xref: /openbmc/qemu/accel/tcg/cputlb.c (revision 3b295bcb)

/*
 *  Common CPU TLB handling
 *
 *  Copyright (c) 2003 Fabrice Bellard
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
 */

#include "qemu/osdep.h"
#include "qemu/main-loop.h"
#include "hw/core/tcg-cpu-ops.h"
#include "exec/exec-all.h"
#include "exec/memory.h"
#include "exec/cpu_ldst.h"
#include "exec/cputlb.h"
#include "exec/memory-internal.h"
#include "exec/ram_addr.h"
#include "tcg/tcg.h"
#include "qemu/error-report.h"
#include "exec/log.h"
#include "exec/helper-proto-common.h"
#include "qemu/atomic.h"
#include "qemu/atomic128.h"
#include "exec/translate-all.h"
#include "trace.h"
#include "tb-hash.h"
#include "internal.h"
#ifdef CONFIG_PLUGIN
#include "qemu/plugin-memory.h"
#endif
#include "tcg/tcg-ldst.h"
#include "tcg/oversized-guest.h"

/* DEBUG defines, enable DEBUG_TLB_LOG to log to the CPU_LOG_MMU target */
/* #define DEBUG_TLB */
/* #define DEBUG_TLB_LOG */

#ifdef DEBUG_TLB
# define DEBUG_TLB_GATE 1
# ifdef DEBUG_TLB_LOG
#  define DEBUG_TLB_LOG_GATE 1
# else
#  define DEBUG_TLB_LOG_GATE 0
# endif
#else
# define DEBUG_TLB_GATE 0
# define DEBUG_TLB_LOG_GATE 0
#endif

#define tlb_debug(fmt, ...) do { \
    if (DEBUG_TLB_LOG_GATE) { \
        qemu_log_mask(CPU_LOG_MMU, "%s: " fmt, __func__, \
                      ## __VA_ARGS__); \
    } else if (DEBUG_TLB_GATE) { \
        fprintf(stderr, "%s: " fmt, __func__, ## __VA_ARGS__); \
    } \
} while (0)

#define assert_cpu_is_self(cpu) do {                              \
        if (DEBUG_TLB_GATE) {                                     \
            g_assert(!(cpu)->created || qemu_cpu_is_self(cpu));   \
        }                                                         \
    } while (0)

/* run_on_cpu_data.target_ptr should always be big enough for a
 * target_ulong even on 32 bit builds */
QEMU_BUILD_BUG_ON(sizeof(target_ulong) > sizeof(run_on_cpu_data));

/* We currently can't handle more than 16 bits in the MMUIDX bitmask.
 */
QEMU_BUILD_BUG_ON(NB_MMU_MODES > 16);
#define ALL_MMUIDX_BITS ((1 << NB_MMU_MODES) - 1)

static inline size_t tlb_n_entries(CPUTLBDescFast *fast)
{
    return (fast->mask >> CPU_TLB_ENTRY_BITS) + 1;
}

static inline size_t sizeof_tlb(CPUTLBDescFast *fast)
{
    return fast->mask + (1 << CPU_TLB_ENTRY_BITS);
}

static void tlb_window_reset(CPUTLBDesc *desc, int64_t ns,
                             size_t max_entries)
{
    desc->window_begin_ns = ns;
    desc->window_max_entries = max_entries;
}

static void tb_jmp_cache_clear_page(CPUState *cpu, vaddr page_addr)
{
    CPUJumpCache *jc = cpu->tb_jmp_cache;
    int i, i0;

    if (unlikely(!jc)) {
        return;
    }

    i0 = tb_jmp_cache_hash_page(page_addr);
    for (i = 0; i < TB_JMP_PAGE_SIZE; i++) {
        qatomic_set(&jc->array[i0 + i].tb, NULL);
    }
}

/**
 * tlb_mmu_resize_locked() - perform TLB resize bookkeeping; resize if necessary
 * @desc: The CPUTLBDesc portion of the TLB
 * @fast: The CPUTLBDescFast portion of the same TLB
 *
 * Called with tlb_lock_held.
 *
 * We have two main constraints when resizing a TLB: (1) we only resize it
 * on a TLB flush (otherwise we'd have to take a perf hit by either rehashing
 * the array or unnecessarily flushing it), which means we do not control how
 * frequently the resizing can occur; (2) we don't have access to the guest's
 * future scheduling decisions, and therefore have to decide the magnitude of
 * the resize based on past observations.
 *
 * In general, a memory-hungry process can benefit greatly from an appropriately
 * sized TLB, since a guest TLB miss is very expensive. This doesn't mean that
 * we just have to make the TLB as large as possible; while an oversized TLB
 * results in minimal TLB miss rates, it also takes longer to be flushed
 * (flushes can be _very_ frequent), and the reduced locality can also hurt
 * performance.
 *
 * To achieve near-optimal performance for all kinds of workloads, we:
 *
 * 1. Aggressively increase the size of the TLB when the use rate of the
 * TLB being flushed is high, since it is likely that in the near future this
 * memory-hungry process will execute again, and its memory hungriness will
 * probably be similar.
 *
 * 2. Slowly reduce the size of the TLB as the use rate declines over a
 * reasonably large time window. The rationale is that if in such a time window
 * we have not observed a high TLB use rate, it is likely that we won't observe
 * it in the near future. In that case, once a time window expires we downsize
 * the TLB to match the maximum use rate observed in the window.
 *
 * 3. Try to keep the maximum use rate in a time window in the 30-70% range,
 * since in that range performance is likely near-optimal. Recall that the TLB
 * is direct mapped, so we want the use rate to be low (or at least not too
 * high), since otherwise we are likely to have a significant amount of
 * conflict misses.
 */
static void tlb_mmu_resize_locked(CPUTLBDesc *desc, CPUTLBDescFast *fast,
                                  int64_t now)
{
    size_t old_size = tlb_n_entries(fast);
    size_t rate;
    size_t new_size = old_size;
    int64_t window_len_ms = 100;
    int64_t window_len_ns = window_len_ms * 1000 * 1000;
    bool window_expired = now > desc->window_begin_ns + window_len_ns;

    if (desc->n_used_entries > desc->window_max_entries) {
        desc->window_max_entries = desc->n_used_entries;
    }
    rate = desc->window_max_entries * 100 / old_size;

    if (rate > 70) {
        new_size = MIN(old_size << 1, 1 << CPU_TLB_DYN_MAX_BITS);
    } else if (rate < 30 && window_expired) {
        size_t ceil = pow2ceil(desc->window_max_entries);
        size_t expected_rate = desc->window_max_entries * 100 / ceil;

        /*
         * Avoid undersizing when the max number of entries seen is just below
         * a pow2. For instance, if max_entries == 1025, the expected use rate
         * would be 1025/2048==50%. However, if max_entries == 1023, we'd get
         * 1023/1024==99.9% use rate, so we'd likely end up doubling the size
         * later. Thus, make sure that the expected use rate remains below 70%.
         * (and since we double the size, that means the lowest rate we'd
         * expect to get is 35%, which is still in the 30-70% range where
         * we consider that the size is appropriate.)
         */
        if (expected_rate > 70) {
            ceil *= 2;
        }
        new_size = MAX(ceil, 1 << CPU_TLB_DYN_MIN_BITS);
    }

    if (new_size == old_size) {
        if (window_expired) {
            tlb_window_reset(desc, now, desc->n_used_entries);
        }
        return;
    }

    g_free(fast->table);
    g_free(desc->fulltlb);

    tlb_window_reset(desc, now, 0);
    /* desc->n_used_entries is cleared by the caller */
    fast->mask = (new_size - 1) << CPU_TLB_ENTRY_BITS;
    fast->table = g_try_new(CPUTLBEntry, new_size);
    desc->fulltlb = g_try_new(CPUTLBEntryFull, new_size);

    /*
     * If the allocations fail, try smaller sizes. We just freed some
     * memory, so going back to half of new_size has a good chance of working.
     * Increased memory pressure elsewhere in the system might cause the
     * allocations to fail though, so we progressively reduce the allocation
     * size, aborting if we cannot even allocate the smallest TLB we support.
     */
    while (fast->table == NULL || desc->fulltlb == NULL) {
        if (new_size == (1 << CPU_TLB_DYN_MIN_BITS)) {
            error_report("%s: %s", __func__, strerror(errno));
            abort();
        }
        new_size = MAX(new_size >> 1, 1 << CPU_TLB_DYN_MIN_BITS);
        fast->mask = (new_size - 1) << CPU_TLB_ENTRY_BITS;

        g_free(fast->table);
        g_free(desc->fulltlb);
        fast->table = g_try_new(CPUTLBEntry, new_size);
        desc->fulltlb = g_try_new(CPUTLBEntryFull, new_size);
    }
}

static void tlb_mmu_flush_locked(CPUTLBDesc *desc, CPUTLBDescFast *fast)
{
    desc->n_used_entries = 0;
    desc->large_page_addr = -1;
    desc->large_page_mask = -1;
    desc->vindex = 0;
    memset(fast->table, -1, sizeof_tlb(fast));
    memset(desc->vtable, -1, sizeof(desc->vtable));
}

static void tlb_flush_one_mmuidx_locked(CPUArchState *env, int mmu_idx,
                                        int64_t now)
{
    CPUTLBDesc *desc = &env_tlb(env)->d[mmu_idx];
    CPUTLBDescFast *fast = &env_tlb(env)->f[mmu_idx];

    tlb_mmu_resize_locked(desc, fast, now);
    tlb_mmu_flush_locked(desc, fast);
}

static void tlb_mmu_init(CPUTLBDesc *desc, CPUTLBDescFast *fast, int64_t now)
{
    size_t n_entries = 1 << CPU_TLB_DYN_DEFAULT_BITS;

    tlb_window_reset(desc, now, 0);
    desc->n_used_entries = 0;
    fast->mask = (n_entries - 1) << CPU_TLB_ENTRY_BITS;
    fast->table = g_new(CPUTLBEntry, n_entries);
    desc->fulltlb = g_new(CPUTLBEntryFull, n_entries);
    tlb_mmu_flush_locked(desc, fast);
}

static inline void tlb_n_used_entries_inc(CPUArchState *env, uintptr_t mmu_idx)
{
    env_tlb(env)->d[mmu_idx].n_used_entries++;
}

static inline void tlb_n_used_entries_dec(CPUArchState *env, uintptr_t mmu_idx)
{
    env_tlb(env)->d[mmu_idx].n_used_entries--;
}

void tlb_init(CPUState *cpu)
{
    CPUArchState *env = cpu->env_ptr;
    int64_t now = get_clock_realtime();
    int i;

    qemu_spin_init(&env_tlb(env)->c.lock);

    /* All tlbs are initialized flushed. */
    env_tlb(env)->c.dirty = 0;

    for (i = 0; i < NB_MMU_MODES; i++) {
        tlb_mmu_init(&env_tlb(env)->d[i], &env_tlb(env)->f[i], now);
    }
}

void tlb_destroy(CPUState *cpu)
{
    CPUArchState *env = cpu->env_ptr;
    int i;

    qemu_spin_destroy(&env_tlb(env)->c.lock);
    for (i = 0; i < NB_MMU_MODES; i++) {
        CPUTLBDesc *desc = &env_tlb(env)->d[i];
        CPUTLBDescFast *fast = &env_tlb(env)->f[i];

        g_free(fast->table);
        g_free(desc->fulltlb);
    }
}

/* flush_all_helper: run fn across all cpus
 *
 * If the wait flag is set then the src cpu's helper will be queued as
 * "safe" work and the loop exited creating a synchronisation point
 * where all queued work will be finished before execution starts
 * again.
 */
static void flush_all_helper(CPUState *src, run_on_cpu_func fn,
                             run_on_cpu_data d)
{
    CPUState *cpu;

    CPU_FOREACH(cpu) {
        if (cpu != src) {
            async_run_on_cpu(cpu, fn, d);
        }
    }
}

void tlb_flush_counts(size_t *pfull, size_t *ppart, size_t *pelide)
{
    CPUState *cpu;
    size_t full = 0, part = 0, elide = 0;

    CPU_FOREACH(cpu) {
        CPUArchState *env = cpu->env_ptr;

        full += qatomic_read(&env_tlb(env)->c.full_flush_count);
        part += qatomic_read(&env_tlb(env)->c.part_flush_count);
        elide += qatomic_read(&env_tlb(env)->c.elide_flush_count);
    }
    *pfull = full;
    *ppart = part;
    *pelide = elide;
}

static void tlb_flush_by_mmuidx_async_work(CPUState *cpu, run_on_cpu_data data)
{
    CPUArchState *env = cpu->env_ptr;
    uint16_t asked = data.host_int;
    uint16_t all_dirty, work, to_clean;
    int64_t now = get_clock_realtime();

    assert_cpu_is_self(cpu);

    tlb_debug("mmu_idx:0x%04" PRIx16 "\n", asked);

    qemu_spin_lock(&env_tlb(env)->c.lock);

    all_dirty = env_tlb(env)->c.dirty;
    to_clean = asked & all_dirty;
    all_dirty &= ~to_clean;
    env_tlb(env)->c.dirty = all_dirty;

    for (work = to_clean; work != 0; work &= work - 1) {
        int mmu_idx = ctz32(work);
        tlb_flush_one_mmuidx_locked(env, mmu_idx, now);
    }

    qemu_spin_unlock(&env_tlb(env)->c.lock);

    tcg_flush_jmp_cache(cpu);

    if (to_clean == ALL_MMUIDX_BITS) {
        qatomic_set(&env_tlb(env)->c.full_flush_count,
                   env_tlb(env)->c.full_flush_count + 1);
    } else {
        qatomic_set(&env_tlb(env)->c.part_flush_count,
                   env_tlb(env)->c.part_flush_count + ctpop16(to_clean));
        if (to_clean != asked) {
            qatomic_set(&env_tlb(env)->c.elide_flush_count,
                       env_tlb(env)->c.elide_flush_count +
                       ctpop16(asked & ~to_clean));
        }
    }
}

void tlb_flush_by_mmuidx(CPUState *cpu, uint16_t idxmap)
{
    tlb_debug("mmu_idx: 0x%" PRIx16 "\n", idxmap);

    if (cpu->created && !qemu_cpu_is_self(cpu)) {
        async_run_on_cpu(cpu, tlb_flush_by_mmuidx_async_work,
                         RUN_ON_CPU_HOST_INT(idxmap));
    } else {
        tlb_flush_by_mmuidx_async_work(cpu, RUN_ON_CPU_HOST_INT(idxmap));
    }
}

void tlb_flush(CPUState *cpu)
{
    tlb_flush_by_mmuidx(cpu, ALL_MMUIDX_BITS);
}

void tlb_flush_by_mmuidx_all_cpus(CPUState *src_cpu, uint16_t idxmap)
{
    const run_on_cpu_func fn = tlb_flush_by_mmuidx_async_work;

    tlb_debug("mmu_idx: 0x%"PRIx16"\n", idxmap);

    flush_all_helper(src_cpu, fn, RUN_ON_CPU_HOST_INT(idxmap));
    fn(src_cpu, RUN_ON_CPU_HOST_INT(idxmap));
}

void tlb_flush_all_cpus(CPUState *src_cpu)
{
    tlb_flush_by_mmuidx_all_cpus(src_cpu, ALL_MMUIDX_BITS);
}

void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *src_cpu, uint16_t idxmap)
{
    const run_on_cpu_func fn = tlb_flush_by_mmuidx_async_work;

    tlb_debug("mmu_idx: 0x%"PRIx16"\n", idxmap);

    flush_all_helper(src_cpu, fn, RUN_ON_CPU_HOST_INT(idxmap));
    async_safe_run_on_cpu(src_cpu, fn, RUN_ON_CPU_HOST_INT(idxmap));
}

void tlb_flush_all_cpus_synced(CPUState *src_cpu)
{
    tlb_flush_by_mmuidx_all_cpus_synced(src_cpu, ALL_MMUIDX_BITS);
}

static bool tlb_hit_page_mask_anyprot(CPUTLBEntry *tlb_entry,
                                      vaddr page, vaddr mask)
{
    page &= mask;
    mask &= TARGET_PAGE_MASK | TLB_INVALID_MASK;

    return (page == (tlb_entry->addr_read & mask) ||
            page == (tlb_addr_write(tlb_entry) & mask) ||
            page == (tlb_entry->addr_code & mask));
}

static inline bool tlb_hit_page_anyprot(CPUTLBEntry *tlb_entry, vaddr page)
{
    return tlb_hit_page_mask_anyprot(tlb_entry, page, -1);
}

/**
 * tlb_entry_is_empty - return true if the entry is not in use
 * @te: pointer to CPUTLBEntry
 */
static inline bool tlb_entry_is_empty(const CPUTLBEntry *te)
{
    return te->addr_read == -1 && te->addr_write == -1 && te->addr_code == -1;
}

/* Called with tlb_c.lock held */
static bool tlb_flush_entry_mask_locked(CPUTLBEntry *tlb_entry,
                                        vaddr page,
                                        vaddr mask)
{
    if (tlb_hit_page_mask_anyprot(tlb_entry, page, mask)) {
        memset(tlb_entry, -1, sizeof(*tlb_entry));
        return true;
    }
    return false;
}

static inline bool tlb_flush_entry_locked(CPUTLBEntry *tlb_entry, vaddr page)
{
    return tlb_flush_entry_mask_locked(tlb_entry, page, -1);
}

/* Called with tlb_c.lock held */
static void tlb_flush_vtlb_page_mask_locked(CPUArchState *env, int mmu_idx,
                                            vaddr page,
                                            vaddr mask)
{
    CPUTLBDesc *d = &env_tlb(env)->d[mmu_idx];
    int k;

    assert_cpu_is_self(env_cpu(env));
    for (k = 0; k < CPU_VTLB_SIZE; k++) {
        if (tlb_flush_entry_mask_locked(&d->vtable[k], page, mask)) {
            tlb_n_used_entries_dec(env, mmu_idx);
        }
    }
}

static inline void tlb_flush_vtlb_page_locked(CPUArchState *env, int mmu_idx,
                                              vaddr page)
{
    tlb_flush_vtlb_page_mask_locked(env, mmu_idx, page, -1);
}

static void tlb_flush_page_locked(CPUArchState *env, int midx, vaddr page)
{
    vaddr lp_addr = env_tlb(env)->d[midx].large_page_addr;
    vaddr lp_mask = env_tlb(env)->d[midx].large_page_mask;

    /* Check if we need to flush due to large pages.  */
    if ((page & lp_mask) == lp_addr) {
        tlb_debug("forcing full flush midx %d (%"
                  VADDR_PRIx "/%" VADDR_PRIx ")\n",
                  midx, lp_addr, lp_mask);
        tlb_flush_one_mmuidx_locked(env, midx, get_clock_realtime());
    } else {
        if (tlb_flush_entry_locked(tlb_entry(env, midx, page), page)) {
            tlb_n_used_entries_dec(env, midx);
        }
        tlb_flush_vtlb_page_locked(env, midx, page);
    }
}

/**
 * tlb_flush_page_by_mmuidx_async_0:
 * @cpu: cpu on which to flush
 * @addr: page of virtual address to flush
 * @idxmap: set of mmu_idx to flush
 *
 * Helper for tlb_flush_page_by_mmuidx and friends, flush one page
 * at @addr from the tlbs indicated by @idxmap from @cpu.
 */
static void tlb_flush_page_by_mmuidx_async_0(CPUState *cpu,
                                             vaddr addr,
                                             uint16_t idxmap)
{
    CPUArchState *env = cpu->env_ptr;
    int mmu_idx;

    assert_cpu_is_self(cpu);

    tlb_debug("page addr: %" VADDR_PRIx " mmu_map:0x%x\n", addr, idxmap);

    qemu_spin_lock(&env_tlb(env)->c.lock);
    for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) {
        if ((idxmap >> mmu_idx) & 1) {
            tlb_flush_page_locked(env, mmu_idx, addr);
        }
    }
    qemu_spin_unlock(&env_tlb(env)->c.lock);

    /*
     * Discard jump cache entries for any tb which might potentially
     * overlap the flushed page, which includes the previous.
     */
    tb_jmp_cache_clear_page(cpu, addr - TARGET_PAGE_SIZE);
    tb_jmp_cache_clear_page(cpu, addr);
}

/**
 * tlb_flush_page_by_mmuidx_async_1:
 * @cpu: cpu on which to flush
 * @data: encoded addr + idxmap
 *
 * Helper for tlb_flush_page_by_mmuidx and friends, called through
 * async_run_on_cpu.  The idxmap parameter is encoded in the page
 * offset of the target_ptr field.  This limits the set of mmu_idx
 * that can be passed via this method.
 */
static void tlb_flush_page_by_mmuidx_async_1(CPUState *cpu,
                                             run_on_cpu_data data)
{
    vaddr addr_and_idxmap = data.target_ptr;
    vaddr addr = addr_and_idxmap & TARGET_PAGE_MASK;
    uint16_t idxmap = addr_and_idxmap & ~TARGET_PAGE_MASK;

    tlb_flush_page_by_mmuidx_async_0(cpu, addr, idxmap);
}

typedef struct {
    vaddr addr;
    uint16_t idxmap;
} TLBFlushPageByMMUIdxData;

/**
 * tlb_flush_page_by_mmuidx_async_2:
 * @cpu: cpu on which to flush
 * @data: allocated addr + idxmap
 *
 * Helper for tlb_flush_page_by_mmuidx and friends, called through
 * async_run_on_cpu.  The addr+idxmap parameters are stored in a
 * TLBFlushPageByMMUIdxData structure that has been allocated
 * specifically for this helper.  Free the structure when done.
 */
static void tlb_flush_page_by_mmuidx_async_2(CPUState *cpu,
                                             run_on_cpu_data data)
{
    TLBFlushPageByMMUIdxData *d = data.host_ptr;

    tlb_flush_page_by_mmuidx_async_0(cpu, d->addr, d->idxmap);
    g_free(d);
}

void tlb_flush_page_by_mmuidx(CPUState *cpu, vaddr addr, uint16_t idxmap)
{
    tlb_debug("addr: %" VADDR_PRIx " mmu_idx:%" PRIx16 "\n", addr, idxmap);

    /* This should already be page aligned */
    addr &= TARGET_PAGE_MASK;

    if (qemu_cpu_is_self(cpu)) {
        tlb_flush_page_by_mmuidx_async_0(cpu, addr, idxmap);
    } else if (idxmap < TARGET_PAGE_SIZE) {
        /*
         * Most targets have only a few mmu_idx.  In the case where
         * we can stuff idxmap into the low TARGET_PAGE_BITS, avoid
         * allocating memory for this operation.
         */
        async_run_on_cpu(cpu, tlb_flush_page_by_mmuidx_async_1,
                         RUN_ON_CPU_TARGET_PTR(addr | idxmap));
    } else {
        TLBFlushPageByMMUIdxData *d = g_new(TLBFlushPageByMMUIdxData, 1);

        /* Otherwise allocate a structure, freed by the worker.  */
        d->addr = addr;
        d->idxmap = idxmap;
        async_run_on_cpu(cpu, tlb_flush_page_by_mmuidx_async_2,
                         RUN_ON_CPU_HOST_PTR(d));
    }
}

void tlb_flush_page(CPUState *cpu, vaddr addr)
{
    tlb_flush_page_by_mmuidx(cpu, addr, ALL_MMUIDX_BITS);
}

void tlb_flush_page_by_mmuidx_all_cpus(CPUState *src_cpu, vaddr addr,
                                       uint16_t idxmap)
{
    tlb_debug("addr: %" VADDR_PRIx " mmu_idx:%"PRIx16"\n", addr, idxmap);

    /* This should already be page aligned */
    addr &= TARGET_PAGE_MASK;

    /*
     * Allocate memory to hold addr+idxmap only when needed.
     * See tlb_flush_page_by_mmuidx for details.
     */
    if (idxmap < TARGET_PAGE_SIZE) {
        flush_all_helper(src_cpu, tlb_flush_page_by_mmuidx_async_1,
                         RUN_ON_CPU_TARGET_PTR(addr | idxmap));
    } else {
        CPUState *dst_cpu;

        /* Allocate a separate data block for each destination cpu.  */
        CPU_FOREACH(dst_cpu) {
            if (dst_cpu != src_cpu) {
                TLBFlushPageByMMUIdxData *d
                    = g_new(TLBFlushPageByMMUIdxData, 1);

                d->addr = addr;
                d->idxmap = idxmap;
                async_run_on_cpu(dst_cpu, tlb_flush_page_by_mmuidx_async_2,
                                 RUN_ON_CPU_HOST_PTR(d));
            }
        }
    }

    tlb_flush_page_by_mmuidx_async_0(src_cpu, addr, idxmap);
}

void tlb_flush_page_all_cpus(CPUState *src, vaddr addr)
{
    tlb_flush_page_by_mmuidx_all_cpus(src, addr, ALL_MMUIDX_BITS);
}

void tlb_flush_page_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
                                              vaddr addr,
                                              uint16_t idxmap)
{
    tlb_debug("addr: %" VADDR_PRIx " mmu_idx:%"PRIx16"\n", addr, idxmap);

    /* This should already be page aligned */
    addr &= TARGET_PAGE_MASK;

    /*
     * Allocate memory to hold addr+idxmap only when needed.
     * See tlb_flush_page_by_mmuidx for details.
     */
    if (idxmap < TARGET_PAGE_SIZE) {
        flush_all_helper(src_cpu, tlb_flush_page_by_mmuidx_async_1,
                         RUN_ON_CPU_TARGET_PTR(addr | idxmap));
        async_safe_run_on_cpu(src_cpu, tlb_flush_page_by_mmuidx_async_1,
                              RUN_ON_CPU_TARGET_PTR(addr | idxmap));
    } else {
        CPUState *dst_cpu;
        TLBFlushPageByMMUIdxData *d;

        /* Allocate a separate data block for each destination cpu.  */
        CPU_FOREACH(dst_cpu) {
            if (dst_cpu != src_cpu) {
                d = g_new(TLBFlushPageByMMUIdxData, 1);
                d->addr = addr;
                d->idxmap = idxmap;
                async_run_on_cpu(dst_cpu, tlb_flush_page_by_mmuidx_async_2,
                                 RUN_ON_CPU_HOST_PTR(d));
            }
        }

        d = g_new(TLBFlushPageByMMUIdxData, 1);
        d->addr = addr;
        d->idxmap = idxmap;
        async_safe_run_on_cpu(src_cpu, tlb_flush_page_by_mmuidx_async_2,
                              RUN_ON_CPU_HOST_PTR(d));
    }
}

void tlb_flush_page_all_cpus_synced(CPUState *src, vaddr addr)
{
    tlb_flush_page_by_mmuidx_all_cpus_synced(src, addr, ALL_MMUIDX_BITS);
}

static void tlb_flush_range_locked(CPUArchState *env, int midx,
                                   vaddr addr, vaddr len,
                                   unsigned bits)
{
    CPUTLBDesc *d = &env_tlb(env)->d[midx];
    CPUTLBDescFast *f = &env_tlb(env)->f[midx];
    vaddr mask = MAKE_64BIT_MASK(0, bits);

    /*
     * If @bits is smaller than the tlb size, there may be multiple entries
     * within the TLB; otherwise all addresses that match under @mask hit
     * the same TLB entry.
     * TODO: Perhaps allow bits to be a few bits less than the size.
     * For now, just flush the entire TLB.
     *
     * If @len is larger than the tlb size, then it will take longer to
     * test all of the entries in the TLB than it will to flush it all.
     */
    if (mask < f->mask || len > f->mask) {
        tlb_debug("forcing full flush midx %d ("
                  "%" VADDR_PRIx "/%" VADDR_PRIx "+%" VADDR_PRIx ")\n",
                  midx, addr, mask, len);
        tlb_flush_one_mmuidx_locked(env, midx, get_clock_realtime());
        return;
    }

    /*
     * Check if we need to flush due to large pages.
     * Because large_page_mask contains all 1's from the msb,
     * we only need to test the end of the range.
     */
    if (((addr + len - 1) & d->large_page_mask) == d->large_page_addr) {
        tlb_debug("forcing full flush midx %d ("
                  "%" VADDR_PRIx "/%" VADDR_PRIx ")\n",
                  midx, d->large_page_addr, d->large_page_mask);
        tlb_flush_one_mmuidx_locked(env, midx, get_clock_realtime());
        return;
    }

    for (vaddr i = 0; i < len; i += TARGET_PAGE_SIZE) {
        vaddr page = addr + i;
        CPUTLBEntry *entry = tlb_entry(env, midx, page);

        if (tlb_flush_entry_mask_locked(entry, page, mask)) {
            tlb_n_used_entries_dec(env, midx);
        }
        tlb_flush_vtlb_page_mask_locked(env, midx, page, mask);
    }
}

typedef struct {
    vaddr addr;
    vaddr len;
    uint16_t idxmap;
    uint16_t bits;
} TLBFlushRangeData;

static void tlb_flush_range_by_mmuidx_async_0(CPUState *cpu,
                                              TLBFlushRangeData d)
{
    CPUArchState *env = cpu->env_ptr;
    int mmu_idx;

    assert_cpu_is_self(cpu);

    tlb_debug("range: %" VADDR_PRIx "/%u+%" VADDR_PRIx " mmu_map:0x%x\n",
              d.addr, d.bits, d.len, d.idxmap);

    qemu_spin_lock(&env_tlb(env)->c.lock);
    for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) {
        if ((d.idxmap >> mmu_idx) & 1) {
            tlb_flush_range_locked(env, mmu_idx, d.addr, d.len, d.bits);
        }
    }
    qemu_spin_unlock(&env_tlb(env)->c.lock);

    /*
     * If the length is larger than the jump cache size, then it will take
     * longer to clear each entry individually than it will to clear it all.
     */
    if (d.len >= (TARGET_PAGE_SIZE * TB_JMP_CACHE_SIZE)) {
        tcg_flush_jmp_cache(cpu);
        return;
    }

    /*
     * Discard jump cache entries for any tb which might potentially
     * overlap the flushed pages, which includes the previous.
     */
    d.addr -= TARGET_PAGE_SIZE;
    for (vaddr i = 0, n = d.len / TARGET_PAGE_SIZE + 1; i < n; i++) {
        tb_jmp_cache_clear_page(cpu, d.addr);
        d.addr += TARGET_PAGE_SIZE;
    }
}

static void tlb_flush_range_by_mmuidx_async_1(CPUState *cpu,
                                              run_on_cpu_data data)
{
    TLBFlushRangeData *d = data.host_ptr;
    tlb_flush_range_by_mmuidx_async_0(cpu, *d);
    g_free(d);
}

void tlb_flush_range_by_mmuidx(CPUState *cpu, vaddr addr,
                               vaddr len, uint16_t idxmap,
                               unsigned bits)
{
    TLBFlushRangeData d;

    /*
     * If all bits are significant, and len is small,
     * this devolves to tlb_flush_page.
     */
    if (bits >= TARGET_LONG_BITS && len <= TARGET_PAGE_SIZE) {
        tlb_flush_page_by_mmuidx(cpu, addr, idxmap);
        return;
    }
    /* If no page bits are significant, this devolves to tlb_flush. */
    if (bits < TARGET_PAGE_BITS) {
        tlb_flush_by_mmuidx(cpu, idxmap);
        return;
    }

    /* This should already be page aligned */
    d.addr = addr & TARGET_PAGE_MASK;
    d.len = len;
    d.idxmap = idxmap;
    d.bits = bits;

    if (qemu_cpu_is_self(cpu)) {
        tlb_flush_range_by_mmuidx_async_0(cpu, d);
    } else {
        /* Otherwise allocate a structure, freed by the worker.  */
        TLBFlushRangeData *p = g_memdup(&d, sizeof(d));
        async_run_on_cpu(cpu, tlb_flush_range_by_mmuidx_async_1,
                         RUN_ON_CPU_HOST_PTR(p));
    }
}

void tlb_flush_page_bits_by_mmuidx(CPUState *cpu, vaddr addr,
                                   uint16_t idxmap, unsigned bits)
{
    tlb_flush_range_by_mmuidx(cpu, addr, TARGET_PAGE_SIZE, idxmap, bits);
}

void tlb_flush_range_by_mmuidx_all_cpus(CPUState *src_cpu,
                                        vaddr addr, vaddr len,
                                        uint16_t idxmap, unsigned bits)
{
    TLBFlushRangeData d;
    CPUState *dst_cpu;

    /*
     * If all bits are significant, and len is small,
     * this devolves to tlb_flush_page.
     */
    if (bits >= TARGET_LONG_BITS && len <= TARGET_PAGE_SIZE) {
        tlb_flush_page_by_mmuidx_all_cpus(src_cpu, addr, idxmap);
        return;
    }
    /* If no page bits are significant, this devolves to tlb_flush. */
    if (bits < TARGET_PAGE_BITS) {
        tlb_flush_by_mmuidx_all_cpus(src_cpu, idxmap);
        return;
    }

    /* This should already be page aligned */
    d.addr = addr & TARGET_PAGE_MASK;
    d.len = len;
    d.idxmap = idxmap;
    d.bits = bits;

    /* Allocate a separate data block for each destination cpu.  */
    CPU_FOREACH(dst_cpu) {
        if (dst_cpu != src_cpu) {
            TLBFlushRangeData *p = g_memdup(&d, sizeof(d));
            async_run_on_cpu(dst_cpu,
                             tlb_flush_range_by_mmuidx_async_1,
                             RUN_ON_CPU_HOST_PTR(p));
        }
    }

    tlb_flush_range_by_mmuidx_async_0(src_cpu, d);
}

void tlb_flush_page_bits_by_mmuidx_all_cpus(CPUState *src_cpu,
                                            vaddr addr, uint16_t idxmap,
                                            unsigned bits)
{
    tlb_flush_range_by_mmuidx_all_cpus(src_cpu, addr, TARGET_PAGE_SIZE,
                                       idxmap, bits);
}

void tlb_flush_range_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
                                               vaddr addr,
                                               vaddr len,
                                               uint16_t idxmap,
                                               unsigned bits)
{
    TLBFlushRangeData d, *p;
    CPUState *dst_cpu;

    /*
     * If all bits are significant, and len is small,
     * this devolves to tlb_flush_page.
     */
    if (bits >= TARGET_LONG_BITS && len <= TARGET_PAGE_SIZE) {
        tlb_flush_page_by_mmuidx_all_cpus_synced(src_cpu, addr, idxmap);
        return;
    }
    /* If no page bits are significant, this devolves to tlb_flush. */
    if (bits < TARGET_PAGE_BITS) {
        tlb_flush_by_mmuidx_all_cpus_synced(src_cpu, idxmap);
        return;
    }

    /* This should already be page aligned */
    d.addr = addr & TARGET_PAGE_MASK;
    d.len = len;
    d.idxmap = idxmap;
    d.bits = bits;

    /* Allocate a separate data block for each destination cpu.  */
    CPU_FOREACH(dst_cpu) {
        if (dst_cpu != src_cpu) {
            p = g_memdup(&d, sizeof(d));
            async_run_on_cpu(dst_cpu, tlb_flush_range_by_mmuidx_async_1,
                             RUN_ON_CPU_HOST_PTR(p));
        }
    }

    p = g_memdup(&d, sizeof(d));
    async_safe_run_on_cpu(src_cpu, tlb_flush_range_by_mmuidx_async_1,
                          RUN_ON_CPU_HOST_PTR(p));
}

void tlb_flush_page_bits_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
                                                   vaddr addr,
                                                   uint16_t idxmap,
                                                   unsigned bits)
{
    tlb_flush_range_by_mmuidx_all_cpus_synced(src_cpu, addr, TARGET_PAGE_SIZE,
                                              idxmap, bits);
}

/* update the TLBs so that writes to code in the virtual page 'addr'
   can be detected */
void tlb_protect_code(ram_addr_t ram_addr)
{
    cpu_physical_memory_test_and_clear_dirty(ram_addr & TARGET_PAGE_MASK,
                                             TARGET_PAGE_SIZE,
                                             DIRTY_MEMORY_CODE);
}

/* update the TLB so that writes in physical page 'phys_addr' are no longer
   tested for self modifying code */
void tlb_unprotect_code(ram_addr_t ram_addr)
{
    cpu_physical_memory_set_dirty_flag(ram_addr, DIRTY_MEMORY_CODE);
}


/*
 * Dirty write flag handling
 *
 * When the TCG code writes to a location it looks up the address in
 * the TLB and uses that data to compute the final address. If any of
 * the lower bits of the address are set then the slow path is forced.
 * There are a number of reasons to do this but for normal RAM the
 * most usual is detecting writes to code regions which may invalidate
 * generated code.
 *
 * Other vCPUs might be reading their TLBs during guest execution, so we update
 * te->addr_write with qatomic_set. We don't need to worry about this for
 * oversized guests as MTTCG is disabled for them.
 *
 * Called with tlb_c.lock held.
 */
static void tlb_reset_dirty_range_locked(CPUTLBEntry *tlb_entry,
                                         uintptr_t start, uintptr_t length)
{
    uintptr_t addr = tlb_entry->addr_write;

    if ((addr & (TLB_INVALID_MASK | TLB_MMIO |
                 TLB_DISCARD_WRITE | TLB_NOTDIRTY)) == 0) {
        addr &= TARGET_PAGE_MASK;
        addr += tlb_entry->addend;
        if ((addr - start) < length) {
#if TARGET_LONG_BITS == 32
            uint32_t *ptr_write = (uint32_t *)&tlb_entry->addr_write;
            ptr_write += HOST_BIG_ENDIAN;
            qatomic_set(ptr_write, *ptr_write | TLB_NOTDIRTY);
#elif TCG_OVERSIZED_GUEST
            tlb_entry->addr_write |= TLB_NOTDIRTY;
#else
            qatomic_set(&tlb_entry->addr_write,
                        tlb_entry->addr_write | TLB_NOTDIRTY);
#endif
        }
    }
}

/*
 * Called with tlb_c.lock held.
 * Called only from the vCPU context, i.e. the TLB's owner thread.
 */
static inline void copy_tlb_helper_locked(CPUTLBEntry *d, const CPUTLBEntry *s)
{
    *d = *s;
}

/* This is a cross vCPU call (i.e. another vCPU resetting the flags of
 * the target vCPU).
 * We must take tlb_c.lock to avoid racing with another vCPU update. The only
 * thing actually updated is the target TLB entry ->addr_write flags.
 */
void tlb_reset_dirty(CPUState *cpu, ram_addr_t start1, ram_addr_t length)
{
    CPUArchState *env;

    int mmu_idx;

    env = cpu->env_ptr;
    qemu_spin_lock(&env_tlb(env)->c.lock);
    for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) {
        unsigned int i;
        unsigned int n = tlb_n_entries(&env_tlb(env)->f[mmu_idx]);

        for (i = 0; i < n; i++) {
            tlb_reset_dirty_range_locked(&env_tlb(env)->f[mmu_idx].table[i],
                                         start1, length);
        }

        for (i = 0; i < CPU_VTLB_SIZE; i++) {
            tlb_reset_dirty_range_locked(&env_tlb(env)->d[mmu_idx].vtable[i],
                                         start1, length);
        }
    }
    qemu_spin_unlock(&env_tlb(env)->c.lock);
}

/* Called with tlb_c.lock held */
static inline void tlb_set_dirty1_locked(CPUTLBEntry *tlb_entry,
                                         vaddr addr)
{
    if (tlb_entry->addr_write == (addr | TLB_NOTDIRTY)) {
        tlb_entry->addr_write = addr;
    }
}

/* update the TLB corresponding to virtual page vaddr
   so that it is no longer dirty */
void tlb_set_dirty(CPUState *cpu, vaddr addr)
{
    CPUArchState *env = cpu->env_ptr;
    int mmu_idx;

    assert_cpu_is_self(cpu);

    addr &= TARGET_PAGE_MASK;
    qemu_spin_lock(&env_tlb(env)->c.lock);
    for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) {
        tlb_set_dirty1_locked(tlb_entry(env, mmu_idx, addr), addr);
    }

    for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) {
        int k;
        for (k = 0; k < CPU_VTLB_SIZE; k++) {
            tlb_set_dirty1_locked(&env_tlb(env)->d[mmu_idx].vtable[k], addr);
        }
    }
    qemu_spin_unlock(&env_tlb(env)->c.lock);
}

/* Our TLB does not support large pages, so remember the area covered by
   large pages and trigger a full TLB flush if these are invalidated.  */
static void tlb_add_large_page(CPUArchState *env, int mmu_idx,
                               vaddr addr, uint64_t size)
{
    vaddr lp_addr = env_tlb(env)->d[mmu_idx].large_page_addr;
    vaddr lp_mask = ~(size - 1);

    if (lp_addr == (vaddr)-1) {
        /* No previous large page.  */
        lp_addr = addr;
    } else {
        /* Extend the existing region to include the new page.
           This is a compromise between unnecessary flushes and
           the cost of maintaining a full variable size TLB.  */
        lp_mask &= env_tlb(env)->d[mmu_idx].large_page_mask;
        while (((lp_addr ^ addr) & lp_mask) != 0) {
            lp_mask <<= 1;
        }
    }
    env_tlb(env)->d[mmu_idx].large_page_addr = lp_addr & lp_mask;
    env_tlb(env)->d[mmu_idx].large_page_mask = lp_mask;
}

static inline void tlb_set_compare(CPUTLBEntryFull *full, CPUTLBEntry *ent,
                                   target_ulong address, int flags,
                                   MMUAccessType access_type, bool enable)
{
    if (enable) {
        address |= flags & TLB_FLAGS_MASK;
        flags &= TLB_SLOW_FLAGS_MASK;
        if (flags) {
            address |= TLB_FORCE_SLOW;
        }
    } else {
        address = -1;
        flags = 0;
    }
    ent->addr_idx[access_type] = address;
    full->slow_flags[access_type] = flags;
}

/*
 * Add a new TLB entry. At most one entry for a given virtual address
 * is permitted. Only a single TARGET_PAGE_SIZE region is mapped, the
 * supplied size is only used by tlb_flush_page.
 *
 * Called from TCG-generated code, which is under an RCU read-side
 * critical section.
 */
void tlb_set_page_full(CPUState *cpu, int mmu_idx,
                       vaddr addr, CPUTLBEntryFull *full)
{
    CPUArchState *env = cpu->env_ptr;
    CPUTLB *tlb = env_tlb(env);
    CPUTLBDesc *desc = &tlb->d[mmu_idx];
    MemoryRegionSection *section;
    unsigned int index, read_flags, write_flags;
    uintptr_t addend;
    CPUTLBEntry *te, tn;
    hwaddr iotlb, xlat, sz, paddr_page;
    vaddr addr_page;
    int asidx, wp_flags, prot;
    bool is_ram, is_romd;

    assert_cpu_is_self(cpu);

    if (full->lg_page_size <= TARGET_PAGE_BITS) {
        sz = TARGET_PAGE_SIZE;
    } else {
        sz = (hwaddr)1 << full->lg_page_size;
        tlb_add_large_page(env, mmu_idx, addr, sz);
    }
    addr_page = addr & TARGET_PAGE_MASK;
    paddr_page = full->phys_addr & TARGET_PAGE_MASK;

    prot = full->prot;
    asidx = cpu_asidx_from_attrs(cpu, full->attrs);
    section = address_space_translate_for_iotlb(cpu, asidx, paddr_page,
                                                &xlat, &sz, full->attrs, &prot);
    assert(sz >= TARGET_PAGE_SIZE);

    tlb_debug("vaddr=%" VADDR_PRIx " paddr=0x" HWADDR_FMT_plx
              " prot=%x idx=%d\n",
              addr, full->phys_addr, prot, mmu_idx);

    read_flags = 0;
    if (full->lg_page_size < TARGET_PAGE_BITS) {
        /* Repeat the MMU check and TLB fill on every access.  */
        read_flags |= TLB_INVALID_MASK;
    }
    if (full->attrs.byte_swap) {
        read_flags |= TLB_BSWAP;
    }

    is_ram = memory_region_is_ram(section->mr);
    is_romd = memory_region_is_romd(section->mr);

    if (is_ram || is_romd) {
        /* RAM and ROMD both have associated host memory. */
        addend = (uintptr_t)memory_region_get_ram_ptr(section->mr) + xlat;
    } else {
        /* I/O does not; force the host address to NULL. */
        addend = 0;
    }

    write_flags = read_flags;
    if (is_ram) {
        iotlb = memory_region_get_ram_addr(section->mr) + xlat;
        /*
         * Computing is_clean is expensive; avoid all that unless
         * the page is actually writable.
         */
        if (prot & PAGE_WRITE) {
            if (section->readonly) {
                write_flags |= TLB_DISCARD_WRITE;
            } else if (cpu_physical_memory_is_clean(iotlb)) {
                write_flags |= TLB_NOTDIRTY;
            }
        }
    } else {
        /* I/O or ROMD */
        iotlb = memory_region_section_get_iotlb(cpu, section) + xlat;
        /*
         * Writes to romd devices must go through MMIO to enable write.
         * Reads to romd devices go through the ram_ptr found above,
         * but of course reads to I/O must go through MMIO.
         */
        write_flags |= TLB_MMIO;
        if (!is_romd) {
            read_flags = write_flags;
        }
    }

    wp_flags = cpu_watchpoint_address_matches(cpu, addr_page,
                                              TARGET_PAGE_SIZE);

    index = tlb_index(env, mmu_idx, addr_page);
    te = tlb_entry(env, mmu_idx, addr_page);

    /*
     * Hold the TLB lock for the rest of the function. We could acquire/release
     * the lock several times in the function, but it is faster to amortize the
     * acquisition cost by acquiring it just once. Note that this leads to
     * a longer critical section, but this is not a concern since the TLB lock
     * is unlikely to be contended.
     */
    qemu_spin_lock(&tlb->c.lock);

    /* Note that the tlb is no longer clean.  */
    tlb->c.dirty |= 1 << mmu_idx;

    /* Make sure there's no cached translation for the new page.  */
    tlb_flush_vtlb_page_locked(env, mmu_idx, addr_page);

    /*
     * Only evict the old entry to the victim tlb if it's for a
     * different page; otherwise just overwrite the stale data.
     */
    if (!tlb_hit_page_anyprot(te, addr_page) && !tlb_entry_is_empty(te)) {
        unsigned vidx = desc->vindex++ % CPU_VTLB_SIZE;
        CPUTLBEntry *tv = &desc->vtable[vidx];

        /* Evict the old entry into the victim tlb.  */
        copy_tlb_helper_locked(tv, te);
        desc->vfulltlb[vidx] = desc->fulltlb[index];
        tlb_n_used_entries_dec(env, mmu_idx);
    }

    /* refill the tlb */
    /*
     * At this point iotlb contains a physical section number in the lower
     * TARGET_PAGE_BITS, and either
     *  + the ram_addr_t of the page base of the target RAM (RAM)
     *  + the offset within section->mr of the page base (I/O, ROMD)
     * We subtract addr_page (which is page aligned and thus won't
     * disturb the low bits) to give an offset which can be added to the
     * (non-page-aligned) vaddr of the eventual memory access to get
     * the MemoryRegion offset for the access. Note that the vaddr we
     * subtract here is that of the page base, and not the same as the
     * vaddr we add back in io_readx()/io_writex()/get_page_addr_code().
     */
    desc->fulltlb[index] = *full;
    full = &desc->fulltlb[index];
    full->xlat_section = iotlb - addr_page;
    full->phys_addr = paddr_page;

    /* Now calculate the new entry */
    tn.addend = addend - addr_page;

    tlb_set_compare(full, &tn, addr_page, read_flags,
                    MMU_INST_FETCH, prot & PAGE_EXEC);

    if (wp_flags & BP_MEM_READ) {
        read_flags |= TLB_WATCHPOINT;
    }
    tlb_set_compare(full, &tn, addr_page, read_flags,
                    MMU_DATA_LOAD, prot & PAGE_READ);

    if (prot & PAGE_WRITE_INV) {
        write_flags |= TLB_INVALID_MASK;
    }
    if (wp_flags & BP_MEM_WRITE) {
        write_flags |= TLB_WATCHPOINT;
    }
    tlb_set_compare(full, &tn, addr_page, write_flags,
                    MMU_DATA_STORE, prot & PAGE_WRITE);

    copy_tlb_helper_locked(te, &tn);
    tlb_n_used_entries_inc(env, mmu_idx);
    qemu_spin_unlock(&tlb->c.lock);
}

void tlb_set_page_with_attrs(CPUState *cpu, vaddr addr,
                             hwaddr paddr, MemTxAttrs attrs, int prot,
                             int mmu_idx, uint64_t size)
{
    CPUTLBEntryFull full = {
        .phys_addr = paddr,
        .attrs = attrs,
        .prot = prot,
        .lg_page_size = ctz64(size)
    };

    assert(is_power_of_2(size));
    tlb_set_page_full(cpu, mmu_idx, addr, &full);
}

void tlb_set_page(CPUState *cpu, vaddr addr,
                  hwaddr paddr, int prot,
                  int mmu_idx, uint64_t size)
{
    tlb_set_page_with_attrs(cpu, addr, paddr, MEMTXATTRS_UNSPECIFIED,
                            prot, mmu_idx, size);
}

/*
 * Note: tlb_fill() can trigger a resize of the TLB. This means that all of the
 * caller's prior references to the TLB table (e.g. CPUTLBEntry pointers) must
 * be discarded and looked up again (e.g. via tlb_entry()).
 */
static void tlb_fill(CPUState *cpu, vaddr addr, int size,
                     MMUAccessType access_type, int mmu_idx, uintptr_t retaddr)
{
    bool ok;

    /*
     * This is not a probe, so only valid return is success; failure
     * should result in exception + longjmp to the cpu loop.
     */
    ok = cpu->cc->tcg_ops->tlb_fill(cpu, addr, size,
                                    access_type, mmu_idx, false, retaddr);
    assert(ok);
}

static inline void cpu_unaligned_access(CPUState *cpu, vaddr addr,
                                        MMUAccessType access_type,
                                        int mmu_idx, uintptr_t retaddr)
{
    cpu->cc->tcg_ops->do_unaligned_access(cpu, addr, access_type,
                                          mmu_idx, retaddr);
}

static inline void cpu_transaction_failed(CPUState *cpu, hwaddr physaddr,
                                          vaddr addr, unsigned size,
                                          MMUAccessType access_type,
                                          int mmu_idx, MemTxAttrs attrs,
                                          MemTxResult response,
                                          uintptr_t retaddr)
{
    CPUClass *cc = CPU_GET_CLASS(cpu);

    if (!cpu->ignore_memory_transaction_failures &&
        cc->tcg_ops->do_transaction_failed) {
        cc->tcg_ops->do_transaction_failed(cpu, physaddr, addr, size,
                                           access_type, mmu_idx, attrs,
                                           response, retaddr);
    }
}

static uint64_t io_readx(CPUArchState *env, CPUTLBEntryFull *full,
                         int mmu_idx, vaddr addr, uintptr_t retaddr,
                         MMUAccessType access_type, MemOp op)
{
    CPUState *cpu = env_cpu(env);
    hwaddr mr_offset;
    MemoryRegionSection *section;
    MemoryRegion *mr;
    uint64_t val;
    MemTxResult r;

    section = iotlb_to_section(cpu, full->xlat_section, full->attrs);
    mr = section->mr;
    mr_offset = (full->xlat_section & TARGET_PAGE_MASK) + addr;
    cpu->mem_io_pc = retaddr;
    if (!cpu->can_do_io) {
        cpu_io_recompile(cpu, retaddr);
    }

    {
        QEMU_IOTHREAD_LOCK_GUARD();
        r = memory_region_dispatch_read(mr, mr_offset, &val, op, full->attrs);
    }

    if (r != MEMTX_OK) {
        hwaddr physaddr = mr_offset +
            section->offset_within_address_space -
            section->offset_within_region;

        cpu_transaction_failed(cpu, physaddr, addr, memop_size(op), access_type,
                               mmu_idx, full->attrs, r, retaddr);
    }
    return val;
}

/*
 * Save a potentially trashed CPUTLBEntryFull for later lookup by plugin.
 * This is read by tlb_plugin_lookup if the fulltlb entry doesn't match
 * because of the side effect of io_writex changing memory layout.
 */
static void save_iotlb_data(CPUState *cs, MemoryRegionSection *section,
                            hwaddr mr_offset)
{
#ifdef CONFIG_PLUGIN
    SavedIOTLB *saved = &cs->saved_iotlb;
    saved->section = section;
    saved->mr_offset = mr_offset;
#endif
}

static void io_writex(CPUArchState *env, CPUTLBEntryFull *full,
                      int mmu_idx, uint64_t val, vaddr addr,
                      uintptr_t retaddr, MemOp op)
{
    CPUState *cpu = env_cpu(env);
    hwaddr mr_offset;
    MemoryRegionSection *section;
    MemoryRegion *mr;
    MemTxResult r;

    section = iotlb_to_section(cpu, full->xlat_section, full->attrs);
    mr = section->mr;
    mr_offset = (full->xlat_section & TARGET_PAGE_MASK) + addr;
    if (!cpu->can_do_io) {
        cpu_io_recompile(cpu, retaddr);
    }
    cpu->mem_io_pc = retaddr;

    /*
     * The memory_region_dispatch may trigger a flush/resize
     * so for plugins we save the iotlb_data just in case.
     */
    save_iotlb_data(cpu, section, mr_offset);

    {
        QEMU_IOTHREAD_LOCK_GUARD();
        r = memory_region_dispatch_write(mr, mr_offset, val, op, full->attrs);
    }

    if (r != MEMTX_OK) {
        hwaddr physaddr = mr_offset +
            section->offset_within_address_space -
            section->offset_within_region;

        cpu_transaction_failed(cpu, physaddr, addr, memop_size(op),
                               MMU_DATA_STORE, mmu_idx, full->attrs, r,
                               retaddr);
    }
}

/* Return true if ADDR is present in the victim tlb, and has been copied
   back to the main tlb.  */
static bool victim_tlb_hit(CPUArchState *env, size_t mmu_idx, size_t index,
                           MMUAccessType access_type, vaddr page)
{
    size_t vidx;

    assert_cpu_is_self(env_cpu(env));
    for (vidx = 0; vidx < CPU_VTLB_SIZE; ++vidx) {
        CPUTLBEntry *vtlb = &env_tlb(env)->d[mmu_idx].vtable[vidx];
        uint64_t cmp = tlb_read_idx(vtlb, access_type);

        if (cmp == page) {
            /* Found entry in victim tlb, swap tlb and iotlb.  */
            CPUTLBEntry tmptlb, *tlb = &env_tlb(env)->f[mmu_idx].table[index];

            qemu_spin_lock(&env_tlb(env)->c.lock);
            copy_tlb_helper_locked(&tmptlb, tlb);
            copy_tlb_helper_locked(tlb, vtlb);
            copy_tlb_helper_locked(vtlb, &tmptlb);
            qemu_spin_unlock(&env_tlb(env)->c.lock);

            CPUTLBEntryFull *f1 = &env_tlb(env)->d[mmu_idx].fulltlb[index];
            CPUTLBEntryFull *f2 = &env_tlb(env)->d[mmu_idx].vfulltlb[vidx];
            CPUTLBEntryFull tmpf;
            tmpf = *f1; *f1 = *f2; *f2 = tmpf;
            return true;
        }
    }
    return false;
}

static void notdirty_write(CPUState *cpu, vaddr mem_vaddr, unsigned size,
                           CPUTLBEntryFull *full, uintptr_t retaddr)
{
    ram_addr_t ram_addr = mem_vaddr + full->xlat_section;

    trace_memory_notdirty_write_access(mem_vaddr, ram_addr, size);

    if (!cpu_physical_memory_get_dirty_flag(ram_addr, DIRTY_MEMORY_CODE)) {
        tb_invalidate_phys_range_fast(ram_addr, size, retaddr);
    }

    /*
     * Set both VGA and migration bits for simplicity and to remove
     * the notdirty callback faster.
     */
    cpu_physical_memory_set_dirty_range(ram_addr, size, DIRTY_CLIENTS_NOCODE);

    /* We remove the notdirty callback only if the code has been flushed. */
    if (!cpu_physical_memory_is_clean(ram_addr)) {
        trace_memory_notdirty_set_dirty(mem_vaddr);
        tlb_set_dirty(cpu, mem_vaddr);
    }
}

static int probe_access_internal(CPUArchState *env, vaddr addr,
                                 int fault_size, MMUAccessType access_type,
                                 int mmu_idx, bool nonfault,
                                 void **phost, CPUTLBEntryFull **pfull,
                                 uintptr_t retaddr)
{
    uintptr_t index = tlb_index(env, mmu_idx, addr);
    CPUTLBEntry *entry = tlb_entry(env, mmu_idx, addr);
    uint64_t tlb_addr = tlb_read_idx(entry, access_type);
    vaddr page_addr = addr & TARGET_PAGE_MASK;
    int flags = TLB_FLAGS_MASK & ~TLB_FORCE_SLOW;
    CPUTLBEntryFull *full;

    if (!tlb_hit_page(tlb_addr, page_addr)) {
        if (!victim_tlb_hit(env, mmu_idx, index, access_type, page_addr)) {
            CPUState *cs = env_cpu(env);

            if (!cs->cc->tcg_ops->tlb_fill(cs, addr, fault_size, access_type,
                                           mmu_idx, nonfault, retaddr)) {
                /* Non-faulting page table read failed.  */
                *phost = NULL;
                *pfull = NULL;
                return TLB_INVALID_MASK;
            }

            /* TLB resize via tlb_fill may have moved the entry.  */
            index = tlb_index(env, mmu_idx, addr);
            entry = tlb_entry(env, mmu_idx, addr);

            /*
             * With PAGE_WRITE_INV, we set TLB_INVALID_MASK immediately,
             * to force the next access through tlb_fill.  We've just
             * called tlb_fill, so we know that this entry *is* valid.
             */
            flags &= ~TLB_INVALID_MASK;
        }
        tlb_addr = tlb_read_idx(entry, access_type);
    }
    flags &= tlb_addr;

    *pfull = full = &env_tlb(env)->d[mmu_idx].fulltlb[index];
    flags |= full->slow_flags[access_type];

    /* Fold all "mmio-like" bits into TLB_MMIO.  This is not RAM.  */
    if (unlikely(flags & ~(TLB_WATCHPOINT | TLB_NOTDIRTY))) {
        *phost = NULL;
        return TLB_MMIO;
    }

    /* Everything else is RAM. */
    *phost = (void *)((uintptr_t)addr + entry->addend);
    return flags;
}

int probe_access_full(CPUArchState *env, vaddr addr, int size,
                      MMUAccessType access_type, int mmu_idx,
                      bool nonfault, void **phost, CPUTLBEntryFull **pfull,
                      uintptr_t retaddr)
{
    int flags = probe_access_internal(env, addr, size, access_type, mmu_idx,
                                      nonfault, phost, pfull, retaddr);

    /* Handle clean RAM pages.  */
    if (unlikely(flags & TLB_NOTDIRTY)) {
        notdirty_write(env_cpu(env), addr, 1, *pfull, retaddr);
        flags &= ~TLB_NOTDIRTY;
    }

    return flags;
}

int probe_access_flags(CPUArchState *env, vaddr addr, int size,
                       MMUAccessType access_type, int mmu_idx,
                       bool nonfault, void **phost, uintptr_t retaddr)
{
    CPUTLBEntryFull *full;
    int flags;

    g_assert(-(addr | TARGET_PAGE_MASK) >= size);

    flags = probe_access_internal(env, addr, size, access_type, mmu_idx,
                                  nonfault, phost, &full, retaddr);

    /* Handle clean RAM pages. */
    if (unlikely(flags & TLB_NOTDIRTY)) {
        notdirty_write(env_cpu(env), addr, 1, full, retaddr);
        flags &= ~TLB_NOTDIRTY;
    }

    return flags;
}

void *probe_access(CPUArchState *env, vaddr addr, int size,
                   MMUAccessType access_type, int mmu_idx, uintptr_t retaddr)
{
    CPUTLBEntryFull *full;
    void *host;
    int flags;

    g_assert(-(addr | TARGET_PAGE_MASK) >= size);

    flags = probe_access_internal(env, addr, size, access_type, mmu_idx,
                                  false, &host, &full, retaddr);

    /* Per the interface, size == 0 merely faults the access. */
    if (size == 0) {
        return NULL;
    }

    if (unlikely(flags & (TLB_NOTDIRTY | TLB_WATCHPOINT))) {
        /* Handle watchpoints.  */
        if (flags & TLB_WATCHPOINT) {
            int wp_access = (access_type == MMU_DATA_STORE
                             ? BP_MEM_WRITE : BP_MEM_READ);
            cpu_check_watchpoint(env_cpu(env), addr, size,
                                 full->attrs, wp_access, retaddr);
        }

        /* Handle clean RAM pages.  */
        if (flags & TLB_NOTDIRTY) {
            notdirty_write(env_cpu(env), addr, 1, full, retaddr);
        }
    }

    return host;
}

void *tlb_vaddr_to_host(CPUArchState *env, abi_ptr addr,
                        MMUAccessType access_type, int mmu_idx)
{
    CPUTLBEntryFull *full;
    void *host;
    int flags;

    flags = probe_access_internal(env, addr, 0, access_type,
                                  mmu_idx, true, &host, &full, 0);

    /* No combination of flags are expected by the caller. */
    return flags ? NULL : host;
}

/*
 * Return a ram_addr_t for the virtual address for execution.
 *
 * Return -1 if we can't translate and execute from an entire page
 * of RAM.  This will force us to execute by loading and translating
 * one insn at a time, without caching.
 *
 * NOTE: This function will trigger an exception if the page is
 * not executable.
 */
tb_page_addr_t get_page_addr_code_hostp(CPUArchState *env, vaddr addr,
                                        void **hostp)
{
    CPUTLBEntryFull *full;
    void *p;

    (void)probe_access_internal(env, addr, 1, MMU_INST_FETCH,
                                cpu_mmu_index(env, true), false, &p, &full, 0);
    if (p == NULL) {
        return -1;
    }

    if (full->lg_page_size < TARGET_PAGE_BITS) {
        return -1;
    }

    if (hostp) {
        *hostp = p;
    }
    return qemu_ram_addr_from_host_nofail(p);
}

/* Load/store with atomicity primitives. */
#include "ldst_atomicity.c.inc"

#ifdef CONFIG_PLUGIN
/*
 * Perform a TLB lookup and populate the qemu_plugin_hwaddr structure.
 * This should be a hot path as we will have just looked this path up
 * in the softmmu lookup code (or helper). We don't handle re-fills or
 * checking the victim table. This is purely informational.
 *
 * This almost never fails as the memory access being instrumented
 * should have just filled the TLB. The one corner case is io_writex
 * which can cause TLB flushes and potential resizing of the TLBs
 * losing the information we need. In those cases we need to recover
 * data from a copy of the CPUTLBEntryFull. As long as this always occurs
 * from the same thread (which a mem callback will be) this is safe.
 */

bool tlb_plugin_lookup(CPUState *cpu, vaddr addr, int mmu_idx,
                       bool is_store, struct qemu_plugin_hwaddr *data)
{
    CPUArchState *env = cpu->env_ptr;
    CPUTLBEntry *tlbe = tlb_entry(env, mmu_idx, addr);
    uintptr_t index = tlb_index(env, mmu_idx, addr);
    uint64_t tlb_addr = is_store ? tlb_addr_write(tlbe) : tlbe->addr_read;

    if (likely(tlb_hit(tlb_addr, addr))) {
        /* We must have an iotlb entry for MMIO */
        if (tlb_addr & TLB_MMIO) {
            CPUTLBEntryFull *full;
            full = &env_tlb(env)->d[mmu_idx].fulltlb[index];
            data->is_io = true;
            data->v.io.section =
                iotlb_to_section(cpu, full->xlat_section, full->attrs);
            data->v.io.offset = (full->xlat_section & TARGET_PAGE_MASK) + addr;
        } else {
            data->is_io = false;
            data->v.ram.hostaddr = (void *)((uintptr_t)addr + tlbe->addend);
        }
        return true;
    } else {
        SavedIOTLB *saved = &cpu->saved_iotlb;
        data->is_io = true;
        data->v.io.section = saved->section;
        data->v.io.offset = saved->mr_offset;
        return true;
    }
}

#endif

/*
 * Probe for a load/store operation.
 * Return the host address and into @flags.
 */

typedef struct MMULookupPageData {
    CPUTLBEntryFull *full;
    void *haddr;
    vaddr addr;
    int flags;
    int size;
} MMULookupPageData;

typedef struct MMULookupLocals {
    MMULookupPageData page[2];
    MemOp memop;
    int mmu_idx;
} MMULookupLocals;

/**
 * mmu_lookup1: translate one page
 * @env: cpu context
 * @data: lookup parameters
 * @mmu_idx: virtual address context
 * @access_type: load/store/code
 * @ra: return address into tcg generated code, or 0
 *
 * Resolve the translation for the one page at @data.addr, filling in
 * the rest of @data with the results.  If the translation fails,
 * tlb_fill will longjmp out.  Return true if the softmmu tlb for
 * @mmu_idx may have resized.
 */
static bool mmu_lookup1(CPUArchState *env, MMULookupPageData *data,
                        int mmu_idx, MMUAccessType access_type, uintptr_t ra)
{
    vaddr addr = data->addr;
    uintptr_t index = tlb_index(env, mmu_idx, addr);
    CPUTLBEntry *entry = tlb_entry(env, mmu_idx, addr);
    uint64_t tlb_addr = tlb_read_idx(entry, access_type);
    bool maybe_resized = false;
    CPUTLBEntryFull *full;
    int flags;

    /* If the TLB entry is for a different page, reload and try again.  */
    if (!tlb_hit(tlb_addr, addr)) {
        if (!victim_tlb_hit(env, mmu_idx, index, access_type,
                            addr & TARGET_PAGE_MASK)) {
            tlb_fill(env_cpu(env), addr, data->size, access_type, mmu_idx, ra);
            maybe_resized = true;
            index = tlb_index(env, mmu_idx, addr);
            entry = tlb_entry(env, mmu_idx, addr);
        }
        tlb_addr = tlb_read_idx(entry, access_type) & ~TLB_INVALID_MASK;
    }

    full = &env_tlb(env)->d[mmu_idx].fulltlb[index];
    flags = tlb_addr & (TLB_FLAGS_MASK & ~TLB_FORCE_SLOW);
    flags |= full->slow_flags[access_type];

    data->full = full;
    data->flags = flags;
    /* Compute haddr speculatively; depending on flags it might be invalid. */
    data->haddr = (void *)((uintptr_t)addr + entry->addend);

    return maybe_resized;
}

/**
 * mmu_watch_or_dirty
 * @env: cpu context
 * @data: lookup parameters
 * @access_type: load/store/code
 * @ra: return address into tcg generated code, or 0
 *
 * Trigger watchpoints for @data.addr:@data.size;
 * record writes to protected clean pages.
 */
static void mmu_watch_or_dirty(CPUArchState *env, MMULookupPageData *data,
                               MMUAccessType access_type, uintptr_t ra)
{
    CPUTLBEntryFull *full = data->full;
    vaddr addr = data->addr;
    int flags = data->flags;
    int size = data->size;

    /* On watchpoint hit, this will longjmp out.  */
    if (flags & TLB_WATCHPOINT) {
        int wp = access_type == MMU_DATA_STORE ? BP_MEM_WRITE : BP_MEM_READ;
        cpu_check_watchpoint(env_cpu(env), addr, size, full->attrs, wp, ra);
        flags &= ~TLB_WATCHPOINT;
    }

    /* Note that notdirty is only set for writes. */
    if (flags & TLB_NOTDIRTY) {
        notdirty_write(env_cpu(env), addr, size, full, ra);
        flags &= ~TLB_NOTDIRTY;
    }
    data->flags = flags;
}

/**
 * mmu_lookup: translate page(s)
 * @env: cpu context
 * @addr: virtual address
 * @oi: combined mmu_idx and MemOp
 * @ra: return address into tcg generated code, or 0
 * @access_type: load/store/code
 * @l: output result
 *
 * Resolve the translation for the page(s) beginning at @addr, for MemOp.size
 * bytes.  Return true if the lookup crosses a page boundary.
 */
static bool mmu_lookup(CPUArchState *env, vaddr addr, MemOpIdx oi,
                       uintptr_t ra, MMUAccessType type, MMULookupLocals *l)
{
    unsigned a_bits;
    bool crosspage;
    int flags;

    l->memop = get_memop(oi);
    l->mmu_idx = get_mmuidx(oi);

    tcg_debug_assert(l->mmu_idx < NB_MMU_MODES);

    /* Handle CPU specific unaligned behaviour */
    a_bits = get_alignment_bits(l->memop);
    if (addr & ((1 << a_bits) - 1)) {
        cpu_unaligned_access(env_cpu(env), addr, type, l->mmu_idx, ra);
    }

    l->page[0].addr = addr;
    l->page[0].size = memop_size(l->memop);
    l->page[1].addr = (addr + l->page[0].size - 1) & TARGET_PAGE_MASK;
    l->page[1].size = 0;
    crosspage = (addr ^ l->page[1].addr) & TARGET_PAGE_MASK;

    if (likely(!crosspage)) {
        mmu_lookup1(env, &l->page[0], l->mmu_idx, type, ra);

        flags = l->page[0].flags;
        if (unlikely(flags & (TLB_WATCHPOINT | TLB_NOTDIRTY))) {
            mmu_watch_or_dirty(env, &l->page[0], type, ra);
        }
        if (unlikely(flags & TLB_BSWAP)) {
            l->memop ^= MO_BSWAP;
        }
    } else {
        /* Finish compute of page crossing. */
        int size0 = l->page[1].addr - addr;
        l->page[1].size = l->page[0].size - size0;
        l->page[0].size = size0;

        /*
         * Lookup both pages, recognizing exceptions from either.  If the
         * second lookup potentially resized, refresh first CPUTLBEntryFull.
         */
        mmu_lookup1(env, &l->page[0], l->mmu_idx, type, ra);
        if (mmu_lookup1(env, &l->page[1], l->mmu_idx, type, ra)) {
            uintptr_t index = tlb_index(env, l->mmu_idx, addr);
            l->page[0].full = &env_tlb(env)->d[l->mmu_idx].fulltlb[index];
        }

        flags = l->page[0].flags | l->page[1].flags;
        if (unlikely(flags & (TLB_WATCHPOINT | TLB_NOTDIRTY))) {
            mmu_watch_or_dirty(env, &l->page[0], type, ra);
            mmu_watch_or_dirty(env, &l->page[1], type, ra);
        }

        /*
         * Since target/sparc is the only user of TLB_BSWAP, and all
         * Sparc accesses are aligned, any treatment across two pages
         * would be arbitrary.  Refuse it until there's a use.
         */
        tcg_debug_assert((flags & TLB_BSWAP) == 0);
    }

    return crosspage;
}

/*
 * Probe for an atomic operation.  Do not allow unaligned operations,
 * or io operations to proceed.  Return the host address.
 */
static void *atomic_mmu_lookup(CPUArchState *env, vaddr addr, MemOpIdx oi,
                               int size, uintptr_t retaddr)
{
    uintptr_t mmu_idx = get_mmuidx(oi);
    MemOp mop = get_memop(oi);
    int a_bits = get_alignment_bits(mop);
    uintptr_t index;
    CPUTLBEntry *tlbe;
    vaddr tlb_addr;
    void *hostaddr;
    CPUTLBEntryFull *full;

    tcg_debug_assert(mmu_idx < NB_MMU_MODES);

    /* Adjust the given return address.  */
    retaddr -= GETPC_ADJ;

    /* Enforce guest required alignment.  */
    if (unlikely(a_bits > 0 && (addr & ((1 << a_bits) - 1)))) {
        /* ??? Maybe indicate atomic op to cpu_unaligned_access */
        cpu_unaligned_access(env_cpu(env), addr, MMU_DATA_STORE,
                             mmu_idx, retaddr);
    }

    /* Enforce qemu required alignment.  */
    if (unlikely(addr & (size - 1))) {
        /* We get here if guest alignment was not requested,
           or was not enforced by cpu_unaligned_access above.
           We might widen the access and emulate, but for now
           mark an exception and exit the cpu loop.  */
        goto stop_the_world;
    }

    index = tlb_index(env, mmu_idx, addr);
    tlbe = tlb_entry(env, mmu_idx, addr);

    /* Check TLB entry and enforce page permissions.  */
    tlb_addr = tlb_addr_write(tlbe);
    if (!tlb_hit(tlb_addr, addr)) {
        if (!victim_tlb_hit(env, mmu_idx, index, MMU_DATA_STORE,
                            addr & TARGET_PAGE_MASK)) {
            tlb_fill(env_cpu(env), addr, size,
                     MMU_DATA_STORE, mmu_idx, retaddr);
            index = tlb_index(env, mmu_idx, addr);
            tlbe = tlb_entry(env, mmu_idx, addr);
        }
        tlb_addr = tlb_addr_write(tlbe) & ~TLB_INVALID_MASK;
    }

    /*
     * Let the guest notice RMW on a write-only page.
     * We have just verified that the page is writable.
     * Subpage lookups may have left TLB_INVALID_MASK set,
     * but addr_read will only be -1 if PAGE_READ was unset.
     */
    if (unlikely(tlbe->addr_read == -1)) {
        tlb_fill(env_cpu(env), addr, size, MMU_DATA_LOAD, mmu_idx, retaddr);
        /*
         * Since we don't support reads and writes to different
         * addresses, and we do have the proper page loaded for
         * write, this shouldn't ever return.  But just in case,
         * handle via stop-the-world.
         */
        goto stop_the_world;
    }
    /* Collect tlb flags for read. */
    tlb_addr |= tlbe->addr_read;

    /* Notice an IO access or a needs-MMU-lookup access */
    if (unlikely(tlb_addr & (TLB_MMIO | TLB_DISCARD_WRITE))) {
        /* There's really nothing that can be done to
           support this apart from stop-the-world.  */
        goto stop_the_world;
    }

    hostaddr = (void *)((uintptr_t)addr + tlbe->addend);
    full = &env_tlb(env)->d[mmu_idx].fulltlb[index];

    if (unlikely(tlb_addr & TLB_NOTDIRTY)) {
        notdirty_write(env_cpu(env), addr, size, full, retaddr);
    }

    if (unlikely(tlb_addr & TLB_FORCE_SLOW)) {
        int wp_flags = 0;

        if (full->slow_flags[MMU_DATA_STORE] & TLB_WATCHPOINT) {
            wp_flags |= BP_MEM_WRITE;
        }
        if (full->slow_flags[MMU_DATA_LOAD] & TLB_WATCHPOINT) {
            wp_flags |= BP_MEM_READ;
        }
        if (wp_flags) {
            cpu_check_watchpoint(env_cpu(env), addr, size,
                                 full->attrs, wp_flags, retaddr);
        }
    }

    return hostaddr;

 stop_the_world:
    cpu_loop_exit_atomic(env_cpu(env), retaddr);
}

/*
 * Load Helpers
 *
 * We support two different access types. SOFTMMU_CODE_ACCESS is
 * specifically for reading instructions from system memory. It is
 * called by the translation loop and in some helpers where the code
 * is disassembled. It shouldn't be called directly by guest code.
 *
 * For the benefit of TCG generated code, we want to avoid the
 * complication of ABI-specific return type promotion and always
 * return a value extended to the register size of the host. This is
 * tcg_target_long, except in the case of a 32-bit host and 64-bit
 * data, and for that we always have uint64_t.
 *
 * We don't bother with this widened value for SOFTMMU_CODE_ACCESS.
 */

/**
 * do_ld_mmio_beN:
 * @env: cpu context
 * @p: translation parameters
 * @ret_be: accumulated data
 * @mmu_idx: virtual address context
 * @ra: return address into tcg generated code, or 0
 *
 * Load @p->size bytes from @p->addr, which is memory-mapped i/o.
 * The bytes are concatenated in big-endian order with @ret_be.
 */
static uint64_t do_ld_mmio_beN(CPUArchState *env, MMULookupPageData *p,
                               uint64_t ret_be, int mmu_idx,
                               MMUAccessType type, uintptr_t ra)
{
    CPUTLBEntryFull *full = p->full;
    vaddr addr = p->addr;
    int i, size = p->size;

    QEMU_IOTHREAD_LOCK_GUARD();
    for (i = 0; i < size; i++) {
        uint8_t x = io_readx(env, full, mmu_idx, addr + i, ra, type, MO_UB);
        ret_be = (ret_be << 8) | x;
    }
    return ret_be;
}

/**
 * do_ld_bytes_beN
 * @p: translation parameters
 * @ret_be: accumulated data
 *
 * Load @p->size bytes from @p->haddr, which is RAM.
 * The bytes to concatenated in big-endian order with @ret_be.
 */
static uint64_t do_ld_bytes_beN(MMULookupPageData *p, uint64_t ret_be)
{
    uint8_t *haddr = p->haddr;
    int i, size = p->size;

    for (i = 0; i < size; i++) {
        ret_be = (ret_be << 8) | haddr[i];
    }
    return ret_be;
}

/**
 * do_ld_parts_beN
 * @p: translation parameters
 * @ret_be: accumulated data
 *
 * As do_ld_bytes_beN, but atomically on each aligned part.
 */
static uint64_t do_ld_parts_beN(MMULookupPageData *p, uint64_t ret_be)
{
    void *haddr = p->haddr;
    int size = p->size;

    do {
        uint64_t x;
        int n;

        /*
         * Find minimum of alignment and size.
         * This is slightly stronger than required by MO_ATOM_SUBALIGN, which
         * would have only checked the low bits of addr|size once at the start,
         * but is just as easy.
         */
        switch (((uintptr_t)haddr | size) & 7) {
        case 4:
            x = cpu_to_be32(load_atomic4(haddr));
            ret_be = (ret_be << 32) | x;
            n = 4;
            break;
        case 2:
        case 6:
            x = cpu_to_be16(load_atomic2(haddr));
            ret_be = (ret_be << 16) | x;
            n = 2;
            break;
        default:
            x = *(uint8_t *)haddr;
            ret_be = (ret_be << 8) | x;
            n = 1;
            break;
        case 0:
            g_assert_not_reached();
        }
        haddr += n;
        size -= n;
    } while (size != 0);
    return ret_be;
}

/**
 * do_ld_parts_be4
 * @p: translation parameters
 * @ret_be: accumulated data
 *
 * As do_ld_bytes_beN, but with one atomic load.
 * Four aligned bytes are guaranteed to cover the load.
 */
static uint64_t do_ld_whole_be4(MMULookupPageData *p, uint64_t ret_be)
{
    int o = p->addr & 3;
    uint32_t x = load_atomic4(p->haddr - o);

    x = cpu_to_be32(x);
    x <<= o * 8;
    x >>= (4 - p->size) * 8;
    return (ret_be << (p->size * 8)) | x;
}

/**
 * do_ld_parts_be8
 * @p: translation parameters
 * @ret_be: accumulated data
 *
 * As do_ld_bytes_beN, but with one atomic load.
 * Eight aligned bytes are guaranteed to cover the load.
 */
static uint64_t do_ld_whole_be8(CPUArchState *env, uintptr_t ra,
                                MMULookupPageData *p, uint64_t ret_be)
{
    int o = p->addr & 7;
    uint64_t x = load_atomic8_or_exit(env, ra, p->haddr - o);

    x = cpu_to_be64(x);
    x <<= o * 8;
    x >>= (8 - p->size) * 8;
    return (ret_be << (p->size * 8)) | x;
}

/**
 * do_ld_parts_be16
 * @p: translation parameters
 * @ret_be: accumulated data
 *
 * As do_ld_bytes_beN, but with one atomic load.
 * 16 aligned bytes are guaranteed to cover the load.
 */
static Int128 do_ld_whole_be16(CPUArchState *env, uintptr_t ra,
                               MMULookupPageData *p, uint64_t ret_be)
{
    int o = p->addr & 15;
    Int128 x, y = load_atomic16_or_exit(env, ra, p->haddr - o);
    int size = p->size;

    if (!HOST_BIG_ENDIAN) {
        y = bswap128(y);
    }
    y = int128_lshift(y, o * 8);
    y = int128_urshift(y, (16 - size) * 8);
    x = int128_make64(ret_be);
    x = int128_lshift(x, size * 8);
    return int128_or(x, y);
}

/*
 * Wrapper for the above.
 */
static uint64_t do_ld_beN(CPUArchState *env, MMULookupPageData *p,
                          uint64_t ret_be, int mmu_idx, MMUAccessType type,
                          MemOp mop, uintptr_t ra)
{
    MemOp atom;
    unsigned tmp, half_size;

    if (unlikely(p->flags & TLB_MMIO)) {
        return do_ld_mmio_beN(env, p, ret_be, mmu_idx, type, ra);
    }

    /*
     * It is a given that we cross a page and therefore there is no
     * atomicity for the load as a whole, but subobjects may need attention.
     */
    atom = mop & MO_ATOM_MASK;
    switch (atom) {
    case MO_ATOM_SUBALIGN:
        return do_ld_parts_beN(p, ret_be);

    case MO_ATOM_IFALIGN_PAIR:
    case MO_ATOM_WITHIN16_PAIR:
        tmp = mop & MO_SIZE;
        tmp = tmp ? tmp - 1 : 0;
        half_size = 1 << tmp;
        if (atom == MO_ATOM_IFALIGN_PAIR
            ? p->size == half_size
            : p->size >= half_size) {
            if (!HAVE_al8_fast && p->size < 4) {
                return do_ld_whole_be4(p, ret_be);
            } else {
                return do_ld_whole_be8(env, ra, p, ret_be);
            }
        }
        /* fall through */

    case MO_ATOM_IFALIGN:
    case MO_ATOM_WITHIN16:
    case MO_ATOM_NONE:
        return do_ld_bytes_beN(p, ret_be);

    default:
        g_assert_not_reached();
    }
}

/*
 * Wrapper for the above, for 8 < size < 16.
 */
static Int128 do_ld16_beN(CPUArchState *env, MMULookupPageData *p,
                          uint64_t a, int mmu_idx, MemOp mop, uintptr_t ra)
{
    int size = p->size;
    uint64_t b;
    MemOp atom;

    if (unlikely(p->flags & TLB_MMIO)) {
        p->size = size - 8;
        a = do_ld_mmio_beN(env, p, a, mmu_idx, MMU_DATA_LOAD, ra);
        p->addr += p->size;
        p->size = 8;
        b = do_ld_mmio_beN(env, p, 0, mmu_idx, MMU_DATA_LOAD, ra);
        return int128_make128(b, a);
    }

    /*
     * It is a given that we cross a page and therefore there is no
     * atomicity for the load as a whole, but subobjects may need attention.
     */
    atom = mop & MO_ATOM_MASK;
    switch (atom) {
    case MO_ATOM_SUBALIGN:
        p->size = size - 8;
        a = do_ld_parts_beN(p, a);
        p->haddr += size - 8;
        p->size = 8;
        b = do_ld_parts_beN(p, 0);
        break;

    case MO_ATOM_WITHIN16_PAIR:
        /* Since size > 8, this is the half that must be atomic. */
        return do_ld_whole_be16(env, ra, p, a);

    case MO_ATOM_IFALIGN_PAIR:
        /*
         * Since size > 8, both halves are misaligned,
         * and so neither is atomic.
         */
    case MO_ATOM_IFALIGN:
    case MO_ATOM_WITHIN16:
    case MO_ATOM_NONE:
        p->size = size - 8;
        a = do_ld_bytes_beN(p, a);
        b = ldq_be_p(p->haddr + size - 8);
        break;

    default:
        g_assert_not_reached();
    }

    return int128_make128(b, a);
}

static uint8_t do_ld_1(CPUArchState *env, MMULookupPageData *p, int mmu_idx,
                       MMUAccessType type, uintptr_t ra)
{
    if (unlikely(p->flags & TLB_MMIO)) {
        return io_readx(env, p->full, mmu_idx, p->addr, ra, type, MO_UB);
    } else {
        return *(uint8_t *)p->haddr;
    }
}

static uint16_t do_ld_2(CPUArchState *env, MMULookupPageData *p, int mmu_idx,
                        MMUAccessType type, MemOp memop, uintptr_t ra)
{
    uint64_t ret;

    if (unlikely(p->flags & TLB_MMIO)) {
        return io_readx(env, p->full, mmu_idx, p->addr, ra, type, memop);
    }

    /* Perform the load host endian, then swap if necessary. */
    ret = load_atom_2(env, ra, p->haddr, memop);
    if (memop & MO_BSWAP) {
        ret = bswap16(ret);
    }
    return ret;
}

static uint32_t do_ld_4(CPUArchState *env, MMULookupPageData *p, int mmu_idx,
                        MMUAccessType type, MemOp memop, uintptr_t ra)
{
    uint32_t ret;

    if (unlikely(p->flags & TLB_MMIO)) {
        return io_readx(env, p->full, mmu_idx, p->addr, ra, type, memop);
    }

    /* Perform the load host endian. */
    ret = load_atom_4(env, ra, p->haddr, memop);
    if (memop & MO_BSWAP) {
        ret = bswap32(ret);
    }
    return ret;
}

static uint64_t do_ld_8(CPUArchState *env, MMULookupPageData *p, int mmu_idx,
                        MMUAccessType type, MemOp memop, uintptr_t ra)
{
    uint64_t ret;

    if (unlikely(p->flags & TLB_MMIO)) {
        return io_readx(env, p->full, mmu_idx, p->addr, ra, type, memop);
    }

    /* Perform the load host endian. */
    ret = load_atom_8(env, ra, p->haddr, memop);
    if (memop & MO_BSWAP) {
        ret = bswap64(ret);
    }
    return ret;
}

static uint8_t do_ld1_mmu(CPUArchState *env, vaddr addr, MemOpIdx oi,
                          uintptr_t ra, MMUAccessType access_type)
{
    MMULookupLocals l;
    bool crosspage;

    cpu_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD);
    crosspage = mmu_lookup(env, addr, oi, ra, access_type, &l);
    tcg_debug_assert(!crosspage);

    return do_ld_1(env, &l.page[0], l.mmu_idx, access_type, ra);
}

tcg_target_ulong helper_ldub_mmu(CPUArchState *env, uint64_t addr,
                                 MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_8);
    return do_ld1_mmu(env, addr, oi, retaddr, MMU_DATA_LOAD);
}

static uint16_t do_ld2_mmu(CPUArchState *env, vaddr addr, MemOpIdx oi,
                           uintptr_t ra, MMUAccessType access_type)
{
    MMULookupLocals l;
    bool crosspage;
    uint16_t ret;
    uint8_t a, b;

    cpu_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD);
    crosspage = mmu_lookup(env, addr, oi, ra, access_type, &l);
    if (likely(!crosspage)) {
        return do_ld_2(env, &l.page[0], l.mmu_idx, access_type, l.memop, ra);
    }

    a = do_ld_1(env, &l.page[0], l.mmu_idx, access_type, ra);
    b = do_ld_1(env, &l.page[1], l.mmu_idx, access_type, ra);

    if ((l.memop & MO_BSWAP) == MO_LE) {
        ret = a | (b << 8);
    } else {
        ret = b | (a << 8);
    }
    return ret;
}

tcg_target_ulong helper_lduw_mmu(CPUArchState *env, uint64_t addr,
                                 MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_16);
    return do_ld2_mmu(env, addr, oi, retaddr, MMU_DATA_LOAD);
}

static uint32_t do_ld4_mmu(CPUArchState *env, vaddr addr, MemOpIdx oi,
                           uintptr_t ra, MMUAccessType access_type)
{
    MMULookupLocals l;
    bool crosspage;
    uint32_t ret;

    cpu_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD);
    crosspage = mmu_lookup(env, addr, oi, ra, access_type, &l);
    if (likely(!crosspage)) {
        return do_ld_4(env, &l.page[0], l.mmu_idx, access_type, l.memop, ra);
    }

    ret = do_ld_beN(env, &l.page[0], 0, l.mmu_idx, access_type, l.memop, ra);
    ret = do_ld_beN(env, &l.page[1], ret, l.mmu_idx, access_type, l.memop, ra);
    if ((l.memop & MO_BSWAP) == MO_LE) {
        ret = bswap32(ret);
    }
    return ret;
}

tcg_target_ulong helper_ldul_mmu(CPUArchState *env, uint64_t addr,
                                 MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_32);
    return do_ld4_mmu(env, addr, oi, retaddr, MMU_DATA_LOAD);
}

static uint64_t do_ld8_mmu(CPUArchState *env, vaddr addr, MemOpIdx oi,
                           uintptr_t ra, MMUAccessType access_type)
{
    MMULookupLocals l;
    bool crosspage;
    uint64_t ret;

    cpu_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD);
    crosspage = mmu_lookup(env, addr, oi, ra, access_type, &l);
    if (likely(!crosspage)) {
        return do_ld_8(env, &l.page[0], l.mmu_idx, access_type, l.memop, ra);
    }

    ret = do_ld_beN(env, &l.page[0], 0, l.mmu_idx, access_type, l.memop, ra);
    ret = do_ld_beN(env, &l.page[1], ret, l.mmu_idx, access_type, l.memop, ra);
    if ((l.memop & MO_BSWAP) == MO_LE) {
        ret = bswap64(ret);
    }
    return ret;
}

uint64_t helper_ldq_mmu(CPUArchState *env, uint64_t addr,
                        MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_64);
    return do_ld8_mmu(env, addr, oi, retaddr, MMU_DATA_LOAD);
}

/*
 * Provide signed versions of the load routines as well.  We can of course
 * avoid this for 64-bit data, or for 32-bit data on 32-bit host.
 */

tcg_target_ulong helper_ldsb_mmu(CPUArchState *env, uint64_t addr,
                                 MemOpIdx oi, uintptr_t retaddr)
{
    return (int8_t)helper_ldub_mmu(env, addr, oi, retaddr);
}

tcg_target_ulong helper_ldsw_mmu(CPUArchState *env, uint64_t addr,
                                 MemOpIdx oi, uintptr_t retaddr)
{
    return (int16_t)helper_lduw_mmu(env, addr, oi, retaddr);
}

tcg_target_ulong helper_ldsl_mmu(CPUArchState *env, uint64_t addr,
                                 MemOpIdx oi, uintptr_t retaddr)
{
    return (int32_t)helper_ldul_mmu(env, addr, oi, retaddr);
}

static Int128 do_ld16_mmu(CPUArchState *env, vaddr addr,
                          MemOpIdx oi, uintptr_t ra)
{
    MMULookupLocals l;
    bool crosspage;
    uint64_t a, b;
    Int128 ret;
    int first;

    cpu_req_mo(TCG_MO_LD_LD | TCG_MO_ST_LD);
    crosspage = mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD, &l);
    if (likely(!crosspage)) {
        /* Perform the load host endian. */
        if (unlikely(l.page[0].flags & TLB_MMIO)) {
            QEMU_IOTHREAD_LOCK_GUARD();
            a = io_readx(env, l.page[0].full, l.mmu_idx, addr,
                         ra, MMU_DATA_LOAD, MO_64);
            b = io_readx(env, l.page[0].full, l.mmu_idx, addr + 8,
                         ra, MMU_DATA_LOAD, MO_64);
            ret = int128_make128(HOST_BIG_ENDIAN ? b : a,
                                 HOST_BIG_ENDIAN ? a : b);
        } else {
            ret = load_atom_16(env, ra, l.page[0].haddr, l.memop);
        }
        if (l.memop & MO_BSWAP) {
            ret = bswap128(ret);
        }
        return ret;
    }

    first = l.page[0].size;
    if (first == 8) {
        MemOp mop8 = (l.memop & ~MO_SIZE) | MO_64;

        a = do_ld_8(env, &l.page[0], l.mmu_idx, MMU_DATA_LOAD, mop8, ra);
        b = do_ld_8(env, &l.page[1], l.mmu_idx, MMU_DATA_LOAD, mop8, ra);
        if ((mop8 & MO_BSWAP) == MO_LE) {
            ret = int128_make128(a, b);
        } else {
            ret = int128_make128(b, a);
        }
        return ret;
    }

    if (first < 8) {
        a = do_ld_beN(env, &l.page[0], 0, l.mmu_idx,
                      MMU_DATA_LOAD, l.memop, ra);
        ret = do_ld16_beN(env, &l.page[1], a, l.mmu_idx, l.memop, ra);
    } else {
        ret = do_ld16_beN(env, &l.page[0], 0, l.mmu_idx, l.memop, ra);
        b = int128_getlo(ret);
        ret = int128_lshift(ret, l.page[1].size * 8);
        a = int128_gethi(ret);
        b = do_ld_beN(env, &l.page[1], b, l.mmu_idx,
                      MMU_DATA_LOAD, l.memop, ra);
        ret = int128_make128(b, a);
    }
    if ((l.memop & MO_BSWAP) == MO_LE) {
        ret = bswap128(ret);
    }
    return ret;
}

Int128 helper_ld16_mmu(CPUArchState *env, uint64_t addr,
                       uint32_t oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_128);
    return do_ld16_mmu(env, addr, oi, retaddr);
}

Int128 helper_ld_i128(CPUArchState *env, uint64_t addr, uint32_t oi)
{
    return helper_ld16_mmu(env, addr, oi, GETPC());
}

/*
 * Load helpers for cpu_ldst.h.
 */

static void plugin_load_cb(CPUArchState *env, abi_ptr addr, MemOpIdx oi)
{
    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
}

uint8_t cpu_ldb_mmu(CPUArchState *env, abi_ptr addr, MemOpIdx oi, uintptr_t ra)
{
    uint8_t ret;

    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_UB);
    ret = do_ld1_mmu(env, addr, oi, ra, MMU_DATA_LOAD);
    plugin_load_cb(env, addr, oi);
    return ret;
}

uint16_t cpu_ldw_mmu(CPUArchState *env, abi_ptr addr,
                     MemOpIdx oi, uintptr_t ra)
{
    uint16_t ret;

    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_16);
    ret = do_ld2_mmu(env, addr, oi, ra, MMU_DATA_LOAD);
    plugin_load_cb(env, addr, oi);
    return ret;
}

uint32_t cpu_ldl_mmu(CPUArchState *env, abi_ptr addr,
                     MemOpIdx oi, uintptr_t ra)
{
    uint32_t ret;

    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_32);
    ret = do_ld4_mmu(env, addr, oi, ra, MMU_DATA_LOAD);
    plugin_load_cb(env, addr, oi);
    return ret;
}

uint64_t cpu_ldq_mmu(CPUArchState *env, abi_ptr addr,
                     MemOpIdx oi, uintptr_t ra)
{
    uint64_t ret;

    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_64);
    ret = do_ld8_mmu(env, addr, oi, ra, MMU_DATA_LOAD);
    plugin_load_cb(env, addr, oi);
    return ret;
}

Int128 cpu_ld16_mmu(CPUArchState *env, abi_ptr addr,
                    MemOpIdx oi, uintptr_t ra)
{
    Int128 ret;

    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_128);
    ret = do_ld16_mmu(env, addr, oi, ra);
    plugin_load_cb(env, addr, oi);
    return ret;
}

/*
 * Store Helpers
 */

/**
 * do_st_mmio_leN:
 * @env: cpu context
 * @p: translation parameters
 * @val_le: data to store
 * @mmu_idx: virtual address context
 * @ra: return address into tcg generated code, or 0
 *
 * Store @p->size bytes at @p->addr, which is memory-mapped i/o.
 * The bytes to store are extracted in little-endian order from @val_le;
 * return the bytes of @val_le beyond @p->size that have not been stored.
 */
static uint64_t do_st_mmio_leN(CPUArchState *env, MMULookupPageData *p,
                               uint64_t val_le, int mmu_idx, uintptr_t ra)
{
    CPUTLBEntryFull *full = p->full;
    vaddr addr = p->addr;
    int i, size = p->size;

    QEMU_IOTHREAD_LOCK_GUARD();
    for (i = 0; i < size; i++, val_le >>= 8) {
        io_writex(env, full, mmu_idx, val_le, addr + i, ra, MO_UB);
    }
    return val_le;
}

/*
 * Wrapper for the above.
 */
static uint64_t do_st_leN(CPUArchState *env, MMULookupPageData *p,
                          uint64_t val_le, int mmu_idx,
                          MemOp mop, uintptr_t ra)
{
    MemOp atom;
    unsigned tmp, half_size;

    if (unlikely(p->flags & TLB_MMIO)) {
        return do_st_mmio_leN(env, p, val_le, mmu_idx, ra);
    } else if (unlikely(p->flags & TLB_DISCARD_WRITE)) {
        return val_le >> (p->size * 8);
    }

    /*
     * It is a given that we cross a page and therefore there is no atomicity
     * for the store as a whole, but subobjects may need attention.
     */
    atom = mop & MO_ATOM_MASK;
    switch (atom) {
    case MO_ATOM_SUBALIGN:
        return store_parts_leN(p->haddr, p->size, val_le);

    case MO_ATOM_IFALIGN_PAIR:
    case MO_ATOM_WITHIN16_PAIR:
        tmp = mop & MO_SIZE;
        tmp = tmp ? tmp - 1 : 0;
        half_size = 1 << tmp;
        if (atom == MO_ATOM_IFALIGN_PAIR
            ? p->size == half_size
            : p->size >= half_size) {
            if (!HAVE_al8_fast && p->size <= 4) {
                return store_whole_le4(p->haddr, p->size, val_le);
            } else if (HAVE_al8) {
                return store_whole_le8(p->haddr, p->size, val_le);
            } else {
                cpu_loop_exit_atomic(env_cpu(env), ra);
            }
        }
        /* fall through */

    case MO_ATOM_IFALIGN:
    case MO_ATOM_WITHIN16:
    case MO_ATOM_NONE:
        return store_bytes_leN(p->haddr, p->size, val_le);

    default:
        g_assert_not_reached();
    }
}

/*
 * Wrapper for the above, for 8 < size < 16.
 */
static uint64_t do_st16_leN(CPUArchState *env, MMULookupPageData *p,
                            Int128 val_le, int mmu_idx,
                            MemOp mop, uintptr_t ra)
{
    int size = p->size;
    MemOp atom;

    if (unlikely(p->flags & TLB_MMIO)) {
        p->size = 8;
        do_st_mmio_leN(env, p, int128_getlo(val_le), mmu_idx, ra);
        p->size = size - 8;
        p->addr += 8;
        return do_st_mmio_leN(env, p, int128_gethi(val_le), mmu_idx, ra);
    } else if (unlikely(p->flags & TLB_DISCARD_WRITE)) {
        return int128_gethi(val_le) >> ((size - 8) * 8);
    }

    /*
     * It is a given that we cross a page and therefore there is no atomicity
     * for the store as a whole, but subobjects may need attention.
     */
    atom = mop & MO_ATOM_MASK;
    switch (atom) {
    case MO_ATOM_SUBALIGN:
        store_parts_leN(p->haddr, 8, int128_getlo(val_le));
        return store_parts_leN(p->haddr + 8, p->size - 8,
                               int128_gethi(val_le));

    case MO_ATOM_WITHIN16_PAIR:
        /* Since size > 8, this is the half that must be atomic. */
        if (!HAVE_ATOMIC128_RW) {
            cpu_loop_exit_atomic(env_cpu(env), ra);
        }
        return store_whole_le16(p->haddr, p->size, val_le);

    case MO_ATOM_IFALIGN_PAIR:
        /*
         * Since size > 8, both halves are misaligned,
         * and so neither is atomic.
         */
    case MO_ATOM_IFALIGN:
    case MO_ATOM_WITHIN16:
    case MO_ATOM_NONE:
        stq_le_p(p->haddr, int128_getlo(val_le));
        return store_bytes_leN(p->haddr + 8, p->size - 8,
                               int128_gethi(val_le));

    default:
        g_assert_not_reached();
    }
}

static void do_st_1(CPUArchState *env, MMULookupPageData *p, uint8_t val,
                    int mmu_idx, uintptr_t ra)
{
    if (unlikely(p->flags & TLB_MMIO)) {
        io_writex(env, p->full, mmu_idx, val, p->addr, ra, MO_UB);
    } else if (unlikely(p->flags & TLB_DISCARD_WRITE)) {
        /* nothing */
    } else {
        *(uint8_t *)p->haddr = val;
    }
}

static void do_st_2(CPUArchState *env, MMULookupPageData *p, uint16_t val,
                    int mmu_idx, MemOp memop, uintptr_t ra)
{
    if (unlikely(p->flags & TLB_MMIO)) {
        io_writex(env, p->full, mmu_idx, val, p->addr, ra, memop);
    } else if (unlikely(p->flags & TLB_DISCARD_WRITE)) {
        /* nothing */
    } else {
        /* Swap to host endian if necessary, then store. */
        if (memop & MO_BSWAP) {
            val = bswap16(val);
        }
        store_atom_2(env, ra, p->haddr, memop, val);
    }
}

static void do_st_4(CPUArchState *env, MMULookupPageData *p, uint32_t val,
                    int mmu_idx, MemOp memop, uintptr_t ra)
{
    if (unlikely(p->flags & TLB_MMIO)) {
        io_writex(env, p->full, mmu_idx, val, p->addr, ra, memop);
    } else if (unlikely(p->flags & TLB_DISCARD_WRITE)) {
        /* nothing */
    } else {
        /* Swap to host endian if necessary, then store. */
        if (memop & MO_BSWAP) {
            val = bswap32(val);
        }
        store_atom_4(env, ra, p->haddr, memop, val);
    }
}

static void do_st_8(CPUArchState *env, MMULookupPageData *p, uint64_t val,
                    int mmu_idx, MemOp memop, uintptr_t ra)
{
    if (unlikely(p->flags & TLB_MMIO)) {
        io_writex(env, p->full, mmu_idx, val, p->addr, ra, memop);
    } else if (unlikely(p->flags & TLB_DISCARD_WRITE)) {
        /* nothing */
    } else {
        /* Swap to host endian if necessary, then store. */
        if (memop & MO_BSWAP) {
            val = bswap64(val);
        }
        store_atom_8(env, ra, p->haddr, memop, val);
    }
}

void helper_stb_mmu(CPUArchState *env, uint64_t addr, uint32_t val,
                    MemOpIdx oi, uintptr_t ra)
{
    MMULookupLocals l;
    bool crosspage;

    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_8);
    cpu_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST);
    crosspage = mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE, &l);
    tcg_debug_assert(!crosspage);

    do_st_1(env, &l.page[0], val, l.mmu_idx, ra);
}

static void do_st2_mmu(CPUArchState *env, vaddr addr, uint16_t val,
                       MemOpIdx oi, uintptr_t ra)
{
    MMULookupLocals l;
    bool crosspage;
    uint8_t a, b;

    cpu_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST);
    crosspage = mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE, &l);
    if (likely(!crosspage)) {
        do_st_2(env, &l.page[0], val, l.mmu_idx, l.memop, ra);
        return;
    }

    if ((l.memop & MO_BSWAP) == MO_LE) {
        a = val, b = val >> 8;
    } else {
        b = val, a = val >> 8;
    }
    do_st_1(env, &l.page[0], a, l.mmu_idx, ra);
    do_st_1(env, &l.page[1], b, l.mmu_idx, ra);
}

void helper_stw_mmu(CPUArchState *env, uint64_t addr, uint32_t val,
                    MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_16);
    do_st2_mmu(env, addr, val, oi, retaddr);
}

static void do_st4_mmu(CPUArchState *env, vaddr addr, uint32_t val,
                       MemOpIdx oi, uintptr_t ra)
{
    MMULookupLocals l;
    bool crosspage;

    cpu_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST);
    crosspage = mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE, &l);
    if (likely(!crosspage)) {
        do_st_4(env, &l.page[0], val, l.mmu_idx, l.memop, ra);
        return;
    }

    /* Swap to little endian for simplicity, then store by bytes. */
    if ((l.memop & MO_BSWAP) != MO_LE) {
        val = bswap32(val);
    }
    val = do_st_leN(env, &l.page[0], val, l.mmu_idx, l.memop, ra);
    (void) do_st_leN(env, &l.page[1], val, l.mmu_idx, l.memop, ra);
}

void helper_stl_mmu(CPUArchState *env, uint64_t addr, uint32_t val,
                    MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_32);
    do_st4_mmu(env, addr, val, oi, retaddr);
}

static void do_st8_mmu(CPUArchState *env, vaddr addr, uint64_t val,
                       MemOpIdx oi, uintptr_t ra)
{
    MMULookupLocals l;
    bool crosspage;

    cpu_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST);
    crosspage = mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE, &l);
    if (likely(!crosspage)) {
        do_st_8(env, &l.page[0], val, l.mmu_idx, l.memop, ra);
        return;
    }

    /* Swap to little endian for simplicity, then store by bytes. */
    if ((l.memop & MO_BSWAP) != MO_LE) {
        val = bswap64(val);
    }
    val = do_st_leN(env, &l.page[0], val, l.mmu_idx, l.memop, ra);
    (void) do_st_leN(env, &l.page[1], val, l.mmu_idx, l.memop, ra);
}

void helper_stq_mmu(CPUArchState *env, uint64_t addr, uint64_t val,
                    MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_64);
    do_st8_mmu(env, addr, val, oi, retaddr);
}

static void do_st16_mmu(CPUArchState *env, vaddr addr, Int128 val,
                        MemOpIdx oi, uintptr_t ra)
{
    MMULookupLocals l;
    bool crosspage;
    uint64_t a, b;
    int first;

    cpu_req_mo(TCG_MO_LD_ST | TCG_MO_ST_ST);
    crosspage = mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE, &l);
    if (likely(!crosspage)) {
        /* Swap to host endian if necessary, then store. */
        if (l.memop & MO_BSWAP) {
            val = bswap128(val);
        }
        if (unlikely(l.page[0].flags & TLB_MMIO)) {
            QEMU_IOTHREAD_LOCK_GUARD();
            if (HOST_BIG_ENDIAN) {
                b = int128_getlo(val), a = int128_gethi(val);
            } else {
                a = int128_getlo(val), b = int128_gethi(val);
            }
            io_writex(env, l.page[0].full, l.mmu_idx, a, addr, ra, MO_64);
            io_writex(env, l.page[0].full, l.mmu_idx, b, addr + 8, ra, MO_64);
        } else if (unlikely(l.page[0].flags & TLB_DISCARD_WRITE)) {
            /* nothing */
        } else {
            store_atom_16(env, ra, l.page[0].haddr, l.memop, val);
        }
        return;
    }

    first = l.page[0].size;
    if (first == 8) {
        MemOp mop8 = (l.memop & ~(MO_SIZE | MO_BSWAP)) | MO_64;

        if (l.memop & MO_BSWAP) {
            val = bswap128(val);
        }
        if (HOST_BIG_ENDIAN) {
            b = int128_getlo(val), a = int128_gethi(val);
        } else {
            a = int128_getlo(val), b = int128_gethi(val);
        }
        do_st_8(env, &l.page[0], a, l.mmu_idx, mop8, ra);
        do_st_8(env, &l.page[1], b, l.mmu_idx, mop8, ra);
        return;
    }

    if ((l.memop & MO_BSWAP) != MO_LE) {
        val = bswap128(val);
    }
    if (first < 8) {
        do_st_leN(env, &l.page[0], int128_getlo(val), l.mmu_idx, l.memop, ra);
        val = int128_urshift(val, first * 8);
        do_st16_leN(env, &l.page[1], val, l.mmu_idx, l.memop, ra);
    } else {
        b = do_st16_leN(env, &l.page[0], val, l.mmu_idx, l.memop, ra);
        do_st_leN(env, &l.page[1], b, l.mmu_idx, l.memop, ra);
    }
}

void helper_st16_mmu(CPUArchState *env, uint64_t addr, Int128 val,
                     MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_128);
    do_st16_mmu(env, addr, val, oi, retaddr);
}

void helper_st_i128(CPUArchState *env, uint64_t addr, Int128 val, MemOpIdx oi)
{
    helper_st16_mmu(env, addr, val, oi, GETPC());
}

/*
 * Store Helpers for cpu_ldst.h
 */

static void plugin_store_cb(CPUArchState *env, abi_ptr addr, MemOpIdx oi)
{
    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
}

void cpu_stb_mmu(CPUArchState *env, target_ulong addr, uint8_t val,
                 MemOpIdx oi, uintptr_t retaddr)
{
    helper_stb_mmu(env, addr, val, oi, retaddr);
    plugin_store_cb(env, addr, oi);
}

void cpu_stw_mmu(CPUArchState *env, target_ulong addr, uint16_t val,
                 MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_16);
    do_st2_mmu(env, addr, val, oi, retaddr);
    plugin_store_cb(env, addr, oi);
}

void cpu_stl_mmu(CPUArchState *env, target_ulong addr, uint32_t val,
                    MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_32);
    do_st4_mmu(env, addr, val, oi, retaddr);
    plugin_store_cb(env, addr, oi);
}

void cpu_stq_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
                 MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_64);
    do_st8_mmu(env, addr, val, oi, retaddr);
    plugin_store_cb(env, addr, oi);
}

void cpu_st16_mmu(CPUArchState *env, target_ulong addr, Int128 val,
                  MemOpIdx oi, uintptr_t retaddr)
{
    tcg_debug_assert((get_memop(oi) & MO_SIZE) == MO_128);
    do_st16_mmu(env, addr, val, oi, retaddr);
    plugin_store_cb(env, addr, oi);
}

#include "ldst_common.c.inc"

/*
 * First set of functions passes in OI and RETADDR.
 * This makes them callable from other helpers.
 */

#define ATOMIC_NAME(X) \
    glue(glue(glue(cpu_atomic_ ## X, SUFFIX), END), _mmu)

#define ATOMIC_MMU_CLEANUP

#include "atomic_common.c.inc"

#define DATA_SIZE 1
#include "atomic_template.h"

#define DATA_SIZE 2
#include "atomic_template.h"

#define DATA_SIZE 4
#include "atomic_template.h"

#ifdef CONFIG_ATOMIC64
#define DATA_SIZE 8
#include "atomic_template.h"
#endif

#if defined(CONFIG_ATOMIC128) || defined(CONFIG_CMPXCHG128)
#define DATA_SIZE 16
#include "atomic_template.h"
#endif

/* Code access functions.  */

uint32_t cpu_ldub_code(CPUArchState *env, abi_ptr addr)
{
    MemOpIdx oi = make_memop_idx(MO_UB, cpu_mmu_index(env, true));
    return do_ld1_mmu(env, addr, oi, 0, MMU_INST_FETCH);
}

uint32_t cpu_lduw_code(CPUArchState *env, abi_ptr addr)
{
    MemOpIdx oi = make_memop_idx(MO_TEUW, cpu_mmu_index(env, true));
    return do_ld2_mmu(env, addr, oi, 0, MMU_INST_FETCH);
}

uint32_t cpu_ldl_code(CPUArchState *env, abi_ptr addr)
{
    MemOpIdx oi = make_memop_idx(MO_TEUL, cpu_mmu_index(env, true));
    return do_ld4_mmu(env, addr, oi, 0, MMU_INST_FETCH);
}

uint64_t cpu_ldq_code(CPUArchState *env, abi_ptr addr)
{
    MemOpIdx oi = make_memop_idx(MO_TEUQ, cpu_mmu_index(env, true));
    return do_ld8_mmu(env, addr, oi, 0, MMU_INST_FETCH);
}

uint8_t cpu_ldb_code_mmu(CPUArchState *env, abi_ptr addr,
                         MemOpIdx oi, uintptr_t retaddr)
{
    return do_ld1_mmu(env, addr, oi, retaddr, MMU_INST_FETCH);
}

uint16_t cpu_ldw_code_mmu(CPUArchState *env, abi_ptr addr,
                          MemOpIdx oi, uintptr_t retaddr)
{
    return do_ld2_mmu(env, addr, oi, retaddr, MMU_INST_FETCH);
}

uint32_t cpu_ldl_code_mmu(CPUArchState *env, abi_ptr addr,
                          MemOpIdx oi, uintptr_t retaddr)
{
    return do_ld4_mmu(env, addr, oi, retaddr, MMU_INST_FETCH);
}

uint64_t cpu_ldq_code_mmu(CPUArchState *env, abi_ptr addr,
                          MemOpIdx oi, uintptr_t retaddr)
{
    return do_ld8_mmu(env, addr, oi, retaddr, MMU_INST_FETCH);
}