kernel - Fix excessive call stack depth on stuck interrupt

[dragonfly.git] / sys / kern / kern_plimit.c

/*
 * Copyright (c) 2006 The DragonFly Project.  All rights reserved.
 * 
 * This code is derived from software contributed to The DragonFly Project
 * by Matthew Dillon <dillon@backplane.com>
 * 
 * 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 DragonFly Project 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 COPYRIGHT HOLDERS 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
 * COPYRIGHT HOLDERS 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.
 */
/*
 * Copyright (c) 1982, 1986, 1991, 1993
 *      The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * 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.
 *
 *      @(#)kern_resource.c     8.5 (Berkeley) 1/21/94
 * $FreeBSD: src/sys/kern/kern_resource.c,v 1.55.2.5 2001/11/03 01:41:08 ps Exp $
 */

#include <sys/resource.h>
#include <sys/spinlock.h>
#include <sys/proc.h>
#include <sys/priv.h>
#include <sys/file.h>
#include <sys/lockf.h>
#include <sys/kern_syscall.h>

#include <vm/vm_param.h>
#include <vm/vm.h>
#include <vm/vm_map.h>

#include <machine/pmap.h>

#include <sys/spinlock2.h>

static void plimit_copy(struct plimit *olimit, struct plimit *nlimit);

/*
 * Initialize proc0's plimit structure.  All later plimit structures
 * are inherited through fork.
 */
void
plimit_init0(struct plimit *limit)
{
        int i;
        rlim_t lim;

        for (i = 0; i < RLIM_NLIMITS; ++i) {
                limit->pl_rlimit[i].rlim_cur = RLIM_INFINITY;
                limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
        }
        limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = maxfiles;
        limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
        limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = maxproc;
        limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
        lim = ptoa((rlim_t)vmstats.v_free_count);
        limit->pl_rlimit[RLIMIT_RSS].rlim_max = lim;
        limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
        limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
        limit->p_cpulimit = RLIM_INFINITY;
        limit->p_refcnt = 1;
        spin_init(&limit->p_spin, "plimitinit");
}

/*
 * Return a plimit for use by a new forked process given the one
 * contained in the parent process.
 *
 * MPSAFE
 */
struct plimit *
plimit_fork(struct proc *p1)
{
        struct plimit *olimit = p1->p_limit;
        struct plimit *nlimit = NULL;
        struct plimit *rlimit;

        /*
         * If we are exclusive (but not threaded-exclusive), but have only
         * one reference, we can convert the structure to copy-on-write
         * again.
         *
         * If we were threaded but are no longer threaded we can do the same
         * thing.
         */
        if (olimit->p_exclusive == 1) {
                KKASSERT(olimit->p_refcnt == 1);
                olimit->p_exclusive = 0;
        } else if (olimit->p_exclusive == 2 && p1->p_nthreads == 1) {
                KKASSERT(olimit->p_refcnt == 1);
                olimit->p_exclusive = 0;
        }

        /*
         * Take a short-cut that requires limited spin locks.  If we aren't
         * exclusive we will not be threaded and we can just bump the ref
         * count.  If that is true and we also have only one ref then there
         * can be no other accessors.
         */
        if (olimit->p_exclusive == 0) {
                if (olimit->p_refcnt == 1) {
                        ++olimit->p_refcnt;
                } else {
                        spin_lock(&olimit->p_spin);
                        ++olimit->p_refcnt;
                        spin_unlock(&olimit->p_spin);
                }
                return(olimit);
        }

        /*
         * Full-blown code-up.
         */
        nlimit = NULL;
        spin_lock(&olimit->p_spin);

        for (;;) {
                if (olimit->p_exclusive == 0) {
                        ++olimit->p_refcnt;
                        rlimit = olimit;
                        break;
                }
                if (nlimit) {
                        plimit_copy(olimit, nlimit);
                        rlimit = nlimit;
                        nlimit = NULL;
                        break;
                }
                spin_unlock(&olimit->p_spin);
                nlimit = kmalloc(sizeof(*nlimit), M_SUBPROC, M_WAITOK);
                spin_lock(&olimit->p_spin);
        }
        spin_unlock(&olimit->p_spin);
        if (nlimit)
                kfree(nlimit, M_SUBPROC);
        return(rlimit);
}

/*
 * This routine is called when a new LWP is created for a process.  We
 * must force exclusivity (=2) so p->p_limit remains stable.
 *
 * LWPs share the same process structure so this does not bump refcnt.
 */
void
plimit_lwp_fork(struct proc *p)
{
        struct plimit *olimit;

        for (;;) {
                olimit = p->p_limit;
                if (olimit->p_exclusive == 2) {
                        KKASSERT(olimit->p_refcnt == 1);
                        break;
                }
                if (olimit->p_refcnt == 1) {
                        olimit->p_exclusive = 2;
                        break;
                }
                plimit_modify(p, -1, NULL);
        }
}

/*
 * This routine is called to fixup a proces's p_limit structure prior
 * to it being modified.  If index >= 0 the specified modification is also
 * made.
 *
 * This routine must make the limit structure exclusive.  A later fork
 * will convert it back to copy-on-write if possible.
 *
 * We can count on p->p_limit being stable since if we had created any
 * threads it will have already been made exclusive (=2).
 *
 * MPSAFE
 */
void
plimit_modify(struct proc *p, int index, struct rlimit *rlim)
{
        struct plimit *olimit;
        struct plimit *nlimit;
        struct plimit *rlimit;

        /*
         * Shortcut.  If we are not threaded we may be able to trivially
         * set the structure to exclusive access without needing to acquire
         * any spinlocks.   The p_limit structure will be stable.
         */
        olimit = p->p_limit;
        if (p->p_nthreads == 1) {
                if (olimit->p_exclusive == 0 && olimit->p_refcnt == 1)
                        olimit->p_exclusive = 1;
                if (olimit->p_exclusive) {
                        if (index >= 0)
                                p->p_limit->pl_rlimit[index] = *rlim;
                        return;
                }
        }

        /*
         * Full-blown code-up.  Make a copy if we aren't exclusive.  If
         * we have only one ref we can safely convert the structure to
         * exclusive without copying.
         */
        nlimit = NULL;
        spin_lock(&olimit->p_spin);

        for (;;) {
                if (olimit->p_refcnt == 1) {
                        if (olimit->p_exclusive == 0)
                                olimit->p_exclusive = 1;
                        rlimit = olimit;
                        break;
                }
                KKASSERT(olimit->p_exclusive == 0);
                if (nlimit) {
                        plimit_copy(olimit, nlimit);
                        nlimit->p_exclusive = 1;
                        p->p_limit = nlimit;
                        rlimit = nlimit;
                        nlimit = NULL;
                        break;
                }
                spin_unlock(&olimit->p_spin);
                nlimit = kmalloc(sizeof(*nlimit), M_SUBPROC, M_WAITOK);
                spin_lock(&olimit->p_spin);
        }
        if (index >= 0)
                rlimit->pl_rlimit[index] = *rlim;
        spin_unlock(&olimit->p_spin);
        if (nlimit)
                kfree(nlimit, M_SUBPROC);
}

/*
 * Destroy a process's plimit structure.
 *
 * MPSAFE
 */
void
plimit_free(struct proc *p)
{
        struct plimit *limit;

        if ((limit = p->p_limit) != NULL) {
                p->p_limit = NULL;

                if (limit->p_refcnt == 1) {
                        limit->p_refcnt = -999;
                        kfree(limit, M_SUBPROC);
                } else {
                        spin_lock(&limit->p_spin);
                        if (--limit->p_refcnt == 0) {
                                spin_unlock(&limit->p_spin);
                                kfree(limit, M_SUBPROC);
                        } else {
                                spin_unlock(&limit->p_spin);
                        }
                }
        }
}

/*
 * Modify a resource limit (from system call)
 *
 * MPSAFE
 */
int
kern_setrlimit(u_int which, struct rlimit *limp)
{
        struct proc *p = curproc;
        struct plimit *limit;
        struct rlimit *alimp;
        int error;

        if (which >= RLIM_NLIMITS)
                return (EINVAL);

        /*
         * We will be modifying a resource, make a copy if necessary.
         */
        plimit_modify(p, -1, NULL);
        limit = p->p_limit;
        alimp = &limit->pl_rlimit[which];

        /*
         * Preserve historical bugs by treating negative limits as unsigned.
         */
        if (limp->rlim_cur < 0)
                limp->rlim_cur = RLIM_INFINITY;
        if (limp->rlim_max < 0)
                limp->rlim_max = RLIM_INFINITY;

        spin_lock(&limit->p_spin);
        if (limp->rlim_cur > alimp->rlim_max ||
            limp->rlim_max > alimp->rlim_max) {
                spin_unlock(&limit->p_spin);
                error = priv_check_cred(p->p_ucred, PRIV_PROC_SETRLIMIT, 0);
                if (error)
                        return (error);
        } else {
                spin_unlock(&limit->p_spin);
        }
        if (limp->rlim_cur > limp->rlim_max)
                limp->rlim_cur = limp->rlim_max;

        switch (which) {
        case RLIMIT_CPU:
                spin_lock(&limit->p_spin);
                if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000)
                        limit->p_cpulimit = RLIM_INFINITY;
                else
                        limit->p_cpulimit = (rlim_t)1000000 * limp->rlim_cur;
                spin_unlock(&limit->p_spin);
                break;
        case RLIMIT_DATA:
                if (limp->rlim_cur > maxdsiz)
                        limp->rlim_cur = maxdsiz;
                if (limp->rlim_max > maxdsiz)
                        limp->rlim_max = maxdsiz;
                break;

        case RLIMIT_STACK:
                if (limp->rlim_cur > maxssiz)
                        limp->rlim_cur = maxssiz;
                if (limp->rlim_max > maxssiz)
                        limp->rlim_max = maxssiz;
                /*
                 * Stack is allocated to the max at exec time with only
                 * "rlim_cur" bytes accessible.  If stack limit is going
                 * up make more accessible, if going down make inaccessible.
                 */
                spin_lock(&limit->p_spin);
                if (limp->rlim_cur != alimp->rlim_cur) {
                        vm_offset_t addr;
                        vm_size_t size;
                        vm_prot_t prot;

                        if (limp->rlim_cur > alimp->rlim_cur) {
                                prot = VM_PROT_ALL;
                                size = limp->rlim_cur - alimp->rlim_cur;
                                addr = USRSTACK - limp->rlim_cur;
                        } else {
                                prot = VM_PROT_NONE;
                                size = alimp->rlim_cur - limp->rlim_cur;
                                addr = USRSTACK - alimp->rlim_cur;
                        }
                        spin_unlock(&limit->p_spin);
                        addr = trunc_page(addr);
                        size = round_page(size);
                        vm_map_protect(&p->p_vmspace->vm_map,
                                       addr, addr+size, prot, FALSE);
                } else {
                        spin_unlock(&limit->p_spin);
                }
                break;

        case RLIMIT_NOFILE:
                if (limp->rlim_cur > maxfilesperproc)
                        limp->rlim_cur = maxfilesperproc;
                if (limp->rlim_max > maxfilesperproc)
                        limp->rlim_max = maxfilesperproc;
                break;

        case RLIMIT_NPROC:
                if (limp->rlim_cur > maxprocperuid)
                        limp->rlim_cur = maxprocperuid;
                if (limp->rlim_max > maxprocperuid)
                        limp->rlim_max = maxprocperuid;
                if (limp->rlim_cur < 1)
                        limp->rlim_cur = 1;
                if (limp->rlim_max < 1)
                        limp->rlim_max = 1;
                break;
        case RLIMIT_POSIXLOCKS:
                if (limp->rlim_cur > maxposixlocksperuid)
                        limp->rlim_cur = maxposixlocksperuid;
                if (limp->rlim_max > maxposixlocksperuid)
                        limp->rlim_max = maxposixlocksperuid;
                break;
        }
        spin_lock(&limit->p_spin);
        *alimp = *limp;
        spin_unlock(&limit->p_spin);
        return (0);
}

/*
 * The rlimit indexed by which is returned in the second argument.
 *
 * MPSAFE
 */
int
kern_getrlimit(u_int which, struct rlimit *limp)
{
        struct proc *p = curproc;
        struct plimit *limit;

        /*
         * p is NULL when kern_getrlimit is called from a
         * kernel thread. In this case as the calling proc
         * isn't available we just skip the limit check.
         */
        if (p == NULL)
                return 0;

        if (which >= RLIM_NLIMITS)
                return (EINVAL);

        limit = p->p_limit;
        spin_lock(&limit->p_spin);
        *limp = p->p_rlimit[which];
        spin_unlock(&limit->p_spin);
        return (0);
}

/*
 * Determine if the cpu limit has been reached and return an operations
 * code for the caller to perform.
 *
 * MPSAFE
 */
int
plimit_testcpulimit(struct plimit *limit, u_int64_t ttime)
{
        struct rlimit *rlim;
        int mode;

        /*
         * Initial tests without the spinlock.  This is the fast path.
         * Any 32/64 bit glitches will fall through and retest with
         * the spinlock.
         */
        if (limit->p_cpulimit == RLIM_INFINITY)
                return(PLIMIT_TESTCPU_OK);
        if (ttime <= limit->p_cpulimit)
                return(PLIMIT_TESTCPU_OK);

        spin_lock(&limit->p_spin);
        if (ttime > limit->p_cpulimit) {
                rlim = &limit->pl_rlimit[RLIMIT_CPU];
                if (ttime / (rlim_t)1000000 >= rlim->rlim_max + 5)
                        mode = PLIMIT_TESTCPU_KILL;
                else
                        mode = PLIMIT_TESTCPU_XCPU;
        } else {
                mode = PLIMIT_TESTCPU_OK;
        }
        spin_unlock(&limit->p_spin);
        return(mode);
}

/*
 * Helper routine to copy olimit to nlimit and initialize nlimit for
 * use.  nlimit's reference count will be set to 1 and its exclusive bit
 * will be cleared.
 *
 * MPSAFE
 */
static
void
plimit_copy(struct plimit *olimit, struct plimit *nlimit)
{
        *nlimit = *olimit;

        spin_init(&nlimit->p_spin, "plimitcopy");
        nlimit->p_refcnt = 1;
        nlimit->p_exclusive = 0;
}