sun4v/promif/promif_emul.c

	promif_emul.c revision 0a9b4b43b38c9b7673e8233a30d5b4bb9b75fe83
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

#include <sys/promif_impl.h>
#include <sys/machsystm.h>
#include <sys/lpad.h>
#include <sys/vmsystm.h>
#include <sys/prom_plat.h>
#include <sys/ldoms.h>
#include <sys/kobj.h>
#include <sys/reboot.h>
#include <sys/hypervisor_api.h>
#include <sys/mdesc.h>
#include <sys/mach_descrip.h>
#include <sys/cpu_module.h>

#ifndef _KMDB
#include <sys/pte.h>
#include <vm/hat_sfmmu.h>
#include <sys/memlist_impl.h>

static processorid_t cif_cpu;
static struct translation *cif_prom_trans;
static size_t cif_prom_ntrans;

int cif_cpu_mp_ready;
int (*prom_cif_handler)(void *) = NULL;

extern struct memlist *phys_avail;
extern struct vnode prom_ppages;
extern void kdi_tlb_page_unlock(caddr_t, int);

#define COMBINE(hi, lo) (((uint64_t)(uint32_t)(hi) << 32) | (uint32_t)(lo))
#define OFW_PT_START_ADDR   0xfffffffc00000000  /* OBP PT start */
#define OFW_PT_END_ADDR     0xffffffffffffffff  /* OBP PT end */

#define PROM_ADDR(a)    (((a) >= OFW_START_ADDR && (a) <= OFW_END_ADDR) || \
            ((a) >= OFW_PT_START_ADDR && (a) <= OFW_PT_END_ADDR))
#endif

#ifdef DEBUG
uint_t cif_debug;
int prom_free_debug;
#define PMFREE_DEBUG(args...) if (prom_free_debug) printf(args)
#else
#define PMFREE_DEBUG(args...)
#endif

extern int (*cif_handler)(void *);

typedef struct {
    char        *name;
    cif_func_t  func;
} cif_callback_t;

static cif_callback_t cb_table[] = {
    { "getprop",            promif_getprop          },
    { "getproplen",         promif_getproplen       },
    { "nextprop",           promif_nextprop         },
    { "peer",           promif_nextnode         },
    { "child",          promif_childnode        },
    { "parent",         promif_parentnode       },
    { "enter",          promif_enter_mon        },
    { "exit",           promif_exit_to_mon      },
    { "boot",           promif_reboot           },
    { "write",          promif_write            },
    { "read",           promif_read         },
    { "interpret",          promif_interpret        },
    { "finddevice",         promif_finddevice       },
    { "instance-to-package",    promif_instance_to_package  },
#ifndef _KMDB
    { "setprop",            promif_setprop          },
    { "test",           promif_test         },
    { "instance-to-path",       promif_instance_to_path     },
    { "SUNW,power-off",     promif_power_off        },
    { "SUNW,asr-list-keys-len", promif_asr_list_keys_len    },
    { "SUNW,asr-list-keys",     promif_asr_list_keys        },
    { "SUNW,asr-export-len",    promif_asr_export_len       },
    { "SUNW,asr-export",        promif_asr_export       },
    { "SUNW,set-security-key",  promif_set_security_key     },
    { "SUNW,get-security-key",  promif_get_security_key     },
    { "SUNW,start-cpu-by-cpuid",    promif_start_cpu        },
    { "SUNW,set-trap-table",    promif_set_mmfsa_traptable  },
    { "SUNW,set-sun4v-api-version", promif_set_sun4v_api_version },
    { "SUNW,get-sun4v-api-version", promif_get_sun4v_api_version },
#endif
    { NULL,             NULL                }
};

cif_func_t
promif_find_cif_callback(char *opname)
{
    cif_callback_t  *cb;

    if (opname == NULL)
        return (NULL);

    for (cb = cb_table; cb->name; cb++) {
        if (prom_strcmp(cb->name, opname) == 0)
            break;
    }

    return (cb->func);
}

static int
kern_cif_handler(void *p)
{
    cell_t      *ci = (cell_t *)p;
    char        *opname;
    cif_func_t  func;
    int     rv;

    ASSERT(cif_handler == kern_cif_handler);

#ifndef _KMDB
    cif_cpu = getprocessorid();
#endif

    opname = p1275_cell2ptr(ci[0]);

    /* lookup the callback for the desired operation */
    func = promif_find_cif_callback(opname);

    if (func == NULL) {
#ifdef _KMDB
        prom_fatal_error("sun4v unsupported CIFs\n");
#else
        cmn_err(CE_CONT, "!sun4v unsupported CIF: %s\n", opname);
        return (-1);
#endif
    }

    /* callback found, execute it */
    rv = func(p);

#ifndef _KMDB
    cif_cpu = -1;
#endif

    return (rv);
}

#ifdef _KMDB

void
cif_init(char *pgmname, caddr_t root, ihandle_t in, ihandle_t out,
    phandle_t pin, phandle_t pout, pnode_t chosen, pnode_t options)
{
    /* initialize pointer to a copy of OBP device tree */
    promif_stree_setroot(root);

    promif_set_nodes(chosen, options);

    /* initialize io parameters */
    promif_io_init(in, out, pin, pout);

    /*
     * Switch CIF handler to the kernel.
     */
    if (pgmname != NULL)
        prom_init(pgmname, (void *)kern_cif_handler);
    else
        cif_handler = kern_cif_handler;
}

#else

static struct translation *
read_prom_mappings(size_t *ntransp)
{
    char *prop = "translations";
    pnode_t node;
    size_t translen;
    ihandle_t immu;
    struct translation *transroot;

    *ntransp = 0;

    /*
     * the "translations" property is associated with the mmu node
     */
    if ((immu = prom_mmu_ihandle()) == (ihandle_t)-1) {
        PMFREE_DEBUG("no mmu ihandle");
        return (NULL);
    }
    node = (pnode_t)prom_getphandle(immu);
    if (node == OBP_NONODE || node == OBP_BADNODE) {
        PMFREE_DEBUG("no mmu node");
        return (NULL);
    }

    if ((translen = prom_getproplen(node, prop)) == -1) {
        PMFREE_DEBUG("no translations property");
        return (NULL);
    }
    transroot = (struct translation *)kmem_zalloc(translen, KM_SLEEP);

    if (prom_getprop(node, prop, (caddr_t)transroot) == -1) {
        PMFREE_DEBUG("translations getprop failed");
        kmem_free(transroot, translen);
        return (NULL);
    }
    *ntransp = translen / sizeof (*transroot);

    return (transroot);
}

static void
unmap_prom_mappings(struct translation *transroot, size_t ntransroot)
{
    int i, j, rv;
    int npgs, nunmapped, nfreed, nskipped, nskipped_io;
    char *p;
    tte_t tte;
    pfn_t pfn;
    page_t *pp;
    uint64_t vaddr;
    struct translation *promt;
    cpuset_t other_cpus;

    /*
     * During startup isa_list is allocated in OBP address space
     * so it needs to be re-allocated in kernel address space
     * before OBP memory is unmapped.
     *
     * see cpu_setup_common().
     */
    p = kmem_zalloc(strlen(isa_list) + 1, KM_SLEEP);
    (void) strcpy(p, isa_list);
    isa_list = p;

    nfreed = 0;
    nunmapped = 0;
    nskipped = 0;
    nskipped_io = 0;

    for (i = 0, promt = transroot; i < ntransroot; i++, promt++) {
        ASSERT(promt->tte_hi != 0);
        ASSERT32(promt->virt_hi == 0 && promt->size_hi == 0);

        vaddr = COMBINE(promt->virt_hi, promt->virt_lo);

        if (!PROM_ADDR(vaddr)) {
            nskipped++;
            continue;
        }

        npgs = mmu_btopr(COMBINE(promt->size_hi, promt->size_lo));

        if (npgs > 1) {
            PMFREE_DEBUG("large trans vaddr=0x%lx, npgs=%d\n",
                vaddr, npgs);
        }
        for (j = 0; j < npgs; j++) {

            pfn = sfmmu_vatopfn((caddr_t)vaddr, KHATID, &tte);

            if (pfn == PFN_INVALID) {
                tte.tte_inthi = promt->tte_hi;
                tte.tte_intlo = promt->tte_lo;
                pfn = TTE_TO_PFN((caddr_t)COMBINE(
                    promt->virt_hi, promt->virt_lo), &tte);
                PMFREE_DEBUG(
                    "no mapping for vaddr=0x%lx (opfn=0x%lx)\n",
                    vaddr, pfn);
                break;
            }
            ASSERT(!TTE_IS_LOCKED(&tte));
            ASSERT(TTE_IS_8K(&tte));

            /*
             * Unload the current mapping for the pfn and
             * if it is the last mapping for a memory page,
             * free the page.
             */
            PMFREE_DEBUG("unmap vaddr=0x%lx pfn=0x%lx", vaddr, pfn);

            hat_unload(kas.a_hat, (caddr_t)vaddr, PAGESIZE,
                HAT_UNLOAD_UNLOCK);

            if (pf_is_memory(pfn)) {
                pp = page_numtopp_nolock(pfn);
                PMFREE_DEBUG(" pp=0x%p", (void *)pp);
                ASSERT(pp);
                ASSERT(PAGE_EXCL(pp));
                ASSERT(PP_ISNORELOC(pp));
                ASSERT(!PP_ISFREE(pp));
                ASSERT(page_find(&prom_ppages, pfn));
                ASSERT(page_get_pagecnt(pp->p_szc) == 1);

                if (pp->p_mapping) {
                    PMFREE_DEBUG(" skip\n");
                } else {
                    PP_CLRNORELOC(pp);
                    page_destroy(pp, 0);
                    memlist_write_lock();
                    rv = memlist_add_span(pfn << PAGESHIFT,
                        PAGESIZE, &phys_avail);
                    ASSERT(rv == MEML_SPANOP_OK);
                    memlist_write_unlock();
                    PMFREE_DEBUG(" free\n");
                    nfreed++;
                }
            } else {
                nskipped_io++;
                PMFREE_DEBUG(" skip IO\n");
            }
            nunmapped++;
            vaddr += PAGESIZE;
        }
    }

    if (transroot) {
        PMFREE_DEBUG(
            "nunmapped=%d nfreed=%d nskipped=%d nskipped_io=%d\n",
            nunmapped, nfreed, nskipped, nskipped_io);
        kmem_free(transroot, ntransroot * sizeof (*transroot));
    }

    /*
     * Unload OBP permanent mappings.
     */
    kdi_tlb_page_unlock((caddr_t)OFW_START_ADDR, 1);
    kpreempt_disable();
    other_cpus = cpu_ready_set;
    CPUSET_DEL(other_cpus, CPU->cpu_id);
    xt_some(other_cpus, vtag_unmap_perm_tl1, (uint64_t)OFW_START_ADDR,
        KCONTEXT);
    kpreempt_enable();
}

static void cache_prom_data(void);

/*
 * This function returns 1 if the current thread is executing in
 * the CIF and 0 otherwise. This is useful information to know
 * since code that implements CIF handlers can assume that it has
 * gone through the kern_preprom() entry point, implying it is
 * running single threaded, has preemption disabled, etc.
 */
int
promif_in_cif(void)
{
    int mycpuid = getprocessorid();

    return ((cif_cpu == mycpuid) ? 1 : 0);
}

/*
 * Check that all cpus in the MD are within range (< NCPU).  Attempt
 * to stop any that aren't.
 */
static void
cif_check_cpus(void)
{
    md_t        *mdp;
    mde_cookie_t    rootnode;
    size_t      listsz;
    int     i;
    mde_cookie_t    *listp = NULL;
    int     num_nodes;
    uint64_t    cpuid;
    int     status;

    mdp = md_get_handle();
    ASSERT(mdp);

    rootnode = md_root_node(mdp);
    ASSERT(rootnode != MDE_INVAL_ELEM_COOKIE);

    num_nodes = md_node_count(mdp);
    ASSERT(num_nodes > 0);

    listsz = num_nodes * sizeof (mde_cookie_t);
    listp = kmem_zalloc(listsz, KM_SLEEP);

    num_nodes = md_scan_dag(mdp, rootnode, md_find_name(mdp, "cpu"),
        md_find_name(mdp, "fwd"), listp);

    if (num_nodes <= 0)
        goto done;

    for (i = 0; i < num_nodes; i++) {
        if (md_get_prop_val(mdp, listp[i], "id", &cpuid)) {
            cmn_err(CE_WARN, "cif_check_cpus: "
                "CPU instance %d has no 'id' property", i);
            continue;
        }

        mutex_enter(&cpu_lock);

        if (cpuid >= NCPU) {
            status = stopcpu_bycpuid(cpuid);
            if (status != 0 && status != ENOTSUP)
                cmn_err(CE_PANIC, "failed to stop cpu %lu (%d)",
                    cpuid, status);
        }

        mutex_exit(&cpu_lock);
    }

done:
    kmem_free(listp, listsz);
    (void) md_fini_handle(mdp);
}

void
cif_init(void)
{
    void (*kmdb_cb)(void);
    uint64_t rtba;
    uint64_t rv;
    size_t ntransroot;
    struct translation *transroot;

    /*
     * Check if domaining is enabled. If not, do not
     * initialize the kernel CIF handler.
     */
    if (!domaining_enabled())
        return;

    transroot = read_prom_mappings(&ntransroot);

    /*
     * Cache PROM data that is needed later, e.g. a shadow
     * copy of the device tree, IO mappings, etc.
     */
    cache_prom_data();

    /*
     * Prepare to take over the get/set of environmental variables.
     */
    promif_prop_init();

    /*
     * Switch CIF handler to the kernel.
     */
    prom_cif_handler = cif_handler;

    promif_preprom();
    cif_handler = kern_cif_handler;

    /*
     * Take over rtba for the boot CPU. The rtba for
     * all other CPUs are set as they enter the system.
     */
    rtba = va_to_pa(&trap_table);
    if ((rv = hv_cpu_set_rtba(&rtba)) != H_EOK)
        panic("hv_cpu_set_rtba failed: %ld\n", rv);

    promif_postprom();

    /*
     * If the system has been booted with kmdb we need kmdb to
     * use the kernel cif handler instead of the PROM cif handler.
     */
    if (boothowto & RB_KMDB) {
        kmdb_cb = (void (*)(void))modlookup("misc/kmdbmod",
            "kctl_switch_promif");
        ASSERT(kmdb_cb != NULL);
        (*kmdb_cb)();
    }

    cif_check_cpus();

    if (transroot != NULL)
        unmap_prom_mappings(transroot, ntransroot);
}

static void
cache_prom_data(void)
{
    /* initialize copy of OBP device tree */
    promif_stree_init();

    /* initialize io parameters */
    promif_io_init();
}

#endif  /* _KMDB */