/*      $NetBSD: pmap_tlb.c,v 1.62 2024/01/01 16:56:30 skrll Exp $      */

/*-
* Copyright (c) 2010 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Matt Thomas at 3am Software Foundry.
*
* 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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 <sys/cdefs.h>

__KERNEL_RCSID(0, "$NetBSD: pmap_tlb.c,v 1.62 2024/01/01 16:56:30 skrll Exp $");

/*
* Manages address spaces in a TLB.
*
* Normally there is a 1:1 mapping between a TLB and a CPU.  However, some
* implementations may share a TLB between multiple CPUs (really CPU thread
* contexts).  This requires the TLB abstraction to be separated from the
* CPU abstraction.  It also requires that the TLB be locked while doing
* TLB activities.
*
* For each TLB, we track the ASIDs in use in a bitmap and a list of pmaps
* that have a valid ASID.
*
* We allocate ASIDs in increasing order until we have exhausted the supply,
* then reinitialize the ASID space, and start allocating again at 1.  When
* allocating from the ASID bitmap, we skip any ASID who has a corresponding
* bit set in the ASID bitmap.  Eventually this causes the ASID bitmap to fill
* and, when completely filled, a reinitialization of the ASID space.
*
* To reinitialize the ASID space, the ASID bitmap is reset and then the ASIDs
* of non-kernel TLB entries get recorded in the ASID bitmap.  If the entries
* in TLB consume more than half of the ASID space, all ASIDs are invalidated,
* the ASID bitmap is recleared, and the list of pmaps is emptied.  Otherwise,
* (the normal case), any ASID present in the TLB (even those which are no
* longer used by a pmap) will remain active (allocated) and all other ASIDs
* will be freed.  If the size of the TLB is much smaller than the ASID space,
* this algorithm completely avoids TLB invalidation.
*
* For multiprocessors, we also have to deal TLB invalidation requests from
* other CPUs, some of which are dealt with the reinitialization of the ASID
* space.  Whereas above we keep the ASIDs of those pmaps which have active
* TLB entries, this type of reinitialization preserves the ASIDs of any
* "onproc" user pmap and all other ASIDs will be freed.  We must do this
* since we can't change the current ASID.
*
* Each pmap has two bitmaps: pm_active and pm_onproc.  Each bit in pm_active
* indicates whether that pmap has an allocated ASID for a CPU.  Each bit in
* pm_onproc indicates that the pmap's ASID is in use, i.e. a CPU has it in its
* "current ASID" field, e.g. the ASID field of the COP 0 register EntryHi for
* MIPS, or the ASID field of TTBR0 for AA64.  The bit number used in these
* bitmaps comes from the CPU's cpu_index().  Even though these bitmaps contain
* the bits for all CPUs, the bits that correspond to the bits belonging to
* the CPUs sharing a TLB can only be manipulated while holding that TLB's
* lock.  Atomic ops must be used to update them since multiple CPUs may be
* changing different sets of bits at same time but these sets never overlap.
*
* When a change to the local TLB may require a change in the TLB's of other
* CPUs, we try to avoid sending an IPI if at all possible.  For instance, if
* we are updating a PTE and that PTE previously was invalid and therefore
* couldn't support an active mapping, there's no need for an IPI since there
* can't be a TLB entry to invalidate.  The other case is when we change a PTE
* to be modified we just update the local TLB.  If another TLB has a stale
* entry, a TLB MOD exception will be raised and that will cause the local TLB
* to be updated.
*
* We never need to update a non-local TLB if the pmap doesn't have a valid
* ASID for that TLB.  If it does have a valid ASID but isn't current "onproc"
* we simply reset its ASID for that TLB and then when it goes "onproc" it
* will allocate a new ASID and any existing TLB entries will be orphaned.
* Only in the case that pmap has an "onproc" ASID do we actually have to send
* an IPI.
*
* Once we determined we must send an IPI to shootdown a TLB, we need to send
* it to one of CPUs that share that TLB.  We choose the lowest numbered CPU
* that has one of the pmap's ASID "onproc".  In reality, any CPU sharing that
* TLB would do, but interrupting an active CPU seems best.
*
* A TLB might have multiple shootdowns active concurrently.  The shootdown
* logic compresses these into a few cases:
*      0) nobody needs to have its TLB entries invalidated
*      1) one ASID needs to have its TLB entries invalidated
*      2) more than one ASID needs to have its TLB entries invalidated
*      3) the kernel needs to have its TLB entries invalidated
*      4) the kernel and one or more ASID need their TLB entries invalidated.
*
* And for each case we do:
*      0) nothing,
*      1) if that ASID is still "onproc", we invalidate the TLB entries for
*         that single ASID.  If not, just reset the pmap's ASID to invalidate
*         and let it allocate a new ASID the next time it goes "onproc",
*      2) we reinitialize the ASID space (preserving any "onproc" ASIDs) and
*         invalidate all non-wired non-global TLB entries,
*      3) we invalidate all of the non-wired global TLB entries,
*      4) we reinitialize the ASID space (again preserving any "onproc" ASIDs)
*         invalidate all non-wired TLB entries.
*
* As you can see, shootdowns are not concerned with addresses, just address
* spaces.  Since the number of TLB entries is usually quite small, this avoids
* a lot of overhead for not much gain.
*/

#define __PMAP_PRIVATE

#include "opt_multiprocessor.h"

#include <sys/param.h>

#include <sys/atomic.h>
#include <sys/cpu.h>
#include <sys/kernel.h>                 /* for cold */
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/systm.h>

#include <uvm/uvm.h>

static kmutex_t pmap_tlb0_lock __cacheline_aligned;

#define IFCONSTANT(x)   (__builtin_constant_p((x)) ? (x) : 0)

#if KERNEL_PID > 31
#error "KERNEL_PID expected in range 0-31"
#endif

#define TLBINFO_ASID_MARK_UNUSED(ti, asid) \
       __BITMAP_CLR((asid), &(ti)->ti_asid_bitmap)
#define TLBINFO_ASID_MARK_USED(ti, asid) \
       __BITMAP_SET((asid), &(ti)->ti_asid_bitmap)
#define TLBINFO_ASID_INUSE_P(ti, asid) \
       __BITMAP_ISSET((asid), &(ti)->ti_asid_bitmap)
#define TLBINFO_ASID_RESET(ti) \
       do {                                                            \
               __BITMAP_ZERO(&ti->ti_asid_bitmap);                     \
               for (tlb_asid_t asid = 0; asid <= KERNEL_PID; asid++)   \
                       TLBINFO_ASID_MARK_USED(ti, asid);               \
       } while (0)
#define TLBINFO_ASID_INITIAL_FREE(asid_max) \
       (asid_max + 1 /* 0 */ - (1 + KERNEL_PID))

struct pmap_tlb_info pmap_tlb0_info = {
       .ti_name = "tlb0",
       .ti_asid_hint = KERNEL_PID + 1,
#ifdef PMAP_TLB_NUM_PIDS
       .ti_asid_max = IFCONSTANT(PMAP_TLB_NUM_PIDS - 1),
       .ti_asids_free = IFCONSTANT(
               TLBINFO_ASID_INITIAL_FREE(PMAP_TLB_NUM_PIDS - 1)),
#endif
       .ti_asid_bitmap._b[0] = __BITS(0, KERNEL_PID),
#ifdef PMAP_TLB_WIRED_UPAGES
       .ti_wired = PMAP_TLB_WIRED_UPAGES,
#endif
       .ti_lock = &pmap_tlb0_lock,
       .ti_pais = LIST_HEAD_INITIALIZER(pmap_tlb0_info.ti_pais),
#if defined(MULTIPROCESSOR) && PMAP_TLB_MAX > 1
       .ti_tlbinvop = TLBINV_NOBODY,
#endif
};

#undef IFCONSTANT

#if defined(MULTIPROCESSOR) && PMAP_TLB_MAX > 1
struct pmap_tlb_info *pmap_tlbs[PMAP_TLB_MAX] = {
       [0] = &pmap_tlb0_info,
};
u_int pmap_ntlbs = 1;
#endif

#ifdef MULTIPROCESSOR
__unused static inline bool
pmap_tlb_intersecting_active_p(pmap_t pm, struct pmap_tlb_info *ti)
{
#if PMAP_TLB_MAX == 1
       return !kcpuset_iszero(pm->pm_active);
#else
       return kcpuset_intersecting_p(pm->pm_active, ti->ti_kcpuset);
#endif
}

static inline bool
pmap_tlb_intersecting_onproc_p(pmap_t pm, struct pmap_tlb_info *ti)
{
#if PMAP_TLB_MAX == 1
       return !kcpuset_iszero(pm->pm_onproc);
#else
       return kcpuset_intersecting_p(pm->pm_onproc, ti->ti_kcpuset);
#endif
}
#endif

static void
pmap_tlb_pai_check(struct pmap_tlb_info *ti, bool locked_p)
{
       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "(ti=%#jx)", (uintptr_t)ti, 0, 0, 0);

#ifdef DIAGNOSTIC
       struct pmap_asid_info *pai;
       if (!locked_p)
               TLBINFO_LOCK(ti);
       LIST_FOREACH(pai, &ti->ti_pais, pai_link) {
               KASSERT(pai != NULL);
               KASSERT(PAI_PMAP(pai, ti) != pmap_kernel());
               KASSERT(pai->pai_asid > KERNEL_PID);
               KASSERTMSG(pai->pai_asid <= ti->ti_asid_max,
                   "pm %p asid %#x", PAI_PMAP(pai, ti), pai->pai_asid);
               KASSERTMSG(TLBINFO_ASID_INUSE_P(ti, pai->pai_asid),
                   "pm %p asid %u", PAI_PMAP(pai, ti), pai->pai_asid);
#ifdef MULTIPROCESSOR
               KASSERT(pmap_tlb_intersecting_active_p(PAI_PMAP(pai, ti), ti));
#endif
       }
       if (!locked_p)
               TLBINFO_UNLOCK(ti);
#endif
       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

static void
pmap_tlb_pai_reset(struct pmap_tlb_info *ti, struct pmap_asid_info *pai,
       struct pmap *pm)
{
       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "(ti=%#jx, pai=%#jx, pm=%#jx): asid %u",
           (uintptr_t)ti, (uintptr_t)pai, (uintptr_t)pm, pai->pai_asid);

       /*
        * We must have an ASID but it must not be onproc (on a processor).
        */
       KASSERT(pai->pai_asid > KERNEL_PID);
       KASSERT(pai->pai_asid <= ti->ti_asid_max);
#if defined(MULTIPROCESSOR)
       KASSERT(pmap_tlb_intersecting_active_p(pm, ti));
       KASSERT(!pmap_tlb_intersecting_onproc_p(pm, ti));
#endif
       LIST_REMOVE(pai, pai_link);
#ifdef DIAGNOSTIC
       pai->pai_link.le_prev = NULL;   /* tagged as unlinked */
#endif
       /*
        * If the platform has a cheap way to flush ASIDs then free the ASID
        * back into the pool.  On multiprocessor systems, we will flush the
        * ASID from the TLB when it's allocated.  That way we know the flush
        * was always done in the correct TLB space.  On uniprocessor systems,
        * just do the flush now since we know that it has been used.  This has
        * a bit less overhead.  Either way, this will mean that we will only
        * need to flush all ASIDs if all ASIDs are in use and we need to
        * allocate a new one.
        */
       if (PMAP_TLB_FLUSH_ASID_ON_RESET) {
#ifndef MULTIPROCESSOR
               UVMHIST_LOG(maphist, " ... asid %u flushed", pai->pai_asid, 0,
                   0, 0);
               tlb_invalidate_asids(pai->pai_asid, pai->pai_asid);
#endif
               if (TLBINFO_ASID_INUSE_P(ti, pai->pai_asid)) {
                       UVMHIST_LOG(maphist, " ... asid marked unused",
                           pai->pai_asid, 0, 0, 0);
                       TLBINFO_ASID_MARK_UNUSED(ti, pai->pai_asid);
                       ti->ti_asids_free++;
               }
       }
       /*
        * Note that we don't mark the ASID as not in use in the TLB's ASID
        * bitmap (thus it can't be allocated until the ASID space is exhausted
        * and therefore reinitialized).  We don't want to flush the TLB for
        * entries belonging to this ASID so we will let natural TLB entry
        * replacement flush them out of the TLB.  Any new entries for this
        * pmap will need a new ASID allocated.
        */
       pai->pai_asid = 0;

#if defined(MULTIPROCESSOR)
       /*
        * The bits in pm_active belonging to this TLB can only be changed
        * while this TLB's lock is held.
        */
#if PMAP_TLB_MAX == 1
       kcpuset_zero(pm->pm_active);
#else
       kcpuset_remove(pm->pm_active, ti->ti_kcpuset);
#endif
       KASSERT(!pmap_tlb_intersecting_active_p(pm, ti));
#endif /* MULTIPROCESSOR */

       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

void
pmap_tlb_info_evcnt_attach(struct pmap_tlb_info *ti)
{
#if defined(MULTIPROCESSOR) && !defined(PMAP_TLB_NO_SYNCI_EVCNT)
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_synci_desired,
           EVCNT_TYPE_MISC, NULL,
           ti->ti_name, "icache syncs desired");
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_synci_asts,
           EVCNT_TYPE_MISC, &ti->ti_evcnt_synci_desired,
           ti->ti_name, "icache sync asts");
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_synci_all,
           EVCNT_TYPE_MISC, &ti->ti_evcnt_synci_asts,
           ti->ti_name, "icache full syncs");
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_synci_pages,
           EVCNT_TYPE_MISC, &ti->ti_evcnt_synci_asts,
           ti->ti_name, "icache pages synced");
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_synci_duplicate,
           EVCNT_TYPE_MISC, &ti->ti_evcnt_synci_desired,
           ti->ti_name, "icache dup pages skipped");
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_synci_deferred,
           EVCNT_TYPE_MISC, &ti->ti_evcnt_synci_desired,
           ti->ti_name, "icache pages deferred");
#endif /* MULTIPROCESSOR && !PMAP_TLB_NO_SYNCI_EVCNT */
       evcnt_attach_dynamic_nozero(&ti->ti_evcnt_asid_reinits,
           EVCNT_TYPE_MISC, NULL,
           ti->ti_name, "asid pool reinit");
}

void
pmap_tlb_info_init(struct pmap_tlb_info *ti)
{
#if defined(MULTIPROCESSOR)
#if PMAP_TLB_MAX == 1
       KASSERT(ti == &pmap_tlb0_info);
#else
       if (ti != &pmap_tlb0_info) {
               KASSERT(pmap_ntlbs < PMAP_TLB_MAX);

               KASSERT(pmap_tlbs[pmap_ntlbs] == NULL);

               ti->ti_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SCHED);
               TLBINFO_ASID_RESET(ti);
               ti->ti_asid_hint = KERNEL_PID + 1;
               ti->ti_asid_max = pmap_tlbs[0]->ti_asid_max;
               ti->ti_asids_free = TLBINFO_ASID_INITIAL_FREE(ti->ti_asid_max);
               ti->ti_tlbinvop = TLBINV_NOBODY;
               ti->ti_victim = NULL;
               kcpuset_create(&ti->ti_kcpuset, true);
               ti->ti_index = pmap_ntlbs++;
               ti->ti_wired = 0;
               pmap_tlbs[ti->ti_index] = ti;
               snprintf(ti->ti_name, sizeof(ti->ti_name), "tlb%u",
                   ti->ti_index);
               pmap_tlb_info_evcnt_attach(ti);

               KASSERT(ti->ti_asid_max < PMAP_TLB_BITMAP_LENGTH);
               return;
       }
#endif
#endif /* MULTIPROCESSOR */
       KASSERT(ti == &pmap_tlb0_info);
       KASSERT(ti->ti_lock == &pmap_tlb0_lock);

       mutex_init(ti->ti_lock, MUTEX_DEFAULT, IPL_SCHED);
#if defined(MULTIPROCESSOR) && PMAP_TLB_MAX > 1
       kcpuset_create(&ti->ti_kcpuset, true);
       kcpuset_set(ti->ti_kcpuset, cpu_index(curcpu()));
#endif

       const tlb_asid_t asid_max = pmap_md_tlb_asid_max();
       if (ti->ti_asid_max == 0 || asid_max < ti->ti_asid_max) {
               ti->ti_asid_max = asid_max;
               ti->ti_asids_free = TLBINFO_ASID_INITIAL_FREE(ti->ti_asid_max);
       }

       KASSERT(__type_fit(tlb_asid_t, ti->ti_asid_max + 1));
       KASSERT(ti->ti_asid_max < PMAP_TLB_BITMAP_LENGTH);
}

#if defined(MULTIPROCESSOR)
void
pmap_tlb_info_attach(struct pmap_tlb_info *ti, struct cpu_info *ci)
{
       KASSERT(!CPU_IS_PRIMARY(ci));
       KASSERT(ci->ci_data.cpu_idlelwp != NULL);
       KASSERT(cold);

       TLBINFO_LOCK(ti);
#if PMAP_TLB_MAX > 1
       kcpuset_set(ti->ti_kcpuset, cpu_index(ci));
       cpu_set_tlb_info(ci, ti);
#endif

       /*
        * Do any MD tlb info init.
        */
       pmap_md_tlb_info_attach(ti, ci);

       /*
        * The kernel pmap uses the kcpuset_running set so it's always
        * up-to-date.
        */
       TLBINFO_UNLOCK(ti);
}
#endif /* MULTIPROCESSOR */

#ifdef DIAGNOSTIC
static size_t
pmap_tlb_asid_count(struct pmap_tlb_info *ti)
{
       size_t count = 0;
       for (tlb_asid_t asid = 1; asid <= ti->ti_asid_max; asid++) {
               if (TLBINFO_ASID_INUSE_P(ti, asid))
                       count++;
       }
       return count;
}
#endif

static void
pmap_tlb_asid_reinitialize(struct pmap_tlb_info *ti, enum tlb_invalidate_op op)
{
       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "(ti=%#jx, op=%ju)", (uintptr_t)ti, op, 0, 0);

       pmap_tlb_pai_check(ti, true);

       ti->ti_evcnt_asid_reinits.ev_count++;

       /*
        * First, clear the ASID bitmap (except for ASID 0 which belongs
        * to the kernel).
        */
       ti->ti_asids_free = TLBINFO_ASID_INITIAL_FREE(ti->ti_asid_max);
       ti->ti_asid_hint = KERNEL_PID + 1;
       TLBINFO_ASID_RESET(ti);

       switch (op) {
#if defined(MULTIPROCESSOR) && defined(PMAP_TLB_NEED_SHOOTDOWN)
       case TLBINV_ALL:
               tlb_invalidate_all();
               break;
       case TLBINV_ALLUSER:
               tlb_invalidate_asids(KERNEL_PID + 1, ti->ti_asid_max);
               break;
#endif /* MULTIPROCESSOR && PMAP_TLB_NEED_SHOOTDOWN */
       case TLBINV_NOBODY: {
               /*
                * If we are just reclaiming ASIDs in the TLB, let's go find
                * what ASIDs are in use in the TLB.  Since this is a
                * semi-expensive operation, we don't want to do it too often.
                * So if more half of the ASIDs are in use, we don't have
                * enough free ASIDs so invalidate the TLB entries with ASIDs
                * and clear the ASID bitmap.  That will force everyone to
                * allocate a new ASID.
                */
#if !defined(MULTIPROCESSOR) || defined(PMAP_TLB_NEED_SHOOTDOWN)
               pmap_tlb_asid_check();
               const u_int asids_found = tlb_record_asids(
                   ti->ti_asid_bitmap._b, ti->ti_asid_max);
               pmap_tlb_asid_check();
#ifdef DIAGNOSTIC
               const u_int asids_count = pmap_tlb_asid_count(ti);
               KASSERTMSG(asids_found == asids_count,
                   "found %u != count %u", asids_found, asids_count);
#endif
               if (__predict_false(asids_found >= ti->ti_asid_max / 2)) {
                       tlb_invalidate_asids(KERNEL_PID + 1, ti->ti_asid_max);
#else /* MULTIPROCESSOR && !PMAP_TLB_NEED_SHOOTDOWN */
                       /*
                        * For those systems (PowerPC) that don't require
                        * cross cpu TLB shootdowns, we have to invalidate the
                        * entire TLB because we can't record the ASIDs in use
                        * on the other CPUs.  This is hopefully cheaper than
                        * than trying to use an IPI to record all the ASIDs
                        * on all the CPUs (which would be a synchronization
                        * nightmare).
                        */
                       tlb_invalidate_all();
#endif /* MULTIPROCESSOR && !PMAP_TLB_NEED_SHOOTDOWN */
                       TLBINFO_ASID_RESET(ti);
                       ti->ti_asids_free = TLBINFO_ASID_INITIAL_FREE(
                               ti->ti_asid_max);
#if !defined(MULTIPROCESSOR) || defined(PMAP_TLB_NEED_SHOOTDOWN)
               } else {
                       ti->ti_asids_free -= asids_found;
               }
#endif /* !MULTIPROCESSOR || PMAP_TLB_NEED_SHOOTDOWN */
               KASSERTMSG(ti->ti_asids_free <= ti->ti_asid_max, "%u",
                   ti->ti_asids_free);
               break;
       }
       default:
               panic("%s: unexpected op %d", __func__, op);
       }

       /*
        * Now go through the active ASIDs.  If the ASID is on a processor or
        * we aren't invalidating all ASIDs and the TLB has an entry owned by
        * that ASID, mark it as in use.  Otherwise release the ASID.
        */
       struct pmap_asid_info *pai, *next;
       for (pai = LIST_FIRST(&ti->ti_pais); pai != NULL; pai = next) {
               struct pmap * const pm = PAI_PMAP(pai, ti);
               next = LIST_NEXT(pai, pai_link);
               KASSERT(pm != pmap_kernel());
               KASSERT(pai->pai_asid > KERNEL_PID);
#if defined(MULTIPROCESSOR)
               if (pmap_tlb_intersecting_onproc_p(pm, ti)) {
                       if (!TLBINFO_ASID_INUSE_P(ti, pai->pai_asid)) {
                               TLBINFO_ASID_MARK_USED(ti, pai->pai_asid);
                               ti->ti_asids_free--;
                       }
                       continue;
               }
#endif /* MULTIPROCESSOR */
               if (TLBINFO_ASID_INUSE_P(ti, pai->pai_asid)) {
                       KASSERT(op == TLBINV_NOBODY);
               } else {
                       pmap_tlb_pai_reset(ti, pai, pm);
               }
       }
#ifdef DIAGNOSTIC
       size_t free_count __diagused = ti->ti_asid_max - pmap_tlb_asid_count(ti);
       KASSERTMSG(free_count == ti->ti_asids_free,
           "bitmap error: %zu != %u", free_count, ti->ti_asids_free);
#endif
       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

#if defined(MULTIPROCESSOR) && defined(PMAP_TLB_NEED_SHOOTDOWN)
#if PMAP_TLB_MAX == 1
#error shootdown not required for single TLB systems
#endif
void
pmap_tlb_shootdown_process(void)
{
       struct cpu_info * const ci = curcpu();
       struct pmap_tlb_info * const ti = cpu_tlb_info(ci);

       UVMHIST_FUNC(__func__);
       UVMHIST_CALLED(maphist);

       KASSERT(cpu_intr_p());
       KASSERTMSG(ci->ci_cpl >= IPL_SCHED, "%s: cpl (%d) < IPL_SCHED (%d)",
           __func__, ci->ci_cpl, IPL_SCHED);

       TLBINFO_LOCK(ti);
       UVMHIST_LOG(maphist, "ti %#jx", (uintptr_t)ti, 0, 0, 0);

       switch (ti->ti_tlbinvop) {
       case TLBINV_ONE: {
               /*
                * We only need to invalidate one user ASID.
                */
               UVMHIST_LOG(maphist, "TLBINV_ONE ti->ti_victim %#jx", (uintptr_t)ti->ti_victim, 0, 0, 0);
               struct pmap_asid_info * const pai = PMAP_PAI(ti->ti_victim, ti);
               KASSERT(ti->ti_victim != pmap_kernel());
               if (pmap_tlb_intersecting_onproc_p(ti->ti_victim, ti)) {
                       UVMHIST_LOG(maphist, "... onproc asid %jd", pai->pai_asid, 0, 0, 0);
                       /*
                        * The victim is an active pmap so we will just
                        * invalidate its TLB entries.
                        */
                       KASSERT(pai->pai_asid > KERNEL_PID);
                       pmap_tlb_asid_check();
                       tlb_invalidate_asids(pai->pai_asid, pai->pai_asid);
                       pmap_tlb_asid_check();
               } else if (pai->pai_asid) {
                       UVMHIST_LOG(maphist, "... not active asid %jd", pai->pai_asid, 0, 0, 0);
                       /*
                        * The victim is no longer an active pmap for this TLB.
                        * So simply clear its ASID and when pmap_activate is
                        * next called for this pmap, it will allocate a new
                        * ASID.
                        */
                       pmap_tlb_pai_reset(ti, pai, PAI_PMAP(pai, ti));
               }
               break;
       }
       case TLBINV_ALLUSER:
               /*
                * Flush all user TLB entries.
                */
               pmap_tlb_asid_reinitialize(ti, TLBINV_ALLUSER);
               break;
       case TLBINV_ALLKERNEL:
               /*
                * We need to invalidate all global TLB entries.
                */
               pmap_tlb_asid_check();
               tlb_invalidate_globals();
               pmap_tlb_asid_check();
               break;
       case TLBINV_ALL:
               /*
                * Flush all the TLB entries (user and kernel).
                */
               pmap_tlb_asid_reinitialize(ti, TLBINV_ALL);
               break;
       case TLBINV_NOBODY:
               /*
                * Might be spurious or another SMT CPU sharing this TLB
                * could have already done the work.
                */
               break;
       }

       /*
        * Indicate we are done with shutdown event.
        */
       ti->ti_victim = NULL;
       ti->ti_tlbinvop = TLBINV_NOBODY;
       TLBINFO_UNLOCK(ti);
}

/*
* This state machine could be encoded into an array of integers but since all
* the values fit in 3 bits, the 5 entry "table" fits in a 16 bit value which
* can be loaded in a single instruction.
*/
#define TLBINV_MAP(op, nobody, one, alluser, allkernel, all)    \
       ((((   (nobody) << 3 * TLBINV_NOBODY)                   \
        | (      (one) << 3 * TLBINV_ONE)                      \
        | (  (alluser) << 3 * TLBINV_ALLUSER)                  \
        | ((allkernel) << 3 * TLBINV_ALLKERNEL)                \
        | (      (all) << 3 * TLBINV_ALL)) >> 3 * (op)) & 7)

#define TLBINV_USER_MAP(op)     \
       TLBINV_MAP(op, TLBINV_ONE, TLBINV_ALLUSER, TLBINV_ALLUSER,      \
           TLBINV_ALL, TLBINV_ALL)

#define TLBINV_KERNEL_MAP(op)   \
       TLBINV_MAP(op, TLBINV_ALLKERNEL, TLBINV_ALL, TLBINV_ALL,        \
           TLBINV_ALLKERNEL, TLBINV_ALL)

bool
pmap_tlb_shootdown_bystanders(pmap_t pm)
{
       /*
        * We don't need to deal with our own TLB.
        */

       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "pm %#jx", (uintptr_t)pm, 0, 0, 0);

       KASSERT(kpreempt_disabled());

       const struct cpu_info * const ci = curcpu();

       kcpuset_t *pm_active = ci->ci_shootdowncpus;
       kcpuset_copy(pm_active, pm->pm_active);
       kcpuset_remove(pm_active, cpu_tlb_info(curcpu())->ti_kcpuset);
       const bool kernel_p = (pm == pmap_kernel());
       bool ipi_sent = false;

       /*
        * If pm_active gets more bits set, then it's after all our changes
        * have been made so they will already be cognizant of them.
        */

       for (size_t i = 0; !kcpuset_iszero(pm_active); i++) {
               KASSERT(i < pmap_ntlbs);
               struct pmap_tlb_info * const ti = pmap_tlbs[i];
               KASSERT(tlbinfo_index(ti) == i);
               UVMHIST_LOG(maphist, "ti %#jx", (uintptr_t)ti, 0, 0, 0);
               /*
                * Skip this TLB if there are no active mappings for it.
                */
               if (!kcpuset_intersecting_p(pm_active, ti->ti_kcpuset))
                       continue;
               struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);
               kcpuset_remove(pm_active, ti->ti_kcpuset);
               TLBINFO_LOCK(ti);
               cpuid_t j = kcpuset_ffs_intersecting(pm->pm_onproc,
                   ti->ti_kcpuset);
               // post decrement since ffs returns bit + 1 or 0 if no bit
               if (j-- > 0) {
                       if (kernel_p) {
                               ti->ti_tlbinvop =
                                   TLBINV_KERNEL_MAP(ti->ti_tlbinvop);
                               ti->ti_victim = NULL;
                       } else {
                               KASSERT(pai->pai_asid);
                               if (__predict_false(ti->ti_victim == pm)) {
                                       KASSERT(ti->ti_tlbinvop == TLBINV_ONE);
                                       /*
                                        * We still need to invalidate this one
                                        * ASID so there's nothing to change.
                                        */
                               } else {
                                       ti->ti_tlbinvop =
                                           TLBINV_USER_MAP(ti->ti_tlbinvop);
                                       if (ti->ti_tlbinvop == TLBINV_ONE)
                                               ti->ti_victim = pm;
                                       else
                                               ti->ti_victim = NULL;
                               }
                       }
                       UVMHIST_LOG(maphist, "tlbinvop %jx victim %#jx", ti->ti_tlbinvop,
                           (uintptr_t)ti->ti_victim, 0, 0);
                       TLBINFO_UNLOCK(ti);
                       /*
                        * Now we can send out the shootdown IPIs to a CPU
                        * that shares this TLB and is currently using this
                        * pmap.  That CPU will process the IPI and do the
                        * all the work.  Any other CPUs sharing that TLB
                        * will take advantage of that work.  pm_onproc might
                        * change now that we have released the lock but we
                        * can tolerate spurious shootdowns.
                        */
                       cpu_send_ipi(cpu_lookup(j), IPI_SHOOTDOWN);
                       ipi_sent = true;
                       continue;
               }
               if (!pmap_tlb_intersecting_active_p(pm, ti)) {
                       UVMHIST_LOG(maphist, "pm %#jx not active", (uintptr_t)pm, 0, 0, 0);
                       /*
                        * If this pmap has an ASID assigned but it's not
                        * currently running, nuke its ASID.  Next time the
                        * pmap is activated, it will allocate a new ASID.
                        * And best of all, we avoid an IPI.
                        */
                       KASSERT(!kernel_p);
                       pmap_tlb_pai_reset(ti, pai, pm);
                       //ti->ti_evcnt_lazy_shots.ev_count++;
               }
               TLBINFO_UNLOCK(ti);
       }

       UVMHIST_LOG(maphist, " <-- done (ipi_sent=%jd)", ipi_sent, 0, 0, 0);

       return ipi_sent;
}
#endif /* MULTIPROCESSOR && PMAP_TLB_NEED_SHOOTDOWN */

int
pmap_tlb_update_addr(pmap_t pm, vaddr_t va, pt_entry_t pte, u_int flags)
{
       KASSERT(kpreempt_disabled());

       struct pmap_tlb_info * const ti = cpu_tlb_info(curcpu());
       struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);
       int rv = -1;

       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, " (pm=%#jx va=%#jx, pte=%#jx flags=%#jx)",
           (uintptr_t)pm, va, pte_value(pte), flags);

       KASSERTMSG(pte_valid_p(pte), "va %#"PRIxVADDR" %#"PRIxPTE,
           va, pte_value(pte));

       TLBINFO_LOCK(ti);
       if (pm == pmap_kernel() || PMAP_PAI_ASIDVALID_P(pai, ti)) {
               pmap_tlb_asid_check();
               rv = tlb_update_addr(va, pai->pai_asid, pte,
                   (flags & PMAP_TLB_INSERT) != 0);
               pmap_tlb_asid_check();
               UVMHIST_LOG(maphist,
                   "   %jd <-- tlb_update_addr(%#jx, %#jx, %#jx, ...)",
                   rv, va, pai->pai_asid, pte_value(pte));
               KASSERTMSG((flags & PMAP_TLB_INSERT) == 0 || rv == 1,
                   "pmap %p (asid %u) va %#"PRIxVADDR" pte %#"PRIxPTE" rv %d",
                   pm, pai->pai_asid, va, pte_value(pte), rv);
       }
#if defined(MULTIPROCESSOR) && defined(PMAP_TLB_NEED_SHOOTDOWN)
       if (flags & PMAP_TLB_NEED_IPI)
               pm->pm_shootdown_pending = 1;
#endif
       TLBINFO_UNLOCK(ti);

       UVMHIST_LOG(maphist, "   <-- done (rv=%jd)", rv, 0, 0, 0);

       return rv;
}

void
pmap_tlb_invalidate_addr(pmap_t pm, vaddr_t va)
{
       KASSERT(kpreempt_disabled());

       struct pmap_tlb_info * const ti = cpu_tlb_info(curcpu());
       struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);

       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, " (pm=%#jx va=%#jx) ti=%#jx asid=%#jx",
           (uintptr_t)pm, va, (uintptr_t)ti, pai->pai_asid);

       TLBINFO_LOCK(ti);
       if (pm == pmap_kernel() || PMAP_PAI_ASIDVALID_P(pai, ti)) {
               pmap_tlb_asid_check();
               UVMHIST_LOG(maphist, " invalidating %#jx asid %#jx",
                   va, pai->pai_asid, 0, 0);
               tlb_invalidate_addr(va, pai->pai_asid);
               pmap_tlb_asid_check();
       }
#if defined(MULTIPROCESSOR) && defined(PMAP_TLB_NEED_SHOOTDOWN)
       pm->pm_shootdown_pending = 1;
#endif
       TLBINFO_UNLOCK(ti);
       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

static inline void
pmap_tlb_asid_alloc(struct pmap_tlb_info *ti, pmap_t pm,
       struct pmap_asid_info *pai)
{
       /*
        * We shouldn't have an ASID assigned, and thusly must not be onproc
        * nor active.
        */
       KASSERT(pm != pmap_kernel());
       KASSERT(pai->pai_asid == 0);
       KASSERT(pai->pai_link.le_prev == NULL);
#if defined(MULTIPROCESSOR)
       KASSERT(!pmap_tlb_intersecting_onproc_p(pm, ti));
       KASSERT(!pmap_tlb_intersecting_active_p(pm, ti));
#endif
       KASSERT(ti->ti_asids_free > 0);
       KASSERT(ti->ti_asid_hint > KERNEL_PID);

       /*
        * If the last ASID allocated was the maximum ASID, then the
        * hint will be out of range.  Reset the hint to first
        * available ASID.
        */
       if (PMAP_TLB_FLUSH_ASID_ON_RESET
           && ti->ti_asid_hint > ti->ti_asid_max) {
               ti->ti_asid_hint = KERNEL_PID + 1;
       }
       KASSERTMSG(ti->ti_asid_hint <= ti->ti_asid_max, "hint %u",
           ti->ti_asid_hint);

       /*
        * Let's see if the hinted ASID is free.  If not search for
        * a new one.
        */
       if (__predict_true(TLBINFO_ASID_INUSE_P(ti, ti->ti_asid_hint))) {
               const size_t nbpw = NBBY * sizeof(ti->ti_asid_bitmap._b[0]);
               size_t i;
               u_long bits;
               for (i = 0; (bits = ~ti->ti_asid_bitmap._b[i]) == 0; i++) {
                       KASSERT(i < __arraycount(ti->ti_asid_bitmap._b) - 1);
               }
               /*
                * ffs wants to find the first bit set while we want
                * to find the first bit cleared.
                */
               const u_int n = __builtin_ffsl(bits) - 1;
               KASSERTMSG((bits << (nbpw - (n+1))) == (1ul << (nbpw-1)),
                   "n %u bits %#lx", n, bits);
               KASSERT(n < nbpw);
               ti->ti_asid_hint = n + i * nbpw;
       }

       KASSERT(ti->ti_asid_hint > KERNEL_PID);
       KASSERT(ti->ti_asid_hint <= ti->ti_asid_max);
       KASSERTMSG(PMAP_TLB_FLUSH_ASID_ON_RESET
           || TLBINFO_ASID_INUSE_P(ti, ti->ti_asid_hint - 1),
           "hint %u bitmap %p", ti->ti_asid_hint, &ti->ti_asid_bitmap);
       KASSERTMSG(!TLBINFO_ASID_INUSE_P(ti, ti->ti_asid_hint),
           "hint %u bitmap %p", ti->ti_asid_hint, &ti->ti_asid_bitmap);

       /*
        * The hint contains our next ASID so take it and advance the hint.
        * Mark it as used and insert the pai into the list of active asids.
        * There is also one less asid free in this TLB.
        */
       pai->pai_asid = ti->ti_asid_hint++;
#ifdef MULTIPROCESSOR
       if (PMAP_TLB_FLUSH_ASID_ON_RESET) {
               /*
                * Clean the new ASID from the TLB.
                */
               tlb_invalidate_asids(pai->pai_asid, pai->pai_asid);
       }
#endif
       TLBINFO_ASID_MARK_USED(ti, pai->pai_asid);
       LIST_INSERT_HEAD(&ti->ti_pais, pai, pai_link);
       ti->ti_asids_free--;

#if defined(MULTIPROCESSOR)
       /*
        * Mark that we now have an active ASID for all CPUs sharing this TLB.
        * The bits in pm_active belonging to this TLB can only be changed
        * while this TLBs lock is held.
        */
#if PMAP_TLB_MAX == 1
       kcpuset_copy(pm->pm_active, kcpuset_running);
#else
       kcpuset_merge(pm->pm_active, ti->ti_kcpuset);
#endif
#endif
}

/*
* Acquire a TLB address space tag (called ASID or TLBPID) and return it.
* ASID might have already been previously acquired.
*/
void
pmap_tlb_asid_acquire(pmap_t pm, struct lwp *l)
{
       KASSERT(kpreempt_disabled());

       struct cpu_info * const ci = l->l_cpu;
       struct pmap_tlb_info * const ti = cpu_tlb_info(ci);
       struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);

       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "(pm=%#jx, l=%#jx, ti=%#jx)", (uintptr_t)pm,
           (uintptr_t)l, (uintptr_t)ti, 0);

       /*
        * Kernels use a fixed ASID and thus doesn't need to acquire one.
        */
       if (pm == pmap_kernel()) {
               UVMHIST_LOG(maphist, " <-- done (kernel)", 0, 0, 0, 0);
               return;
       }

       TLBINFO_LOCK(ti);
       KASSERT(pai->pai_asid <= KERNEL_PID || pai->pai_link.le_prev != NULL);
       KASSERT(pai->pai_asid > KERNEL_PID || pai->pai_link.le_prev == NULL);
       pmap_tlb_pai_check(ti, true);

       if (__predict_false(!tlbinfo_asids_p(ti))) {
#if defined(MULTIPROCESSOR)
               /*
                * Mark that we are active for all CPUs sharing this TLB.
                * The bits in pm_active belonging to this TLB can only
                * be changed  while this TLBs lock is held.
                */
#if PMAP_TLB_MAX == 1
               kcpuset_copy(pm->pm_active, kcpuset_running);
#else
               kcpuset_merge(pm->pm_active, ti->ti_kcpuset);
#endif
#endif
       } else if (__predict_false(!PMAP_PAI_ASIDVALID_P(pai, ti))) {
               /*
                * If we've run out ASIDs, reinitialize the ASID space.
                */
               if (__predict_false(tlbinfo_noasids_p(ti))) {
                       KASSERT(l == curlwp);
                       UVMHIST_LOG(maphist, " asid reinit", 0, 0, 0, 0);
                       pmap_tlb_asid_reinitialize(ti, TLBINV_NOBODY);
                       KASSERT(!tlbinfo_noasids_p(ti));
               }

               /*
                * Get an ASID.
                */
               pmap_tlb_asid_alloc(ti, pm, pai);
               UVMHIST_LOG(maphist, "allocated asid %#jx", pai->pai_asid,
                   0, 0, 0);
       }
       pmap_tlb_pai_check(ti, true);
#if defined(MULTIPROCESSOR)
       KASSERT(kcpuset_isset(pm->pm_active, cpu_index(ci)));
#endif

       if (l == curlwp) {
#if defined(MULTIPROCESSOR)
               /*
                * The bits in pm_onproc belonging to this TLB can only
                * be changed while this TLBs lock is held unless atomic
                * operations are used.
                */
               KASSERT(pm != pmap_kernel());
               kcpuset_atomic_set(pm->pm_onproc, cpu_index(ci));
#endif
               ci->ci_pmap_asid_cur = pai->pai_asid;
               UVMHIST_LOG(maphist, "setting asid to %#jx", pai->pai_asid,
                   0, 0, 0);
               tlb_set_asid(pai->pai_asid, pm);
               pmap_tlb_asid_check();
       } else {
               printf("%s: l (%p) != curlwp %p\n", __func__, l, curlwp);
       }
       TLBINFO_UNLOCK(ti);
       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

void
pmap_tlb_asid_deactivate(pmap_t pm)
{
       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "pm %#jx", (uintptr_t)pm, 0, 0, 0);

       KASSERT(kpreempt_disabled());
#if defined(MULTIPROCESSOR)
       /*
        * The kernel pmap is aways onproc and active and must never have
        * those bits cleared.  If pmap_remove_all was called, it has already
        * deactivated the pmap and thusly onproc will be 0 so there's nothing
        * to do.
        */
       if (pm != pmap_kernel() && !kcpuset_iszero(pm->pm_onproc)) {
               struct cpu_info * const ci = curcpu();
               KASSERT(!cpu_intr_p());
               KASSERTMSG(kcpuset_isset(pm->pm_onproc, cpu_index(ci)),
                   "%s: pmap %p onproc %p doesn't include cpu %d (%p)",
                   __func__, pm, pm->pm_onproc, cpu_index(ci), ci);
               /*
                * The bits in pm_onproc that belong to this TLB can
                * be changed while this TLBs lock is not held as long
                * as we use atomic ops.
                */
               kcpuset_atomic_clear(pm->pm_onproc, cpu_index(ci));
       }
#endif
       curcpu()->ci_pmap_asid_cur = KERNEL_PID;
       tlb_set_asid(KERNEL_PID, pmap_kernel());

       pmap_tlb_pai_check(cpu_tlb_info(curcpu()), false);
#if defined(DEBUG)
       pmap_tlb_asid_check();
#endif
       UVMHIST_LOG(maphist, " <-- done (pm=%#jx)", (uintptr_t)pm, 0, 0, 0);
}

void
pmap_tlb_asid_release_all(struct pmap *pm)
{
       UVMHIST_FUNC(__func__);
       UVMHIST_CALLARGS(maphist, "(pm=%#jx)", (uintptr_t)pm, 0, 0, 0);

       KASSERT(pm != pmap_kernel());
#if defined(MULTIPROCESSOR)
       //KASSERT(!kcpuset_iszero(pm->pm_onproc)); // XXX
       struct cpu_info * const ci __diagused = curcpu();
       KASSERT(!kcpuset_isotherset(pm->pm_onproc, cpu_index(ci)));
#if PMAP_TLB_MAX > 1
       for (u_int i = 0; !kcpuset_iszero(pm->pm_active); i++) {
               KASSERT(i < pmap_ntlbs);
               struct pmap_tlb_info * const ti = pmap_tlbs[i];
#else
               struct pmap_tlb_info * const ti = &pmap_tlb0_info;
#endif
               struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);
               TLBINFO_LOCK(ti);
               if (PMAP_PAI_ASIDVALID_P(pai, ti)) {
                       /*
                        * This pmap should not be in use by any other cpu so
                        * we can just reset and be happy.
                        */
                       if (ti->ti_victim == pm)
                               ti->ti_victim = NULL;
                       if (__predict_true(tlbinfo_asids_p(ti)))
                               pmap_tlb_pai_reset(ti, pai, pm);
               }
               KASSERT(pai->pai_link.le_prev == NULL);
               TLBINFO_UNLOCK(ti);
#if PMAP_TLB_MAX > 1
       }
#endif
#ifdef DIAGNOSTIC
       for (size_t i = 0; i < (PMAP_TLB_MAX > 1 ? pmap_ntlbs : 1); i++) {
               KASSERTMSG(pm->pm_pai[i].pai_asid == 0,
                   "pm %p i %zu asid %u",
                   pm, i, pm->pm_pai[i].pai_asid);
       }
#endif
#else
       /*
        * Handle the case of an UP kernel which only has, at most, one TLB.
        * If the pmap has an ASID allocated, free it.
        */
       struct pmap_tlb_info * const ti = &pmap_tlb0_info;
       struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);
       TLBINFO_LOCK(ti);
       if (pai->pai_asid > KERNEL_PID) {
               if (curcpu()->ci_pmap_asid_cur == pai->pai_asid) {
                       tlb_invalidate_asids(pai->pai_asid, pai->pai_asid);
               } else {
                       pmap_tlb_pai_reset(ti, pai, pm);
               }
       }
       TLBINFO_UNLOCK(ti);
#endif /* MULTIPROCESSOR */
       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

void
pmap_tlb_asid_check(void)
{
       UVMHIST_FUNC(__func__);
       UVMHIST_CALLED(maphist);

#ifdef DEBUG
       kpreempt_disable();
       const tlb_asid_t asid __debugused = tlb_get_asid();
       UVMHIST_LOG(maphist, " asid %u vs pmap_cur_asid %u", asid,
           curcpu()->ci_pmap_asid_cur, 0, 0);
       KDASSERTMSG(asid == curcpu()->ci_pmap_asid_cur,
          "%s: asid (%#x) != current asid (%#x)",
           __func__, asid, curcpu()->ci_pmap_asid_cur);
       kpreempt_enable();
#endif
       UVMHIST_LOG(maphist, " <-- done", 0, 0, 0, 0);
}

#ifdef DEBUG
void
pmap_tlb_check(pmap_t pm, bool (*func)(void *, vaddr_t, tlb_asid_t, pt_entry_t))
{
       struct pmap_tlb_info * const ti = cpu_tlb_info(curcpu());
       struct pmap_asid_info * const pai = PMAP_PAI(pm, ti);
       TLBINFO_LOCK(ti);
       if (pm == pmap_kernel() || pai->pai_asid > KERNEL_PID)
               tlb_walk(pm, func);
       TLBINFO_UNLOCK(ti);
}
#endif /* DEBUG */

#ifdef DDB
void
pmap_db_tlb_print(struct pmap *pm,
   void (*pr)(const char *, ...) __printflike(1, 2))
{
#if !defined(MULTIPROCESSOR) || PMAP_TLB_MAX == 1
       pr(" asid %5u\n", pm->pm_pai[0].pai_asid);
#else
       for (size_t i = 0; i < (PMAP_TLB_MAX > 1 ? pmap_ntlbs : 1); i++) {
               pr(" tlb %zu  asid %5u\n", i, pm->pm_pai[i].pai_asid);
       }
#endif
}
#endif /* DDB */