ -738,7 +738,9 @@ xpc_process_disconnect(struct xpc_channel *ch, unsigned long *irq_flags) /* make sure all activity has settled down first */ - if (atomic_read(&ch->references) > 0) { + if (atomic_read(&ch->references) > 0 || + ((ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) && + !(ch->flags & XPC_C_DISCONNECTINGCALLOUT_MADE))) { return; } DBUG_ON(atomic_read(&ch->kthreads_assigned) != 0); @@ -775,7 +777,7 @@ xpc_process_disconnect(struct xpc_channel *ch, unsigned long *irq_flags) /* both sides are disconnected now */ - if (ch->flags & XPC_C_CONNECTCALLOUT) { + if (ch->flags & XPC_C_DISCONNECTINGCALLOUT_MADE) { spin_unlock_irqrestore(&ch->lock, *irq_flags); xpc_disconnect_callout(ch, xpcDisconnected); spin_lock_irqsave(&ch->lock, *irq_flags); @@ -1300,7 +1302,7 @@ xpc_process_msg_IPI(struct xpc_partition *part, int ch_number) "delivered=%d, partid=%d, channel=%d\n", nmsgs_sent, ch->partid, ch->number); - if (ch->flags & XPC_C_CONNECTCALLOUT) { + if (ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) { xpc_activate_kthreads(ch, nmsgs_sent); } } diff --git a/arch/ia64/sn/kernel/xpc_main.c b/arch/ia64/sn/kernel/xpc_main.c index 9cd460dfe27..8cbf1643257 100644 --- a/arch/ia64/sn/kernel/xpc_main.c +++ b/arch/ia64/sn/kernel/xpc_main.c @@ -750,12 +750,16 @@ xpc_daemonize_kthread(void *args) /* let registerer know that connection has been established */ spin_lock_irqsave(&ch->lock, irq_flags); - if (!(ch->flags & XPC_C_CONNECTCALLOUT)) { - ch->flags |= XPC_C_CONNECTCALLOUT; + if (!(ch->flags & XPC_C_CONNECTEDCALLOUT)) { + ch->flags |= XPC_C_CONNECTEDCALLOUT; spin_unlock_irqrestore(&ch->lock, irq_flags); xpc_connected_callout(ch); + spin_lock_irqsave(&ch->lock, irq_flags); + ch->flags |= XPC_C_CONNECTEDCALLOUT_MADE; + spin_unlock_irqrestore(&ch->lock, irq_flags); + /* * It is possible that while the callout was being * made that the remote partition sent some messages. @@ -777,15 +781,17 @@ xpc_daemonize_kthread(void *args) if (atomic_dec_return(&ch->kthreads_assigned) == 0) { spin_lock_irqsave(&ch->lock, irq_flags); - if ((ch->flags & XPC_C_CONNECTCALLOUT) && - !(ch->flags & XPC_C_DISCONNECTCALLOUT)) { - ch->flags |= XPC_C_DISCONNECTCALLOUT; + if ((ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) && + !(ch->flags & XPC_C_DISCONNECTINGCALLOUT)) { + ch->flags |= XPC_C_DISCONNECTINGCALLOUT; spin_unlock_irqrestore(&ch->lock, irq_flags); xpc_disconnect_callout(ch, xpcDisconnecting); - } else { - spin_unlock_irqrestore(&ch->lock, irq_flags); + + spin_lock_irqsave(&ch->lock, irq_flags); + ch->flags |= XPC_C_DISCONNECTINGCALLOUT_MADE; } + spin_unlock_irqrestore(&ch->lock, irq_flags); if (atomic_dec_return(&part->nchannels_engaged) == 0) { xpc_mark_partition_disengaged(part); xpc_IPI_send_disengage(part); diff --git a/arch/ia64/sn/pci/pci_dma.c b/arch/ia64/sn/pci/pci_dma.c index 5a36292388e..b4b84c26921 100644 --- a/arch/ia64/sn/pci/pci_dma.c +++ b/arch/ia64/sn/pci/pci_dma.c @@ -335,10 +335,10 @@ int sn_pci_legacy_read(struct pci_bus *bus, u16 port, u32 *val, u8 size) */ SAL_CALL(isrv, SN_SAL_IOIF_PCI_SAFE, - pci_domain_nr(bus), bus->number, - 0, /* io */ - 0, /* read */ - port, size, __pa(val)); + pci_domain_nr(bus), bus->number, + 0, /* io */ + 0, /* read */ + port, size, __pa(val)); if (isrv.status == 0) return size; @@ -381,10 +381,10 @@ int sn_pci_legacy_write(struct pci_bus *bus, u16 port, u32 val, u8 size) */ SAL_CALL(isrv, SN_SAL_IOIF_PCI_SAFE, - pci_domain_nr(bus), bus->number, - 0, /* io */ - 1, /* write */ - port, size, __pa(&val)); + pci_domain_nr(bus), bus->number, + 0, /* io */ + 1, /* write */ + port, size, __pa(&val)); if (isrv.status == 0) return size; diff --git a/arch/ia64/sn/pci/pcibr/pcibr_ate.c b/arch/ia64/sn/pci/pcibr/pcibr_ate.c index aa3fa5152a3..1f0253bfe0a 100644 --- a/arch/ia64/sn/pci/pcibr/pcibr_ate.c +++ b/arch/ia64/sn/pci/pcibr/pcibr_ate.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2001-2004 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 2001-2006 Silicon Graphics, Inc. All rights reserved. */ #include <linux/types.h> @@ -12,7 +12,7 @@ #include <asm/sn/pcibus_provider_defs.h> #include <asm/sn/pcidev.h> -int pcibr_invalidate_ate = 0; /* by default don't invalidate ATE on free */ +int pcibr_invalidate_ate; /* by default don't invalidate ATE on free */ /* * mark_ate: Mark the ate as either free or inuse. @@ -20,14 +20,12 @@ int pcibr_invalidate_ate = 0; /* by default don't invalidate ATE on free */ static void mark_ate(struct ate_resource *ate_resource, int start, int number, u64 value) { - u64 *ate = ate_resource->ate; int index; int length = 0; for (index = start; length < number; index++, length++) ate[index] = value; - } /* @@ -37,7 +35,6 @@ static void mark_ate(struct ate_resource *ate_resource, int start, int number, static int find_free_ate(struct ate_resource *ate_resource, int start, int count) { - u64 *ate = ate_resource->ate; int index; int start_free; @@ -70,12 +67,10 @@ static int find_free_ate(struct ate_resource *ate_resource, int start, static inline void free_ate_resource(struct ate_resource *ate_resource, int start) { - mark_ate(ate_resource, start, ate_resource->ate[start], 0); if ((ate_resource->lowest_free_index > start) || (ate_resource->lowest_free_index < 0)) ate_resource->lowest_free_index = start; - } /* @@ -84,7 +79,6 @@ static inline void free_ate_resource(struct ate_resource *ate_resource, static inline int alloc_ate_resource(struct ate_resource *ate_resource, int ate_needed) { - int start_index; /* @@ -118,19 +112,12 @@ static inline int alloc_ate_resource(struct ate_resource *ate_resource, */ int pcibr_ate_alloc(struct pcibus_info *pcibus_info, int count) { - int status = 0; - u64 flag; + int status; + unsigned long flags; - flag = pcibr_lock(pcibus_info); + spin_lock_irqsave(&pcibus_info->pbi_lock, flags); status = alloc_ate_resource(&pcibus_info->pbi_int_ate_resource, count); - - if (status < 0) { - /* Failed to allocate */ - pcibr_unlock(pcibus_info, flag); - return -1; - } - - pcibr_unlock(pcibus_info, flag); + spin_unlock_irqrestore(&pcibus_info->pbi_lock, flags); return status; } @@ -182,7 +169,7 @@ void pcibr_ate_free(struct pcibus_info *pcibus_info, int index) ate_write(pcibus_info, index, count, (ate & ~PCI32_ATE_V)); } - flags = pcibr_lock(pcibus_info); + spin_lock_irqsave(&pcibus_info->pbi_lock, flags); free_ate_resource(&pcibus_info->pbi_int_ate_resource, index); - pcibr_unlock(pcibus_info, flags); + spin_unlock_irqrestore(&pcibus_info->pbi_lock, flags); } diff --git a/arch/ia64/sn/pci/pcibr/pcibr_dma.c b/arch/ia64/sn/pci/pcibr/pcibr_dma.c index 54ce5b7ceed..9f86bb6519a 100644 --- a/arch/ia64/sn/pci/pcibr/pcibr_dma.c +++ b/arch/ia64/sn/pci/pcibr/pcibr_dma.c @@ -137,14 +137,12 @@ pcibr_dmatrans_direct64(struct pcidev_info * info, u64 paddr, pci_addr |= PCI64_ATTR_VIRTUAL; return pci_addr; - } static dma_addr_t pcibr_dmatrans_direct32(struct pcidev_info * info, u64 paddr, size_t req_size, u64 flags) { - struct pcidev_info *pcidev_info = info->pdi_host_pcidev_info; struct pcibus_info *pcibus_info = (struct pcibus_info *)pcidev_info-> pdi_pcibus_info; @@ -171,7 +169,6 @@ pcibr_dmatrans_direct32(struct pcidev_info * info, } return PCI32_DIRECT_BASE | offset; - } /* @@ -218,9 +215,8 @@ void sn_dma_flush(u64 addr) u64 flags; u64 itte; struct hubdev_info *hubinfo; - volatile struct sn_flush_device_kernel *p; - volatile struct sn_flush_device_common *common; - + struct sn_flush_device_kernel *p; + struct sn_flush_device_common *common; struct sn_flush_nasid_entry *flush_nasid_list; if (!sn_ioif_inited) @@ -310,8 +306,7 @@ void sn_dma_flush(u64 addr) (common->sfdl_slot - 1)); } } else { - spin_lock_irqsave((spinlock_t *)&p->sfdl_flush_lock, - flags); + spin_lock_irqsave(&p->sfdl_flush_lock, flags); *common->sfdl_flush_addr = 0; /* force an interrupt. */ @@ -322,8 +317,7 @@ void sn_dma_flush(u64 addr) cpu_relax(); /* okay, everything is synched up. */ - spin_unlock_irqrestore((spinlock_t *)&p->sfdl_flush_lock, - flags); + spin_unlock_irqrestore(&p->sfdl_flush_lock, flags); } return; } diff --git a/arch/ia64/sn/pci/pcibr/pcibr_provider.c b/arch/ia64/sn/pci/pcibr/pcibr_provider.c index 2fac27049bf..98f716bd92f 100644 --- a/arch/ia64/sn/pci/pcibr/pcibr_provider.c +++ b/arch/ia64/sn/pci/pcibr/pcibr_provider.c @@ -163,9 +163,12 @@ pcibr_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont /* Setup the PMU ATE map */ soft->pbi_int_ate_resource.lowest_free_index = 0; soft->pbi_int_ate_resource.ate = - kmalloc(soft->pbi_int_ate_size * sizeof(u64), GFP_KERNEL); - memset(soft->pbi_int_ate_resource.ate, 0, - (soft->pbi_int_ate_size * sizeof(u64))); + kzalloc(soft->pbi_int_ate_size * sizeof(u64), GFP_KERNEL); + + if (!soft->pbi_int_ate_resource.ate) { + kfree(soft); + return NULL; + } if (prom_bussoft->bs_asic_type == PCIIO_ASIC_TYPE_TIOCP) { /* TIO PCI Bridge: find nearest node with CPUs */ diff --git a/arch/ia64/sn/pci/tioce_provider.c b/arch/ia64/sn/pci/tioce_provider.c index e52831ed93e..fa073cc4b56 100644 --- a/arch/ia64/sn/pci/tioce_provider.c +++ b/arch/ia64/sn/pci/tioce_provider.c @@ -15,6 +15,124 @@ #include <asm/sn/pcidev.h> #include <asm/sn/pcibus_provider_defs.h> #include <asm/sn/tioce_provider.h> +#include <asm/sn/sn2/sn_hwperf.h> + +/* + * 1/26/2006 + * + * WAR for SGI PV 944642. For revA TIOCE, need to use the following recipe + * (taken from the above PV) before and after accessing tioce internal MMR's + * to avoid tioce lockups. + * + * The recipe as taken from the PV: + * + * if(mmr address < 0x45000) { + * if(mmr address == 0 or 0x80) + * mmr wrt or read address 0xc0 + * else if(mmr address == 0x148 or 0x200) + * mmr wrt or read address 0x28 + * else + * mmr wrt or read address 0x158 + * + * do desired mmr access (rd or wrt) + * + * if(mmr address == 0x100) + * mmr wrt or read address 0x38 + * mmr wrt or read address 0xb050 + * } else + * do desired mmr access + * + * According to hw, we can use reads instead of writes to the above addres + * + * Note this WAR can only to be used for accessing internal MMR's in the + * TIOCE Coretalk Address Range 0x0 - 0x07ff_ffff. This includes the + * "Local CE Registers and Memories" and "PCI Compatible Config Space" address + * spaces from table 2-1 of the "CE Programmer's Reference Overview" document. + * + * All registers defined in struct tioce will meet that criteria. + */ + +static void inline +tioce_mmr_war_pre(struct tioce_kernel *kern, void *mmr_addr) +{ + u64 mmr_base; + u64 mmr_offset; + + if (kern->ce_common->ce_rev != TIOCE_REV_A) + return; + + mmr_base = kern->ce_common->ce_pcibus.bs_base; + mmr_offset = (u64)mmr_addr - mmr_base; + + if (mmr_offset < 0x45000) { + u64 mmr_war_offset; + + if (mmr_offset == 0 || mmr_offset == 0x80) + mmr_war_offset = 0xc0; + else if (mmr_offset == 0x148 || mmr_offset == 0x200) + mmr_war_offset = 0x28; + else + mmr_war_offset = 0x158; + + readq_relaxed((void *)(mmr_base + mmr_war_offset)); + } +} + +static void inline +tioce_mmr_war_post(struct tioce_kernel *kern, void *mmr_addr) +{ + u64 mmr_base; + u64 mmr_offset; + + if (kern->ce_common->ce_rev != TIOCE_REV_A) + return; + + mmr_base = kern->ce_common->ce_pcibus.bs_base; + mmr_offset = (u64)mmr_addr - mmr_base; + + if (mmr_offset < 0x45000) { + if (mmr_offset == 0x100) + readq_relaxed((void *)(mmr_base + 0x38)); + readq_relaxed((void *)(mmr_base + 0xb050)); + } +} + +/* load mmr contents into a variable */ +#define tioce_mmr_load(kern, mmrp, varp) do {\ + tioce_mmr_war_pre(kern, mmrp); \ + *(varp) = readq_relaxed(mmrp); \ + tioce_mmr_war_post(kern, mmrp); \ +} while (0) + +/* store variable contents into mmr */ +#define tioce_mmr_store(kern, mmrp, varp) do {\ + tioce_mmr_war_pre(kern, mmrp); \ + writeq(*varp, mmrp); \ + tioce_mmr_war_post(kern, mmrp); \ +} while (0) + +/* store immediate value into mmr */ +#define tioce_mmr_storei(kern, mmrp, val) do {\ + tioce_mmr_war_pre(kern, mmrp); \ + writeq(val, mmrp); \ + tioce_mmr_war_post(kern, mmrp); \ +} while (0) + +/* set bits (immediate value) into mmr */ +#define tioce_mmr_seti(kern, mmrp, bits) do {\ + u64 tmp; \ + tioce_mmr_load(kern, mmrp, &tmp); \ + tmp |= (bits); \ + tioce_mmr_store(kern, mmrp, &tmp); \ +} while (0) + +/* clear bits (immediate value) into mmr */ +#define tioce_mmr_clri(kern, mmrp, bits) do { \ + u64 tmp; \ + tioce_mmr_load(kern, mmrp, &tmp); \ + tmp &= ~(bits); \ + tioce_mmr_store(kern, mmrp, &tmp); \ +} while (0) /** * Bus address ranges for the 5 flavors of TIOCE DMA @@ -62,9 +180,9 @@ #define TIOCE_ATE_M40 2 #define TIOCE_ATE_M40S 3 -#define KB(x) ((x) << 10) -#define MB(x) ((x) << 20) -#define GB(x) ((x) << 30) +#define KB(x) ((u64)(x) << 10) +#define MB(x) ((u64)(x) << 20) +#define GB(x) ((u64)(x) << 30) /** * tioce_dma_d64 - create a DMA mapping using 64-bit direct mode @@ -151,7 +269,7 @@ tioce_alloc_map(struct tioce_kernel *ce_kern, int type, int port, int last; int entries; int nates; - int pagesize; + u64 pagesize; u64 *ate_shadow; u64 *ate_reg; u64 addr; @@ -228,7 +346,7 @@ tioce_alloc_map(struct tioce_kernel *ce_kern, int type, int port, ate = ATE_MAKE(addr, pagesize); ate_shadow[i + j] = ate; - writeq(ate, &ate_reg[i + j]); + tioce_mmr_storei(ce_kern, &ate_reg[i + j], ate); addr += pagesize; } @@ -272,7 +390,8 @@ tioce_dma_d32(struct pci_dev *pdev, u64 ct_addr) u64 tmp; ce_kern->ce_port[port].dirmap_shadow = ct_upper; - writeq(ct_upper, &ce_mmr->ce_ure_dir_map[port]); + tioce_mmr_storei(ce_kern, &ce_mmr->ce_ure_dir_map[port], + ct_upper); tmp = ce_mmr->ce_ure_dir_map[port]; dma_ok = 1; } else @@ -344,7 +463,8 @@ tioce_dma_unmap(struct pci_dev *pdev, dma_addr_t bus_addr, int dir) if (TIOCE_D32_ADDR(bus_addr)) { if (--ce_kern->ce_port[port].dirmap_refcnt == 0) { ce_kern->ce_port[port].dirmap_shadow = 0; - writeq(0, &ce_mmr->ce_ure_dir_map[port]); + tioce_mmr_storei(ce_kern, &ce_mmr->ce_ure_dir_map[port], + 0); } } else { struct tioce_dmamap *map; @@ -365,7 +485,7 @@ tioce_dma_unmap(struct pci_dev *pdev, dma_addr_t bus_addr, int dir) } else if (--map->refcnt == 0) { for (i = 0; i < map->ate_count; i++) { map->ate_shadow[i] = 0; - map->ate_hw[i] = 0; + tioce_mmr_storei(ce_kern, &map->ate_hw[i], 0); } list_del(&map->ce_dmamap_list); @@ -486,7 +606,7 @@ tioce_do_dma_map(struct pci_dev *pdev, u64 paddr, size_t byte_count, spin_unlock_irqrestore(&ce_kern->ce_lock, flags); dma_map_done: - if (mapaddr & barrier) + if (mapaddr && barrier) mapaddr = tioce_dma_barrier(mapaddr, 1); return mapaddr; @@ -541,17 +661,61 @@ tioce_error_intr_handler(int irq, void *arg, struct pt_regs *pt) soft->ce_pcibus.bs_persist_segment, soft->ce_pcibus.bs_persist_busnum, 0, 0, 0, 0, 0); + if (ret_stuff.v0) + panic("tioce_error_intr_handler: Fatal TIOCE error"); + return IRQ_HANDLED; } /** + * tioce_reserve_m32 - reserve M32 ate's for the indicated address range + * @tioce_kernel: TIOCE context to reserve ate's for + * @base: starting bus address to reserve + * @limit: last bus address to reserve + * + * If base/limit falls within the range of bus space mapped through the + * M32 space, reserve the resources corresponding to the range. + */ +static void +tioce_reserve_m32(struct tioce_kernel *ce_kern, u64 base, u64 limit) +{ + int ate_index, last_ate, ps; + struct tioce *ce_mmr; + + if (!TIOCE_M32_ADDR(base)) + return; + + ce_mmr = (struct tioce *)ce_kern->ce_common->ce_pcibus.bs_base; + ps = ce_kern->ce_ate3240_pagesize; + ate_index = ATE_PAGE(base, ps); + last_ate = ate_index + ATE_NPAGES(base, limit-base+1, ps) - 1; + + if (ate_index < 64) + ate_index = 64; + + while (ate_index <= last_ate) { + u64 ate; + + ate = ATE_MAKE(0xdeadbeef, ps); + ce_kern->ce_ate3240_shadow[ate_index] = ate; + tioce_mmr_storei(ce_kern, &ce_mmr->ce_ure_ate3240[ate_index], + ate); + ate_index++; + } +} + +/** * tioce_kern_init - init kernel structures related to a given TIOCE * @tioce_common: ptr to a cached tioce_common struct that originated in prom - */ static struct tioce_kernel * + */ +static struct tioce_kernel * tioce_kern_init(struct tioce_common *tioce_common) { int i; + int ps; + int dev; u32 tmp; + unsigned int seg, bus; struct tioce *tioce_mmr; struct tioce_kernel *tioce_kern; @@ -572,9 +736,10 @@ tioce_kern_init(struct tioce_common *tioce_common) * here to use pci_read_config_xxx() so use the raw_pci_ops vector. */ - raw_pci_ops->read(tioce_common->ce_pcibus.bs_persist_segment, - tioce_common->ce_pcibus.bs_persist_busnum, - PCI_DEVFN(2, 0), PCI_SECONDARY_BUS, 1, &tmp); + seg = tioce_common->ce_pcibus.bs_persist_segment; + bus = tioce_common->ce_pcibus.bs_persist_busnum; + + raw_pci_ops->read(seg, bus, PCI_DEVFN(2, 0), PCI_SECONDARY_BUS, 1,&tmp); tioce_kern->ce_port1_secondary = (u8) tmp; /* @@ -583,18 +748,76 @@ tioce_kern_init(struct tioce_common *tioce_common) */ tioce_mmr = (struct tioce *)tioce_common->ce_pcibus.bs_base; - __sn_clrq_relaxed(&tioce_mmr->ce_ure_page_map, CE_URE_PAGESIZE_MASK); - __sn_setq_relaxed(&tioce_mmr->ce_ure_page_map, CE_URE_256K_PAGESIZE); - tioce_kern->ce_ate3240_pagesize = KB(256); + tioce_mmr_clri(tioce_kern, &tioce_mmr->ce_ure_page_map, + CE_URE_PAGESIZE_MASK); + tioce_mmr_seti(tioce_kern, &tioce_mmr->ce_ure_page_map, + CE_URE_256K_PAGESIZE); + ps = tioce_kern->ce_ate3240_pagesize = KB(256); for (i = 0; i < TIOCE_NUM_M40_ATES; i++) { tioce_kern->ce_ate40_shadow[i] = 0; - writeq(0, &tioce_mmr->ce_ure_ate40[i]); + tioce_mmr_storei(tioce_kern, &tioce_mmr->ce_ure_ate40[i], 0); } for (i = 0; i < TIOCE_NUM_M3240_ATES; i++) { tioce_kern->ce_ate3240_shadow[i] = 0; - writeq(0, &tioce_mmr->ce_ure_ate3240[i]); + tioce_mmr_storei(tioce_kern, &tioce_mmr->ce_ure_ate3240[i], 0); + } + + /* + * Reserve ATE's corresponding to reserved address ranges. These + * include: + * + * Memory space covered by each PPB mem base/limit register + * Memory space covered by each PPB prefetch base/limit register + * + * These bus ranges are for pio (downstream) traffic only, and so + * cannot be used for DMA. + */ + + for (dev = 1; dev <= 2; dev++) { + u64 base, limit; + + /* mem base/limit */ + + raw_pci_ops->read(seg, bus, PCI_DEVFN(dev, 0), + PCI_MEMORY_BASE, 2, &tmp); + base = (u64)tmp << 16; + + raw_pci_ops->read(seg, bus, PCI_DEVFN(dev, 0), + PCI_MEMORY_LIMIT, 2, &tmp); + limit = (u64)tmp << 16; + limit |= 0xfffffUL; + + if (base < limit) + tioce_reserve_m32(tioce_kern, base, limit); + + /* + * prefetch mem base/limit. The tioce ppb's have 64-bit + * decoders, so read the upper portions w/o checking the + * attributes. + */ + + raw_pci_ops->read(seg, bus, PCI_DEVFN(dev, 0), + PCI_PREF_MEMORY_BASE, 2, &tmp); + base = ((u64)tmp & PCI_PREF_RANGE_MASK) << 16; + + raw_pci_ops->read(seg, bus, PCI_DEVFN(dev, 0), + PCI_PREF_BASE_UPPER32, 4, &tmp); + base |= (u64)tmp << 32; + + raw_pci_ops->read(seg, bus, PCI_DEVFN(dev, 0), + PCI_PREF_MEMORY_LIMIT, 2, &tmp); + + limit = ((u64)tmp & PCI_PREF_RANGE_MASK) << 16; + limit |= 0xfffffUL; + + raw_pci_ops->read(seg, bus, PCI_DEVFN(dev, 0), + PCI_PREF_LIMIT_UPPER32, 4, &tmp); + limit |= (u64)tmp << 32; + + if ((base < limit) && TIOCE_M32_ADDR(base)) + tioce_reserve_m32(tioce_kern, base, limit); } return tioce_kern; @@ -614,6 +837,7 @@ tioce_force_interrupt(struct sn_irq_info *sn_irq_info) { struct pcidev_info *pcidev_info; struct tioce_common *ce_common; + struct tioce_kernel *ce_kern; struct tioce *ce_mmr; u64 force_int_val; @@ -629,6 +853,29 @@ tioce_force_interrupt(struct sn_irq_info *sn_irq_info) ce_common = (struct tioce_common *)pcidev_info->pdi_pcibus_info; ce_mmr = (struct tioce *)ce_common->ce_pcibus.bs_base; + ce_kern = (struct tioce_kernel *)ce_common->ce_kernel_private; + + /* + * TIOCE Rev A workaround (PV 945826), force an interrupt by writing + * the TIO_INTx register directly (1/26/2006) + */ + if (ce_common->ce_rev == TIOCE_REV_A) { + u64 int_bit_mask = (1ULL << sn_irq_info->irq_int_bit); + u64 status; + + tioce_mmr_load(ce_kern, &ce_mmr->ce_adm_int_status, &status); + if (status & int_bit_mask) { + u64 force_irq = (1 << 8) | sn_irq_info->irq_irq; + u64 ctalk = sn_irq_info->irq_xtalkaddr; + u64 nasid, offset; + + nasid = (ctalk & CTALK_NASID_MASK) >> CTALK_NASID_SHFT; + offset = (ctalk & CTALK_NODE_OFFSET); + HUB_S(TIO_IOSPACE_ADDR(nasid, offset), force_irq); + } + + return; + } /* * irq_int_bit is originally set up by prom, and holds the interrupt @@ -666,7 +913,7 @@ tioce_force_interrupt(struct sn_irq_info *sn_irq_info) default: return; } - writeq(force_int_val, &ce_mmr->ce_adm_force_int); + tioce_mmr_storei(ce_kern, &ce_mmr->ce_adm_force_int, force_int_val); } /** @@ -685,6 +932,7 @@ tioce_target_interrupt(struct sn_irq_info *sn_irq_info) { struct pcidev_info *pcidev_info; struct tioce_common *ce_common; + struct tioce_kernel *ce_kern; struct tioce *ce_mmr; int bit; u64 vector; @@ -695,14 +943,15 @@ tioce_target_interrupt(struct sn_irq_info *sn_irq_info) ce_common = (struct tioce_common *)pcidev_info->pdi_pcibus_info; ce_mmr = (struct tioce *)ce_common->ce_pcibus.bs_base; + ce_kern = (struct tioce_kernel *)ce_common->ce_kernel_private; bit = sn_irq_info->irq_int_bit; - __sn_setq_relaxed(&ce_mmr->ce_adm_int_mask, (1UL << bit)); + tioce_mmr_seti(ce_kern, &ce_mmr->ce_adm_int_mask, (1UL << bit)); vector = (u64)sn_irq_info->irq_irq << INTR_VECTOR_SHFT; vector |= sn_irq_info->irq_xtalkaddr; - writeq(vector, &ce_mmr->ce_adm_int_dest[bit]); - __sn_clrq_relaxed(&ce_mmr->ce_adm_int_mask, (1UL << bit)); + tioce_mmr_storei(ce_kern, &ce_mmr->ce_adm_int_dest[bit], vector); + tioce_mmr_clri(ce_kern, &ce_mmr->ce_adm_int_mask, (1UL << bit)); tioce_force_interrupt(sn_irq_info); } @@ -721,7 +970,11 @@ tioce_target_interrupt(struct sn_irq_info *sn_irq_info) static void * tioce_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *controller) { + int my_nasid; + cnodeid_t my_cnode, mem_cnode; struct tioce_common *tioce_common; + struct tioce_kernel *tioce_kern; + struct tioce *tioce_mmr; /* * Allocate kernel bus soft and copy from prom. @@ -734,11 +987,23 @@ tioce_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont memcpy(tioce_common, prom_bussoft, sizeof(struct tioce_common)); tioce_common->ce_pcibus.bs_base |= __IA64_UNCACHED_OFFSET; - if (tioce_kern_init(tioce_common) == NULL) { + tioce_kern = tioce_kern_init(tioce_common); + if (tioce_kern == NULL) { kfree(tioce_common); return NULL; } + /* + * Clear out any transient errors before registering the error + * interrupt handler. + */ + + tioce_mmr = (struct tioce *)tioce_common->ce_pcibus.bs_base; + tioce_mmr_seti(tioce_kern, &tioce_mmr->ce_adm_int_status_alias, ~0ULL); + tioce_mmr_seti(tioce_kern, &tioce_mmr->ce_adm_error_summary_alias, + ~0ULL); + tioce_mmr_seti(tioce_kern, &tioce_mmr->ce_dre_comp_err_addr, ~0ULL); + if (request_irq(SGI_PCIASIC_ERROR, tioce_error_intr_handler, SA_SHIRQ, "TIOCE error", (void *)tioce_common)) @@ -750,6 +1015,21 @@ tioce_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont tioce_common->ce_pcibus.bs_persist_segment, tioce_common->ce_pcibus.bs_persist_busnum); + /* + * identify closest nasid for memory allocations + */ + + my_nasid = NASID_GET(tioce_common->ce_pcibus.bs_base); + my_cnode = nasid_to_cnodeid(my_nasid); + + if (sn_hwperf_get_nearest_node(my_cnode, &mem_cnode, NULL) < 0) { + printk(KERN_WARNING "tioce_bus_fixup: failed to find " + "closest node with MEM to TIO node %d\n", my_cnode); + mem_cnode = (cnodeid_t)-1; /* use any node */ + } + + controller->node = mem_cnode; + return tioce_common; } |