No include/asm-ia64/mmzone.h in 2.4.16-ia64-011128
I tried compiling 2.4.16-ia64-011128 as type SGI-SN1. It broke at include/linux/mmzone.h:138 trying to include asm/mmzone.h which does not exist on ia64. Is ia64-011128 supposed to compile sn1/sn2 out of the box or are there extra patches required?
Also this does not look right in arch/ia64/Makefile. GENERIC and SGI_SN do not agree on the core files and subirs, does anybody ever use generic?
ifdef CONFIG_IA64_GENERIC
CORE_FILES := arch/$(ARCH)/hp/hp.a \
arch/$(ARCH)/sn/sn.o \
arch/$(ARCH)/dig/dig.a \
arch/$(ARCH)/sn/io/sgiio.o \
$(CORE_FILES)
SUBDIRS := arch/$(ARCH)/hp \
arch/$(ARCH)/sn/sn1 \
arch/$(ARCH)/sn \
arch/$(ARCH)/dig \
arch/$(ARCH)/sn/io \
$(SUBDIRS)else # !GENERIC
....
ifdef CONFIG_IA64_SGI_SN
CFLAGS += -DBRINGUP
SUBDIRS := arch/$(ARCH)/sn/kernel \
arch/$(ARCH)/sn/io \
arch/$(ARCH)/sn/fakeprom \
$(SUBDIRS)
CORE_FILES := arch/$(ARCH)/sn/kernel/sn.o \
arch/$(ARCH)/sn/io/sgiio.o \
$(CORE_FILES)endif
At the subject of arch/ia64/Makefile, it tests for CONFIG_IA64_SOFTSDV but nothing else in the kernel refers to that variable, it cannot be set.
Quoted 4 lines · Keith Owens
I tried compiling 2.4.16-ia64-011128 as type SGI-SN1. It broke at
include/linux/mmzone.h:138trying to includeasm/mmzone.hwhich does not exist on ia64. Is ia64-011128 supposed to compile sn1/sn2 out of the box or are there extra patches required?
No, as you've discovered, it's missing mmzone.h.
Quoted 4 lines
Also this does not look right in
arch/ia64/Makefile. GENERIC andSGI_SNdo not agree on the core files and subirs, does anybody ever use generic?
I don't think it's currently used, but it's still potentially a good idea down the road. We've probably let it go a little stale right now.
John
Quoted 7 lines · John Hesterberg
Quoted 4 lines · Keith Owens
I tried compiling 2.4.16-ia64-011128 as type SGI-SN1. It broke at
include/linux/mmzone.h:138trying to includeasm/mmzone.hwhich does not exist on ia64. Is ia64-011128 supposed to compile sn1/sn2 out of the box or are there extra patches required?No, as you've discovered, it's missing
mmzone.h.
Is that intentional or a mistake? I have converted ia64 to kbuild 2.5, ready to send the patch to Linus. But I could only test platform DIG, all of the other platforms (HP_SIM, SGI_SN1, SGI_SN2, GENERIC) got compile errors, even under kbuild 2.4.
I don't have time to fix up the platform code. Unless somebody can supply patches against ia64-011128 to make all the platforms compile, I will have to hope that the conversion for the failing platforms is correct. OTOH, they don't work now, so why worry ;).
Quoted 11 lines · Keith Owens
Quoted 7 lines · John Hesterberg
Quoted 4 lines · Keith Owens
I tried compiling 2.4.16-ia64-011128 as type SGI-SN1. It broke at
include/linux/mmzone.h:138trying to includeasm/mmzone.hwhich does not exist on ia64. Is ia64-011128 supposed to compile sn1/sn2 out of the box or are there extra patches required?No, as you've discovered, it's missing
mmzone.h.Is that intentional or a mistake?
Neither, really. We'd like it to be at least buildable, but right now our SN specific code is tied to changes that are not yet in the non-SN specific code. That makes it hard to come up with a version of the code that is reasonably current and still buildable.
I think I have a patch for you to use, but the machine it's on just went belly up (grrr), so I'll have to wait until I get in to work tomorrow to recover the machine. If you're ready to send to Linus, you'll have to decide if you want to wait or not.
John
Quoted 18 lines · John Hesterberg
Quoted 11 lines · Keith Owens
Quoted 7 lines · John Hesterberg
Quoted 4 lines · Keith Owens
I tried compiling 2.4.16-ia64-011128 as type SGI-SN1. It broke at
include/linux/mmzone.h:138trying to includeasm/mmzone.hwhich does not exist on ia64. Is ia64-011128 supposed to compile sn1/sn2 out of the box or are there extra patches required?No, as you've discovered, it's missing
mmzone.h.Is that intentional or a mistake?
Neither, really. We'd like it to be at least buildable, but right now our SN specific code is tied to changes that are not yet in the non-SN specific code. That makes it hard to come up with a version of the code that is reasonably current and still buildable.
Quoted 4 lines
I think I have a patch for you to use, but the machine it's on just went belly up (grrr), so I'll have to wait until I get in to work tomorrow to recover the machine. If you're ready to send to Linus, you'll have to decide if you want to wait or not.
Here's the patch. It builds for SN1 and DIG, I didn't try anything else. It boots DIG, but probably won't boot SN1.
Some, most, or all of this patch will change before it makes it into David's ia64 patch, so don't base other changes on it. This is for test building only. It's as unofficial as it gets.
To build for SN1, use arch/ia64/sn/configs/sn1/defconfig-sn1-mp, do make oldconfig, and take the default values for everything.
John
linux/arch/ia64/kernel/smp.c +13 -0
@@ -72,6 +72,11 @@#define IPI_CALL_FUNC 0#define IPI_CPU_STOP 1 +#ifdef CONFIG_IA64_SGI_SN1 +#define IPI_FLUSH_TLB 3 +#endif + +static void stop_this_cpu (void ) {
@@ -315,6 +320,14 @@send_IPI_allbutself (IPI_CPU_STOP ); smp_num_cpus =1 ; } + +#ifdef CONFIG_IA64_SGI_SN1 +void +smp_send_flush_tlb (void ) +{ +send_IPI_allbutself (IPI_FLUSH_TLB ); +} +#endif int __init setup_profiling_timer (unsigned int multiplier)
linux/arch/ia64/sn/io/klconflib.c +1 -4
@@ -302,10 +302,7 @@int get_cpu_slice (cpuid_t cpu) { -klcpu_t *acpu; -if ((acpu =get_cpuinfo (cpu)) =NULL ) -return -1 ; -return acpu->cpu_info.physid; +return -1 ; }
linux/arch/ia64/sn/kernel/setup.c +19 -19
@@ -40,7 +40,6 @@#include <asm/sn/sn2/shub.h> #endif -void platform_smp_callin (int );long sn_rtc_cycles_per_second;/*
@@ -249,13 +248,6 @@for (cnode=0 ; cnode<num_compact_nodes; cnode++)memcpy (nodepdaindr[cnode]->pernode_pdaindr, nodepdaindr,sizeof (nodepdaindr)); - -/* -* For the bootcpu, we do this here. All other cpus will make the -* call as part of smp_callin initialization. -*/ -platform_smp_callin (smp_processor_id ()); -/* * Set up IO related platform-dependent nodepda fields. * The following routine actually sets up the hubinfo struct
@@ -265,21 +257,32 @@init_platform_nodepda (nodepdaindr[cnode], cnode); } -/* -* platform_smp_callin +/** +* sn_cpu_init - initialize per-cpu data areas +* @cpuid: cpuid of the caller * * Called during cpu initialization on each cpu as it starts. * Currently, initializes the per-cpu data area for SNIA. -* Also setup up a few fields in the nodepda. +* Also sets up a few fields in the nodepda. Also known as +* platform_cpu_init() by the ia64 machvec code. */ void __init -platform_smp_callin (int cpuid) +sn_cpu_init (void ) { +int cpuid;int cpuphyid;int nasid;int slice;int cnode; +/* +* The boot cpu makes this call again after platform initialization is +* complete. +*/ +if (nodepdaindr[0 ] =NULL ) +return ; + + cpuid =smp_processor_id (); cpuphyid = ((ia64_get_lid () >>16 ) &0xffff ); nasid =cpu_physical_id_to_nasid (cpuphyid); cnode =nasid_to_cnodeid (nasid);
@@ -292,10 +295,6 @@ pda.hb_state =0 ; pda.idle_flag =0 ; -if (nodepda->node_num_cpus =0 ) - nodepda->node_first_cpu = cpuid; - nodepda->node_num_cpus++; -#ifdef CONFIG_IA64_SGI_SN1 {int synergy;
@@ -305,14 +304,15 @@#endif #ifdef CONFIG_IA64_SGI_SN2 +/* -* We must use different memory allocators for first cpu (bootmem allocator) -* than for the other cpus (regular allocator). +* We must use different memory allocators for first cpu (bootmem +* allocator) than for the other cpus (regular allocator). */ if (cpuid =0 ) irqpdaindr[cpuid] =alloc_bootmem_node (NODE_DATA (cpuid_to_cnodeid (cpuid)),sizeof (irqpda_t));else - irqpdaindr[cpuid] =page_address (alloc_pages_node (numa_node_id (),GFP_KERNEL ,get_order (sizeof (irqpda_t)))); + irqpdaindr[cpuid] =page_address (alloc_pages_node (local_cnodeid (),GFP_KERNEL ,get_order (sizeof (irqpda_t))));memset (irqpdaindr[cpuid],0 ,sizeof (irqpda_t)); pda.p_irqpda = irqpdaindr[cpuid];#endif
linux/drivers/net/ioc3-eth.c +273 -335
896 lines not on this page · raw message
linux/include/asm-ia64/acpi-ext.h +1 -1
@@ -297,7 +297,7 @@extern int pxm_to_nid_map[MAX_PXM_DOMAINS ];/* _PXM to logical node ID map */ extern int nid_to_pxm_map[PLAT_MAX_COMPACT_NODES ];/* logical node ID to _PXM map */ -extern int numnodes;/* total number of nodes in system */ +extern int numnodes;/* total number of nodes in system */ extern int num_memory_chunks;/* total number of memory chunks */ /*
linux/include/asm-ia64/sn/mmzone_temp.h +4 -0
@@ -22,6 +22,10 @@#error "Unknown architecture"#endif +/* +* Temporary hack +*/ +#define num_compact_nodes numnodes/* * General Concepts:
linux/drivers/char/sgi-l1_protocol.c new file +2577 -0
1,966 lines not on this page · raw message
linux/drivers/sgi/sn/fetchop.c new file +348 -0
348 lines not on this page · raw message
linux/drivers/sgi/sn/hires_clock.c new file +119 -0
119 lines not on this page · raw message
linux/drivers/sgi/sn/Makefile new file +20 -0
Lines of this file are not shown. View the raw message.
linux/include/asm-ia64/mmzone.h new file +17 -0
Lines of this file are not shown. View the raw message.
The rest of this diff is not shown. View the raw message.
Part of the diff is not shown here. View the raw message.