Index: head/sys/boot/i386/libi386/bootinfo64.c =================================================================== --- head/sys/boot/i386/libi386/bootinfo64.c (revision 277417) +++ head/sys/boot/i386/libi386/bootinfo64.c (revision 277418) @@ -1,262 +1,263 @@ /*- * Copyright (c) 1998 Michael Smith * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include "bootstrap.h" #include "libi386.h" #include "btxv86.h" /* * Copy module-related data into the load area, where it can be * used as a directory for loaded modules. * * Module data is presented in a self-describing format. Each datum * is preceded by a 32-bit identifier and a 32-bit size field. * * Currently, the following data are saved: * * MOD_NAME (variable) module name (string) * MOD_TYPE (variable) module type (string) * MOD_ARGS (variable) module parameters (string) * MOD_ADDR sizeof(vm_offset_t) module load address * MOD_SIZE sizeof(size_t) module size * MOD_METADATA (variable) type-specific metadata */ #define COPY32(v, a, c) { \ u_int32_t x = (v); \ if (c) \ i386_copyin(&x, a, sizeof(x)); \ a += sizeof(x); \ } #define MOD_STR(t, a, s, c) { \ COPY32(t, a, c); \ COPY32(strlen(s) + 1, a, c); \ if (c) \ i386_copyin(s, a, strlen(s) + 1); \ a += roundup(strlen(s) + 1, sizeof(u_int64_t));\ } #define MOD_NAME(a, s, c) MOD_STR(MODINFO_NAME, a, s, c) #define MOD_TYPE(a, s, c) MOD_STR(MODINFO_TYPE, a, s, c) #define MOD_ARGS(a, s, c) MOD_STR(MODINFO_ARGS, a, s, c) #define MOD_VAR(t, a, s, c) { \ COPY32(t, a, c); \ COPY32(sizeof(s), a, c); \ if (c) \ i386_copyin(&s, a, sizeof(s)); \ a += roundup(sizeof(s), sizeof(u_int64_t)); \ } #define MOD_ADDR(a, s, c) MOD_VAR(MODINFO_ADDR, a, s, c) #define MOD_SIZE(a, s, c) MOD_VAR(MODINFO_SIZE, a, s, c) #define MOD_METADATA(a, mm, c) { \ COPY32(MODINFO_METADATA | mm->md_type, a, c); \ COPY32(mm->md_size, a, c); \ if (c) \ i386_copyin(mm->md_data, a, mm->md_size); \ a += roundup(mm->md_size, sizeof(u_int64_t));\ } #define MOD_END(a, c) { \ COPY32(MODINFO_END, a, c); \ COPY32(0, a, c); \ } static vm_offset_t bi_copymodules64(vm_offset_t addr) { struct preloaded_file *fp; struct file_metadata *md; int c; u_int64_t v; c = addr != 0; /* start with the first module on the list, should be the kernel */ for (fp = file_findfile(NULL, NULL); fp != NULL; fp = fp->f_next) { MOD_NAME(addr, fp->f_name, c); /* this field must come first */ MOD_TYPE(addr, fp->f_type, c); if (fp->f_args) MOD_ARGS(addr, fp->f_args, c); v = fp->f_addr; MOD_ADDR(addr, v, c); v = fp->f_size; MOD_SIZE(addr, v, c); for (md = fp->f_metadata; md != NULL; md = md->md_next) if (!(md->md_type & MODINFOMD_NOCOPY)) MOD_METADATA(addr, md, c); } MOD_END(addr, c); return(addr); } /* * Check to see if this CPU supports long mode. */ static int bi_checkcpu(void) { char *cpu_vendor; int vendor[3]; int eflags; unsigned int regs[4]; /* Check for presence of "cpuid". */ eflags = read_eflags(); write_eflags(eflags ^ PSL_ID); if (!((eflags ^ read_eflags()) & PSL_ID)) return (0); /* Fetch the vendor string. */ do_cpuid(0, regs); vendor[0] = regs[1]; vendor[1] = regs[3]; vendor[2] = regs[2]; cpu_vendor = (char *)vendor; /* Check for vendors that support AMD features. */ if (strncmp(cpu_vendor, INTEL_VENDOR_ID, 12) != 0 && strncmp(cpu_vendor, AMD_VENDOR_ID, 12) != 0 && strncmp(cpu_vendor, CENTAUR_VENDOR_ID, 12) != 0) return (0); /* Has to support AMD features. */ do_cpuid(0x80000000, regs); if (!(regs[0] >= 0x80000001)) return (0); /* Check for long mode. */ do_cpuid(0x80000001, regs); return (regs[3] & AMDID_LM); } /* * Load the information expected by an amd64 kernel. * * - The 'boothowto' argument is constructed * - The 'bootdev' argument is constructed * - The 'bootinfo' struct is constructed, and copied into the kernel space. * - The kernel environment is copied into kernel space. * - Module metadata are formatted and placed in kernel space. */ int bi_load64(char *args, vm_offset_t addr, vm_offset_t *modulep, vm_offset_t *kernendp, int add_smap) { struct preloaded_file *xp, *kfp; struct i386_devdesc *rootdev; struct file_metadata *md; u_int64_t kernend; u_int64_t envp; + u_int64_t module; vm_offset_t size; char *rootdevname; int howto; if (!bi_checkcpu()) { printf("CPU doesn't support long mode\n"); return (EINVAL); } howto = bi_getboothowto(args); /* * Allow the environment variable 'rootdev' to override the supplied device * This should perhaps go to MI code and/or have $rootdev tested/set by * MI code before launching the kernel. */ rootdevname = getenv("rootdev"); i386_getdev((void **)(&rootdev), rootdevname, NULL); if (rootdev == NULL) { /* bad $rootdev/$currdev */ printf("can't determine root device\n"); return(EINVAL); } /* Try reading the /etc/fstab file to select the root device */ getrootmount(i386_fmtdev((void *)rootdev)); if (addr == 0) { /* find the last module in the chain */ for (xp = file_findfile(NULL, NULL); xp != NULL; xp = xp->f_next) { if (addr < (xp->f_addr + xp->f_size)) addr = xp->f_addr + xp->f_size; } } /* pad to a page boundary */ addr = roundup(addr, PAGE_SIZE); /* place the metadata before anything */ - *modulep = addr; + module = *modulep = addr; kfp = file_findfile(NULL, "elf kernel"); if (kfp == NULL) kfp = file_findfile(NULL, "elf64 kernel"); if (kfp == NULL) panic("can't find kernel file"); kernend = 0; /* fill it in later */ file_addmetadata(kfp, MODINFOMD_HOWTO, sizeof howto, &howto); file_addmetadata(kfp, MODINFOMD_ENVP, sizeof envp, &envp); file_addmetadata(kfp, MODINFOMD_KERNEND, sizeof kernend, &kernend); - file_addmetadata(kfp, MODINFOMD_MODULEP, sizeof modulep, modulep); + file_addmetadata(kfp, MODINFOMD_MODULEP, sizeof module, &module); if (add_smap != 0) bios_addsmapdata(kfp); size = bi_copymodules64(0); /* copy our environment */ envp = roundup(addr + size, PAGE_SIZE); addr = bi_copyenv(envp); /* set kernend */ kernend = roundup(addr, PAGE_SIZE); *kernendp = kernend; /* patch MODINFOMD_KERNEND */ md = file_findmetadata(kfp, MODINFOMD_KERNEND); bcopy(&kernend, md->md_data, sizeof kernend); /* patch MODINFOMD_ENVP */ md = file_findmetadata(kfp, MODINFOMD_ENVP); bcopy(&envp, md->md_data, sizeof envp); /* copy module list and metadata */ (void)bi_copymodules64(*modulep); return(0); } Index: head/sys/x86/xen/pv.c =================================================================== --- head/sys/x86/xen/pv.c (revision 277417) +++ head/sys/x86/xen/pv.c (revision 277418) @@ -1,442 +1,442 @@ /* * Copyright (c) 2004 Christian Limpach. * Copyright (c) 2004-2006,2008 Kip Macy * Copyright (c) 2008 The NetBSD Foundation, Inc. * Copyright (c) 2013 Roger Pau Monné * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif /* Native initial function */ extern u_int64_t hammer_time(u_int64_t, u_int64_t); /* Xen initial function */ uint64_t hammer_time_xen(start_info_t *, uint64_t); #define MAX_E820_ENTRIES 128 /*--------------------------- Forward Declarations ---------------------------*/ static caddr_t xen_pv_parse_preload_data(u_int64_t); static void xen_pv_parse_memmap(caddr_t, vm_paddr_t *, int *); #ifdef SMP static int xen_pv_start_all_aps(void); #endif /*---------------------------- Extern Declarations ---------------------------*/ #ifdef SMP /* Variables used by amd64 mp_machdep to start APs */ extern struct mtx ap_boot_mtx; extern void *bootstacks[]; extern char *doublefault_stack; extern char *nmi_stack; extern void *dpcpu; extern int bootAP; extern char *bootSTK; #endif /* * Placed by the linker at the end of the bss section, which is the last * section loaded by Xen before loading the symtab and strtab. */ extern uint32_t end; /*-------------------------------- Global Data -------------------------------*/ /* Xen init_ops implementation. */ struct init_ops xen_init_ops = { .parse_preload_data = xen_pv_parse_preload_data, .early_clock_source_init = xen_clock_init, .early_delay = xen_delay, .parse_memmap = xen_pv_parse_memmap, #ifdef SMP .start_all_aps = xen_pv_start_all_aps, #endif .msi_init = xen_msi_init, }; static struct bios_smap xen_smap[MAX_E820_ENTRIES]; /*-------------------------------- Xen PV init -------------------------------*/ /* * First function called by the Xen PVH boot sequence. * * Set some Xen global variables and prepare the environment so it is * as similar as possible to what native FreeBSD init function expects. */ uint64_t hammer_time_xen(start_info_t *si, uint64_t xenstack) { uint64_t physfree; uint64_t *PT4 = (u_int64_t *)xenstack; uint64_t *PT3 = (u_int64_t *)(xenstack + PAGE_SIZE); uint64_t *PT2 = (u_int64_t *)(xenstack + 2 * PAGE_SIZE); int i; xen_domain_type = XEN_PV_DOMAIN; vm_guest = VM_GUEST_XEN; if ((si == NULL) || (xenstack == 0)) { xc_printf("ERROR: invalid start_info or xen stack, halting\n"); HYPERVISOR_shutdown(SHUTDOWN_crash); } xc_printf("FreeBSD PVH running on %s\n", si->magic); /* We use 3 pages of xen stack for the boot pagetables */ physfree = xenstack + 3 * PAGE_SIZE - KERNBASE; /* Setup Xen global variables */ HYPERVISOR_start_info = si; HYPERVISOR_shared_info = (shared_info_t *)(si->shared_info + KERNBASE); /* * Setup some misc global variables for Xen devices * * XXX: Devices that need these specific variables should * be rewritten to fetch this info by themselves from the * start_info page. */ xen_store = (struct xenstore_domain_interface *) (ptoa(si->store_mfn) + KERNBASE); console_page = (char *)(ptoa(si->console.domU.mfn) + KERNBASE); /* * Use the stack Xen gives us to build the page tables * as native FreeBSD expects to find them (created * by the boot trampoline). */ for (i = 0; i < (PAGE_SIZE / sizeof(uint64_t)); i++) { /* * Each slot of the level 4 pages points * to the same level 3 page */ PT4[i] = ((uint64_t)&PT3[0]) - KERNBASE; PT4[i] |= PG_V | PG_RW | PG_U; /* * Each slot of the level 3 pages points * to the same level 2 page */ PT3[i] = ((uint64_t)&PT2[0]) - KERNBASE; PT3[i] |= PG_V | PG_RW | PG_U; /* * The level 2 page slots are mapped with * 2MB pages for 1GB. */ PT2[i] = i * (2 * 1024 * 1024); PT2[i] |= PG_V | PG_RW | PG_PS | PG_U; } load_cr3(((uint64_t)&PT4[0]) - KERNBASE); /* Set the hooks for early functions that diverge from bare metal */ init_ops = xen_init_ops; apic_ops = xen_apic_ops; /* Now we can jump into the native init function */ return (hammer_time(0, physfree)); } /*-------------------------------- PV specific -------------------------------*/ #ifdef SMP static bool start_xen_ap(int cpu) { struct vcpu_guest_context *ctxt; int ms, cpus = mp_naps; const size_t stacksize = KSTACK_PAGES * PAGE_SIZE; /* allocate and set up an idle stack data page */ bootstacks[cpu] = (void *)kmem_malloc(kernel_arena, stacksize, M_WAITOK | M_ZERO); doublefault_stack = (char *)kmem_malloc(kernel_arena, PAGE_SIZE, M_WAITOK | M_ZERO); nmi_stack = (char *)kmem_malloc(kernel_arena, PAGE_SIZE, M_WAITOK | M_ZERO); dpcpu = (void *)kmem_malloc(kernel_arena, DPCPU_SIZE, M_WAITOK | M_ZERO); bootSTK = (char *)bootstacks[cpu] + KSTACK_PAGES * PAGE_SIZE - 8; bootAP = cpu; ctxt = malloc(sizeof(*ctxt), M_TEMP, M_WAITOK | M_ZERO); if (ctxt == NULL) panic("unable to allocate memory"); ctxt->flags = VGCF_IN_KERNEL; ctxt->user_regs.rip = (unsigned long) init_secondary; ctxt->user_regs.rsp = (unsigned long) bootSTK; /* Set the AP to use the same page tables */ ctxt->ctrlreg[3] = KPML4phys; if (HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, ctxt)) panic("unable to initialize AP#%d", cpu); free(ctxt, M_TEMP); /* Launch the vCPU */ if (HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL)) panic("unable to start AP#%d", cpu); /* Wait up to 5 seconds for it to start. */ for (ms = 0; ms < 5000; ms++) { if (mp_naps > cpus) return (true); DELAY(1000); } return (false); } static int xen_pv_start_all_aps(void) { int cpu; mtx_init(&ap_boot_mtx, "ap boot", NULL, MTX_SPIN); for (cpu = 1; cpu < mp_ncpus; cpu++) { /* attempt to start the Application Processor */ if (!start_xen_ap(cpu)) panic("AP #%d failed to start!", cpu); CPU_SET(cpu, &all_cpus); /* record AP in CPU map */ } return (mp_naps); } #endif /* SMP */ /* * Functions to convert the "extra" parameters passed by Xen * into FreeBSD boot options. */ static void xen_pv_set_env(void) { char *cmd_line_next, *cmd_line; size_t env_size; cmd_line = HYPERVISOR_start_info->cmd_line; env_size = sizeof(HYPERVISOR_start_info->cmd_line); /* Skip leading spaces */ for (; isspace(*cmd_line) && (env_size != 0); cmd_line++) env_size--; /* Replace ',' with '\0' */ for (cmd_line_next = cmd_line; strsep(&cmd_line_next, ",") != NULL;) ; init_static_kenv(cmd_line, env_size); } static void xen_pv_set_boothowto(void) { int i; char *env; /* get equivalents from the environment */ for (i = 0; howto_names[i].ev != NULL; i++) { if ((env = kern_getenv(howto_names[i].ev)) != NULL) { boothowto |= howto_names[i].mask; freeenv(env); } } } #ifdef DDB /* * The way Xen loads the symtab is different from the native boot loader, * because it's tailored for NetBSD. So we have to adapt and use the same * method as NetBSD. Portions of the code below have been picked from NetBSD: * sys/kern/kern_ksyms.c CVS Revision 1.71. */ static void xen_pv_parse_symtab(void) { Elf_Ehdr *ehdr; Elf_Shdr *shdr; vm_offset_t sym_end; uint32_t size; int i, j; size = end; sym_end = HYPERVISOR_start_info->mod_start != 0 ? HYPERVISOR_start_info->mod_start : HYPERVISOR_start_info->mfn_list; /* * Make sure the size is right headed, sym_end is just a * high boundary, but at least allows us to fail earlier. */ if ((vm_offset_t)&end + size > sym_end) { xc_printf("Unable to load ELF symtab: size mismatch\n"); return; } ehdr = (Elf_Ehdr *)(&end + 1); if (memcmp(ehdr->e_ident, ELFMAG, SELFMAG) || ehdr->e_ident[EI_CLASS] != ELF_TARG_CLASS || ehdr->e_version > 1) { xc_printf("Unable to load ELF symtab: invalid symbol table\n"); return; } shdr = (Elf_Shdr *)((uint8_t *)ehdr + ehdr->e_shoff); /* Find the symbol table and the corresponding string table. */ for (i = 1; i < ehdr->e_shnum; i++) { if (shdr[i].sh_type != SHT_SYMTAB) continue; if (shdr[i].sh_offset == 0) continue; ksymtab = (uintptr_t)((uint8_t *)ehdr + shdr[i].sh_offset); ksymtab_size = shdr[i].sh_size; j = shdr[i].sh_link; if (shdr[j].sh_offset == 0) continue; /* Can this happen? */ kstrtab = (uintptr_t)((uint8_t *)ehdr + shdr[j].sh_offset); break; } if (ksymtab == 0 || kstrtab == 0) { xc_printf( "Unable to load ELF symtab: could not find symtab or strtab\n"); return; } } #endif static caddr_t xen_pv_parse_preload_data(u_int64_t modulep) { caddr_t kmdp; vm_ooffset_t off; vm_paddr_t metadata; if (HYPERVISOR_start_info->mod_start != 0) { preload_metadata = (caddr_t)(HYPERVISOR_start_info->mod_start); kmdp = preload_search_by_type("elf kernel"); if (kmdp == NULL) kmdp = preload_search_by_type("elf64 kernel"); KASSERT(kmdp != NULL, ("unable to find kernel")); /* * Xen has relocated the metadata and the modules, * so we need to recalculate it's position. This is * done by saving the original modulep address and * then calculating the offset with mod_start, * which contains the relocated modulep address. */ - metadata = MD_FETCH(kmdp, MODINFOMD_MODULEP, int); + metadata = MD_FETCH(kmdp, MODINFOMD_MODULEP, vm_paddr_t); off = HYPERVISOR_start_info->mod_start - metadata; preload_bootstrap_relocate(off); boothowto = MD_FETCH(kmdp, MODINFOMD_HOWTO, int); kern_envp = MD_FETCH(kmdp, MODINFOMD_ENVP, char *); kern_envp += off; } else { /* Parse the extra boot information given by Xen */ xen_pv_set_env(); xen_pv_set_boothowto(); kmdp = NULL; } #ifdef DDB xen_pv_parse_symtab(); #endif return (kmdp); } static void xen_pv_parse_memmap(caddr_t kmdp, vm_paddr_t *physmap, int *physmap_idx) { struct xen_memory_map memmap; u_int32_t size; int rc; /* Fetch the E820 map from Xen */ memmap.nr_entries = MAX_E820_ENTRIES; set_xen_guest_handle(memmap.buffer, xen_smap); rc = HYPERVISOR_memory_op(XENMEM_memory_map, &memmap); if (rc) panic("unable to fetch Xen E820 memory map"); size = memmap.nr_entries * sizeof(xen_smap[0]); bios_add_smap_entries(xen_smap, size, physmap, physmap_idx); }