17c478bd9Sstevel@tonic-gate /* 27c478bd9Sstevel@tonic-gate * CDDL HEADER START 37c478bd9Sstevel@tonic-gate * 47c478bd9Sstevel@tonic-gate * The contents of this file are subject to the terms of the 5100b72f4Sandrei * Common Development and Distribution License (the "License"). 6100b72f4Sandrei * You may not use this file except in compliance with the License. 77c478bd9Sstevel@tonic-gate * 87c478bd9Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 97c478bd9Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 107c478bd9Sstevel@tonic-gate * See the License for the specific language governing permissions 117c478bd9Sstevel@tonic-gate * and limitations under the License. 127c478bd9Sstevel@tonic-gate * 137c478bd9Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 147c478bd9Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 157c478bd9Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 167c478bd9Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 177c478bd9Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 187c478bd9Sstevel@tonic-gate * 197c478bd9Sstevel@tonic-gate * CDDL HEADER END 207c478bd9Sstevel@tonic-gate */ 217c478bd9Sstevel@tonic-gate /* 22fb2f18f8Sesaxe * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 237c478bd9Sstevel@tonic-gate * Use is subject to license terms. 247c478bd9Sstevel@tonic-gate */ 257c478bd9Sstevel@tonic-gate 267c478bd9Sstevel@tonic-gate #pragma ident "%Z%%M% %I% %E% SMI" 277c478bd9Sstevel@tonic-gate 287c478bd9Sstevel@tonic-gate #include <sys/types.h> 297c478bd9Sstevel@tonic-gate #include <sys/t_lock.h> 307c478bd9Sstevel@tonic-gate #include <sys/param.h> 317c478bd9Sstevel@tonic-gate #include <sys/sysmacros.h> 327c478bd9Sstevel@tonic-gate #include <sys/signal.h> 337c478bd9Sstevel@tonic-gate #include <sys/systm.h> 347c478bd9Sstevel@tonic-gate #include <sys/user.h> 357c478bd9Sstevel@tonic-gate #include <sys/mman.h> 367c478bd9Sstevel@tonic-gate #include <sys/vm.h> 377c478bd9Sstevel@tonic-gate #include <sys/conf.h> 387c478bd9Sstevel@tonic-gate #include <sys/avintr.h> 397c478bd9Sstevel@tonic-gate #include <sys/autoconf.h> 407c478bd9Sstevel@tonic-gate #include <sys/disp.h> 417c478bd9Sstevel@tonic-gate #include <sys/class.h> 427c478bd9Sstevel@tonic-gate #include <sys/bitmap.h> 437c478bd9Sstevel@tonic-gate 447c478bd9Sstevel@tonic-gate #include <sys/privregs.h> 457c478bd9Sstevel@tonic-gate 467c478bd9Sstevel@tonic-gate #include <sys/proc.h> 477c478bd9Sstevel@tonic-gate #include <sys/buf.h> 487c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 49*ae115bc7Smrj #include <sys/mem.h> 507c478bd9Sstevel@tonic-gate #include <sys/kstat.h> 517c478bd9Sstevel@tonic-gate 527c478bd9Sstevel@tonic-gate #include <sys/reboot.h> 537c478bd9Sstevel@tonic-gate 547c478bd9Sstevel@tonic-gate #include <sys/cred.h> 557c478bd9Sstevel@tonic-gate #include <sys/vnode.h> 567c478bd9Sstevel@tonic-gate #include <sys/file.h> 577c478bd9Sstevel@tonic-gate 587c478bd9Sstevel@tonic-gate #include <sys/procfs.h> 597c478bd9Sstevel@tonic-gate 607c478bd9Sstevel@tonic-gate #include <sys/vfs.h> 617c478bd9Sstevel@tonic-gate #include <sys/cmn_err.h> 627c478bd9Sstevel@tonic-gate #include <sys/utsname.h> 637c478bd9Sstevel@tonic-gate #include <sys/debug.h> 647c478bd9Sstevel@tonic-gate #include <sys/kdi.h> 657c478bd9Sstevel@tonic-gate 667c478bd9Sstevel@tonic-gate #include <sys/dumphdr.h> 677c478bd9Sstevel@tonic-gate #include <sys/bootconf.h> 687c478bd9Sstevel@tonic-gate #include <sys/varargs.h> 697c478bd9Sstevel@tonic-gate #include <sys/promif.h> 707c478bd9Sstevel@tonic-gate #include <sys/modctl.h> /* for "procfs" hack */ 717c478bd9Sstevel@tonic-gate 727c478bd9Sstevel@tonic-gate #include <sys/sunddi.h> 737c478bd9Sstevel@tonic-gate #include <sys/sunndi.h> 747c478bd9Sstevel@tonic-gate #include <sys/ndi_impldefs.h> 757c478bd9Sstevel@tonic-gate #include <sys/ddidmareq.h> 767c478bd9Sstevel@tonic-gate #include <sys/psw.h> 777c478bd9Sstevel@tonic-gate #include <sys/regset.h> 787c478bd9Sstevel@tonic-gate #include <sys/clock.h> 797c478bd9Sstevel@tonic-gate #include <sys/pte.h> 807c478bd9Sstevel@tonic-gate #include <sys/tss.h> 817c478bd9Sstevel@tonic-gate #include <sys/stack.h> 827c478bd9Sstevel@tonic-gate #include <sys/trap.h> 837c478bd9Sstevel@tonic-gate #include <sys/fp.h> 847c478bd9Sstevel@tonic-gate #include <vm/anon.h> 857c478bd9Sstevel@tonic-gate #include <vm/as.h> 867c478bd9Sstevel@tonic-gate #include <vm/page.h> 877c478bd9Sstevel@tonic-gate #include <vm/seg.h> 887c478bd9Sstevel@tonic-gate #include <vm/seg_dev.h> 897c478bd9Sstevel@tonic-gate #include <vm/seg_kmem.h> 907c478bd9Sstevel@tonic-gate #include <vm/seg_kpm.h> 917c478bd9Sstevel@tonic-gate #include <vm/seg_map.h> 927c478bd9Sstevel@tonic-gate #include <vm/seg_vn.h> 937c478bd9Sstevel@tonic-gate #include <vm/seg_kp.h> 947c478bd9Sstevel@tonic-gate #include <sys/memnode.h> 957c478bd9Sstevel@tonic-gate #include <vm/vm_dep.h> 967c478bd9Sstevel@tonic-gate #include <sys/thread.h> 977c478bd9Sstevel@tonic-gate #include <sys/sysconf.h> 987c478bd9Sstevel@tonic-gate #include <sys/vm_machparam.h> 997c478bd9Sstevel@tonic-gate #include <sys/archsystm.h> 1007c478bd9Sstevel@tonic-gate #include <sys/machsystm.h> 1017c478bd9Sstevel@tonic-gate #include <vm/hat.h> 1027c478bd9Sstevel@tonic-gate #include <vm/hat_i86.h> 1037c478bd9Sstevel@tonic-gate #include <sys/pmem.h> 1047c478bd9Sstevel@tonic-gate #include <sys/smp_impldefs.h> 1057c478bd9Sstevel@tonic-gate #include <sys/x86_archext.h> 1067c478bd9Sstevel@tonic-gate #include <sys/segments.h> 1077c478bd9Sstevel@tonic-gate #include <sys/clconf.h> 1087c478bd9Sstevel@tonic-gate #include <sys/kobj.h> 1097c478bd9Sstevel@tonic-gate #include <sys/kobj_lex.h> 1107c478bd9Sstevel@tonic-gate #include <sys/cpc_impl.h> 111fb2f18f8Sesaxe #include <sys/pg.h> 1127c478bd9Sstevel@tonic-gate #include <sys/x86_archext.h> 1137aec1d6eScindi #include <sys/cpu_module.h> 11484ab085aSmws #include <sys/smbios.h> 115*ae115bc7Smrj #include <sys/debug_info.h> 116*ae115bc7Smrj 117*ae115bc7Smrj 118*ae115bc7Smrj #include <sys/bootinfo.h> 119*ae115bc7Smrj #include <vm/kboot_mmu.h> 1207c478bd9Sstevel@tonic-gate 1217c478bd9Sstevel@tonic-gate extern void progressbar_init(void); 1227c478bd9Sstevel@tonic-gate extern void progressbar_start(void); 1239acbbeafSnn extern void brand_init(void); 1247c478bd9Sstevel@tonic-gate 1257c478bd9Sstevel@tonic-gate /* 1267c478bd9Sstevel@tonic-gate * XXX make declaration below "static" when drivers no longer use this 1277c478bd9Sstevel@tonic-gate * interface. 1287c478bd9Sstevel@tonic-gate */ 1297c478bd9Sstevel@tonic-gate extern caddr_t p0_va; /* Virtual address for accessing physical page 0 */ 1307c478bd9Sstevel@tonic-gate 1317c478bd9Sstevel@tonic-gate /* 1327c478bd9Sstevel@tonic-gate * segkp 1337c478bd9Sstevel@tonic-gate */ 1347c478bd9Sstevel@tonic-gate extern int segkp_fromheap; 1357c478bd9Sstevel@tonic-gate 1367c478bd9Sstevel@tonic-gate static void kvm_init(void); 1377c478bd9Sstevel@tonic-gate static void startup_init(void); 1387c478bd9Sstevel@tonic-gate static void startup_memlist(void); 139*ae115bc7Smrj static void startup_kmem(void); 1407c478bd9Sstevel@tonic-gate static void startup_modules(void); 1417c478bd9Sstevel@tonic-gate static void startup_vm(void); 1427c478bd9Sstevel@tonic-gate static void startup_end(void); 1437c478bd9Sstevel@tonic-gate 1447c478bd9Sstevel@tonic-gate /* 1457c478bd9Sstevel@tonic-gate * Declare these as initialized data so we can patch them. 1467c478bd9Sstevel@tonic-gate */ 147f53ad214Skchow #ifdef __i386 148f53ad214Skchow /* 149f53ad214Skchow * Due to virtual address space limitations running in 32 bit mode, restrict 150f53ad214Skchow * the amount of physical memory configured to a max of PHYSMEM32 pages (16g). 151f53ad214Skchow * 152f53ad214Skchow * If the physical max memory size of 64g were allowed to be configured, the 153f53ad214Skchow * size of user virtual address space will be less than 1g. A limited user 154f53ad214Skchow * address space greatly reduces the range of applications that can run. 155f53ad214Skchow * 156f53ad214Skchow * If more physical memory than PHYSMEM32 is required, users should preferably 157f53ad214Skchow * run in 64 bit mode which has no virtual address space limitation issues. 158f53ad214Skchow * 159f53ad214Skchow * If 64 bit mode is not available (as in IA32) and/or more physical memory 160f53ad214Skchow * than PHYSMEM32 is required in 32 bit mode, physmem can be set to the desired 161f53ad214Skchow * value or to 0 (to configure all available memory) via eeprom(1M). kernelbase 162f53ad214Skchow * should also be carefully tuned to balance out the need of the user 163f53ad214Skchow * application while minimizing the risk of kernel heap exhaustion due to 164f53ad214Skchow * kernelbase being set too high. 165f53ad214Skchow */ 166f53ad214Skchow #define PHYSMEM32 0x400000 167f53ad214Skchow 168f53ad214Skchow pgcnt_t physmem = PHYSMEM32; 169f53ad214Skchow #else 1707c478bd9Sstevel@tonic-gate pgcnt_t physmem = 0; /* memory size in pages, patch if you want less */ 171f53ad214Skchow #endif 1727c478bd9Sstevel@tonic-gate pgcnt_t obp_pages; /* Memory used by PROM for its text and data */ 1737c478bd9Sstevel@tonic-gate 1747c478bd9Sstevel@tonic-gate char *kobj_file_buf; 1757c478bd9Sstevel@tonic-gate int kobj_file_bufsize; /* set in /etc/system */ 1767c478bd9Sstevel@tonic-gate 1777c478bd9Sstevel@tonic-gate /* Global variables for MP support. Used in mp_startup */ 1787c478bd9Sstevel@tonic-gate caddr_t rm_platter_va; 1797c478bd9Sstevel@tonic-gate uint32_t rm_platter_pa; 1807c478bd9Sstevel@tonic-gate 18183f9b804Skchow int auto_lpg_disable = 1; 18283f9b804Skchow 1837c478bd9Sstevel@tonic-gate /* 1847c478bd9Sstevel@tonic-gate * Some CPUs have holes in the middle of the 64-bit virtual address range. 1857c478bd9Sstevel@tonic-gate */ 1867c478bd9Sstevel@tonic-gate uintptr_t hole_start, hole_end; 1877c478bd9Sstevel@tonic-gate 1887c478bd9Sstevel@tonic-gate /* 1897c478bd9Sstevel@tonic-gate * kpm mapping window 1907c478bd9Sstevel@tonic-gate */ 1917c478bd9Sstevel@tonic-gate caddr_t kpm_vbase; 1927c478bd9Sstevel@tonic-gate size_t kpm_size; 1937c478bd9Sstevel@tonic-gate static int kpm_desired = 0; /* Do we want to try to use segkpm? */ 1947c478bd9Sstevel@tonic-gate 1957c478bd9Sstevel@tonic-gate /* 1967c478bd9Sstevel@tonic-gate * Configuration parameters set at boot time. 1977c478bd9Sstevel@tonic-gate */ 1987c478bd9Sstevel@tonic-gate 1997c478bd9Sstevel@tonic-gate caddr_t econtig; /* end of first block of contiguous kernel */ 2007c478bd9Sstevel@tonic-gate 2017c478bd9Sstevel@tonic-gate struct bootops *bootops = 0; /* passed in from boot */ 2027c478bd9Sstevel@tonic-gate struct bootops **bootopsp; 2037c478bd9Sstevel@tonic-gate struct boot_syscalls *sysp; /* passed in from boot */ 2047c478bd9Sstevel@tonic-gate 2057c478bd9Sstevel@tonic-gate char bootblock_fstype[16]; 2067c478bd9Sstevel@tonic-gate 2077c478bd9Sstevel@tonic-gate char kern_bootargs[OBP_MAXPATHLEN]; 2087c478bd9Sstevel@tonic-gate 209ad23a2dbSjohansen /* 210ad23a2dbSjohansen * ZFS zio segment. This allows us to exclude large portions of ZFS data that 211ad23a2dbSjohansen * gets cached in kmem caches on the heap. If this is set to zero, we allocate 212ad23a2dbSjohansen * zio buffers from their own segment, otherwise they are allocated from the 213ad23a2dbSjohansen * heap. The optimization of allocating zio buffers from their own segment is 214ad23a2dbSjohansen * only valid on 64-bit kernels. 215ad23a2dbSjohansen */ 216ad23a2dbSjohansen #if defined(__amd64) 217ad23a2dbSjohansen int segzio_fromheap = 0; 218ad23a2dbSjohansen #else 219ad23a2dbSjohansen int segzio_fromheap = 1; 220ad23a2dbSjohansen #endif 221ad23a2dbSjohansen 2227c478bd9Sstevel@tonic-gate /* 2237c478bd9Sstevel@tonic-gate * new memory fragmentations are possible in startup() due to BOP_ALLOCs. this 2247c478bd9Sstevel@tonic-gate * depends on number of BOP_ALLOC calls made and requested size, memory size 2257c478bd9Sstevel@tonic-gate * combination and whether boot.bin memory needs to be freed. 2267c478bd9Sstevel@tonic-gate */ 2277c478bd9Sstevel@tonic-gate #define POSS_NEW_FRAGMENTS 12 2287c478bd9Sstevel@tonic-gate 2297c478bd9Sstevel@tonic-gate /* 2307c478bd9Sstevel@tonic-gate * VM data structures 2317c478bd9Sstevel@tonic-gate */ 2327c478bd9Sstevel@tonic-gate long page_hashsz; /* Size of page hash table (power of two) */ 2337c478bd9Sstevel@tonic-gate struct page *pp_base; /* Base of initial system page struct array */ 2347c478bd9Sstevel@tonic-gate struct page **page_hash; /* Page hash table */ 2357c478bd9Sstevel@tonic-gate struct seg ktextseg; /* Segment used for kernel executable image */ 2367c478bd9Sstevel@tonic-gate struct seg kvalloc; /* Segment used for "valloc" mapping */ 2377c478bd9Sstevel@tonic-gate struct seg kpseg; /* Segment used for pageable kernel virt mem */ 2387c478bd9Sstevel@tonic-gate struct seg kmapseg; /* Segment used for generic kernel mappings */ 2397c478bd9Sstevel@tonic-gate struct seg kdebugseg; /* Segment used for the kernel debugger */ 2407c478bd9Sstevel@tonic-gate 2417c478bd9Sstevel@tonic-gate struct seg *segkmap = &kmapseg; /* Kernel generic mapping segment */ 242*ae115bc7Smrj static struct seg *segmap = &kmapseg; /* easier to use name for in here */ 243*ae115bc7Smrj 2447c478bd9Sstevel@tonic-gate struct seg *segkp = &kpseg; /* Pageable kernel virtual memory segment */ 2457c478bd9Sstevel@tonic-gate 2467c478bd9Sstevel@tonic-gate #if defined(__amd64) 2477c478bd9Sstevel@tonic-gate struct seg kvseg_core; /* Segment used for the core heap */ 2487c478bd9Sstevel@tonic-gate struct seg kpmseg; /* Segment used for physical mapping */ 2497c478bd9Sstevel@tonic-gate struct seg *segkpm = &kpmseg; /* 64bit kernel physical mapping segment */ 2507c478bd9Sstevel@tonic-gate #else 2517c478bd9Sstevel@tonic-gate struct seg *segkpm = NULL; /* Unused on IA32 */ 2527c478bd9Sstevel@tonic-gate #endif 2537c478bd9Sstevel@tonic-gate 2547c478bd9Sstevel@tonic-gate caddr_t segkp_base; /* Base address of segkp */ 255ad23a2dbSjohansen caddr_t segzio_base; /* Base address of segzio */ 2567c478bd9Sstevel@tonic-gate #if defined(__amd64) 2577c478bd9Sstevel@tonic-gate pgcnt_t segkpsize = btop(SEGKPDEFSIZE); /* size of segkp segment in pages */ 2587c478bd9Sstevel@tonic-gate #else 2597c478bd9Sstevel@tonic-gate pgcnt_t segkpsize = 0; 2607c478bd9Sstevel@tonic-gate #endif 261ad23a2dbSjohansen pgcnt_t segziosize = 0; /* size of zio segment in pages */ 2627c478bd9Sstevel@tonic-gate 263*ae115bc7Smrj /* 264*ae115bc7Smrj * VA range available to the debugger 265*ae115bc7Smrj */ 266*ae115bc7Smrj const caddr_t kdi_segdebugbase = (const caddr_t)SEGDEBUGBASE; 267*ae115bc7Smrj const size_t kdi_segdebugsize = SEGDEBUGSIZE; 268*ae115bc7Smrj 2697c478bd9Sstevel@tonic-gate struct memseg *memseg_base; 2707c478bd9Sstevel@tonic-gate struct vnode unused_pages_vp; 2717c478bd9Sstevel@tonic-gate 2727c478bd9Sstevel@tonic-gate #define FOURGB 0x100000000LL 2737c478bd9Sstevel@tonic-gate 2747c478bd9Sstevel@tonic-gate struct memlist *memlist; 2757c478bd9Sstevel@tonic-gate 2767c478bd9Sstevel@tonic-gate caddr_t s_text; /* start of kernel text segment */ 2777c478bd9Sstevel@tonic-gate caddr_t e_text; /* end of kernel text segment */ 2787c478bd9Sstevel@tonic-gate caddr_t s_data; /* start of kernel data segment */ 2797c478bd9Sstevel@tonic-gate caddr_t e_data; /* end of kernel data segment */ 2807c478bd9Sstevel@tonic-gate caddr_t modtext; /* start of loadable module text reserved */ 2817c478bd9Sstevel@tonic-gate caddr_t e_modtext; /* end of loadable module text reserved */ 2827c478bd9Sstevel@tonic-gate caddr_t moddata; /* start of loadable module data reserved */ 2837c478bd9Sstevel@tonic-gate caddr_t e_moddata; /* end of loadable module data reserved */ 2847c478bd9Sstevel@tonic-gate 2857c478bd9Sstevel@tonic-gate struct memlist *phys_install; /* Total installed physical memory */ 2867c478bd9Sstevel@tonic-gate struct memlist *phys_avail; /* Total available physical memory */ 2877c478bd9Sstevel@tonic-gate 2887c478bd9Sstevel@tonic-gate /* 2897c478bd9Sstevel@tonic-gate * kphysm_init returns the number of pages that were processed 2907c478bd9Sstevel@tonic-gate */ 291*ae115bc7Smrj static pgcnt_t kphysm_init(page_t *, pgcnt_t); 2927c478bd9Sstevel@tonic-gate 2937c478bd9Sstevel@tonic-gate #define IO_PROP_SIZE 64 /* device property size */ 2947c478bd9Sstevel@tonic-gate 2957c478bd9Sstevel@tonic-gate /* 2967c478bd9Sstevel@tonic-gate * a couple useful roundup macros 2977c478bd9Sstevel@tonic-gate */ 2987c478bd9Sstevel@tonic-gate #define ROUND_UP_PAGE(x) \ 2997c478bd9Sstevel@tonic-gate ((uintptr_t)P2ROUNDUP((uintptr_t)(x), (uintptr_t)MMU_PAGESIZE)) 3007c478bd9Sstevel@tonic-gate #define ROUND_UP_LPAGE(x) \ 3017c478bd9Sstevel@tonic-gate ((uintptr_t)P2ROUNDUP((uintptr_t)(x), mmu.level_size[1])) 3027c478bd9Sstevel@tonic-gate #define ROUND_UP_4MEG(x) \ 303*ae115bc7Smrj ((uintptr_t)P2ROUNDUP((uintptr_t)(x), (uintptr_t)FOUR_MEG)) 3047c478bd9Sstevel@tonic-gate #define ROUND_UP_TOPLEVEL(x) \ 3057c478bd9Sstevel@tonic-gate ((uintptr_t)P2ROUNDUP((uintptr_t)(x), mmu.level_size[mmu.max_level])) 3067c478bd9Sstevel@tonic-gate 3077c478bd9Sstevel@tonic-gate /* 3087c478bd9Sstevel@tonic-gate * 32-bit Kernel's Virtual memory layout. 3097c478bd9Sstevel@tonic-gate * +-----------------------+ 310*ae115bc7Smrj * | | 3117c478bd9Sstevel@tonic-gate * 0xFFC00000 -|-----------------------|- ARGSBASE 3127c478bd9Sstevel@tonic-gate * | debugger | 3137c478bd9Sstevel@tonic-gate * 0xFF800000 -|-----------------------|- SEGDEBUGBASE 3147c478bd9Sstevel@tonic-gate * | Kernel Data | 3157c478bd9Sstevel@tonic-gate * 0xFEC00000 -|-----------------------| 3167c478bd9Sstevel@tonic-gate * | Kernel Text | 3177c478bd9Sstevel@tonic-gate * 0xFE800000 -|-----------------------|- KERNEL_TEXT 318*ae115bc7Smrj * |--- GDT ---|- GDT page (GDT_VA) 319*ae115bc7Smrj * |--- debug info ---|- debug info (DEBUG_INFO_VA) 320*ae115bc7Smrj * | | 321*ae115bc7Smrj * | page_t structures | 3227c478bd9Sstevel@tonic-gate * | memsegs, memlists, | 3237c478bd9Sstevel@tonic-gate * | page hash, etc. | 324*ae115bc7Smrj * --- -|-----------------------|- ekernelheap, valloc_base (floating) 325*ae115bc7Smrj * | | (segkp is just an arena in the heap) 3267c478bd9Sstevel@tonic-gate * | | 3277c478bd9Sstevel@tonic-gate * | kvseg | 3287c478bd9Sstevel@tonic-gate * | | 3297c478bd9Sstevel@tonic-gate * | | 3307c478bd9Sstevel@tonic-gate * --- -|-----------------------|- kernelheap (floating) 3317c478bd9Sstevel@tonic-gate * | Segkmap | 332*ae115bc7Smrj * 0xC3002000 -|-----------------------|- segmap_start (floating) 3337c478bd9Sstevel@tonic-gate * | Red Zone | 3347c478bd9Sstevel@tonic-gate * 0xC3000000 -|-----------------------|- kernelbase / userlimit (floating) 3357c478bd9Sstevel@tonic-gate * | | || 3367c478bd9Sstevel@tonic-gate * | Shared objects | \/ 3377c478bd9Sstevel@tonic-gate * | | 3387c478bd9Sstevel@tonic-gate * : : 3397c478bd9Sstevel@tonic-gate * | user data | 3407c478bd9Sstevel@tonic-gate * |-----------------------| 3417c478bd9Sstevel@tonic-gate * | user text | 3427c478bd9Sstevel@tonic-gate * 0x08048000 -|-----------------------| 3437c478bd9Sstevel@tonic-gate * | user stack | 3447c478bd9Sstevel@tonic-gate * : : 3457c478bd9Sstevel@tonic-gate * | invalid | 3467c478bd9Sstevel@tonic-gate * 0x00000000 +-----------------------+ 3477c478bd9Sstevel@tonic-gate * 3487c478bd9Sstevel@tonic-gate * 3497c478bd9Sstevel@tonic-gate * 64-bit Kernel's Virtual memory layout. (assuming 64 bit app) 3507c478bd9Sstevel@tonic-gate * +-----------------------+ 351*ae115bc7Smrj * | | 3527c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FFC00000 |-----------------------|- ARGSBASE 3537c478bd9Sstevel@tonic-gate * | debugger (?) | 3547c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FF800000 |-----------------------|- SEGDEBUGBASE 3557c478bd9Sstevel@tonic-gate * | unused | 3567c478bd9Sstevel@tonic-gate * +-----------------------+ 3577c478bd9Sstevel@tonic-gate * | Kernel Data | 3587c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FBC00000 |-----------------------| 3597c478bd9Sstevel@tonic-gate * | Kernel Text | 3607c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FB800000 |-----------------------|- KERNEL_TEXT 361*ae115bc7Smrj * |--- GDT ---|- GDT page (GDT_VA) 362*ae115bc7Smrj * |--- debug info ---|- debug info (DEBUG_INFO_VA) 363*ae115bc7Smrj * | | 3647c478bd9Sstevel@tonic-gate * | Core heap | (used for loadable modules) 3657c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.C0000000 |-----------------------|- core_base / ekernelheap 3667c478bd9Sstevel@tonic-gate * | Kernel | 3677c478bd9Sstevel@tonic-gate * | heap | 3687c478bd9Sstevel@tonic-gate * 0xFFFFFXXX.XXX00000 |-----------------------|- kernelheap (floating) 369*ae115bc7Smrj * | segmap | 370*ae115bc7Smrj * 0xFFFFFXXX.XXX00000 |-----------------------|- segmap_start (floating) 3717c478bd9Sstevel@tonic-gate * | device mappings | 3727c478bd9Sstevel@tonic-gate * 0xFFFFFXXX.XXX00000 |-----------------------|- toxic_addr (floating) 373ad23a2dbSjohansen * | segzio | 374ad23a2dbSjohansen * 0xFFFFFXXX.XXX00000 |-----------------------|- segzio_base (floating) 3757c478bd9Sstevel@tonic-gate * | segkp | 376*ae115bc7Smrj * --- |-----------------------|- segkp_base (floating) 377*ae115bc7Smrj * | page_t structures | valloc_base + valloc_sz 378*ae115bc7Smrj * | memsegs, memlists, | 379*ae115bc7Smrj * | page hash, etc. | 380*ae115bc7Smrj * 0xFFFFFF00.00000000 |-----------------------|- valloc_base 3817c478bd9Sstevel@tonic-gate * | segkpm | 3827c478bd9Sstevel@tonic-gate * 0xFFFFFE00.00000000 |-----------------------| 3837c478bd9Sstevel@tonic-gate * | Red Zone | 3847c478bd9Sstevel@tonic-gate * 0xFFFFFD80.00000000 |-----------------------|- KERNELBASE 3857c478bd9Sstevel@tonic-gate * | User stack |- User space memory 3867c478bd9Sstevel@tonic-gate * | | 3877c478bd9Sstevel@tonic-gate * | shared objects, etc | (grows downwards) 3887c478bd9Sstevel@tonic-gate * : : 3897c478bd9Sstevel@tonic-gate * | | 3907c478bd9Sstevel@tonic-gate * 0xFFFF8000.00000000 |-----------------------| 3917c478bd9Sstevel@tonic-gate * | | 3927c478bd9Sstevel@tonic-gate * | VA Hole / unused | 3937c478bd9Sstevel@tonic-gate * | | 3947c478bd9Sstevel@tonic-gate * 0x00008000.00000000 |-----------------------| 3957c478bd9Sstevel@tonic-gate * | | 3967c478bd9Sstevel@tonic-gate * | | 3977c478bd9Sstevel@tonic-gate * : : 3987c478bd9Sstevel@tonic-gate * | user heap | (grows upwards) 3997c478bd9Sstevel@tonic-gate * | | 4007c478bd9Sstevel@tonic-gate * | user data | 4017c478bd9Sstevel@tonic-gate * |-----------------------| 4027c478bd9Sstevel@tonic-gate * | user text | 4037c478bd9Sstevel@tonic-gate * 0x00000000.04000000 |-----------------------| 4047c478bd9Sstevel@tonic-gate * | invalid | 4057c478bd9Sstevel@tonic-gate * 0x00000000.00000000 +-----------------------+ 4067c478bd9Sstevel@tonic-gate * 4077c478bd9Sstevel@tonic-gate * A 32 bit app on the 64 bit kernel sees the same layout as on the 32 bit 4087c478bd9Sstevel@tonic-gate * kernel, except that userlimit is raised to 0xfe000000 4097c478bd9Sstevel@tonic-gate * 4107c478bd9Sstevel@tonic-gate * Floating values: 4117c478bd9Sstevel@tonic-gate * 4127c478bd9Sstevel@tonic-gate * valloc_base: start of the kernel's memory management/tracking data 413*ae115bc7Smrj * structures. This region contains page_t structures for 414*ae115bc7Smrj * physical memory, memsegs, memlists, and the page hash. 4157c478bd9Sstevel@tonic-gate * 4167c478bd9Sstevel@tonic-gate * core_base: start of the kernel's "core" heap area on 64-bit systems. 4177c478bd9Sstevel@tonic-gate * This area is intended to be used for global data as well as for module 4187c478bd9Sstevel@tonic-gate * text/data that does not fit into the nucleus pages. The core heap is 4197c478bd9Sstevel@tonic-gate * restricted to a 2GB range, allowing every address within it to be 4207c478bd9Sstevel@tonic-gate * accessed using rip-relative addressing 4217c478bd9Sstevel@tonic-gate * 4227c478bd9Sstevel@tonic-gate * ekernelheap: end of kernelheap and start of segmap. 4237c478bd9Sstevel@tonic-gate * 4247c478bd9Sstevel@tonic-gate * kernelheap: start of kernel heap. On 32-bit systems, this starts right 4257c478bd9Sstevel@tonic-gate * above a red zone that separates the user's address space from the 4267c478bd9Sstevel@tonic-gate * kernel's. On 64-bit systems, it sits above segkp and segkpm. 4277c478bd9Sstevel@tonic-gate * 428*ae115bc7Smrj * segmap_start: start of segmap. The length of segmap can be modified 4297c478bd9Sstevel@tonic-gate * by changing segmapsize in /etc/system (preferred) or eeprom (deprecated). 4307c478bd9Sstevel@tonic-gate * The default length is 16MB on 32-bit systems and 64MB on 64-bit systems. 4317c478bd9Sstevel@tonic-gate * 4327c478bd9Sstevel@tonic-gate * kernelbase: On a 32-bit kernel the default value of 0xd4000000 will be 4337c478bd9Sstevel@tonic-gate * decreased by 2X the size required for page_t. This allows the kernel 4347c478bd9Sstevel@tonic-gate * heap to grow in size with physical memory. With sizeof(page_t) == 80 4357c478bd9Sstevel@tonic-gate * bytes, the following shows the values of kernelbase and kernel heap 4367c478bd9Sstevel@tonic-gate * sizes for different memory configurations (assuming default segmap and 4377c478bd9Sstevel@tonic-gate * segkp sizes). 4387c478bd9Sstevel@tonic-gate * 4397c478bd9Sstevel@tonic-gate * mem size for kernelbase kernel heap 4407c478bd9Sstevel@tonic-gate * size page_t's size 4417c478bd9Sstevel@tonic-gate * ---- --------- ---------- ----------- 4427c478bd9Sstevel@tonic-gate * 1gb 0x01400000 0xd1800000 684MB 4437c478bd9Sstevel@tonic-gate * 2gb 0x02800000 0xcf000000 704MB 4447c478bd9Sstevel@tonic-gate * 4gb 0x05000000 0xca000000 744MB 4457c478bd9Sstevel@tonic-gate * 6gb 0x07800000 0xc5000000 784MB 4467c478bd9Sstevel@tonic-gate * 8gb 0x0a000000 0xc0000000 824MB 4477c478bd9Sstevel@tonic-gate * 16gb 0x14000000 0xac000000 984MB 4487c478bd9Sstevel@tonic-gate * 32gb 0x28000000 0x84000000 1304MB 4497c478bd9Sstevel@tonic-gate * 64gb 0x50000000 0x34000000 1944MB (*) 4507c478bd9Sstevel@tonic-gate * 4517c478bd9Sstevel@tonic-gate * kernelbase is less than the abi minimum of 0xc0000000 for memory 4527c478bd9Sstevel@tonic-gate * configurations above 8gb. 4537c478bd9Sstevel@tonic-gate * 4547c478bd9Sstevel@tonic-gate * (*) support for memory configurations above 32gb will require manual tuning 4557c478bd9Sstevel@tonic-gate * of kernelbase to balance out the need of user applications. 4567c478bd9Sstevel@tonic-gate */ 4577c478bd9Sstevel@tonic-gate 4587c478bd9Sstevel@tonic-gate /* real-time-clock initialization parameters */ 459*ae115bc7Smrj extern time_t process_rtc_config_file(void); 4607c478bd9Sstevel@tonic-gate 4617c478bd9Sstevel@tonic-gate char *final_kernelheap; 4627c478bd9Sstevel@tonic-gate char *boot_kernelheap; 4637c478bd9Sstevel@tonic-gate uintptr_t kernelbase; 464*ae115bc7Smrj uintptr_t postbootkernelbase; /* not set till boot loader is gone */ 4657c478bd9Sstevel@tonic-gate uintptr_t eprom_kernelbase; 4667c478bd9Sstevel@tonic-gate size_t segmapsize; 4677c478bd9Sstevel@tonic-gate static uintptr_t segmap_reserved; 468*ae115bc7Smrj uintptr_t segmap_start; 4697c478bd9Sstevel@tonic-gate int segmapfreelists; 4707c478bd9Sstevel@tonic-gate pgcnt_t npages; 471*ae115bc7Smrj pgcnt_t orig_npages; 4727c478bd9Sstevel@tonic-gate size_t core_size; /* size of "core" heap */ 4737c478bd9Sstevel@tonic-gate uintptr_t core_base; /* base address of "core" heap */ 4747c478bd9Sstevel@tonic-gate 4757c478bd9Sstevel@tonic-gate /* 4767c478bd9Sstevel@tonic-gate * List of bootstrap pages. We mark these as allocated in startup. 4777c478bd9Sstevel@tonic-gate * release_bootstrap() will free them when we're completely done with 4787c478bd9Sstevel@tonic-gate * the bootstrap. 4797c478bd9Sstevel@tonic-gate */ 480*ae115bc7Smrj static page_t *bootpages; 481*ae115bc7Smrj 482*ae115bc7Smrj /* 483*ae115bc7Smrj * boot time pages that have a vnode from the ramdisk will keep that forever. 484*ae115bc7Smrj */ 485*ae115bc7Smrj static page_t *rd_pages; 4867c478bd9Sstevel@tonic-gate 4877c478bd9Sstevel@tonic-gate struct system_hardware system_hardware; 4887c478bd9Sstevel@tonic-gate 4897c478bd9Sstevel@tonic-gate /* 4907c478bd9Sstevel@tonic-gate * Enable some debugging messages concerning memory usage... 4917c478bd9Sstevel@tonic-gate */ 4927c478bd9Sstevel@tonic-gate static void 493*ae115bc7Smrj print_memlist(char *title, struct memlist *mp) 4947c478bd9Sstevel@tonic-gate { 4957c478bd9Sstevel@tonic-gate prom_printf("MEMLIST: %s:\n", title); 4967c478bd9Sstevel@tonic-gate while (mp != NULL) { 4977c478bd9Sstevel@tonic-gate prom_printf("\tAddress 0x%" PRIx64 ", size 0x%" PRIx64 "\n", 4987c478bd9Sstevel@tonic-gate mp->address, mp->size); 4997c478bd9Sstevel@tonic-gate mp = mp->next; 5007c478bd9Sstevel@tonic-gate } 5017c478bd9Sstevel@tonic-gate } 5027c478bd9Sstevel@tonic-gate 5037c478bd9Sstevel@tonic-gate /* 5047c478bd9Sstevel@tonic-gate * XX64 need a comment here.. are these just default values, surely 5057c478bd9Sstevel@tonic-gate * we read the "cpuid" type information to figure this out. 5067c478bd9Sstevel@tonic-gate */ 5077c478bd9Sstevel@tonic-gate int l2cache_sz = 0x80000; 5087c478bd9Sstevel@tonic-gate int l2cache_linesz = 0x40; 5097c478bd9Sstevel@tonic-gate int l2cache_assoc = 1; 5107c478bd9Sstevel@tonic-gate 5117c478bd9Sstevel@tonic-gate /* 5127c478bd9Sstevel@tonic-gate * on 64 bit we use a predifined VA range for mapping devices in the kernel 5137c478bd9Sstevel@tonic-gate * on 32 bit the mappings are intermixed in the heap, so we use a bit map 5147c478bd9Sstevel@tonic-gate */ 5157c478bd9Sstevel@tonic-gate #ifdef __amd64 5167c478bd9Sstevel@tonic-gate 5177c478bd9Sstevel@tonic-gate vmem_t *device_arena; 5187c478bd9Sstevel@tonic-gate uintptr_t toxic_addr = (uintptr_t)NULL; 519*ae115bc7Smrj size_t toxic_size = 1024 * 1024 * 1024; /* Sparc uses 1 gig too */ 5207c478bd9Sstevel@tonic-gate 5217c478bd9Sstevel@tonic-gate #else /* __i386 */ 5227c478bd9Sstevel@tonic-gate 5237c478bd9Sstevel@tonic-gate ulong_t *toxic_bit_map; /* one bit for each 4k of VA in heap_arena */ 5247c478bd9Sstevel@tonic-gate size_t toxic_bit_map_len = 0; /* in bits */ 5257c478bd9Sstevel@tonic-gate 5267c478bd9Sstevel@tonic-gate #endif /* __i386 */ 5277c478bd9Sstevel@tonic-gate 5287c478bd9Sstevel@tonic-gate /* 5297c478bd9Sstevel@tonic-gate * Simple boot time debug facilities 5307c478bd9Sstevel@tonic-gate */ 5317c478bd9Sstevel@tonic-gate static char *prm_dbg_str[] = { 5327c478bd9Sstevel@tonic-gate "%s:%d: '%s' is 0x%x\n", 5337c478bd9Sstevel@tonic-gate "%s:%d: '%s' is 0x%llx\n" 5347c478bd9Sstevel@tonic-gate }; 5357c478bd9Sstevel@tonic-gate 5367c478bd9Sstevel@tonic-gate int prom_debug; 5377c478bd9Sstevel@tonic-gate 5387c478bd9Sstevel@tonic-gate #define PRM_DEBUG(q) if (prom_debug) \ 5397c478bd9Sstevel@tonic-gate prom_printf(prm_dbg_str[sizeof (q) >> 3], "startup.c", __LINE__, #q, q); 5407c478bd9Sstevel@tonic-gate #define PRM_POINT(q) if (prom_debug) \ 5417c478bd9Sstevel@tonic-gate prom_printf("%s:%d: %s\n", "startup.c", __LINE__, q); 5427c478bd9Sstevel@tonic-gate 5437c478bd9Sstevel@tonic-gate /* 5447c478bd9Sstevel@tonic-gate * This structure is used to keep track of the intial allocations 5457c478bd9Sstevel@tonic-gate * done in startup_memlist(). The value of NUM_ALLOCATIONS needs to 5467c478bd9Sstevel@tonic-gate * be >= the number of ADD_TO_ALLOCATIONS() executed in the code. 5477c478bd9Sstevel@tonic-gate */ 5487c478bd9Sstevel@tonic-gate #define NUM_ALLOCATIONS 7 5497c478bd9Sstevel@tonic-gate int num_allocations = 0; 5507c478bd9Sstevel@tonic-gate struct { 5517c478bd9Sstevel@tonic-gate void **al_ptr; 5527c478bd9Sstevel@tonic-gate size_t al_size; 5537c478bd9Sstevel@tonic-gate } allocations[NUM_ALLOCATIONS]; 5547c478bd9Sstevel@tonic-gate size_t valloc_sz = 0; 5557c478bd9Sstevel@tonic-gate uintptr_t valloc_base; 5567c478bd9Sstevel@tonic-gate 5577c478bd9Sstevel@tonic-gate #define ADD_TO_ALLOCATIONS(ptr, size) { \ 5587c478bd9Sstevel@tonic-gate size = ROUND_UP_PAGE(size); \ 5597c478bd9Sstevel@tonic-gate if (num_allocations == NUM_ALLOCATIONS) \ 5607c478bd9Sstevel@tonic-gate panic("too many ADD_TO_ALLOCATIONS()"); \ 5617c478bd9Sstevel@tonic-gate allocations[num_allocations].al_ptr = (void**)&ptr; \ 5627c478bd9Sstevel@tonic-gate allocations[num_allocations].al_size = size; \ 5637c478bd9Sstevel@tonic-gate valloc_sz += size; \ 5647c478bd9Sstevel@tonic-gate ++num_allocations; \ 5657c478bd9Sstevel@tonic-gate } 5667c478bd9Sstevel@tonic-gate 567*ae115bc7Smrj /* 568*ae115bc7Smrj * Allocate all the initial memory needed by the page allocator. 569*ae115bc7Smrj */ 5707c478bd9Sstevel@tonic-gate static void 5717c478bd9Sstevel@tonic-gate perform_allocations(void) 5727c478bd9Sstevel@tonic-gate { 5737c478bd9Sstevel@tonic-gate caddr_t mem; 5747c478bd9Sstevel@tonic-gate int i; 575*ae115bc7Smrj int valloc_align; 5767c478bd9Sstevel@tonic-gate 577*ae115bc7Smrj PRM_DEBUG(valloc_base); 578*ae115bc7Smrj PRM_DEBUG(valloc_sz); 579*ae115bc7Smrj valloc_align = mmu.level_size[mmu.max_page_level > 0]; 580*ae115bc7Smrj mem = BOP_ALLOC(bootops, (caddr_t)valloc_base, valloc_sz, valloc_align); 5817c478bd9Sstevel@tonic-gate if (mem != (caddr_t)valloc_base) 5827c478bd9Sstevel@tonic-gate panic("BOP_ALLOC() failed"); 5837c478bd9Sstevel@tonic-gate bzero(mem, valloc_sz); 5847c478bd9Sstevel@tonic-gate for (i = 0; i < num_allocations; ++i) { 5857c478bd9Sstevel@tonic-gate *allocations[i].al_ptr = (void *)mem; 5867c478bd9Sstevel@tonic-gate mem += allocations[i].al_size; 5877c478bd9Sstevel@tonic-gate } 5887c478bd9Sstevel@tonic-gate } 5897c478bd9Sstevel@tonic-gate 5907c478bd9Sstevel@tonic-gate /* 5917c478bd9Sstevel@tonic-gate * Our world looks like this at startup time. 5927c478bd9Sstevel@tonic-gate * 5937c478bd9Sstevel@tonic-gate * In a 32-bit OS, boot loads the kernel text at 0xfe800000 and kernel data 5947c478bd9Sstevel@tonic-gate * at 0xfec00000. On a 64-bit OS, kernel text and data are loaded at 5957c478bd9Sstevel@tonic-gate * 0xffffffff.fe800000 and 0xffffffff.fec00000 respectively. Those 5967c478bd9Sstevel@tonic-gate * addresses are fixed in the binary at link time. 5977c478bd9Sstevel@tonic-gate * 5987c478bd9Sstevel@tonic-gate * On the text page: 5997c478bd9Sstevel@tonic-gate * unix/genunix/krtld/module text loads. 6007c478bd9Sstevel@tonic-gate * 6017c478bd9Sstevel@tonic-gate * On the data page: 602*ae115bc7Smrj * unix/genunix/krtld/module data loads. 603*ae115bc7Smrj * 6047c478bd9Sstevel@tonic-gate * Machine-dependent startup code 6057c478bd9Sstevel@tonic-gate */ 6067c478bd9Sstevel@tonic-gate void 6077c478bd9Sstevel@tonic-gate startup(void) 6087c478bd9Sstevel@tonic-gate { 60975bcd456Sjg extern void startup_bios_disk(void); 61075bcd456Sjg extern void startup_pci_bios(void); 6117c478bd9Sstevel@tonic-gate /* 6127c478bd9Sstevel@tonic-gate * Make sure that nobody tries to use sekpm until we have 6137c478bd9Sstevel@tonic-gate * initialized it properly. 6147c478bd9Sstevel@tonic-gate */ 6157c478bd9Sstevel@tonic-gate #if defined(__amd64) 6167c478bd9Sstevel@tonic-gate kpm_desired = kpm_enable; 6177c478bd9Sstevel@tonic-gate #endif 6187c478bd9Sstevel@tonic-gate kpm_enable = 0; 6197c478bd9Sstevel@tonic-gate 6207c478bd9Sstevel@tonic-gate progressbar_init(); 6217c478bd9Sstevel@tonic-gate startup_init(); 6227c478bd9Sstevel@tonic-gate startup_memlist(); 623*ae115bc7Smrj startup_kmem(); 62475bcd456Sjg startup_pci_bios(); 6257c478bd9Sstevel@tonic-gate startup_modules(); 6267c478bd9Sstevel@tonic-gate startup_bios_disk(); 6277c478bd9Sstevel@tonic-gate startup_vm(); 6287c478bd9Sstevel@tonic-gate startup_end(); 6297c478bd9Sstevel@tonic-gate progressbar_start(); 6307c478bd9Sstevel@tonic-gate } 6317c478bd9Sstevel@tonic-gate 6327c478bd9Sstevel@tonic-gate static void 6337c478bd9Sstevel@tonic-gate startup_init() 6347c478bd9Sstevel@tonic-gate { 6357c478bd9Sstevel@tonic-gate PRM_POINT("startup_init() starting..."); 6367c478bd9Sstevel@tonic-gate 6377c478bd9Sstevel@tonic-gate /* 6387c478bd9Sstevel@tonic-gate * Complete the extraction of cpuid data 6397c478bd9Sstevel@tonic-gate */ 6407c478bd9Sstevel@tonic-gate cpuid_pass2(CPU); 6417c478bd9Sstevel@tonic-gate 6427c478bd9Sstevel@tonic-gate (void) check_boot_version(BOP_GETVERSION(bootops)); 6437c478bd9Sstevel@tonic-gate 6447c478bd9Sstevel@tonic-gate /* 6457c478bd9Sstevel@tonic-gate * Check for prom_debug in boot environment 6467c478bd9Sstevel@tonic-gate */ 6477c478bd9Sstevel@tonic-gate if (BOP_GETPROPLEN(bootops, "prom_debug") >= 0) { 6487c478bd9Sstevel@tonic-gate ++prom_debug; 6497c478bd9Sstevel@tonic-gate PRM_POINT("prom_debug found in boot enviroment"); 6507c478bd9Sstevel@tonic-gate } 6517c478bd9Sstevel@tonic-gate 6527c478bd9Sstevel@tonic-gate /* 6537c478bd9Sstevel@tonic-gate * Collect node, cpu and memory configuration information. 6547c478bd9Sstevel@tonic-gate */ 6557c478bd9Sstevel@tonic-gate get_system_configuration(); 6567c478bd9Sstevel@tonic-gate 6577c478bd9Sstevel@tonic-gate /* 6587c478bd9Sstevel@tonic-gate * Halt if this is an unsupported processor. 6597c478bd9Sstevel@tonic-gate */ 6607c478bd9Sstevel@tonic-gate if (x86_type == X86_TYPE_486 || x86_type == X86_TYPE_CYRIX_486) { 6617c478bd9Sstevel@tonic-gate printf("\n486 processor (\"%s\") detected.\n", 6627c478bd9Sstevel@tonic-gate CPU->cpu_brandstr); 6637c478bd9Sstevel@tonic-gate halt("This processor is not supported by this release " 6647c478bd9Sstevel@tonic-gate "of Solaris."); 6657c478bd9Sstevel@tonic-gate } 6667c478bd9Sstevel@tonic-gate 6677c478bd9Sstevel@tonic-gate PRM_POINT("startup_init() done"); 6687c478bd9Sstevel@tonic-gate } 6697c478bd9Sstevel@tonic-gate 6707c478bd9Sstevel@tonic-gate /* 6717c478bd9Sstevel@tonic-gate * Callback for copy_memlist_filter() to filter nucleus, kadb/kmdb, (ie. 6727c478bd9Sstevel@tonic-gate * everything mapped above KERNEL_TEXT) pages from phys_avail. Note it 6737c478bd9Sstevel@tonic-gate * also filters out physical page zero. There is some reliance on the 6747c478bd9Sstevel@tonic-gate * boot loader allocating only a few contiguous physical memory chunks. 6757c478bd9Sstevel@tonic-gate */ 6767c478bd9Sstevel@tonic-gate static void 6777c478bd9Sstevel@tonic-gate avail_filter(uint64_t *addr, uint64_t *size) 6787c478bd9Sstevel@tonic-gate { 6797c478bd9Sstevel@tonic-gate uintptr_t va; 6807c478bd9Sstevel@tonic-gate uintptr_t next_va; 6817c478bd9Sstevel@tonic-gate pfn_t pfn; 6827c478bd9Sstevel@tonic-gate uint64_t pfn_addr; 6837c478bd9Sstevel@tonic-gate uint64_t pfn_eaddr; 6847c478bd9Sstevel@tonic-gate uint_t prot; 6857c478bd9Sstevel@tonic-gate size_t len; 6867c478bd9Sstevel@tonic-gate uint_t change; 6877c478bd9Sstevel@tonic-gate 6887c478bd9Sstevel@tonic-gate if (prom_debug) 6897c478bd9Sstevel@tonic-gate prom_printf("\tFilter: in: a=%" PRIx64 ", s=%" PRIx64 "\n", 6907c478bd9Sstevel@tonic-gate *addr, *size); 6917c478bd9Sstevel@tonic-gate 6927c478bd9Sstevel@tonic-gate /* 6937c478bd9Sstevel@tonic-gate * page zero is required for BIOS.. never make it available 6947c478bd9Sstevel@tonic-gate */ 6957c478bd9Sstevel@tonic-gate if (*addr == 0) { 6967c478bd9Sstevel@tonic-gate *addr += MMU_PAGESIZE; 6977c478bd9Sstevel@tonic-gate *size -= MMU_PAGESIZE; 6987c478bd9Sstevel@tonic-gate } 6997c478bd9Sstevel@tonic-gate 7007c478bd9Sstevel@tonic-gate /* 701*ae115bc7Smrj * First we trim from the front of the range. Since kbm_probe() 7027c478bd9Sstevel@tonic-gate * walks ranges in virtual order, but addr/size are physical, we need 7037c478bd9Sstevel@tonic-gate * to the list until no changes are seen. This deals with the case 7047c478bd9Sstevel@tonic-gate * where page "p" is mapped at v, page "p + PAGESIZE" is mapped at w 7057c478bd9Sstevel@tonic-gate * but w < v. 7067c478bd9Sstevel@tonic-gate */ 7077c478bd9Sstevel@tonic-gate do { 7087c478bd9Sstevel@tonic-gate change = 0; 7097c478bd9Sstevel@tonic-gate for (va = KERNEL_TEXT; 710*ae115bc7Smrj *size > 0 && kbm_probe(&va, &len, &pfn, &prot) != 0; 7117c478bd9Sstevel@tonic-gate va = next_va) { 7127c478bd9Sstevel@tonic-gate 7137c478bd9Sstevel@tonic-gate next_va = va + len; 714*ae115bc7Smrj pfn_addr = pfn_to_pa(pfn); 7157c478bd9Sstevel@tonic-gate pfn_eaddr = pfn_addr + len; 7167c478bd9Sstevel@tonic-gate 7177c478bd9Sstevel@tonic-gate if (pfn_addr <= *addr && pfn_eaddr > *addr) { 7187c478bd9Sstevel@tonic-gate change = 1; 7197c478bd9Sstevel@tonic-gate while (*size > 0 && len > 0) { 7207c478bd9Sstevel@tonic-gate *addr += MMU_PAGESIZE; 7217c478bd9Sstevel@tonic-gate *size -= MMU_PAGESIZE; 7227c478bd9Sstevel@tonic-gate len -= MMU_PAGESIZE; 7237c478bd9Sstevel@tonic-gate } 7247c478bd9Sstevel@tonic-gate } 7257c478bd9Sstevel@tonic-gate } 7267c478bd9Sstevel@tonic-gate if (change && prom_debug) 7277c478bd9Sstevel@tonic-gate prom_printf("\t\ttrim: a=%" PRIx64 ", s=%" PRIx64 "\n", 7287c478bd9Sstevel@tonic-gate *addr, *size); 7297c478bd9Sstevel@tonic-gate } while (change); 7307c478bd9Sstevel@tonic-gate 7317c478bd9Sstevel@tonic-gate /* 7327c478bd9Sstevel@tonic-gate * Trim pages from the end of the range. 7337c478bd9Sstevel@tonic-gate */ 7347c478bd9Sstevel@tonic-gate for (va = KERNEL_TEXT; 735*ae115bc7Smrj *size > 0 && kbm_probe(&va, &len, &pfn, &prot) != 0; 7367c478bd9Sstevel@tonic-gate va = next_va) { 7377c478bd9Sstevel@tonic-gate 7387c478bd9Sstevel@tonic-gate next_va = va + len; 739*ae115bc7Smrj pfn_addr = pfn_to_pa(pfn); 7407c478bd9Sstevel@tonic-gate 7417c478bd9Sstevel@tonic-gate if (pfn_addr >= *addr && pfn_addr < *addr + *size) 7427c478bd9Sstevel@tonic-gate *size = pfn_addr - *addr; 7437c478bd9Sstevel@tonic-gate } 7447c478bd9Sstevel@tonic-gate 7457c478bd9Sstevel@tonic-gate if (prom_debug) 7467c478bd9Sstevel@tonic-gate prom_printf("\tFilter out: a=%" PRIx64 ", s=%" PRIx64 "\n", 7477c478bd9Sstevel@tonic-gate *addr, *size); 7487c478bd9Sstevel@tonic-gate } 7497c478bd9Sstevel@tonic-gate 7507c478bd9Sstevel@tonic-gate static void 7517c478bd9Sstevel@tonic-gate kpm_init() 7527c478bd9Sstevel@tonic-gate { 7537c478bd9Sstevel@tonic-gate struct segkpm_crargs b; 7547c478bd9Sstevel@tonic-gate 7557c478bd9Sstevel@tonic-gate /* 7567c478bd9Sstevel@tonic-gate * These variables were all designed for sfmmu in which segkpm is 7577c478bd9Sstevel@tonic-gate * mapped using a single pagesize - either 8KB or 4MB. On x86, we 7587c478bd9Sstevel@tonic-gate * might use 2+ page sizes on a single machine, so none of these 7597c478bd9Sstevel@tonic-gate * variables have a single correct value. They are set up as if we 7607c478bd9Sstevel@tonic-gate * always use a 4KB pagesize, which should do no harm. In the long 7617c478bd9Sstevel@tonic-gate * run, we should get rid of KPM's assumption that only a single 7627c478bd9Sstevel@tonic-gate * pagesize is used. 7637c478bd9Sstevel@tonic-gate */ 7647c478bd9Sstevel@tonic-gate kpm_pgshft = MMU_PAGESHIFT; 7657c478bd9Sstevel@tonic-gate kpm_pgsz = MMU_PAGESIZE; 7667c478bd9Sstevel@tonic-gate kpm_pgoff = MMU_PAGEOFFSET; 7677c478bd9Sstevel@tonic-gate kpmp2pshft = 0; 7687c478bd9Sstevel@tonic-gate kpmpnpgs = 1; 7697c478bd9Sstevel@tonic-gate ASSERT(((uintptr_t)kpm_vbase & (kpm_pgsz - 1)) == 0); 7707c478bd9Sstevel@tonic-gate 7717c478bd9Sstevel@tonic-gate PRM_POINT("about to create segkpm"); 7727c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 7737c478bd9Sstevel@tonic-gate 7747c478bd9Sstevel@tonic-gate if (seg_attach(&kas, kpm_vbase, kpm_size, segkpm) < 0) 7757c478bd9Sstevel@tonic-gate panic("cannot attach segkpm"); 7767c478bd9Sstevel@tonic-gate 7777c478bd9Sstevel@tonic-gate b.prot = PROT_READ | PROT_WRITE; 7787c478bd9Sstevel@tonic-gate b.nvcolors = 1; 7797c478bd9Sstevel@tonic-gate 7807c478bd9Sstevel@tonic-gate if (segkpm_create(segkpm, (caddr_t)&b) != 0) 7817c478bd9Sstevel@tonic-gate panic("segkpm_create segkpm"); 7827c478bd9Sstevel@tonic-gate 7837c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 784*ae115bc7Smrj } 7857c478bd9Sstevel@tonic-gate 786*ae115bc7Smrj /* 787*ae115bc7Smrj * The debug info page provides enough information to allow external 788*ae115bc7Smrj * inspectors (e.g. when running under a hypervisor) to bootstrap 789*ae115bc7Smrj * themselves into allowing full-blown kernel debugging. 790*ae115bc7Smrj */ 791*ae115bc7Smrj static void 792*ae115bc7Smrj init_debug_info(void) 793*ae115bc7Smrj { 794*ae115bc7Smrj caddr_t mem; 795*ae115bc7Smrj debug_info_t *di; 796*ae115bc7Smrj 797*ae115bc7Smrj #ifndef __lint 798*ae115bc7Smrj ASSERT(sizeof (debug_info_t) < MMU_PAGESIZE); 799*ae115bc7Smrj #endif 800*ae115bc7Smrj 801*ae115bc7Smrj mem = BOP_ALLOC(bootops, (caddr_t)DEBUG_INFO_VA, MMU_PAGESIZE, 802*ae115bc7Smrj MMU_PAGESIZE); 803*ae115bc7Smrj 804*ae115bc7Smrj if (mem != (caddr_t)DEBUG_INFO_VA) 805*ae115bc7Smrj panic("BOP_ALLOC() failed"); 806*ae115bc7Smrj bzero(mem, MMU_PAGESIZE); 807*ae115bc7Smrj 808*ae115bc7Smrj di = (debug_info_t *)mem; 809*ae115bc7Smrj 810*ae115bc7Smrj di->di_magic = DEBUG_INFO_MAGIC; 811*ae115bc7Smrj di->di_version = DEBUG_INFO_VERSION; 8127c478bd9Sstevel@tonic-gate } 8137c478bd9Sstevel@tonic-gate 8147c478bd9Sstevel@tonic-gate /* 815*ae115bc7Smrj * Build the memlists and other kernel essential memory system data structures. 816*ae115bc7Smrj * This is everything at valloc_base. 8177c478bd9Sstevel@tonic-gate */ 8187c478bd9Sstevel@tonic-gate static void 8197c478bd9Sstevel@tonic-gate startup_memlist(void) 8207c478bd9Sstevel@tonic-gate { 8217c478bd9Sstevel@tonic-gate size_t memlist_sz; 8227c478bd9Sstevel@tonic-gate size_t memseg_sz; 8237c478bd9Sstevel@tonic-gate size_t pagehash_sz; 8247c478bd9Sstevel@tonic-gate size_t pp_sz; 8257c478bd9Sstevel@tonic-gate uintptr_t va; 8267c478bd9Sstevel@tonic-gate size_t len; 8277c478bd9Sstevel@tonic-gate uint_t prot; 8287c478bd9Sstevel@tonic-gate pfn_t pfn; 8297c478bd9Sstevel@tonic-gate int memblocks; 8307c478bd9Sstevel@tonic-gate caddr_t pagecolor_mem; 8317c478bd9Sstevel@tonic-gate size_t pagecolor_memsz; 8327c478bd9Sstevel@tonic-gate caddr_t page_ctrs_mem; 8337c478bd9Sstevel@tonic-gate size_t page_ctrs_size; 8347c478bd9Sstevel@tonic-gate struct memlist *current; 8357c478bd9Sstevel@tonic-gate extern void startup_build_mem_nodes(struct memlist *); 8367c478bd9Sstevel@tonic-gate 8377c478bd9Sstevel@tonic-gate /* XX64 fix these - they should be in include files */ 8387c478bd9Sstevel@tonic-gate extern size_t page_coloring_init(uint_t, int, int); 8397c478bd9Sstevel@tonic-gate extern void page_coloring_setup(caddr_t); 8407c478bd9Sstevel@tonic-gate 8417c478bd9Sstevel@tonic-gate PRM_POINT("startup_memlist() starting..."); 8427c478bd9Sstevel@tonic-gate 8437c478bd9Sstevel@tonic-gate /* 8447c478bd9Sstevel@tonic-gate * Use leftover large page nucleus text/data space for loadable modules. 8457c478bd9Sstevel@tonic-gate * Use at most MODTEXT/MODDATA. 8467c478bd9Sstevel@tonic-gate */ 847*ae115bc7Smrj len = kbm_nucleus_size; 848*ae115bc7Smrj ASSERT(len > MMU_PAGESIZE); 8497c478bd9Sstevel@tonic-gate 850*ae115bc7Smrj moddata = (caddr_t)ROUND_UP_PAGE(e_data); 851*ae115bc7Smrj e_moddata = (caddr_t)P2ROUNDUP((uintptr_t)e_data, (uintptr_t)len); 852*ae115bc7Smrj if (e_moddata - moddata > MODDATA) 853*ae115bc7Smrj e_moddata = moddata + MODDATA; 8547c478bd9Sstevel@tonic-gate 855*ae115bc7Smrj modtext = (caddr_t)ROUND_UP_PAGE(e_text); 856*ae115bc7Smrj e_modtext = (caddr_t)P2ROUNDUP((uintptr_t)e_text, (uintptr_t)len); 857*ae115bc7Smrj if (e_modtext - modtext > MODTEXT) 858*ae115bc7Smrj e_modtext = modtext + MODTEXT; 8597c478bd9Sstevel@tonic-gate 8607c478bd9Sstevel@tonic-gate econtig = e_moddata; 8617c478bd9Sstevel@tonic-gate 8627c478bd9Sstevel@tonic-gate PRM_DEBUG(modtext); 8637c478bd9Sstevel@tonic-gate PRM_DEBUG(e_modtext); 8647c478bd9Sstevel@tonic-gate PRM_DEBUG(moddata); 8657c478bd9Sstevel@tonic-gate PRM_DEBUG(e_moddata); 8667c478bd9Sstevel@tonic-gate PRM_DEBUG(econtig); 8677c478bd9Sstevel@tonic-gate 8687c478bd9Sstevel@tonic-gate /* 869*ae115bc7Smrj * Examine the boot loader physical memory map to find out: 8707c478bd9Sstevel@tonic-gate * - total memory in system - physinstalled 8717c478bd9Sstevel@tonic-gate * - the max physical address - physmax 872*ae115bc7Smrj * - the number of discontiguous segments of memory. 8737c478bd9Sstevel@tonic-gate */ 8747c478bd9Sstevel@tonic-gate if (prom_debug) 875*ae115bc7Smrj print_memlist("boot physinstalled", 8767c478bd9Sstevel@tonic-gate bootops->boot_mem->physinstalled); 8777c478bd9Sstevel@tonic-gate installed_top_size(bootops->boot_mem->physinstalled, &physmax, 8787c478bd9Sstevel@tonic-gate &physinstalled, &memblocks); 8797c478bd9Sstevel@tonic-gate PRM_DEBUG(physmax); 8807c478bd9Sstevel@tonic-gate PRM_DEBUG(physinstalled); 8817c478bd9Sstevel@tonic-gate PRM_DEBUG(memblocks); 8827c478bd9Sstevel@tonic-gate 8837c478bd9Sstevel@tonic-gate /* 8847c478bd9Sstevel@tonic-gate * Initialize hat's mmu parameters. 8857c478bd9Sstevel@tonic-gate * Check for enforce-prot-exec in boot environment. It's used to 8867c478bd9Sstevel@tonic-gate * enable/disable support for the page table entry NX bit. 8877c478bd9Sstevel@tonic-gate * The default is to enforce PROT_EXEC on processors that support NX. 8887c478bd9Sstevel@tonic-gate * Boot seems to round up the "len", but 8 seems to be big enough. 8897c478bd9Sstevel@tonic-gate */ 8907c478bd9Sstevel@tonic-gate mmu_init(); 8917c478bd9Sstevel@tonic-gate 8927c478bd9Sstevel@tonic-gate #ifdef __i386 8937c478bd9Sstevel@tonic-gate /* 8947c478bd9Sstevel@tonic-gate * physmax is lowered if there is more memory than can be 8957c478bd9Sstevel@tonic-gate * physically addressed in 32 bit (PAE/non-PAE) modes. 8967c478bd9Sstevel@tonic-gate */ 8977c478bd9Sstevel@tonic-gate if (mmu.pae_hat) { 8987c478bd9Sstevel@tonic-gate if (PFN_ABOVE64G(physmax)) { 8997c478bd9Sstevel@tonic-gate physinstalled -= (physmax - (PFN_64G - 1)); 9007c478bd9Sstevel@tonic-gate physmax = PFN_64G - 1; 9017c478bd9Sstevel@tonic-gate } 9027c478bd9Sstevel@tonic-gate } else { 9037c478bd9Sstevel@tonic-gate if (PFN_ABOVE4G(physmax)) { 9047c478bd9Sstevel@tonic-gate physinstalled -= (physmax - (PFN_4G - 1)); 9057c478bd9Sstevel@tonic-gate physmax = PFN_4G - 1; 9067c478bd9Sstevel@tonic-gate } 9077c478bd9Sstevel@tonic-gate } 9087c478bd9Sstevel@tonic-gate #endif 9097c478bd9Sstevel@tonic-gate 9107c478bd9Sstevel@tonic-gate startup_build_mem_nodes(bootops->boot_mem->physinstalled); 9117c478bd9Sstevel@tonic-gate 9127c478bd9Sstevel@tonic-gate if (BOP_GETPROPLEN(bootops, "enforce-prot-exec") >= 0) { 9137c478bd9Sstevel@tonic-gate int len = BOP_GETPROPLEN(bootops, "enforce-prot-exec"); 9147c478bd9Sstevel@tonic-gate char value[8]; 9157c478bd9Sstevel@tonic-gate 9167c478bd9Sstevel@tonic-gate if (len < 8) 9177c478bd9Sstevel@tonic-gate (void) BOP_GETPROP(bootops, "enforce-prot-exec", value); 9187c478bd9Sstevel@tonic-gate else 9197c478bd9Sstevel@tonic-gate (void) strcpy(value, ""); 9207c478bd9Sstevel@tonic-gate if (strcmp(value, "off") == 0) 9217c478bd9Sstevel@tonic-gate mmu.pt_nx = 0; 9227c478bd9Sstevel@tonic-gate } 9237c478bd9Sstevel@tonic-gate PRM_DEBUG(mmu.pt_nx); 9247c478bd9Sstevel@tonic-gate 9257c478bd9Sstevel@tonic-gate /* 9267c478bd9Sstevel@tonic-gate * We will need page_t's for every page in the system, except for 9277c478bd9Sstevel@tonic-gate * memory mapped at or above above the start of the kernel text segment. 9287c478bd9Sstevel@tonic-gate * 9297c478bd9Sstevel@tonic-gate * pages above e_modtext are attributed to kernel debugger (obp_pages) 9307c478bd9Sstevel@tonic-gate */ 9317c478bd9Sstevel@tonic-gate npages = physinstalled - 1; /* avail_filter() skips page 0, so "- 1" */ 9327c478bd9Sstevel@tonic-gate obp_pages = 0; 9337c478bd9Sstevel@tonic-gate va = KERNEL_TEXT; 934*ae115bc7Smrj while (kbm_probe(&va, &len, &pfn, &prot) != 0) { 9357c478bd9Sstevel@tonic-gate npages -= len >> MMU_PAGESHIFT; 9367c478bd9Sstevel@tonic-gate if (va >= (uintptr_t)e_moddata) 9377c478bd9Sstevel@tonic-gate obp_pages += len >> MMU_PAGESHIFT; 9387c478bd9Sstevel@tonic-gate va += len; 9397c478bd9Sstevel@tonic-gate } 9407c478bd9Sstevel@tonic-gate PRM_DEBUG(npages); 9417c478bd9Sstevel@tonic-gate PRM_DEBUG(obp_pages); 9427c478bd9Sstevel@tonic-gate 9437c478bd9Sstevel@tonic-gate /* 9447c478bd9Sstevel@tonic-gate * If physmem is patched to be non-zero, use it instead of 9457c478bd9Sstevel@tonic-gate * the computed value unless it is larger than the real 9467c478bd9Sstevel@tonic-gate * amount of memory on hand. 9477c478bd9Sstevel@tonic-gate */ 9484944b02eSkchow if (physmem == 0 || physmem > npages) { 9497c478bd9Sstevel@tonic-gate physmem = npages; 9504944b02eSkchow } else if (physmem < npages) { 951f53ad214Skchow orig_npages = npages; 9527c478bd9Sstevel@tonic-gate npages = physmem; 9534944b02eSkchow } 9547c478bd9Sstevel@tonic-gate PRM_DEBUG(physmem); 9557c478bd9Sstevel@tonic-gate 9567c478bd9Sstevel@tonic-gate /* 9577c478bd9Sstevel@tonic-gate * We now compute the sizes of all the initial allocations for 9587c478bd9Sstevel@tonic-gate * structures the kernel needs in order do kmem_alloc(). These 9597c478bd9Sstevel@tonic-gate * include: 9607c478bd9Sstevel@tonic-gate * memsegs 9617c478bd9Sstevel@tonic-gate * memlists 9627c478bd9Sstevel@tonic-gate * page hash table 9637c478bd9Sstevel@tonic-gate * page_t's 9647c478bd9Sstevel@tonic-gate * page coloring data structs 9657c478bd9Sstevel@tonic-gate */ 9667c478bd9Sstevel@tonic-gate memseg_sz = sizeof (struct memseg) * (memblocks + POSS_NEW_FRAGMENTS); 9677c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(memseg_base, memseg_sz); 9687c478bd9Sstevel@tonic-gate PRM_DEBUG(memseg_sz); 9697c478bd9Sstevel@tonic-gate 9707c478bd9Sstevel@tonic-gate /* 971*ae115bc7Smrj * Reserve space for memlists. There's no real good way to know exactly 972*ae115bc7Smrj * how much room we'll need, but this should be a good upper bound. 9737c478bd9Sstevel@tonic-gate */ 9747c478bd9Sstevel@tonic-gate memlist_sz = ROUND_UP_PAGE(2 * sizeof (struct memlist) * 9757c478bd9Sstevel@tonic-gate (memblocks + POSS_NEW_FRAGMENTS)); 9767c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(memlist, memlist_sz); 9777c478bd9Sstevel@tonic-gate PRM_DEBUG(memlist_sz); 9787c478bd9Sstevel@tonic-gate 9797c478bd9Sstevel@tonic-gate /* 9807c478bd9Sstevel@tonic-gate * The page structure hash table size is a power of 2 9817c478bd9Sstevel@tonic-gate * such that the average hash chain length is PAGE_HASHAVELEN. 9827c478bd9Sstevel@tonic-gate */ 9837c478bd9Sstevel@tonic-gate page_hashsz = npages / PAGE_HASHAVELEN; 9847c478bd9Sstevel@tonic-gate page_hashsz = 1 << highbit(page_hashsz); 9857c478bd9Sstevel@tonic-gate pagehash_sz = sizeof (struct page *) * page_hashsz; 9867c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(page_hash, pagehash_sz); 9877c478bd9Sstevel@tonic-gate PRM_DEBUG(pagehash_sz); 9887c478bd9Sstevel@tonic-gate 9897c478bd9Sstevel@tonic-gate /* 990*ae115bc7Smrj * Set aside room for the page structures themselves. 9917c478bd9Sstevel@tonic-gate */ 992*ae115bc7Smrj PRM_DEBUG(npages); 993*ae115bc7Smrj pp_sz = sizeof (struct page) * npages; 9947c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(pp_base, pp_sz); 9957c478bd9Sstevel@tonic-gate PRM_DEBUG(pp_sz); 9967c478bd9Sstevel@tonic-gate 9977c478bd9Sstevel@tonic-gate /* 9987c478bd9Sstevel@tonic-gate * determine l2 cache info and memory size for page coloring 9997c478bd9Sstevel@tonic-gate */ 10007c478bd9Sstevel@tonic-gate (void) getl2cacheinfo(CPU, 10017c478bd9Sstevel@tonic-gate &l2cache_sz, &l2cache_linesz, &l2cache_assoc); 10027c478bd9Sstevel@tonic-gate pagecolor_memsz = 10037c478bd9Sstevel@tonic-gate page_coloring_init(l2cache_sz, l2cache_linesz, l2cache_assoc); 10047c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(pagecolor_mem, pagecolor_memsz); 10057c478bd9Sstevel@tonic-gate PRM_DEBUG(pagecolor_memsz); 10067c478bd9Sstevel@tonic-gate 10077c478bd9Sstevel@tonic-gate page_ctrs_size = page_ctrs_sz(); 10087c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(page_ctrs_mem, page_ctrs_size); 10097c478bd9Sstevel@tonic-gate PRM_DEBUG(page_ctrs_size); 10107c478bd9Sstevel@tonic-gate 1011*ae115bc7Smrj #if defined(__amd64) 10127c478bd9Sstevel@tonic-gate valloc_sz = ROUND_UP_LPAGE(valloc_sz); 1013*ae115bc7Smrj valloc_base = VALLOC_BASE; 1014*ae115bc7Smrj #else /* __i386 */ 1015*ae115bc7Smrj valloc_base = (uintptr_t)(MISC_VA_BASE - valloc_sz); 1016*ae115bc7Smrj valloc_base = P2ALIGN(valloc_base, mmu.level_size[1]); 1017*ae115bc7Smrj #endif /* __i386 */ 10187c478bd9Sstevel@tonic-gate PRM_DEBUG(valloc_base); 1019*ae115bc7Smrj 1020*ae115bc7Smrj /* 1021*ae115bc7Smrj * do all the initial allocations 1022*ae115bc7Smrj */ 1023*ae115bc7Smrj perform_allocations(); 1024*ae115bc7Smrj 1025*ae115bc7Smrj /* 1026*ae115bc7Smrj * Build phys_install and phys_avail in kernel memspace. 1027*ae115bc7Smrj * - phys_install should be all memory in the system. 1028*ae115bc7Smrj * - phys_avail is phys_install minus any memory mapped before this 1029*ae115bc7Smrj * point above KERNEL_TEXT. 1030*ae115bc7Smrj */ 1031*ae115bc7Smrj current = phys_install = memlist; 1032*ae115bc7Smrj copy_memlist_filter(bootops->boot_mem->physinstalled, ¤t, NULL); 1033*ae115bc7Smrj if ((caddr_t)current > (caddr_t)memlist + memlist_sz) 1034*ae115bc7Smrj panic("physinstalled was too big!"); 1035*ae115bc7Smrj if (prom_debug) 1036*ae115bc7Smrj print_memlist("phys_install", phys_install); 1037*ae115bc7Smrj 1038*ae115bc7Smrj phys_avail = current; 1039*ae115bc7Smrj PRM_POINT("Building phys_avail:\n"); 1040*ae115bc7Smrj copy_memlist_filter(bootops->boot_mem->physinstalled, ¤t, 1041*ae115bc7Smrj avail_filter); 1042*ae115bc7Smrj if ((caddr_t)current > (caddr_t)memlist + memlist_sz) 1043*ae115bc7Smrj panic("physavail was too big!"); 1044*ae115bc7Smrj if (prom_debug) 1045*ae115bc7Smrj print_memlist("phys_avail", phys_avail); 1046*ae115bc7Smrj 1047*ae115bc7Smrj /* 1048*ae115bc7Smrj * setup page coloring 1049*ae115bc7Smrj */ 1050*ae115bc7Smrj page_coloring_setup(pagecolor_mem); 1051*ae115bc7Smrj page_lock_init(); /* currently a no-op */ 1052*ae115bc7Smrj 1053*ae115bc7Smrj /* 1054*ae115bc7Smrj * free page list counters 1055*ae115bc7Smrj */ 1056*ae115bc7Smrj (void) page_ctrs_alloc(page_ctrs_mem); 1057*ae115bc7Smrj 1058*ae115bc7Smrj /* 1059*ae115bc7Smrj * Initialize the page structures from the memory lists. 1060*ae115bc7Smrj */ 1061*ae115bc7Smrj availrmem_initial = availrmem = freemem = 0; 1062*ae115bc7Smrj PRM_POINT("Calling kphysm_init()..."); 1063*ae115bc7Smrj npages = kphysm_init(pp_base, npages); 1064*ae115bc7Smrj PRM_POINT("kphysm_init() done"); 1065*ae115bc7Smrj PRM_DEBUG(npages); 1066*ae115bc7Smrj 1067*ae115bc7Smrj init_debug_info(); 1068*ae115bc7Smrj 1069*ae115bc7Smrj /* 1070*ae115bc7Smrj * Now that page_t's have been initialized, remove all the 1071*ae115bc7Smrj * initial allocation pages from the kernel free page lists. 1072*ae115bc7Smrj */ 1073*ae115bc7Smrj boot_mapin((caddr_t)valloc_base, valloc_sz); 1074*ae115bc7Smrj boot_mapin((caddr_t)GDT_VA, MMU_PAGESIZE); 1075*ae115bc7Smrj boot_mapin((caddr_t)DEBUG_INFO_VA, MMU_PAGESIZE); 1076*ae115bc7Smrj PRM_POINT("startup_memlist() done"); 1077*ae115bc7Smrj 1078*ae115bc7Smrj PRM_DEBUG(valloc_sz); 1079*ae115bc7Smrj } 1080*ae115bc7Smrj 1081*ae115bc7Smrj /* 1082*ae115bc7Smrj * Layout the kernel's part of address space and initialize kmem allocator. 1083*ae115bc7Smrj */ 1084*ae115bc7Smrj static void 1085*ae115bc7Smrj startup_kmem(void) 1086*ae115bc7Smrj { 1087*ae115bc7Smrj PRM_POINT("startup_kmem() starting..."); 10887c478bd9Sstevel@tonic-gate 10897c478bd9Sstevel@tonic-gate #if defined(__amd64) 10907c478bd9Sstevel@tonic-gate if (eprom_kernelbase && eprom_kernelbase != KERNELBASE) 10917c478bd9Sstevel@tonic-gate cmn_err(CE_NOTE, "!kernelbase cannot be changed on 64-bit " 10927c478bd9Sstevel@tonic-gate "systems."); 10937c478bd9Sstevel@tonic-gate kernelbase = (uintptr_t)KERNELBASE; 10947c478bd9Sstevel@tonic-gate core_base = (uintptr_t)COREHEAP_BASE; 1095*ae115bc7Smrj core_size = (size_t)MISC_VA_BASE - COREHEAP_BASE; 10967c478bd9Sstevel@tonic-gate #else /* __i386 */ 10977c478bd9Sstevel@tonic-gate /* 10987c478bd9Sstevel@tonic-gate * We configure kernelbase based on: 10997c478bd9Sstevel@tonic-gate * 11007c478bd9Sstevel@tonic-gate * 1. user specified kernelbase via eeprom command. Value cannot exceed 11017c478bd9Sstevel@tonic-gate * KERNELBASE_MAX. we large page align eprom_kernelbase 11027c478bd9Sstevel@tonic-gate * 11037c478bd9Sstevel@tonic-gate * 2. Default to KERNELBASE and adjust to 2X less the size for page_t. 11047c478bd9Sstevel@tonic-gate * On large memory systems we must lower kernelbase to allow 11057c478bd9Sstevel@tonic-gate * enough room for page_t's for all of memory. 11067c478bd9Sstevel@tonic-gate * 11077c478bd9Sstevel@tonic-gate * The value set here, might be changed a little later. 11087c478bd9Sstevel@tonic-gate */ 11097c478bd9Sstevel@tonic-gate if (eprom_kernelbase) { 11107c478bd9Sstevel@tonic-gate kernelbase = eprom_kernelbase & mmu.level_mask[1]; 11117c478bd9Sstevel@tonic-gate if (kernelbase > KERNELBASE_MAX) 11127c478bd9Sstevel@tonic-gate kernelbase = KERNELBASE_MAX; 11137c478bd9Sstevel@tonic-gate } else { 11147c478bd9Sstevel@tonic-gate kernelbase = (uintptr_t)KERNELBASE; 11157c478bd9Sstevel@tonic-gate kernelbase -= ROUND_UP_4MEG(2 * valloc_sz); 11167c478bd9Sstevel@tonic-gate } 11177c478bd9Sstevel@tonic-gate ASSERT((kernelbase & mmu.level_offset[1]) == 0); 1118*ae115bc7Smrj core_base = valloc_base; 11197c478bd9Sstevel@tonic-gate core_size = 0; 1120*ae115bc7Smrj #endif /* __i386 */ 11217c478bd9Sstevel@tonic-gate 11227c478bd9Sstevel@tonic-gate PRM_DEBUG(core_base); 11237c478bd9Sstevel@tonic-gate PRM_DEBUG(core_size); 1124*ae115bc7Smrj PRM_DEBUG(kernelbase); 11257c478bd9Sstevel@tonic-gate 11267c478bd9Sstevel@tonic-gate /* 11277c478bd9Sstevel@tonic-gate * At this point, we can only use a portion of the kernelheap that 1128*ae115bc7Smrj * will be available after we boot. 32-bit systems have this 1129*ae115bc7Smrj * limitation. 11307c478bd9Sstevel@tonic-gate * 11317c478bd9Sstevel@tonic-gate * On 32-bit systems we have to leave room to place segmap below 11327c478bd9Sstevel@tonic-gate * the heap. We don't yet know how large segmap will be, so we 11337c478bd9Sstevel@tonic-gate * have to be very conservative. 1134*ae115bc7Smrj * 1135*ae115bc7Smrj * On 64 bit systems there should be LOTS of room so just use 1136*ae115bc7Smrj * the next 4Gig below core_base. 11377c478bd9Sstevel@tonic-gate */ 11387c478bd9Sstevel@tonic-gate #if defined(__amd64) 1139*ae115bc7Smrj 1140*ae115bc7Smrj boot_kernelheap = (caddr_t)core_base - FOURGB; 11417c478bd9Sstevel@tonic-gate segmap_reserved = 0; 11427c478bd9Sstevel@tonic-gate 11437c478bd9Sstevel@tonic-gate #else /* __i386 */ 1144*ae115bc7Smrj 11457c478bd9Sstevel@tonic-gate segkp_fromheap = 1; 11467c478bd9Sstevel@tonic-gate segmap_reserved = ROUND_UP_LPAGE(MAX(segmapsize, SEGMAPMAX)); 1147*ae115bc7Smrj boot_kernelheap = 1148*ae115bc7Smrj (caddr_t)ROUND_UP_LPAGE(kernelbase) + segmap_reserved; 1149*ae115bc7Smrj 1150*ae115bc7Smrj #endif /* __i386 */ 11517c478bd9Sstevel@tonic-gate PRM_DEBUG(boot_kernelheap); 11527c478bd9Sstevel@tonic-gate ekernelheap = (char *)core_base; 1153*ae115bc7Smrj PRM_DEBUG(ekernelheap); 1154*ae115bc7Smrj kernelheap = boot_kernelheap; 11557c478bd9Sstevel@tonic-gate 11567c478bd9Sstevel@tonic-gate /* 11577c478bd9Sstevel@tonic-gate * If segmap is too large we can push the bottom of the kernel heap 11587c478bd9Sstevel@tonic-gate * higher than the base. Or worse, it could exceed the top of the 11597c478bd9Sstevel@tonic-gate * VA space entirely, causing it to wrap around. 11607c478bd9Sstevel@tonic-gate */ 11617c478bd9Sstevel@tonic-gate if (kernelheap >= ekernelheap || (uintptr_t)kernelheap < kernelbase) 11627c478bd9Sstevel@tonic-gate panic("too little memory available for kernelheap," 11637c478bd9Sstevel@tonic-gate " use a different kernelbase"); 11647c478bd9Sstevel@tonic-gate 11657c478bd9Sstevel@tonic-gate /* 11667c478bd9Sstevel@tonic-gate * Now that we know the real value of kernelbase, 11677c478bd9Sstevel@tonic-gate * update variables that were initialized with a value of 11687c478bd9Sstevel@tonic-gate * KERNELBASE (in common/conf/param.c). 11697c478bd9Sstevel@tonic-gate * 11707c478bd9Sstevel@tonic-gate * XXX The problem with this sort of hackery is that the 11717c478bd9Sstevel@tonic-gate * compiler just may feel like putting the const declarations 11727c478bd9Sstevel@tonic-gate * (in param.c) into the .text section. Perhaps they should 11737c478bd9Sstevel@tonic-gate * just be declared as variables there? 11747c478bd9Sstevel@tonic-gate */ 11757c478bd9Sstevel@tonic-gate 11767c478bd9Sstevel@tonic-gate #if defined(__amd64) 11777c478bd9Sstevel@tonic-gate ASSERT(_kernelbase == KERNELBASE); 11787c478bd9Sstevel@tonic-gate ASSERT(_userlimit == USERLIMIT); 11797c478bd9Sstevel@tonic-gate #else 11807c478bd9Sstevel@tonic-gate *(uintptr_t *)&_kernelbase = kernelbase; 11817c478bd9Sstevel@tonic-gate *(uintptr_t *)&_userlimit = kernelbase; 11827c478bd9Sstevel@tonic-gate *(uintptr_t *)&_userlimit32 = _userlimit; 11837c478bd9Sstevel@tonic-gate #endif 11847c478bd9Sstevel@tonic-gate PRM_DEBUG(_kernelbase); 11857c478bd9Sstevel@tonic-gate PRM_DEBUG(_userlimit); 11867c478bd9Sstevel@tonic-gate PRM_DEBUG(_userlimit32); 11877c478bd9Sstevel@tonic-gate 11887c478bd9Sstevel@tonic-gate /* 11897c478bd9Sstevel@tonic-gate * Initialize the kernel heap. Note 3rd argument must be > 1st. 11907c478bd9Sstevel@tonic-gate */ 1191*ae115bc7Smrj kernelheap_init(boot_kernelheap, ekernelheap, 1192*ae115bc7Smrj boot_kernelheap + MMU_PAGESIZE, 1193*ae115bc7Smrj (void *)core_base, (void *)(core_base + core_size)); 11947c478bd9Sstevel@tonic-gate 11957c478bd9Sstevel@tonic-gate /* 11967c478bd9Sstevel@tonic-gate * Initialize kernel memory allocator. 11977c478bd9Sstevel@tonic-gate */ 11987c478bd9Sstevel@tonic-gate kmem_init(); 11997c478bd9Sstevel@tonic-gate 12007c478bd9Sstevel@tonic-gate /* 12017c478bd9Sstevel@tonic-gate * print this out early so that we know what's going on 12027c478bd9Sstevel@tonic-gate */ 12037c478bd9Sstevel@tonic-gate cmn_err(CE_CONT, "?features: %b\n", x86_feature, FMT_X86_FEATURE); 12047c478bd9Sstevel@tonic-gate 12057c478bd9Sstevel@tonic-gate /* 12067c478bd9Sstevel@tonic-gate * Initialize bp_mapin(). 12077c478bd9Sstevel@tonic-gate */ 12087c478bd9Sstevel@tonic-gate bp_init(MMU_PAGESIZE, HAT_STORECACHING_OK); 12097c478bd9Sstevel@tonic-gate 1210f53ad214Skchow /* 1211f53ad214Skchow * orig_npages is non-zero if physmem has been configured for less 1212f53ad214Skchow * than the available memory. 1213f53ad214Skchow */ 1214f53ad214Skchow if (orig_npages) { 1215f53ad214Skchow #ifdef __i386 1216f53ad214Skchow /* 1217f53ad214Skchow * use npages for physmem in case it has been temporarily 1218f53ad214Skchow * modified via /etc/system in kmem_init/mod_read_system_file. 1219f53ad214Skchow */ 1220f53ad214Skchow if (npages == PHYSMEM32) { 1221*ae115bc7Smrj cmn_err(CE_WARN, "!Due to 32-bit virtual" 1222f53ad214Skchow " address space limitations, limiting" 1223f53ad214Skchow " physmem to 0x%lx of 0x%lx available pages", 1224f53ad214Skchow npages, orig_npages); 1225f53ad214Skchow } else { 1226f53ad214Skchow cmn_err(CE_WARN, "!limiting physmem to 0x%lx of" 1227f53ad214Skchow " 0x%lx available pages", npages, orig_npages); 1228f53ad214Skchow } 1229f53ad214Skchow #else 1230f53ad214Skchow cmn_err(CE_WARN, "!limiting physmem to 0x%lx of" 1231f53ad214Skchow " 0x%lx available pages", npages, orig_npages); 1232f53ad214Skchow #endif 1233f53ad214Skchow } 12347c478bd9Sstevel@tonic-gate #if defined(__i386) 12357c478bd9Sstevel@tonic-gate if (eprom_kernelbase && (eprom_kernelbase != kernelbase)) 12367c478bd9Sstevel@tonic-gate cmn_err(CE_WARN, "kernelbase value, User specified 0x%lx, " 12377c478bd9Sstevel@tonic-gate "System using 0x%lx", 12387c478bd9Sstevel@tonic-gate (uintptr_t)eprom_kernelbase, (uintptr_t)kernelbase); 12397c478bd9Sstevel@tonic-gate #endif 12407c478bd9Sstevel@tonic-gate 12417c478bd9Sstevel@tonic-gate #ifdef KERNELBASE_ABI_MIN 12427c478bd9Sstevel@tonic-gate if (kernelbase < (uintptr_t)KERNELBASE_ABI_MIN) { 12437c478bd9Sstevel@tonic-gate cmn_err(CE_NOTE, "!kernelbase set to 0x%lx, system is not " 12447c478bd9Sstevel@tonic-gate "i386 ABI compliant.", (uintptr_t)kernelbase); 12457c478bd9Sstevel@tonic-gate } 12467c478bd9Sstevel@tonic-gate #endif 12477c478bd9Sstevel@tonic-gate 1248*ae115bc7Smrj PRM_POINT("startup_kmem() done"); 12497c478bd9Sstevel@tonic-gate } 12507c478bd9Sstevel@tonic-gate 12517c478bd9Sstevel@tonic-gate static void 12527c478bd9Sstevel@tonic-gate startup_modules(void) 12537c478bd9Sstevel@tonic-gate { 12547c478bd9Sstevel@tonic-gate unsigned int i; 12557c478bd9Sstevel@tonic-gate extern void prom_setup(void); 12567c478bd9Sstevel@tonic-gate 12577c478bd9Sstevel@tonic-gate PRM_POINT("startup_modules() starting..."); 12587c478bd9Sstevel@tonic-gate /* 12597c478bd9Sstevel@tonic-gate * Initialize ten-micro second timer so that drivers will 12607c478bd9Sstevel@tonic-gate * not get short changed in their init phase. This was 12617c478bd9Sstevel@tonic-gate * not getting called until clkinit which, on fast cpu's 12627c478bd9Sstevel@tonic-gate * caused the drv_usecwait to be way too short. 12637c478bd9Sstevel@tonic-gate */ 12647c478bd9Sstevel@tonic-gate microfind(); 12657c478bd9Sstevel@tonic-gate 12667c478bd9Sstevel@tonic-gate /* 12677c478bd9Sstevel@tonic-gate * Read the GMT lag from /etc/rtc_config. 12687c478bd9Sstevel@tonic-gate */ 1269*ae115bc7Smrj sgmtl(process_rtc_config_file()); 12707c478bd9Sstevel@tonic-gate 12717c478bd9Sstevel@tonic-gate /* 12727c478bd9Sstevel@tonic-gate * Calculate default settings of system parameters based upon 12737c478bd9Sstevel@tonic-gate * maxusers, yet allow to be overridden via the /etc/system file. 12747c478bd9Sstevel@tonic-gate */ 12757c478bd9Sstevel@tonic-gate param_calc(0); 12767c478bd9Sstevel@tonic-gate 12777c478bd9Sstevel@tonic-gate mod_setup(); 12787c478bd9Sstevel@tonic-gate 12797c478bd9Sstevel@tonic-gate /* 12807c478bd9Sstevel@tonic-gate * Initialize system parameters. 12817c478bd9Sstevel@tonic-gate */ 12827c478bd9Sstevel@tonic-gate param_init(); 12837c478bd9Sstevel@tonic-gate 12849acbbeafSnn /* 12859acbbeafSnn * Initialize the default brands 12869acbbeafSnn */ 12879acbbeafSnn brand_init(); 12889acbbeafSnn 12897c478bd9Sstevel@tonic-gate /* 12907c478bd9Sstevel@tonic-gate * maxmem is the amount of physical memory we're playing with. 12917c478bd9Sstevel@tonic-gate */ 12927c478bd9Sstevel@tonic-gate maxmem = physmem; 12937c478bd9Sstevel@tonic-gate 12947c478bd9Sstevel@tonic-gate /* 12957c478bd9Sstevel@tonic-gate * Initialize the hat layer. 12967c478bd9Sstevel@tonic-gate */ 12977c478bd9Sstevel@tonic-gate hat_init(); 12987c478bd9Sstevel@tonic-gate 12997c478bd9Sstevel@tonic-gate /* 13007c478bd9Sstevel@tonic-gate * Initialize segment management stuff. 13017c478bd9Sstevel@tonic-gate */ 13027c478bd9Sstevel@tonic-gate seg_init(); 13037c478bd9Sstevel@tonic-gate 13047c478bd9Sstevel@tonic-gate if (modload("fs", "specfs") == -1) 13057c478bd9Sstevel@tonic-gate halt("Can't load specfs"); 13067c478bd9Sstevel@tonic-gate 13077c478bd9Sstevel@tonic-gate if (modload("fs", "devfs") == -1) 13087c478bd9Sstevel@tonic-gate halt("Can't load devfs"); 13097c478bd9Sstevel@tonic-gate 1310facf4a8dSllai if (modload("fs", "dev") == -1) 1311facf4a8dSllai halt("Can't load dev"); 1312facf4a8dSllai 131345916cd2Sjpk (void) modloadonly("sys", "lbl_edition"); 131445916cd2Sjpk 13157c478bd9Sstevel@tonic-gate dispinit(); 13167c478bd9Sstevel@tonic-gate 13177c478bd9Sstevel@tonic-gate /* 13187c478bd9Sstevel@tonic-gate * This is needed here to initialize hw_serial[] for cluster booting. 13197c478bd9Sstevel@tonic-gate */ 13207c478bd9Sstevel@tonic-gate if ((i = modload("misc", "sysinit")) != (unsigned int)-1) 13217c478bd9Sstevel@tonic-gate (void) modunload(i); 13227c478bd9Sstevel@tonic-gate else 13237c478bd9Sstevel@tonic-gate cmn_err(CE_CONT, "sysinit load failed"); 13247c478bd9Sstevel@tonic-gate 13257c478bd9Sstevel@tonic-gate /* Read cluster configuration data. */ 13267c478bd9Sstevel@tonic-gate clconf_init(); 13277c478bd9Sstevel@tonic-gate 13287c478bd9Sstevel@tonic-gate /* 13297c478bd9Sstevel@tonic-gate * Create a kernel device tree. First, create rootnex and 13307c478bd9Sstevel@tonic-gate * then invoke bus specific code to probe devices. 13317c478bd9Sstevel@tonic-gate */ 13327c478bd9Sstevel@tonic-gate setup_ddi(); 13337aec1d6eScindi 13347aec1d6eScindi /* 13357aec1d6eScindi * Set up the CPU module subsystem. Modifies the device tree, so it 13367aec1d6eScindi * must be done after setup_ddi(). 13377aec1d6eScindi */ 13387aec1d6eScindi cmi_init(); 13397aec1d6eScindi 13407aec1d6eScindi /* 13417aec1d6eScindi * Initialize the MCA handlers 13427aec1d6eScindi */ 13437aec1d6eScindi if (x86_feature & X86_MCA) 13447aec1d6eScindi cmi_mca_init(); 13457aec1d6eScindi 13467c478bd9Sstevel@tonic-gate /* 13477c478bd9Sstevel@tonic-gate * Fake a prom tree such that /dev/openprom continues to work 13487c478bd9Sstevel@tonic-gate */ 1349*ae115bc7Smrj PRM_POINT("startup_modules: calling prom_setup..."); 13507c478bd9Sstevel@tonic-gate prom_setup(); 1351*ae115bc7Smrj PRM_POINT("startup_modules: done"); 13527c478bd9Sstevel@tonic-gate 13537c478bd9Sstevel@tonic-gate /* 13547c478bd9Sstevel@tonic-gate * Load all platform specific modules 13557c478bd9Sstevel@tonic-gate */ 1356*ae115bc7Smrj PRM_POINT("startup_modules: calling psm_modload..."); 13577c478bd9Sstevel@tonic-gate psm_modload(); 13587c478bd9Sstevel@tonic-gate 13597c478bd9Sstevel@tonic-gate PRM_POINT("startup_modules() done"); 13607c478bd9Sstevel@tonic-gate } 13617c478bd9Sstevel@tonic-gate 1362*ae115bc7Smrj /* 1363*ae115bc7Smrj * claim a "setaside" boot page for use in the kernel 1364*ae115bc7Smrj */ 1365*ae115bc7Smrj page_t * 1366*ae115bc7Smrj boot_claim_page(pfn_t pfn) 13677c478bd9Sstevel@tonic-gate { 1368*ae115bc7Smrj page_t *pp; 13697c478bd9Sstevel@tonic-gate 1370*ae115bc7Smrj pp = page_numtopp_nolock(pfn); 1371*ae115bc7Smrj ASSERT(pp != NULL); 13727c478bd9Sstevel@tonic-gate 1373*ae115bc7Smrj if (PP_ISBOOTPAGES(pp)) { 1374*ae115bc7Smrj if (pp->p_next != NULL) 1375*ae115bc7Smrj pp->p_next->p_prev = pp->p_prev; 1376*ae115bc7Smrj if (pp->p_prev == NULL) 1377*ae115bc7Smrj bootpages = pp->p_next; 1378*ae115bc7Smrj else 1379*ae115bc7Smrj pp->p_prev->p_next = pp->p_next; 1380*ae115bc7Smrj } else { 1381*ae115bc7Smrj /* 1382*ae115bc7Smrj * htable_attach() expects a base pagesize page 1383*ae115bc7Smrj */ 1384*ae115bc7Smrj if (pp->p_szc != 0) 1385*ae115bc7Smrj page_boot_demote(pp); 1386*ae115bc7Smrj pp = page_numtopp(pfn, SE_EXCL); 1387*ae115bc7Smrj } 1388*ae115bc7Smrj return (pp); 13897c478bd9Sstevel@tonic-gate } 13907c478bd9Sstevel@tonic-gate 13917c478bd9Sstevel@tonic-gate /* 13927c478bd9Sstevel@tonic-gate * Walk through the pagetables looking for pages mapped in by boot. If the 13937c478bd9Sstevel@tonic-gate * setaside flag is set the pages are expected to be returned to the 13947c478bd9Sstevel@tonic-gate * kernel later in boot, so we add them to the bootpages list. 13957c478bd9Sstevel@tonic-gate */ 13967c478bd9Sstevel@tonic-gate static void 13977c478bd9Sstevel@tonic-gate protect_boot_range(uintptr_t low, uintptr_t high, int setaside) 13987c478bd9Sstevel@tonic-gate { 13997c478bd9Sstevel@tonic-gate uintptr_t va = low; 14007c478bd9Sstevel@tonic-gate size_t len; 14017c478bd9Sstevel@tonic-gate uint_t prot; 14027c478bd9Sstevel@tonic-gate pfn_t pfn; 14037c478bd9Sstevel@tonic-gate page_t *pp; 14047c478bd9Sstevel@tonic-gate pgcnt_t boot_protect_cnt = 0; 14057c478bd9Sstevel@tonic-gate 1406*ae115bc7Smrj while (kbm_probe(&va, &len, &pfn, &prot) != 0 && va < high) { 14077c478bd9Sstevel@tonic-gate if (va + len >= high) 14087c478bd9Sstevel@tonic-gate panic("0x%lx byte mapping at 0x%p exceeds boot's " 14097c478bd9Sstevel@tonic-gate "legal range.", len, (void *)va); 14107c478bd9Sstevel@tonic-gate 14117c478bd9Sstevel@tonic-gate while (len > 0) { 14127c478bd9Sstevel@tonic-gate pp = page_numtopp_alloc(pfn); 14137c478bd9Sstevel@tonic-gate if (pp != NULL) { 14147c478bd9Sstevel@tonic-gate if (setaside == 0) 14157c478bd9Sstevel@tonic-gate panic("Unexpected mapping by boot. " 14167c478bd9Sstevel@tonic-gate "addr=%p pfn=%lx\n", 14177c478bd9Sstevel@tonic-gate (void *)va, pfn); 14187c478bd9Sstevel@tonic-gate 14197c478bd9Sstevel@tonic-gate pp->p_next = bootpages; 1420*ae115bc7Smrj pp->p_prev = NULL; 1421*ae115bc7Smrj PP_SETBOOTPAGES(pp); 1422*ae115bc7Smrj if (bootpages != NULL) { 1423*ae115bc7Smrj bootpages->p_prev = pp; 1424*ae115bc7Smrj } 14257c478bd9Sstevel@tonic-gate bootpages = pp; 14267c478bd9Sstevel@tonic-gate ++boot_protect_cnt; 14277c478bd9Sstevel@tonic-gate } 14287c478bd9Sstevel@tonic-gate 14297c478bd9Sstevel@tonic-gate ++pfn; 14307c478bd9Sstevel@tonic-gate len -= MMU_PAGESIZE; 14317c478bd9Sstevel@tonic-gate va += MMU_PAGESIZE; 14327c478bd9Sstevel@tonic-gate } 14337c478bd9Sstevel@tonic-gate } 14347c478bd9Sstevel@tonic-gate PRM_DEBUG(boot_protect_cnt); 14357c478bd9Sstevel@tonic-gate } 14367c478bd9Sstevel@tonic-gate 1437*ae115bc7Smrj /* 1438*ae115bc7Smrj * Finish initializing the VM system, now that we are no longer 1439*ae115bc7Smrj * relying on the boot time memory allocators. 1440*ae115bc7Smrj */ 14417c478bd9Sstevel@tonic-gate static void 14427c478bd9Sstevel@tonic-gate startup_vm(void) 14437c478bd9Sstevel@tonic-gate { 14447c478bd9Sstevel@tonic-gate struct segmap_crargs a; 14457c478bd9Sstevel@tonic-gate 1446ec25b48fSsusans extern int use_brk_lpg, use_stk_lpg; 1447beb1bda0Sdavemq 14487c478bd9Sstevel@tonic-gate PRM_POINT("startup_vm() starting..."); 14497c478bd9Sstevel@tonic-gate 14507c478bd9Sstevel@tonic-gate /* 1451*ae115bc7Smrj * Establish the final size of the kernel's heap, size of segmap, 1452*ae115bc7Smrj * segkp, etc. 14537c478bd9Sstevel@tonic-gate */ 14547c478bd9Sstevel@tonic-gate 14557c478bd9Sstevel@tonic-gate #if defined(__amd64) 14567c478bd9Sstevel@tonic-gate 14577c478bd9Sstevel@tonic-gate /* 1458*ae115bc7Smrj * Check if there is enough virtual address space in KPM region to 1459*ae115bc7Smrj * map physmax. 14607c478bd9Sstevel@tonic-gate */ 1461*ae115bc7Smrj kpm_vbase = (caddr_t)(uintptr_t)SEGKPM_BASE; 1462*ae115bc7Smrj kpm_size = 0; 1463*ae115bc7Smrj if (kpm_desired) { 1464*ae115bc7Smrj kpm_size = ROUND_UP_LPAGE(mmu_ptob(physmax + 1)); 1465*ae115bc7Smrj if ((uintptr_t)kpm_vbase + kpm_size > (uintptr_t)VALLOC_BASE) { 1466*ae115bc7Smrj kpm_size = 0; 1467*ae115bc7Smrj kpm_desired = 0; 1468*ae115bc7Smrj } 1469*ae115bc7Smrj } 14707c478bd9Sstevel@tonic-gate 1471*ae115bc7Smrj PRM_DEBUG(kpm_size); 1472*ae115bc7Smrj PRM_DEBUG(kpm_vbase); 14737c478bd9Sstevel@tonic-gate 14747c478bd9Sstevel@tonic-gate /* 1475*ae115bc7Smrj * By default we create a seg_kp in 64 bit kernels, it's a little 1476*ae115bc7Smrj * faster to access than embedding it in the heap. 14777c478bd9Sstevel@tonic-gate */ 1478*ae115bc7Smrj segkp_base = (caddr_t)valloc_base + valloc_sz; 14797c478bd9Sstevel@tonic-gate if (!segkp_fromheap) { 14807c478bd9Sstevel@tonic-gate size_t sz = mmu_ptob(segkpsize); 14817c478bd9Sstevel@tonic-gate 14827c478bd9Sstevel@tonic-gate /* 1483*ae115bc7Smrj * determine size of segkp 14847c478bd9Sstevel@tonic-gate */ 14857c478bd9Sstevel@tonic-gate if (sz < SEGKPMINSIZE || sz > SEGKPMAXSIZE) { 14867c478bd9Sstevel@tonic-gate sz = SEGKPDEFSIZE; 14877c478bd9Sstevel@tonic-gate cmn_err(CE_WARN, "!Illegal value for segkpsize. " 14887c478bd9Sstevel@tonic-gate "segkpsize has been reset to %ld pages", 14897c478bd9Sstevel@tonic-gate mmu_btop(sz)); 14907c478bd9Sstevel@tonic-gate } 14917c478bd9Sstevel@tonic-gate sz = MIN(sz, MAX(SEGKPMINSIZE, mmu_ptob(physmem))); 14927c478bd9Sstevel@tonic-gate 14937c478bd9Sstevel@tonic-gate segkpsize = mmu_btop(ROUND_UP_LPAGE(sz)); 14947c478bd9Sstevel@tonic-gate } 1495*ae115bc7Smrj PRM_DEBUG(segkp_base); 1496*ae115bc7Smrj PRM_DEBUG(segkpsize); 14977c478bd9Sstevel@tonic-gate 1498*ae115bc7Smrj segzio_base = segkp_base + mmu_ptob(segkpsize); 1499*ae115bc7Smrj if (segzio_fromheap) { 1500*ae115bc7Smrj segziosize = 0; 1501*ae115bc7Smrj } else { 1502ad23a2dbSjohansen size_t size; 150390e485acSjohansen size_t maxsize; 1504ad23a2dbSjohansen 1505ad23a2dbSjohansen /* size is in bytes, segziosize is in pages */ 1506ad23a2dbSjohansen if (segziosize == 0) { 150790e485acSjohansen size = mmu_ptob(physmem / 2); 1508ad23a2dbSjohansen } else { 1509ad23a2dbSjohansen size = mmu_ptob(segziosize); 1510ad23a2dbSjohansen } 1511ad23a2dbSjohansen 151290e485acSjohansen /* max size is 3/4ths of physmem */ 151390e485acSjohansen maxsize = mmu_ptob(physmem) - mmu_ptob(physmem / 4); 151490e485acSjohansen 1515ad23a2dbSjohansen if (size < SEGZIOMINSIZE) { 1516ad23a2dbSjohansen size = SEGZIOMINSIZE; 151790e485acSjohansen } else if (size > maxsize) { 151890e485acSjohansen size = maxsize; 1519ad23a2dbSjohansen } 1520ad23a2dbSjohansen segziosize = mmu_btop(ROUND_UP_LPAGE(size)); 1521ad23a2dbSjohansen } 1522*ae115bc7Smrj PRM_DEBUG(segziosize); 1523*ae115bc7Smrj PRM_DEBUG(segzio_base); 1524ad23a2dbSjohansen 15257c478bd9Sstevel@tonic-gate /* 1526*ae115bc7Smrj * Put the range of VA for device mappings next, kmdb knows to not 1527*ae115bc7Smrj * grep in this range of addresses. 15287c478bd9Sstevel@tonic-gate */ 1529*ae115bc7Smrj toxic_addr = 1530*ae115bc7Smrj ROUND_UP_LPAGE((uintptr_t)segzio_base + mmu_ptob(segziosize)); 15317c478bd9Sstevel@tonic-gate PRM_DEBUG(toxic_addr); 1532*ae115bc7Smrj segmap_start = ROUND_UP_LPAGE(toxic_addr + toxic_size); 1533*ae115bc7Smrj #else /* __i386 */ 1534*ae115bc7Smrj segmap_start = ROUND_UP_LPAGE(kernelbase); 1535*ae115bc7Smrj #endif /* __i386 */ 1536*ae115bc7Smrj PRM_DEBUG(segmap_start); 1537*ae115bc7Smrj ASSERT((caddr_t)segmap_start < boot_kernelheap); 15387c478bd9Sstevel@tonic-gate 15397c478bd9Sstevel@tonic-gate /* 15407c478bd9Sstevel@tonic-gate * Users can change segmapsize through eeprom or /etc/system. 15417c478bd9Sstevel@tonic-gate * If the variable is tuned through eeprom, there is no upper 15427c478bd9Sstevel@tonic-gate * bound on the size of segmap. If it is tuned through 15437c478bd9Sstevel@tonic-gate * /etc/system on 32-bit systems, it must be no larger than we 15447c478bd9Sstevel@tonic-gate * planned for in startup_memlist(). 15457c478bd9Sstevel@tonic-gate */ 15467c478bd9Sstevel@tonic-gate segmapsize = MAX(ROUND_UP_LPAGE(segmapsize), SEGMAPDEFAULT); 15477c478bd9Sstevel@tonic-gate 15487c478bd9Sstevel@tonic-gate #if defined(__i386) 15497c478bd9Sstevel@tonic-gate if (segmapsize > segmap_reserved) { 15507c478bd9Sstevel@tonic-gate cmn_err(CE_NOTE, "!segmapsize may not be set > 0x%lx in " 15517c478bd9Sstevel@tonic-gate "/etc/system. Use eeprom.", (long)SEGMAPMAX); 15527c478bd9Sstevel@tonic-gate segmapsize = segmap_reserved; 15537c478bd9Sstevel@tonic-gate } 15547c478bd9Sstevel@tonic-gate /* 15557c478bd9Sstevel@tonic-gate * 32-bit systems don't have segkpm or segkp, so segmap appears at 15567c478bd9Sstevel@tonic-gate * the bottom of the kernel's address range. Set aside space for a 1557*ae115bc7Smrj * small red zone just below the start of segmap. 15587c478bd9Sstevel@tonic-gate */ 1559*ae115bc7Smrj segmap_start += KERNEL_REDZONE_SIZE; 15607c478bd9Sstevel@tonic-gate segmapsize -= KERNEL_REDZONE_SIZE; 15617c478bd9Sstevel@tonic-gate #endif 15627c478bd9Sstevel@tonic-gate 1563*ae115bc7Smrj PRM_DEBUG(segmap_start); 15647c478bd9Sstevel@tonic-gate PRM_DEBUG(segmapsize); 1565*ae115bc7Smrj final_kernelheap = (caddr_t)ROUND_UP_LPAGE(segmap_start + segmapsize); 15667c478bd9Sstevel@tonic-gate PRM_DEBUG(final_kernelheap); 15677c478bd9Sstevel@tonic-gate 1568*ae115bc7Smrj /* 1569*ae115bc7Smrj * Do final allocations of HAT data structures that need to 1570*ae115bc7Smrj * be allocated before quiescing the boot loader. 1571*ae115bc7Smrj */ 1572*ae115bc7Smrj PRM_POINT("Calling hat_kern_alloc()..."); 1573*ae115bc7Smrj hat_kern_alloc((caddr_t)segmap_start, segmapsize, ekernelheap); 1574*ae115bc7Smrj PRM_POINT("hat_kern_alloc() done"); 1575*ae115bc7Smrj 1576*ae115bc7Smrj /* 1577*ae115bc7Smrj * Setup MTRR (Memory type range registers) 1578*ae115bc7Smrj */ 1579*ae115bc7Smrj setup_mtrr(); 1580*ae115bc7Smrj 1581*ae115bc7Smrj /* 1582*ae115bc7Smrj * The next two loops are done in distinct steps in order 1583*ae115bc7Smrj * to be sure that any page that is doubly mapped (both above 1584*ae115bc7Smrj * KERNEL_TEXT and below kernelbase) is dealt with correctly. 1585*ae115bc7Smrj * Note this may never happen, but it might someday. 1586*ae115bc7Smrj */ 1587*ae115bc7Smrj bootpages = NULL; 1588*ae115bc7Smrj PRM_POINT("Protecting boot pages"); 1589*ae115bc7Smrj 1590*ae115bc7Smrj /* 1591*ae115bc7Smrj * Protect any pages mapped above KERNEL_TEXT that somehow have 1592*ae115bc7Smrj * page_t's. This can only happen if something weird allocated 1593*ae115bc7Smrj * in this range (like kadb/kmdb). 1594*ae115bc7Smrj */ 1595*ae115bc7Smrj protect_boot_range(KERNEL_TEXT, (uintptr_t)-1, 0); 1596*ae115bc7Smrj 1597*ae115bc7Smrj /* 1598*ae115bc7Smrj * Before we can take over memory allocation/mapping from the boot 1599*ae115bc7Smrj * loader we must remove from our free page lists any boot allocated 1600*ae115bc7Smrj * pages that stay mapped until release_bootstrap(). 1601*ae115bc7Smrj */ 1602*ae115bc7Smrj protect_boot_range(0, kernelbase, 1); 1603*ae115bc7Smrj 1604*ae115bc7Smrj /* 1605*ae115bc7Smrj * Switch to running on regular HAT (not boot_mmu) 1606*ae115bc7Smrj */ 1607*ae115bc7Smrj PRM_POINT("Calling hat_kern_setup()..."); 1608*ae115bc7Smrj hat_kern_setup(); 1609*ae115bc7Smrj 1610*ae115bc7Smrj /* 1611*ae115bc7Smrj * It is no longer safe to call BOP_ALLOC(), so make sure we don't. 1612*ae115bc7Smrj */ 1613*ae115bc7Smrj bop_no_more_mem(); 1614*ae115bc7Smrj 1615*ae115bc7Smrj PRM_POINT("hat_kern_setup() done"); 1616*ae115bc7Smrj 1617*ae115bc7Smrj hat_cpu_online(CPU); 1618*ae115bc7Smrj 16197c478bd9Sstevel@tonic-gate /* 16207c478bd9Sstevel@tonic-gate * Initialize VM system 16217c478bd9Sstevel@tonic-gate */ 16227c478bd9Sstevel@tonic-gate PRM_POINT("Calling kvm_init()..."); 16237c478bd9Sstevel@tonic-gate kvm_init(); 16247c478bd9Sstevel@tonic-gate PRM_POINT("kvm_init() done"); 16257c478bd9Sstevel@tonic-gate 16267c478bd9Sstevel@tonic-gate /* 16277c478bd9Sstevel@tonic-gate * Tell kmdb that the VM system is now working 16287c478bd9Sstevel@tonic-gate */ 16297c478bd9Sstevel@tonic-gate if (boothowto & RB_DEBUG) 16307c478bd9Sstevel@tonic-gate kdi_dvec_vmready(); 16317c478bd9Sstevel@tonic-gate 16327c478bd9Sstevel@tonic-gate /* 16337c478bd9Sstevel@tonic-gate * Mangle the brand string etc. 16347c478bd9Sstevel@tonic-gate */ 16357c478bd9Sstevel@tonic-gate cpuid_pass3(CPU); 16367c478bd9Sstevel@tonic-gate 16377c478bd9Sstevel@tonic-gate /* 16387c478bd9Sstevel@tonic-gate * Now that we can use memory outside the top 4GB (on 64-bit 16397c478bd9Sstevel@tonic-gate * systems) and we know the size of segmap, we can set the final 1640*ae115bc7Smrj * size of the kernel's heap. 16417c478bd9Sstevel@tonic-gate */ 16427c478bd9Sstevel@tonic-gate if (final_kernelheap < boot_kernelheap) { 1643*ae115bc7Smrj PRM_POINT("kernelheap_extend()"); 1644*ae115bc7Smrj PRM_DEBUG(boot_kernelheap); 1645*ae115bc7Smrj PRM_DEBUG(final_kernelheap); 16467c478bd9Sstevel@tonic-gate kernelheap_extend(final_kernelheap, boot_kernelheap); 16477c478bd9Sstevel@tonic-gate } 16487c478bd9Sstevel@tonic-gate 16497c478bd9Sstevel@tonic-gate #if defined(__amd64) 16507c478bd9Sstevel@tonic-gate 16517c478bd9Sstevel@tonic-gate /* 16527c478bd9Sstevel@tonic-gate * Create the device arena for toxic (to dtrace/kmdb) mappings. 16537c478bd9Sstevel@tonic-gate */ 16547c478bd9Sstevel@tonic-gate device_arena = vmem_create("device", (void *)toxic_addr, 16557c478bd9Sstevel@tonic-gate toxic_size, MMU_PAGESIZE, NULL, NULL, NULL, 0, VM_SLEEP); 16567c478bd9Sstevel@tonic-gate 16577c478bd9Sstevel@tonic-gate #else /* __i386 */ 16587c478bd9Sstevel@tonic-gate 16597c478bd9Sstevel@tonic-gate /* 16607c478bd9Sstevel@tonic-gate * allocate the bit map that tracks toxic pages 16617c478bd9Sstevel@tonic-gate */ 1662*ae115bc7Smrj toxic_bit_map_len = btop((ulong_t)(valloc_base - kernelbase)); 16637c478bd9Sstevel@tonic-gate PRM_DEBUG(toxic_bit_map_len); 16647c478bd9Sstevel@tonic-gate toxic_bit_map = 16657c478bd9Sstevel@tonic-gate kmem_zalloc(BT_SIZEOFMAP(toxic_bit_map_len), KM_NOSLEEP); 16667c478bd9Sstevel@tonic-gate ASSERT(toxic_bit_map != NULL); 16677c478bd9Sstevel@tonic-gate PRM_DEBUG(toxic_bit_map); 16687c478bd9Sstevel@tonic-gate 16697c478bd9Sstevel@tonic-gate #endif /* __i386 */ 16707c478bd9Sstevel@tonic-gate 16717c478bd9Sstevel@tonic-gate 16727c478bd9Sstevel@tonic-gate /* 16737c478bd9Sstevel@tonic-gate * Now that we've got more VA, as well as the ability to allocate from 16747c478bd9Sstevel@tonic-gate * it, tell the debugger. 16757c478bd9Sstevel@tonic-gate */ 16767c478bd9Sstevel@tonic-gate if (boothowto & RB_DEBUG) 16777c478bd9Sstevel@tonic-gate kdi_dvec_memavail(); 16787c478bd9Sstevel@tonic-gate 16797c478bd9Sstevel@tonic-gate /* 16807c478bd9Sstevel@tonic-gate * The following code installs a special page fault handler (#pf) 16817c478bd9Sstevel@tonic-gate * to work around a pentium bug. 16827c478bd9Sstevel@tonic-gate */ 16837c478bd9Sstevel@tonic-gate #if !defined(__amd64) 16847c478bd9Sstevel@tonic-gate if (x86_type == X86_TYPE_P5) { 1685*ae115bc7Smrj desctbr_t idtr; 16867c478bd9Sstevel@tonic-gate gate_desc_t *newidt; 1687*ae115bc7Smrj struct machcpu *mcpu = &CPU->cpu_m; 16887c478bd9Sstevel@tonic-gate 16897c478bd9Sstevel@tonic-gate if ((newidt = kmem_zalloc(MMU_PAGESIZE, KM_NOSLEEP)) == NULL) 16907c478bd9Sstevel@tonic-gate panic("failed to install pentium_pftrap"); 16917c478bd9Sstevel@tonic-gate 16927c478bd9Sstevel@tonic-gate bcopy(idt0, newidt, sizeof (idt0)); 16937c478bd9Sstevel@tonic-gate set_gatesegd(&newidt[T_PGFLT], &pentium_pftrap, 1694*ae115bc7Smrj KCS_SEL, SDT_SYSIGT, SEL_KPL); 16957c478bd9Sstevel@tonic-gate 16967c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, (caddr_t)newidt, MMU_PAGESIZE, 16977c478bd9Sstevel@tonic-gate PROT_READ|PROT_EXEC); 16987c478bd9Sstevel@tonic-gate 1699*ae115bc7Smrj mcpu->mcpu_idt = newidt; 1700*ae115bc7Smrj idtr.dtr_base = (uintptr_t)mcpu->mcpu_idt; 1701*ae115bc7Smrj idtr.dtr_limit = sizeof (idt0) - 1; 1702*ae115bc7Smrj wr_idtr(&idtr); 17037c478bd9Sstevel@tonic-gate } 17047c478bd9Sstevel@tonic-gate #endif /* !__amd64 */ 17057c478bd9Sstevel@tonic-gate 17067c478bd9Sstevel@tonic-gate /* 17077c478bd9Sstevel@tonic-gate * Map page pfn=0 for drivers, such as kd, that need to pick up 17087c478bd9Sstevel@tonic-gate * parameters left there by controllers/BIOS. 17097c478bd9Sstevel@tonic-gate */ 17107c478bd9Sstevel@tonic-gate PRM_POINT("setup up p0_va"); 17117c478bd9Sstevel@tonic-gate p0_va = i86devmap(0, 1, PROT_READ); 17127c478bd9Sstevel@tonic-gate PRM_DEBUG(p0_va); 17137c478bd9Sstevel@tonic-gate 17147c478bd9Sstevel@tonic-gate cmn_err(CE_CONT, "?mem = %luK (0x%lx)\n", 17157c478bd9Sstevel@tonic-gate physinstalled << (MMU_PAGESHIFT - 10), ptob(physinstalled)); 17167c478bd9Sstevel@tonic-gate 171783f9b804Skchow /* 171883f9b804Skchow * disable automatic large pages for small memory systems or 171983f9b804Skchow * when the disable flag is set. 172083f9b804Skchow */ 1721ec25b48fSsusans if (!auto_lpg_disable && mmu.max_page_level > 0) { 1722ec25b48fSsusans max_uheap_lpsize = LEVEL_SIZE(1); 1723ec25b48fSsusans max_ustack_lpsize = LEVEL_SIZE(1); 1724ec25b48fSsusans max_privmap_lpsize = LEVEL_SIZE(1); 1725ec25b48fSsusans max_uidata_lpsize = LEVEL_SIZE(1); 1726ec25b48fSsusans max_utext_lpsize = LEVEL_SIZE(1); 1727ec25b48fSsusans max_shm_lpsize = LEVEL_SIZE(1); 1728ec25b48fSsusans } 1729ec25b48fSsusans if (physmem < privm_lpg_min_physmem || mmu.max_page_level == 0 || 1730ec25b48fSsusans auto_lpg_disable) { 1731beb1bda0Sdavemq use_brk_lpg = 0; 1732beb1bda0Sdavemq use_stk_lpg = 0; 1733ec25b48fSsusans } 1734ec25b48fSsusans if (mmu.max_page_level > 0) { 1735ec25b48fSsusans mcntl0_lpsize = LEVEL_SIZE(1); 1736beb1bda0Sdavemq } 1737beb1bda0Sdavemq 17387c478bd9Sstevel@tonic-gate PRM_POINT("Calling hat_init_finish()..."); 17397c478bd9Sstevel@tonic-gate hat_init_finish(); 17407c478bd9Sstevel@tonic-gate PRM_POINT("hat_init_finish() done"); 17417c478bd9Sstevel@tonic-gate 17427c478bd9Sstevel@tonic-gate /* 17437c478bd9Sstevel@tonic-gate * Initialize the segkp segment type. 17447c478bd9Sstevel@tonic-gate */ 17457c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 1746*ae115bc7Smrj PRM_POINT("Attaching segkp"); 1747*ae115bc7Smrj if (segkp_fromheap) { 17487c478bd9Sstevel@tonic-gate segkp->s_as = &kas; 1749*ae115bc7Smrj } else if (seg_attach(&kas, (caddr_t)segkp_base, mmu_ptob(segkpsize), 1750*ae115bc7Smrj segkp) < 0) { 1751*ae115bc7Smrj panic("startup: cannot attach segkp"); 1752*ae115bc7Smrj /*NOTREACHED*/ 17537c478bd9Sstevel@tonic-gate } 1754*ae115bc7Smrj PRM_POINT("Doing segkp_create()"); 17557c478bd9Sstevel@tonic-gate if (segkp_create(segkp) != 0) { 17567c478bd9Sstevel@tonic-gate panic("startup: segkp_create failed"); 17577c478bd9Sstevel@tonic-gate /*NOTREACHED*/ 17587c478bd9Sstevel@tonic-gate } 17597c478bd9Sstevel@tonic-gate PRM_DEBUG(segkp); 17607c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 17617c478bd9Sstevel@tonic-gate 17627c478bd9Sstevel@tonic-gate /* 17637c478bd9Sstevel@tonic-gate * kpm segment 17647c478bd9Sstevel@tonic-gate */ 17657c478bd9Sstevel@tonic-gate segmap_kpm = 0; 17667c478bd9Sstevel@tonic-gate if (kpm_desired) { 17677c478bd9Sstevel@tonic-gate kpm_init(); 17687c478bd9Sstevel@tonic-gate kpm_enable = 1; 1769a5652762Spraks vpm_enable = 1; 17707c478bd9Sstevel@tonic-gate } 17717c478bd9Sstevel@tonic-gate 17727c478bd9Sstevel@tonic-gate /* 17737c478bd9Sstevel@tonic-gate * Now create segmap segment. 17747c478bd9Sstevel@tonic-gate */ 17757c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 1776*ae115bc7Smrj if (seg_attach(&kas, (caddr_t)segmap_start, segmapsize, segmap) < 0) { 1777*ae115bc7Smrj panic("cannot attach segmap"); 17787c478bd9Sstevel@tonic-gate /*NOTREACHED*/ 17797c478bd9Sstevel@tonic-gate } 1780*ae115bc7Smrj PRM_DEBUG(segmap); 17817c478bd9Sstevel@tonic-gate 17827c478bd9Sstevel@tonic-gate a.prot = PROT_READ | PROT_WRITE; 17837c478bd9Sstevel@tonic-gate a.shmsize = 0; 17847c478bd9Sstevel@tonic-gate a.nfreelist = segmapfreelists; 17857c478bd9Sstevel@tonic-gate 1786*ae115bc7Smrj if (segmap_create(segmap, (caddr_t)&a) != 0) 1787*ae115bc7Smrj panic("segmap_create segmap"); 17887c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 17897c478bd9Sstevel@tonic-gate 17907c478bd9Sstevel@tonic-gate setup_vaddr_for_ppcopy(CPU); 17917c478bd9Sstevel@tonic-gate 17927c478bd9Sstevel@tonic-gate segdev_init(); 17937c478bd9Sstevel@tonic-gate pmem_init(); 1794*ae115bc7Smrj 17957c478bd9Sstevel@tonic-gate PRM_POINT("startup_vm() done"); 17967c478bd9Sstevel@tonic-gate } 17977c478bd9Sstevel@tonic-gate 1798*ae115bc7Smrj /* 1799*ae115bc7Smrj * Load a tod module for the non-standard tod part found on this system. 1800*ae115bc7Smrj */ 1801*ae115bc7Smrj static void 1802*ae115bc7Smrj load_tod_module(char *todmod) 1803*ae115bc7Smrj { 1804*ae115bc7Smrj if (modload("tod", todmod) == -1) 1805*ae115bc7Smrj halt("Can't load TOD module"); 1806*ae115bc7Smrj } 1807*ae115bc7Smrj 18087c478bd9Sstevel@tonic-gate static void 18097c478bd9Sstevel@tonic-gate startup_end(void) 18107c478bd9Sstevel@tonic-gate { 18117c478bd9Sstevel@tonic-gate extern void setx86isalist(void); 18127c478bd9Sstevel@tonic-gate 18137c478bd9Sstevel@tonic-gate PRM_POINT("startup_end() starting..."); 18147c478bd9Sstevel@tonic-gate 18157c478bd9Sstevel@tonic-gate /* 18167c478bd9Sstevel@tonic-gate * Perform tasks that get done after most of the VM 18177c478bd9Sstevel@tonic-gate * initialization has been done but before the clock 18187c478bd9Sstevel@tonic-gate * and other devices get started. 18197c478bd9Sstevel@tonic-gate */ 18207c478bd9Sstevel@tonic-gate kern_setup1(); 18217c478bd9Sstevel@tonic-gate 18227c478bd9Sstevel@tonic-gate /* 18237c478bd9Sstevel@tonic-gate * Perform CPC initialization for this CPU. 18247c478bd9Sstevel@tonic-gate */ 18257c478bd9Sstevel@tonic-gate kcpc_hw_init(CPU); 18267c478bd9Sstevel@tonic-gate 1827ee88d2b9Skchow #if defined(OPTERON_WORKAROUND_6323525) 1828ee88d2b9Skchow if (opteron_workaround_6323525) 1829ee88d2b9Skchow patch_workaround_6323525(); 1830ee88d2b9Skchow #endif 1831*ae115bc7Smrj /* 1832*ae115bc7Smrj * If needed, load TOD module now so that ddi_get_time(9F) etc. work 1833*ae115bc7Smrj * (For now, "needed" is defined as set tod_module_name in /etc/system) 1834*ae115bc7Smrj */ 1835*ae115bc7Smrj if (tod_module_name != NULL) { 1836*ae115bc7Smrj PRM_POINT("load_tod_module()"); 1837*ae115bc7Smrj load_tod_module(tod_module_name); 1838*ae115bc7Smrj } 1839*ae115bc7Smrj 18407c478bd9Sstevel@tonic-gate /* 18417c478bd9Sstevel@tonic-gate * Configure the system. 18427c478bd9Sstevel@tonic-gate */ 18437c478bd9Sstevel@tonic-gate PRM_POINT("Calling configure()..."); 18447c478bd9Sstevel@tonic-gate configure(); /* set up devices */ 18457c478bd9Sstevel@tonic-gate PRM_POINT("configure() done"); 18467c478bd9Sstevel@tonic-gate 18477c478bd9Sstevel@tonic-gate /* 18487c478bd9Sstevel@tonic-gate * Set the isa_list string to the defined instruction sets we 18497c478bd9Sstevel@tonic-gate * support. 18507c478bd9Sstevel@tonic-gate */ 18517c478bd9Sstevel@tonic-gate setx86isalist(); 1852100b72f4Sandrei cpu_intr_alloc(CPU, NINTR_THREADS); 18537c478bd9Sstevel@tonic-gate psm_install(); 18547c478bd9Sstevel@tonic-gate 18557c478bd9Sstevel@tonic-gate /* 18567c478bd9Sstevel@tonic-gate * We're done with bootops. We don't unmap the bootstrap yet because 18577c478bd9Sstevel@tonic-gate * we're still using bootsvcs. 18587c478bd9Sstevel@tonic-gate */ 1859*ae115bc7Smrj PRM_POINT("NULLing out bootops"); 1860*ae115bc7Smrj *bootopsp = (struct bootops *)NULL; 18617c478bd9Sstevel@tonic-gate bootops = (struct bootops *)NULL; 18627c478bd9Sstevel@tonic-gate 18637c478bd9Sstevel@tonic-gate PRM_POINT("Enabling interrupts"); 18647c478bd9Sstevel@tonic-gate (*picinitf)(); 18657c478bd9Sstevel@tonic-gate sti(); 18667c478bd9Sstevel@tonic-gate 18677c478bd9Sstevel@tonic-gate (void) add_avsoftintr((void *)&softlevel1_hdl, 1, softlevel1, 18687c478bd9Sstevel@tonic-gate "softlevel1", NULL, NULL); /* XXX to be moved later */ 18697c478bd9Sstevel@tonic-gate 18707c478bd9Sstevel@tonic-gate PRM_POINT("startup_end() done"); 18717c478bd9Sstevel@tonic-gate } 18727c478bd9Sstevel@tonic-gate 18737c478bd9Sstevel@tonic-gate extern char hw_serial[]; 18747c478bd9Sstevel@tonic-gate char *_hs1107 = hw_serial; 18757c478bd9Sstevel@tonic-gate ulong_t _bdhs34; 18767c478bd9Sstevel@tonic-gate 18777c478bd9Sstevel@tonic-gate void 18787c478bd9Sstevel@tonic-gate post_startup(void) 18797c478bd9Sstevel@tonic-gate { 18807c478bd9Sstevel@tonic-gate /* 18817c478bd9Sstevel@tonic-gate * Set the system wide, processor-specific flags to be passed 18827c478bd9Sstevel@tonic-gate * to userland via the aux vector for performance hints and 18837c478bd9Sstevel@tonic-gate * instruction set extensions. 18847c478bd9Sstevel@tonic-gate */ 18857c478bd9Sstevel@tonic-gate bind_hwcap(); 18867c478bd9Sstevel@tonic-gate 18877c478bd9Sstevel@tonic-gate /* 1888*ae115bc7Smrj * Load the System Management BIOS into the global ksmbios 1889*ae115bc7Smrj * handle, if an SMBIOS is present on this system. 18907c478bd9Sstevel@tonic-gate */ 189184ab085aSmws ksmbios = smbios_open(NULL, SMB_VERSION, ksmbios_flags, NULL); 18927c478bd9Sstevel@tonic-gate 18937c478bd9Sstevel@tonic-gate /* 1894*ae115bc7Smrj * Startup the memory scrubber. 18957c478bd9Sstevel@tonic-gate */ 189684ab085aSmws memscrub_init(); 18977c478bd9Sstevel@tonic-gate 18987aec1d6eScindi /* 18997aec1d6eScindi * Complete CPU module initialization 19007aec1d6eScindi */ 19017aec1d6eScindi cmi_post_init(); 19027aec1d6eScindi 19037c478bd9Sstevel@tonic-gate /* 190484ab085aSmws * Perform forceloading tasks for /etc/system. 19057c478bd9Sstevel@tonic-gate */ 190684ab085aSmws (void) mod_sysctl(SYS_FORCELOAD, NULL); 19077c478bd9Sstevel@tonic-gate 19087c478bd9Sstevel@tonic-gate /* 19097c478bd9Sstevel@tonic-gate * ON4.0: Force /proc module in until clock interrupt handle fixed 19107c478bd9Sstevel@tonic-gate * ON4.0: This must be fixed or restated in /etc/systems. 19117c478bd9Sstevel@tonic-gate */ 19127c478bd9Sstevel@tonic-gate (void) modload("fs", "procfs"); 19137c478bd9Sstevel@tonic-gate 19147c478bd9Sstevel@tonic-gate #if defined(__i386) 19157c478bd9Sstevel@tonic-gate /* 19167c478bd9Sstevel@tonic-gate * Check for required functional Floating Point hardware, 19177c478bd9Sstevel@tonic-gate * unless FP hardware explicitly disabled. 19187c478bd9Sstevel@tonic-gate */ 19197c478bd9Sstevel@tonic-gate if (fpu_exists && (fpu_pentium_fdivbug || fp_kind == FP_NO)) 19207c478bd9Sstevel@tonic-gate halt("No working FP hardware found"); 19217c478bd9Sstevel@tonic-gate #endif 19227c478bd9Sstevel@tonic-gate 19237c478bd9Sstevel@tonic-gate maxmem = freemem; 19247c478bd9Sstevel@tonic-gate 19257c478bd9Sstevel@tonic-gate add_cpunode2devtree(CPU->cpu_id, CPU->cpu_m.mcpu_cpi); 19267c478bd9Sstevel@tonic-gate } 19277c478bd9Sstevel@tonic-gate 19287c478bd9Sstevel@tonic-gate static int 19297c478bd9Sstevel@tonic-gate pp_in_ramdisk(page_t *pp) 19307c478bd9Sstevel@tonic-gate { 19317c478bd9Sstevel@tonic-gate extern uint64_t ramdisk_start, ramdisk_end; 19327c478bd9Sstevel@tonic-gate 19337c478bd9Sstevel@tonic-gate return ((pp->p_pagenum >= btop(ramdisk_start)) && 19347c478bd9Sstevel@tonic-gate (pp->p_pagenum < btopr(ramdisk_end))); 19357c478bd9Sstevel@tonic-gate } 19367c478bd9Sstevel@tonic-gate 19377c478bd9Sstevel@tonic-gate void 19387c478bd9Sstevel@tonic-gate release_bootstrap(void) 19397c478bd9Sstevel@tonic-gate { 19407c478bd9Sstevel@tonic-gate int root_is_ramdisk; 19417c478bd9Sstevel@tonic-gate page_t *pp; 19427c478bd9Sstevel@tonic-gate extern void kobj_boot_unmountroot(void); 19437c478bd9Sstevel@tonic-gate extern dev_t rootdev; 19447c478bd9Sstevel@tonic-gate 19457c478bd9Sstevel@tonic-gate /* unmount boot ramdisk and release kmem usage */ 19467c478bd9Sstevel@tonic-gate kobj_boot_unmountroot(); 19477c478bd9Sstevel@tonic-gate 19487c478bd9Sstevel@tonic-gate /* 19497c478bd9Sstevel@tonic-gate * We're finished using the boot loader so free its pages. 19507c478bd9Sstevel@tonic-gate */ 19517c478bd9Sstevel@tonic-gate PRM_POINT("Unmapping lower boot pages"); 1952*ae115bc7Smrj clear_boot_mappings(0, _userlimit); 1953*ae115bc7Smrj postbootkernelbase = kernelbase; 19547c478bd9Sstevel@tonic-gate 19557c478bd9Sstevel@tonic-gate /* 19567c478bd9Sstevel@tonic-gate * If root isn't on ramdisk, destroy the hardcoded 19577c478bd9Sstevel@tonic-gate * ramdisk node now and release the memory. Else, 19587c478bd9Sstevel@tonic-gate * ramdisk memory is kept in rd_pages. 19597c478bd9Sstevel@tonic-gate */ 19607c478bd9Sstevel@tonic-gate root_is_ramdisk = (getmajor(rootdev) == ddi_name_to_major("ramdisk")); 19617c478bd9Sstevel@tonic-gate if (!root_is_ramdisk) { 19627c478bd9Sstevel@tonic-gate dev_info_t *dip = ddi_find_devinfo("ramdisk", -1, 0); 19637c478bd9Sstevel@tonic-gate ASSERT(dip && ddi_get_parent(dip) == ddi_root_node()); 19647c478bd9Sstevel@tonic-gate ndi_rele_devi(dip); /* held from ddi_find_devinfo */ 19657c478bd9Sstevel@tonic-gate (void) ddi_remove_child(dip, 0); 19667c478bd9Sstevel@tonic-gate } 19677c478bd9Sstevel@tonic-gate 19687c478bd9Sstevel@tonic-gate PRM_POINT("Releasing boot pages"); 19697c478bd9Sstevel@tonic-gate while (bootpages) { 19707c478bd9Sstevel@tonic-gate pp = bootpages; 19717c478bd9Sstevel@tonic-gate bootpages = pp->p_next; 19727c478bd9Sstevel@tonic-gate if (root_is_ramdisk && pp_in_ramdisk(pp)) { 19737c478bd9Sstevel@tonic-gate pp->p_next = rd_pages; 19747c478bd9Sstevel@tonic-gate rd_pages = pp; 19757c478bd9Sstevel@tonic-gate continue; 19767c478bd9Sstevel@tonic-gate } 19777c478bd9Sstevel@tonic-gate pp->p_next = (struct page *)0; 1978*ae115bc7Smrj pp->p_prev = (struct page *)0; 1979*ae115bc7Smrj PP_CLRBOOTPAGES(pp); 19807c478bd9Sstevel@tonic-gate page_free(pp, 1); 19817c478bd9Sstevel@tonic-gate } 1982*ae115bc7Smrj PRM_POINT("Boot pages released"); 19837c478bd9Sstevel@tonic-gate 19847c478bd9Sstevel@tonic-gate /* 19857c478bd9Sstevel@tonic-gate * Find 1 page below 1 MB so that other processors can boot up. 19867c478bd9Sstevel@tonic-gate * Make sure it has a kernel VA as well as a 1:1 mapping. 19877c478bd9Sstevel@tonic-gate * We should have just free'd one up. 19887c478bd9Sstevel@tonic-gate */ 19897c478bd9Sstevel@tonic-gate if (use_mp) { 1990*ae115bc7Smrj pfn_t pfn; 1991*ae115bc7Smrj 19927c478bd9Sstevel@tonic-gate for (pfn = 1; pfn < btop(1*1024*1024); pfn++) { 19937c478bd9Sstevel@tonic-gate if (page_numtopp_alloc(pfn) == NULL) 19947c478bd9Sstevel@tonic-gate continue; 19957c478bd9Sstevel@tonic-gate rm_platter_va = i86devmap(pfn, 1, 19967c478bd9Sstevel@tonic-gate PROT_READ | PROT_WRITE | PROT_EXEC); 19977c478bd9Sstevel@tonic-gate rm_platter_pa = ptob(pfn); 19987c478bd9Sstevel@tonic-gate hat_devload(kas.a_hat, 19997c478bd9Sstevel@tonic-gate (caddr_t)(uintptr_t)rm_platter_pa, MMU_PAGESIZE, 20007c478bd9Sstevel@tonic-gate pfn, PROT_READ | PROT_WRITE | PROT_EXEC, 20017c478bd9Sstevel@tonic-gate HAT_LOAD_NOCONSIST); 20027c478bd9Sstevel@tonic-gate break; 20037c478bd9Sstevel@tonic-gate } 20047c478bd9Sstevel@tonic-gate if (pfn == btop(1*1024*1024)) 20057c478bd9Sstevel@tonic-gate panic("No page available for starting " 20067c478bd9Sstevel@tonic-gate "other processors"); 20077c478bd9Sstevel@tonic-gate } 20087c478bd9Sstevel@tonic-gate 20097c478bd9Sstevel@tonic-gate } 20107c478bd9Sstevel@tonic-gate 20117c478bd9Sstevel@tonic-gate /* 20127c478bd9Sstevel@tonic-gate * Initialize the platform-specific parts of a page_t. 20137c478bd9Sstevel@tonic-gate */ 20147c478bd9Sstevel@tonic-gate void 20157c478bd9Sstevel@tonic-gate add_physmem_cb(page_t *pp, pfn_t pnum) 20167c478bd9Sstevel@tonic-gate { 20177c478bd9Sstevel@tonic-gate pp->p_pagenum = pnum; 20187c478bd9Sstevel@tonic-gate pp->p_mapping = NULL; 20197c478bd9Sstevel@tonic-gate pp->p_embed = 0; 20207c478bd9Sstevel@tonic-gate pp->p_share = 0; 20217c478bd9Sstevel@tonic-gate pp->p_mlentry = 0; 20227c478bd9Sstevel@tonic-gate } 20237c478bd9Sstevel@tonic-gate 20247c478bd9Sstevel@tonic-gate /* 20257c478bd9Sstevel@tonic-gate * kphysm_init() initializes physical memory. 20267c478bd9Sstevel@tonic-gate */ 20277c478bd9Sstevel@tonic-gate static pgcnt_t 20287c478bd9Sstevel@tonic-gate kphysm_init( 2029*ae115bc7Smrj page_t *pp, 20307c478bd9Sstevel@tonic-gate pgcnt_t npages) 20317c478bd9Sstevel@tonic-gate { 20327c478bd9Sstevel@tonic-gate struct memlist *pmem; 20337c478bd9Sstevel@tonic-gate struct memseg *cur_memseg; 20347c478bd9Sstevel@tonic-gate pfn_t base_pfn; 20357c478bd9Sstevel@tonic-gate pgcnt_t num; 20367c478bd9Sstevel@tonic-gate pgcnt_t pages_done = 0; 20377c478bd9Sstevel@tonic-gate uint64_t addr; 20387c478bd9Sstevel@tonic-gate uint64_t size; 20397c478bd9Sstevel@tonic-gate extern pfn_t ddiphysmin; 20407c478bd9Sstevel@tonic-gate 20417c478bd9Sstevel@tonic-gate ASSERT(page_hash != NULL && page_hashsz != 0); 20427c478bd9Sstevel@tonic-gate 2043*ae115bc7Smrj cur_memseg = memseg_base; 20447c478bd9Sstevel@tonic-gate for (pmem = phys_avail; pmem && npages; pmem = pmem->next) { 20457c478bd9Sstevel@tonic-gate /* 20467c478bd9Sstevel@tonic-gate * In a 32 bit kernel can't use higher memory if we're 20477c478bd9Sstevel@tonic-gate * not booting in PAE mode. This check takes care of that. 20487c478bd9Sstevel@tonic-gate */ 20497c478bd9Sstevel@tonic-gate addr = pmem->address; 20507c478bd9Sstevel@tonic-gate size = pmem->size; 20517c478bd9Sstevel@tonic-gate if (btop(addr) > physmax) 20527c478bd9Sstevel@tonic-gate continue; 20537c478bd9Sstevel@tonic-gate 20547c478bd9Sstevel@tonic-gate /* 20557c478bd9Sstevel@tonic-gate * align addr and size - they may not be at page boundaries 20567c478bd9Sstevel@tonic-gate */ 20577c478bd9Sstevel@tonic-gate if ((addr & MMU_PAGEOFFSET) != 0) { 20587c478bd9Sstevel@tonic-gate addr += MMU_PAGEOFFSET; 20597c478bd9Sstevel@tonic-gate addr &= ~(uint64_t)MMU_PAGEOFFSET; 20607c478bd9Sstevel@tonic-gate size -= addr - pmem->address; 20617c478bd9Sstevel@tonic-gate } 20627c478bd9Sstevel@tonic-gate 20636bb54764Skchow /* only process pages below or equal to physmax */ 20646bb54764Skchow if ((btop(addr + size) - 1) > physmax) 20656bb54764Skchow size = ptob(physmax - btop(addr) + 1); 20667c478bd9Sstevel@tonic-gate 20677c478bd9Sstevel@tonic-gate num = btop(size); 20687c478bd9Sstevel@tonic-gate if (num == 0) 20697c478bd9Sstevel@tonic-gate continue; 20707c478bd9Sstevel@tonic-gate 20717c478bd9Sstevel@tonic-gate if (num > npages) 20727c478bd9Sstevel@tonic-gate num = npages; 20737c478bd9Sstevel@tonic-gate 20747c478bd9Sstevel@tonic-gate npages -= num; 20757c478bd9Sstevel@tonic-gate pages_done += num; 20767c478bd9Sstevel@tonic-gate base_pfn = btop(addr); 20777c478bd9Sstevel@tonic-gate 20787c478bd9Sstevel@tonic-gate if (prom_debug) 20797c478bd9Sstevel@tonic-gate prom_printf("MEMSEG addr=0x%" PRIx64 20807c478bd9Sstevel@tonic-gate " pgs=0x%lx pfn 0x%lx-0x%lx\n", 20817c478bd9Sstevel@tonic-gate addr, num, base_pfn, base_pfn + num); 20827c478bd9Sstevel@tonic-gate 20837c478bd9Sstevel@tonic-gate /* 2084*ae115bc7Smrj * Ignore pages below ddiphysmin to simplify ddi memory 20857c478bd9Sstevel@tonic-gate * allocation with non-zero addr_lo requests. 20867c478bd9Sstevel@tonic-gate */ 20877c478bd9Sstevel@tonic-gate if (base_pfn < ddiphysmin) { 2088*ae115bc7Smrj if (base_pfn + num <= ddiphysmin) 20897c478bd9Sstevel@tonic-gate continue; 20907c478bd9Sstevel@tonic-gate pp += (ddiphysmin - base_pfn); 20917c478bd9Sstevel@tonic-gate num -= (ddiphysmin - base_pfn); 20927c478bd9Sstevel@tonic-gate base_pfn = ddiphysmin; 20937c478bd9Sstevel@tonic-gate } 2094*ae115bc7Smrj 20957c478bd9Sstevel@tonic-gate /* 20967c478bd9Sstevel@tonic-gate * Build the memsegs entry 20977c478bd9Sstevel@tonic-gate */ 20987c478bd9Sstevel@tonic-gate cur_memseg->pages = pp; 20997c478bd9Sstevel@tonic-gate cur_memseg->epages = pp + num; 21007c478bd9Sstevel@tonic-gate cur_memseg->pages_base = base_pfn; 21017c478bd9Sstevel@tonic-gate cur_memseg->pages_end = base_pfn + num; 21027c478bd9Sstevel@tonic-gate 21037c478bd9Sstevel@tonic-gate /* 2104*ae115bc7Smrj * Insert into memseg list in decreasing pfn range order. 21057c478bd9Sstevel@tonic-gate * Low memory is typically more fragmented such that this 21067c478bd9Sstevel@tonic-gate * ordering keeps the larger ranges at the front of the list 21077c478bd9Sstevel@tonic-gate * for code that searches memseg. 2108*ae115bc7Smrj * This ASSERTS that the memsegs coming in from boot are in 2109*ae115bc7Smrj * increasing physical address order and not contiguous. 21107c478bd9Sstevel@tonic-gate */ 2111*ae115bc7Smrj if (memsegs != NULL) { 2112*ae115bc7Smrj ASSERT(cur_memseg->pages_base >= memsegs->pages_end); 2113*ae115bc7Smrj cur_memseg->next = memsegs; 21147c478bd9Sstevel@tonic-gate } 2115*ae115bc7Smrj memsegs = cur_memseg; 21167c478bd9Sstevel@tonic-gate 21177c478bd9Sstevel@tonic-gate /* 21187c478bd9Sstevel@tonic-gate * add_physmem() initializes the PSM part of the page 21197c478bd9Sstevel@tonic-gate * struct by calling the PSM back with add_physmem_cb(). 21207c478bd9Sstevel@tonic-gate * In addition it coalesces pages into larger pages as 21217c478bd9Sstevel@tonic-gate * it initializes them. 21227c478bd9Sstevel@tonic-gate */ 21237c478bd9Sstevel@tonic-gate add_physmem(pp, num, base_pfn); 21247c478bd9Sstevel@tonic-gate cur_memseg++; 21257c478bd9Sstevel@tonic-gate availrmem_initial += num; 21267c478bd9Sstevel@tonic-gate availrmem += num; 21277c478bd9Sstevel@tonic-gate 2128*ae115bc7Smrj pp += num; 21297c478bd9Sstevel@tonic-gate } 21307c478bd9Sstevel@tonic-gate 21317c478bd9Sstevel@tonic-gate PRM_DEBUG(availrmem_initial); 21327c478bd9Sstevel@tonic-gate PRM_DEBUG(availrmem); 21337c478bd9Sstevel@tonic-gate PRM_DEBUG(freemem); 21347c478bd9Sstevel@tonic-gate build_pfn_hash(); 21357c478bd9Sstevel@tonic-gate return (pages_done); 21367c478bd9Sstevel@tonic-gate } 21377c478bd9Sstevel@tonic-gate 21387c478bd9Sstevel@tonic-gate /* 21397c478bd9Sstevel@tonic-gate * Kernel VM initialization. 21407c478bd9Sstevel@tonic-gate */ 21417c478bd9Sstevel@tonic-gate static void 21427c478bd9Sstevel@tonic-gate kvm_init(void) 21437c478bd9Sstevel@tonic-gate { 21447c478bd9Sstevel@tonic-gate ASSERT((((uintptr_t)s_text) & MMU_PAGEOFFSET) == 0); 21457c478bd9Sstevel@tonic-gate 21467c478bd9Sstevel@tonic-gate /* 21477c478bd9Sstevel@tonic-gate * Put the kernel segments in kernel address space. 21487c478bd9Sstevel@tonic-gate */ 21497c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 21507c478bd9Sstevel@tonic-gate as_avlinit(&kas); 21517c478bd9Sstevel@tonic-gate 21527c478bd9Sstevel@tonic-gate (void) seg_attach(&kas, s_text, e_moddata - s_text, &ktextseg); 21537c478bd9Sstevel@tonic-gate (void) segkmem_create(&ktextseg); 21547c478bd9Sstevel@tonic-gate 21557c478bd9Sstevel@tonic-gate (void) seg_attach(&kas, (caddr_t)valloc_base, valloc_sz, &kvalloc); 21567c478bd9Sstevel@tonic-gate (void) segkmem_create(&kvalloc); 21577c478bd9Sstevel@tonic-gate 21587c478bd9Sstevel@tonic-gate /* 21597c478bd9Sstevel@tonic-gate * We're about to map out /boot. This is the beginning of the 21607c478bd9Sstevel@tonic-gate * system resource management transition. We can no longer 21617c478bd9Sstevel@tonic-gate * call into /boot for I/O or memory allocations. 21627c478bd9Sstevel@tonic-gate */ 21637c478bd9Sstevel@tonic-gate (void) seg_attach(&kas, final_kernelheap, 21647c478bd9Sstevel@tonic-gate ekernelheap - final_kernelheap, &kvseg); 21657c478bd9Sstevel@tonic-gate (void) segkmem_create(&kvseg); 21667c478bd9Sstevel@tonic-gate 2167*ae115bc7Smrj if (core_size > 0) { 2168*ae115bc7Smrj PRM_POINT("attaching kvseg_core"); 2169*ae115bc7Smrj (void) seg_attach(&kas, (caddr_t)core_base, core_size, 2170*ae115bc7Smrj &kvseg_core); 2171*ae115bc7Smrj (void) segkmem_create(&kvseg_core); 2172*ae115bc7Smrj } 2173ad23a2dbSjohansen 2174*ae115bc7Smrj if (segziosize > 0) { 2175*ae115bc7Smrj PRM_POINT("attaching segzio"); 2176ad23a2dbSjohansen (void) seg_attach(&kas, segzio_base, mmu_ptob(segziosize), 2177ad23a2dbSjohansen &kzioseg); 2178ad23a2dbSjohansen (void) segkmem_zio_create(&kzioseg); 2179ad23a2dbSjohansen 2180ad23a2dbSjohansen /* create zio area covering new segment */ 2181ad23a2dbSjohansen segkmem_zio_init(segzio_base, mmu_ptob(segziosize)); 2182ad23a2dbSjohansen } 21837c478bd9Sstevel@tonic-gate 2184*ae115bc7Smrj (void) seg_attach(&kas, kdi_segdebugbase, kdi_segdebugsize, &kdebugseg); 21857c478bd9Sstevel@tonic-gate (void) segkmem_create(&kdebugseg); 21867c478bd9Sstevel@tonic-gate 21877c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 21887c478bd9Sstevel@tonic-gate 21897c478bd9Sstevel@tonic-gate /* 21907c478bd9Sstevel@tonic-gate * Ensure that the red zone at kernelbase is never accessible. 21917c478bd9Sstevel@tonic-gate */ 2192*ae115bc7Smrj PRM_POINT("protecting redzone"); 21937c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, (caddr_t)kernelbase, KERNEL_REDZONE_SIZE, 0); 21947c478bd9Sstevel@tonic-gate 21957c478bd9Sstevel@tonic-gate /* 21967c478bd9Sstevel@tonic-gate * Make the text writable so that it can be hot patched by DTrace. 21977c478bd9Sstevel@tonic-gate */ 21987c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, s_text, e_modtext - s_text, 21997c478bd9Sstevel@tonic-gate PROT_READ | PROT_WRITE | PROT_EXEC); 22007c478bd9Sstevel@tonic-gate 22017c478bd9Sstevel@tonic-gate /* 22027c478bd9Sstevel@tonic-gate * Make data writable until end. 22037c478bd9Sstevel@tonic-gate */ 22047c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, s_data, e_moddata - s_data, 22057c478bd9Sstevel@tonic-gate PROT_READ | PROT_WRITE | PROT_EXEC); 22067c478bd9Sstevel@tonic-gate } 22077c478bd9Sstevel@tonic-gate 22087c478bd9Sstevel@tonic-gate /* 22097c478bd9Sstevel@tonic-gate * These are MTTR registers supported by P6 22107c478bd9Sstevel@tonic-gate */ 22117c478bd9Sstevel@tonic-gate static struct mtrrvar mtrrphys_arr[MAX_MTRRVAR]; 22127c478bd9Sstevel@tonic-gate static uint64_t mtrr64k, mtrr16k1, mtrr16k2; 22137c478bd9Sstevel@tonic-gate static uint64_t mtrr4k1, mtrr4k2, mtrr4k3; 22147c478bd9Sstevel@tonic-gate static uint64_t mtrr4k4, mtrr4k5, mtrr4k6; 22157c478bd9Sstevel@tonic-gate static uint64_t mtrr4k7, mtrr4k8, mtrrcap; 22167c478bd9Sstevel@tonic-gate uint64_t mtrrdef, pat_attr_reg; 22177c478bd9Sstevel@tonic-gate 22187c478bd9Sstevel@tonic-gate /* 22197c478bd9Sstevel@tonic-gate * Disable reprogramming of MTRRs by default. 22207c478bd9Sstevel@tonic-gate */ 22217c478bd9Sstevel@tonic-gate int enable_relaxed_mtrr = 0; 22227c478bd9Sstevel@tonic-gate 22237c478bd9Sstevel@tonic-gate void 222445916cd2Sjpk setup_mtrr(void) 22257c478bd9Sstevel@tonic-gate { 22267c478bd9Sstevel@tonic-gate int i, ecx; 22277c478bd9Sstevel@tonic-gate int vcnt; 22287c478bd9Sstevel@tonic-gate struct mtrrvar *mtrrphys; 22297c478bd9Sstevel@tonic-gate 22307c478bd9Sstevel@tonic-gate if (!(x86_feature & X86_MTRR)) 22317c478bd9Sstevel@tonic-gate return; 22327c478bd9Sstevel@tonic-gate 22330ac7d7d8Skucharsk mtrrcap = rdmsr(REG_MTRRCAP); 22340ac7d7d8Skucharsk mtrrdef = rdmsr(REG_MTRRDEF); 22357c478bd9Sstevel@tonic-gate if (mtrrcap & MTRRCAP_FIX) { 22360ac7d7d8Skucharsk mtrr64k = rdmsr(REG_MTRR64K); 22370ac7d7d8Skucharsk mtrr16k1 = rdmsr(REG_MTRR16K1); 22380ac7d7d8Skucharsk mtrr16k2 = rdmsr(REG_MTRR16K2); 22390ac7d7d8Skucharsk mtrr4k1 = rdmsr(REG_MTRR4K1); 22400ac7d7d8Skucharsk mtrr4k2 = rdmsr(REG_MTRR4K2); 22410ac7d7d8Skucharsk mtrr4k3 = rdmsr(REG_MTRR4K3); 22420ac7d7d8Skucharsk mtrr4k4 = rdmsr(REG_MTRR4K4); 22430ac7d7d8Skucharsk mtrr4k5 = rdmsr(REG_MTRR4K5); 22440ac7d7d8Skucharsk mtrr4k6 = rdmsr(REG_MTRR4K6); 22450ac7d7d8Skucharsk mtrr4k7 = rdmsr(REG_MTRR4K7); 22460ac7d7d8Skucharsk mtrr4k8 = rdmsr(REG_MTRR4K8); 22477c478bd9Sstevel@tonic-gate } 22487c478bd9Sstevel@tonic-gate if ((vcnt = (mtrrcap & MTRRCAP_VCNTMASK)) > MAX_MTRRVAR) 22497c478bd9Sstevel@tonic-gate vcnt = MAX_MTRRVAR; 22507c478bd9Sstevel@tonic-gate 22517c478bd9Sstevel@tonic-gate for (i = 0, ecx = REG_MTRRPHYSBASE0, mtrrphys = mtrrphys_arr; 22527c478bd9Sstevel@tonic-gate i < vcnt - 1; i++, ecx += 2, mtrrphys++) { 22530ac7d7d8Skucharsk mtrrphys->mtrrphys_base = rdmsr(ecx); 22540ac7d7d8Skucharsk mtrrphys->mtrrphys_mask = rdmsr(ecx + 1); 22557c478bd9Sstevel@tonic-gate if ((x86_feature & X86_PAT) && enable_relaxed_mtrr) { 22567c478bd9Sstevel@tonic-gate mtrrphys->mtrrphys_mask &= ~MTRRPHYSMASK_V; 22577c478bd9Sstevel@tonic-gate } 22587c478bd9Sstevel@tonic-gate } 22597c478bd9Sstevel@tonic-gate if (x86_feature & X86_PAT) { 22607c478bd9Sstevel@tonic-gate if (enable_relaxed_mtrr) 22617c478bd9Sstevel@tonic-gate mtrrdef = MTRR_TYPE_WB|MTRRDEF_FE|MTRRDEF_E; 22627c478bd9Sstevel@tonic-gate pat_attr_reg = PAT_DEFAULT_ATTRIBUTE; 22637c478bd9Sstevel@tonic-gate } 22647c478bd9Sstevel@tonic-gate 22657c478bd9Sstevel@tonic-gate mtrr_sync(); 22667c478bd9Sstevel@tonic-gate } 22677c478bd9Sstevel@tonic-gate 22687c478bd9Sstevel@tonic-gate /* 22697c478bd9Sstevel@tonic-gate * Sync current cpu mtrr with the incore copy of mtrr. 22707c478bd9Sstevel@tonic-gate * This function has to be invoked with interrupts disabled 22717c478bd9Sstevel@tonic-gate * Currently we do not capture other cpu's. This is invoked on cpu0 22727c478bd9Sstevel@tonic-gate * just after reading /etc/system. 22737c478bd9Sstevel@tonic-gate * On other cpu's its invoked from mp_startup(). 22747c478bd9Sstevel@tonic-gate */ 22757c478bd9Sstevel@tonic-gate void 227645916cd2Sjpk mtrr_sync(void) 22777c478bd9Sstevel@tonic-gate { 22787c478bd9Sstevel@tonic-gate uint_t crvalue, cr0_orig; 22797c478bd9Sstevel@tonic-gate int vcnt, i, ecx; 22807c478bd9Sstevel@tonic-gate struct mtrrvar *mtrrphys; 22817c478bd9Sstevel@tonic-gate 22827c478bd9Sstevel@tonic-gate cr0_orig = crvalue = getcr0(); 22837c478bd9Sstevel@tonic-gate crvalue |= CR0_CD; 22847c478bd9Sstevel@tonic-gate crvalue &= ~CR0_NW; 22857c478bd9Sstevel@tonic-gate setcr0(crvalue); 22867c478bd9Sstevel@tonic-gate invalidate_cache(); 22877c478bd9Sstevel@tonic-gate 2288*ae115bc7Smrj reload_cr3(); 22890ac7d7d8Skucharsk if (x86_feature & X86_PAT) 22900ac7d7d8Skucharsk wrmsr(REG_MTRRPAT, pat_attr_reg); 22910ac7d7d8Skucharsk 22920ac7d7d8Skucharsk wrmsr(REG_MTRRDEF, rdmsr(REG_MTRRDEF) & 22930ac7d7d8Skucharsk ~((uint64_t)(uintptr_t)MTRRDEF_E)); 22940ac7d7d8Skucharsk 22957c478bd9Sstevel@tonic-gate if (mtrrcap & MTRRCAP_FIX) { 22960ac7d7d8Skucharsk wrmsr(REG_MTRR64K, mtrr64k); 22970ac7d7d8Skucharsk wrmsr(REG_MTRR16K1, mtrr16k1); 22980ac7d7d8Skucharsk wrmsr(REG_MTRR16K2, mtrr16k2); 22990ac7d7d8Skucharsk wrmsr(REG_MTRR4K1, mtrr4k1); 23000ac7d7d8Skucharsk wrmsr(REG_MTRR4K2, mtrr4k2); 23010ac7d7d8Skucharsk wrmsr(REG_MTRR4K3, mtrr4k3); 23020ac7d7d8Skucharsk wrmsr(REG_MTRR4K4, mtrr4k4); 23030ac7d7d8Skucharsk wrmsr(REG_MTRR4K5, mtrr4k5); 23040ac7d7d8Skucharsk wrmsr(REG_MTRR4K6, mtrr4k6); 23050ac7d7d8Skucharsk wrmsr(REG_MTRR4K7, mtrr4k7); 23060ac7d7d8Skucharsk wrmsr(REG_MTRR4K8, mtrr4k8); 23077c478bd9Sstevel@tonic-gate } 23087c478bd9Sstevel@tonic-gate if ((vcnt = (mtrrcap & MTRRCAP_VCNTMASK)) > MAX_MTRRVAR) 23097c478bd9Sstevel@tonic-gate vcnt = MAX_MTRRVAR; 23107c478bd9Sstevel@tonic-gate for (i = 0, ecx = REG_MTRRPHYSBASE0, mtrrphys = mtrrphys_arr; 23110ac7d7d8Skucharsk i < vcnt - 1; i++, ecx += 2, mtrrphys++) { 23120ac7d7d8Skucharsk wrmsr(ecx, mtrrphys->mtrrphys_base); 23130ac7d7d8Skucharsk wrmsr(ecx + 1, mtrrphys->mtrrphys_mask); 23147c478bd9Sstevel@tonic-gate } 23150ac7d7d8Skucharsk wrmsr(REG_MTRRDEF, mtrrdef); 2316*ae115bc7Smrj 2317*ae115bc7Smrj reload_cr3(); 23187c478bd9Sstevel@tonic-gate invalidate_cache(); 23197c478bd9Sstevel@tonic-gate setcr0(cr0_orig); 23207c478bd9Sstevel@tonic-gate } 23217c478bd9Sstevel@tonic-gate 23227c478bd9Sstevel@tonic-gate /* 23237c478bd9Sstevel@tonic-gate * resync mtrr so that BIOS is happy. Called from mdboot 23247c478bd9Sstevel@tonic-gate */ 23257c478bd9Sstevel@tonic-gate void 232645916cd2Sjpk mtrr_resync(void) 23277c478bd9Sstevel@tonic-gate { 23287c478bd9Sstevel@tonic-gate if ((x86_feature & X86_PAT) && enable_relaxed_mtrr) { 23297c478bd9Sstevel@tonic-gate /* 23307c478bd9Sstevel@tonic-gate * We could have changed the default mtrr definition. 23317c478bd9Sstevel@tonic-gate * Put it back to uncached which is what it is at power on 23327c478bd9Sstevel@tonic-gate */ 23337c478bd9Sstevel@tonic-gate mtrrdef = MTRR_TYPE_UC|MTRRDEF_FE|MTRRDEF_E; 23347c478bd9Sstevel@tonic-gate mtrr_sync(); 23357c478bd9Sstevel@tonic-gate } 23367c478bd9Sstevel@tonic-gate } 23377c478bd9Sstevel@tonic-gate 23387c478bd9Sstevel@tonic-gate void 233945916cd2Sjpk get_system_configuration(void) 23407c478bd9Sstevel@tonic-gate { 23417c478bd9Sstevel@tonic-gate char prop[32]; 23427c478bd9Sstevel@tonic-gate u_longlong_t nodes_ll, cpus_pernode_ll, lvalue; 23437c478bd9Sstevel@tonic-gate 23447c478bd9Sstevel@tonic-gate if (((BOP_GETPROPLEN(bootops, "nodes") > sizeof (prop)) || 23457c478bd9Sstevel@tonic-gate (BOP_GETPROP(bootops, "nodes", prop) < 0) || 23467c478bd9Sstevel@tonic-gate (kobj_getvalue(prop, &nodes_ll) == -1) || 23477c478bd9Sstevel@tonic-gate (nodes_ll > MAXNODES)) || 23487c478bd9Sstevel@tonic-gate ((BOP_GETPROPLEN(bootops, "cpus_pernode") > sizeof (prop)) || 23497c478bd9Sstevel@tonic-gate (BOP_GETPROP(bootops, "cpus_pernode", prop) < 0) || 23507c478bd9Sstevel@tonic-gate (kobj_getvalue(prop, &cpus_pernode_ll) == -1))) { 23517c478bd9Sstevel@tonic-gate 23527c478bd9Sstevel@tonic-gate system_hardware.hd_nodes = 1; 23537c478bd9Sstevel@tonic-gate system_hardware.hd_cpus_per_node = 0; 23547c478bd9Sstevel@tonic-gate } else { 23557c478bd9Sstevel@tonic-gate system_hardware.hd_nodes = (int)nodes_ll; 23567c478bd9Sstevel@tonic-gate system_hardware.hd_cpus_per_node = (int)cpus_pernode_ll; 23577c478bd9Sstevel@tonic-gate } 23587c478bd9Sstevel@tonic-gate if ((BOP_GETPROPLEN(bootops, "kernelbase") > sizeof (prop)) || 23597c478bd9Sstevel@tonic-gate (BOP_GETPROP(bootops, "kernelbase", prop) < 0) || 23607c478bd9Sstevel@tonic-gate (kobj_getvalue(prop, &lvalue) == -1)) 23617c478bd9Sstevel@tonic-gate eprom_kernelbase = NULL; 23627c478bd9Sstevel@tonic-gate else 23637c478bd9Sstevel@tonic-gate eprom_kernelbase = (uintptr_t)lvalue; 23647c478bd9Sstevel@tonic-gate 23657c478bd9Sstevel@tonic-gate if ((BOP_GETPROPLEN(bootops, "segmapsize") > sizeof (prop)) || 23667c478bd9Sstevel@tonic-gate (BOP_GETPROP(bootops, "segmapsize", prop) < 0) || 23677c478bd9Sstevel@tonic-gate (kobj_getvalue(prop, &lvalue) == -1)) { 23687c478bd9Sstevel@tonic-gate segmapsize = SEGMAPDEFAULT; 23697c478bd9Sstevel@tonic-gate } else { 23707c478bd9Sstevel@tonic-gate segmapsize = (uintptr_t)lvalue; 23717c478bd9Sstevel@tonic-gate } 23727c478bd9Sstevel@tonic-gate 23737c478bd9Sstevel@tonic-gate if ((BOP_GETPROPLEN(bootops, "segmapfreelists") > sizeof (prop)) || 23747c478bd9Sstevel@tonic-gate (BOP_GETPROP(bootops, "segmapfreelists", prop) < 0) || 23757c478bd9Sstevel@tonic-gate (kobj_getvalue(prop, &lvalue) == -1)) { 23767c478bd9Sstevel@tonic-gate segmapfreelists = 0; /* use segmap driver default */ 23777c478bd9Sstevel@tonic-gate } else { 23787c478bd9Sstevel@tonic-gate segmapfreelists = (int)lvalue; 23797c478bd9Sstevel@tonic-gate } 23804944b02eSkchow 23814944b02eSkchow if ((BOP_GETPROPLEN(bootops, "physmem") <= sizeof (prop)) && 23824944b02eSkchow (BOP_GETPROP(bootops, "physmem", prop) >= 0) && 23834944b02eSkchow (kobj_getvalue(prop, &lvalue) != -1)) { 23844944b02eSkchow physmem = (uintptr_t)lvalue; 23854944b02eSkchow } 23867c478bd9Sstevel@tonic-gate } 23877c478bd9Sstevel@tonic-gate 23887c478bd9Sstevel@tonic-gate /* 23897c478bd9Sstevel@tonic-gate * Add to a memory list. 23907c478bd9Sstevel@tonic-gate * start = start of new memory segment 23917c478bd9Sstevel@tonic-gate * len = length of new memory segment in bytes 23927c478bd9Sstevel@tonic-gate * new = pointer to a new struct memlist 23937c478bd9Sstevel@tonic-gate * memlistp = memory list to which to add segment. 23947c478bd9Sstevel@tonic-gate */ 2395*ae115bc7Smrj void 23967c478bd9Sstevel@tonic-gate memlist_add( 23977c478bd9Sstevel@tonic-gate uint64_t start, 23987c478bd9Sstevel@tonic-gate uint64_t len, 23997c478bd9Sstevel@tonic-gate struct memlist *new, 24007c478bd9Sstevel@tonic-gate struct memlist **memlistp) 24017c478bd9Sstevel@tonic-gate { 24027c478bd9Sstevel@tonic-gate struct memlist *cur; 24037c478bd9Sstevel@tonic-gate uint64_t end = start + len; 24047c478bd9Sstevel@tonic-gate 24057c478bd9Sstevel@tonic-gate new->address = start; 24067c478bd9Sstevel@tonic-gate new->size = len; 24077c478bd9Sstevel@tonic-gate 24087c478bd9Sstevel@tonic-gate cur = *memlistp; 24097c478bd9Sstevel@tonic-gate 24107c478bd9Sstevel@tonic-gate while (cur) { 24117c478bd9Sstevel@tonic-gate if (cur->address >= end) { 24127c478bd9Sstevel@tonic-gate new->next = cur; 24137c478bd9Sstevel@tonic-gate *memlistp = new; 24147c478bd9Sstevel@tonic-gate new->prev = cur->prev; 24157c478bd9Sstevel@tonic-gate cur->prev = new; 24167c478bd9Sstevel@tonic-gate return; 24177c478bd9Sstevel@tonic-gate } 24187c478bd9Sstevel@tonic-gate ASSERT(cur->address + cur->size <= start); 24197c478bd9Sstevel@tonic-gate if (cur->next == NULL) { 24207c478bd9Sstevel@tonic-gate cur->next = new; 24217c478bd9Sstevel@tonic-gate new->prev = cur; 24227c478bd9Sstevel@tonic-gate new->next = NULL; 24237c478bd9Sstevel@tonic-gate return; 24247c478bd9Sstevel@tonic-gate } 24257c478bd9Sstevel@tonic-gate memlistp = &cur->next; 24267c478bd9Sstevel@tonic-gate cur = cur->next; 24277c478bd9Sstevel@tonic-gate } 24287c478bd9Sstevel@tonic-gate } 24297c478bd9Sstevel@tonic-gate 24307c478bd9Sstevel@tonic-gate void 24317c478bd9Sstevel@tonic-gate kobj_vmem_init(vmem_t **text_arena, vmem_t **data_arena) 24327c478bd9Sstevel@tonic-gate { 24337c478bd9Sstevel@tonic-gate size_t tsize = e_modtext - modtext; 24347c478bd9Sstevel@tonic-gate size_t dsize = e_moddata - moddata; 24357c478bd9Sstevel@tonic-gate 24367c478bd9Sstevel@tonic-gate *text_arena = vmem_create("module_text", tsize ? modtext : NULL, tsize, 24377c478bd9Sstevel@tonic-gate 1, segkmem_alloc, segkmem_free, heaptext_arena, 0, VM_SLEEP); 24387c478bd9Sstevel@tonic-gate *data_arena = vmem_create("module_data", dsize ? moddata : NULL, dsize, 24397c478bd9Sstevel@tonic-gate 1, segkmem_alloc, segkmem_free, heap32_arena, 0, VM_SLEEP); 24407c478bd9Sstevel@tonic-gate } 24417c478bd9Sstevel@tonic-gate 24427c478bd9Sstevel@tonic-gate caddr_t 24437c478bd9Sstevel@tonic-gate kobj_text_alloc(vmem_t *arena, size_t size) 24447c478bd9Sstevel@tonic-gate { 24457c478bd9Sstevel@tonic-gate return (vmem_alloc(arena, size, VM_SLEEP | VM_BESTFIT)); 24467c478bd9Sstevel@tonic-gate } 24477c478bd9Sstevel@tonic-gate 24487c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 24497c478bd9Sstevel@tonic-gate caddr_t 24507c478bd9Sstevel@tonic-gate kobj_texthole_alloc(caddr_t addr, size_t size) 24517c478bd9Sstevel@tonic-gate { 24527c478bd9Sstevel@tonic-gate panic("unexpected call to kobj_texthole_alloc()"); 24537c478bd9Sstevel@tonic-gate /*NOTREACHED*/ 24547c478bd9Sstevel@tonic-gate return (0); 24557c478bd9Sstevel@tonic-gate } 24567c478bd9Sstevel@tonic-gate 24577c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 24587c478bd9Sstevel@tonic-gate void 24597c478bd9Sstevel@tonic-gate kobj_texthole_free(caddr_t addr, size_t size) 24607c478bd9Sstevel@tonic-gate { 24617c478bd9Sstevel@tonic-gate panic("unexpected call to kobj_texthole_free()"); 24627c478bd9Sstevel@tonic-gate } 24637c478bd9Sstevel@tonic-gate 24647c478bd9Sstevel@tonic-gate /* 24657c478bd9Sstevel@tonic-gate * This is called just after configure() in startup(). 24667c478bd9Sstevel@tonic-gate * 24677c478bd9Sstevel@tonic-gate * The ISALIST concept is a bit hopeless on Intel, because 24687c478bd9Sstevel@tonic-gate * there's no guarantee of an ever-more-capable processor 24697c478bd9Sstevel@tonic-gate * given that various parts of the instruction set may appear 24707c478bd9Sstevel@tonic-gate * and disappear between different implementations. 24717c478bd9Sstevel@tonic-gate * 24727c478bd9Sstevel@tonic-gate * While it would be possible to correct it and even enhance 24737c478bd9Sstevel@tonic-gate * it somewhat, the explicit hardware capability bitmask allows 24747c478bd9Sstevel@tonic-gate * more flexibility. 24757c478bd9Sstevel@tonic-gate * 24767c478bd9Sstevel@tonic-gate * So, we just leave this alone. 24777c478bd9Sstevel@tonic-gate */ 24787c478bd9Sstevel@tonic-gate void 24797c478bd9Sstevel@tonic-gate setx86isalist(void) 24807c478bd9Sstevel@tonic-gate { 24817c478bd9Sstevel@tonic-gate char *tp; 24827c478bd9Sstevel@tonic-gate size_t len; 24837c478bd9Sstevel@tonic-gate extern char *isa_list; 24847c478bd9Sstevel@tonic-gate 24857c478bd9Sstevel@tonic-gate #define TBUFSIZE 1024 24867c478bd9Sstevel@tonic-gate 24877c478bd9Sstevel@tonic-gate tp = kmem_alloc(TBUFSIZE, KM_SLEEP); 24887c478bd9Sstevel@tonic-gate *tp = '\0'; 24897c478bd9Sstevel@tonic-gate 24907c478bd9Sstevel@tonic-gate #if defined(__amd64) 24917c478bd9Sstevel@tonic-gate (void) strcpy(tp, "amd64 "); 24927c478bd9Sstevel@tonic-gate #endif 24937c478bd9Sstevel@tonic-gate 24947c478bd9Sstevel@tonic-gate switch (x86_vendor) { 24957c478bd9Sstevel@tonic-gate case X86_VENDOR_Intel: 24967c478bd9Sstevel@tonic-gate case X86_VENDOR_AMD: 24977c478bd9Sstevel@tonic-gate case X86_VENDOR_TM: 24987c478bd9Sstevel@tonic-gate if (x86_feature & X86_CMOV) { 24997c478bd9Sstevel@tonic-gate /* 25007c478bd9Sstevel@tonic-gate * Pentium Pro or later 25017c478bd9Sstevel@tonic-gate */ 25027c478bd9Sstevel@tonic-gate (void) strcat(tp, "pentium_pro"); 25037c478bd9Sstevel@tonic-gate (void) strcat(tp, x86_feature & X86_MMX ? 25047c478bd9Sstevel@tonic-gate "+mmx pentium_pro " : " "); 25057c478bd9Sstevel@tonic-gate } 25067c478bd9Sstevel@tonic-gate /*FALLTHROUGH*/ 25077c478bd9Sstevel@tonic-gate case X86_VENDOR_Cyrix: 25087c478bd9Sstevel@tonic-gate /* 25097c478bd9Sstevel@tonic-gate * The Cyrix 6x86 does not have any Pentium features 25107c478bd9Sstevel@tonic-gate * accessible while not at privilege level 0. 25117c478bd9Sstevel@tonic-gate */ 25127c478bd9Sstevel@tonic-gate if (x86_feature & X86_CPUID) { 25137c478bd9Sstevel@tonic-gate (void) strcat(tp, "pentium"); 25147c478bd9Sstevel@tonic-gate (void) strcat(tp, x86_feature & X86_MMX ? 25157c478bd9Sstevel@tonic-gate "+mmx pentium " : " "); 25167c478bd9Sstevel@tonic-gate } 25177c478bd9Sstevel@tonic-gate break; 25187c478bd9Sstevel@tonic-gate default: 25197c478bd9Sstevel@tonic-gate break; 25207c478bd9Sstevel@tonic-gate } 25217c478bd9Sstevel@tonic-gate (void) strcat(tp, "i486 i386 i86"); 25227c478bd9Sstevel@tonic-gate len = strlen(tp) + 1; /* account for NULL at end of string */ 25237c478bd9Sstevel@tonic-gate isa_list = strcpy(kmem_alloc(len, KM_SLEEP), tp); 25247c478bd9Sstevel@tonic-gate kmem_free(tp, TBUFSIZE); 25257c478bd9Sstevel@tonic-gate 25267c478bd9Sstevel@tonic-gate #undef TBUFSIZE 25277c478bd9Sstevel@tonic-gate } 25287c478bd9Sstevel@tonic-gate 25297c478bd9Sstevel@tonic-gate 25307c478bd9Sstevel@tonic-gate #ifdef __amd64 25317c478bd9Sstevel@tonic-gate 25327c478bd9Sstevel@tonic-gate void * 25337c478bd9Sstevel@tonic-gate device_arena_alloc(size_t size, int vm_flag) 25347c478bd9Sstevel@tonic-gate { 25357c478bd9Sstevel@tonic-gate return (vmem_alloc(device_arena, size, vm_flag)); 25367c478bd9Sstevel@tonic-gate } 25377c478bd9Sstevel@tonic-gate 25387c478bd9Sstevel@tonic-gate void 25397c478bd9Sstevel@tonic-gate device_arena_free(void *vaddr, size_t size) 25407c478bd9Sstevel@tonic-gate { 25417c478bd9Sstevel@tonic-gate vmem_free(device_arena, vaddr, size); 25427c478bd9Sstevel@tonic-gate } 25437c478bd9Sstevel@tonic-gate 2544*ae115bc7Smrj #else /* __i386 */ 25457c478bd9Sstevel@tonic-gate 25467c478bd9Sstevel@tonic-gate void * 25477c478bd9Sstevel@tonic-gate device_arena_alloc(size_t size, int vm_flag) 25487c478bd9Sstevel@tonic-gate { 25497c478bd9Sstevel@tonic-gate caddr_t vaddr; 25507c478bd9Sstevel@tonic-gate uintptr_t v; 25517c478bd9Sstevel@tonic-gate size_t start; 25527c478bd9Sstevel@tonic-gate size_t end; 25537c478bd9Sstevel@tonic-gate 25547c478bd9Sstevel@tonic-gate vaddr = vmem_alloc(heap_arena, size, vm_flag); 25557c478bd9Sstevel@tonic-gate if (vaddr == NULL) 25567c478bd9Sstevel@tonic-gate return (NULL); 25577c478bd9Sstevel@tonic-gate 25587c478bd9Sstevel@tonic-gate v = (uintptr_t)vaddr; 25597c478bd9Sstevel@tonic-gate ASSERT(v >= kernelbase); 2560*ae115bc7Smrj ASSERT(v + size <= valloc_base); 25617c478bd9Sstevel@tonic-gate 25627c478bd9Sstevel@tonic-gate start = btop(v - kernelbase); 25637c478bd9Sstevel@tonic-gate end = btop(v + size - 1 - kernelbase); 25647c478bd9Sstevel@tonic-gate ASSERT(start < toxic_bit_map_len); 25657c478bd9Sstevel@tonic-gate ASSERT(end < toxic_bit_map_len); 25667c478bd9Sstevel@tonic-gate 25677c478bd9Sstevel@tonic-gate while (start <= end) { 25687c478bd9Sstevel@tonic-gate BT_ATOMIC_SET(toxic_bit_map, start); 25697c478bd9Sstevel@tonic-gate ++start; 25707c478bd9Sstevel@tonic-gate } 25717c478bd9Sstevel@tonic-gate return (vaddr); 25727c478bd9Sstevel@tonic-gate } 25737c478bd9Sstevel@tonic-gate 25747c478bd9Sstevel@tonic-gate void 25757c478bd9Sstevel@tonic-gate device_arena_free(void *vaddr, size_t size) 25767c478bd9Sstevel@tonic-gate { 25777c478bd9Sstevel@tonic-gate uintptr_t v = (uintptr_t)vaddr; 25787c478bd9Sstevel@tonic-gate size_t start; 25797c478bd9Sstevel@tonic-gate size_t end; 25807c478bd9Sstevel@tonic-gate 25817c478bd9Sstevel@tonic-gate ASSERT(v >= kernelbase); 2582*ae115bc7Smrj ASSERT(v + size <= valloc_base); 25837c478bd9Sstevel@tonic-gate 25847c478bd9Sstevel@tonic-gate start = btop(v - kernelbase); 25857c478bd9Sstevel@tonic-gate end = btop(v + size - 1 - kernelbase); 25867c478bd9Sstevel@tonic-gate ASSERT(start < toxic_bit_map_len); 25877c478bd9Sstevel@tonic-gate ASSERT(end < toxic_bit_map_len); 25887c478bd9Sstevel@tonic-gate 25897c478bd9Sstevel@tonic-gate while (start <= end) { 25907c478bd9Sstevel@tonic-gate ASSERT(BT_TEST(toxic_bit_map, start) != 0); 25917c478bd9Sstevel@tonic-gate BT_ATOMIC_CLEAR(toxic_bit_map, start); 25927c478bd9Sstevel@tonic-gate ++start; 25937c478bd9Sstevel@tonic-gate } 25947c478bd9Sstevel@tonic-gate vmem_free(heap_arena, vaddr, size); 25957c478bd9Sstevel@tonic-gate } 25967c478bd9Sstevel@tonic-gate 25977c478bd9Sstevel@tonic-gate /* 25987c478bd9Sstevel@tonic-gate * returns 1st address in range that is in device arena, or NULL 25997c478bd9Sstevel@tonic-gate * if len is not NULL it returns the length of the toxic range 26007c478bd9Sstevel@tonic-gate */ 26017c478bd9Sstevel@tonic-gate void * 26027c478bd9Sstevel@tonic-gate device_arena_contains(void *vaddr, size_t size, size_t *len) 26037c478bd9Sstevel@tonic-gate { 26047c478bd9Sstevel@tonic-gate uintptr_t v = (uintptr_t)vaddr; 26057c478bd9Sstevel@tonic-gate uintptr_t eaddr = v + size; 26067c478bd9Sstevel@tonic-gate size_t start; 26077c478bd9Sstevel@tonic-gate size_t end; 26087c478bd9Sstevel@tonic-gate 26097c478bd9Sstevel@tonic-gate /* 26107c478bd9Sstevel@tonic-gate * if called very early by kmdb, just return NULL 26117c478bd9Sstevel@tonic-gate */ 26127c478bd9Sstevel@tonic-gate if (toxic_bit_map == NULL) 26137c478bd9Sstevel@tonic-gate return (NULL); 26147c478bd9Sstevel@tonic-gate 26157c478bd9Sstevel@tonic-gate /* 26167c478bd9Sstevel@tonic-gate * First check if we're completely outside the bitmap range. 26177c478bd9Sstevel@tonic-gate */ 2618*ae115bc7Smrj if (v >= valloc_base || eaddr < kernelbase) 26197c478bd9Sstevel@tonic-gate return (NULL); 26207c478bd9Sstevel@tonic-gate 26217c478bd9Sstevel@tonic-gate /* 26227c478bd9Sstevel@tonic-gate * Trim ends of search to look at only what the bitmap covers. 26237c478bd9Sstevel@tonic-gate */ 26247c478bd9Sstevel@tonic-gate if (v < kernelbase) 26257c478bd9Sstevel@tonic-gate v = kernelbase; 26267c478bd9Sstevel@tonic-gate start = btop(v - kernelbase); 26277c478bd9Sstevel@tonic-gate end = btop(eaddr - kernelbase); 26287c478bd9Sstevel@tonic-gate if (end >= toxic_bit_map_len) 26297c478bd9Sstevel@tonic-gate end = toxic_bit_map_len; 26307c478bd9Sstevel@tonic-gate 26317c478bd9Sstevel@tonic-gate if (bt_range(toxic_bit_map, &start, &end, end) == 0) 26327c478bd9Sstevel@tonic-gate return (NULL); 26337c478bd9Sstevel@tonic-gate 26347c478bd9Sstevel@tonic-gate v = kernelbase + ptob(start); 26357c478bd9Sstevel@tonic-gate if (len != NULL) 26367c478bd9Sstevel@tonic-gate *len = ptob(end - start); 26377c478bd9Sstevel@tonic-gate return ((void *)v); 26387c478bd9Sstevel@tonic-gate } 26397c478bd9Sstevel@tonic-gate 2640*ae115bc7Smrj #endif /* __i386 */ 2641