1 files changed, 266 insertions, 9 deletions

diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
index 2bbbb3a..42ab79c 100644
--- a/accel/tcg/cputlb.c
+++ b/accel/tcg/cputlb.c
@@ -409,12 +409,21 @@ 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,
+                                      target_ulong page, target_ulong 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,
                                         target_ulong page)
 {
-    return tlb_hit_page(tlb_entry->addr_read, page) ||
-           tlb_hit_page(tlb_addr_write(tlb_entry), page) ||
-           tlb_hit_page(tlb_entry->addr_code, page);
+    return tlb_hit_page_mask_anyprot(tlb_entry, page, -1);
 }
 
 /**
@@ -427,31 +436,45 @@ static inline bool tlb_entry_is_empty(const CPUTLBEntry *te)
 }
 
 /* Called with tlb_c.lock held */
-static inline bool tlb_flush_entry_locked(CPUTLBEntry *tlb_entry,
-                                          target_ulong page)
+static bool tlb_flush_entry_mask_locked(CPUTLBEntry *tlb_entry,
+                                        target_ulong page,
+                                        target_ulong mask)
 {
-    if (tlb_hit_page_anyprot(tlb_entry, page)) {
+    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,
+                                          target_ulong page)
+{
+    return tlb_flush_entry_mask_locked(tlb_entry, page, -1);
+}
+
 /* Called with tlb_c.lock held */
-static inline void tlb_flush_vtlb_page_locked(CPUArchState *env, int mmu_idx,
-                                              target_ulong page)
+static void tlb_flush_vtlb_page_mask_locked(CPUArchState *env, int mmu_idx,
+                                            target_ulong page,
+                                            target_ulong 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_locked(&d->vtable[k], page)) {
+        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,
+                                              target_ulong page)
+{
+    tlb_flush_vtlb_page_mask_locked(env, mmu_idx, page, -1);
+}
+
 static void tlb_flush_page_locked(CPUArchState *env, int midx,
                                   target_ulong page)
 {
@@ -666,6 +689,240 @@ void tlb_flush_page_all_cpus_synced(CPUState *src, target_ulong addr)
     tlb_flush_page_by_mmuidx_all_cpus_synced(src, addr, ALL_MMUIDX_BITS);
 }
 
+static void tlb_flush_page_bits_locked(CPUArchState *env, int midx,
+                                       target_ulong page, unsigned bits)
+{
+    CPUTLBDesc *d = &env_tlb(env)->d[midx];
+    CPUTLBDescFast *f = &env_tlb(env)->f[midx];
+    target_ulong 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 (mask < f->mask) {
+        tlb_debug("forcing full flush midx %d ("
+                  TARGET_FMT_lx "/" TARGET_FMT_lx ")\n",
+                  midx, page, mask);
+        tlb_flush_one_mmuidx_locked(env, midx, get_clock_realtime());
+        return;
+    }
+
+    /* Check if we need to flush due to large pages.  */
+    if ((page & d->large_page_mask) == d->large_page_addr) {
+        tlb_debug("forcing full flush midx %d ("
+                  TARGET_FMT_lx "/" TARGET_FMT_lx ")\n",
+                  midx, d->large_page_addr, d->large_page_mask);
+        tlb_flush_one_mmuidx_locked(env, midx, get_clock_realtime());
+        return;
+    }
+
+    if (tlb_flush_entry_mask_locked(tlb_entry(env, midx, page), page, mask)) {
+        tlb_n_used_entries_dec(env, midx);
+    }
+    tlb_flush_vtlb_page_mask_locked(env, midx, page, mask);
+}
+
+typedef struct {
+    target_ulong addr;
+    uint16_t idxmap;
+    uint16_t bits;
+} TLBFlushPageBitsByMMUIdxData;
+
+static void
+tlb_flush_page_bits_by_mmuidx_async_0(CPUState *cpu,
+                                      TLBFlushPageBitsByMMUIdxData d)
+{
+    CPUArchState *env = cpu->env_ptr;
+    int mmu_idx;
+
+    assert_cpu_is_self(cpu);
+
+    tlb_debug("page addr:" TARGET_FMT_lx "/%u mmu_map:0x%x\n",
+              d.addr, d.bits, 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_page_bits_locked(env, mmu_idx, d.addr, d.bits);
+        }
+    }
+    qemu_spin_unlock(&env_tlb(env)->c.lock);
+
+    tb_flush_jmp_cache(cpu, d.addr);
+}
+
+static bool encode_pbm_to_runon(run_on_cpu_data *out,
+                                TLBFlushPageBitsByMMUIdxData d)
+{
+    /* We need 6 bits to hold to hold @bits up to 63. */
+    if (d.idxmap <= MAKE_64BIT_MASK(0, TARGET_PAGE_BITS - 6)) {
+        *out = RUN_ON_CPU_TARGET_PTR(d.addr | (d.idxmap << 6) | d.bits);
+        return true;
+    }
+    return false;
+}
+
+static TLBFlushPageBitsByMMUIdxData
+decode_runon_to_pbm(run_on_cpu_data data)
+{
+    target_ulong addr_map_bits = (target_ulong) data.target_ptr;
+    return (TLBFlushPageBitsByMMUIdxData){
+        .addr = addr_map_bits & TARGET_PAGE_MASK,
+        .idxmap = (addr_map_bits & ~TARGET_PAGE_MASK) >> 6,
+        .bits = addr_map_bits & 0x3f
+    };
+}
+
+static void tlb_flush_page_bits_by_mmuidx_async_1(CPUState *cpu,
+                                                  run_on_cpu_data runon)
+{
+    tlb_flush_page_bits_by_mmuidx_async_0(cpu, decode_runon_to_pbm(runon));
+}
+
+static void tlb_flush_page_bits_by_mmuidx_async_2(CPUState *cpu,
+                                                  run_on_cpu_data data)
+{
+    TLBFlushPageBitsByMMUIdxData *d = data.host_ptr;
+    tlb_flush_page_bits_by_mmuidx_async_0(cpu, *d);
+    g_free(d);
+}
+
+void tlb_flush_page_bits_by_mmuidx(CPUState *cpu, target_ulong addr,
+                                   uint16_t idxmap, unsigned bits)
+{
+    TLBFlushPageBitsByMMUIdxData d;
+    run_on_cpu_data runon;
+
+    /* If all bits are significant, this devolves to tlb_flush_page. */
+    if (bits >= TARGET_LONG_BITS) {
+        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.idxmap = idxmap;
+    d.bits = bits;
+
+    if (qemu_cpu_is_self(cpu)) {
+        tlb_flush_page_bits_by_mmuidx_async_0(cpu, d);
+    } else if (encode_pbm_to_runon(&runon, d)) {
+        async_run_on_cpu(cpu, tlb_flush_page_bits_by_mmuidx_async_1, runon);
+    } else {
+        TLBFlushPageBitsByMMUIdxData *p
+            = g_new(TLBFlushPageBitsByMMUIdxData, 1);
+
+        /* Otherwise allocate a structure, freed by the worker.  */
+        *p = d;
+        async_run_on_cpu(cpu, tlb_flush_page_bits_by_mmuidx_async_2,
+                         RUN_ON_CPU_HOST_PTR(p));
+    }
+}
+
+void tlb_flush_page_bits_by_mmuidx_all_cpus(CPUState *src_cpu,
+                                            target_ulong addr,
+                                            uint16_t idxmap,
+                                            unsigned bits)
+{
+    TLBFlushPageBitsByMMUIdxData d;
+    run_on_cpu_data runon;
+
+    /* If all bits are significant, this devolves to tlb_flush_page. */
+    if (bits >= TARGET_LONG_BITS) {
+        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.idxmap = idxmap;
+    d.bits = bits;
+
+    if (encode_pbm_to_runon(&runon, d)) {
+        flush_all_helper(src_cpu, tlb_flush_page_bits_by_mmuidx_async_1, runon);
+    } else {
+        CPUState *dst_cpu;
+        TLBFlushPageBitsByMMUIdxData *p;
+
+        /* Allocate a separate data block for each destination cpu.  */
+        CPU_FOREACH(dst_cpu) {
+            if (dst_cpu != src_cpu) {
+                p = g_new(TLBFlushPageBitsByMMUIdxData, 1);
+                *p = d;
+                async_run_on_cpu(dst_cpu,
+                                 tlb_flush_page_bits_by_mmuidx_async_2,
+                                 RUN_ON_CPU_HOST_PTR(p));
+            }
+        }
+    }
+
+    tlb_flush_page_bits_by_mmuidx_async_0(src_cpu, d);
+}
+
+void tlb_flush_page_bits_by_mmuidx_all_cpus_synced(CPUState *src_cpu,
+                                                   target_ulong addr,
+                                                   uint16_t idxmap,
+                                                   unsigned bits)
+{
+    TLBFlushPageBitsByMMUIdxData d;
+    run_on_cpu_data runon;
+
+    /* If all bits are significant, this devolves to tlb_flush_page. */
+    if (bits >= TARGET_LONG_BITS) {
+        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.idxmap = idxmap;
+    d.bits = bits;
+
+    if (encode_pbm_to_runon(&runon, d)) {
+        flush_all_helper(src_cpu, tlb_flush_page_bits_by_mmuidx_async_1, runon);
+        async_safe_run_on_cpu(src_cpu, tlb_flush_page_bits_by_mmuidx_async_1,
+                              runon);
+    } else {
+        CPUState *dst_cpu;
+        TLBFlushPageBitsByMMUIdxData *p;
+
+        /* Allocate a separate data block for each destination cpu.  */
+        CPU_FOREACH(dst_cpu) {
+            if (dst_cpu != src_cpu) {
+                p = g_new(TLBFlushPageBitsByMMUIdxData, 1);
+                *p = d;
+                async_run_on_cpu(dst_cpu, tlb_flush_page_bits_by_mmuidx_async_2,
+                                 RUN_ON_CPU_HOST_PTR(p));
+            }
+        }
+
+        p = g_new(TLBFlushPageBitsByMMUIdxData, 1);
+        *p = d;
+        async_safe_run_on_cpu(src_cpu, tlb_flush_page_bits_by_mmuidx_async_2,
+                              RUN_ON_CPU_HOST_PTR(p));
+    }
+}
+
 /* 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)