usched: Allow process to change self cpu affinity

[dragonfly.git] / sys / vm / vm_glue.c

/*
 * (MPSAFE)
 *
 * Copyright (c) 1991, 1993
 *      The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * The Mach Operating System project at Carnegie-Mellon University.
 *
 * 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.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
 *
 *      from: @(#)vm_glue.c     8.6 (Berkeley) 1/5/94
 *
 *
 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
 * All rights reserved.
 *
 * Permission to use, copy, modify and distribute this software and
 * its documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 *
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 *
 * Carnegie Mellon requests users of this software to return to
 *
 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 *
 * any improvements or extensions that they make and grant Carnegie the
 * rights to redistribute these changes.
 *
 * $FreeBSD: src/sys/vm/vm_glue.c,v 1.94.2.4 2003/01/13 22:51:17 dillon Exp $
 */

#include "opt_vm.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/buf.h>
#include <sys/shm.h>
#include <sys/vmmeter.h>
#include <sys/sysctl.h>

#include <sys/kernel.h>
#include <sys/unistd.h>

#include <machine/limits.h>
#include <machine/vmm.h>

#include <vm/vm.h>
#include <vm/vm_param.h>
#include <sys/lock.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>

#include <sys/user.h>
#include <vm/vm_page2.h>
#include <sys/thread2.h>
#include <sys/sysref2.h>

/*
 * THIS MUST BE THE LAST INITIALIZATION ITEM!!!
 *
 * Note: run scheduling should be divorced from the vm system.
 */
static void scheduler (void *);
SYSINIT(scheduler, SI_SUB_RUN_SCHEDULER, SI_ORDER_FIRST, scheduler, NULL);

#ifdef INVARIANTS

static int swap_debug = 0;
SYSCTL_INT(_vm, OID_AUTO, swap_debug,
        CTLFLAG_RW, &swap_debug, 0, "");

#endif

static int scheduler_notify;

static void swapout (struct proc *);

/*
 * No requirements.
 */
int
kernacc(c_caddr_t addr, int len, int rw)
{
        boolean_t rv;
        vm_offset_t saddr, eaddr;
        vm_prot_t prot;

        KASSERT((rw & (~VM_PROT_ALL)) == 0,
            ("illegal ``rw'' argument to kernacc (%x)", rw));

        /*
         * The globaldata space is not part of the kernel_map proper,
         * check access separately.
         */
        if (is_globaldata_space((vm_offset_t)addr, (vm_offset_t)(addr + len)))
                return (TRUE);

        /*
         * Nominal kernel memory access - check access via kernel_map.
         */
        if ((vm_offset_t)addr + len > kernel_map.max_offset ||
            (vm_offset_t)addr + len < (vm_offset_t)addr) {
                return (FALSE);
        }
        prot = rw;
        saddr = trunc_page((vm_offset_t)addr);
        eaddr = round_page((vm_offset_t)addr + len);
        rv = vm_map_check_protection(&kernel_map, saddr, eaddr, prot, FALSE);

        return (rv == TRUE);
}

/*
 * No requirements.
 */
int
useracc(c_caddr_t addr, int len, int rw)
{
        boolean_t rv;
        vm_prot_t prot;
        vm_map_t map;
        vm_map_entry_t save_hint;
        vm_offset_t wrap;
        vm_offset_t gpa;

        KASSERT((rw & (~VM_PROT_ALL)) == 0,
            ("illegal ``rw'' argument to useracc (%x)", rw));
        prot = rw;

        if (curthread->td_vmm) {
                if (vmm_vm_get_gpa(curproc, (register_t *)&gpa, (register_t) addr))
                        panic("%s: could not get GPA\n", __func__);
                addr = (c_caddr_t) gpa;
        }

        /*
         * XXX - check separately to disallow access to user area and user
         * page tables - they are in the map.
         */
        wrap = (vm_offset_t)addr + len;
        if (wrap > VM_MAX_USER_ADDRESS || wrap < (vm_offset_t)addr) {
                return (FALSE);
        }
        map = &curproc->p_vmspace->vm_map;
        vm_map_lock_read(map);
        /*
         * We save the map hint, and restore it.  Useracc appears to distort
         * the map hint unnecessarily.
         */
        save_hint = map->hint;
        rv = vm_map_check_protection(map, trunc_page((vm_offset_t)addr),
                                     round_page(wrap), prot, TRUE);
        map->hint = save_hint;
        vm_map_unlock_read(map);
        
        return (rv == TRUE);
}

/*
 * No requirements.
 */
void
vslock(caddr_t addr, u_int len)
{
        if (len) {
                vm_map_wire(&curproc->p_vmspace->vm_map,
                            trunc_page((vm_offset_t)addr),
                            round_page((vm_offset_t)addr + len), 0);
        }
}

/*
 * No requirements.
 */
void
vsunlock(caddr_t addr, u_int len)
{
        if (len) {
                vm_map_wire(&curproc->p_vmspace->vm_map,
                            trunc_page((vm_offset_t)addr),
                            round_page((vm_offset_t)addr + len),
                            KM_PAGEABLE);
        }
}

/*
 * Implement fork's actions on an address space.
 * Here we arrange for the address space to be copied or referenced,
 * allocate a user struct (pcb and kernel stack), then call the
 * machine-dependent layer to fill those in and make the new process
 * ready to run.  The new process is set up so that it returns directly
 * to user mode to avoid stack copying and relocation problems.
 *
 * No requirements.
 */
void
vm_fork(struct proc *p1, struct proc *p2, int flags)
{
        if ((flags & RFPROC) == 0) {
                /*
                 * Divorce the memory, if it is shared, essentially
                 * this changes shared memory amongst threads, into
                 * COW locally.
                 */
                if ((flags & RFMEM) == 0) {
                        if (vmspace_getrefs(p1->p_vmspace) > 1) {
                                vmspace_unshare(p1);
                        }
                }
                cpu_fork(ONLY_LWP_IN_PROC(p1), NULL, flags);
                return;
        }

        if (flags & RFMEM) {
                vmspace_ref(p1->p_vmspace);
                p2->p_vmspace = p1->p_vmspace;
        }

        while (vm_page_count_severe()) {
                vm_wait(0);
        }

        if ((flags & RFMEM) == 0) {
                p2->p_vmspace = vmspace_fork(p1->p_vmspace);

                pmap_pinit2(vmspace_pmap(p2->p_vmspace));

                if (p1->p_vmspace->vm_shm)
                        shmfork(p1, p2);
        }

        pmap_init_proc(p2);
}

/*
 * Set default limits for VM system.  Call during proc0's initialization.
 *
 * Called from the low level boot code only.
 */
void
vm_init_limits(struct proc *p)
{
        int rss_limit;

        /*
         * Set up the initial limits on process VM. Set the maximum resident
         * set size to be half of (reasonably) available memory.  Since this
         * is a soft limit, it comes into effect only when the system is out
         * of memory - half of main memory helps to favor smaller processes,
         * and reduces thrashing of the object cache.
         */
        p->p_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
        p->p_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
        p->p_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
        p->p_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
        /* limit the limit to no less than 2MB */
        rss_limit = max(vmstats.v_free_count, 512);
        p->p_rlimit[RLIMIT_RSS].rlim_cur = ptoa(rss_limit);
        p->p_rlimit[RLIMIT_RSS].rlim_max = RLIM_INFINITY;
}

/*
 * Faultin the specified process.  Note that the process can be in any
 * state.  Just clear P_SWAPPEDOUT and call wakeup in case the process is
 * sleeping.
 *
 * No requirements.
 */
void
faultin(struct proc *p)
{
        if (p->p_flags & P_SWAPPEDOUT) {
                /*
                 * The process is waiting in the kernel to return to user
                 * mode but cannot until P_SWAPPEDOUT gets cleared.
                 */
                lwkt_gettoken(&p->p_token);
                p->p_flags &= ~(P_SWAPPEDOUT | P_SWAPWAIT);
#ifdef INVARIANTS
                if (swap_debug)
                        kprintf("swapping in %d (%s)\n", p->p_pid, p->p_comm);
#endif
                wakeup(p);
                lwkt_reltoken(&p->p_token);
        }
}

/*
 * Kernel initialization eventually falls through to this function,
 * which is process 0.
 *
 * This swapin algorithm attempts to swap-in processes only if there
 * is enough space for them.  Of course, if a process waits for a long
 * time, it will be swapped in anyway.
 */
struct scheduler_info {
        struct proc *pp;
        int ppri;
};

static int scheduler_callback(struct proc *p, void *data);

static void
scheduler(void *dummy)
{
        struct scheduler_info info;
        struct proc *p;

        KKASSERT(!IN_CRITICAL_SECT(curthread));
loop:
        scheduler_notify = 0;
        /*
         * Don't try to swap anything in if we are low on memory.
         */
        if (vm_page_count_severe()) {
                vm_wait(0);
                goto loop;
        }

        /*
         * Look for a good candidate to wake up
         */
        info.pp = NULL;
        info.ppri = INT_MIN;
        allproc_scan(scheduler_callback, &info);

        /*
         * Nothing to do, back to sleep for at least 1/10 of a second.  If
         * we are woken up, immediately process the next request.  If
         * multiple requests have built up the first is processed 
         * immediately and the rest are staggered.
         */
        if ((p = info.pp) == NULL) {
                tsleep(&proc0, 0, "nowork", hz / 10);
                if (scheduler_notify == 0)
                        tsleep(&scheduler_notify, 0, "nowork", 0);
                goto loop;
        }

        /*
         * Fault the selected process in, then wait for a short period of
         * time and loop up.
         *
         * XXX we need a heuristic to get a measure of system stress and
         * then adjust our stagger wakeup delay accordingly.
         */
        lwkt_gettoken(&p->p_token);
        faultin(p);
        p->p_swtime = 0;
        lwkt_reltoken(&p->p_token);
        PRELE(p);
        tsleep(&proc0, 0, "swapin", hz / 10);
        goto loop;
}

/*
 * Process only has its hold count bumped, we need the token
 * to safely scan the LWPs
 */
static int
scheduler_callback(struct proc *p, void *data)
{
        struct scheduler_info *info = data;
        struct vmspace *vm;
        struct lwp *lp;
        segsz_t pgs;
        int pri;

        /*
         * We only care about processes in swap-wait.  Interlock test with
         * token if the flag is found set.
         */
        if ((p->p_flags & P_SWAPWAIT) == 0)
                return 0;
        lwkt_gettoken_shared(&p->p_token);
        if ((p->p_flags & P_SWAPWAIT) == 0) {
                lwkt_reltoken(&p->p_token);
                return 0;
        }

        /*
         * Calculate priority for swap-in
         */
        pri = 0;
        FOREACH_LWP_IN_PROC(lp, p) {
                /* XXX lwp might need a different metric */
                pri += lp->lwp_slptime;
        }
        pri += p->p_swtime - p->p_nice * 8;

        /*
         * The more pages paged out while we were swapped,
         * the more work we have to do to get up and running
         * again and the lower our wakeup priority.
         *
         * Each second of sleep time is worth ~1MB
         */
        if ((vm = p->p_vmspace) != NULL) {
                vmspace_hold(vm);
                pgs = vmspace_resident_count(vm);
                if (pgs < vm->vm_swrss) {
                        pri -= (vm->vm_swrss - pgs) /
                               (1024 * 1024 / PAGE_SIZE);
                }
                vmspace_drop(vm);
        }
        lwkt_reltoken(&p->p_token);

        /*
         * If this process is higher priority and there is
         * enough space, then select this process instead of
         * the previous selection.
         */
        if (pri > info->ppri) {
                if (info->pp)
                        PRELE(info->pp);
                PHOLD(p);
                info->pp = p;
                info->ppri = pri;
        }
        return(0);
}

/*
 * SMP races ok.
 * No requirements.
 */
void
swapin_request(void)
{
        if (scheduler_notify == 0) {
                scheduler_notify = 1;
                wakeup(&scheduler_notify);
        }
}

#ifndef NO_SWAPPING

#define swappable(p) \
        (((p)->p_lock == 0) && \
        ((p)->p_flags & (P_TRACED|P_SYSTEM|P_SWAPPEDOUT|P_WEXIT)) == 0)


/*
 * Swap_idle_threshold1 is the guaranteed swapped in time for a process
 */
static int swap_idle_threshold1 = 15;
SYSCTL_INT(_vm, OID_AUTO, swap_idle_threshold1,
        CTLFLAG_RW, &swap_idle_threshold1, 0, "Guaranteed process resident time (sec)");

/*
 * Swap_idle_threshold2 is the time that a process can be idle before
 * it will be swapped out, if idle swapping is enabled.  Default is
 * one minute.
 */
static int swap_idle_threshold2 = 60;
SYSCTL_INT(_vm, OID_AUTO, swap_idle_threshold2,
        CTLFLAG_RW, &swap_idle_threshold2, 0, "Time (sec) a process can idle before being swapped");

/*
 * Swapout is driven by the pageout daemon.  Very simple, we find eligible
 * procs and mark them as being swapped out.  This will cause the kernel
 * to prefer to pageout those proc's pages first and the procs in question 
 * will not return to user mode until the swapper tells them they can.
 *
 * If any procs have been sleeping/stopped for at least maxslp seconds,
 * they are swapped.  Else, we swap the longest-sleeping or stopped process,
 * if any, otherwise the longest-resident process.
 */

static int swapout_procs_callback(struct proc *p, void *data);

/*
 * No requirements.
 */
void
swapout_procs(int action)
{
        allproc_scan(swapout_procs_callback, &action);
}

static int
swapout_procs_callback(struct proc *p, void *data)
{
        struct lwp *lp;
        int action = *(int *)data;
        int minslp = -1;

        if (!swappable(p))
                return(0);

        lwkt_gettoken(&p->p_token);

        /*
         * We only consider active processes.
         */
        if (p->p_stat != SACTIVE && p->p_stat != SSTOP) {
                lwkt_reltoken(&p->p_token);
                return(0);
        }

        FOREACH_LWP_IN_PROC(lp, p) {
                /*
                 * do not swap out a realtime process
                 */
                if (RTP_PRIO_IS_REALTIME(lp->lwp_rtprio.type)) {
                        lwkt_reltoken(&p->p_token);
                        return(0);
                }

                /*
                 * Guarentee swap_idle_threshold time in memory
                 */
                if (lp->lwp_slptime < swap_idle_threshold1) {
                        lwkt_reltoken(&p->p_token);
                        return(0);
                }

                /*
                 * If the system is under memory stress, or if we
                 * are swapping idle processes >= swap_idle_threshold2,
                 * then swap the process out.
                 */
                if (((action & VM_SWAP_NORMAL) == 0) &&
                    (((action & VM_SWAP_IDLE) == 0) ||
                     (lp->lwp_slptime < swap_idle_threshold2))) {
                        lwkt_reltoken(&p->p_token);
                        return(0);
                }

                if (minslp == -1 || lp->lwp_slptime < minslp)
                        minslp = lp->lwp_slptime;
        }

        /*
         * If the process has been asleep for awhile, swap
         * it out.
         */
        if ((action & VM_SWAP_NORMAL) ||
            ((action & VM_SWAP_IDLE) &&
             (minslp > swap_idle_threshold2))) {
                swapout(p);
        }

        /*
         * cleanup our reference
         */
        lwkt_reltoken(&p->p_token);

        return(0);
}

/*
 * The caller must hold p->p_token
 */
static void
swapout(struct proc *p)
{
#ifdef INVARIANTS
        if (swap_debug)
                kprintf("swapping out %d (%s)\n", p->p_pid, p->p_comm);
#endif
        ++p->p_ru.ru_nswap;

        /*
         * remember the process resident count
         */
        p->p_vmspace->vm_swrss = vmspace_resident_count(p->p_vmspace);
        p->p_flags |= P_SWAPPEDOUT;
        p->p_swtime = 0;
}

#endif /* !NO_SWAPPING */