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 /* 22*a563a037Sbholler * Copyright 2008 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> 49ae115bc7Smrj #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> 70843e1988Sjohnlev #include <sys/modctl.h> 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> 841d03c31eSjohnlev #include <vm/kboot_mmu.h> 857c478bd9Sstevel@tonic-gate #include <vm/anon.h> 867c478bd9Sstevel@tonic-gate #include <vm/as.h> 877c478bd9Sstevel@tonic-gate #include <vm/page.h> 887c478bd9Sstevel@tonic-gate #include <vm/seg.h> 897c478bd9Sstevel@tonic-gate #include <vm/seg_dev.h> 907c478bd9Sstevel@tonic-gate #include <vm/seg_kmem.h> 917c478bd9Sstevel@tonic-gate #include <vm/seg_kpm.h> 927c478bd9Sstevel@tonic-gate #include <vm/seg_map.h> 937c478bd9Sstevel@tonic-gate #include <vm/seg_vn.h> 947c478bd9Sstevel@tonic-gate #include <vm/seg_kp.h> 957c478bd9Sstevel@tonic-gate #include <sys/memnode.h> 967c478bd9Sstevel@tonic-gate #include <vm/vm_dep.h> 977c478bd9Sstevel@tonic-gate #include <sys/thread.h> 987c478bd9Sstevel@tonic-gate #include <sys/sysconf.h> 997c478bd9Sstevel@tonic-gate #include <sys/vm_machparam.h> 1007c478bd9Sstevel@tonic-gate #include <sys/archsystm.h> 1017c478bd9Sstevel@tonic-gate #include <sys/machsystm.h> 1027c478bd9Sstevel@tonic-gate #include <vm/hat.h> 1037c478bd9Sstevel@tonic-gate #include <vm/hat_i86.h> 1047c478bd9Sstevel@tonic-gate #include <sys/pmem.h> 1057c478bd9Sstevel@tonic-gate #include <sys/smp_impldefs.h> 1067c478bd9Sstevel@tonic-gate #include <sys/x86_archext.h> 107*a563a037Sbholler #include <sys/cpuvar.h> 1087c478bd9Sstevel@tonic-gate #include <sys/segments.h> 1097c478bd9Sstevel@tonic-gate #include <sys/clconf.h> 1107c478bd9Sstevel@tonic-gate #include <sys/kobj.h> 1117c478bd9Sstevel@tonic-gate #include <sys/kobj_lex.h> 1127c478bd9Sstevel@tonic-gate #include <sys/cpc_impl.h> 1137c478bd9Sstevel@tonic-gate #include <sys/x86_archext.h> 1147aec1d6eScindi #include <sys/cpu_module.h> 11584ab085aSmws #include <sys/smbios.h> 116ae115bc7Smrj #include <sys/debug_info.h> 1171d03c31eSjohnlev #include <sys/bootinfo.h> 118dd4eeefdSeota #include <sys/ddi_timer.h> 119ae115bc7Smrj 120843e1988Sjohnlev #ifdef __xpv 1211d03c31eSjohnlev 122843e1988Sjohnlev #include <sys/hypervisor.h> 123843e1988Sjohnlev #include <sys/xen_mmu.h> 124843e1988Sjohnlev #include <sys/evtchn_impl.h> 125843e1988Sjohnlev #include <sys/gnttab.h> 126843e1988Sjohnlev #include <sys/xpv_panic.h> 127843e1988Sjohnlev #include <xen/sys/xenbus_comms.h> 128843e1988Sjohnlev #include <xen/public/physdev.h> 1291d03c31eSjohnlev 130843e1988Sjohnlev extern void xen_late_startup(void); 131ae115bc7Smrj 1321d03c31eSjohnlev struct xen_evt_data cpu0_evt_data; 1331d03c31eSjohnlev 1341d03c31eSjohnlev #endif /* __xpv */ 1357c478bd9Sstevel@tonic-gate 1367c478bd9Sstevel@tonic-gate extern void progressbar_init(void); 1377c478bd9Sstevel@tonic-gate extern void progressbar_start(void); 1389acbbeafSnn extern void brand_init(void); 1397c478bd9Sstevel@tonic-gate 140d7d93655Sblakej extern int size_pse_array(pgcnt_t, int); 141d7d93655Sblakej 1427c478bd9Sstevel@tonic-gate /* 1437c478bd9Sstevel@tonic-gate * XXX make declaration below "static" when drivers no longer use this 1447c478bd9Sstevel@tonic-gate * interface. 1457c478bd9Sstevel@tonic-gate */ 1467c478bd9Sstevel@tonic-gate extern caddr_t p0_va; /* Virtual address for accessing physical page 0 */ 1477c478bd9Sstevel@tonic-gate 1487c478bd9Sstevel@tonic-gate /* 1497c478bd9Sstevel@tonic-gate * segkp 1507c478bd9Sstevel@tonic-gate */ 1517c478bd9Sstevel@tonic-gate extern int segkp_fromheap; 1527c478bd9Sstevel@tonic-gate 1537c478bd9Sstevel@tonic-gate static void kvm_init(void); 1547c478bd9Sstevel@tonic-gate static void startup_init(void); 1557c478bd9Sstevel@tonic-gate static void startup_memlist(void); 156ae115bc7Smrj static void startup_kmem(void); 1577c478bd9Sstevel@tonic-gate static void startup_modules(void); 1587c478bd9Sstevel@tonic-gate static void startup_vm(void); 1597c478bd9Sstevel@tonic-gate static void startup_end(void); 16035b1ab99Sjosephb static void layout_kernel_va(void); 1617c478bd9Sstevel@tonic-gate 1627c478bd9Sstevel@tonic-gate /* 1637c478bd9Sstevel@tonic-gate * Declare these as initialized data so we can patch them. 1647c478bd9Sstevel@tonic-gate */ 165f53ad214Skchow #ifdef __i386 16635b1ab99Sjosephb 167f53ad214Skchow /* 168f53ad214Skchow * Due to virtual address space limitations running in 32 bit mode, restrict 16935b1ab99Sjosephb * the amount of physical memory configured to a max of PHYSMEM pages (16g). 170f53ad214Skchow * 171f53ad214Skchow * If the physical max memory size of 64g were allowed to be configured, the 172f53ad214Skchow * size of user virtual address space will be less than 1g. A limited user 173f53ad214Skchow * address space greatly reduces the range of applications that can run. 174f53ad214Skchow * 17535b1ab99Sjosephb * If more physical memory than PHYSMEM is required, users should preferably 17635b1ab99Sjosephb * run in 64 bit mode which has far looser virtual address space limitations. 177f53ad214Skchow * 178f53ad214Skchow * If 64 bit mode is not available (as in IA32) and/or more physical memory 17935b1ab99Sjosephb * than PHYSMEM is required in 32 bit mode, physmem can be set to the desired 180f53ad214Skchow * value or to 0 (to configure all available memory) via eeprom(1M). kernelbase 181f53ad214Skchow * should also be carefully tuned to balance out the need of the user 182f53ad214Skchow * application while minimizing the risk of kernel heap exhaustion due to 183f53ad214Skchow * kernelbase being set too high. 184f53ad214Skchow */ 18535b1ab99Sjosephb #define PHYSMEM 0x400000 186f53ad214Skchow 18735b1ab99Sjosephb #else /* __amd64 */ 18835b1ab99Sjosephb 18935b1ab99Sjosephb /* 19035b1ab99Sjosephb * For now we can handle memory with physical addresses up to about 19135b1ab99Sjosephb * 64 Terabytes. This keeps the kernel above the VA hole, leaving roughly 19235b1ab99Sjosephb * half the VA space for seg_kpm. When systems get bigger than 64TB this 19335b1ab99Sjosephb * code will need revisiting. There is an implicit assumption that there 19435b1ab99Sjosephb * are no *huge* holes in the physical address space too. 19535b1ab99Sjosephb */ 19635b1ab99Sjosephb #define TERABYTE (1ul << 40) 19735b1ab99Sjosephb #define PHYSMEM_MAX64 mmu_btop(64 * TERABYTE) 19835b1ab99Sjosephb #define PHYSMEM PHYSMEM_MAX64 19935b1ab99Sjosephb #define AMD64_VA_HOLE_END 0xFFFF800000000000ul 20035b1ab99Sjosephb 20135b1ab99Sjosephb #endif /* __amd64 */ 20235b1ab99Sjosephb 20335b1ab99Sjosephb pgcnt_t physmem = PHYSMEM; 2047c478bd9Sstevel@tonic-gate pgcnt_t obp_pages; /* Memory used by PROM for its text and data */ 2057c478bd9Sstevel@tonic-gate 2067c478bd9Sstevel@tonic-gate char *kobj_file_buf; 2077c478bd9Sstevel@tonic-gate int kobj_file_bufsize; /* set in /etc/system */ 2087c478bd9Sstevel@tonic-gate 2097c478bd9Sstevel@tonic-gate /* Global variables for MP support. Used in mp_startup */ 2107c478bd9Sstevel@tonic-gate caddr_t rm_platter_va; 2117c478bd9Sstevel@tonic-gate uint32_t rm_platter_pa; 2127c478bd9Sstevel@tonic-gate 21383f9b804Skchow int auto_lpg_disable = 1; 21483f9b804Skchow 2157c478bd9Sstevel@tonic-gate /* 2167c478bd9Sstevel@tonic-gate * Some CPUs have holes in the middle of the 64-bit virtual address range. 2177c478bd9Sstevel@tonic-gate */ 2187c478bd9Sstevel@tonic-gate uintptr_t hole_start, hole_end; 2197c478bd9Sstevel@tonic-gate 2207c478bd9Sstevel@tonic-gate /* 2217c478bd9Sstevel@tonic-gate * kpm mapping window 2227c478bd9Sstevel@tonic-gate */ 2237c478bd9Sstevel@tonic-gate caddr_t kpm_vbase; 2247c478bd9Sstevel@tonic-gate size_t kpm_size; 22535b1ab99Sjosephb static int kpm_desired; 22635b1ab99Sjosephb #ifdef __amd64 22735b1ab99Sjosephb static uintptr_t segkpm_base = (uintptr_t)SEGKPM_BASE; 22835b1ab99Sjosephb #endif 2297c478bd9Sstevel@tonic-gate 2307c478bd9Sstevel@tonic-gate /* 2317c478bd9Sstevel@tonic-gate * Configuration parameters set at boot time. 2327c478bd9Sstevel@tonic-gate */ 2337c478bd9Sstevel@tonic-gate 2347c478bd9Sstevel@tonic-gate caddr_t econtig; /* end of first block of contiguous kernel */ 2357c478bd9Sstevel@tonic-gate 2367c478bd9Sstevel@tonic-gate struct bootops *bootops = 0; /* passed in from boot */ 2377c478bd9Sstevel@tonic-gate struct bootops **bootopsp; 2387c478bd9Sstevel@tonic-gate struct boot_syscalls *sysp; /* passed in from boot */ 2397c478bd9Sstevel@tonic-gate 2407c478bd9Sstevel@tonic-gate char bootblock_fstype[16]; 2417c478bd9Sstevel@tonic-gate 2427c478bd9Sstevel@tonic-gate char kern_bootargs[OBP_MAXPATHLEN]; 2437c478bd9Sstevel@tonic-gate 244ad23a2dbSjohansen /* 245ad23a2dbSjohansen * ZFS zio segment. This allows us to exclude large portions of ZFS data that 246ad23a2dbSjohansen * gets cached in kmem caches on the heap. If this is set to zero, we allocate 247ad23a2dbSjohansen * zio buffers from their own segment, otherwise they are allocated from the 248ad23a2dbSjohansen * heap. The optimization of allocating zio buffers from their own segment is 249ad23a2dbSjohansen * only valid on 64-bit kernels. 250ad23a2dbSjohansen */ 251ad23a2dbSjohansen #if defined(__amd64) 252ad23a2dbSjohansen int segzio_fromheap = 0; 253ad23a2dbSjohansen #else 254ad23a2dbSjohansen int segzio_fromheap = 1; 255ad23a2dbSjohansen #endif 256ad23a2dbSjohansen 2577c478bd9Sstevel@tonic-gate /* 2587c478bd9Sstevel@tonic-gate * new memory fragmentations are possible in startup() due to BOP_ALLOCs. this 2597c478bd9Sstevel@tonic-gate * depends on number of BOP_ALLOC calls made and requested size, memory size 2607c478bd9Sstevel@tonic-gate * combination and whether boot.bin memory needs to be freed. 2617c478bd9Sstevel@tonic-gate */ 2627c478bd9Sstevel@tonic-gate #define POSS_NEW_FRAGMENTS 12 2637c478bd9Sstevel@tonic-gate 2647c478bd9Sstevel@tonic-gate /* 2657c478bd9Sstevel@tonic-gate * VM data structures 2667c478bd9Sstevel@tonic-gate */ 2677c478bd9Sstevel@tonic-gate long page_hashsz; /* Size of page hash table (power of two) */ 2687c478bd9Sstevel@tonic-gate struct page *pp_base; /* Base of initial system page struct array */ 2697c478bd9Sstevel@tonic-gate struct page **page_hash; /* Page hash table */ 270d7d93655Sblakej pad_mutex_t *pse_mutex; /* Locks protecting pp->p_selock */ 271d7d93655Sblakej size_t pse_table_size; /* Number of mutexes in pse_mutex[] */ 272d7d93655Sblakej int pse_shift; /* log2(pse_table_size) */ 2737c478bd9Sstevel@tonic-gate struct seg ktextseg; /* Segment used for kernel executable image */ 2747c478bd9Sstevel@tonic-gate struct seg kvalloc; /* Segment used for "valloc" mapping */ 2757c478bd9Sstevel@tonic-gate struct seg kpseg; /* Segment used for pageable kernel virt mem */ 2767c478bd9Sstevel@tonic-gate struct seg kmapseg; /* Segment used for generic kernel mappings */ 2777c478bd9Sstevel@tonic-gate struct seg kdebugseg; /* Segment used for the kernel debugger */ 2787c478bd9Sstevel@tonic-gate 2797c478bd9Sstevel@tonic-gate struct seg *segkmap = &kmapseg; /* Kernel generic mapping segment */ 280ae115bc7Smrj static struct seg *segmap = &kmapseg; /* easier to use name for in here */ 281ae115bc7Smrj 2827c478bd9Sstevel@tonic-gate struct seg *segkp = &kpseg; /* Pageable kernel virtual memory segment */ 2837c478bd9Sstevel@tonic-gate 2847c478bd9Sstevel@tonic-gate #if defined(__amd64) 2857c478bd9Sstevel@tonic-gate struct seg kvseg_core; /* Segment used for the core heap */ 2867c478bd9Sstevel@tonic-gate struct seg kpmseg; /* Segment used for physical mapping */ 2877c478bd9Sstevel@tonic-gate struct seg *segkpm = &kpmseg; /* 64bit kernel physical mapping segment */ 2887c478bd9Sstevel@tonic-gate #else 2897c478bd9Sstevel@tonic-gate struct seg *segkpm = NULL; /* Unused on IA32 */ 2907c478bd9Sstevel@tonic-gate #endif 2917c478bd9Sstevel@tonic-gate 2927c478bd9Sstevel@tonic-gate caddr_t segkp_base; /* Base address of segkp */ 293ad23a2dbSjohansen caddr_t segzio_base; /* Base address of segzio */ 2947c478bd9Sstevel@tonic-gate #if defined(__amd64) 2957c478bd9Sstevel@tonic-gate pgcnt_t segkpsize = btop(SEGKPDEFSIZE); /* size of segkp segment in pages */ 2967c478bd9Sstevel@tonic-gate #else 2977c478bd9Sstevel@tonic-gate pgcnt_t segkpsize = 0; 2987c478bd9Sstevel@tonic-gate #endif 299ad23a2dbSjohansen pgcnt_t segziosize = 0; /* size of zio segment in pages */ 3007c478bd9Sstevel@tonic-gate 301ae115bc7Smrj /* 302ae115bc7Smrj * VA range available to the debugger 303ae115bc7Smrj */ 304ae115bc7Smrj const caddr_t kdi_segdebugbase = (const caddr_t)SEGDEBUGBASE; 305ae115bc7Smrj const size_t kdi_segdebugsize = SEGDEBUGSIZE; 306ae115bc7Smrj 3077c478bd9Sstevel@tonic-gate struct memseg *memseg_base; 3087c478bd9Sstevel@tonic-gate struct vnode unused_pages_vp; 3097c478bd9Sstevel@tonic-gate 3107c478bd9Sstevel@tonic-gate #define FOURGB 0x100000000LL 3117c478bd9Sstevel@tonic-gate 3127c478bd9Sstevel@tonic-gate struct memlist *memlist; 3137c478bd9Sstevel@tonic-gate 3147c478bd9Sstevel@tonic-gate caddr_t s_text; /* start of kernel text segment */ 3157c478bd9Sstevel@tonic-gate caddr_t e_text; /* end of kernel text segment */ 3167c478bd9Sstevel@tonic-gate caddr_t s_data; /* start of kernel data segment */ 3177c478bd9Sstevel@tonic-gate caddr_t e_data; /* end of kernel data segment */ 3187c478bd9Sstevel@tonic-gate caddr_t modtext; /* start of loadable module text reserved */ 3197c478bd9Sstevel@tonic-gate caddr_t e_modtext; /* end of loadable module text reserved */ 3207c478bd9Sstevel@tonic-gate caddr_t moddata; /* start of loadable module data reserved */ 3217c478bd9Sstevel@tonic-gate caddr_t e_moddata; /* end of loadable module data reserved */ 3227c478bd9Sstevel@tonic-gate 3237c478bd9Sstevel@tonic-gate struct memlist *phys_install; /* Total installed physical memory */ 3247c478bd9Sstevel@tonic-gate struct memlist *phys_avail; /* Total available physical memory */ 3257c478bd9Sstevel@tonic-gate 3267c478bd9Sstevel@tonic-gate /* 3277c478bd9Sstevel@tonic-gate * kphysm_init returns the number of pages that were processed 3287c478bd9Sstevel@tonic-gate */ 329ae115bc7Smrj static pgcnt_t kphysm_init(page_t *, pgcnt_t); 3307c478bd9Sstevel@tonic-gate 3317c478bd9Sstevel@tonic-gate #define IO_PROP_SIZE 64 /* device property size */ 3327c478bd9Sstevel@tonic-gate 3337c478bd9Sstevel@tonic-gate /* 3347c478bd9Sstevel@tonic-gate * a couple useful roundup macros 3357c478bd9Sstevel@tonic-gate */ 3367c478bd9Sstevel@tonic-gate #define ROUND_UP_PAGE(x) \ 3377c478bd9Sstevel@tonic-gate ((uintptr_t)P2ROUNDUP((uintptr_t)(x), (uintptr_t)MMU_PAGESIZE)) 3387c478bd9Sstevel@tonic-gate #define ROUND_UP_LPAGE(x) \ 3397c478bd9Sstevel@tonic-gate ((uintptr_t)P2ROUNDUP((uintptr_t)(x), mmu.level_size[1])) 3407c478bd9Sstevel@tonic-gate #define ROUND_UP_4MEG(x) \ 341ae115bc7Smrj ((uintptr_t)P2ROUNDUP((uintptr_t)(x), (uintptr_t)FOUR_MEG)) 3427c478bd9Sstevel@tonic-gate #define ROUND_UP_TOPLEVEL(x) \ 3437c478bd9Sstevel@tonic-gate ((uintptr_t)P2ROUNDUP((uintptr_t)(x), mmu.level_size[mmu.max_level])) 3447c478bd9Sstevel@tonic-gate 3457c478bd9Sstevel@tonic-gate /* 3467c478bd9Sstevel@tonic-gate * 32-bit Kernel's Virtual memory layout. 3477c478bd9Sstevel@tonic-gate * +-----------------------+ 348ae115bc7Smrj * | | 3497c478bd9Sstevel@tonic-gate * 0xFFC00000 -|-----------------------|- ARGSBASE 3507c478bd9Sstevel@tonic-gate * | debugger | 3517c478bd9Sstevel@tonic-gate * 0xFF800000 -|-----------------------|- SEGDEBUGBASE 3527c478bd9Sstevel@tonic-gate * | Kernel Data | 3537c478bd9Sstevel@tonic-gate * 0xFEC00000 -|-----------------------| 3547c478bd9Sstevel@tonic-gate * | Kernel Text | 355843e1988Sjohnlev * 0xFE800000 -|-----------------------|- KERNEL_TEXT (0xFB400000 on Xen) 356ae115bc7Smrj * |--- GDT ---|- GDT page (GDT_VA) 357ae115bc7Smrj * |--- debug info ---|- debug info (DEBUG_INFO_VA) 358ae115bc7Smrj * | | 359ae115bc7Smrj * | page_t structures | 3607c478bd9Sstevel@tonic-gate * | memsegs, memlists, | 3617c478bd9Sstevel@tonic-gate * | page hash, etc. | 362ae115bc7Smrj * --- -|-----------------------|- ekernelheap, valloc_base (floating) 363ae115bc7Smrj * | | (segkp is just an arena in the heap) 3647c478bd9Sstevel@tonic-gate * | | 3657c478bd9Sstevel@tonic-gate * | kvseg | 3667c478bd9Sstevel@tonic-gate * | | 3677c478bd9Sstevel@tonic-gate * | | 3687c478bd9Sstevel@tonic-gate * --- -|-----------------------|- kernelheap (floating) 3697c478bd9Sstevel@tonic-gate * | Segkmap | 370ae115bc7Smrj * 0xC3002000 -|-----------------------|- segmap_start (floating) 3717c478bd9Sstevel@tonic-gate * | Red Zone | 3727c478bd9Sstevel@tonic-gate * 0xC3000000 -|-----------------------|- kernelbase / userlimit (floating) 3737c478bd9Sstevel@tonic-gate * | | || 3747c478bd9Sstevel@tonic-gate * | Shared objects | \/ 3757c478bd9Sstevel@tonic-gate * | | 3767c478bd9Sstevel@tonic-gate * : : 3777c478bd9Sstevel@tonic-gate * | user data | 3787c478bd9Sstevel@tonic-gate * |-----------------------| 3797c478bd9Sstevel@tonic-gate * | user text | 3807c478bd9Sstevel@tonic-gate * 0x08048000 -|-----------------------| 3817c478bd9Sstevel@tonic-gate * | user stack | 3827c478bd9Sstevel@tonic-gate * : : 3837c478bd9Sstevel@tonic-gate * | invalid | 3847c478bd9Sstevel@tonic-gate * 0x00000000 +-----------------------+ 3857c478bd9Sstevel@tonic-gate * 3867c478bd9Sstevel@tonic-gate * 3877c478bd9Sstevel@tonic-gate * 64-bit Kernel's Virtual memory layout. (assuming 64 bit app) 3887c478bd9Sstevel@tonic-gate * +-----------------------+ 389ae115bc7Smrj * | | 3907c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FFC00000 |-----------------------|- ARGSBASE 3917c478bd9Sstevel@tonic-gate * | debugger (?) | 3927c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FF800000 |-----------------------|- SEGDEBUGBASE 3937c478bd9Sstevel@tonic-gate * | unused | 3947c478bd9Sstevel@tonic-gate * +-----------------------+ 3957c478bd9Sstevel@tonic-gate * | Kernel Data | 3967c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FBC00000 |-----------------------| 3977c478bd9Sstevel@tonic-gate * | Kernel Text | 3987c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.FB800000 |-----------------------|- KERNEL_TEXT 399ae115bc7Smrj * |--- GDT ---|- GDT page (GDT_VA) 400ae115bc7Smrj * |--- debug info ---|- debug info (DEBUG_INFO_VA) 401ae115bc7Smrj * | | 4027c478bd9Sstevel@tonic-gate * | Core heap | (used for loadable modules) 4037c478bd9Sstevel@tonic-gate * 0xFFFFFFFF.C0000000 |-----------------------|- core_base / ekernelheap 4047c478bd9Sstevel@tonic-gate * | Kernel | 4057c478bd9Sstevel@tonic-gate * | heap | 4067c478bd9Sstevel@tonic-gate * 0xFFFFFXXX.XXX00000 |-----------------------|- kernelheap (floating) 407ae115bc7Smrj * | segmap | 408ae115bc7Smrj * 0xFFFFFXXX.XXX00000 |-----------------------|- segmap_start (floating) 4097c478bd9Sstevel@tonic-gate * | device mappings | 4107c478bd9Sstevel@tonic-gate * 0xFFFFFXXX.XXX00000 |-----------------------|- toxic_addr (floating) 411ad23a2dbSjohansen * | segzio | 412ad23a2dbSjohansen * 0xFFFFFXXX.XXX00000 |-----------------------|- segzio_base (floating) 4137c478bd9Sstevel@tonic-gate * | segkp | 414ae115bc7Smrj * --- |-----------------------|- segkp_base (floating) 415ae115bc7Smrj * | page_t structures | valloc_base + valloc_sz 416ae115bc7Smrj * | memsegs, memlists, | 417ae115bc7Smrj * | page hash, etc. | 41835b1ab99Sjosephb * 0xFFFFFF00.00000000 |-----------------------|- valloc_base (lower if > 1TB) 4197c478bd9Sstevel@tonic-gate * | segkpm | 4207c478bd9Sstevel@tonic-gate * 0xFFFFFE00.00000000 |-----------------------| 4217c478bd9Sstevel@tonic-gate * | Red Zone | 42235b1ab99Sjosephb * 0xFFFFFD80.00000000 |-----------------------|- KERNELBASE (lower if > 1TB) 4237c478bd9Sstevel@tonic-gate * | User stack |- User space memory 4247c478bd9Sstevel@tonic-gate * | | 4257c478bd9Sstevel@tonic-gate * | shared objects, etc | (grows downwards) 4267c478bd9Sstevel@tonic-gate * : : 4277c478bd9Sstevel@tonic-gate * | | 4287c478bd9Sstevel@tonic-gate * 0xFFFF8000.00000000 |-----------------------| 4297c478bd9Sstevel@tonic-gate * | | 4307c478bd9Sstevel@tonic-gate * | VA Hole / unused | 4317c478bd9Sstevel@tonic-gate * | | 4327c478bd9Sstevel@tonic-gate * 0x00008000.00000000 |-----------------------| 4337c478bd9Sstevel@tonic-gate * | | 4347c478bd9Sstevel@tonic-gate * | | 4357c478bd9Sstevel@tonic-gate * : : 4367c478bd9Sstevel@tonic-gate * | user heap | (grows upwards) 4377c478bd9Sstevel@tonic-gate * | | 4387c478bd9Sstevel@tonic-gate * | user data | 4397c478bd9Sstevel@tonic-gate * |-----------------------| 4407c478bd9Sstevel@tonic-gate * | user text | 4417c478bd9Sstevel@tonic-gate * 0x00000000.04000000 |-----------------------| 4427c478bd9Sstevel@tonic-gate * | invalid | 4437c478bd9Sstevel@tonic-gate * 0x00000000.00000000 +-----------------------+ 4447c478bd9Sstevel@tonic-gate * 4457c478bd9Sstevel@tonic-gate * A 32 bit app on the 64 bit kernel sees the same layout as on the 32 bit 4467c478bd9Sstevel@tonic-gate * kernel, except that userlimit is raised to 0xfe000000 4477c478bd9Sstevel@tonic-gate * 4487c478bd9Sstevel@tonic-gate * Floating values: 4497c478bd9Sstevel@tonic-gate * 4507c478bd9Sstevel@tonic-gate * valloc_base: start of the kernel's memory management/tracking data 451ae115bc7Smrj * structures. This region contains page_t structures for 452ae115bc7Smrj * physical memory, memsegs, memlists, and the page hash. 4537c478bd9Sstevel@tonic-gate * 4547c478bd9Sstevel@tonic-gate * core_base: start of the kernel's "core" heap area on 64-bit systems. 4557c478bd9Sstevel@tonic-gate * This area is intended to be used for global data as well as for module 4567c478bd9Sstevel@tonic-gate * text/data that does not fit into the nucleus pages. The core heap is 4577c478bd9Sstevel@tonic-gate * restricted to a 2GB range, allowing every address within it to be 4587c478bd9Sstevel@tonic-gate * accessed using rip-relative addressing 4597c478bd9Sstevel@tonic-gate * 4607c478bd9Sstevel@tonic-gate * ekernelheap: end of kernelheap and start of segmap. 4617c478bd9Sstevel@tonic-gate * 4627c478bd9Sstevel@tonic-gate * kernelheap: start of kernel heap. On 32-bit systems, this starts right 4637c478bd9Sstevel@tonic-gate * above a red zone that separates the user's address space from the 4647c478bd9Sstevel@tonic-gate * kernel's. On 64-bit systems, it sits above segkp and segkpm. 4657c478bd9Sstevel@tonic-gate * 466ae115bc7Smrj * segmap_start: start of segmap. The length of segmap can be modified 4677c478bd9Sstevel@tonic-gate * by changing segmapsize in /etc/system (preferred) or eeprom (deprecated). 4687c478bd9Sstevel@tonic-gate * The default length is 16MB on 32-bit systems and 64MB on 64-bit systems. 4697c478bd9Sstevel@tonic-gate * 4707c478bd9Sstevel@tonic-gate * kernelbase: On a 32-bit kernel the default value of 0xd4000000 will be 4717c478bd9Sstevel@tonic-gate * decreased by 2X the size required for page_t. This allows the kernel 4727c478bd9Sstevel@tonic-gate * heap to grow in size with physical memory. With sizeof(page_t) == 80 4737c478bd9Sstevel@tonic-gate * bytes, the following shows the values of kernelbase and kernel heap 4747c478bd9Sstevel@tonic-gate * sizes for different memory configurations (assuming default segmap and 4757c478bd9Sstevel@tonic-gate * segkp sizes). 4767c478bd9Sstevel@tonic-gate * 4777c478bd9Sstevel@tonic-gate * mem size for kernelbase kernel heap 4787c478bd9Sstevel@tonic-gate * size page_t's size 4797c478bd9Sstevel@tonic-gate * ---- --------- ---------- ----------- 4807c478bd9Sstevel@tonic-gate * 1gb 0x01400000 0xd1800000 684MB 4817c478bd9Sstevel@tonic-gate * 2gb 0x02800000 0xcf000000 704MB 4827c478bd9Sstevel@tonic-gate * 4gb 0x05000000 0xca000000 744MB 4837c478bd9Sstevel@tonic-gate * 6gb 0x07800000 0xc5000000 784MB 4847c478bd9Sstevel@tonic-gate * 8gb 0x0a000000 0xc0000000 824MB 4857c478bd9Sstevel@tonic-gate * 16gb 0x14000000 0xac000000 984MB 4867c478bd9Sstevel@tonic-gate * 32gb 0x28000000 0x84000000 1304MB 4877c478bd9Sstevel@tonic-gate * 64gb 0x50000000 0x34000000 1944MB (*) 4887c478bd9Sstevel@tonic-gate * 4897c478bd9Sstevel@tonic-gate * kernelbase is less than the abi minimum of 0xc0000000 for memory 4907c478bd9Sstevel@tonic-gate * configurations above 8gb. 4917c478bd9Sstevel@tonic-gate * 4927c478bd9Sstevel@tonic-gate * (*) support for memory configurations above 32gb will require manual tuning 4937c478bd9Sstevel@tonic-gate * of kernelbase to balance out the need of user applications. 4947c478bd9Sstevel@tonic-gate */ 4957c478bd9Sstevel@tonic-gate 4967c478bd9Sstevel@tonic-gate /* real-time-clock initialization parameters */ 497ae115bc7Smrj extern time_t process_rtc_config_file(void); 4987c478bd9Sstevel@tonic-gate 4997c478bd9Sstevel@tonic-gate uintptr_t kernelbase; 500ae115bc7Smrj uintptr_t postbootkernelbase; /* not set till boot loader is gone */ 5017c478bd9Sstevel@tonic-gate uintptr_t eprom_kernelbase; 5027c478bd9Sstevel@tonic-gate size_t segmapsize; 503ae115bc7Smrj uintptr_t segmap_start; 5047c478bd9Sstevel@tonic-gate int segmapfreelists; 5057c478bd9Sstevel@tonic-gate pgcnt_t npages; 506ae115bc7Smrj pgcnt_t orig_npages; 5077c478bd9Sstevel@tonic-gate size_t core_size; /* size of "core" heap */ 5087c478bd9Sstevel@tonic-gate uintptr_t core_base; /* base address of "core" heap */ 5097c478bd9Sstevel@tonic-gate 5107c478bd9Sstevel@tonic-gate /* 5117c478bd9Sstevel@tonic-gate * List of bootstrap pages. We mark these as allocated in startup. 5127c478bd9Sstevel@tonic-gate * release_bootstrap() will free them when we're completely done with 5137c478bd9Sstevel@tonic-gate * the bootstrap. 5147c478bd9Sstevel@tonic-gate */ 515ae115bc7Smrj static page_t *bootpages; 516ae115bc7Smrj 517ae115bc7Smrj /* 518ae115bc7Smrj * boot time pages that have a vnode from the ramdisk will keep that forever. 519ae115bc7Smrj */ 520ae115bc7Smrj static page_t *rd_pages; 5217c478bd9Sstevel@tonic-gate 5227c478bd9Sstevel@tonic-gate struct system_hardware system_hardware; 5237c478bd9Sstevel@tonic-gate 524551bc2a6Smrj /* 525551bc2a6Smrj * Is this Solaris instance running in a fully virtualized xVM domain? 526551bc2a6Smrj */ 527551bc2a6Smrj int xpv_is_hvm = 0; 528551bc2a6Smrj 5297c478bd9Sstevel@tonic-gate /* 5307c478bd9Sstevel@tonic-gate * Enable some debugging messages concerning memory usage... 5317c478bd9Sstevel@tonic-gate */ 5327c478bd9Sstevel@tonic-gate static void 533ae115bc7Smrj print_memlist(char *title, struct memlist *mp) 5347c478bd9Sstevel@tonic-gate { 5357c478bd9Sstevel@tonic-gate prom_printf("MEMLIST: %s:\n", title); 5367c478bd9Sstevel@tonic-gate while (mp != NULL) { 5377c478bd9Sstevel@tonic-gate prom_printf("\tAddress 0x%" PRIx64 ", size 0x%" PRIx64 "\n", 5387c478bd9Sstevel@tonic-gate mp->address, mp->size); 5397c478bd9Sstevel@tonic-gate mp = mp->next; 5407c478bd9Sstevel@tonic-gate } 5417c478bd9Sstevel@tonic-gate } 5427c478bd9Sstevel@tonic-gate 5437c478bd9Sstevel@tonic-gate /* 5447c478bd9Sstevel@tonic-gate * XX64 need a comment here.. are these just default values, surely 5457c478bd9Sstevel@tonic-gate * we read the "cpuid" type information to figure this out. 5467c478bd9Sstevel@tonic-gate */ 5477c478bd9Sstevel@tonic-gate int l2cache_sz = 0x80000; 5487c478bd9Sstevel@tonic-gate int l2cache_linesz = 0x40; 5497c478bd9Sstevel@tonic-gate int l2cache_assoc = 1; 5507c478bd9Sstevel@tonic-gate 5512cb27123Saguzovsk static size_t textrepl_min_gb = 10; 5522cb27123Saguzovsk 5537c478bd9Sstevel@tonic-gate /* 5547c478bd9Sstevel@tonic-gate * on 64 bit we use a predifined VA range for mapping devices in the kernel 5557c478bd9Sstevel@tonic-gate * on 32 bit the mappings are intermixed in the heap, so we use a bit map 5567c478bd9Sstevel@tonic-gate */ 5577c478bd9Sstevel@tonic-gate #ifdef __amd64 5587c478bd9Sstevel@tonic-gate 5597c478bd9Sstevel@tonic-gate vmem_t *device_arena; 5607c478bd9Sstevel@tonic-gate uintptr_t toxic_addr = (uintptr_t)NULL; 561ae115bc7Smrj size_t toxic_size = 1024 * 1024 * 1024; /* Sparc uses 1 gig too */ 5627c478bd9Sstevel@tonic-gate 5637c478bd9Sstevel@tonic-gate #else /* __i386 */ 5647c478bd9Sstevel@tonic-gate 5657c478bd9Sstevel@tonic-gate ulong_t *toxic_bit_map; /* one bit for each 4k of VA in heap_arena */ 5667c478bd9Sstevel@tonic-gate size_t toxic_bit_map_len = 0; /* in bits */ 5677c478bd9Sstevel@tonic-gate 5687c478bd9Sstevel@tonic-gate #endif /* __i386 */ 5697c478bd9Sstevel@tonic-gate 5707c478bd9Sstevel@tonic-gate /* 5717c478bd9Sstevel@tonic-gate * Simple boot time debug facilities 5727c478bd9Sstevel@tonic-gate */ 5737c478bd9Sstevel@tonic-gate static char *prm_dbg_str[] = { 5747c478bd9Sstevel@tonic-gate "%s:%d: '%s' is 0x%x\n", 5757c478bd9Sstevel@tonic-gate "%s:%d: '%s' is 0x%llx\n" 5767c478bd9Sstevel@tonic-gate }; 5777c478bd9Sstevel@tonic-gate 5787c478bd9Sstevel@tonic-gate int prom_debug; 5797c478bd9Sstevel@tonic-gate 5807c478bd9Sstevel@tonic-gate #define PRM_DEBUG(q) if (prom_debug) \ 5817c478bd9Sstevel@tonic-gate prom_printf(prm_dbg_str[sizeof (q) >> 3], "startup.c", __LINE__, #q, q); 5827c478bd9Sstevel@tonic-gate #define PRM_POINT(q) if (prom_debug) \ 5837c478bd9Sstevel@tonic-gate prom_printf("%s:%d: %s\n", "startup.c", __LINE__, q); 5847c478bd9Sstevel@tonic-gate 5857c478bd9Sstevel@tonic-gate /* 5867c478bd9Sstevel@tonic-gate * This structure is used to keep track of the intial allocations 5877c478bd9Sstevel@tonic-gate * done in startup_memlist(). The value of NUM_ALLOCATIONS needs to 5887c478bd9Sstevel@tonic-gate * be >= the number of ADD_TO_ALLOCATIONS() executed in the code. 5897c478bd9Sstevel@tonic-gate */ 5907c478bd9Sstevel@tonic-gate #define NUM_ALLOCATIONS 7 5917c478bd9Sstevel@tonic-gate int num_allocations = 0; 5927c478bd9Sstevel@tonic-gate struct { 5937c478bd9Sstevel@tonic-gate void **al_ptr; 5947c478bd9Sstevel@tonic-gate size_t al_size; 5957c478bd9Sstevel@tonic-gate } allocations[NUM_ALLOCATIONS]; 5967c478bd9Sstevel@tonic-gate size_t valloc_sz = 0; 5977c478bd9Sstevel@tonic-gate uintptr_t valloc_base; 5987c478bd9Sstevel@tonic-gate 5997c478bd9Sstevel@tonic-gate #define ADD_TO_ALLOCATIONS(ptr, size) { \ 6007c478bd9Sstevel@tonic-gate size = ROUND_UP_PAGE(size); \ 6017c478bd9Sstevel@tonic-gate if (num_allocations == NUM_ALLOCATIONS) \ 6027c478bd9Sstevel@tonic-gate panic("too many ADD_TO_ALLOCATIONS()"); \ 6037c478bd9Sstevel@tonic-gate allocations[num_allocations].al_ptr = (void**)&ptr; \ 6047c478bd9Sstevel@tonic-gate allocations[num_allocations].al_size = size; \ 6057c478bd9Sstevel@tonic-gate valloc_sz += size; \ 6067c478bd9Sstevel@tonic-gate ++num_allocations; \ 6077c478bd9Sstevel@tonic-gate } 6087c478bd9Sstevel@tonic-gate 609ae115bc7Smrj /* 610ae115bc7Smrj * Allocate all the initial memory needed by the page allocator. 611ae115bc7Smrj */ 6127c478bd9Sstevel@tonic-gate static void 6137c478bd9Sstevel@tonic-gate perform_allocations(void) 6147c478bd9Sstevel@tonic-gate { 6157c478bd9Sstevel@tonic-gate caddr_t mem; 6167c478bd9Sstevel@tonic-gate int i; 617ae115bc7Smrj int valloc_align; 6187c478bd9Sstevel@tonic-gate 619ae115bc7Smrj PRM_DEBUG(valloc_base); 620ae115bc7Smrj PRM_DEBUG(valloc_sz); 621ae115bc7Smrj valloc_align = mmu.level_size[mmu.max_page_level > 0]; 622ae115bc7Smrj mem = BOP_ALLOC(bootops, (caddr_t)valloc_base, valloc_sz, valloc_align); 6237c478bd9Sstevel@tonic-gate if (mem != (caddr_t)valloc_base) 6247c478bd9Sstevel@tonic-gate panic("BOP_ALLOC() failed"); 6257c478bd9Sstevel@tonic-gate bzero(mem, valloc_sz); 6267c478bd9Sstevel@tonic-gate for (i = 0; i < num_allocations; ++i) { 6277c478bd9Sstevel@tonic-gate *allocations[i].al_ptr = (void *)mem; 6287c478bd9Sstevel@tonic-gate mem += allocations[i].al_size; 6297c478bd9Sstevel@tonic-gate } 6307c478bd9Sstevel@tonic-gate } 6317c478bd9Sstevel@tonic-gate 6327c478bd9Sstevel@tonic-gate /* 6337c478bd9Sstevel@tonic-gate * Our world looks like this at startup time. 6347c478bd9Sstevel@tonic-gate * 6357c478bd9Sstevel@tonic-gate * In a 32-bit OS, boot loads the kernel text at 0xfe800000 and kernel data 6367c478bd9Sstevel@tonic-gate * at 0xfec00000. On a 64-bit OS, kernel text and data are loaded at 6377c478bd9Sstevel@tonic-gate * 0xffffffff.fe800000 and 0xffffffff.fec00000 respectively. Those 6387c478bd9Sstevel@tonic-gate * addresses are fixed in the binary at link time. 6397c478bd9Sstevel@tonic-gate * 6407c478bd9Sstevel@tonic-gate * On the text page: 6417c478bd9Sstevel@tonic-gate * unix/genunix/krtld/module text loads. 6427c478bd9Sstevel@tonic-gate * 6437c478bd9Sstevel@tonic-gate * On the data page: 644ae115bc7Smrj * unix/genunix/krtld/module data loads. 645ae115bc7Smrj * 6467c478bd9Sstevel@tonic-gate * Machine-dependent startup code 6477c478bd9Sstevel@tonic-gate */ 6487c478bd9Sstevel@tonic-gate void 6497c478bd9Sstevel@tonic-gate startup(void) 6507c478bd9Sstevel@tonic-gate { 651843e1988Sjohnlev #if !defined(__xpv) 65275bcd456Sjg extern void startup_bios_disk(void); 65375bcd456Sjg extern void startup_pci_bios(void); 654843e1988Sjohnlev #endif 655*a563a037Sbholler extern cpuset_t cpu_ready_set; 656*a563a037Sbholler 657843e1988Sjohnlev /* 658843e1988Sjohnlev * Make sure that nobody tries to use sekpm until we have 659843e1988Sjohnlev * initialized it properly. 660843e1988Sjohnlev */ 6617c478bd9Sstevel@tonic-gate #if defined(__amd64) 66235b1ab99Sjosephb kpm_desired = 1; 6637c478bd9Sstevel@tonic-gate #endif 6647c478bd9Sstevel@tonic-gate kpm_enable = 0; 665*a563a037Sbholler CPUSET_ONLY(cpu_ready_set, 0); /* cpu 0 is boot cpu */ 6667c478bd9Sstevel@tonic-gate 667843e1988Sjohnlev #if defined(__xpv) /* XXPV fix me! */ 668843e1988Sjohnlev { 669843e1988Sjohnlev extern int segvn_use_regions; 670843e1988Sjohnlev segvn_use_regions = 0; 671843e1988Sjohnlev } 672843e1988Sjohnlev #endif 6737c478bd9Sstevel@tonic-gate progressbar_init(); 6747c478bd9Sstevel@tonic-gate startup_init(); 675ab4a9bebSjohnlev #if defined(__xpv) 676ab4a9bebSjohnlev startup_xen_version(); 677ab4a9bebSjohnlev #endif 6787c478bd9Sstevel@tonic-gate startup_memlist(); 679ae115bc7Smrj startup_kmem(); 68035b1ab99Sjosephb startup_vm(); 681843e1988Sjohnlev #if !defined(__xpv) 68275bcd456Sjg startup_pci_bios(); 683843e1988Sjohnlev #endif 6847c478bd9Sstevel@tonic-gate startup_modules(); 685843e1988Sjohnlev #if !defined(__xpv) 6867c478bd9Sstevel@tonic-gate startup_bios_disk(); 687843e1988Sjohnlev #endif 6887c478bd9Sstevel@tonic-gate startup_end(); 6897c478bd9Sstevel@tonic-gate progressbar_start(); 6907c478bd9Sstevel@tonic-gate } 6917c478bd9Sstevel@tonic-gate 6927c478bd9Sstevel@tonic-gate static void 6937c478bd9Sstevel@tonic-gate startup_init() 6947c478bd9Sstevel@tonic-gate { 6957c478bd9Sstevel@tonic-gate PRM_POINT("startup_init() starting..."); 6967c478bd9Sstevel@tonic-gate 6977c478bd9Sstevel@tonic-gate /* 6987c478bd9Sstevel@tonic-gate * Complete the extraction of cpuid data 6997c478bd9Sstevel@tonic-gate */ 7007c478bd9Sstevel@tonic-gate cpuid_pass2(CPU); 7017c478bd9Sstevel@tonic-gate 7027c478bd9Sstevel@tonic-gate (void) check_boot_version(BOP_GETVERSION(bootops)); 7037c478bd9Sstevel@tonic-gate 7047c478bd9Sstevel@tonic-gate /* 7057c478bd9Sstevel@tonic-gate * Check for prom_debug in boot environment 7067c478bd9Sstevel@tonic-gate */ 7077c478bd9Sstevel@tonic-gate if (BOP_GETPROPLEN(bootops, "prom_debug") >= 0) { 7087c478bd9Sstevel@tonic-gate ++prom_debug; 7097c478bd9Sstevel@tonic-gate PRM_POINT("prom_debug found in boot enviroment"); 7107c478bd9Sstevel@tonic-gate } 7117c478bd9Sstevel@tonic-gate 7127c478bd9Sstevel@tonic-gate /* 7137c478bd9Sstevel@tonic-gate * Collect node, cpu and memory configuration information. 7147c478bd9Sstevel@tonic-gate */ 7157c478bd9Sstevel@tonic-gate get_system_configuration(); 7167c478bd9Sstevel@tonic-gate 7177c478bd9Sstevel@tonic-gate /* 7187c478bd9Sstevel@tonic-gate * Halt if this is an unsupported processor. 7197c478bd9Sstevel@tonic-gate */ 7207c478bd9Sstevel@tonic-gate if (x86_type == X86_TYPE_486 || x86_type == X86_TYPE_CYRIX_486) { 7217c478bd9Sstevel@tonic-gate printf("\n486 processor (\"%s\") detected.\n", 7227c478bd9Sstevel@tonic-gate CPU->cpu_brandstr); 7237c478bd9Sstevel@tonic-gate halt("This processor is not supported by this release " 7247c478bd9Sstevel@tonic-gate "of Solaris."); 7257c478bd9Sstevel@tonic-gate } 7267c478bd9Sstevel@tonic-gate 7277c478bd9Sstevel@tonic-gate PRM_POINT("startup_init() done"); 7287c478bd9Sstevel@tonic-gate } 7297c478bd9Sstevel@tonic-gate 7307c478bd9Sstevel@tonic-gate /* 7317c478bd9Sstevel@tonic-gate * Callback for copy_memlist_filter() to filter nucleus, kadb/kmdb, (ie. 7327c478bd9Sstevel@tonic-gate * everything mapped above KERNEL_TEXT) pages from phys_avail. Note it 7337c478bd9Sstevel@tonic-gate * also filters out physical page zero. There is some reliance on the 7347c478bd9Sstevel@tonic-gate * boot loader allocating only a few contiguous physical memory chunks. 7357c478bd9Sstevel@tonic-gate */ 7367c478bd9Sstevel@tonic-gate static void 7377c478bd9Sstevel@tonic-gate avail_filter(uint64_t *addr, uint64_t *size) 7387c478bd9Sstevel@tonic-gate { 7397c478bd9Sstevel@tonic-gate uintptr_t va; 7407c478bd9Sstevel@tonic-gate uintptr_t next_va; 7417c478bd9Sstevel@tonic-gate pfn_t pfn; 7427c478bd9Sstevel@tonic-gate uint64_t pfn_addr; 7437c478bd9Sstevel@tonic-gate uint64_t pfn_eaddr; 7447c478bd9Sstevel@tonic-gate uint_t prot; 7457c478bd9Sstevel@tonic-gate size_t len; 7467c478bd9Sstevel@tonic-gate uint_t change; 7477c478bd9Sstevel@tonic-gate 7487c478bd9Sstevel@tonic-gate if (prom_debug) 7497c478bd9Sstevel@tonic-gate prom_printf("\tFilter: in: a=%" PRIx64 ", s=%" PRIx64 "\n", 7507c478bd9Sstevel@tonic-gate *addr, *size); 7517c478bd9Sstevel@tonic-gate 7527c478bd9Sstevel@tonic-gate /* 7537c478bd9Sstevel@tonic-gate * page zero is required for BIOS.. never make it available 7547c478bd9Sstevel@tonic-gate */ 7557c478bd9Sstevel@tonic-gate if (*addr == 0) { 7567c478bd9Sstevel@tonic-gate *addr += MMU_PAGESIZE; 7577c478bd9Sstevel@tonic-gate *size -= MMU_PAGESIZE; 7587c478bd9Sstevel@tonic-gate } 7597c478bd9Sstevel@tonic-gate 7607c478bd9Sstevel@tonic-gate /* 761ae115bc7Smrj * First we trim from the front of the range. Since kbm_probe() 7627c478bd9Sstevel@tonic-gate * walks ranges in virtual order, but addr/size are physical, we need 7637c478bd9Sstevel@tonic-gate * to the list until no changes are seen. This deals with the case 7647c478bd9Sstevel@tonic-gate * where page "p" is mapped at v, page "p + PAGESIZE" is mapped at w 7657c478bd9Sstevel@tonic-gate * but w < v. 7667c478bd9Sstevel@tonic-gate */ 7677c478bd9Sstevel@tonic-gate do { 7687c478bd9Sstevel@tonic-gate change = 0; 7697c478bd9Sstevel@tonic-gate for (va = KERNEL_TEXT; 770ae115bc7Smrj *size > 0 && kbm_probe(&va, &len, &pfn, &prot) != 0; 7717c478bd9Sstevel@tonic-gate va = next_va) { 7727c478bd9Sstevel@tonic-gate 7737c478bd9Sstevel@tonic-gate next_va = va + len; 774ae115bc7Smrj pfn_addr = pfn_to_pa(pfn); 7757c478bd9Sstevel@tonic-gate pfn_eaddr = pfn_addr + len; 7767c478bd9Sstevel@tonic-gate 7777c478bd9Sstevel@tonic-gate if (pfn_addr <= *addr && pfn_eaddr > *addr) { 7787c478bd9Sstevel@tonic-gate change = 1; 7797c478bd9Sstevel@tonic-gate while (*size > 0 && len > 0) { 7807c478bd9Sstevel@tonic-gate *addr += MMU_PAGESIZE; 7817c478bd9Sstevel@tonic-gate *size -= MMU_PAGESIZE; 7827c478bd9Sstevel@tonic-gate len -= MMU_PAGESIZE; 7837c478bd9Sstevel@tonic-gate } 7847c478bd9Sstevel@tonic-gate } 7857c478bd9Sstevel@tonic-gate } 7867c478bd9Sstevel@tonic-gate if (change && prom_debug) 7877c478bd9Sstevel@tonic-gate prom_printf("\t\ttrim: a=%" PRIx64 ", s=%" PRIx64 "\n", 7887c478bd9Sstevel@tonic-gate *addr, *size); 7897c478bd9Sstevel@tonic-gate } while (change); 7907c478bd9Sstevel@tonic-gate 7917c478bd9Sstevel@tonic-gate /* 7927c478bd9Sstevel@tonic-gate * Trim pages from the end of the range. 7937c478bd9Sstevel@tonic-gate */ 7947c478bd9Sstevel@tonic-gate for (va = KERNEL_TEXT; 795ae115bc7Smrj *size > 0 && kbm_probe(&va, &len, &pfn, &prot) != 0; 7967c478bd9Sstevel@tonic-gate va = next_va) { 7977c478bd9Sstevel@tonic-gate 7987c478bd9Sstevel@tonic-gate next_va = va + len; 799ae115bc7Smrj pfn_addr = pfn_to_pa(pfn); 8007c478bd9Sstevel@tonic-gate 8017c478bd9Sstevel@tonic-gate if (pfn_addr >= *addr && pfn_addr < *addr + *size) 8027c478bd9Sstevel@tonic-gate *size = pfn_addr - *addr; 8037c478bd9Sstevel@tonic-gate } 8047c478bd9Sstevel@tonic-gate 8057c478bd9Sstevel@tonic-gate if (prom_debug) 8067c478bd9Sstevel@tonic-gate prom_printf("\tFilter out: a=%" PRIx64 ", s=%" PRIx64 "\n", 8077c478bd9Sstevel@tonic-gate *addr, *size); 8087c478bd9Sstevel@tonic-gate } 8097c478bd9Sstevel@tonic-gate 8107c478bd9Sstevel@tonic-gate static void 8117c478bd9Sstevel@tonic-gate kpm_init() 8127c478bd9Sstevel@tonic-gate { 8137c478bd9Sstevel@tonic-gate struct segkpm_crargs b; 8147c478bd9Sstevel@tonic-gate 8157c478bd9Sstevel@tonic-gate /* 8167c478bd9Sstevel@tonic-gate * These variables were all designed for sfmmu in which segkpm is 8177c478bd9Sstevel@tonic-gate * mapped using a single pagesize - either 8KB or 4MB. On x86, we 8187c478bd9Sstevel@tonic-gate * might use 2+ page sizes on a single machine, so none of these 8197c478bd9Sstevel@tonic-gate * variables have a single correct value. They are set up as if we 8207c478bd9Sstevel@tonic-gate * always use a 4KB pagesize, which should do no harm. In the long 8217c478bd9Sstevel@tonic-gate * run, we should get rid of KPM's assumption that only a single 8227c478bd9Sstevel@tonic-gate * pagesize is used. 8237c478bd9Sstevel@tonic-gate */ 8247c478bd9Sstevel@tonic-gate kpm_pgshft = MMU_PAGESHIFT; 8257c478bd9Sstevel@tonic-gate kpm_pgsz = MMU_PAGESIZE; 8267c478bd9Sstevel@tonic-gate kpm_pgoff = MMU_PAGEOFFSET; 8277c478bd9Sstevel@tonic-gate kpmp2pshft = 0; 8287c478bd9Sstevel@tonic-gate kpmpnpgs = 1; 8297c478bd9Sstevel@tonic-gate ASSERT(((uintptr_t)kpm_vbase & (kpm_pgsz - 1)) == 0); 8307c478bd9Sstevel@tonic-gate 8317c478bd9Sstevel@tonic-gate PRM_POINT("about to create segkpm"); 8327c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 8337c478bd9Sstevel@tonic-gate 8347c478bd9Sstevel@tonic-gate if (seg_attach(&kas, kpm_vbase, kpm_size, segkpm) < 0) 8357c478bd9Sstevel@tonic-gate panic("cannot attach segkpm"); 8367c478bd9Sstevel@tonic-gate 8377c478bd9Sstevel@tonic-gate b.prot = PROT_READ | PROT_WRITE; 8387c478bd9Sstevel@tonic-gate b.nvcolors = 1; 8397c478bd9Sstevel@tonic-gate 8407c478bd9Sstevel@tonic-gate if (segkpm_create(segkpm, (caddr_t)&b) != 0) 8417c478bd9Sstevel@tonic-gate panic("segkpm_create segkpm"); 8427c478bd9Sstevel@tonic-gate 8437c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 844ae115bc7Smrj } 8457c478bd9Sstevel@tonic-gate 846ae115bc7Smrj /* 847ae115bc7Smrj * The debug info page provides enough information to allow external 848ae115bc7Smrj * inspectors (e.g. when running under a hypervisor) to bootstrap 849ae115bc7Smrj * themselves into allowing full-blown kernel debugging. 850ae115bc7Smrj */ 851ae115bc7Smrj static void 852ae115bc7Smrj init_debug_info(void) 853ae115bc7Smrj { 854ae115bc7Smrj caddr_t mem; 855ae115bc7Smrj debug_info_t *di; 856ae115bc7Smrj 857ae115bc7Smrj #ifndef __lint 858ae115bc7Smrj ASSERT(sizeof (debug_info_t) < MMU_PAGESIZE); 859ae115bc7Smrj #endif 860ae115bc7Smrj 861ae115bc7Smrj mem = BOP_ALLOC(bootops, (caddr_t)DEBUG_INFO_VA, MMU_PAGESIZE, 862ae115bc7Smrj MMU_PAGESIZE); 863ae115bc7Smrj 864ae115bc7Smrj if (mem != (caddr_t)DEBUG_INFO_VA) 865ae115bc7Smrj panic("BOP_ALLOC() failed"); 866ae115bc7Smrj bzero(mem, MMU_PAGESIZE); 867ae115bc7Smrj 868ae115bc7Smrj di = (debug_info_t *)mem; 869ae115bc7Smrj 870ae115bc7Smrj di->di_magic = DEBUG_INFO_MAGIC; 871ae115bc7Smrj di->di_version = DEBUG_INFO_VERSION; 872843e1988Sjohnlev di->di_modules = (uintptr_t)&modules; 873843e1988Sjohnlev di->di_s_text = (uintptr_t)s_text; 874843e1988Sjohnlev di->di_e_text = (uintptr_t)e_text; 875843e1988Sjohnlev di->di_s_data = (uintptr_t)s_data; 876843e1988Sjohnlev di->di_e_data = (uintptr_t)e_data; 877843e1988Sjohnlev di->di_hat_htable_off = offsetof(hat_t, hat_htable); 878843e1988Sjohnlev di->di_ht_pfn_off = offsetof(htable_t, ht_pfn); 8797c478bd9Sstevel@tonic-gate } 8807c478bd9Sstevel@tonic-gate 8817c478bd9Sstevel@tonic-gate /* 882ae115bc7Smrj * Build the memlists and other kernel essential memory system data structures. 883ae115bc7Smrj * This is everything at valloc_base. 8847c478bd9Sstevel@tonic-gate */ 8857c478bd9Sstevel@tonic-gate static void 8867c478bd9Sstevel@tonic-gate startup_memlist(void) 8877c478bd9Sstevel@tonic-gate { 8887c478bd9Sstevel@tonic-gate size_t memlist_sz; 8897c478bd9Sstevel@tonic-gate size_t memseg_sz; 8907c478bd9Sstevel@tonic-gate size_t pagehash_sz; 8917c478bd9Sstevel@tonic-gate size_t pp_sz; 8927c478bd9Sstevel@tonic-gate uintptr_t va; 8937c478bd9Sstevel@tonic-gate size_t len; 8947c478bd9Sstevel@tonic-gate uint_t prot; 8957c478bd9Sstevel@tonic-gate pfn_t pfn; 8967c478bd9Sstevel@tonic-gate int memblocks; 8977c478bd9Sstevel@tonic-gate caddr_t pagecolor_mem; 8987c478bd9Sstevel@tonic-gate size_t pagecolor_memsz; 8997c478bd9Sstevel@tonic-gate caddr_t page_ctrs_mem; 9007c478bd9Sstevel@tonic-gate size_t page_ctrs_size; 901d7d93655Sblakej size_t pse_table_alloc_size; 9027c478bd9Sstevel@tonic-gate struct memlist *current; 9037c478bd9Sstevel@tonic-gate extern void startup_build_mem_nodes(struct memlist *); 9047c478bd9Sstevel@tonic-gate 9057c478bd9Sstevel@tonic-gate /* XX64 fix these - they should be in include files */ 9067c478bd9Sstevel@tonic-gate extern size_t page_coloring_init(uint_t, int, int); 9077c478bd9Sstevel@tonic-gate extern void page_coloring_setup(caddr_t); 9087c478bd9Sstevel@tonic-gate 9097c478bd9Sstevel@tonic-gate PRM_POINT("startup_memlist() starting..."); 9107c478bd9Sstevel@tonic-gate 9117c478bd9Sstevel@tonic-gate /* 9127c478bd9Sstevel@tonic-gate * Use leftover large page nucleus text/data space for loadable modules. 9137c478bd9Sstevel@tonic-gate * Use at most MODTEXT/MODDATA. 9147c478bd9Sstevel@tonic-gate */ 915ae115bc7Smrj len = kbm_nucleus_size; 916ae115bc7Smrj ASSERT(len > MMU_PAGESIZE); 9177c478bd9Sstevel@tonic-gate 918ae115bc7Smrj moddata = (caddr_t)ROUND_UP_PAGE(e_data); 919ae115bc7Smrj e_moddata = (caddr_t)P2ROUNDUP((uintptr_t)e_data, (uintptr_t)len); 920ae115bc7Smrj if (e_moddata - moddata > MODDATA) 921ae115bc7Smrj e_moddata = moddata + MODDATA; 9227c478bd9Sstevel@tonic-gate 923ae115bc7Smrj modtext = (caddr_t)ROUND_UP_PAGE(e_text); 924ae115bc7Smrj e_modtext = (caddr_t)P2ROUNDUP((uintptr_t)e_text, (uintptr_t)len); 925ae115bc7Smrj if (e_modtext - modtext > MODTEXT) 926ae115bc7Smrj e_modtext = modtext + MODTEXT; 9277c478bd9Sstevel@tonic-gate 9287c478bd9Sstevel@tonic-gate econtig = e_moddata; 9297c478bd9Sstevel@tonic-gate 9307c478bd9Sstevel@tonic-gate PRM_DEBUG(modtext); 9317c478bd9Sstevel@tonic-gate PRM_DEBUG(e_modtext); 9327c478bd9Sstevel@tonic-gate PRM_DEBUG(moddata); 9337c478bd9Sstevel@tonic-gate PRM_DEBUG(e_moddata); 9347c478bd9Sstevel@tonic-gate PRM_DEBUG(econtig); 9357c478bd9Sstevel@tonic-gate 9367c478bd9Sstevel@tonic-gate /* 937ae115bc7Smrj * Examine the boot loader physical memory map to find out: 9387c478bd9Sstevel@tonic-gate * - total memory in system - physinstalled 9397c478bd9Sstevel@tonic-gate * - the max physical address - physmax 940ae115bc7Smrj * - the number of discontiguous segments of memory. 9417c478bd9Sstevel@tonic-gate */ 9427c478bd9Sstevel@tonic-gate if (prom_debug) 943ae115bc7Smrj print_memlist("boot physinstalled", 9447c478bd9Sstevel@tonic-gate bootops->boot_mem->physinstalled); 9457c478bd9Sstevel@tonic-gate installed_top_size(bootops->boot_mem->physinstalled, &physmax, 9467c478bd9Sstevel@tonic-gate &physinstalled, &memblocks); 9477c478bd9Sstevel@tonic-gate PRM_DEBUG(physmax); 9487c478bd9Sstevel@tonic-gate PRM_DEBUG(physinstalled); 9497c478bd9Sstevel@tonic-gate PRM_DEBUG(memblocks); 9507c478bd9Sstevel@tonic-gate 9517c478bd9Sstevel@tonic-gate /* 9527c478bd9Sstevel@tonic-gate * Initialize hat's mmu parameters. 9537c478bd9Sstevel@tonic-gate * Check for enforce-prot-exec in boot environment. It's used to 9547c478bd9Sstevel@tonic-gate * enable/disable support for the page table entry NX bit. 9557c478bd9Sstevel@tonic-gate * The default is to enforce PROT_EXEC on processors that support NX. 9567c478bd9Sstevel@tonic-gate * Boot seems to round up the "len", but 8 seems to be big enough. 9577c478bd9Sstevel@tonic-gate */ 9587c478bd9Sstevel@tonic-gate mmu_init(); 9597c478bd9Sstevel@tonic-gate 9607c478bd9Sstevel@tonic-gate #ifdef __i386 9617c478bd9Sstevel@tonic-gate /* 9627c478bd9Sstevel@tonic-gate * physmax is lowered if there is more memory than can be 9637c478bd9Sstevel@tonic-gate * physically addressed in 32 bit (PAE/non-PAE) modes. 9647c478bd9Sstevel@tonic-gate */ 9657c478bd9Sstevel@tonic-gate if (mmu.pae_hat) { 9667c478bd9Sstevel@tonic-gate if (PFN_ABOVE64G(physmax)) { 9677c478bd9Sstevel@tonic-gate physinstalled -= (physmax - (PFN_64G - 1)); 9687c478bd9Sstevel@tonic-gate physmax = PFN_64G - 1; 9697c478bd9Sstevel@tonic-gate } 9707c478bd9Sstevel@tonic-gate } else { 9717c478bd9Sstevel@tonic-gate if (PFN_ABOVE4G(physmax)) { 9727c478bd9Sstevel@tonic-gate physinstalled -= (physmax - (PFN_4G - 1)); 9737c478bd9Sstevel@tonic-gate physmax = PFN_4G - 1; 9747c478bd9Sstevel@tonic-gate } 9757c478bd9Sstevel@tonic-gate } 9767c478bd9Sstevel@tonic-gate #endif 9777c478bd9Sstevel@tonic-gate 9787c478bd9Sstevel@tonic-gate startup_build_mem_nodes(bootops->boot_mem->physinstalled); 9797c478bd9Sstevel@tonic-gate 9807c478bd9Sstevel@tonic-gate if (BOP_GETPROPLEN(bootops, "enforce-prot-exec") >= 0) { 9817c478bd9Sstevel@tonic-gate int len = BOP_GETPROPLEN(bootops, "enforce-prot-exec"); 9827c478bd9Sstevel@tonic-gate char value[8]; 9837c478bd9Sstevel@tonic-gate 9847c478bd9Sstevel@tonic-gate if (len < 8) 9857c478bd9Sstevel@tonic-gate (void) BOP_GETPROP(bootops, "enforce-prot-exec", value); 9867c478bd9Sstevel@tonic-gate else 9877c478bd9Sstevel@tonic-gate (void) strcpy(value, ""); 9887c478bd9Sstevel@tonic-gate if (strcmp(value, "off") == 0) 9897c478bd9Sstevel@tonic-gate mmu.pt_nx = 0; 9907c478bd9Sstevel@tonic-gate } 9917c478bd9Sstevel@tonic-gate PRM_DEBUG(mmu.pt_nx); 9927c478bd9Sstevel@tonic-gate 9937c478bd9Sstevel@tonic-gate /* 9947c478bd9Sstevel@tonic-gate * We will need page_t's for every page in the system, except for 9957c478bd9Sstevel@tonic-gate * memory mapped at or above above the start of the kernel text segment. 9967c478bd9Sstevel@tonic-gate * 9977c478bd9Sstevel@tonic-gate * pages above e_modtext are attributed to kernel debugger (obp_pages) 9987c478bd9Sstevel@tonic-gate */ 9997c478bd9Sstevel@tonic-gate npages = physinstalled - 1; /* avail_filter() skips page 0, so "- 1" */ 10007c478bd9Sstevel@tonic-gate obp_pages = 0; 10017c478bd9Sstevel@tonic-gate va = KERNEL_TEXT; 1002ae115bc7Smrj while (kbm_probe(&va, &len, &pfn, &prot) != 0) { 10037c478bd9Sstevel@tonic-gate npages -= len >> MMU_PAGESHIFT; 10047c478bd9Sstevel@tonic-gate if (va >= (uintptr_t)e_moddata) 10057c478bd9Sstevel@tonic-gate obp_pages += len >> MMU_PAGESHIFT; 10067c478bd9Sstevel@tonic-gate va += len; 10077c478bd9Sstevel@tonic-gate } 10087c478bd9Sstevel@tonic-gate PRM_DEBUG(npages); 10097c478bd9Sstevel@tonic-gate PRM_DEBUG(obp_pages); 10107c478bd9Sstevel@tonic-gate 10117c478bd9Sstevel@tonic-gate /* 101235b1ab99Sjosephb * If physmem is patched to be non-zero, use it instead of the computed 101335b1ab99Sjosephb * value unless it is larger than the actual amount of memory on hand. 10147c478bd9Sstevel@tonic-gate */ 10154944b02eSkchow if (physmem == 0 || physmem > npages) { 10167c478bd9Sstevel@tonic-gate physmem = npages; 10174944b02eSkchow } else if (physmem < npages) { 1018f53ad214Skchow orig_npages = npages; 10197c478bd9Sstevel@tonic-gate npages = physmem; 10204944b02eSkchow } 10217c478bd9Sstevel@tonic-gate PRM_DEBUG(physmem); 10227c478bd9Sstevel@tonic-gate 10237c478bd9Sstevel@tonic-gate /* 10247c478bd9Sstevel@tonic-gate * We now compute the sizes of all the initial allocations for 10257c478bd9Sstevel@tonic-gate * structures the kernel needs in order do kmem_alloc(). These 10267c478bd9Sstevel@tonic-gate * include: 10277c478bd9Sstevel@tonic-gate * memsegs 10287c478bd9Sstevel@tonic-gate * memlists 10297c478bd9Sstevel@tonic-gate * page hash table 10307c478bd9Sstevel@tonic-gate * page_t's 10317c478bd9Sstevel@tonic-gate * page coloring data structs 10327c478bd9Sstevel@tonic-gate */ 10337c478bd9Sstevel@tonic-gate memseg_sz = sizeof (struct memseg) * (memblocks + POSS_NEW_FRAGMENTS); 10347c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(memseg_base, memseg_sz); 10357c478bd9Sstevel@tonic-gate PRM_DEBUG(memseg_sz); 10367c478bd9Sstevel@tonic-gate 10377c478bd9Sstevel@tonic-gate /* 1038ae115bc7Smrj * Reserve space for memlists. There's no real good way to know exactly 1039ae115bc7Smrj * how much room we'll need, but this should be a good upper bound. 10407c478bd9Sstevel@tonic-gate */ 10417c478bd9Sstevel@tonic-gate memlist_sz = ROUND_UP_PAGE(2 * sizeof (struct memlist) * 10427c478bd9Sstevel@tonic-gate (memblocks + POSS_NEW_FRAGMENTS)); 10437c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(memlist, memlist_sz); 10447c478bd9Sstevel@tonic-gate PRM_DEBUG(memlist_sz); 10457c478bd9Sstevel@tonic-gate 10467c478bd9Sstevel@tonic-gate /* 10477c478bd9Sstevel@tonic-gate * The page structure hash table size is a power of 2 10487c478bd9Sstevel@tonic-gate * such that the average hash chain length is PAGE_HASHAVELEN. 10497c478bd9Sstevel@tonic-gate */ 10507c478bd9Sstevel@tonic-gate page_hashsz = npages / PAGE_HASHAVELEN; 10517c478bd9Sstevel@tonic-gate page_hashsz = 1 << highbit(page_hashsz); 10527c478bd9Sstevel@tonic-gate pagehash_sz = sizeof (struct page *) * page_hashsz; 10537c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(page_hash, pagehash_sz); 10547c478bd9Sstevel@tonic-gate PRM_DEBUG(pagehash_sz); 10557c478bd9Sstevel@tonic-gate 10567c478bd9Sstevel@tonic-gate /* 1057ae115bc7Smrj * Set aside room for the page structures themselves. 10587c478bd9Sstevel@tonic-gate */ 1059ae115bc7Smrj PRM_DEBUG(npages); 1060ae115bc7Smrj pp_sz = sizeof (struct page) * npages; 10617c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(pp_base, pp_sz); 10627c478bd9Sstevel@tonic-gate PRM_DEBUG(pp_sz); 10637c478bd9Sstevel@tonic-gate 10647c478bd9Sstevel@tonic-gate /* 10657c478bd9Sstevel@tonic-gate * determine l2 cache info and memory size for page coloring 10667c478bd9Sstevel@tonic-gate */ 10677c478bd9Sstevel@tonic-gate (void) getl2cacheinfo(CPU, 10687c478bd9Sstevel@tonic-gate &l2cache_sz, &l2cache_linesz, &l2cache_assoc); 10697c478bd9Sstevel@tonic-gate pagecolor_memsz = 10707c478bd9Sstevel@tonic-gate page_coloring_init(l2cache_sz, l2cache_linesz, l2cache_assoc); 10717c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(pagecolor_mem, pagecolor_memsz); 10727c478bd9Sstevel@tonic-gate PRM_DEBUG(pagecolor_memsz); 10737c478bd9Sstevel@tonic-gate 10747c478bd9Sstevel@tonic-gate page_ctrs_size = page_ctrs_sz(); 10757c478bd9Sstevel@tonic-gate ADD_TO_ALLOCATIONS(page_ctrs_mem, page_ctrs_size); 10767c478bd9Sstevel@tonic-gate PRM_DEBUG(page_ctrs_size); 10777c478bd9Sstevel@tonic-gate 1078d7d93655Sblakej /* 1079d7d93655Sblakej * Allocate the array that protects pp->p_selock. 1080d7d93655Sblakej */ 1081d7d93655Sblakej pse_shift = size_pse_array(physmem, max_ncpus); 1082d7d93655Sblakej pse_table_size = 1 << pse_shift; 1083d7d93655Sblakej pse_table_alloc_size = pse_table_size * sizeof (pad_mutex_t); 1084d7d93655Sblakej ADD_TO_ALLOCATIONS(pse_mutex, pse_table_alloc_size); 1085d7d93655Sblakej 1086ae115bc7Smrj #if defined(__amd64) 10877c478bd9Sstevel@tonic-gate valloc_sz = ROUND_UP_LPAGE(valloc_sz); 1088ae115bc7Smrj valloc_base = VALLOC_BASE; 108935b1ab99Sjosephb 109035b1ab99Sjosephb /* 109135b1ab99Sjosephb * The default values of VALLOC_BASE and SEGKPM_BASE should work 109235b1ab99Sjosephb * for values of physmax up to 1 Terabyte. They need adjusting when 109335b1ab99Sjosephb * memory is at addresses above 1 TB. 109435b1ab99Sjosephb */ 109535b1ab99Sjosephb if (physmax + 1 > mmu_btop(TERABYTE)) { 109635b1ab99Sjosephb uint64_t kpm_resv_amount = mmu_ptob(physmax + 1); 109735b1ab99Sjosephb 109835b1ab99Sjosephb /* Round to largest possible pagesize for now */ 109935b1ab99Sjosephb kpm_resv_amount = P2ROUNDUP(kpm_resv_amount, ONE_GIG); 110035b1ab99Sjosephb 110135b1ab99Sjosephb segkpm_base = -(2 * kpm_resv_amount); /* down from top VA */ 110235b1ab99Sjosephb 110335b1ab99Sjosephb /* make sure we leave some space for user apps above hole */ 110435b1ab99Sjosephb segkpm_base = MAX(segkpm_base, AMD64_VA_HOLE_END + TERABYTE); 110535b1ab99Sjosephb if (segkpm_base > SEGKPM_BASE) 110635b1ab99Sjosephb segkpm_base = SEGKPM_BASE; 110735b1ab99Sjosephb PRM_DEBUG(segkpm_base); 110835b1ab99Sjosephb 110935b1ab99Sjosephb valloc_base = segkpm_base + kpm_resv_amount; 111035b1ab99Sjosephb PRM_DEBUG(valloc_base); 111135b1ab99Sjosephb } 1112ae115bc7Smrj #else /* __i386 */ 1113ae115bc7Smrj valloc_base = (uintptr_t)(MISC_VA_BASE - valloc_sz); 1114ae115bc7Smrj valloc_base = P2ALIGN(valloc_base, mmu.level_size[1]); 11157c478bd9Sstevel@tonic-gate PRM_DEBUG(valloc_base); 111635b1ab99Sjosephb #endif /* __i386 */ 1117ae115bc7Smrj 1118ae115bc7Smrj /* 1119ae115bc7Smrj * do all the initial allocations 1120ae115bc7Smrj */ 1121ae115bc7Smrj perform_allocations(); 1122ae115bc7Smrj 1123ae115bc7Smrj /* 1124ae115bc7Smrj * Build phys_install and phys_avail in kernel memspace. 1125ae115bc7Smrj * - phys_install should be all memory in the system. 1126ae115bc7Smrj * - phys_avail is phys_install minus any memory mapped before this 1127ae115bc7Smrj * point above KERNEL_TEXT. 1128ae115bc7Smrj */ 1129ae115bc7Smrj current = phys_install = memlist; 1130ae115bc7Smrj copy_memlist_filter(bootops->boot_mem->physinstalled, ¤t, NULL); 1131ae115bc7Smrj if ((caddr_t)current > (caddr_t)memlist + memlist_sz) 1132ae115bc7Smrj panic("physinstalled was too big!"); 1133ae115bc7Smrj if (prom_debug) 1134ae115bc7Smrj print_memlist("phys_install", phys_install); 1135ae115bc7Smrj 1136ae115bc7Smrj phys_avail = current; 1137ae115bc7Smrj PRM_POINT("Building phys_avail:\n"); 1138ae115bc7Smrj copy_memlist_filter(bootops->boot_mem->physinstalled, ¤t, 1139ae115bc7Smrj avail_filter); 1140ae115bc7Smrj if ((caddr_t)current > (caddr_t)memlist + memlist_sz) 1141ae115bc7Smrj panic("physavail was too big!"); 1142ae115bc7Smrj if (prom_debug) 1143ae115bc7Smrj print_memlist("phys_avail", phys_avail); 1144ae115bc7Smrj 1145ae115bc7Smrj /* 1146ae115bc7Smrj * setup page coloring 1147ae115bc7Smrj */ 1148ae115bc7Smrj page_coloring_setup(pagecolor_mem); 1149ae115bc7Smrj page_lock_init(); /* currently a no-op */ 1150ae115bc7Smrj 1151ae115bc7Smrj /* 1152ae115bc7Smrj * free page list counters 1153ae115bc7Smrj */ 1154ae115bc7Smrj (void) page_ctrs_alloc(page_ctrs_mem); 1155ae115bc7Smrj 1156ae115bc7Smrj /* 1157ae115bc7Smrj * Initialize the page structures from the memory lists. 1158ae115bc7Smrj */ 1159ae115bc7Smrj availrmem_initial = availrmem = freemem = 0; 1160ae115bc7Smrj PRM_POINT("Calling kphysm_init()..."); 1161ae115bc7Smrj npages = kphysm_init(pp_base, npages); 1162ae115bc7Smrj PRM_POINT("kphysm_init() done"); 1163ae115bc7Smrj PRM_DEBUG(npages); 1164ae115bc7Smrj 1165ae115bc7Smrj init_debug_info(); 1166ae115bc7Smrj 1167ae115bc7Smrj /* 1168ae115bc7Smrj * Now that page_t's have been initialized, remove all the 1169ae115bc7Smrj * initial allocation pages from the kernel free page lists. 1170ae115bc7Smrj */ 1171ae115bc7Smrj boot_mapin((caddr_t)valloc_base, valloc_sz); 11720cfdb603Sjosephb boot_mapin((caddr_t)MISC_VA_BASE, MISC_VA_SIZE); 1173ae115bc7Smrj PRM_POINT("startup_memlist() done"); 1174ae115bc7Smrj 1175ae115bc7Smrj PRM_DEBUG(valloc_sz); 11762cb27123Saguzovsk 1177567d55e1Saguzovsk #if defined(__amd64) 1178567d55e1Saguzovsk if ((availrmem >> (30 - MMU_PAGESHIFT)) >= 1179567d55e1Saguzovsk textrepl_min_gb && l2cache_sz <= 2 << 20) { 1180567d55e1Saguzovsk extern size_t textrepl_size_thresh; 11812cb27123Saguzovsk textrepl_size_thresh = (16 << 20) - 1; 11822cb27123Saguzovsk } 1183567d55e1Saguzovsk #endif 1184ae115bc7Smrj } 1185ae115bc7Smrj 1186ae115bc7Smrj /* 1187ae115bc7Smrj * Layout the kernel's part of address space and initialize kmem allocator. 1188ae115bc7Smrj */ 1189ae115bc7Smrj static void 1190ae115bc7Smrj startup_kmem(void) 1191ae115bc7Smrj { 1192932dc8e5Sdp extern void page_set_colorequiv_arr(void); 1193932dc8e5Sdp 1194ae115bc7Smrj PRM_POINT("startup_kmem() starting..."); 11957c478bd9Sstevel@tonic-gate 11967c478bd9Sstevel@tonic-gate #if defined(__amd64) 11977c478bd9Sstevel@tonic-gate if (eprom_kernelbase && eprom_kernelbase != KERNELBASE) 11987c478bd9Sstevel@tonic-gate cmn_err(CE_NOTE, "!kernelbase cannot be changed on 64-bit " 11997c478bd9Sstevel@tonic-gate "systems."); 120035b1ab99Sjosephb kernelbase = segkpm_base - KERNEL_REDZONE_SIZE; 12017c478bd9Sstevel@tonic-gate core_base = (uintptr_t)COREHEAP_BASE; 1202ae115bc7Smrj core_size = (size_t)MISC_VA_BASE - COREHEAP_BASE; 12037c478bd9Sstevel@tonic-gate #else /* __i386 */ 12047c478bd9Sstevel@tonic-gate /* 12057c478bd9Sstevel@tonic-gate * We configure kernelbase based on: 12067c478bd9Sstevel@tonic-gate * 12077c478bd9Sstevel@tonic-gate * 1. user specified kernelbase via eeprom command. Value cannot exceed 12087c478bd9Sstevel@tonic-gate * KERNELBASE_MAX. we large page align eprom_kernelbase 12097c478bd9Sstevel@tonic-gate * 12107c478bd9Sstevel@tonic-gate * 2. Default to KERNELBASE and adjust to 2X less the size for page_t. 12117c478bd9Sstevel@tonic-gate * On large memory systems we must lower kernelbase to allow 12127c478bd9Sstevel@tonic-gate * enough room for page_t's for all of memory. 12137c478bd9Sstevel@tonic-gate * 12147c478bd9Sstevel@tonic-gate * The value set here, might be changed a little later. 12157c478bd9Sstevel@tonic-gate */ 12167c478bd9Sstevel@tonic-gate if (eprom_kernelbase) { 12177c478bd9Sstevel@tonic-gate kernelbase = eprom_kernelbase & mmu.level_mask[1]; 12187c478bd9Sstevel@tonic-gate if (kernelbase > KERNELBASE_MAX) 12197c478bd9Sstevel@tonic-gate kernelbase = KERNELBASE_MAX; 12207c478bd9Sstevel@tonic-gate } else { 12217c478bd9Sstevel@tonic-gate kernelbase = (uintptr_t)KERNELBASE; 12227c478bd9Sstevel@tonic-gate kernelbase -= ROUND_UP_4MEG(2 * valloc_sz); 12237c478bd9Sstevel@tonic-gate } 12247c478bd9Sstevel@tonic-gate ASSERT((kernelbase & mmu.level_offset[1]) == 0); 1225ae115bc7Smrj core_base = valloc_base; 12267c478bd9Sstevel@tonic-gate core_size = 0; 1227ae115bc7Smrj #endif /* __i386 */ 12287c478bd9Sstevel@tonic-gate 12297c478bd9Sstevel@tonic-gate PRM_DEBUG(core_base); 12307c478bd9Sstevel@tonic-gate PRM_DEBUG(core_size); 1231ae115bc7Smrj PRM_DEBUG(kernelbase); 12327c478bd9Sstevel@tonic-gate 123335b1ab99Sjosephb #if defined(__i386) 12347c478bd9Sstevel@tonic-gate segkp_fromheap = 1; 1235ae115bc7Smrj #endif /* __i386 */ 123635b1ab99Sjosephb 12377c478bd9Sstevel@tonic-gate ekernelheap = (char *)core_base; 1238ae115bc7Smrj PRM_DEBUG(ekernelheap); 12397c478bd9Sstevel@tonic-gate 12407c478bd9Sstevel@tonic-gate /* 12417c478bd9Sstevel@tonic-gate * Now that we know the real value of kernelbase, 12427c478bd9Sstevel@tonic-gate * update variables that were initialized with a value of 12437c478bd9Sstevel@tonic-gate * KERNELBASE (in common/conf/param.c). 12447c478bd9Sstevel@tonic-gate * 12457c478bd9Sstevel@tonic-gate * XXX The problem with this sort of hackery is that the 12467c478bd9Sstevel@tonic-gate * compiler just may feel like putting the const declarations 12477c478bd9Sstevel@tonic-gate * (in param.c) into the .text section. Perhaps they should 12487c478bd9Sstevel@tonic-gate * just be declared as variables there? 12497c478bd9Sstevel@tonic-gate */ 12507c478bd9Sstevel@tonic-gate 12517c478bd9Sstevel@tonic-gate *(uintptr_t *)&_kernelbase = kernelbase; 12527c478bd9Sstevel@tonic-gate *(uintptr_t *)&_userlimit = kernelbase; 1253ccbaea4fSjosephb #if defined(__amd64) 1254ccbaea4fSjosephb *(uintptr_t *)&_userlimit -= KERNELBASE - USERLIMIT; 1255ccbaea4fSjosephb #else 12567c478bd9Sstevel@tonic-gate *(uintptr_t *)&_userlimit32 = _userlimit; 12577c478bd9Sstevel@tonic-gate #endif 12587c478bd9Sstevel@tonic-gate PRM_DEBUG(_kernelbase); 12597c478bd9Sstevel@tonic-gate PRM_DEBUG(_userlimit); 12607c478bd9Sstevel@tonic-gate PRM_DEBUG(_userlimit32); 12617c478bd9Sstevel@tonic-gate 126235b1ab99Sjosephb layout_kernel_va(); 126335b1ab99Sjosephb 126435b1ab99Sjosephb #if defined(__i386) 126535b1ab99Sjosephb /* 126635b1ab99Sjosephb * If segmap is too large we can push the bottom of the kernel heap 126735b1ab99Sjosephb * higher than the base. Or worse, it could exceed the top of the 126835b1ab99Sjosephb * VA space entirely, causing it to wrap around. 126935b1ab99Sjosephb */ 127035b1ab99Sjosephb if (kernelheap >= ekernelheap || (uintptr_t)kernelheap < kernelbase) 127135b1ab99Sjosephb panic("too little address space available for kernelheap," 127235b1ab99Sjosephb " use eeprom for lower kernelbase or smaller segmapsize"); 127335b1ab99Sjosephb #endif /* __i386 */ 127435b1ab99Sjosephb 12757c478bd9Sstevel@tonic-gate /* 12767c478bd9Sstevel@tonic-gate * Initialize the kernel heap. Note 3rd argument must be > 1st. 12777c478bd9Sstevel@tonic-gate */ 127835b1ab99Sjosephb kernelheap_init(kernelheap, ekernelheap, 127935b1ab99Sjosephb kernelheap + MMU_PAGESIZE, 1280ae115bc7Smrj (void *)core_base, (void *)(core_base + core_size)); 12817c478bd9Sstevel@tonic-gate 1282843e1988Sjohnlev #if defined(__xpv) 1283843e1988Sjohnlev /* 1284843e1988Sjohnlev * Link pending events struct into cpu struct 1285843e1988Sjohnlev */ 1286843e1988Sjohnlev CPU->cpu_m.mcpu_evt_pend = &cpu0_evt_data; 1287843e1988Sjohnlev #endif 12887c478bd9Sstevel@tonic-gate /* 12897c478bd9Sstevel@tonic-gate * Initialize kernel memory allocator. 12907c478bd9Sstevel@tonic-gate */ 12917c478bd9Sstevel@tonic-gate kmem_init(); 12927c478bd9Sstevel@tonic-gate 1293932dc8e5Sdp /* 1294932dc8e5Sdp * Factor in colorequiv to check additional 'equivalent' bins 1295932dc8e5Sdp */ 1296932dc8e5Sdp page_set_colorequiv_arr(); 1297932dc8e5Sdp 12987c478bd9Sstevel@tonic-gate /* 12997c478bd9Sstevel@tonic-gate * print this out early so that we know what's going on 13007c478bd9Sstevel@tonic-gate */ 13017c478bd9Sstevel@tonic-gate cmn_err(CE_CONT, "?features: %b\n", x86_feature, FMT_X86_FEATURE); 13027c478bd9Sstevel@tonic-gate 13037c478bd9Sstevel@tonic-gate /* 13047c478bd9Sstevel@tonic-gate * Initialize bp_mapin(). 13057c478bd9Sstevel@tonic-gate */ 13067c478bd9Sstevel@tonic-gate bp_init(MMU_PAGESIZE, HAT_STORECACHING_OK); 13077c478bd9Sstevel@tonic-gate 1308f53ad214Skchow /* 1309f53ad214Skchow * orig_npages is non-zero if physmem has been configured for less 1310f53ad214Skchow * than the available memory. 1311f53ad214Skchow */ 1312f53ad214Skchow if (orig_npages) { 131335b1ab99Sjosephb cmn_err(CE_WARN, "!%slimiting physmem to 0x%lx of 0x%lx pages", 131435b1ab99Sjosephb (npages == PHYSMEM ? "Due to virtual address space " : ""), 131535b1ab99Sjosephb npages, orig_npages); 1316f53ad214Skchow } 13177c478bd9Sstevel@tonic-gate #if defined(__i386) 13187c478bd9Sstevel@tonic-gate if (eprom_kernelbase && (eprom_kernelbase != kernelbase)) 13197c478bd9Sstevel@tonic-gate cmn_err(CE_WARN, "kernelbase value, User specified 0x%lx, " 13207c478bd9Sstevel@tonic-gate "System using 0x%lx", 13217c478bd9Sstevel@tonic-gate (uintptr_t)eprom_kernelbase, (uintptr_t)kernelbase); 13227c478bd9Sstevel@tonic-gate #endif 13237c478bd9Sstevel@tonic-gate 13247c478bd9Sstevel@tonic-gate #ifdef KERNELBASE_ABI_MIN 13257c478bd9Sstevel@tonic-gate if (kernelbase < (uintptr_t)KERNELBASE_ABI_MIN) { 13267c478bd9Sstevel@tonic-gate cmn_err(CE_NOTE, "!kernelbase set to 0x%lx, system is not " 13277c478bd9Sstevel@tonic-gate "i386 ABI compliant.", (uintptr_t)kernelbase); 13287c478bd9Sstevel@tonic-gate } 13297c478bd9Sstevel@tonic-gate #endif 13307c478bd9Sstevel@tonic-gate 1331843e1988Sjohnlev #ifdef __xpv 1332843e1988Sjohnlev /* 1333843e1988Sjohnlev * Some of the xen start information has to be relocated up 1334843e1988Sjohnlev * into the kernel's permanent address space. 1335843e1988Sjohnlev */ 1336843e1988Sjohnlev PRM_POINT("calling xen_relocate_start_info()"); 1337843e1988Sjohnlev xen_relocate_start_info(); 1338843e1988Sjohnlev PRM_POINT("xen_relocate_start_info() done"); 1339843e1988Sjohnlev 1340843e1988Sjohnlev /* 1341843e1988Sjohnlev * (Update the vcpu pointer in our cpu structure to point into 1342843e1988Sjohnlev * the relocated shared info.) 1343843e1988Sjohnlev */ 1344843e1988Sjohnlev CPU->cpu_m.mcpu_vcpu_info = 1345843e1988Sjohnlev &HYPERVISOR_shared_info->vcpu_info[CPU->cpu_id]; 1346843e1988Sjohnlev #endif 1347843e1988Sjohnlev 1348ae115bc7Smrj PRM_POINT("startup_kmem() done"); 13497c478bd9Sstevel@tonic-gate } 13507c478bd9Sstevel@tonic-gate 1351551bc2a6Smrj #ifndef __xpv 1352551bc2a6Smrj /* 1353551bc2a6Smrj * If we have detected that we are running in an HVM environment, we need 1354551bc2a6Smrj * to prepend the PV driver directory to the module search path. 1355551bc2a6Smrj */ 1356551bc2a6Smrj #define HVM_MOD_DIR "/platform/i86hvm/kernel" 1357551bc2a6Smrj static void 1358551bc2a6Smrj update_default_path() 1359551bc2a6Smrj { 1360551bc2a6Smrj char *current, *newpath; 1361551bc2a6Smrj int newlen; 1362551bc2a6Smrj 1363551bc2a6Smrj /* 1364551bc2a6Smrj * We are about to resync with krtld. krtld will reset its 1365551bc2a6Smrj * internal module search path iff Solaris has set default_path. 1366551bc2a6Smrj * We want to be sure we're prepending this new directory to the 1367551bc2a6Smrj * right search path. 1368551bc2a6Smrj */ 1369551bc2a6Smrj current = (default_path == NULL) ? kobj_module_path : default_path; 1370551bc2a6Smrj 1371551bc2a6Smrj newlen = strlen(HVM_MOD_DIR) + strlen(current) + 1; 1372551bc2a6Smrj newpath = kmem_alloc(newlen, KM_SLEEP); 1373551bc2a6Smrj (void) strcpy(newpath, HVM_MOD_DIR); 1374551bc2a6Smrj (void) strcat(newpath, " "); 1375551bc2a6Smrj (void) strcat(newpath, current); 1376551bc2a6Smrj 1377551bc2a6Smrj default_path = newpath; 1378551bc2a6Smrj } 1379551bc2a6Smrj #endif 1380551bc2a6Smrj 13817c478bd9Sstevel@tonic-gate static void 13827c478bd9Sstevel@tonic-gate startup_modules(void) 13837c478bd9Sstevel@tonic-gate { 13847c478bd9Sstevel@tonic-gate unsigned int i; 13857c478bd9Sstevel@tonic-gate extern void prom_setup(void); 13867c478bd9Sstevel@tonic-gate 13877c478bd9Sstevel@tonic-gate PRM_POINT("startup_modules() starting..."); 1388843e1988Sjohnlev 1389843e1988Sjohnlev #ifndef __xpv 13907c478bd9Sstevel@tonic-gate /* 13917c478bd9Sstevel@tonic-gate * Initialize ten-micro second timer so that drivers will 13927c478bd9Sstevel@tonic-gate * not get short changed in their init phase. This was 13937c478bd9Sstevel@tonic-gate * not getting called until clkinit which, on fast cpu's 13947c478bd9Sstevel@tonic-gate * caused the drv_usecwait to be way too short. 13957c478bd9Sstevel@tonic-gate */ 13967c478bd9Sstevel@tonic-gate microfind(); 1397551bc2a6Smrj 1398551bc2a6Smrj if (xpv_is_hvm) 1399551bc2a6Smrj update_default_path(); 1400843e1988Sjohnlev #endif 14017c478bd9Sstevel@tonic-gate 14027c478bd9Sstevel@tonic-gate /* 14037c478bd9Sstevel@tonic-gate * Read the GMT lag from /etc/rtc_config. 14047c478bd9Sstevel@tonic-gate */ 1405ae115bc7Smrj sgmtl(process_rtc_config_file()); 14067c478bd9Sstevel@tonic-gate 14077c478bd9Sstevel@tonic-gate /* 14087c478bd9Sstevel@tonic-gate * Calculate default settings of system parameters based upon 14097c478bd9Sstevel@tonic-gate * maxusers, yet allow to be overridden via the /etc/system file. 14107c478bd9Sstevel@tonic-gate */ 14117c478bd9Sstevel@tonic-gate param_calc(0); 14127c478bd9Sstevel@tonic-gate 14137c478bd9Sstevel@tonic-gate mod_setup(); 14147c478bd9Sstevel@tonic-gate 14157c478bd9Sstevel@tonic-gate /* 14167c478bd9Sstevel@tonic-gate * Initialize system parameters. 14177c478bd9Sstevel@tonic-gate */ 14187c478bd9Sstevel@tonic-gate param_init(); 14197c478bd9Sstevel@tonic-gate 14209acbbeafSnn /* 14219acbbeafSnn * Initialize the default brands 14229acbbeafSnn */ 14239acbbeafSnn brand_init(); 14249acbbeafSnn 14257c478bd9Sstevel@tonic-gate /* 14267c478bd9Sstevel@tonic-gate * maxmem is the amount of physical memory we're playing with. 14277c478bd9Sstevel@tonic-gate */ 14287c478bd9Sstevel@tonic-gate maxmem = physmem; 14297c478bd9Sstevel@tonic-gate 14307c478bd9Sstevel@tonic-gate /* 14317c478bd9Sstevel@tonic-gate * Initialize segment management stuff. 14327c478bd9Sstevel@tonic-gate */ 14337c478bd9Sstevel@tonic-gate seg_init(); 14347c478bd9Sstevel@tonic-gate 14357c478bd9Sstevel@tonic-gate if (modload("fs", "specfs") == -1) 14367c478bd9Sstevel@tonic-gate halt("Can't load specfs"); 14377c478bd9Sstevel@tonic-gate 14387c478bd9Sstevel@tonic-gate if (modload("fs", "devfs") == -1) 14397c478bd9Sstevel@tonic-gate halt("Can't load devfs"); 14407c478bd9Sstevel@tonic-gate 1441facf4a8dSllai if (modload("fs", "dev") == -1) 1442facf4a8dSllai halt("Can't load dev"); 1443facf4a8dSllai 144445916cd2Sjpk (void) modloadonly("sys", "lbl_edition"); 144545916cd2Sjpk 14467c478bd9Sstevel@tonic-gate dispinit(); 14477c478bd9Sstevel@tonic-gate 14487c478bd9Sstevel@tonic-gate /* 14497c478bd9Sstevel@tonic-gate * This is needed here to initialize hw_serial[] for cluster booting. 14507c478bd9Sstevel@tonic-gate */ 14517c478bd9Sstevel@tonic-gate if ((i = modload("misc", "sysinit")) != (unsigned int)-1) 14527c478bd9Sstevel@tonic-gate (void) modunload(i); 14537c478bd9Sstevel@tonic-gate else 14547c478bd9Sstevel@tonic-gate cmn_err(CE_CONT, "sysinit load failed"); 14557c478bd9Sstevel@tonic-gate 14567c478bd9Sstevel@tonic-gate /* Read cluster configuration data. */ 14577c478bd9Sstevel@tonic-gate clconf_init(); 14587c478bd9Sstevel@tonic-gate 1459843e1988Sjohnlev #if defined(__xpv) 1460843e1988Sjohnlev ec_init(); 1461843e1988Sjohnlev gnttab_init(); 1462843e1988Sjohnlev (void) xs_early_init(); 1463843e1988Sjohnlev #endif /* __xpv */ 1464843e1988Sjohnlev 14657c478bd9Sstevel@tonic-gate /* 14667c478bd9Sstevel@tonic-gate * Create a kernel device tree. First, create rootnex and 14677c478bd9Sstevel@tonic-gate * then invoke bus specific code to probe devices. 14687c478bd9Sstevel@tonic-gate */ 14697c478bd9Sstevel@tonic-gate setup_ddi(); 14707aec1d6eScindi 147120c794b3Sgavinm #ifndef __xpv 147220c794b3Sgavinm { 147320c794b3Sgavinm /* 147420c794b3Sgavinm * Set up the CPU module subsystem. Modifies the device tree, 147520c794b3Sgavinm * so it must be done after setup_ddi(). 147620c794b3Sgavinm */ 14777aec1d6eScindi 147820c794b3Sgavinm cmi_hdl_t hdl; 147920c794b3Sgavinm 148020c794b3Sgavinm if ((hdl = cmi_init(CMI_HDL_NATIVE, cmi_ntv_hwchipid(CPU), 148120c794b3Sgavinm cmi_ntv_hwcoreid(CPU), cmi_ntv_hwstrandid(CPU))) != NULL) { 148220c794b3Sgavinm if (x86_feature & X86_MCA) 148320c794b3Sgavinm cmi_mca_init(hdl); 148420c794b3Sgavinm } 148520c794b3Sgavinm } 148620c794b3Sgavinm #endif /* __xpv */ 14877aec1d6eScindi 14887c478bd9Sstevel@tonic-gate /* 14897c478bd9Sstevel@tonic-gate * Fake a prom tree such that /dev/openprom continues to work 14907c478bd9Sstevel@tonic-gate */ 1491ae115bc7Smrj PRM_POINT("startup_modules: calling prom_setup..."); 14927c478bd9Sstevel@tonic-gate prom_setup(); 1493ae115bc7Smrj PRM_POINT("startup_modules: done"); 14947c478bd9Sstevel@tonic-gate 14957c478bd9Sstevel@tonic-gate /* 14967c478bd9Sstevel@tonic-gate * Load all platform specific modules 14977c478bd9Sstevel@tonic-gate */ 1498ae115bc7Smrj PRM_POINT("startup_modules: calling psm_modload..."); 14997c478bd9Sstevel@tonic-gate psm_modload(); 15007c478bd9Sstevel@tonic-gate 15017c478bd9Sstevel@tonic-gate PRM_POINT("startup_modules() done"); 15027c478bd9Sstevel@tonic-gate } 15037c478bd9Sstevel@tonic-gate 1504ae115bc7Smrj /* 1505ae115bc7Smrj * claim a "setaside" boot page for use in the kernel 1506ae115bc7Smrj */ 1507ae115bc7Smrj page_t * 1508ae115bc7Smrj boot_claim_page(pfn_t pfn) 15097c478bd9Sstevel@tonic-gate { 1510ae115bc7Smrj page_t *pp; 15117c478bd9Sstevel@tonic-gate 1512ae115bc7Smrj pp = page_numtopp_nolock(pfn); 1513ae115bc7Smrj ASSERT(pp != NULL); 15147c478bd9Sstevel@tonic-gate 1515ae115bc7Smrj if (PP_ISBOOTPAGES(pp)) { 1516ae115bc7Smrj if (pp->p_next != NULL) 1517ae115bc7Smrj pp->p_next->p_prev = pp->p_prev; 1518ae115bc7Smrj if (pp->p_prev == NULL) 1519ae115bc7Smrj bootpages = pp->p_next; 1520ae115bc7Smrj else 1521ae115bc7Smrj pp->p_prev->p_next = pp->p_next; 1522ae115bc7Smrj } else { 1523ae115bc7Smrj /* 1524ae115bc7Smrj * htable_attach() expects a base pagesize page 1525ae115bc7Smrj */ 1526ae115bc7Smrj if (pp->p_szc != 0) 1527ae115bc7Smrj page_boot_demote(pp); 1528ae115bc7Smrj pp = page_numtopp(pfn, SE_EXCL); 1529ae115bc7Smrj } 1530ae115bc7Smrj return (pp); 15317c478bd9Sstevel@tonic-gate } 15327c478bd9Sstevel@tonic-gate 15337c478bd9Sstevel@tonic-gate /* 15347c478bd9Sstevel@tonic-gate * Walk through the pagetables looking for pages mapped in by boot. If the 15357c478bd9Sstevel@tonic-gate * setaside flag is set the pages are expected to be returned to the 15367c478bd9Sstevel@tonic-gate * kernel later in boot, so we add them to the bootpages list. 15377c478bd9Sstevel@tonic-gate */ 15387c478bd9Sstevel@tonic-gate static void 15397c478bd9Sstevel@tonic-gate protect_boot_range(uintptr_t low, uintptr_t high, int setaside) 15407c478bd9Sstevel@tonic-gate { 15417c478bd9Sstevel@tonic-gate uintptr_t va = low; 15427c478bd9Sstevel@tonic-gate size_t len; 15437c478bd9Sstevel@tonic-gate uint_t prot; 15447c478bd9Sstevel@tonic-gate pfn_t pfn; 15457c478bd9Sstevel@tonic-gate page_t *pp; 15467c478bd9Sstevel@tonic-gate pgcnt_t boot_protect_cnt = 0; 15477c478bd9Sstevel@tonic-gate 1548ae115bc7Smrj while (kbm_probe(&va, &len, &pfn, &prot) != 0 && va < high) { 15497c478bd9Sstevel@tonic-gate if (va + len >= high) 15507c478bd9Sstevel@tonic-gate panic("0x%lx byte mapping at 0x%p exceeds boot's " 15517c478bd9Sstevel@tonic-gate "legal range.", len, (void *)va); 15527c478bd9Sstevel@tonic-gate 15537c478bd9Sstevel@tonic-gate while (len > 0) { 15547c478bd9Sstevel@tonic-gate pp = page_numtopp_alloc(pfn); 15557c478bd9Sstevel@tonic-gate if (pp != NULL) { 15567c478bd9Sstevel@tonic-gate if (setaside == 0) 15577c478bd9Sstevel@tonic-gate panic("Unexpected mapping by boot. " 15587c478bd9Sstevel@tonic-gate "addr=%p pfn=%lx\n", 15597c478bd9Sstevel@tonic-gate (void *)va, pfn); 15607c478bd9Sstevel@tonic-gate 15617c478bd9Sstevel@tonic-gate pp->p_next = bootpages; 1562ae115bc7Smrj pp->p_prev = NULL; 1563ae115bc7Smrj PP_SETBOOTPAGES(pp); 1564ae115bc7Smrj if (bootpages != NULL) { 1565ae115bc7Smrj bootpages->p_prev = pp; 1566ae115bc7Smrj } 15677c478bd9Sstevel@tonic-gate bootpages = pp; 15687c478bd9Sstevel@tonic-gate ++boot_protect_cnt; 15697c478bd9Sstevel@tonic-gate } 15707c478bd9Sstevel@tonic-gate 15717c478bd9Sstevel@tonic-gate ++pfn; 15727c478bd9Sstevel@tonic-gate len -= MMU_PAGESIZE; 15737c478bd9Sstevel@tonic-gate va += MMU_PAGESIZE; 15747c478bd9Sstevel@tonic-gate } 15757c478bd9Sstevel@tonic-gate } 15767c478bd9Sstevel@tonic-gate PRM_DEBUG(boot_protect_cnt); 15777c478bd9Sstevel@tonic-gate } 15787c478bd9Sstevel@tonic-gate 1579ae115bc7Smrj /* 158035b1ab99Sjosephb * 1581ae115bc7Smrj */ 15827c478bd9Sstevel@tonic-gate static void 158335b1ab99Sjosephb layout_kernel_va(void) 15847c478bd9Sstevel@tonic-gate { 158535b1ab99Sjosephb PRM_POINT("layout_kernel_va() starting..."); 15867c478bd9Sstevel@tonic-gate /* 1587ae115bc7Smrj * Establish the final size of the kernel's heap, size of segmap, 1588ae115bc7Smrj * segkp, etc. 15897c478bd9Sstevel@tonic-gate */ 15907c478bd9Sstevel@tonic-gate 15917c478bd9Sstevel@tonic-gate #if defined(__amd64) 15927c478bd9Sstevel@tonic-gate 159335b1ab99Sjosephb kpm_vbase = (caddr_t)segkpm_base; 159435b1ab99Sjosephb kpm_size = ROUND_UP_LPAGE(mmu_ptob(physmax + 1)); 159535b1ab99Sjosephb if ((uintptr_t)kpm_vbase + kpm_size > (uintptr_t)valloc_base) 159635b1ab99Sjosephb panic("not enough room for kpm!"); 1597ae115bc7Smrj PRM_DEBUG(kpm_size); 1598ae115bc7Smrj PRM_DEBUG(kpm_vbase); 15997c478bd9Sstevel@tonic-gate 16007c478bd9Sstevel@tonic-gate /* 1601ae115bc7Smrj * By default we create a seg_kp in 64 bit kernels, it's a little 1602ae115bc7Smrj * faster to access than embedding it in the heap. 16037c478bd9Sstevel@tonic-gate */ 1604ae115bc7Smrj segkp_base = (caddr_t)valloc_base + valloc_sz; 16057c478bd9Sstevel@tonic-gate if (!segkp_fromheap) { 16067c478bd9Sstevel@tonic-gate size_t sz = mmu_ptob(segkpsize); 16077c478bd9Sstevel@tonic-gate 16087c478bd9Sstevel@tonic-gate /* 1609ae115bc7Smrj * determine size of segkp 16107c478bd9Sstevel@tonic-gate */ 16117c478bd9Sstevel@tonic-gate if (sz < SEGKPMINSIZE || sz > SEGKPMAXSIZE) { 16127c478bd9Sstevel@tonic-gate sz = SEGKPDEFSIZE; 16137c478bd9Sstevel@tonic-gate cmn_err(CE_WARN, "!Illegal value for segkpsize. " 16147c478bd9Sstevel@tonic-gate "segkpsize has been reset to %ld pages", 16157c478bd9Sstevel@tonic-gate mmu_btop(sz)); 16167c478bd9Sstevel@tonic-gate } 16177c478bd9Sstevel@tonic-gate sz = MIN(sz, MAX(SEGKPMINSIZE, mmu_ptob(physmem))); 16187c478bd9Sstevel@tonic-gate 16197c478bd9Sstevel@tonic-gate segkpsize = mmu_btop(ROUND_UP_LPAGE(sz)); 16207c478bd9Sstevel@tonic-gate } 1621ae115bc7Smrj PRM_DEBUG(segkp_base); 1622ae115bc7Smrj PRM_DEBUG(segkpsize); 16237c478bd9Sstevel@tonic-gate 162435b1ab99Sjosephb /* 162535b1ab99Sjosephb * segzio is used for ZFS cached data. It uses a distinct VA 162635b1ab99Sjosephb * segment (from kernel heap) so that we can easily tell not to 162735b1ab99Sjosephb * include it in kernel crash dumps on 64 bit kernels. The trick is 162835b1ab99Sjosephb * to give it lots of VA, but not constrain the kernel heap. 162935b1ab99Sjosephb * We scale the size of segzio linearly with physmem up to 163035b1ab99Sjosephb * SEGZIOMAXSIZE. Above that amount it scales at 50% of physmem. 163135b1ab99Sjosephb */ 1632ae115bc7Smrj segzio_base = segkp_base + mmu_ptob(segkpsize); 1633ae115bc7Smrj if (segzio_fromheap) { 1634ae115bc7Smrj segziosize = 0; 1635ae115bc7Smrj } else { 163635b1ab99Sjosephb size_t physmem_size = mmu_ptob(physmem); 163735b1ab99Sjosephb size_t size = (segziosize == 0) ? 163835b1ab99Sjosephb physmem_size : mmu_ptob(segziosize); 1639ad23a2dbSjohansen 164035b1ab99Sjosephb if (size < SEGZIOMINSIZE) 1641ad23a2dbSjohansen size = SEGZIOMINSIZE; 164235b1ab99Sjosephb if (size > SEGZIOMAXSIZE) { 1643a778305aSjohansen size = SEGZIOMAXSIZE; 164435b1ab99Sjosephb if (physmem_size > size) 164535b1ab99Sjosephb size += (physmem_size - size) / 2; 1646ad23a2dbSjohansen } 1647ad23a2dbSjohansen segziosize = mmu_btop(ROUND_UP_LPAGE(size)); 1648ad23a2dbSjohansen } 1649ae115bc7Smrj PRM_DEBUG(segziosize); 1650ae115bc7Smrj PRM_DEBUG(segzio_base); 1651ad23a2dbSjohansen 16527c478bd9Sstevel@tonic-gate /* 1653ae115bc7Smrj * Put the range of VA for device mappings next, kmdb knows to not 1654ae115bc7Smrj * grep in this range of addresses. 16557c478bd9Sstevel@tonic-gate */ 1656ae115bc7Smrj toxic_addr = 1657ae115bc7Smrj ROUND_UP_LPAGE((uintptr_t)segzio_base + mmu_ptob(segziosize)); 16587c478bd9Sstevel@tonic-gate PRM_DEBUG(toxic_addr); 1659ae115bc7Smrj segmap_start = ROUND_UP_LPAGE(toxic_addr + toxic_size); 1660ae115bc7Smrj #else /* __i386 */ 1661ae115bc7Smrj segmap_start = ROUND_UP_LPAGE(kernelbase); 1662ae115bc7Smrj #endif /* __i386 */ 1663ae115bc7Smrj PRM_DEBUG(segmap_start); 16647c478bd9Sstevel@tonic-gate 16657c478bd9Sstevel@tonic-gate /* 16667c478bd9Sstevel@tonic-gate * Users can change segmapsize through eeprom or /etc/system. 16677c478bd9Sstevel@tonic-gate * If the variable is tuned through eeprom, there is no upper 16687c478bd9Sstevel@tonic-gate * bound on the size of segmap. If it is tuned through 16697c478bd9Sstevel@tonic-gate * /etc/system on 32-bit systems, it must be no larger than we 16707c478bd9Sstevel@tonic-gate * planned for in startup_memlist(). 16717c478bd9Sstevel@tonic-gate */ 16727c478bd9Sstevel@tonic-gate segmapsize = MAX(ROUND_UP_LPAGE(segmapsize), SEGMAPDEFAULT); 16737c478bd9Sstevel@tonic-gate 16747c478bd9Sstevel@tonic-gate #if defined(__i386) 16757c478bd9Sstevel@tonic-gate /* 16767c478bd9Sstevel@tonic-gate * 32-bit systems don't have segkpm or segkp, so segmap appears at 16777c478bd9Sstevel@tonic-gate * the bottom of the kernel's address range. Set aside space for a 1678ae115bc7Smrj * small red zone just below the start of segmap. 16797c478bd9Sstevel@tonic-gate */ 1680ae115bc7Smrj segmap_start += KERNEL_REDZONE_SIZE; 16817c478bd9Sstevel@tonic-gate segmapsize -= KERNEL_REDZONE_SIZE; 16827c478bd9Sstevel@tonic-gate #endif 16837c478bd9Sstevel@tonic-gate 1684ae115bc7Smrj PRM_DEBUG(segmap_start); 16857c478bd9Sstevel@tonic-gate PRM_DEBUG(segmapsize); 168635b1ab99Sjosephb kernelheap = (caddr_t)ROUND_UP_LPAGE(segmap_start + segmapsize); 168735b1ab99Sjosephb PRM_DEBUG(kernelheap); 168835b1ab99Sjosephb PRM_POINT("layout_kernel_va() done..."); 168935b1ab99Sjosephb } 169035b1ab99Sjosephb 169135b1ab99Sjosephb /* 169235b1ab99Sjosephb * Finish initializing the VM system, now that we are no longer 169335b1ab99Sjosephb * relying on the boot time memory allocators. 169435b1ab99Sjosephb */ 169535b1ab99Sjosephb static void 169635b1ab99Sjosephb startup_vm(void) 169735b1ab99Sjosephb { 169835b1ab99Sjosephb struct segmap_crargs a; 169935b1ab99Sjosephb 170035b1ab99Sjosephb extern int use_brk_lpg, use_stk_lpg; 170135b1ab99Sjosephb 170235b1ab99Sjosephb PRM_POINT("startup_vm() starting..."); 170335b1ab99Sjosephb 170435b1ab99Sjosephb /* 170535b1ab99Sjosephb * Initialize the hat layer. 170635b1ab99Sjosephb */ 170735b1ab99Sjosephb hat_init(); 17087c478bd9Sstevel@tonic-gate 1709ae115bc7Smrj /* 1710ae115bc7Smrj * Do final allocations of HAT data structures that need to 1711ae115bc7Smrj * be allocated before quiescing the boot loader. 1712ae115bc7Smrj */ 1713ae115bc7Smrj PRM_POINT("Calling hat_kern_alloc()..."); 1714ae115bc7Smrj hat_kern_alloc((caddr_t)segmap_start, segmapsize, ekernelheap); 1715ae115bc7Smrj PRM_POINT("hat_kern_alloc() done"); 1716ae115bc7Smrj 1717843e1988Sjohnlev #ifndef __xpv 1718ae115bc7Smrj /* 17191d03c31eSjohnlev * Setup Page Attribute Table 1720ae115bc7Smrj */ 17211d03c31eSjohnlev pat_sync(); 1722843e1988Sjohnlev #endif 1723ae115bc7Smrj 1724ae115bc7Smrj /* 1725ae115bc7Smrj * The next two loops are done in distinct steps in order 1726ae115bc7Smrj * to be sure that any page that is doubly mapped (both above 1727ae115bc7Smrj * KERNEL_TEXT and below kernelbase) is dealt with correctly. 1728ae115bc7Smrj * Note this may never happen, but it might someday. 1729ae115bc7Smrj */ 1730ae115bc7Smrj bootpages = NULL; 1731ae115bc7Smrj PRM_POINT("Protecting boot pages"); 1732ae115bc7Smrj 1733ae115bc7Smrj /* 1734ae115bc7Smrj * Protect any pages mapped above KERNEL_TEXT that somehow have 1735ae115bc7Smrj * page_t's. This can only happen if something weird allocated 1736ae115bc7Smrj * in this range (like kadb/kmdb). 1737ae115bc7Smrj */ 1738ae115bc7Smrj protect_boot_range(KERNEL_TEXT, (uintptr_t)-1, 0); 1739ae115bc7Smrj 1740ae115bc7Smrj /* 1741ae115bc7Smrj * Before we can take over memory allocation/mapping from the boot 1742ae115bc7Smrj * loader we must remove from our free page lists any boot allocated 1743ae115bc7Smrj * pages that stay mapped until release_bootstrap(). 1744ae115bc7Smrj */ 1745ae115bc7Smrj protect_boot_range(0, kernelbase, 1); 1746ae115bc7Smrj 1747843e1988Sjohnlev 1748ae115bc7Smrj /* 1749ae115bc7Smrj * Switch to running on regular HAT (not boot_mmu) 1750ae115bc7Smrj */ 1751ae115bc7Smrj PRM_POINT("Calling hat_kern_setup()..."); 1752ae115bc7Smrj hat_kern_setup(); 1753ae115bc7Smrj 1754ae115bc7Smrj /* 1755ae115bc7Smrj * It is no longer safe to call BOP_ALLOC(), so make sure we don't. 1756ae115bc7Smrj */ 1757ae115bc7Smrj bop_no_more_mem(); 1758ae115bc7Smrj 1759ae115bc7Smrj PRM_POINT("hat_kern_setup() done"); 1760ae115bc7Smrj 1761ae115bc7Smrj hat_cpu_online(CPU); 1762ae115bc7Smrj 17637c478bd9Sstevel@tonic-gate /* 17647c478bd9Sstevel@tonic-gate * Initialize VM system 17657c478bd9Sstevel@tonic-gate */ 17667c478bd9Sstevel@tonic-gate PRM_POINT("Calling kvm_init()..."); 17677c478bd9Sstevel@tonic-gate kvm_init(); 17687c478bd9Sstevel@tonic-gate PRM_POINT("kvm_init() done"); 17697c478bd9Sstevel@tonic-gate 17707c478bd9Sstevel@tonic-gate /* 17717c478bd9Sstevel@tonic-gate * Tell kmdb that the VM system is now working 17727c478bd9Sstevel@tonic-gate */ 17737c478bd9Sstevel@tonic-gate if (boothowto & RB_DEBUG) 17747c478bd9Sstevel@tonic-gate kdi_dvec_vmready(); 17757c478bd9Sstevel@tonic-gate 1776843e1988Sjohnlev #if defined(__xpv) 1777843e1988Sjohnlev /* 1778843e1988Sjohnlev * Populate the I/O pool on domain 0 1779843e1988Sjohnlev */ 1780843e1988Sjohnlev if (DOMAIN_IS_INITDOMAIN(xen_info)) { 1781843e1988Sjohnlev extern long populate_io_pool(void); 1782843e1988Sjohnlev long init_io_pool_cnt; 1783843e1988Sjohnlev 1784843e1988Sjohnlev PRM_POINT("Populating reserve I/O page pool"); 1785843e1988Sjohnlev init_io_pool_cnt = populate_io_pool(); 1786843e1988Sjohnlev PRM_DEBUG(init_io_pool_cnt); 1787843e1988Sjohnlev } 1788843e1988Sjohnlev #endif 17897c478bd9Sstevel@tonic-gate /* 17907c478bd9Sstevel@tonic-gate * Mangle the brand string etc. 17917c478bd9Sstevel@tonic-gate */ 17927c478bd9Sstevel@tonic-gate cpuid_pass3(CPU); 17937c478bd9Sstevel@tonic-gate 17947c478bd9Sstevel@tonic-gate #if defined(__amd64) 17957c478bd9Sstevel@tonic-gate 17967c478bd9Sstevel@tonic-gate /* 17977c478bd9Sstevel@tonic-gate * Create the device arena for toxic (to dtrace/kmdb) mappings. 17987c478bd9Sstevel@tonic-gate */ 17997c478bd9Sstevel@tonic-gate device_arena = vmem_create("device", (void *)toxic_addr, 18007c478bd9Sstevel@tonic-gate toxic_size, MMU_PAGESIZE, NULL, NULL, NULL, 0, VM_SLEEP); 18017c478bd9Sstevel@tonic-gate 18027c478bd9Sstevel@tonic-gate #else /* __i386 */ 18037c478bd9Sstevel@tonic-gate 18047c478bd9Sstevel@tonic-gate /* 18057c478bd9Sstevel@tonic-gate * allocate the bit map that tracks toxic pages 18067c478bd9Sstevel@tonic-gate */ 1807ae115bc7Smrj toxic_bit_map_len = btop((ulong_t)(valloc_base - kernelbase)); 18087c478bd9Sstevel@tonic-gate PRM_DEBUG(toxic_bit_map_len); 18097c478bd9Sstevel@tonic-gate toxic_bit_map = 18107c478bd9Sstevel@tonic-gate kmem_zalloc(BT_SIZEOFMAP(toxic_bit_map_len), KM_NOSLEEP); 18117c478bd9Sstevel@tonic-gate ASSERT(toxic_bit_map != NULL); 18127c478bd9Sstevel@tonic-gate PRM_DEBUG(toxic_bit_map); 18137c478bd9Sstevel@tonic-gate 18147c478bd9Sstevel@tonic-gate #endif /* __i386 */ 18157c478bd9Sstevel@tonic-gate 18167c478bd9Sstevel@tonic-gate 18177c478bd9Sstevel@tonic-gate /* 18187c478bd9Sstevel@tonic-gate * Now that we've got more VA, as well as the ability to allocate from 18197c478bd9Sstevel@tonic-gate * it, tell the debugger. 18207c478bd9Sstevel@tonic-gate */ 18217c478bd9Sstevel@tonic-gate if (boothowto & RB_DEBUG) 18227c478bd9Sstevel@tonic-gate kdi_dvec_memavail(); 18237c478bd9Sstevel@tonic-gate 18247c478bd9Sstevel@tonic-gate /* 18257c478bd9Sstevel@tonic-gate * The following code installs a special page fault handler (#pf) 18267c478bd9Sstevel@tonic-gate * to work around a pentium bug. 18277c478bd9Sstevel@tonic-gate */ 1828843e1988Sjohnlev #if !defined(__amd64) && !defined(__xpv) 18297c478bd9Sstevel@tonic-gate if (x86_type == X86_TYPE_P5) { 1830ae115bc7Smrj desctbr_t idtr; 18317c478bd9Sstevel@tonic-gate gate_desc_t *newidt; 18327c478bd9Sstevel@tonic-gate 18337c478bd9Sstevel@tonic-gate if ((newidt = kmem_zalloc(MMU_PAGESIZE, KM_NOSLEEP)) == NULL) 18347c478bd9Sstevel@tonic-gate panic("failed to install pentium_pftrap"); 18357c478bd9Sstevel@tonic-gate 18360cfdb603Sjosephb bcopy(idt0, newidt, NIDT * sizeof (*idt0)); 18377c478bd9Sstevel@tonic-gate set_gatesegd(&newidt[T_PGFLT], &pentium_pftrap, 1838843e1988Sjohnlev KCS_SEL, SDT_SYSIGT, TRP_KPL); 18397c478bd9Sstevel@tonic-gate 18407c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, (caddr_t)newidt, MMU_PAGESIZE, 18410cfdb603Sjosephb PROT_READ | PROT_EXEC); 18427c478bd9Sstevel@tonic-gate 18430cfdb603Sjosephb CPU->cpu_idt = newidt; 18440cfdb603Sjosephb idtr.dtr_base = (uintptr_t)CPU->cpu_idt; 18450cfdb603Sjosephb idtr.dtr_limit = (NIDT * sizeof (*idt0)) - 1; 1846ae115bc7Smrj wr_idtr(&idtr); 18477c478bd9Sstevel@tonic-gate } 18487c478bd9Sstevel@tonic-gate #endif /* !__amd64 */ 18497c478bd9Sstevel@tonic-gate 1850843e1988Sjohnlev #if !defined(__xpv) 18517c478bd9Sstevel@tonic-gate /* 18527c478bd9Sstevel@tonic-gate * Map page pfn=0 for drivers, such as kd, that need to pick up 18537c478bd9Sstevel@tonic-gate * parameters left there by controllers/BIOS. 18547c478bd9Sstevel@tonic-gate */ 18557c478bd9Sstevel@tonic-gate PRM_POINT("setup up p0_va"); 18567c478bd9Sstevel@tonic-gate p0_va = i86devmap(0, 1, PROT_READ); 18577c478bd9Sstevel@tonic-gate PRM_DEBUG(p0_va); 1858843e1988Sjohnlev #endif 18597c478bd9Sstevel@tonic-gate 18607c478bd9Sstevel@tonic-gate cmn_err(CE_CONT, "?mem = %luK (0x%lx)\n", 18617c478bd9Sstevel@tonic-gate physinstalled << (MMU_PAGESHIFT - 10), ptob(physinstalled)); 18627c478bd9Sstevel@tonic-gate 186383f9b804Skchow /* 186483f9b804Skchow * disable automatic large pages for small memory systems or 186583f9b804Skchow * when the disable flag is set. 186602bc52beSkchow * 186702bc52beSkchow * Do not yet consider page sizes larger than 2m/4m. 186883f9b804Skchow */ 1869ec25b48fSsusans if (!auto_lpg_disable && mmu.max_page_level > 0) { 1870ec25b48fSsusans max_uheap_lpsize = LEVEL_SIZE(1); 1871ec25b48fSsusans max_ustack_lpsize = LEVEL_SIZE(1); 1872ec25b48fSsusans max_privmap_lpsize = LEVEL_SIZE(1); 1873ec25b48fSsusans max_uidata_lpsize = LEVEL_SIZE(1); 1874ec25b48fSsusans max_utext_lpsize = LEVEL_SIZE(1); 1875ec25b48fSsusans max_shm_lpsize = LEVEL_SIZE(1); 1876ec25b48fSsusans } 1877ec25b48fSsusans if (physmem < privm_lpg_min_physmem || mmu.max_page_level == 0 || 1878ec25b48fSsusans auto_lpg_disable) { 1879beb1bda0Sdavemq use_brk_lpg = 0; 1880beb1bda0Sdavemq use_stk_lpg = 0; 1881ec25b48fSsusans } 188202bc52beSkchow mcntl0_lpsize = LEVEL_SIZE(mmu.umax_page_level); 1883beb1bda0Sdavemq 18847c478bd9Sstevel@tonic-gate PRM_POINT("Calling hat_init_finish()..."); 18857c478bd9Sstevel@tonic-gate hat_init_finish(); 18867c478bd9Sstevel@tonic-gate PRM_POINT("hat_init_finish() done"); 18877c478bd9Sstevel@tonic-gate 18887c478bd9Sstevel@tonic-gate /* 18897c478bd9Sstevel@tonic-gate * Initialize the segkp segment type. 18907c478bd9Sstevel@tonic-gate */ 18917c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 1892ae115bc7Smrj PRM_POINT("Attaching segkp"); 1893ae115bc7Smrj if (segkp_fromheap) { 18947c478bd9Sstevel@tonic-gate segkp->s_as = &kas; 1895ae115bc7Smrj } else if (seg_attach(&kas, (caddr_t)segkp_base, mmu_ptob(segkpsize), 1896ae115bc7Smrj segkp) < 0) { 1897ae115bc7Smrj panic("startup: cannot attach segkp"); 1898ae115bc7Smrj /*NOTREACHED*/ 18997c478bd9Sstevel@tonic-gate } 1900ae115bc7Smrj PRM_POINT("Doing segkp_create()"); 19017c478bd9Sstevel@tonic-gate if (segkp_create(segkp) != 0) { 19027c478bd9Sstevel@tonic-gate panic("startup: segkp_create failed"); 19037c478bd9Sstevel@tonic-gate /*NOTREACHED*/ 19047c478bd9Sstevel@tonic-gate } 19057c478bd9Sstevel@tonic-gate PRM_DEBUG(segkp); 19067c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 19077c478bd9Sstevel@tonic-gate 19087c478bd9Sstevel@tonic-gate /* 19097c478bd9Sstevel@tonic-gate * kpm segment 19107c478bd9Sstevel@tonic-gate */ 19117c478bd9Sstevel@tonic-gate segmap_kpm = 0; 19127c478bd9Sstevel@tonic-gate if (kpm_desired) { 19137c478bd9Sstevel@tonic-gate kpm_init(); 19147c478bd9Sstevel@tonic-gate kpm_enable = 1; 1915a5652762Spraks vpm_enable = 1; 19167c478bd9Sstevel@tonic-gate } 19177c478bd9Sstevel@tonic-gate 19187c478bd9Sstevel@tonic-gate /* 19197c478bd9Sstevel@tonic-gate * Now create segmap segment. 19207c478bd9Sstevel@tonic-gate */ 19217c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 1922ae115bc7Smrj if (seg_attach(&kas, (caddr_t)segmap_start, segmapsize, segmap) < 0) { 1923ae115bc7Smrj panic("cannot attach segmap"); 19247c478bd9Sstevel@tonic-gate /*NOTREACHED*/ 19257c478bd9Sstevel@tonic-gate } 1926ae115bc7Smrj PRM_DEBUG(segmap); 19277c478bd9Sstevel@tonic-gate 19287c478bd9Sstevel@tonic-gate a.prot = PROT_READ | PROT_WRITE; 19297c478bd9Sstevel@tonic-gate a.shmsize = 0; 19307c478bd9Sstevel@tonic-gate a.nfreelist = segmapfreelists; 19317c478bd9Sstevel@tonic-gate 1932ae115bc7Smrj if (segmap_create(segmap, (caddr_t)&a) != 0) 1933ae115bc7Smrj panic("segmap_create segmap"); 19347c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 19357c478bd9Sstevel@tonic-gate 19367c478bd9Sstevel@tonic-gate setup_vaddr_for_ppcopy(CPU); 19377c478bd9Sstevel@tonic-gate 19387c478bd9Sstevel@tonic-gate segdev_init(); 1939843e1988Sjohnlev #if defined(__xpv) 1940843e1988Sjohnlev if (DOMAIN_IS_INITDOMAIN(xen_info)) 1941843e1988Sjohnlev #endif 1942843e1988Sjohnlev pmem_init(); 1943ae115bc7Smrj 19447c478bd9Sstevel@tonic-gate PRM_POINT("startup_vm() done"); 19457c478bd9Sstevel@tonic-gate } 19467c478bd9Sstevel@tonic-gate 1947ae115bc7Smrj /* 1948ae115bc7Smrj * Load a tod module for the non-standard tod part found on this system. 1949ae115bc7Smrj */ 1950ae115bc7Smrj static void 1951ae115bc7Smrj load_tod_module(char *todmod) 1952ae115bc7Smrj { 1953ae115bc7Smrj if (modload("tod", todmod) == -1) 1954ae115bc7Smrj halt("Can't load TOD module"); 1955ae115bc7Smrj } 1956ae115bc7Smrj 19577c478bd9Sstevel@tonic-gate static void 19587c478bd9Sstevel@tonic-gate startup_end(void) 19597c478bd9Sstevel@tonic-gate { 1960dd4eeefdSeota int i; 19617c478bd9Sstevel@tonic-gate extern void setx86isalist(void); 19627c478bd9Sstevel@tonic-gate 19637c478bd9Sstevel@tonic-gate PRM_POINT("startup_end() starting..."); 19647c478bd9Sstevel@tonic-gate 19657c478bd9Sstevel@tonic-gate /* 19667c478bd9Sstevel@tonic-gate * Perform tasks that get done after most of the VM 19677c478bd9Sstevel@tonic-gate * initialization has been done but before the clock 19687c478bd9Sstevel@tonic-gate * and other devices get started. 19697c478bd9Sstevel@tonic-gate */ 19707c478bd9Sstevel@tonic-gate kern_setup1(); 19717c478bd9Sstevel@tonic-gate 19727c478bd9Sstevel@tonic-gate /* 19737c478bd9Sstevel@tonic-gate * Perform CPC initialization for this CPU. 19747c478bd9Sstevel@tonic-gate */ 19757c478bd9Sstevel@tonic-gate kcpc_hw_init(CPU); 19767c478bd9Sstevel@tonic-gate 1977ee88d2b9Skchow #if defined(OPTERON_WORKAROUND_6323525) 1978ee88d2b9Skchow if (opteron_workaround_6323525) 1979ee88d2b9Skchow patch_workaround_6323525(); 1980ee88d2b9Skchow #endif 1981ae115bc7Smrj /* 1982ae115bc7Smrj * If needed, load TOD module now so that ddi_get_time(9F) etc. work 1983ae115bc7Smrj * (For now, "needed" is defined as set tod_module_name in /etc/system) 1984ae115bc7Smrj */ 1985ae115bc7Smrj if (tod_module_name != NULL) { 1986ae115bc7Smrj PRM_POINT("load_tod_module()"); 1987ae115bc7Smrj load_tod_module(tod_module_name); 1988ae115bc7Smrj } 1989ae115bc7Smrj 1990843e1988Sjohnlev #if defined(__xpv) 1991843e1988Sjohnlev /* 1992843e1988Sjohnlev * Forceload interposing TOD module for the hypervisor. 1993843e1988Sjohnlev */ 1994843e1988Sjohnlev PRM_POINT("load_tod_module()"); 1995843e1988Sjohnlev load_tod_module("xpvtod"); 1996843e1988Sjohnlev #endif 1997843e1988Sjohnlev 19987c478bd9Sstevel@tonic-gate /* 19997c478bd9Sstevel@tonic-gate * Configure the system. 20007c478bd9Sstevel@tonic-gate */ 20017c478bd9Sstevel@tonic-gate PRM_POINT("Calling configure()..."); 20027c478bd9Sstevel@tonic-gate configure(); /* set up devices */ 20037c478bd9Sstevel@tonic-gate PRM_POINT("configure() done"); 20047c478bd9Sstevel@tonic-gate 20057c478bd9Sstevel@tonic-gate /* 20067c478bd9Sstevel@tonic-gate * Set the isa_list string to the defined instruction sets we 20077c478bd9Sstevel@tonic-gate * support. 20087c478bd9Sstevel@tonic-gate */ 20097c478bd9Sstevel@tonic-gate setx86isalist(); 2010100b72f4Sandrei cpu_intr_alloc(CPU, NINTR_THREADS); 20117c478bd9Sstevel@tonic-gate psm_install(); 20127c478bd9Sstevel@tonic-gate 20137c478bd9Sstevel@tonic-gate /* 20147c478bd9Sstevel@tonic-gate * We're done with bootops. We don't unmap the bootstrap yet because 20157c478bd9Sstevel@tonic-gate * we're still using bootsvcs. 20167c478bd9Sstevel@tonic-gate */ 2017ae115bc7Smrj PRM_POINT("NULLing out bootops"); 2018ae115bc7Smrj *bootopsp = (struct bootops *)NULL; 20197c478bd9Sstevel@tonic-gate bootops = (struct bootops *)NULL; 20207c478bd9Sstevel@tonic-gate 2021843e1988Sjohnlev #if defined(__xpv) 2022843e1988Sjohnlev ec_init_debug_irq(); 2023843e1988Sjohnlev xs_domu_init(); 2024843e1988Sjohnlev #endif 20257c478bd9Sstevel@tonic-gate PRM_POINT("Enabling interrupts"); 20267c478bd9Sstevel@tonic-gate (*picinitf)(); 20277c478bd9Sstevel@tonic-gate sti(); 2028843e1988Sjohnlev #if defined(__xpv) 2029843e1988Sjohnlev ASSERT(CPU->cpu_m.mcpu_vcpu_info->evtchn_upcall_mask == 0); 2030843e1988Sjohnlev xen_late_startup(); 2031843e1988Sjohnlev #endif 20327c478bd9Sstevel@tonic-gate 20337c478bd9Sstevel@tonic-gate (void) add_avsoftintr((void *)&softlevel1_hdl, 1, softlevel1, 203435b1ab99Sjosephb "softlevel1", NULL, NULL); /* XXX to be moved later */ 20357c478bd9Sstevel@tonic-gate 2036dd4eeefdSeota /* 2037dd4eeefdSeota * Register these software interrupts for ddi timer. 2038dd4eeefdSeota * Software interrupts up to the level 10 are supported. 2039dd4eeefdSeota */ 2040dd4eeefdSeota for (i = DDI_IPL_1; i <= DDI_IPL_10; i++) { 2041dd4eeefdSeota char name[sizeof ("timer_softintr") + 2]; 2042dd4eeefdSeota (void) sprintf(name, "timer_softintr%02d", i); 2043dd4eeefdSeota (void) add_avsoftintr((void *)&softlevel_hdl[i-1], i, 2044dd4eeefdSeota (avfunc)timer_softintr, name, (caddr_t)(uintptr_t)i, NULL); 2045dd4eeefdSeota } 2046dd4eeefdSeota 20477c478bd9Sstevel@tonic-gate PRM_POINT("startup_end() done"); 20487c478bd9Sstevel@tonic-gate } 20497c478bd9Sstevel@tonic-gate 20507c478bd9Sstevel@tonic-gate extern char hw_serial[]; 20517c478bd9Sstevel@tonic-gate char *_hs1107 = hw_serial; 20527c478bd9Sstevel@tonic-gate ulong_t _bdhs34; 20537c478bd9Sstevel@tonic-gate 20547c478bd9Sstevel@tonic-gate void 20557c478bd9Sstevel@tonic-gate post_startup(void) 20567c478bd9Sstevel@tonic-gate { 20577c478bd9Sstevel@tonic-gate /* 20587c478bd9Sstevel@tonic-gate * Set the system wide, processor-specific flags to be passed 20597c478bd9Sstevel@tonic-gate * to userland via the aux vector for performance hints and 20607c478bd9Sstevel@tonic-gate * instruction set extensions. 20617c478bd9Sstevel@tonic-gate */ 20627c478bd9Sstevel@tonic-gate bind_hwcap(); 20637c478bd9Sstevel@tonic-gate 2064843e1988Sjohnlev #ifdef __xpv 2065843e1988Sjohnlev if (DOMAIN_IS_INITDOMAIN(xen_info)) 2066843e1988Sjohnlev #endif 2067843e1988Sjohnlev { 2068843e1988Sjohnlev /* 2069843e1988Sjohnlev * Load the System Management BIOS into the global ksmbios 2070843e1988Sjohnlev * handle, if an SMBIOS is present on this system. 2071843e1988Sjohnlev */ 2072843e1988Sjohnlev ksmbios = smbios_open(NULL, SMB_VERSION, ksmbios_flags, NULL); 20737c478bd9Sstevel@tonic-gate 2074843e1988Sjohnlev #if defined(__xpv) 2075843e1988Sjohnlev xpv_panic_init(); 2076843e1988Sjohnlev #else 2077843e1988Sjohnlev /* 2078843e1988Sjohnlev * Startup the memory scrubber. 2079843e1988Sjohnlev * XXPV This should be running somewhere .. 2080843e1988Sjohnlev */ 2081843e1988Sjohnlev memscrub_init(); 2082843e1988Sjohnlev #endif 2083843e1988Sjohnlev } 20847c478bd9Sstevel@tonic-gate 20857aec1d6eScindi /* 20867aec1d6eScindi * Complete CPU module initialization 20877aec1d6eScindi */ 208820c794b3Sgavinm cmi_post_startup(); 20897aec1d6eScindi 20907c478bd9Sstevel@tonic-gate /* 209184ab085aSmws * Perform forceloading tasks for /etc/system. 20927c478bd9Sstevel@tonic-gate */ 209384ab085aSmws (void) mod_sysctl(SYS_FORCELOAD, NULL); 20947c478bd9Sstevel@tonic-gate 20957c478bd9Sstevel@tonic-gate /* 20967c478bd9Sstevel@tonic-gate * ON4.0: Force /proc module in until clock interrupt handle fixed 20977c478bd9Sstevel@tonic-gate * ON4.0: This must be fixed or restated in /etc/systems. 20987c478bd9Sstevel@tonic-gate */ 20997c478bd9Sstevel@tonic-gate (void) modload("fs", "procfs"); 21007c478bd9Sstevel@tonic-gate 2101c35aa225Smarx (void) i_ddi_attach_hw_nodes("pit_beep"); 2102c35aa225Smarx 21037c478bd9Sstevel@tonic-gate #if defined(__i386) 21047c478bd9Sstevel@tonic-gate /* 21057c478bd9Sstevel@tonic-gate * Check for required functional Floating Point hardware, 21067c478bd9Sstevel@tonic-gate * unless FP hardware explicitly disabled. 21077c478bd9Sstevel@tonic-gate */ 21087c478bd9Sstevel@tonic-gate if (fpu_exists && (fpu_pentium_fdivbug || fp_kind == FP_NO)) 21097c478bd9Sstevel@tonic-gate halt("No working FP hardware found"); 21107c478bd9Sstevel@tonic-gate #endif 21117c478bd9Sstevel@tonic-gate 21127c478bd9Sstevel@tonic-gate maxmem = freemem; 21137c478bd9Sstevel@tonic-gate 21147c478bd9Sstevel@tonic-gate add_cpunode2devtree(CPU->cpu_id, CPU->cpu_m.mcpu_cpi); 21157c478bd9Sstevel@tonic-gate } 21167c478bd9Sstevel@tonic-gate 21177c478bd9Sstevel@tonic-gate static int 21187c478bd9Sstevel@tonic-gate pp_in_ramdisk(page_t *pp) 21197c478bd9Sstevel@tonic-gate { 21207c478bd9Sstevel@tonic-gate extern uint64_t ramdisk_start, ramdisk_end; 21217c478bd9Sstevel@tonic-gate 21227c478bd9Sstevel@tonic-gate return ((pp->p_pagenum >= btop(ramdisk_start)) && 21237c478bd9Sstevel@tonic-gate (pp->p_pagenum < btopr(ramdisk_end))); 21247c478bd9Sstevel@tonic-gate } 21257c478bd9Sstevel@tonic-gate 21267c478bd9Sstevel@tonic-gate void 21277c478bd9Sstevel@tonic-gate release_bootstrap(void) 21287c478bd9Sstevel@tonic-gate { 21297c478bd9Sstevel@tonic-gate int root_is_ramdisk; 21307c478bd9Sstevel@tonic-gate page_t *pp; 21317c478bd9Sstevel@tonic-gate extern void kobj_boot_unmountroot(void); 21327c478bd9Sstevel@tonic-gate extern dev_t rootdev; 21337c478bd9Sstevel@tonic-gate 21347c478bd9Sstevel@tonic-gate /* unmount boot ramdisk and release kmem usage */ 21357c478bd9Sstevel@tonic-gate kobj_boot_unmountroot(); 21367c478bd9Sstevel@tonic-gate 21377c478bd9Sstevel@tonic-gate /* 21387c478bd9Sstevel@tonic-gate * We're finished using the boot loader so free its pages. 21397c478bd9Sstevel@tonic-gate */ 21407c478bd9Sstevel@tonic-gate PRM_POINT("Unmapping lower boot pages"); 2141ae115bc7Smrj clear_boot_mappings(0, _userlimit); 2142ae115bc7Smrj postbootkernelbase = kernelbase; 21437c478bd9Sstevel@tonic-gate 21447c478bd9Sstevel@tonic-gate /* 21457c478bd9Sstevel@tonic-gate * If root isn't on ramdisk, destroy the hardcoded 21467c478bd9Sstevel@tonic-gate * ramdisk node now and release the memory. Else, 21477c478bd9Sstevel@tonic-gate * ramdisk memory is kept in rd_pages. 21487c478bd9Sstevel@tonic-gate */ 21497c478bd9Sstevel@tonic-gate root_is_ramdisk = (getmajor(rootdev) == ddi_name_to_major("ramdisk")); 21507c478bd9Sstevel@tonic-gate if (!root_is_ramdisk) { 21517c478bd9Sstevel@tonic-gate dev_info_t *dip = ddi_find_devinfo("ramdisk", -1, 0); 21527c478bd9Sstevel@tonic-gate ASSERT(dip && ddi_get_parent(dip) == ddi_root_node()); 21537c478bd9Sstevel@tonic-gate ndi_rele_devi(dip); /* held from ddi_find_devinfo */ 21547c478bd9Sstevel@tonic-gate (void) ddi_remove_child(dip, 0); 21557c478bd9Sstevel@tonic-gate } 21567c478bd9Sstevel@tonic-gate 21577c478bd9Sstevel@tonic-gate PRM_POINT("Releasing boot pages"); 21587c478bd9Sstevel@tonic-gate while (bootpages) { 21597c478bd9Sstevel@tonic-gate pp = bootpages; 21607c478bd9Sstevel@tonic-gate bootpages = pp->p_next; 21617c478bd9Sstevel@tonic-gate if (root_is_ramdisk && pp_in_ramdisk(pp)) { 21627c478bd9Sstevel@tonic-gate pp->p_next = rd_pages; 21637c478bd9Sstevel@tonic-gate rd_pages = pp; 21647c478bd9Sstevel@tonic-gate continue; 21657c478bd9Sstevel@tonic-gate } 21667c478bd9Sstevel@tonic-gate pp->p_next = (struct page *)0; 2167ae115bc7Smrj pp->p_prev = (struct page *)0; 2168ae115bc7Smrj PP_CLRBOOTPAGES(pp); 21697c478bd9Sstevel@tonic-gate page_free(pp, 1); 21707c478bd9Sstevel@tonic-gate } 2171ae115bc7Smrj PRM_POINT("Boot pages released"); 21727c478bd9Sstevel@tonic-gate 2173843e1988Sjohnlev #if !defined(__xpv) 2174843e1988Sjohnlev /* XXPV -- note this following bunch of code needs to be revisited in Xen 3.0 */ 21757c478bd9Sstevel@tonic-gate /* 21767c478bd9Sstevel@tonic-gate * Find 1 page below 1 MB so that other processors can boot up. 21777c478bd9Sstevel@tonic-gate * Make sure it has a kernel VA as well as a 1:1 mapping. 21787c478bd9Sstevel@tonic-gate * We should have just free'd one up. 21797c478bd9Sstevel@tonic-gate */ 21807c478bd9Sstevel@tonic-gate if (use_mp) { 2181ae115bc7Smrj pfn_t pfn; 2182ae115bc7Smrj 21837c478bd9Sstevel@tonic-gate for (pfn = 1; pfn < btop(1*1024*1024); pfn++) { 21847c478bd9Sstevel@tonic-gate if (page_numtopp_alloc(pfn) == NULL) 21857c478bd9Sstevel@tonic-gate continue; 21867c478bd9Sstevel@tonic-gate rm_platter_va = i86devmap(pfn, 1, 21877c478bd9Sstevel@tonic-gate PROT_READ | PROT_WRITE | PROT_EXEC); 21887c478bd9Sstevel@tonic-gate rm_platter_pa = ptob(pfn); 21897c478bd9Sstevel@tonic-gate hat_devload(kas.a_hat, 21907c478bd9Sstevel@tonic-gate (caddr_t)(uintptr_t)rm_platter_pa, MMU_PAGESIZE, 21917c478bd9Sstevel@tonic-gate pfn, PROT_READ | PROT_WRITE | PROT_EXEC, 21927c478bd9Sstevel@tonic-gate HAT_LOAD_NOCONSIST); 21937c478bd9Sstevel@tonic-gate break; 21947c478bd9Sstevel@tonic-gate } 21957c478bd9Sstevel@tonic-gate if (pfn == btop(1*1024*1024)) 21967c478bd9Sstevel@tonic-gate panic("No page available for starting " 21977c478bd9Sstevel@tonic-gate "other processors"); 21987c478bd9Sstevel@tonic-gate } 2199843e1988Sjohnlev #endif /* !__xpv */ 22007c478bd9Sstevel@tonic-gate } 22017c478bd9Sstevel@tonic-gate 22027c478bd9Sstevel@tonic-gate /* 22037c478bd9Sstevel@tonic-gate * Initialize the platform-specific parts of a page_t. 22047c478bd9Sstevel@tonic-gate */ 22057c478bd9Sstevel@tonic-gate void 22067c478bd9Sstevel@tonic-gate add_physmem_cb(page_t *pp, pfn_t pnum) 22077c478bd9Sstevel@tonic-gate { 22087c478bd9Sstevel@tonic-gate pp->p_pagenum = pnum; 22097c478bd9Sstevel@tonic-gate pp->p_mapping = NULL; 22107c478bd9Sstevel@tonic-gate pp->p_embed = 0; 22117c478bd9Sstevel@tonic-gate pp->p_share = 0; 22127c478bd9Sstevel@tonic-gate pp->p_mlentry = 0; 22137c478bd9Sstevel@tonic-gate } 22147c478bd9Sstevel@tonic-gate 22157c478bd9Sstevel@tonic-gate /* 22167c478bd9Sstevel@tonic-gate * kphysm_init() initializes physical memory. 22177c478bd9Sstevel@tonic-gate */ 22187c478bd9Sstevel@tonic-gate static pgcnt_t 22197c478bd9Sstevel@tonic-gate kphysm_init( 2220ae115bc7Smrj page_t *pp, 22217c478bd9Sstevel@tonic-gate pgcnt_t npages) 22227c478bd9Sstevel@tonic-gate { 22237c478bd9Sstevel@tonic-gate struct memlist *pmem; 22247c478bd9Sstevel@tonic-gate struct memseg *cur_memseg; 22257c478bd9Sstevel@tonic-gate pfn_t base_pfn; 22267c478bd9Sstevel@tonic-gate pgcnt_t num; 22277c478bd9Sstevel@tonic-gate pgcnt_t pages_done = 0; 22287c478bd9Sstevel@tonic-gate uint64_t addr; 22297c478bd9Sstevel@tonic-gate uint64_t size; 22307c478bd9Sstevel@tonic-gate extern pfn_t ddiphysmin; 22317c478bd9Sstevel@tonic-gate 22327c478bd9Sstevel@tonic-gate ASSERT(page_hash != NULL && page_hashsz != 0); 22337c478bd9Sstevel@tonic-gate 2234ae115bc7Smrj cur_memseg = memseg_base; 22357c478bd9Sstevel@tonic-gate for (pmem = phys_avail; pmem && npages; pmem = pmem->next) { 22367c478bd9Sstevel@tonic-gate /* 22377c478bd9Sstevel@tonic-gate * In a 32 bit kernel can't use higher memory if we're 22387c478bd9Sstevel@tonic-gate * not booting in PAE mode. This check takes care of that. 22397c478bd9Sstevel@tonic-gate */ 22407c478bd9Sstevel@tonic-gate addr = pmem->address; 22417c478bd9Sstevel@tonic-gate size = pmem->size; 22427c478bd9Sstevel@tonic-gate if (btop(addr) > physmax) 22437c478bd9Sstevel@tonic-gate continue; 22447c478bd9Sstevel@tonic-gate 22457c478bd9Sstevel@tonic-gate /* 22467c478bd9Sstevel@tonic-gate * align addr and size - they may not be at page boundaries 22477c478bd9Sstevel@tonic-gate */ 22487c478bd9Sstevel@tonic-gate if ((addr & MMU_PAGEOFFSET) != 0) { 22497c478bd9Sstevel@tonic-gate addr += MMU_PAGEOFFSET; 22507c478bd9Sstevel@tonic-gate addr &= ~(uint64_t)MMU_PAGEOFFSET; 22517c478bd9Sstevel@tonic-gate size -= addr - pmem->address; 22527c478bd9Sstevel@tonic-gate } 22537c478bd9Sstevel@tonic-gate 22546bb54764Skchow /* only process pages below or equal to physmax */ 22556bb54764Skchow if ((btop(addr + size) - 1) > physmax) 22566bb54764Skchow size = ptob(physmax - btop(addr) + 1); 22577c478bd9Sstevel@tonic-gate 22587c478bd9Sstevel@tonic-gate num = btop(size); 22597c478bd9Sstevel@tonic-gate if (num == 0) 22607c478bd9Sstevel@tonic-gate continue; 22617c478bd9Sstevel@tonic-gate 22627c478bd9Sstevel@tonic-gate if (num > npages) 22637c478bd9Sstevel@tonic-gate num = npages; 22647c478bd9Sstevel@tonic-gate 22657c478bd9Sstevel@tonic-gate npages -= num; 22667c478bd9Sstevel@tonic-gate pages_done += num; 22677c478bd9Sstevel@tonic-gate base_pfn = btop(addr); 22687c478bd9Sstevel@tonic-gate 22697c478bd9Sstevel@tonic-gate if (prom_debug) 22707c478bd9Sstevel@tonic-gate prom_printf("MEMSEG addr=0x%" PRIx64 22717c478bd9Sstevel@tonic-gate " pgs=0x%lx pfn 0x%lx-0x%lx\n", 22727c478bd9Sstevel@tonic-gate addr, num, base_pfn, base_pfn + num); 22737c478bd9Sstevel@tonic-gate 22747c478bd9Sstevel@tonic-gate /* 2275ae115bc7Smrj * Ignore pages below ddiphysmin to simplify ddi memory 22767c478bd9Sstevel@tonic-gate * allocation with non-zero addr_lo requests. 22777c478bd9Sstevel@tonic-gate */ 22787c478bd9Sstevel@tonic-gate if (base_pfn < ddiphysmin) { 2279ae115bc7Smrj if (base_pfn + num <= ddiphysmin) 22807c478bd9Sstevel@tonic-gate continue; 22817c478bd9Sstevel@tonic-gate pp += (ddiphysmin - base_pfn); 22827c478bd9Sstevel@tonic-gate num -= (ddiphysmin - base_pfn); 22837c478bd9Sstevel@tonic-gate base_pfn = ddiphysmin; 22847c478bd9Sstevel@tonic-gate } 2285ae115bc7Smrj 22867c478bd9Sstevel@tonic-gate /* 22877c478bd9Sstevel@tonic-gate * Build the memsegs entry 22887c478bd9Sstevel@tonic-gate */ 22897c478bd9Sstevel@tonic-gate cur_memseg->pages = pp; 22907c478bd9Sstevel@tonic-gate cur_memseg->epages = pp + num; 22917c478bd9Sstevel@tonic-gate cur_memseg->pages_base = base_pfn; 22927c478bd9Sstevel@tonic-gate cur_memseg->pages_end = base_pfn + num; 22937c478bd9Sstevel@tonic-gate 22947c478bd9Sstevel@tonic-gate /* 2295ae115bc7Smrj * Insert into memseg list in decreasing pfn range order. 22967c478bd9Sstevel@tonic-gate * Low memory is typically more fragmented such that this 22977c478bd9Sstevel@tonic-gate * ordering keeps the larger ranges at the front of the list 22987c478bd9Sstevel@tonic-gate * for code that searches memseg. 2299ae115bc7Smrj * This ASSERTS that the memsegs coming in from boot are in 2300ae115bc7Smrj * increasing physical address order and not contiguous. 23017c478bd9Sstevel@tonic-gate */ 2302ae115bc7Smrj if (memsegs != NULL) { 2303ae115bc7Smrj ASSERT(cur_memseg->pages_base >= memsegs->pages_end); 2304ae115bc7Smrj cur_memseg->next = memsegs; 23057c478bd9Sstevel@tonic-gate } 2306ae115bc7Smrj memsegs = cur_memseg; 23077c478bd9Sstevel@tonic-gate 23087c478bd9Sstevel@tonic-gate /* 23097c478bd9Sstevel@tonic-gate * add_physmem() initializes the PSM part of the page 23107c478bd9Sstevel@tonic-gate * struct by calling the PSM back with add_physmem_cb(). 23117c478bd9Sstevel@tonic-gate * In addition it coalesces pages into larger pages as 23127c478bd9Sstevel@tonic-gate * it initializes them. 23137c478bd9Sstevel@tonic-gate */ 23147c478bd9Sstevel@tonic-gate add_physmem(pp, num, base_pfn); 23157c478bd9Sstevel@tonic-gate cur_memseg++; 23167c478bd9Sstevel@tonic-gate availrmem_initial += num; 23177c478bd9Sstevel@tonic-gate availrmem += num; 23187c478bd9Sstevel@tonic-gate 2319ae115bc7Smrj pp += num; 23207c478bd9Sstevel@tonic-gate } 23217c478bd9Sstevel@tonic-gate 23227c478bd9Sstevel@tonic-gate PRM_DEBUG(availrmem_initial); 23237c478bd9Sstevel@tonic-gate PRM_DEBUG(availrmem); 23247c478bd9Sstevel@tonic-gate PRM_DEBUG(freemem); 23257c478bd9Sstevel@tonic-gate build_pfn_hash(); 23267c478bd9Sstevel@tonic-gate return (pages_done); 23277c478bd9Sstevel@tonic-gate } 23287c478bd9Sstevel@tonic-gate 23297c478bd9Sstevel@tonic-gate /* 23307c478bd9Sstevel@tonic-gate * Kernel VM initialization. 23317c478bd9Sstevel@tonic-gate */ 23327c478bd9Sstevel@tonic-gate static void 23337c478bd9Sstevel@tonic-gate kvm_init(void) 23347c478bd9Sstevel@tonic-gate { 23357c478bd9Sstevel@tonic-gate ASSERT((((uintptr_t)s_text) & MMU_PAGEOFFSET) == 0); 23367c478bd9Sstevel@tonic-gate 23377c478bd9Sstevel@tonic-gate /* 23387c478bd9Sstevel@tonic-gate * Put the kernel segments in kernel address space. 23397c478bd9Sstevel@tonic-gate */ 23407c478bd9Sstevel@tonic-gate rw_enter(&kas.a_lock, RW_WRITER); 23417c478bd9Sstevel@tonic-gate as_avlinit(&kas); 23427c478bd9Sstevel@tonic-gate 23437c478bd9Sstevel@tonic-gate (void) seg_attach(&kas, s_text, e_moddata - s_text, &ktextseg); 23447c478bd9Sstevel@tonic-gate (void) segkmem_create(&ktextseg); 23457c478bd9Sstevel@tonic-gate 23467c478bd9Sstevel@tonic-gate (void) seg_attach(&kas, (caddr_t)valloc_base, valloc_sz, &kvalloc); 23477c478bd9Sstevel@tonic-gate (void) segkmem_create(&kvalloc); 23487c478bd9Sstevel@tonic-gate 234935b1ab99Sjosephb (void) seg_attach(&kas, kernelheap, 235035b1ab99Sjosephb ekernelheap - kernelheap, &kvseg); 23517c478bd9Sstevel@tonic-gate (void) segkmem_create(&kvseg); 23527c478bd9Sstevel@tonic-gate 2353ae115bc7Smrj if (core_size > 0) { 2354ae115bc7Smrj PRM_POINT("attaching kvseg_core"); 2355ae115bc7Smrj (void) seg_attach(&kas, (caddr_t)core_base, core_size, 2356ae115bc7Smrj &kvseg_core); 2357ae115bc7Smrj (void) segkmem_create(&kvseg_core); 2358ae115bc7Smrj } 2359ad23a2dbSjohansen 2360ae115bc7Smrj if (segziosize > 0) { 2361ae115bc7Smrj PRM_POINT("attaching segzio"); 2362ad23a2dbSjohansen (void) seg_attach(&kas, segzio_base, mmu_ptob(segziosize), 2363ad23a2dbSjohansen &kzioseg); 2364ad23a2dbSjohansen (void) segkmem_zio_create(&kzioseg); 2365ad23a2dbSjohansen 2366ad23a2dbSjohansen /* create zio area covering new segment */ 2367ad23a2dbSjohansen segkmem_zio_init(segzio_base, mmu_ptob(segziosize)); 2368ad23a2dbSjohansen } 23697c478bd9Sstevel@tonic-gate 2370ae115bc7Smrj (void) seg_attach(&kas, kdi_segdebugbase, kdi_segdebugsize, &kdebugseg); 23717c478bd9Sstevel@tonic-gate (void) segkmem_create(&kdebugseg); 23727c478bd9Sstevel@tonic-gate 23737c478bd9Sstevel@tonic-gate rw_exit(&kas.a_lock); 23747c478bd9Sstevel@tonic-gate 23757c478bd9Sstevel@tonic-gate /* 23767c478bd9Sstevel@tonic-gate * Ensure that the red zone at kernelbase is never accessible. 23777c478bd9Sstevel@tonic-gate */ 2378ae115bc7Smrj PRM_POINT("protecting redzone"); 23797c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, (caddr_t)kernelbase, KERNEL_REDZONE_SIZE, 0); 23807c478bd9Sstevel@tonic-gate 23817c478bd9Sstevel@tonic-gate /* 23827c478bd9Sstevel@tonic-gate * Make the text writable so that it can be hot patched by DTrace. 23837c478bd9Sstevel@tonic-gate */ 23847c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, s_text, e_modtext - s_text, 23857c478bd9Sstevel@tonic-gate PROT_READ | PROT_WRITE | PROT_EXEC); 23867c478bd9Sstevel@tonic-gate 23877c478bd9Sstevel@tonic-gate /* 23887c478bd9Sstevel@tonic-gate * Make data writable until end. 23897c478bd9Sstevel@tonic-gate */ 23907c478bd9Sstevel@tonic-gate (void) as_setprot(&kas, s_data, e_moddata - s_data, 23917c478bd9Sstevel@tonic-gate PROT_READ | PROT_WRITE | PROT_EXEC); 23927c478bd9Sstevel@tonic-gate } 23937c478bd9Sstevel@tonic-gate 2394843e1988Sjohnlev #ifndef __xpv 23957c478bd9Sstevel@tonic-gate /* 23961d03c31eSjohnlev * Solaris adds an entry for Write Combining caching to the PAT 23977c478bd9Sstevel@tonic-gate */ 23981d03c31eSjohnlev static uint64_t pat_attr_reg = PAT_DEFAULT_ATTRIBUTE; 23997c478bd9Sstevel@tonic-gate 24007c478bd9Sstevel@tonic-gate void 24011d03c31eSjohnlev pat_sync(void) 24027c478bd9Sstevel@tonic-gate { 24031d03c31eSjohnlev ulong_t cr0, cr0_orig, cr4; 24047c478bd9Sstevel@tonic-gate 24051d03c31eSjohnlev if (!(x86_feature & X86_PAT)) 24067c478bd9Sstevel@tonic-gate return; 24071d03c31eSjohnlev cr0_orig = cr0 = getcr0(); 24081d03c31eSjohnlev cr4 = getcr4(); 24097c478bd9Sstevel@tonic-gate 24101d03c31eSjohnlev /* disable caching and flush all caches and TLBs */ 24111d03c31eSjohnlev cr0 |= CR0_CD; 24121d03c31eSjohnlev cr0 &= ~CR0_NW; 24131d03c31eSjohnlev setcr0(cr0); 24141d03c31eSjohnlev invalidate_cache(); 24151d03c31eSjohnlev if (cr4 & CR4_PGE) { 24161d03c31eSjohnlev setcr4(cr4 & ~(ulong_t)CR4_PGE); 24171d03c31eSjohnlev setcr4(cr4); 24181d03c31eSjohnlev } else { 24191d03c31eSjohnlev reload_cr3(); 24207c478bd9Sstevel@tonic-gate } 24217c478bd9Sstevel@tonic-gate 24221d03c31eSjohnlev /* add our entry to the PAT */ 24231d03c31eSjohnlev wrmsr(REG_PAT, pat_attr_reg); 24247c478bd9Sstevel@tonic-gate 24251d03c31eSjohnlev /* flush TLBs and cache again, then reenable cr0 caching */ 24261d03c31eSjohnlev if (cr4 & CR4_PGE) { 24271d03c31eSjohnlev setcr4(cr4 & ~(ulong_t)CR4_PGE); 24281d03c31eSjohnlev setcr4(cr4); 24291d03c31eSjohnlev } else { 24301d03c31eSjohnlev reload_cr3(); 24317c478bd9Sstevel@tonic-gate } 24327c478bd9Sstevel@tonic-gate invalidate_cache(); 24337c478bd9Sstevel@tonic-gate setcr0(cr0_orig); 24347c478bd9Sstevel@tonic-gate } 24357c478bd9Sstevel@tonic-gate 24361d03c31eSjohnlev #endif /* !__xpv */ 24377c478bd9Sstevel@tonic-gate 24387c478bd9Sstevel@tonic-gate void 243945916cd2Sjpk get_system_configuration(void) 24407c478bd9Sstevel@tonic-gate { 24417c478bd9Sstevel@tonic-gate char prop[32]; 24427c478bd9Sstevel@tonic-gate u_longlong_t nodes_ll, cpus_pernode_ll, lvalue; 24437c478bd9Sstevel@tonic-gate 244435b1ab99Sjosephb if (BOP_GETPROPLEN(bootops, "nodes") > sizeof (prop) || 244535b1ab99Sjosephb BOP_GETPROP(bootops, "nodes", prop) < 0 || 244635b1ab99Sjosephb kobj_getvalue(prop, &nodes_ll) == -1 || 244735b1ab99Sjosephb nodes_ll > MAXNODES || 244835b1ab99Sjosephb BOP_GETPROPLEN(bootops, "cpus_pernode") > sizeof (prop) || 244935b1ab99Sjosephb BOP_GETPROP(bootops, "cpus_pernode", prop) < 0 || 245035b1ab99Sjosephb kobj_getvalue(prop, &cpus_pernode_ll) == -1) { 24517c478bd9Sstevel@tonic-gate system_hardware.hd_nodes = 1; 24527c478bd9Sstevel@tonic-gate system_hardware.hd_cpus_per_node = 0; 24537c478bd9Sstevel@tonic-gate } else { 24547c478bd9Sstevel@tonic-gate system_hardware.hd_nodes = (int)nodes_ll; 24557c478bd9Sstevel@tonic-gate system_hardware.hd_cpus_per_node = (int)cpus_pernode_ll; 24567c478bd9Sstevel@tonic-gate } 245735b1ab99Sjosephb 245835b1ab99Sjosephb if (BOP_GETPROPLEN(bootops, "kernelbase") > sizeof (prop) || 245935b1ab99Sjosephb BOP_GETPROP(bootops, "kernelbase", prop) < 0 || 246035b1ab99Sjosephb kobj_getvalue(prop, &lvalue) == -1) 246135b1ab99Sjosephb eprom_kernelbase = NULL; 24627c478bd9Sstevel@tonic-gate else 246335b1ab99Sjosephb eprom_kernelbase = (uintptr_t)lvalue; 24647c478bd9Sstevel@tonic-gate 246535b1ab99Sjosephb if (BOP_GETPROPLEN(bootops, "segmapsize") > sizeof (prop) || 246635b1ab99Sjosephb BOP_GETPROP(bootops, "segmapsize", prop) < 0 || 246735b1ab99Sjosephb kobj_getvalue(prop, &lvalue) == -1) 24687c478bd9Sstevel@tonic-gate segmapsize = SEGMAPDEFAULT; 246935b1ab99Sjosephb else 24707c478bd9Sstevel@tonic-gate segmapsize = (uintptr_t)lvalue; 24717c478bd9Sstevel@tonic-gate 247235b1ab99Sjosephb if (BOP_GETPROPLEN(bootops, "segmapfreelists") > sizeof (prop) || 247335b1ab99Sjosephb BOP_GETPROP(bootops, "segmapfreelists", prop) < 0 || 247435b1ab99Sjosephb kobj_getvalue(prop, &lvalue) == -1) 24757c478bd9Sstevel@tonic-gate segmapfreelists = 0; /* use segmap driver default */ 247635b1ab99Sjosephb else 24777c478bd9Sstevel@tonic-gate segmapfreelists = (int)lvalue; 24784944b02eSkchow 2479aac11643Sjosephb /* physmem used to be here, but moved much earlier to fakebop.c */ 24807c478bd9Sstevel@tonic-gate } 24817c478bd9Sstevel@tonic-gate 24827c478bd9Sstevel@tonic-gate /* 24837c478bd9Sstevel@tonic-gate * Add to a memory list. 24847c478bd9Sstevel@tonic-gate * start = start of new memory segment 24857c478bd9Sstevel@tonic-gate * len = length of new memory segment in bytes 24867c478bd9Sstevel@tonic-gate * new = pointer to a new struct memlist 24877c478bd9Sstevel@tonic-gate * memlistp = memory list to which to add segment. 24887c478bd9Sstevel@tonic-gate */ 2489ae115bc7Smrj void 24907c478bd9Sstevel@tonic-gate memlist_add( 24917c478bd9Sstevel@tonic-gate uint64_t start, 24927c478bd9Sstevel@tonic-gate uint64_t len, 24937c478bd9Sstevel@tonic-gate struct memlist *new, 24947c478bd9Sstevel@tonic-gate struct memlist **memlistp) 24957c478bd9Sstevel@tonic-gate { 24967c478bd9Sstevel@tonic-gate struct memlist *cur; 24977c478bd9Sstevel@tonic-gate uint64_t end = start + len; 24987c478bd9Sstevel@tonic-gate 24997c478bd9Sstevel@tonic-gate new->address = start; 25007c478bd9Sstevel@tonic-gate new->size = len; 25017c478bd9Sstevel@tonic-gate 25027c478bd9Sstevel@tonic-gate cur = *memlistp; 25037c478bd9Sstevel@tonic-gate 25047c478bd9Sstevel@tonic-gate while (cur) { 25057c478bd9Sstevel@tonic-gate if (cur->address >= end) { 25067c478bd9Sstevel@tonic-gate new->next = cur; 25077c478bd9Sstevel@tonic-gate *memlistp = new; 25087c478bd9Sstevel@tonic-gate new->prev = cur->prev; 25097c478bd9Sstevel@tonic-gate cur->prev = new; 25107c478bd9Sstevel@tonic-gate return; 25117c478bd9Sstevel@tonic-gate } 25127c478bd9Sstevel@tonic-gate ASSERT(cur->address + cur->size <= start); 25137c478bd9Sstevel@tonic-gate if (cur->next == NULL) { 25147c478bd9Sstevel@tonic-gate cur->next = new; 25157c478bd9Sstevel@tonic-gate new->prev = cur; 25167c478bd9Sstevel@tonic-gate new->next = NULL; 25177c478bd9Sstevel@tonic-gate return; 25187c478bd9Sstevel@tonic-gate } 25197c478bd9Sstevel@tonic-gate memlistp = &cur->next; 25207c478bd9Sstevel@tonic-gate cur = cur->next; 25217c478bd9Sstevel@tonic-gate } 25227c478bd9Sstevel@tonic-gate } 25237c478bd9Sstevel@tonic-gate 25247c478bd9Sstevel@tonic-gate void 25257c478bd9Sstevel@tonic-gate kobj_vmem_init(vmem_t **text_arena, vmem_t **data_arena) 25267c478bd9Sstevel@tonic-gate { 25277c478bd9Sstevel@tonic-gate size_t tsize = e_modtext - modtext; 25287c478bd9Sstevel@tonic-gate size_t dsize = e_moddata - moddata; 25297c478bd9Sstevel@tonic-gate 25307c478bd9Sstevel@tonic-gate *text_arena = vmem_create("module_text", tsize ? modtext : NULL, tsize, 25317c478bd9Sstevel@tonic-gate 1, segkmem_alloc, segkmem_free, heaptext_arena, 0, VM_SLEEP); 25327c478bd9Sstevel@tonic-gate *data_arena = vmem_create("module_data", dsize ? moddata : NULL, dsize, 25337c478bd9Sstevel@tonic-gate 1, segkmem_alloc, segkmem_free, heap32_arena, 0, VM_SLEEP); 25347c478bd9Sstevel@tonic-gate } 25357c478bd9Sstevel@tonic-gate 25367c478bd9Sstevel@tonic-gate caddr_t 25377c478bd9Sstevel@tonic-gate kobj_text_alloc(vmem_t *arena, size_t size) 25387c478bd9Sstevel@tonic-gate { 25397c478bd9Sstevel@tonic-gate return (vmem_alloc(arena, size, VM_SLEEP | VM_BESTFIT)); 25407c478bd9Sstevel@tonic-gate } 25417c478bd9Sstevel@tonic-gate 25427c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 25437c478bd9Sstevel@tonic-gate caddr_t 25447c478bd9Sstevel@tonic-gate kobj_texthole_alloc(caddr_t addr, size_t size) 25457c478bd9Sstevel@tonic-gate { 25467c478bd9Sstevel@tonic-gate panic("unexpected call to kobj_texthole_alloc()"); 25477c478bd9Sstevel@tonic-gate /*NOTREACHED*/ 25487c478bd9Sstevel@tonic-gate return (0); 25497c478bd9Sstevel@tonic-gate } 25507c478bd9Sstevel@tonic-gate 25517c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 25527c478bd9Sstevel@tonic-gate void 25537c478bd9Sstevel@tonic-gate kobj_texthole_free(caddr_t addr, size_t size) 25547c478bd9Sstevel@tonic-gate { 25557c478bd9Sstevel@tonic-gate panic("unexpected call to kobj_texthole_free()"); 25567c478bd9Sstevel@tonic-gate } 25577c478bd9Sstevel@tonic-gate 25587c478bd9Sstevel@tonic-gate /* 25597c478bd9Sstevel@tonic-gate * This is called just after configure() in startup(). 25607c478bd9Sstevel@tonic-gate * 25617c478bd9Sstevel@tonic-gate * The ISALIST concept is a bit hopeless on Intel, because 25627c478bd9Sstevel@tonic-gate * there's no guarantee of an ever-more-capable processor 25637c478bd9Sstevel@tonic-gate * given that various parts of the instruction set may appear 25647c478bd9Sstevel@tonic-gate * and disappear between different implementations. 25657c478bd9Sstevel@tonic-gate * 25667c478bd9Sstevel@tonic-gate * While it would be possible to correct it and even enhance 25677c478bd9Sstevel@tonic-gate * it somewhat, the explicit hardware capability bitmask allows 25687c478bd9Sstevel@tonic-gate * more flexibility. 25697c478bd9Sstevel@tonic-gate * 25707c478bd9Sstevel@tonic-gate * So, we just leave this alone. 25717c478bd9Sstevel@tonic-gate */ 25727c478bd9Sstevel@tonic-gate void 25737c478bd9Sstevel@tonic-gate setx86isalist(void) 25747c478bd9Sstevel@tonic-gate { 25757c478bd9Sstevel@tonic-gate char *tp; 25767c478bd9Sstevel@tonic-gate size_t len; 25777c478bd9Sstevel@tonic-gate extern char *isa_list; 25787c478bd9Sstevel@tonic-gate 25797c478bd9Sstevel@tonic-gate #define TBUFSIZE 1024 25807c478bd9Sstevel@tonic-gate 25817c478bd9Sstevel@tonic-gate tp = kmem_alloc(TBUFSIZE, KM_SLEEP); 25827c478bd9Sstevel@tonic-gate *tp = '\0'; 25837c478bd9Sstevel@tonic-gate 25847c478bd9Sstevel@tonic-gate #if defined(__amd64) 25857c478bd9Sstevel@tonic-gate (void) strcpy(tp, "amd64 "); 25867c478bd9Sstevel@tonic-gate #endif 25877c478bd9Sstevel@tonic-gate 25887c478bd9Sstevel@tonic-gate switch (x86_vendor) { 25897c478bd9Sstevel@tonic-gate case X86_VENDOR_Intel: 25907c478bd9Sstevel@tonic-gate case X86_VENDOR_AMD: 25917c478bd9Sstevel@tonic-gate case X86_VENDOR_TM: 25927c478bd9Sstevel@tonic-gate if (x86_feature & X86_CMOV) { 25937c478bd9Sstevel@tonic-gate /* 25947c478bd9Sstevel@tonic-gate * Pentium Pro or later 25957c478bd9Sstevel@tonic-gate */ 25967c478bd9Sstevel@tonic-gate (void) strcat(tp, "pentium_pro"); 25977c478bd9Sstevel@tonic-gate (void) strcat(tp, x86_feature & X86_MMX ? 25987c478bd9Sstevel@tonic-gate "+mmx pentium_pro " : " "); 25997c478bd9Sstevel@tonic-gate } 26007c478bd9Sstevel@tonic-gate /*FALLTHROUGH*/ 26017c478bd9Sstevel@tonic-gate case X86_VENDOR_Cyrix: 26027c478bd9Sstevel@tonic-gate /* 26037c478bd9Sstevel@tonic-gate * The Cyrix 6x86 does not have any Pentium features 26047c478bd9Sstevel@tonic-gate * accessible while not at privilege level 0. 26057c478bd9Sstevel@tonic-gate */ 26067c478bd9Sstevel@tonic-gate if (x86_feature & X86_CPUID) { 26077c478bd9Sstevel@tonic-gate (void) strcat(tp, "pentium"); 26087c478bd9Sstevel@tonic-gate (void) strcat(tp, x86_feature & X86_MMX ? 26097c478bd9Sstevel@tonic-gate "+mmx pentium " : " "); 26107c478bd9Sstevel@tonic-gate } 26117c478bd9Sstevel@tonic-gate break; 26127c478bd9Sstevel@tonic-gate default: 26137c478bd9Sstevel@tonic-gate break; 26147c478bd9Sstevel@tonic-gate } 26157c478bd9Sstevel@tonic-gate (void) strcat(tp, "i486 i386 i86"); 26167c478bd9Sstevel@tonic-gate len = strlen(tp) + 1; /* account for NULL at end of string */ 26177c478bd9Sstevel@tonic-gate isa_list = strcpy(kmem_alloc(len, KM_SLEEP), tp); 26187c478bd9Sstevel@tonic-gate kmem_free(tp, TBUFSIZE); 26197c478bd9Sstevel@tonic-gate 26207c478bd9Sstevel@tonic-gate #undef TBUFSIZE 26217c478bd9Sstevel@tonic-gate } 26227c478bd9Sstevel@tonic-gate 26237c478bd9Sstevel@tonic-gate 26247c478bd9Sstevel@tonic-gate #ifdef __amd64 26257c478bd9Sstevel@tonic-gate 26267c478bd9Sstevel@tonic-gate void * 26277c478bd9Sstevel@tonic-gate device_arena_alloc(size_t size, int vm_flag) 26287c478bd9Sstevel@tonic-gate { 26297c478bd9Sstevel@tonic-gate return (vmem_alloc(device_arena, size, vm_flag)); 26307c478bd9Sstevel@tonic-gate } 26317c478bd9Sstevel@tonic-gate 26327c478bd9Sstevel@tonic-gate void 26337c478bd9Sstevel@tonic-gate device_arena_free(void *vaddr, size_t size) 26347c478bd9Sstevel@tonic-gate { 26357c478bd9Sstevel@tonic-gate vmem_free(device_arena, vaddr, size); 26367c478bd9Sstevel@tonic-gate } 26377c478bd9Sstevel@tonic-gate 2638ae115bc7Smrj #else /* __i386 */ 26397c478bd9Sstevel@tonic-gate 26407c478bd9Sstevel@tonic-gate void * 26417c478bd9Sstevel@tonic-gate device_arena_alloc(size_t size, int vm_flag) 26427c478bd9Sstevel@tonic-gate { 26437c478bd9Sstevel@tonic-gate caddr_t vaddr; 26447c478bd9Sstevel@tonic-gate uintptr_t v; 26457c478bd9Sstevel@tonic-gate size_t start; 26467c478bd9Sstevel@tonic-gate size_t end; 26477c478bd9Sstevel@tonic-gate 26487c478bd9Sstevel@tonic-gate vaddr = vmem_alloc(heap_arena, size, vm_flag); 26497c478bd9Sstevel@tonic-gate if (vaddr == NULL) 26507c478bd9Sstevel@tonic-gate return (NULL); 26517c478bd9Sstevel@tonic-gate 26527c478bd9Sstevel@tonic-gate v = (uintptr_t)vaddr; 26537c478bd9Sstevel@tonic-gate ASSERT(v >= kernelbase); 2654ae115bc7Smrj ASSERT(v + size <= valloc_base); 26557c478bd9Sstevel@tonic-gate 26567c478bd9Sstevel@tonic-gate start = btop(v - kernelbase); 26577c478bd9Sstevel@tonic-gate end = btop(v + size - 1 - kernelbase); 26587c478bd9Sstevel@tonic-gate ASSERT(start < toxic_bit_map_len); 26597c478bd9Sstevel@tonic-gate ASSERT(end < toxic_bit_map_len); 26607c478bd9Sstevel@tonic-gate 26617c478bd9Sstevel@tonic-gate while (start <= end) { 26627c478bd9Sstevel@tonic-gate BT_ATOMIC_SET(toxic_bit_map, start); 26637c478bd9Sstevel@tonic-gate ++start; 26647c478bd9Sstevel@tonic-gate } 26657c478bd9Sstevel@tonic-gate return (vaddr); 26667c478bd9Sstevel@tonic-gate } 26677c478bd9Sstevel@tonic-gate 26687c478bd9Sstevel@tonic-gate void 26697c478bd9Sstevel@tonic-gate device_arena_free(void *vaddr, size_t size) 26707c478bd9Sstevel@tonic-gate { 26717c478bd9Sstevel@tonic-gate uintptr_t v = (uintptr_t)vaddr; 26727c478bd9Sstevel@tonic-gate size_t start; 26737c478bd9Sstevel@tonic-gate size_t end; 26747c478bd9Sstevel@tonic-gate 26757c478bd9Sstevel@tonic-gate ASSERT(v >= kernelbase); 2676ae115bc7Smrj ASSERT(v + size <= valloc_base); 26777c478bd9Sstevel@tonic-gate 26787c478bd9Sstevel@tonic-gate start = btop(v - kernelbase); 26797c478bd9Sstevel@tonic-gate end = btop(v + size - 1 - kernelbase); 26807c478bd9Sstevel@tonic-gate ASSERT(start < toxic_bit_map_len); 26817c478bd9Sstevel@tonic-gate ASSERT(end < toxic_bit_map_len); 26827c478bd9Sstevel@tonic-gate 26837c478bd9Sstevel@tonic-gate while (start <= end) { 26847c478bd9Sstevel@tonic-gate ASSERT(BT_TEST(toxic_bit_map, start) != 0); 26857c478bd9Sstevel@tonic-gate BT_ATOMIC_CLEAR(toxic_bit_map, start); 26867c478bd9Sstevel@tonic-gate ++start; 26877c478bd9Sstevel@tonic-gate } 26887c478bd9Sstevel@tonic-gate vmem_free(heap_arena, vaddr, size); 26897c478bd9Sstevel@tonic-gate } 26907c478bd9Sstevel@tonic-gate 26917c478bd9Sstevel@tonic-gate /* 26927c478bd9Sstevel@tonic-gate * returns 1st address in range that is in device arena, or NULL 26937c478bd9Sstevel@tonic-gate * if len is not NULL it returns the length of the toxic range 26947c478bd9Sstevel@tonic-gate */ 26957c478bd9Sstevel@tonic-gate void * 26967c478bd9Sstevel@tonic-gate device_arena_contains(void *vaddr, size_t size, size_t *len) 26977c478bd9Sstevel@tonic-gate { 26987c478bd9Sstevel@tonic-gate uintptr_t v = (uintptr_t)vaddr; 26997c478bd9Sstevel@tonic-gate uintptr_t eaddr = v + size; 27007c478bd9Sstevel@tonic-gate size_t start; 27017c478bd9Sstevel@tonic-gate size_t end; 27027c478bd9Sstevel@tonic-gate 27037c478bd9Sstevel@tonic-gate /* 27047c478bd9Sstevel@tonic-gate * if called very early by kmdb, just return NULL 27057c478bd9Sstevel@tonic-gate */ 27067c478bd9Sstevel@tonic-gate if (toxic_bit_map == NULL) 27077c478bd9Sstevel@tonic-gate return (NULL); 27087c478bd9Sstevel@tonic-gate 27097c478bd9Sstevel@tonic-gate /* 27107c478bd9Sstevel@tonic-gate * First check if we're completely outside the bitmap range. 27117c478bd9Sstevel@tonic-gate */ 2712ae115bc7Smrj if (v >= valloc_base || eaddr < kernelbase) 27137c478bd9Sstevel@tonic-gate return (NULL); 27147c478bd9Sstevel@tonic-gate 27157c478bd9Sstevel@tonic-gate /* 27167c478bd9Sstevel@tonic-gate * Trim ends of search to look at only what the bitmap covers. 27177c478bd9Sstevel@tonic-gate */ 27187c478bd9Sstevel@tonic-gate if (v < kernelbase) 27197c478bd9Sstevel@tonic-gate v = kernelbase; 27207c478bd9Sstevel@tonic-gate start = btop(v - kernelbase); 27217c478bd9Sstevel@tonic-gate end = btop(eaddr - kernelbase); 27227c478bd9Sstevel@tonic-gate if (end >= toxic_bit_map_len) 27237c478bd9Sstevel@tonic-gate end = toxic_bit_map_len; 27247c478bd9Sstevel@tonic-gate 27257c478bd9Sstevel@tonic-gate if (bt_range(toxic_bit_map, &start, &end, end) == 0) 27267c478bd9Sstevel@tonic-gate return (NULL); 27277c478bd9Sstevel@tonic-gate 27287c478bd9Sstevel@tonic-gate v = kernelbase + ptob(start); 27297c478bd9Sstevel@tonic-gate if (len != NULL) 27307c478bd9Sstevel@tonic-gate *len = ptob(end - start); 27317c478bd9Sstevel@tonic-gate return ((void *)v); 27327c478bd9Sstevel@tonic-gate } 27337c478bd9Sstevel@tonic-gate 2734ae115bc7Smrj #endif /* __i386 */ 2735