aboutsummaryrefslogtreecommitdiff
path: root/exec.c
diff options
context:
space:
mode:
Diffstat (limited to 'exec.c')
-rw-r--r--exec.c295
1 files changed, 208 insertions, 87 deletions
diff --git a/exec.c b/exec.c
index 6dff7bc43a..c85321a38b 100644
--- a/exec.c
+++ b/exec.c
@@ -44,7 +44,7 @@
#include "trace.h"
#endif
#include "exec/cpu-all.h"
-
+#include "qemu/rcu_queue.h"
#include "exec/cputlb.h"
#include "translate-all.h"
@@ -58,7 +58,10 @@
#if !defined(CONFIG_USER_ONLY)
static bool in_migration;
-RAMList ram_list = { .blocks = QTAILQ_HEAD_INITIALIZER(ram_list.blocks) };
+/* ram_list is read under rcu_read_lock()/rcu_read_unlock(). Writes
+ * are protected by the ramlist lock.
+ */
+RAMList ram_list = { .blocks = QLIST_HEAD_INITIALIZER(ram_list.blocks) };
static MemoryRegion *system_memory;
static MemoryRegion *system_io;
@@ -115,6 +118,8 @@ struct PhysPageEntry {
typedef PhysPageEntry Node[P_L2_SIZE];
typedef struct PhysPageMap {
+ struct rcu_head rcu;
+
unsigned sections_nb;
unsigned sections_nb_alloc;
unsigned nodes_nb;
@@ -124,6 +129,8 @@ typedef struct PhysPageMap {
} PhysPageMap;
struct AddressSpaceDispatch {
+ struct rcu_head rcu;
+
/* This is a multi-level map on the physical address space.
* The bottom level has pointers to MemoryRegionSections.
*/
@@ -315,6 +322,7 @@ bool memory_region_is_unassigned(MemoryRegion *mr)
&& mr != &io_mem_watch;
}
+/* Called from RCU critical section */
static MemoryRegionSection *address_space_lookup_region(AddressSpaceDispatch *d,
hwaddr addr,
bool resolve_subpage)
@@ -330,6 +338,7 @@ static MemoryRegionSection *address_space_lookup_region(AddressSpaceDispatch *d,
return section;
}
+/* Called from RCU critical section */
static MemoryRegionSection *
address_space_translate_internal(AddressSpaceDispatch *d, hwaddr addr, hwaddr *xlat,
hwaddr *plen, bool resolve_subpage)
@@ -370,8 +379,10 @@ MemoryRegion *address_space_translate(AddressSpace *as, hwaddr addr,
MemoryRegion *mr;
hwaddr len = *plen;
+ rcu_read_lock();
for (;;) {
- section = address_space_translate_internal(as->dispatch, addr, &addr, plen, true);
+ AddressSpaceDispatch *d = atomic_rcu_read(&as->dispatch);
+ section = address_space_translate_internal(d, addr, &addr, plen, true);
mr = section->mr;
if (!mr->iommu_ops) {
@@ -397,15 +408,18 @@ MemoryRegion *address_space_translate(AddressSpace *as, hwaddr addr,
*plen = len;
*xlat = addr;
+ rcu_read_unlock();
return mr;
}
+/* Called from RCU critical section */
MemoryRegionSection *
-address_space_translate_for_iotlb(AddressSpace *as, hwaddr addr, hwaddr *xlat,
- hwaddr *plen)
+address_space_translate_for_iotlb(CPUState *cpu, hwaddr addr,
+ hwaddr *xlat, hwaddr *plen)
{
MemoryRegionSection *section;
- section = address_space_translate_internal(as->dispatch, addr, xlat, plen, false);
+ section = address_space_translate_internal(cpu->memory_dispatch,
+ addr, xlat, plen, false);
assert(!section->mr->iommu_ops);
return section;
@@ -795,16 +809,16 @@ void cpu_abort(CPUState *cpu, const char *fmt, ...)
}
#if !defined(CONFIG_USER_ONLY)
+/* Called from RCU critical section */
static RAMBlock *qemu_get_ram_block(ram_addr_t addr)
{
RAMBlock *block;
- /* The list is protected by the iothread lock here. */
- block = ram_list.mru_block;
+ block = atomic_rcu_read(&ram_list.mru_block);
if (block && addr - block->offset < block->max_length) {
goto found;
}
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
if (addr - block->offset < block->max_length) {
goto found;
}
@@ -814,6 +828,22 @@ static RAMBlock *qemu_get_ram_block(ram_addr_t addr)
abort();
found:
+ /* It is safe to write mru_block outside the iothread lock. This
+ * is what happens:
+ *
+ * mru_block = xxx
+ * rcu_read_unlock()
+ * xxx removed from list
+ * rcu_read_lock()
+ * read mru_block
+ * mru_block = NULL;
+ * call_rcu(reclaim_ramblock, xxx);
+ * rcu_read_unlock()
+ *
+ * atomic_rcu_set is not needed here. The block was already published
+ * when it was placed into the list. Here we're just making an extra
+ * copy of the pointer.
+ */
ram_list.mru_block = block;
return block;
}
@@ -827,10 +857,12 @@ static void tlb_reset_dirty_range_all(ram_addr_t start, ram_addr_t length)
end = TARGET_PAGE_ALIGN(start + length);
start &= TARGET_PAGE_MASK;
+ rcu_read_lock();
block = qemu_get_ram_block(start);
assert(block == qemu_get_ram_block(end - 1));
start1 = (uintptr_t)ramblock_ptr(block, start - block->offset);
cpu_tlb_reset_dirty_all(start1, length);
+ rcu_read_unlock();
}
/* Note: start and end must be within the same ram block. */
@@ -851,6 +883,7 @@ static void cpu_physical_memory_set_dirty_tracking(bool enable)
in_migration = enable;
}
+/* Called from RCU critical section */
hwaddr memory_region_section_get_iotlb(CPUState *cpu,
MemoryRegionSection *section,
target_ulong vaddr,
@@ -1162,6 +1195,7 @@ error:
}
#endif
+/* Called with the ramlist lock held. */
static ram_addr_t find_ram_offset(ram_addr_t size)
{
RAMBlock *block, *next_block;
@@ -1169,15 +1203,16 @@ static ram_addr_t find_ram_offset(ram_addr_t size)
assert(size != 0); /* it would hand out same offset multiple times */
- if (QTAILQ_EMPTY(&ram_list.blocks))
+ if (QLIST_EMPTY_RCU(&ram_list.blocks)) {
return 0;
+ }
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
ram_addr_t end, next = RAM_ADDR_MAX;
end = block->offset + block->max_length;
- QTAILQ_FOREACH(next_block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(next_block, &ram_list.blocks, next) {
if (next_block->offset >= end) {
next = MIN(next, next_block->offset);
}
@@ -1202,9 +1237,11 @@ ram_addr_t last_ram_offset(void)
RAMBlock *block;
ram_addr_t last = 0;
- QTAILQ_FOREACH(block, &ram_list.blocks, next)
+ rcu_read_lock();
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
last = MAX(last, block->offset + block->max_length);
-
+ }
+ rcu_read_unlock();
return last;
}
@@ -1224,11 +1261,14 @@ static void qemu_ram_setup_dump(void *addr, ram_addr_t size)
}
}
+/* Called within an RCU critical section, or while the ramlist lock
+ * is held.
+ */
static RAMBlock *find_ram_block(ram_addr_t addr)
{
RAMBlock *block;
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
if (block->offset == addr) {
return block;
}
@@ -1237,11 +1277,13 @@ static RAMBlock *find_ram_block(ram_addr_t addr)
return NULL;
}
+/* Called with iothread lock held. */
void qemu_ram_set_idstr(ram_addr_t addr, const char *name, DeviceState *dev)
{
- RAMBlock *new_block = find_ram_block(addr);
- RAMBlock *block;
+ RAMBlock *new_block, *block;
+ rcu_read_lock();
+ new_block = find_ram_block(addr);
assert(new_block);
assert(!new_block->idstr[0]);
@@ -1254,25 +1296,32 @@ void qemu_ram_set_idstr(ram_addr_t addr, const char *name, DeviceState *dev)
}
pstrcat(new_block->idstr, sizeof(new_block->idstr), name);
- /* This assumes the iothread lock is taken here too. */
- qemu_mutex_lock_ramlist();
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
if (block != new_block && !strcmp(block->idstr, new_block->idstr)) {
fprintf(stderr, "RAMBlock \"%s\" already registered, abort!\n",
new_block->idstr);
abort();
}
}
- qemu_mutex_unlock_ramlist();
+ rcu_read_unlock();
}
+/* Called with iothread lock held. */
void qemu_ram_unset_idstr(ram_addr_t addr)
{
- RAMBlock *block = find_ram_block(addr);
+ RAMBlock *block;
+
+ /* FIXME: arch_init.c assumes that this is not called throughout
+ * migration. Ignore the problem since hot-unplug during migration
+ * does not work anyway.
+ */
+ rcu_read_lock();
+ block = find_ram_block(addr);
if (block) {
memset(block->idstr, 0, sizeof(block->idstr));
}
+ rcu_read_unlock();
}
static int memory_try_enable_merging(void *addr, size_t len)
@@ -1331,11 +1380,11 @@ int qemu_ram_resize(ram_addr_t base, ram_addr_t newsize, Error **errp)
static ram_addr_t ram_block_add(RAMBlock *new_block, Error **errp)
{
RAMBlock *block;
+ RAMBlock *last_block = NULL;
ram_addr_t old_ram_size, new_ram_size;
old_ram_size = last_ram_offset() >> TARGET_PAGE_BITS;
- /* This assumes the iothread lock is taken here too. */
qemu_mutex_lock_ramlist();
new_block->offset = find_ram_offset(new_block->max_length);
@@ -1357,19 +1406,27 @@ static ram_addr_t ram_block_add(RAMBlock *new_block, Error **errp)
}
}
- /* Keep the list sorted from biggest to smallest block. */
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ /* Keep the list sorted from biggest to smallest block. Unlike QTAILQ,
+ * QLIST (which has an RCU-friendly variant) does not have insertion at
+ * tail, so save the last element in last_block.
+ */
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
+ last_block = block;
if (block->max_length < new_block->max_length) {
break;
}
}
if (block) {
- QTAILQ_INSERT_BEFORE(block, new_block, next);
- } else {
- QTAILQ_INSERT_TAIL(&ram_list.blocks, new_block, next);
+ QLIST_INSERT_BEFORE_RCU(block, new_block, next);
+ } else if (last_block) {
+ QLIST_INSERT_AFTER_RCU(last_block, new_block, next);
+ } else { /* list is empty */
+ QLIST_INSERT_HEAD_RCU(&ram_list.blocks, new_block, next);
}
ram_list.mru_block = NULL;
+ /* Write list before version */
+ smp_wmb();
ram_list.version++;
qemu_mutex_unlock_ramlist();
@@ -1377,6 +1434,8 @@ static ram_addr_t ram_block_add(RAMBlock *new_block, Error **errp)
if (new_ram_size > old_ram_size) {
int i;
+
+ /* ram_list.dirty_memory[] is protected by the iothread lock. */
for (i = 0; i < DIRTY_MEMORY_NUM; i++) {
ram_list.dirty_memory[i] =
bitmap_zero_extend(ram_list.dirty_memory[i],
@@ -1507,49 +1566,55 @@ void qemu_ram_free_from_ptr(ram_addr_t addr)
{
RAMBlock *block;
- /* This assumes the iothread lock is taken here too. */
qemu_mutex_lock_ramlist();
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
if (addr == block->offset) {
- QTAILQ_REMOVE(&ram_list.blocks, block, next);
+ QLIST_REMOVE_RCU(block, next);
ram_list.mru_block = NULL;
+ /* Write list before version */
+ smp_wmb();
ram_list.version++;
- g_free(block);
+ g_free_rcu(block, rcu);
break;
}
}
qemu_mutex_unlock_ramlist();
}
+static void reclaim_ramblock(RAMBlock *block)
+{
+ if (block->flags & RAM_PREALLOC) {
+ ;
+ } else if (xen_enabled()) {
+ xen_invalidate_map_cache_entry(block->host);
+#ifndef _WIN32
+ } else if (block->fd >= 0) {
+ munmap(block->host, block->max_length);
+ close(block->fd);
+#endif
+ } else {
+ qemu_anon_ram_free(block->host, block->max_length);
+ }
+ g_free(block);
+}
+
void qemu_ram_free(ram_addr_t addr)
{
RAMBlock *block;
- /* This assumes the iothread lock is taken here too. */
qemu_mutex_lock_ramlist();
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
if (addr == block->offset) {
- QTAILQ_REMOVE(&ram_list.blocks, block, next);
+ QLIST_REMOVE_RCU(block, next);
ram_list.mru_block = NULL;
+ /* Write list before version */
+ smp_wmb();
ram_list.version++;
- if (block->flags & RAM_PREALLOC) {
- ;
- } else if (xen_enabled()) {
- xen_invalidate_map_cache_entry(block->host);
-#ifndef _WIN32
- } else if (block->fd >= 0) {
- munmap(block->host, block->max_length);
- close(block->fd);
-#endif
- } else {
- qemu_anon_ram_free(block->host, block->max_length);
- }
- g_free(block);
+ call_rcu(block, reclaim_ramblock, rcu);
break;
}
}
qemu_mutex_unlock_ramlist();
-
}
#ifndef _WIN32
@@ -1560,7 +1625,7 @@ void qemu_ram_remap(ram_addr_t addr, ram_addr_t length)
int flags;
void *area, *vaddr;
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
offset = addr - block->offset;
if (offset < block->max_length) {
vaddr = ramblock_ptr(block, offset);
@@ -1597,7 +1662,6 @@ void qemu_ram_remap(ram_addr_t addr, ram_addr_t length)
memory_try_enable_merging(vaddr, length);
qemu_ram_setup_dump(vaddr, length);
}
- return;
}
}
}
@@ -1605,49 +1669,78 @@ void qemu_ram_remap(ram_addr_t addr, ram_addr_t length)
int qemu_get_ram_fd(ram_addr_t addr)
{
- RAMBlock *block = qemu_get_ram_block(addr);
+ RAMBlock *block;
+ int fd;
- return block->fd;
+ rcu_read_lock();
+ block = qemu_get_ram_block(addr);
+ fd = block->fd;
+ rcu_read_unlock();
+ return fd;
}
void *qemu_get_ram_block_host_ptr(ram_addr_t addr)
{
- RAMBlock *block = qemu_get_ram_block(addr);
+ RAMBlock *block;
+ void *ptr;
- return ramblock_ptr(block, 0);
+ rcu_read_lock();
+ block = qemu_get_ram_block(addr);
+ ptr = ramblock_ptr(block, 0);
+ rcu_read_unlock();
+ return ptr;
}
/* Return a host pointer to ram allocated with qemu_ram_alloc.
- With the exception of the softmmu code in this file, this should
- only be used for local memory (e.g. video ram) that the device owns,
- and knows it isn't going to access beyond the end of the block.
-
- It should not be used for general purpose DMA.
- Use cpu_physical_memory_map/cpu_physical_memory_rw instead.
+ * This should not be used for general purpose DMA. Use address_space_map
+ * or address_space_rw instead. For local memory (e.g. video ram) that the
+ * device owns, use memory_region_get_ram_ptr.
+ *
+ * By the time this function returns, the returned pointer is not protected
+ * by RCU anymore. If the caller is not within an RCU critical section and
+ * does not hold the iothread lock, it must have other means of protecting the
+ * pointer, such as a reference to the region that includes the incoming
+ * ram_addr_t.
*/
void *qemu_get_ram_ptr(ram_addr_t addr)
{
- RAMBlock *block = qemu_get_ram_block(addr);
+ RAMBlock *block;
+ void *ptr;
- if (xen_enabled()) {
+ rcu_read_lock();
+ block = qemu_get_ram_block(addr);
+
+ if (xen_enabled() && block->host == NULL) {
/* We need to check if the requested address is in the RAM
* because we don't want to map the entire memory in QEMU.
* In that case just map until the end of the page.
*/
if (block->offset == 0) {
- return xen_map_cache(addr, 0, 0);
- } else if (block->host == NULL) {
- block->host =
- xen_map_cache(block->offset, block->max_length, 1);
+ ptr = xen_map_cache(addr, 0, 0);
+ goto unlock;
}
+
+ block->host = xen_map_cache(block->offset, block->max_length, 1);
}
- return ramblock_ptr(block, addr - block->offset);
+ ptr = ramblock_ptr(block, addr - block->offset);
+
+unlock:
+ rcu_read_unlock();
+ return ptr;
}
/* Return a host pointer to guest's ram. Similar to qemu_get_ram_ptr
- * but takes a size argument */
+ * but takes a size argument.
+ *
+ * By the time this function returns, the returned pointer is not protected
+ * by RCU anymore. If the caller is not within an RCU critical section and
+ * does not hold the iothread lock, it must have other means of protecting the
+ * pointer, such as a reference to the region that includes the incoming
+ * ram_addr_t.
+ */
static void *qemu_ram_ptr_length(ram_addr_t addr, hwaddr *size)
{
+ void *ptr;
if (*size == 0) {
return NULL;
}
@@ -1655,12 +1748,14 @@ static void *qemu_ram_ptr_length(ram_addr_t addr, hwaddr *size)
return xen_map_cache(addr, *size, 1);
} else {
RAMBlock *block;
-
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ rcu_read_lock();
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
if (addr - block->offset < block->max_length) {
if (addr - block->offset + *size > block->max_length)
*size = block->max_length - addr + block->offset;
- return ramblock_ptr(block, addr - block->offset);
+ ptr = ramblock_ptr(block, addr - block->offset);
+ rcu_read_unlock();
+ return ptr;
}
}
@@ -1670,23 +1765,35 @@ static void *qemu_ram_ptr_length(ram_addr_t addr, hwaddr *size)
}
/* Some of the softmmu routines need to translate from a host pointer
- (typically a TLB entry) back to a ram offset. */
+ * (typically a TLB entry) back to a ram offset.
+ *
+ * By the time this function returns, the returned pointer is not protected
+ * by RCU anymore. If the caller is not within an RCU critical section and
+ * does not hold the iothread lock, it must have other means of protecting the
+ * pointer, such as a reference to the region that includes the incoming
+ * ram_addr_t.
+ */
MemoryRegion *qemu_ram_addr_from_host(void *ptr, ram_addr_t *ram_addr)
{
RAMBlock *block;
uint8_t *host = ptr;
+ MemoryRegion *mr;
if (xen_enabled()) {
+ rcu_read_lock();
*ram_addr = xen_ram_addr_from_mapcache(ptr);
- return qemu_get_ram_block(*ram_addr)->mr;
+ mr = qemu_get_ram_block(*ram_addr)->mr;
+ rcu_read_unlock();
+ return mr;
}
- block = ram_list.mru_block;
+ rcu_read_lock();
+ block = atomic_rcu_read(&ram_list.mru_block);
if (block && block->host && host - block->host < block->max_length) {
goto found;
}
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
/* This case append when the block is not mapped. */
if (block->host == NULL) {
continue;
@@ -1696,11 +1803,14 @@ MemoryRegion *qemu_ram_addr_from_host(void *ptr, ram_addr_t *ram_addr)
}
}
+ rcu_read_unlock();
return NULL;
found:
*ram_addr = block->offset + (host - block->host);
- return block->mr;
+ mr = block->mr;
+ rcu_read_unlock();
+ return mr;
}
static void notdirty_mem_write(void *opaque, hwaddr ram_addr,
@@ -1961,9 +2071,12 @@ static uint16_t dummy_section(PhysPageMap *map, AddressSpace *as,
return phys_section_add(map, &section);
}
-MemoryRegion *iotlb_to_region(AddressSpace *as, hwaddr index)
+MemoryRegion *iotlb_to_region(CPUState *cpu, hwaddr index)
{
- return as->dispatch->map.sections[index & ~TARGET_PAGE_MASK].mr;
+ AddressSpaceDispatch *d = atomic_rcu_read(&cpu->memory_dispatch);
+ MemoryRegionSection *sections = d->map.sections;
+
+ return sections[index & ~TARGET_PAGE_MASK].mr;
}
static void io_mem_init(void)
@@ -1997,6 +2110,12 @@ static void mem_begin(MemoryListener *listener)
as->next_dispatch = d;
}
+static void address_space_dispatch_free(AddressSpaceDispatch *d)
+{
+ phys_sections_free(&d->map);
+ g_free(d);
+}
+
static void mem_commit(MemoryListener *listener)
{
AddressSpace *as = container_of(listener, AddressSpace, dispatch_listener);
@@ -2005,11 +2124,9 @@ static void mem_commit(MemoryListener *listener)
phys_page_compact_all(next, next->map.nodes_nb);
- as->dispatch = next;
-
+ atomic_rcu_set(&as->dispatch, next);
if (cur) {
- phys_sections_free(&cur->map);
- g_free(cur);
+ call_rcu(cur, address_space_dispatch_free, rcu);
}
}
@@ -2026,7 +2143,7 @@ static void tcg_commit(MemoryListener *listener)
if (cpu->tcg_as_listener != listener) {
continue;
}
- tlb_flush(cpu, 1);
+ cpu_reload_memory_map(cpu);
}
}
@@ -2068,8 +2185,10 @@ void address_space_destroy_dispatch(AddressSpace *as)
{
AddressSpaceDispatch *d = as->dispatch;
- g_free(d);
- as->dispatch = NULL;
+ atomic_rcu_set(&as->dispatch, NULL);
+ if (d) {
+ call_rcu(d, address_space_dispatch_free, rcu);
+ }
}
static void memory_map_init(void)
@@ -2948,8 +3067,10 @@ void qemu_ram_foreach_block(RAMBlockIterFunc func, void *opaque)
{
RAMBlock *block;
- QTAILQ_FOREACH(block, &ram_list.blocks, next) {
+ rcu_read_lock();
+ QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
func(block->host, block->offset, block->used_length, opaque);
}
+ rcu_read_unlock();
}
#endif
generated by cgit v1.2.3 (git 2.26.2) at 2025-01-02 10:40:47 +0000