]> git.proxmox.com Git - mirror_qemu.git/blobdiff - exec.c
vfio/pci: Pass an error object to vfio_msix_early_setup
[mirror_qemu.git] / exec.c
diff --git a/exec.c b/exec.c
index ce3fb9ec8ef731a74cdf81f378110239ed51e129..e63c5a1bd4d5ef4a14aaae5ea5040751407a8f5a 100644 (file)
--- a/exec.c
+++ b/exec.c
@@ -255,7 +255,7 @@ static void phys_page_set(AddressSpaceDispatch *d,
 /* Compact a non leaf page entry. Simply detect that the entry has a single child,
  * and update our entry so we can skip it and go directly to the destination.
  */
-static void phys_page_compact(PhysPageEntry *lp, Node *nodes, unsigned long *compacted)
+static void phys_page_compact(PhysPageEntry *lp, Node *nodes)
 {
     unsigned valid_ptr = P_L2_SIZE;
     int valid = 0;
@@ -275,7 +275,7 @@ static void phys_page_compact(PhysPageEntry *lp, Node *nodes, unsigned long *com
         valid_ptr = i;
         valid++;
         if (p[i].skip) {
-            phys_page_compact(&p[i], nodes, compacted);
+            phys_page_compact(&p[i], nodes);
         }
     }
 
@@ -307,10 +307,8 @@ static void phys_page_compact(PhysPageEntry *lp, Node *nodes, unsigned long *com
 
 static void phys_page_compact_all(AddressSpaceDispatch *d, int nodes_nb)
 {
-    DECLARE_BITMAP(compacted, nodes_nb);
-
     if (d->phys_map.skip) {
-        phys_page_compact(&d->phys_map, d->map.nodes, compacted);
+        phys_page_compact(&d->phys_map, d->map.nodes);
     }
 }
 
@@ -598,36 +596,11 @@ AddressSpace *cpu_get_address_space(CPUState *cpu, int asidx)
 }
 #endif
 
-static bool cpu_index_auto_assigned;
-
-static int cpu_get_free_index(void)
-{
-    CPUState *some_cpu;
-    int cpu_index = 0;
-
-    cpu_index_auto_assigned = true;
-    CPU_FOREACH(some_cpu) {
-        cpu_index++;
-    }
-    return cpu_index;
-}
-
 void cpu_exec_exit(CPUState *cpu)
 {
     CPUClass *cc = CPU_GET_CLASS(cpu);
 
-    cpu_list_lock();
-    if (!QTAILQ_IN_USE(cpu, node)) {
-        /* there is nothing to undo since cpu_exec_init() hasn't been called */
-        cpu_list_unlock();
-        return;
-    }
-
-    assert(!(cpu_index_auto_assigned && cpu != QTAILQ_LAST(&cpus, CPUTailQ)));
-
-    QTAILQ_REMOVE(&cpus, cpu, node);
-    cpu->cpu_index = UNASSIGNED_CPU_INDEX;
-    cpu_list_unlock();
+    cpu_list_remove(cpu);
 
     if (cc->vmsd != NULL) {
         vmstate_unregister(NULL, cc->vmsd, cpu);
@@ -663,15 +636,7 @@ void cpu_exec_init(CPUState *cpu, Error **errp)
     object_ref(OBJECT(cpu->memory));
 #endif
 
-    cpu_list_lock();
-    if (cpu->cpu_index == UNASSIGNED_CPU_INDEX) {
-        cpu->cpu_index = cpu_get_free_index();
-        assert(cpu->cpu_index != UNASSIGNED_CPU_INDEX);
-    } else {
-        assert(!cpu_index_auto_assigned);
-    }
-    QTAILQ_INSERT_TAIL(&cpus, cpu, node);
-    cpu_list_unlock();
+    cpu_list_add(cpu);
 
 #ifndef CONFIG_USER_ONLY
     if (qdev_get_vmsd(DEVICE(cpu)) == NULL) {
@@ -1234,7 +1199,6 @@ static void *file_ram_alloc(RAMBlock *block,
     char *c;
     void *area = MAP_FAILED;
     int fd = -1;
-    int64_t page_size;
 
     if (kvm_enabled() && !kvm_has_sync_mmu()) {
         error_setg(errp,
@@ -1289,17 +1253,17 @@ static void *file_ram_alloc(RAMBlock *block,
          */
     }
 
-    page_size = qemu_fd_getpagesize(fd);
-    block->mr->align = MAX(page_size, QEMU_VMALLOC_ALIGN);
+    block->page_size = qemu_fd_getpagesize(fd);
+    block->mr->align = MAX(block->page_size, QEMU_VMALLOC_ALIGN);
 
-    if (memory < page_size) {
+    if (memory < block->page_size) {
         error_setg(errp, "memory size 0x" RAM_ADDR_FMT " must be equal to "
-                   "or larger than page size 0x%" PRIx64,
-                   memory, page_size);
+                   "or larger than page size 0x%zx",
+                   memory, block->page_size);
         goto error;
     }
 
-    memory = ROUND_UP(memory, page_size);
+    memory = ROUND_UP(memory, block->page_size);
 
     /*
      * ftruncate is not supported by hugetlbfs in older
@@ -1454,6 +1418,11 @@ void qemu_ram_unset_idstr(RAMBlock *block)
     }
 }
 
+size_t qemu_ram_pagesize(RAMBlock *rb)
+{
+    return rb->page_size;
+}
+
 static int memory_try_enable_merging(void *addr, size_t len)
 {
     if (!machine_mem_merge(current_machine)) {
@@ -1621,10 +1590,8 @@ static void ram_block_add(RAMBlock *new_block, Error **errp)
     if (new_block->host) {
         qemu_ram_setup_dump(new_block->host, new_block->max_length);
         qemu_madvise(new_block->host, new_block->max_length, QEMU_MADV_HUGEPAGE);
+        /* MADV_DONTFORK is also needed by KVM in absence of synchronous MMU */
         qemu_madvise(new_block->host, new_block->max_length, QEMU_MADV_DONTFORK);
-        if (kvm_enabled()) {
-            kvm_setup_guest_memory(new_block->host, new_block->max_length);
-        }
     }
 }
 
@@ -1695,6 +1662,7 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
     new_block->max_length = max_size;
     assert(max_size >= size);
     new_block->fd = -1;
+    new_block->page_size = getpagesize();
     new_block->host = host;
     if (host) {
         new_block->flags |= RAM_PREALLOC;