2003-05-13 02:25:15 +02:00
|
|
|
/*
|
2012-12-02 17:04:43 +01:00
|
|
|
* Virtual page mapping
|
2007-09-16 23:08:06 +02:00
|
|
|
*
|
2003-05-13 02:25:15 +02:00
|
|
|
* 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 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
|
2009-07-16 22:47:01 +02:00
|
|
|
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
|
2003-05-13 02:25:15 +02:00
|
|
|
*/
|
2016-01-26 19:16:56 +01:00
|
|
|
#include "qemu/osdep.h"
|
include/qemu/osdep.h: Don't include qapi/error.h
Commit 57cb38b included qapi/error.h into qemu/osdep.h to get the
Error typedef. Since then, we've moved to include qemu/osdep.h
everywhere. Its file comment explains: "To avoid getting into
possible circular include dependencies, this file should not include
any other QEMU headers, with the exceptions of config-host.h,
compiler.h, os-posix.h and os-win32.h, all of which are doing a
similar job to this file and are under similar constraints."
qapi/error.h doesn't do a similar job, and it doesn't adhere to
similar constraints: it includes qapi-types.h. That's in excess of
100KiB of crap most .c files don't actually need.
Add the typedef to qemu/typedefs.h, and include that instead of
qapi/error.h. Include qapi/error.h in .c files that need it and don't
get it now. Include qapi-types.h in qom/object.h for uint16List.
Update scripts/clean-includes accordingly. Update it further to match
reality: replace config.h by config-target.h, add sysemu/os-posix.h,
sysemu/os-win32.h. Update the list of includes in the qemu/osdep.h
comment quoted above similarly.
This reduces the number of objects depending on qapi/error.h from "all
of them" to less than a third. Unfortunately, the number depending on
qapi-types.h shrinks only a little. More work is needed for that one.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[Fix compilation without the spice devel packages. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-14 09:01:28 +01:00
|
|
|
#include "qapi/error.h"
|
2003-05-13 02:25:15 +02:00
|
|
|
|
2016-03-20 18:16:19 +01:00
|
|
|
#include "qemu/cutils.h"
|
2003-09-30 23:04:53 +02:00
|
|
|
#include "cpu.h"
|
2016-03-15 13:18:37 +01:00
|
|
|
#include "exec/exec-all.h"
|
2017-04-24 20:50:19 +02:00
|
|
|
#include "exec/target_page.h"
|
2008-05-23 11:57:34 +02:00
|
|
|
#include "tcg.h"
|
2014-06-27 08:40:04 +02:00
|
|
|
#include "hw/qdev-core.h"
|
2017-07-14 04:15:08 +02:00
|
|
|
#include "hw/qdev-properties.h"
|
2015-03-11 07:56:34 +01:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2015-02-04 16:43:54 +01:00
|
|
|
#include "hw/boards.h"
|
2016-03-15 16:58:45 +01:00
|
|
|
#include "hw/xen/xen.h"
|
2015-03-11 07:56:34 +01:00
|
|
|
#endif
|
2012-12-17 18:20:04 +01:00
|
|
|
#include "sysemu/kvm.h"
|
2013-07-04 15:09:22 +02:00
|
|
|
#include "sysemu/sysemu.h"
|
2012-12-17 18:20:00 +01:00
|
|
|
#include "qemu/timer.h"
|
|
|
|
#include "qemu/config-file.h"
|
2013-09-02 16:57:02 +02:00
|
|
|
#include "qemu/error-report.h"
|
2006-03-25 20:31:22 +01:00
|
|
|
#if defined(CONFIG_USER_ONLY)
|
2016-06-22 19:11:19 +02:00
|
|
|
#include "qemu.h"
|
2010-08-31 17:41:25 +02:00
|
|
|
#else /* !CONFIG_USER_ONLY */
|
2014-06-27 08:40:04 +02:00
|
|
|
#include "hw/hw.h"
|
|
|
|
#include "exec/memory.h"
|
2016-03-16 10:24:54 +01:00
|
|
|
#include "exec/ioport.h"
|
2014-06-27 08:40:04 +02:00
|
|
|
#include "sysemu/dma.h"
|
2017-03-02 03:36:11 +01:00
|
|
|
#include "sysemu/numa.h"
|
2017-03-07 15:19:08 +01:00
|
|
|
#include "sysemu/hw_accel.h"
|
2014-06-27 08:40:04 +02:00
|
|
|
#include "exec/address-spaces.h"
|
2012-12-17 18:20:04 +01:00
|
|
|
#include "sysemu/xen-mapcache.h"
|
2017-01-25 17:14:15 +01:00
|
|
|
#include "trace-root.h"
|
2017-02-24 19:28:32 +01:00
|
|
|
|
2017-02-24 19:28:33 +01:00
|
|
|
#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
|
|
|
|
#include <linux/falloc.h>
|
|
|
|
#endif
|
|
|
|
|
2006-03-25 20:31:22 +01:00
|
|
|
#endif
|
2013-09-05 20:41:35 +02:00
|
|
|
#include "qemu/rcu_queue.h"
|
2015-06-18 18:47:22 +02:00
|
|
|
#include "qemu/main-loop.h"
|
2012-12-02 17:04:43 +01:00
|
|
|
#include "translate-all.h"
|
2015-09-17 18:25:07 +02:00
|
|
|
#include "sysemu/replay.h"
|
2012-04-09 18:50:52 +02:00
|
|
|
|
2012-12-17 18:19:49 +01:00
|
|
|
#include "exec/memory-internal.h"
|
2013-10-14 17:13:59 +02:00
|
|
|
#include "exec/ram_addr.h"
|
2016-01-07 14:55:28 +01:00
|
|
|
#include "exec/log.h"
|
2011-12-15 14:25:22 +01:00
|
|
|
|
2016-05-12 05:48:12 +02:00
|
|
|
#include "migration/vmstate.h"
|
|
|
|
|
2013-11-11 16:52:07 +01:00
|
|
|
#include "qemu/range.h"
|
2015-09-24 13:41:17 +02:00
|
|
|
#ifndef _WIN32
|
|
|
|
#include "qemu/mmap-alloc.h"
|
|
|
|
#endif
|
2013-11-11 16:52:07 +01:00
|
|
|
|
2017-05-12 06:17:41 +02:00
|
|
|
#include "monitor/monitor.h"
|
|
|
|
|
2007-05-26 19:36:03 +02:00
|
|
|
//#define DEBUG_SUBPAGE
|
2007-03-17 16:17:58 +01:00
|
|
|
|
2008-06-08 03:09:01 +02:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2013-09-05 20:41:35 +02:00
|
|
|
/* ram_list is read under rcu_read_lock()/rcu_read_unlock(). Writes
|
|
|
|
* are protected by the ramlist lock.
|
|
|
|
*/
|
2015-01-21 13:45:24 +01:00
|
|
|
RAMList ram_list = { .blocks = QLIST_HEAD_INITIALIZER(ram_list.blocks) };
|
2011-07-26 13:26:14 +02:00
|
|
|
|
|
|
|
static MemoryRegion *system_memory;
|
2011-08-08 15:09:03 +02:00
|
|
|
static MemoryRegion *system_io;
|
2011-07-26 13:26:14 +02:00
|
|
|
|
2012-10-02 20:13:51 +02:00
|
|
|
AddressSpace address_space_io;
|
|
|
|
AddressSpace address_space_memory;
|
2012-10-02 18:49:28 +02:00
|
|
|
|
2013-05-24 14:37:28 +02:00
|
|
|
MemoryRegion io_mem_rom, io_mem_notdirty;
|
2013-05-26 21:55:37 +02:00
|
|
|
static MemoryRegion io_mem_unassigned;
|
2012-01-01 23:32:15 +01:00
|
|
|
|
2014-05-14 11:43:22 +02:00
|
|
|
/* RAM is pre-allocated and passed into qemu_ram_alloc_from_ptr */
|
|
|
|
#define RAM_PREALLOC (1 << 0)
|
|
|
|
|
2014-06-10 13:15:24 +02:00
|
|
|
/* RAM is mmap-ed with MAP_SHARED */
|
|
|
|
#define RAM_SHARED (1 << 1)
|
|
|
|
|
2014-11-12 13:27:41 +01:00
|
|
|
/* Only a portion of RAM (used_length) is actually used, and migrated.
|
|
|
|
* This used_length size can change across reboots.
|
|
|
|
*/
|
|
|
|
#define RAM_RESIZEABLE (1 << 2)
|
|
|
|
|
2018-03-12 18:20:58 +01:00
|
|
|
/* UFFDIO_ZEROPAGE is available on this RAMBlock to atomically
|
|
|
|
* zero the page and wake waiting processes.
|
|
|
|
* (Set during postcopy)
|
|
|
|
*/
|
|
|
|
#define RAM_UF_ZEROPAGE (1 << 3)
|
2018-05-14 08:57:00 +02:00
|
|
|
|
|
|
|
/* RAM can be migrated */
|
|
|
|
#define RAM_MIGRATABLE (1 << 4)
|
2008-06-08 03:09:01 +02:00
|
|
|
#endif
|
2004-01-04 19:06:42 +01:00
|
|
|
|
2016-10-24 17:26:49 +02:00
|
|
|
#ifdef TARGET_PAGE_BITS_VARY
|
|
|
|
int target_page_bits;
|
|
|
|
bool target_page_bits_decided;
|
|
|
|
#endif
|
|
|
|
|
2013-06-24 23:50:24 +02:00
|
|
|
struct CPUTailQ cpus = QTAILQ_HEAD_INITIALIZER(cpus);
|
2005-11-22 00:25:50 +01:00
|
|
|
/* current CPU in the current thread. It is only valid inside
|
|
|
|
cpu_exec() */
|
2015-08-26 00:17:58 +02:00
|
|
|
__thread CPUState *current_cpu;
|
2008-06-29 03:03:05 +02:00
|
|
|
/* 0 = Do not count executed instructions.
|
2008-06-30 19:22:19 +02:00
|
|
|
1 = Precise instruction counting.
|
2008-06-29 03:03:05 +02:00
|
|
|
2 = Adaptive rate instruction counting. */
|
2012-11-26 15:36:40 +01:00
|
|
|
int use_icount;
|
2005-11-22 00:25:50 +01:00
|
|
|
|
2017-07-03 12:12:13 +02:00
|
|
|
uintptr_t qemu_host_page_size;
|
|
|
|
intptr_t qemu_host_page_mask;
|
|
|
|
|
2016-10-24 17:26:49 +02:00
|
|
|
bool set_preferred_target_page_bits(int bits)
|
|
|
|
{
|
|
|
|
/* The target page size is the lowest common denominator for all
|
|
|
|
* the CPUs in the system, so we can only make it smaller, never
|
|
|
|
* larger. And we can't make it smaller once we've committed to
|
|
|
|
* a particular size.
|
|
|
|
*/
|
|
|
|
#ifdef TARGET_PAGE_BITS_VARY
|
|
|
|
assert(bits >= TARGET_PAGE_BITS_MIN);
|
|
|
|
if (target_page_bits == 0 || target_page_bits > bits) {
|
|
|
|
if (target_page_bits_decided) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
target_page_bits = bits;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2008-06-08 03:09:01 +02:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2012-02-10 16:00:01 +01:00
|
|
|
|
2016-10-24 17:26:49 +02:00
|
|
|
static void finalize_target_page_bits(void)
|
|
|
|
{
|
|
|
|
#ifdef TARGET_PAGE_BITS_VARY
|
|
|
|
if (target_page_bits == 0) {
|
|
|
|
target_page_bits = TARGET_PAGE_BITS_MIN;
|
|
|
|
}
|
|
|
|
target_page_bits_decided = true;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2013-05-21 12:07:21 +02:00
|
|
|
typedef struct PhysPageEntry PhysPageEntry;
|
|
|
|
|
|
|
|
struct PhysPageEntry {
|
2013-11-11 13:42:43 +01:00
|
|
|
/* How many bits skip to next level (in units of L2_SIZE). 0 for a leaf. */
|
2013-11-11 13:51:56 +01:00
|
|
|
uint32_t skip : 6;
|
2013-11-11 13:42:43 +01:00
|
|
|
/* index into phys_sections (!skip) or phys_map_nodes (skip) */
|
2013-11-11 13:51:56 +01:00
|
|
|
uint32_t ptr : 26;
|
2013-05-21 12:07:21 +02:00
|
|
|
};
|
|
|
|
|
2013-11-11 13:51:56 +01:00
|
|
|
#define PHYS_MAP_NODE_NIL (((uint32_t)~0) >> 6)
|
|
|
|
|
2013-11-07 17:14:36 +01:00
|
|
|
/* Size of the L2 (and L3, etc) page tables. */
|
2013-11-07 17:14:37 +01:00
|
|
|
#define ADDR_SPACE_BITS 64
|
2013-11-07 17:14:36 +01:00
|
|
|
|
2013-11-13 19:13:03 +01:00
|
|
|
#define P_L2_BITS 9
|
2013-11-07 17:14:36 +01:00
|
|
|
#define P_L2_SIZE (1 << P_L2_BITS)
|
|
|
|
|
|
|
|
#define P_L2_LEVELS (((ADDR_SPACE_BITS - TARGET_PAGE_BITS - 1) / P_L2_BITS) + 1)
|
|
|
|
|
|
|
|
typedef PhysPageEntry Node[P_L2_SIZE];
|
2013-05-29 12:28:21 +02:00
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
typedef struct PhysPageMap {
|
2015-01-21 12:09:14 +01:00
|
|
|
struct rcu_head rcu;
|
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
unsigned sections_nb;
|
|
|
|
unsigned sections_nb_alloc;
|
|
|
|
unsigned nodes_nb;
|
|
|
|
unsigned nodes_nb_alloc;
|
|
|
|
Node *nodes;
|
|
|
|
MemoryRegionSection *sections;
|
|
|
|
} PhysPageMap;
|
|
|
|
|
2013-05-21 12:07:21 +02:00
|
|
|
struct AddressSpaceDispatch {
|
2016-03-01 07:18:24 +01:00
|
|
|
MemoryRegionSection *mru_section;
|
2013-05-21 12:07:21 +02:00
|
|
|
/* This is a multi-level map on the physical address space.
|
|
|
|
* The bottom level has pointers to MemoryRegionSections.
|
|
|
|
*/
|
|
|
|
PhysPageEntry phys_map;
|
2013-12-01 13:02:23 +01:00
|
|
|
PhysPageMap map;
|
2013-05-21 12:07:21 +02:00
|
|
|
};
|
|
|
|
|
2013-05-26 21:46:51 +02:00
|
|
|
#define SUBPAGE_IDX(addr) ((addr) & ~TARGET_PAGE_MASK)
|
|
|
|
typedef struct subpage_t {
|
|
|
|
MemoryRegion iomem;
|
2017-09-21 10:50:58 +02:00
|
|
|
FlatView *fv;
|
2013-05-26 21:46:51 +02:00
|
|
|
hwaddr base;
|
2016-10-24 17:26:49 +02:00
|
|
|
uint16_t sub_section[];
|
2013-05-26 21:46:51 +02:00
|
|
|
} subpage_t;
|
|
|
|
|
2013-05-29 11:09:17 +02:00
|
|
|
#define PHYS_SECTION_UNASSIGNED 0
|
|
|
|
#define PHYS_SECTION_NOTDIRTY 1
|
|
|
|
#define PHYS_SECTION_ROM 2
|
|
|
|
#define PHYS_SECTION_WATCH 3
|
2012-02-12 17:32:55 +01:00
|
|
|
|
2008-06-08 03:09:01 +02:00
|
|
|
static void io_mem_init(void);
|
2011-07-26 13:26:14 +02:00
|
|
|
static void memory_map_init(void);
|
2013-12-17 04:06:51 +01:00
|
|
|
static void tcg_commit(MemoryListener *listener);
|
2008-06-08 03:09:01 +02:00
|
|
|
|
2012-01-02 11:47:48 +01:00
|
|
|
static MemoryRegion io_mem_watch;
|
2015-10-01 16:29:50 +02:00
|
|
|
|
|
|
|
/**
|
|
|
|
* CPUAddressSpace: all the information a CPU needs about an AddressSpace
|
|
|
|
* @cpu: the CPU whose AddressSpace this is
|
|
|
|
* @as: the AddressSpace itself
|
|
|
|
* @memory_dispatch: its dispatch pointer (cached, RCU protected)
|
|
|
|
* @tcg_as_listener: listener for tracking changes to the AddressSpace
|
|
|
|
*/
|
|
|
|
struct CPUAddressSpace {
|
|
|
|
CPUState *cpu;
|
|
|
|
AddressSpace *as;
|
|
|
|
struct AddressSpaceDispatch *memory_dispatch;
|
|
|
|
MemoryListener tcg_as_listener;
|
|
|
|
};
|
|
|
|
|
2017-04-21 11:16:25 +02:00
|
|
|
struct DirtyBitmapSnapshot {
|
|
|
|
ram_addr_t start;
|
|
|
|
ram_addr_t end;
|
|
|
|
unsigned long dirty[];
|
|
|
|
};
|
|
|
|
|
2007-03-17 00:58:11 +01:00
|
|
|
#endif
|
2003-05-14 21:00:11 +02:00
|
|
|
|
2010-03-01 00:55:53 +01:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2012-02-12 19:12:49 +01:00
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
static void phys_map_node_reserve(PhysPageMap *map, unsigned nodes)
|
2012-02-12 19:12:49 +01:00
|
|
|
{
|
2016-07-15 12:03:50 +02:00
|
|
|
static unsigned alloc_hint = 16;
|
2013-12-01 13:02:23 +01:00
|
|
|
if (map->nodes_nb + nodes > map->nodes_nb_alloc) {
|
2016-07-15 12:03:50 +02:00
|
|
|
map->nodes_nb_alloc = MAX(map->nodes_nb_alloc, alloc_hint);
|
2013-12-01 13:02:23 +01:00
|
|
|
map->nodes_nb_alloc = MAX(map->nodes_nb_alloc, map->nodes_nb + nodes);
|
|
|
|
map->nodes = g_renew(Node, map->nodes, map->nodes_nb_alloc);
|
2016-07-15 12:03:50 +02:00
|
|
|
alloc_hint = map->nodes_nb_alloc;
|
2012-02-12 19:12:49 +01:00
|
|
|
}
|
2012-02-13 19:12:05 +01:00
|
|
|
}
|
|
|
|
|
2015-05-21 15:12:29 +02:00
|
|
|
static uint32_t phys_map_node_alloc(PhysPageMap *map, bool leaf)
|
2012-02-13 19:12:05 +01:00
|
|
|
{
|
|
|
|
unsigned i;
|
2013-11-11 13:51:56 +01:00
|
|
|
uint32_t ret;
|
2015-05-21 15:12:29 +02:00
|
|
|
PhysPageEntry e;
|
|
|
|
PhysPageEntry *p;
|
2012-02-13 19:12:05 +01:00
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
ret = map->nodes_nb++;
|
2015-05-21 15:12:29 +02:00
|
|
|
p = map->nodes[ret];
|
2012-02-13 19:12:05 +01:00
|
|
|
assert(ret != PHYS_MAP_NODE_NIL);
|
2013-12-01 13:02:23 +01:00
|
|
|
assert(ret != map->nodes_nb_alloc);
|
2015-05-21 15:12:29 +02:00
|
|
|
|
|
|
|
e.skip = leaf ? 0 : 1;
|
|
|
|
e.ptr = leaf ? PHYS_SECTION_UNASSIGNED : PHYS_MAP_NODE_NIL;
|
2013-11-07 17:14:36 +01:00
|
|
|
for (i = 0; i < P_L2_SIZE; ++i) {
|
2015-05-21 15:12:29 +02:00
|
|
|
memcpy(&p[i], &e, sizeof(e));
|
2012-02-12 19:12:49 +01:00
|
|
|
}
|
2012-02-13 19:12:05 +01:00
|
|
|
return ret;
|
2012-02-12 19:12:49 +01:00
|
|
|
}
|
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
static void phys_page_set_level(PhysPageMap *map, PhysPageEntry *lp,
|
|
|
|
hwaddr *index, hwaddr *nb, uint16_t leaf,
|
2012-02-13 19:21:20 +01:00
|
|
|
int level)
|
2012-02-13 19:12:05 +01:00
|
|
|
{
|
|
|
|
PhysPageEntry *p;
|
2013-11-07 17:14:36 +01:00
|
|
|
hwaddr step = (hwaddr)1 << (level * P_L2_BITS);
|
2005-07-24 14:55:09 +02:00
|
|
|
|
2013-11-11 13:42:43 +01:00
|
|
|
if (lp->skip && lp->ptr == PHYS_MAP_NODE_NIL) {
|
2015-05-21 15:12:29 +02:00
|
|
|
lp->ptr = phys_map_node_alloc(map, level == 0);
|
2004-05-21 16:52:29 +02:00
|
|
|
}
|
2015-05-21 15:12:29 +02:00
|
|
|
p = map->nodes[lp->ptr];
|
2013-11-07 17:14:36 +01:00
|
|
|
lp = &p[(*index >> (level * P_L2_BITS)) & (P_L2_SIZE - 1)];
|
2012-02-13 19:12:05 +01:00
|
|
|
|
2013-11-07 17:14:36 +01:00
|
|
|
while (*nb && lp < &p[P_L2_SIZE]) {
|
2012-02-13 19:45:32 +01:00
|
|
|
if ((*index & (step - 1)) == 0 && *nb >= step) {
|
2013-11-11 13:42:43 +01:00
|
|
|
lp->skip = 0;
|
2012-02-13 19:25:31 +01:00
|
|
|
lp->ptr = leaf;
|
2012-02-13 19:45:32 +01:00
|
|
|
*index += step;
|
|
|
|
*nb -= step;
|
2012-02-13 19:21:20 +01:00
|
|
|
} else {
|
2013-12-01 13:02:23 +01:00
|
|
|
phys_page_set_level(map, lp, index, nb, leaf, level - 1);
|
2012-02-13 19:21:20 +01:00
|
|
|
}
|
|
|
|
++lp;
|
2012-02-13 19:12:05 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-10-03 16:22:53 +02:00
|
|
|
static void phys_page_set(AddressSpaceDispatch *d,
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr index, hwaddr nb,
|
2012-02-13 19:21:20 +01:00
|
|
|
uint16_t leaf)
|
2012-02-13 19:12:05 +01:00
|
|
|
{
|
2012-02-13 19:21:20 +01:00
|
|
|
/* Wildly overreserve - it doesn't matter much. */
|
2013-12-01 13:02:23 +01:00
|
|
|
phys_map_node_reserve(&d->map, 3 * P_L2_LEVELS);
|
2010-03-11 00:53:37 +01:00
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
phys_page_set_level(&d->map, &d->phys_map, &index, &nb, leaf, P_L2_LEVELS - 1);
|
2004-05-21 16:52:29 +02:00
|
|
|
}
|
|
|
|
|
2013-11-11 16:52:07 +01:00
|
|
|
/* 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.
|
|
|
|
*/
|
2016-09-28 14:37:20 +02:00
|
|
|
static void phys_page_compact(PhysPageEntry *lp, Node *nodes)
|
2013-11-11 16:52:07 +01:00
|
|
|
{
|
|
|
|
unsigned valid_ptr = P_L2_SIZE;
|
|
|
|
int valid = 0;
|
|
|
|
PhysPageEntry *p;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (lp->ptr == PHYS_MAP_NODE_NIL) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
p = nodes[lp->ptr];
|
|
|
|
for (i = 0; i < P_L2_SIZE; i++) {
|
|
|
|
if (p[i].ptr == PHYS_MAP_NODE_NIL) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
valid_ptr = i;
|
|
|
|
valid++;
|
|
|
|
if (p[i].skip) {
|
2016-09-28 14:37:20 +02:00
|
|
|
phys_page_compact(&p[i], nodes);
|
2013-11-11 16:52:07 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We can only compress if there's only one child. */
|
|
|
|
if (valid != 1) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
assert(valid_ptr < P_L2_SIZE);
|
|
|
|
|
|
|
|
/* Don't compress if it won't fit in the # of bits we have. */
|
|
|
|
if (lp->skip + p[valid_ptr].skip >= (1 << 3)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
lp->ptr = p[valid_ptr].ptr;
|
|
|
|
if (!p[valid_ptr].skip) {
|
|
|
|
/* If our only child is a leaf, make this a leaf. */
|
|
|
|
/* By design, we should have made this node a leaf to begin with so we
|
|
|
|
* should never reach here.
|
|
|
|
* But since it's so simple to handle this, let's do it just in case we
|
|
|
|
* change this rule.
|
|
|
|
*/
|
|
|
|
lp->skip = 0;
|
|
|
|
} else {
|
|
|
|
lp->skip += p[valid_ptr].skip;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-09-21 10:51:00 +02:00
|
|
|
void address_space_dispatch_compact(AddressSpaceDispatch *d)
|
2013-11-11 16:52:07 +01:00
|
|
|
{
|
|
|
|
if (d->phys_map.skip) {
|
2016-09-28 14:37:20 +02:00
|
|
|
phys_page_compact(&d->phys_map, d->map.nodes);
|
2013-11-11 16:52:07 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-03-01 07:18:23 +01:00
|
|
|
static inline bool section_covers_addr(const MemoryRegionSection *section,
|
|
|
|
hwaddr addr)
|
|
|
|
{
|
|
|
|
/* Memory topology clips a memory region to [0, 2^64); size.hi > 0 means
|
|
|
|
* the section must cover the entire address space.
|
|
|
|
*/
|
2016-06-30 00:48:03 +02:00
|
|
|
return int128_gethi(section->size) ||
|
2016-03-01 07:18:23 +01:00
|
|
|
range_covers_byte(section->offset_within_address_space,
|
2016-06-30 00:48:03 +02:00
|
|
|
int128_getlo(section->size), addr);
|
2016-03-01 07:18:23 +01:00
|
|
|
}
|
|
|
|
|
2017-05-15 10:50:57 +02:00
|
|
|
static MemoryRegionSection *phys_page_find(AddressSpaceDispatch *d, hwaddr addr)
|
2004-05-21 16:52:29 +02:00
|
|
|
{
|
2017-05-15 10:50:57 +02:00
|
|
|
PhysPageEntry lp = d->phys_map, *p;
|
|
|
|
Node *nodes = d->map.nodes;
|
|
|
|
MemoryRegionSection *sections = d->map.sections;
|
2013-11-13 19:08:19 +01:00
|
|
|
hwaddr index = addr >> TARGET_PAGE_BITS;
|
2012-02-13 15:44:19 +01:00
|
|
|
int i;
|
2011-11-20 16:52:22 +01:00
|
|
|
|
2013-11-11 13:42:43 +01:00
|
|
|
for (i = P_L2_LEVELS; lp.skip && (i -= lp.skip) >= 0;) {
|
2012-02-13 19:25:31 +01:00
|
|
|
if (lp.ptr == PHYS_MAP_NODE_NIL) {
|
2013-05-29 12:09:47 +02:00
|
|
|
return §ions[PHYS_SECTION_UNASSIGNED];
|
2012-02-13 15:44:19 +01:00
|
|
|
}
|
2013-05-29 12:09:47 +02:00
|
|
|
p = nodes[lp.ptr];
|
2013-11-07 17:14:36 +01:00
|
|
|
lp = p[(index >> (i * P_L2_BITS)) & (P_L2_SIZE - 1)];
|
2012-02-12 17:32:55 +01:00
|
|
|
}
|
2013-11-11 16:52:07 +01:00
|
|
|
|
2016-03-01 07:18:23 +01:00
|
|
|
if (section_covers_addr(§ions[lp.ptr], addr)) {
|
2013-11-11 16:52:07 +01:00
|
|
|
return §ions[lp.ptr];
|
|
|
|
} else {
|
|
|
|
return §ions[PHYS_SECTION_UNASSIGNED];
|
|
|
|
}
|
2012-03-08 15:16:34 +01:00
|
|
|
}
|
|
|
|
|
2012-04-21 15:08:33 +02:00
|
|
|
bool memory_region_is_unassigned(MemoryRegion *mr)
|
|
|
|
{
|
2013-05-24 14:34:08 +02:00
|
|
|
return mr != &io_mem_rom && mr != &io_mem_notdirty && !mr->rom_device
|
2012-12-02 17:04:43 +01:00
|
|
|
&& mr != &io_mem_watch;
|
2003-05-14 21:00:11 +02:00
|
|
|
}
|
2013-05-24 12:59:37 +02:00
|
|
|
|
2015-01-21 12:09:14 +01:00
|
|
|
/* Called from RCU critical section */
|
2013-06-02 15:27:39 +02:00
|
|
|
static MemoryRegionSection *address_space_lookup_region(AddressSpaceDispatch *d,
|
2013-05-26 21:46:51 +02:00
|
|
|
hwaddr addr,
|
|
|
|
bool resolve_subpage)
|
2013-05-06 16:48:02 +02:00
|
|
|
{
|
2016-03-01 07:18:24 +01:00
|
|
|
MemoryRegionSection *section = atomic_read(&d->mru_section);
|
2013-05-26 21:46:51 +02:00
|
|
|
subpage_t *subpage;
|
|
|
|
|
2017-11-15 15:11:03 +01:00
|
|
|
if (!section || section == &d->map.sections[PHYS_SECTION_UNASSIGNED] ||
|
|
|
|
!section_covers_addr(section, addr)) {
|
2017-05-15 10:50:57 +02:00
|
|
|
section = phys_page_find(d, addr);
|
2017-11-15 15:11:03 +01:00
|
|
|
atomic_set(&d->mru_section, section);
|
2016-03-01 07:18:24 +01:00
|
|
|
}
|
2013-05-26 21:46:51 +02:00
|
|
|
if (resolve_subpage && section->mr->subpage) {
|
|
|
|
subpage = container_of(section->mr, subpage_t, iomem);
|
2013-12-01 13:02:23 +01:00
|
|
|
section = &d->map.sections[subpage->sub_section[SUBPAGE_IDX(addr)]];
|
2013-05-26 21:46:51 +02:00
|
|
|
}
|
|
|
|
return section;
|
2013-05-06 16:48:02 +02:00
|
|
|
}
|
|
|
|
|
2015-01-21 12:09:14 +01:00
|
|
|
/* Called from RCU critical section */
|
2013-05-26 21:46:51 +02:00
|
|
|
static MemoryRegionSection *
|
2013-06-02 15:27:39 +02:00
|
|
|
address_space_translate_internal(AddressSpaceDispatch *d, hwaddr addr, hwaddr *xlat,
|
2013-05-26 21:46:51 +02:00
|
|
|
hwaddr *plen, bool resolve_subpage)
|
2013-05-24 12:59:37 +02:00
|
|
|
{
|
|
|
|
MemoryRegionSection *section;
|
2015-06-17 10:40:27 +02:00
|
|
|
MemoryRegion *mr;
|
2014-02-07 15:47:46 +01:00
|
|
|
Int128 diff;
|
2013-05-24 12:59:37 +02:00
|
|
|
|
2013-06-02 15:27:39 +02:00
|
|
|
section = address_space_lookup_region(d, addr, resolve_subpage);
|
2013-05-24 12:59:37 +02:00
|
|
|
/* Compute offset within MemoryRegionSection */
|
|
|
|
addr -= section->offset_within_address_space;
|
|
|
|
|
|
|
|
/* Compute offset within MemoryRegion */
|
|
|
|
*xlat = addr + section->offset_within_region;
|
|
|
|
|
2015-06-17 10:40:27 +02:00
|
|
|
mr = section->mr;
|
exec: skip MMIO regions correctly in cpu_physical_memory_write_rom_internal
Loading the BIOS in the mac99 machine is interesting, because there is a
PROM in the middle of the BIOS region (from 16K to 32K). Before memory
region accesses were clamped, when QEMU was asked to load a BIOS from
0xfff00000 to 0xffffffff it would put even those 16K from the BIOS file
into the region. This is weird because those 16K were not actually
visible between 0xfff04000 and 0xfff07fff. However, it worked.
After clamping was added, this also worked. In this case, the
cpu_physical_memory_write_rom_internal function split the write in
three parts: the first 16K were copied, the PROM area (second 16K) were
ignored, then the rest was copied.
Problems then started with commit 965eb2f (exec: do not clamp accesses
to MMIO regions, 2015-06-17). Clamping accesses is not done for MMIO
regions because they can overlap wildly, and MMIO registers can be
expected to perform full-width accesses based only on their address
(with no respect for adjacent registers that could decode to completely
different MemoryRegions). However, this lack of clamping also applied
to the PROM area! cpu_physical_memory_write_rom_internal thus failed
to copy the third range above, i.e. only copied the first 16K of the BIOS.
In effect, address_space_translate is expecting _something else_ to do
the clamping for MMIO regions if the incoming length is large. This
"something else" is memory_access_size in the case of address_space_rw,
so use the same logic in cpu_physical_memory_write_rom_internal.
Reported-by: Alexander Graf <agraf@redhat.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Tested-by: Laurent Vivier <lvivier@redhat.com>
Fixes: 965eb2f
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2015-07-04 00:24:51 +02:00
|
|
|
|
|
|
|
/* MMIO registers can be expected to perform full-width accesses based only
|
|
|
|
* on their address, without considering adjacent registers that could
|
|
|
|
* decode to completely different MemoryRegions. When such registers
|
|
|
|
* exist (e.g. I/O ports 0xcf8 and 0xcf9 on most PC chipsets), MMIO
|
|
|
|
* regions overlap wildly. For this reason we cannot clamp the accesses
|
|
|
|
* here.
|
|
|
|
*
|
|
|
|
* If the length is small (as is the case for address_space_ldl/stl),
|
|
|
|
* everything works fine. If the incoming length is large, however,
|
|
|
|
* the caller really has to do the clamping through memory_access_size.
|
|
|
|
*/
|
2015-06-17 10:40:27 +02:00
|
|
|
if (memory_region_is_ram(mr)) {
|
2015-06-17 10:36:54 +02:00
|
|
|
diff = int128_sub(section->size, int128_make64(addr));
|
2015-06-17 10:40:27 +02:00
|
|
|
*plen = int128_get64(int128_min(diff, int128_make64(*plen)));
|
|
|
|
}
|
2013-05-24 12:59:37 +02:00
|
|
|
return section;
|
|
|
|
}
|
2013-05-26 21:46:51 +02:00
|
|
|
|
2018-03-03 17:24:04 +01:00
|
|
|
/**
|
|
|
|
* address_space_translate_iommu - translate an address through an IOMMU
|
|
|
|
* memory region and then through the target address space.
|
|
|
|
*
|
|
|
|
* @iommu_mr: the IOMMU memory region that we start the translation from
|
|
|
|
* @addr: the address to be translated through the MMU
|
|
|
|
* @xlat: the translated address offset within the destination memory region.
|
|
|
|
* It cannot be %NULL.
|
|
|
|
* @plen_out: valid read/write length of the translated address. It
|
|
|
|
* cannot be %NULL.
|
|
|
|
* @page_mask_out: page mask for the translated address. This
|
|
|
|
* should only be meaningful for IOMMU translated
|
|
|
|
* addresses, since there may be huge pages that this bit
|
|
|
|
* would tell. It can be %NULL if we don't care about it.
|
|
|
|
* @is_write: whether the translation operation is for write
|
|
|
|
* @is_mmio: whether this can be MMIO, set true if it can
|
|
|
|
* @target_as: the address space targeted by the IOMMU
|
2018-05-31 15:50:53 +02:00
|
|
|
* @attrs: transaction attributes
|
2018-03-03 17:24:04 +01:00
|
|
|
*
|
|
|
|
* This function is called from RCU critical section. It is the common
|
|
|
|
* part of flatview_do_translate and address_space_translate_cached.
|
|
|
|
*/
|
|
|
|
static MemoryRegionSection address_space_translate_iommu(IOMMUMemoryRegion *iommu_mr,
|
|
|
|
hwaddr *xlat,
|
|
|
|
hwaddr *plen_out,
|
|
|
|
hwaddr *page_mask_out,
|
|
|
|
bool is_write,
|
|
|
|
bool is_mmio,
|
2018-05-31 15:50:53 +02:00
|
|
|
AddressSpace **target_as,
|
|
|
|
MemTxAttrs attrs)
|
2018-03-03 17:24:04 +01:00
|
|
|
{
|
|
|
|
MemoryRegionSection *section;
|
|
|
|
hwaddr page_mask = (hwaddr)-1;
|
|
|
|
|
|
|
|
do {
|
|
|
|
hwaddr addr = *xlat;
|
|
|
|
IOMMUMemoryRegionClass *imrc = memory_region_get_iommu_class_nocheck(iommu_mr);
|
2018-06-15 15:57:16 +02:00
|
|
|
int iommu_idx = 0;
|
|
|
|
IOMMUTLBEntry iotlb;
|
|
|
|
|
|
|
|
if (imrc->attrs_to_index) {
|
|
|
|
iommu_idx = imrc->attrs_to_index(iommu_mr, attrs);
|
|
|
|
}
|
|
|
|
|
|
|
|
iotlb = imrc->translate(iommu_mr, addr, is_write ?
|
|
|
|
IOMMU_WO : IOMMU_RO, iommu_idx);
|
2018-03-03 17:24:04 +01:00
|
|
|
|
|
|
|
if (!(iotlb.perm & (1 << is_write))) {
|
|
|
|
goto unassigned;
|
|
|
|
}
|
|
|
|
|
|
|
|
addr = ((iotlb.translated_addr & ~iotlb.addr_mask)
|
|
|
|
| (addr & iotlb.addr_mask));
|
|
|
|
page_mask &= iotlb.addr_mask;
|
|
|
|
*plen_out = MIN(*plen_out, (addr | iotlb.addr_mask) - addr + 1);
|
|
|
|
*target_as = iotlb.target_as;
|
|
|
|
|
|
|
|
section = address_space_translate_internal(
|
|
|
|
address_space_to_dispatch(iotlb.target_as), addr, xlat,
|
|
|
|
plen_out, is_mmio);
|
|
|
|
|
|
|
|
iommu_mr = memory_region_get_iommu(section->mr);
|
|
|
|
} while (unlikely(iommu_mr));
|
|
|
|
|
|
|
|
if (page_mask_out) {
|
|
|
|
*page_mask_out = page_mask;
|
|
|
|
}
|
|
|
|
return *section;
|
|
|
|
|
|
|
|
unassigned:
|
|
|
|
return (MemoryRegionSection) { .mr = &io_mem_unassigned };
|
|
|
|
}
|
|
|
|
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
/**
|
|
|
|
* flatview_do_translate - translate an address in FlatView
|
|
|
|
*
|
|
|
|
* @fv: the flat view that we want to translate on
|
|
|
|
* @addr: the address to be translated in above address space
|
|
|
|
* @xlat: the translated address offset within memory region. It
|
|
|
|
* cannot be @NULL.
|
|
|
|
* @plen_out: valid read/write length of the translated address. It
|
|
|
|
* can be @NULL when we don't care about it.
|
|
|
|
* @page_mask_out: page mask for the translated address. This
|
|
|
|
* should only be meaningful for IOMMU translated
|
|
|
|
* addresses, since there may be huge pages that this bit
|
|
|
|
* would tell. It can be @NULL if we don't care about it.
|
|
|
|
* @is_write: whether the translation operation is for write
|
|
|
|
* @is_mmio: whether this can be MMIO, set true if it can
|
2018-04-17 11:39:35 +02:00
|
|
|
* @target_as: the address space targeted by the IOMMU
|
2018-05-31 15:50:53 +02:00
|
|
|
* @attrs: memory transaction attributes
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
*
|
|
|
|
* This function is called from RCU critical section
|
|
|
|
*/
|
2017-09-21 10:50:58 +02:00
|
|
|
static MemoryRegionSection flatview_do_translate(FlatView *fv,
|
|
|
|
hwaddr addr,
|
|
|
|
hwaddr *xlat,
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
hwaddr *plen_out,
|
|
|
|
hwaddr *page_mask_out,
|
2017-09-21 10:50:58 +02:00
|
|
|
bool is_write,
|
|
|
|
bool is_mmio,
|
2018-05-31 15:50:53 +02:00
|
|
|
AddressSpace **target_as,
|
|
|
|
MemTxAttrs attrs)
|
2016-12-30 11:09:13 +01:00
|
|
|
{
|
|
|
|
MemoryRegionSection *section;
|
2017-07-11 05:56:19 +02:00
|
|
|
IOMMUMemoryRegion *iommu_mr;
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
hwaddr plen = (hwaddr)(-1);
|
|
|
|
|
2018-04-17 11:39:35 +02:00
|
|
|
if (!plen_out) {
|
|
|
|
plen_out = &plen;
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
}
|
2016-12-30 11:09:13 +01:00
|
|
|
|
2018-03-03 17:24:04 +01:00
|
|
|
section = address_space_translate_internal(
|
|
|
|
flatview_to_dispatch(fv), addr, xlat,
|
|
|
|
plen_out, is_mmio);
|
2016-12-30 11:09:13 +01:00
|
|
|
|
2018-03-03 17:24:04 +01:00
|
|
|
iommu_mr = memory_region_get_iommu(section->mr);
|
|
|
|
if (unlikely(iommu_mr)) {
|
|
|
|
return address_space_translate_iommu(iommu_mr, xlat,
|
|
|
|
plen_out, page_mask_out,
|
|
|
|
is_write, is_mmio,
|
2018-05-31 15:50:53 +02:00
|
|
|
target_as, attrs);
|
2016-12-30 11:09:13 +01:00
|
|
|
}
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
if (page_mask_out) {
|
2018-03-03 17:24:04 +01:00
|
|
|
/* Not behind an IOMMU, use default page size. */
|
|
|
|
*page_mask_out = ~TARGET_PAGE_MASK;
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
}
|
|
|
|
|
2017-05-17 10:57:42 +02:00
|
|
|
return *section;
|
2016-12-30 11:09:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from RCU critical section */
|
2017-05-17 10:57:42 +02:00
|
|
|
IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr,
|
2018-05-31 15:50:53 +02:00
|
|
|
bool is_write, MemTxAttrs attrs)
|
2013-05-26 21:46:51 +02:00
|
|
|
{
|
2017-05-17 10:57:42 +02:00
|
|
|
MemoryRegionSection section;
|
2017-10-10 11:42:46 +02:00
|
|
|
hwaddr xlat, page_mask;
|
2012-10-30 12:47:46 +01:00
|
|
|
|
2017-10-10 11:42:46 +02:00
|
|
|
/*
|
|
|
|
* This can never be MMIO, and we don't really care about plen,
|
|
|
|
* but page mask.
|
|
|
|
*/
|
|
|
|
section = flatview_do_translate(address_space_to_flatview(as), addr, &xlat,
|
2018-05-31 15:50:53 +02:00
|
|
|
NULL, &page_mask, is_write, false, &as,
|
|
|
|
attrs);
|
2012-10-30 12:47:46 +01:00
|
|
|
|
2017-05-17 10:57:42 +02:00
|
|
|
/* Illegal translation */
|
|
|
|
if (section.mr == &io_mem_unassigned) {
|
|
|
|
goto iotlb_fail;
|
|
|
|
}
|
2012-10-30 12:47:46 +01:00
|
|
|
|
2017-05-17 10:57:42 +02:00
|
|
|
/* Convert memory region offset into address space offset */
|
|
|
|
xlat += section.offset_within_address_space -
|
|
|
|
section.offset_within_region;
|
|
|
|
|
|
|
|
return (IOMMUTLBEntry) {
|
2017-09-21 10:50:53 +02:00
|
|
|
.target_as = as,
|
2017-10-10 11:42:46 +02:00
|
|
|
.iova = addr & ~page_mask,
|
|
|
|
.translated_addr = xlat & ~page_mask,
|
|
|
|
.addr_mask = page_mask,
|
2017-05-17 10:57:42 +02:00
|
|
|
/* IOTLBs are for DMAs, and DMA only allows on RAMs. */
|
|
|
|
.perm = IOMMU_RW,
|
|
|
|
};
|
|
|
|
|
|
|
|
iotlb_fail:
|
|
|
|
return (IOMMUTLBEntry) {0};
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from RCU critical section */
|
2017-09-21 10:50:58 +02:00
|
|
|
MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat,
|
2018-05-31 15:50:52 +02:00
|
|
|
hwaddr *plen, bool is_write,
|
|
|
|
MemTxAttrs attrs)
|
2017-05-17 10:57:42 +02:00
|
|
|
{
|
|
|
|
MemoryRegion *mr;
|
|
|
|
MemoryRegionSection section;
|
2017-09-21 10:50:58 +02:00
|
|
|
AddressSpace *as = NULL;
|
2017-05-17 10:57:42 +02:00
|
|
|
|
|
|
|
/* This can be MMIO, so setup MMIO bit. */
|
exec: add page_mask for flatview_do_translate
The function is originally used for flatview_space_translate() and what
we care about most is (xlat, plen) range. However for iotlb requests, we
don't really care about "plen", but the size of the page that "xlat" is
located on. While, plen cannot really contain this information.
A simple example to show why "plen" is not good for IOTLB translations:
E.g., for huge pages, it is possible that guest mapped 1G huge page on
device side that used this GPA range:
0x100000000 - 0x13fffffff
Then let's say we want to translate one IOVA that finally mapped to GPA
0x13ffffe00 (which is located on this 1G huge page). Then here we'll
get:
(xlat, plen) = (0x13fffe00, 0x200)
So the IOTLB would be only covering a very small range since from
"plen" (which is 0x200 bytes) we cannot tell the size of the page.
Actually we can really know that this is a huge page - we just throw the
information away in flatview_do_translate().
This patch introduced "page_mask" optional parameter to capture that
page mask info. Also, I made "plen" an optional parameter as well, with
some comments for the whole function.
No functional change yet.
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20171010094247.10173-2-maxime.coquelin@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-10-10 11:42:45 +02:00
|
|
|
section = flatview_do_translate(fv, addr, xlat, plen, NULL,
|
2018-05-31 15:50:53 +02:00
|
|
|
is_write, true, &as, attrs);
|
2017-05-17 10:57:42 +02:00
|
|
|
mr = section.mr;
|
|
|
|
|
2014-05-07 15:40:39 +02:00
|
|
|
if (xen_enabled() && memory_access_is_direct(mr, is_write)) {
|
2014-02-07 15:47:46 +01:00
|
|
|
hwaddr page = ((addr & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE) - addr;
|
2015-03-17 06:35:54 +01:00
|
|
|
*plen = MIN(page, *plen);
|
2014-02-07 15:47:46 +01:00
|
|
|
}
|
|
|
|
|
2012-10-30 12:47:46 +01:00
|
|
|
return mr;
|
2013-05-26 21:46:51 +02:00
|
|
|
}
|
|
|
|
|
2018-06-15 15:57:16 +02:00
|
|
|
typedef struct TCGIOMMUNotifier {
|
|
|
|
IOMMUNotifier n;
|
|
|
|
MemoryRegion *mr;
|
|
|
|
CPUState *cpu;
|
|
|
|
int iommu_idx;
|
|
|
|
bool active;
|
|
|
|
} TCGIOMMUNotifier;
|
|
|
|
|
|
|
|
static void tcg_iommu_unmap_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
|
|
|
|
{
|
|
|
|
TCGIOMMUNotifier *notifier = container_of(n, TCGIOMMUNotifier, n);
|
|
|
|
|
|
|
|
if (!notifier->active) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
tlb_flush(notifier->cpu);
|
|
|
|
notifier->active = false;
|
|
|
|
/* We leave the notifier struct on the list to avoid reallocating it later.
|
|
|
|
* Generally the number of IOMMUs a CPU deals with will be small.
|
|
|
|
* In any case we can't unregister the iommu notifier from a notify
|
|
|
|
* callback.
|
|
|
|
*/
|
|
|
|
}
|
|
|
|
|
|
|
|
static void tcg_register_iommu_notifier(CPUState *cpu,
|
|
|
|
IOMMUMemoryRegion *iommu_mr,
|
|
|
|
int iommu_idx)
|
|
|
|
{
|
|
|
|
/* Make sure this CPU has an IOMMU notifier registered for this
|
|
|
|
* IOMMU/IOMMU index combination, so that we can flush its TLB
|
|
|
|
* when the IOMMU tells us the mappings we've cached have changed.
|
|
|
|
*/
|
|
|
|
MemoryRegion *mr = MEMORY_REGION(iommu_mr);
|
|
|
|
TCGIOMMUNotifier *notifier;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < cpu->iommu_notifiers->len; i++) {
|
|
|
|
notifier = &g_array_index(cpu->iommu_notifiers, TCGIOMMUNotifier, i);
|
|
|
|
if (notifier->mr == mr && notifier->iommu_idx == iommu_idx) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (i == cpu->iommu_notifiers->len) {
|
|
|
|
/* Not found, add a new entry at the end of the array */
|
|
|
|
cpu->iommu_notifiers = g_array_set_size(cpu->iommu_notifiers, i + 1);
|
|
|
|
notifier = &g_array_index(cpu->iommu_notifiers, TCGIOMMUNotifier, i);
|
|
|
|
|
|
|
|
notifier->mr = mr;
|
|
|
|
notifier->iommu_idx = iommu_idx;
|
|
|
|
notifier->cpu = cpu;
|
|
|
|
/* Rather than trying to register interest in the specific part
|
|
|
|
* of the iommu's address space that we've accessed and then
|
|
|
|
* expand it later as subsequent accesses touch more of it, we
|
|
|
|
* just register interest in the whole thing, on the assumption
|
|
|
|
* that iommu reconfiguration will be rare.
|
|
|
|
*/
|
|
|
|
iommu_notifier_init(¬ifier->n,
|
|
|
|
tcg_iommu_unmap_notify,
|
|
|
|
IOMMU_NOTIFIER_UNMAP,
|
|
|
|
0,
|
|
|
|
HWADDR_MAX,
|
|
|
|
iommu_idx);
|
|
|
|
memory_region_register_iommu_notifier(notifier->mr, ¬ifier->n);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!notifier->active) {
|
|
|
|
notifier->active = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void tcg_iommu_free_notifier_list(CPUState *cpu)
|
|
|
|
{
|
|
|
|
/* Destroy the CPU's notifier list */
|
|
|
|
int i;
|
|
|
|
TCGIOMMUNotifier *notifier;
|
|
|
|
|
|
|
|
for (i = 0; i < cpu->iommu_notifiers->len; i++) {
|
|
|
|
notifier = &g_array_index(cpu->iommu_notifiers, TCGIOMMUNotifier, i);
|
|
|
|
memory_region_unregister_iommu_notifier(notifier->mr, ¬ifier->n);
|
|
|
|
}
|
|
|
|
g_array_free(cpu->iommu_notifiers, true);
|
|
|
|
}
|
|
|
|
|
2015-01-21 12:09:14 +01:00
|
|
|
/* Called from RCU critical section */
|
2013-05-26 21:46:51 +02:00
|
|
|
MemoryRegionSection *
|
2016-01-21 15:15:05 +01:00
|
|
|
address_space_translate_for_iotlb(CPUState *cpu, int asidx, hwaddr addr,
|
2018-06-15 15:57:16 +02:00
|
|
|
hwaddr *xlat, hwaddr *plen,
|
|
|
|
MemTxAttrs attrs, int *prot)
|
2013-05-26 21:46:51 +02:00
|
|
|
{
|
2012-10-30 12:47:46 +01:00
|
|
|
MemoryRegionSection *section;
|
2018-06-15 15:57:16 +02:00
|
|
|
IOMMUMemoryRegion *iommu_mr;
|
|
|
|
IOMMUMemoryRegionClass *imrc;
|
|
|
|
IOMMUTLBEntry iotlb;
|
|
|
|
int iommu_idx;
|
2016-10-21 17:34:18 +02:00
|
|
|
AddressSpaceDispatch *d = atomic_rcu_read(&cpu->cpu_ases[asidx].memory_dispatch);
|
2016-01-21 15:15:05 +01:00
|
|
|
|
2018-06-15 15:57:16 +02:00
|
|
|
for (;;) {
|
|
|
|
section = address_space_translate_internal(d, addr, &addr, plen, false);
|
|
|
|
|
|
|
|
iommu_mr = memory_region_get_iommu(section->mr);
|
|
|
|
if (!iommu_mr) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
imrc = memory_region_get_iommu_class_nocheck(iommu_mr);
|
|
|
|
|
|
|
|
iommu_idx = imrc->attrs_to_index(iommu_mr, attrs);
|
|
|
|
tcg_register_iommu_notifier(cpu, iommu_mr, iommu_idx);
|
|
|
|
/* We need all the permissions, so pass IOMMU_NONE so the IOMMU
|
|
|
|
* doesn't short-cut its translation table walk.
|
|
|
|
*/
|
|
|
|
iotlb = imrc->translate(iommu_mr, addr, IOMMU_NONE, iommu_idx);
|
|
|
|
addr = ((iotlb.translated_addr & ~iotlb.addr_mask)
|
|
|
|
| (addr & iotlb.addr_mask));
|
|
|
|
/* Update the caller's prot bits to remove permissions the IOMMU
|
|
|
|
* is giving us a failure response for. If we get down to no
|
|
|
|
* permissions left at all we can give up now.
|
|
|
|
*/
|
|
|
|
if (!(iotlb.perm & IOMMU_RO)) {
|
|
|
|
*prot &= ~(PAGE_READ | PAGE_EXEC);
|
|
|
|
}
|
|
|
|
if (!(iotlb.perm & IOMMU_WO)) {
|
|
|
|
*prot &= ~PAGE_WRITE;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!*prot) {
|
|
|
|
goto translate_fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
d = flatview_to_dispatch(address_space_to_flatview(iotlb.target_as));
|
|
|
|
}
|
2012-10-30 12:47:46 +01:00
|
|
|
|
2017-07-11 05:56:19 +02:00
|
|
|
assert(!memory_region_is_iommu(section->mr));
|
2018-06-15 15:57:16 +02:00
|
|
|
*xlat = addr;
|
2012-10-30 12:47:46 +01:00
|
|
|
return section;
|
2018-06-15 15:57:16 +02:00
|
|
|
|
|
|
|
translate_fail:
|
|
|
|
return &d->map.sections[PHYS_SECTION_UNASSIGNED];
|
2013-05-26 21:46:51 +02:00
|
|
|
}
|
2012-12-02 17:04:43 +01:00
|
|
|
#endif
|
2003-05-14 21:00:11 +02:00
|
|
|
|
2013-01-20 20:23:22 +01:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2012-12-02 17:04:43 +01:00
|
|
|
|
|
|
|
static int cpu_common_post_load(void *opaque, int version_id)
|
2003-05-14 21:00:11 +02:00
|
|
|
{
|
2013-01-17 18:51:17 +01:00
|
|
|
CPUState *cpu = opaque;
|
2003-05-28 01:29:48 +02:00
|
|
|
|
2012-12-02 17:04:43 +01:00
|
|
|
/* 0x01 was CPU_INTERRUPT_EXIT. This line can be removed when the
|
|
|
|
version_id is increased. */
|
2013-01-17 18:51:17 +01:00
|
|
|
cpu->interrupt_request &= ~0x01;
|
2016-11-14 15:17:28 +01:00
|
|
|
tlb_flush(cpu);
|
2012-12-02 17:04:43 +01:00
|
|
|
|
2018-01-10 14:48:46 +01:00
|
|
|
/* loadvm has just updated the content of RAM, bypassing the
|
|
|
|
* usual mechanisms that ensure we flush TBs for writes to
|
|
|
|
* memory we've translated code from. So we must flush all TBs,
|
|
|
|
* which will now be stale.
|
|
|
|
*/
|
|
|
|
tb_flush(cpu);
|
|
|
|
|
2012-12-02 17:04:43 +01:00
|
|
|
return 0;
|
2003-05-28 01:29:48 +02:00
|
|
|
}
|
2003-06-21 15:11:07 +02:00
|
|
|
|
2014-07-31 07:41:17 +02:00
|
|
|
static int cpu_common_pre_load(void *opaque)
|
|
|
|
{
|
|
|
|
CPUState *cpu = opaque;
|
|
|
|
|
2014-12-19 12:53:14 +01:00
|
|
|
cpu->exception_index = -1;
|
2014-07-31 07:41:17 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool cpu_common_exception_index_needed(void *opaque)
|
|
|
|
{
|
|
|
|
CPUState *cpu = opaque;
|
|
|
|
|
2014-12-19 12:53:14 +01:00
|
|
|
return tcg_enabled() && cpu->exception_index != -1;
|
2014-07-31 07:41:17 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static const VMStateDescription vmstate_cpu_common_exception_index = {
|
|
|
|
.name = "cpu_common/exception_index",
|
|
|
|
.version_id = 1,
|
|
|
|
.minimum_version_id = 1,
|
2014-09-23 14:09:54 +02:00
|
|
|
.needed = cpu_common_exception_index_needed,
|
2014-07-31 07:41:17 +02:00
|
|
|
.fields = (VMStateField[]) {
|
|
|
|
VMSTATE_INT32(exception_index, CPUState),
|
|
|
|
VMSTATE_END_OF_LIST()
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2015-07-03 14:01:44 +02:00
|
|
|
static bool cpu_common_crash_occurred_needed(void *opaque)
|
|
|
|
{
|
|
|
|
CPUState *cpu = opaque;
|
|
|
|
|
|
|
|
return cpu->crash_occurred;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const VMStateDescription vmstate_cpu_common_crash_occurred = {
|
|
|
|
.name = "cpu_common/crash_occurred",
|
|
|
|
.version_id = 1,
|
|
|
|
.minimum_version_id = 1,
|
|
|
|
.needed = cpu_common_crash_occurred_needed,
|
|
|
|
.fields = (VMStateField[]) {
|
|
|
|
VMSTATE_BOOL(crash_occurred, CPUState),
|
|
|
|
VMSTATE_END_OF_LIST()
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2013-06-17 04:09:11 +02:00
|
|
|
const VMStateDescription vmstate_cpu_common = {
|
2012-12-02 17:04:43 +01:00
|
|
|
.name = "cpu_common",
|
|
|
|
.version_id = 1,
|
|
|
|
.minimum_version_id = 1,
|
2014-07-31 07:41:17 +02:00
|
|
|
.pre_load = cpu_common_pre_load,
|
2012-12-02 17:04:43 +01:00
|
|
|
.post_load = cpu_common_post_load,
|
2014-04-16 16:01:33 +02:00
|
|
|
.fields = (VMStateField[]) {
|
2013-01-17 18:51:17 +01:00
|
|
|
VMSTATE_UINT32(halted, CPUState),
|
|
|
|
VMSTATE_UINT32(interrupt_request, CPUState),
|
2012-12-02 17:04:43 +01:00
|
|
|
VMSTATE_END_OF_LIST()
|
2014-07-31 07:41:17 +02:00
|
|
|
},
|
2014-09-23 14:09:54 +02:00
|
|
|
.subsections = (const VMStateDescription*[]) {
|
|
|
|
&vmstate_cpu_common_exception_index,
|
2015-07-03 14:01:44 +02:00
|
|
|
&vmstate_cpu_common_crash_occurred,
|
2014-09-23 14:09:54 +02:00
|
|
|
NULL
|
2012-12-02 17:04:43 +01:00
|
|
|
}
|
|
|
|
};
|
2013-06-17 04:09:11 +02:00
|
|
|
|
2012-12-02 17:04:43 +01:00
|
|
|
#endif
|
2003-06-25 18:16:50 +02:00
|
|
|
|
2012-12-17 19:47:15 +01:00
|
|
|
CPUState *qemu_get_cpu(int index)
|
2003-06-25 18:16:50 +02:00
|
|
|
{
|
2013-06-24 23:50:24 +02:00
|
|
|
CPUState *cpu;
|
2003-06-25 18:16:50 +02:00
|
|
|
|
2013-06-24 23:50:24 +02:00
|
|
|
CPU_FOREACH(cpu) {
|
2012-12-17 06:18:02 +01:00
|
|
|
if (cpu->cpu_index == index) {
|
2013-06-24 23:50:24 +02:00
|
|
|
return cpu;
|
2012-12-17 06:18:02 +01:00
|
|
|
}
|
2003-06-25 18:16:50 +02:00
|
|
|
}
|
2012-12-02 17:04:43 +01:00
|
|
|
|
2013-06-24 23:50:24 +02:00
|
|
|
return NULL;
|
2003-06-25 18:16:50 +02:00
|
|
|
}
|
|
|
|
|
2013-12-17 04:06:51 +01:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2017-11-23 10:23:32 +01:00
|
|
|
void cpu_address_space_init(CPUState *cpu, int asidx,
|
|
|
|
const char *prefix, MemoryRegion *mr)
|
2013-12-17 04:06:51 +01:00
|
|
|
{
|
2016-01-21 15:15:04 +01:00
|
|
|
CPUAddressSpace *newas;
|
2017-11-23 10:23:32 +01:00
|
|
|
AddressSpace *as = g_new0(AddressSpace, 1);
|
2017-11-23 10:23:33 +01:00
|
|
|
char *as_name;
|
2017-11-23 10:23:32 +01:00
|
|
|
|
|
|
|
assert(mr);
|
2017-11-23 10:23:33 +01:00
|
|
|
as_name = g_strdup_printf("%s-%d", prefix, cpu->cpu_index);
|
|
|
|
address_space_init(as, mr, as_name);
|
|
|
|
g_free(as_name);
|
2016-01-21 15:15:04 +01:00
|
|
|
|
|
|
|
/* Target code should have set num_ases before calling us */
|
|
|
|
assert(asidx < cpu->num_ases);
|
|
|
|
|
2016-01-21 15:15:04 +01:00
|
|
|
if (asidx == 0) {
|
|
|
|
/* address space 0 gets the convenience alias */
|
|
|
|
cpu->as = as;
|
|
|
|
}
|
|
|
|
|
2016-01-21 15:15:04 +01:00
|
|
|
/* KVM cannot currently support multiple address spaces. */
|
|
|
|
assert(asidx == 0 || !kvm_enabled());
|
2013-12-17 04:06:51 +01:00
|
|
|
|
2016-01-21 15:15:04 +01:00
|
|
|
if (!cpu->cpu_ases) {
|
|
|
|
cpu->cpu_ases = g_new0(CPUAddressSpace, cpu->num_ases);
|
2013-12-17 04:06:51 +01:00
|
|
|
}
|
2015-10-01 16:29:50 +02:00
|
|
|
|
2016-01-21 15:15:04 +01:00
|
|
|
newas = &cpu->cpu_ases[asidx];
|
|
|
|
newas->cpu = cpu;
|
|
|
|
newas->as = as;
|
2016-01-21 15:15:04 +01:00
|
|
|
if (tcg_enabled()) {
|
2016-01-21 15:15:04 +01:00
|
|
|
newas->tcg_as_listener.commit = tcg_commit;
|
|
|
|
memory_listener_register(&newas->tcg_as_listener, as);
|
2016-01-21 15:15:04 +01:00
|
|
|
}
|
2013-12-17 04:06:51 +01:00
|
|
|
}
|
2016-01-21 15:15:05 +01:00
|
|
|
|
|
|
|
AddressSpace *cpu_get_address_space(CPUState *cpu, int asidx)
|
|
|
|
{
|
|
|
|
/* Return the AddressSpace corresponding to the specified index */
|
|
|
|
return cpu->cpu_ases[asidx].as;
|
|
|
|
}
|
2013-12-17 04:06:51 +01:00
|
|
|
#endif
|
|
|
|
|
2016-10-20 13:26:04 +02:00
|
|
|
void cpu_exec_unrealizefn(CPUState *cpu)
|
2016-05-12 05:48:11 +02:00
|
|
|
{
|
2016-05-12 05:48:12 +02:00
|
|
|
CPUClass *cc = CPU_GET_CLASS(cpu);
|
|
|
|
|
2016-08-28 03:45:14 +02:00
|
|
|
cpu_list_remove(cpu);
|
2016-05-12 05:48:12 +02:00
|
|
|
|
|
|
|
if (cc->vmsd != NULL) {
|
|
|
|
vmstate_unregister(NULL, cc->vmsd, cpu);
|
|
|
|
}
|
|
|
|
if (qdev_get_vmsd(DEVICE(cpu)) == NULL) {
|
|
|
|
vmstate_unregister(NULL, &vmstate_cpu_common, cpu);
|
|
|
|
}
|
2018-06-15 15:57:16 +02:00
|
|
|
#ifndef CONFIG_USER_ONLY
|
|
|
|
tcg_iommu_free_notifier_list(cpu);
|
|
|
|
#endif
|
2016-05-12 05:48:11 +02:00
|
|
|
}
|
|
|
|
|
2017-07-14 04:15:08 +02:00
|
|
|
Property cpu_common_props[] = {
|
|
|
|
#ifndef CONFIG_USER_ONLY
|
|
|
|
/* Create a memory property for softmmu CPU object,
|
|
|
|
* so users can wire up its memory. (This can't go in qom/cpu.c
|
|
|
|
* because that file is compiled only once for both user-mode
|
|
|
|
* and system builds.) The default if no link is set up is to use
|
|
|
|
* the system address space.
|
|
|
|
*/
|
|
|
|
DEFINE_PROP_LINK("memory", CPUState, memory, TYPE_MEMORY_REGION,
|
|
|
|
MemoryRegion *),
|
|
|
|
#endif
|
|
|
|
DEFINE_PROP_END_OF_LIST(),
|
|
|
|
};
|
|
|
|
|
2016-10-20 13:26:02 +02:00
|
|
|
void cpu_exec_initfn(CPUState *cpu)
|
2003-06-25 18:16:50 +02:00
|
|
|
{
|
2016-01-21 15:15:04 +01:00
|
|
|
cpu->as = NULL;
|
2016-01-21 15:15:04 +01:00
|
|
|
cpu->num_ases = 0;
|
2016-01-21 15:15:04 +01:00
|
|
|
|
2015-04-27 22:00:33 +02:00
|
|
|
#ifndef CONFIG_USER_ONLY
|
|
|
|
cpu->thread_id = qemu_get_thread_id();
|
2016-01-21 15:15:06 +01:00
|
|
|
cpu->memory = system_memory;
|
|
|
|
object_ref(OBJECT(cpu->memory));
|
2015-04-27 22:00:33 +02:00
|
|
|
#endif
|
2016-10-20 13:26:02 +02:00
|
|
|
}
|
|
|
|
|
2016-10-20 13:26:03 +02:00
|
|
|
void cpu_exec_realizefn(CPUState *cpu, Error **errp)
|
2016-10-20 13:26:02 +02:00
|
|
|
{
|
2017-10-16 04:02:42 +02:00
|
|
|
CPUClass *cc = CPU_GET_CLASS(cpu);
|
qom: move CPUClass.tcg_initialize to a global
55c3cee ("qom: Introduce CPUClass.tcg_initialize", 2017-10-24)
introduces a per-CPUClass bool that we check so that the target CPU
is initialized for TCG only once. This works well except when
we end up creating more than one CPUClass, in which case we end
up incorrectly initializing TCG more than once, i.e. once for
each CPUClass.
This can be replicated with:
$ aarch64-softmmu/qemu-system-aarch64 -machine xlnx-zcu102 -smp 6 \
-global driver=xlnx,,zynqmp,property=has_rpu,value=on
In this case the class name of the "RPUs" is prefixed by "cortex-r5-",
whereas the "regular" CPUs are prefixed by "cortex-a53-". This
results in two CPUClass instances being created.
Fix it by introducing a static variable, so that only the first
target CPU being initialized will initialize the target-dependent
part of TCG, regardless of CPUClass instances.
Fixes: 55c3ceef61fcf06fc98ddc752b7cce788ce7680b
Signed-off-by: Emilio G. Cota <cota@braap.org>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 1510343626-25861-2-git-send-email-cota@braap.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2017-11-13 14:55:25 +01:00
|
|
|
static bool tcg_target_initialized;
|
2015-04-27 22:00:33 +02:00
|
|
|
|
2016-08-28 03:45:14 +02:00
|
|
|
cpu_list_add(cpu);
|
2016-07-25 11:59:19 +02:00
|
|
|
|
qom: move CPUClass.tcg_initialize to a global
55c3cee ("qom: Introduce CPUClass.tcg_initialize", 2017-10-24)
introduces a per-CPUClass bool that we check so that the target CPU
is initialized for TCG only once. This works well except when
we end up creating more than one CPUClass, in which case we end
up incorrectly initializing TCG more than once, i.e. once for
each CPUClass.
This can be replicated with:
$ aarch64-softmmu/qemu-system-aarch64 -machine xlnx-zcu102 -smp 6 \
-global driver=xlnx,,zynqmp,property=has_rpu,value=on
In this case the class name of the "RPUs" is prefixed by "cortex-r5-",
whereas the "regular" CPUs are prefixed by "cortex-a53-". This
results in two CPUClass instances being created.
Fix it by introducing a static variable, so that only the first
target CPU being initialized will initialize the target-dependent
part of TCG, regardless of CPUClass instances.
Fixes: 55c3ceef61fcf06fc98ddc752b7cce788ce7680b
Signed-off-by: Emilio G. Cota <cota@braap.org>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 1510343626-25861-2-git-send-email-cota@braap.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2017-11-13 14:55:25 +01:00
|
|
|
if (tcg_enabled() && !tcg_target_initialized) {
|
|
|
|
tcg_target_initialized = true;
|
2017-10-16 04:02:42 +02:00
|
|
|
cc->tcg_initialize();
|
|
|
|
}
|
|
|
|
|
2016-07-25 11:59:19 +02:00
|
|
|
#ifndef CONFIG_USER_ONLY
|
2013-07-29 04:07:50 +02:00
|
|
|
if (qdev_get_vmsd(DEVICE(cpu)) == NULL) {
|
2014-06-27 08:40:04 +02:00
|
|
|
vmstate_register(NULL, cpu->cpu_index, &vmstate_cpu_common, cpu);
|
2013-07-29 04:07:50 +02:00
|
|
|
}
|
2013-01-20 20:23:22 +01:00
|
|
|
if (cc->vmsd != NULL) {
|
2014-06-27 08:40:04 +02:00
|
|
|
vmstate_register(NULL, cpu->cpu_index, cc->vmsd, cpu);
|
2013-01-20 20:23:22 +01:00
|
|
|
}
|
2018-06-15 15:57:16 +02:00
|
|
|
|
|
|
|
cpu->iommu_notifiers = g_array_new(false, true, sizeof(TCGIOMMUNotifier));
|
2014-06-27 08:40:04 +02:00
|
|
|
#endif
|
2003-06-25 18:16:50 +02:00
|
|
|
}
|
|
|
|
|
2018-02-07 11:40:26 +01:00
|
|
|
const char *parse_cpu_model(const char *cpu_model)
|
|
|
|
{
|
|
|
|
ObjectClass *oc;
|
|
|
|
CPUClass *cc;
|
|
|
|
gchar **model_pieces;
|
|
|
|
const char *cpu_type;
|
|
|
|
|
|
|
|
model_pieces = g_strsplit(cpu_model, ",", 2);
|
|
|
|
|
|
|
|
oc = cpu_class_by_name(CPU_RESOLVING_TYPE, model_pieces[0]);
|
|
|
|
if (oc == NULL) {
|
|
|
|
error_report("unable to find CPU model '%s'", model_pieces[0]);
|
|
|
|
g_strfreev(model_pieces);
|
|
|
|
exit(EXIT_FAILURE);
|
|
|
|
}
|
|
|
|
|
|
|
|
cpu_type = object_class_get_name(oc);
|
|
|
|
cc = CPU_CLASS(oc);
|
|
|
|
cc->parse_features(cpu_type, model_pieces[1], &error_fatal);
|
|
|
|
g_strfreev(model_pieces);
|
|
|
|
return cpu_type;
|
|
|
|
}
|
|
|
|
|
2017-07-12 23:51:42 +02:00
|
|
|
#if defined(CONFIG_USER_ONLY)
|
2013-06-29 18:55:54 +02:00
|
|
|
static void breakpoint_invalidate(CPUState *cpu, target_ulong pc)
|
2012-04-10 00:48:17 +02:00
|
|
|
{
|
2017-07-12 23:51:42 +02:00
|
|
|
mmap_lock();
|
|
|
|
tb_invalidate_phys_page_range(pc, pc + 1, 0);
|
|
|
|
mmap_unlock();
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void breakpoint_invalidate(CPUState *cpu, target_ulong pc)
|
|
|
|
{
|
|
|
|
MemTxAttrs attrs;
|
|
|
|
hwaddr phys = cpu_get_phys_page_attrs_debug(cpu, pc, &attrs);
|
|
|
|
int asidx = cpu_asidx_from_attrs(cpu, attrs);
|
|
|
|
if (phys != -1) {
|
|
|
|
/* Locks grabbed by tb_invalidate_phys_addr */
|
|
|
|
tb_invalidate_phys_addr(cpu->cpu_ases[asidx].as,
|
2018-05-31 15:50:52 +02:00
|
|
|
phys | (pc & ~TARGET_PAGE_MASK), attrs);
|
2017-07-12 23:51:42 +02:00
|
|
|
}
|
2012-04-10 00:48:17 +02:00
|
|
|
}
|
2017-07-12 23:51:42 +02:00
|
|
|
#endif
|
2004-04-25 19:57:43 +02:00
|
|
|
|
2010-03-01 04:31:14 +01:00
|
|
|
#if defined(CONFIG_USER_ONLY)
|
2013-09-02 16:57:02 +02:00
|
|
|
void cpu_watchpoint_remove_all(CPUState *cpu, int mask)
|
2010-03-01 04:31:14 +01:00
|
|
|
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2014-09-12 15:06:48 +02:00
|
|
|
int cpu_watchpoint_remove(CPUState *cpu, vaddr addr, vaddr len,
|
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
return -ENOSYS;
|
|
|
|
}
|
|
|
|
|
|
|
|
void cpu_watchpoint_remove_by_ref(CPUState *cpu, CPUWatchpoint *watchpoint)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2013-09-02 16:57:02 +02:00
|
|
|
int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
|
2010-03-01 04:31:14 +01:00
|
|
|
int flags, CPUWatchpoint **watchpoint)
|
|
|
|
{
|
|
|
|
return -ENOSYS;
|
|
|
|
}
|
|
|
|
#else
|
2007-03-17 00:58:11 +01:00
|
|
|
/* Add a watchpoint. */
|
2013-09-02 16:57:02 +02:00
|
|
|
int cpu_watchpoint_insert(CPUState *cpu, vaddr addr, vaddr len,
|
2008-11-18 21:07:32 +01:00
|
|
|
int flags, CPUWatchpoint **watchpoint)
|
2007-03-17 00:58:11 +01:00
|
|
|
{
|
2008-11-25 23:13:57 +01:00
|
|
|
CPUWatchpoint *wp;
|
2007-03-17 00:58:11 +01:00
|
|
|
|
2014-09-12 15:06:48 +02:00
|
|
|
/* forbid ranges which are empty or run off the end of the address space */
|
2014-09-18 07:03:36 +02:00
|
|
|
if (len == 0 || (addr + len - 1) < addr) {
|
2013-09-02 16:57:02 +02:00
|
|
|
error_report("tried to set invalid watchpoint at %"
|
|
|
|
VADDR_PRIx ", len=%" VADDR_PRIu, addr, len);
|
2008-11-18 21:14:20 +01:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
2011-08-21 05:09:37 +02:00
|
|
|
wp = g_malloc(sizeof(*wp));
|
2008-11-18 21:07:32 +01:00
|
|
|
|
|
|
|
wp->vaddr = addr;
|
2014-09-12 15:06:48 +02:00
|
|
|
wp->len = len;
|
2008-11-18 21:07:32 +01:00
|
|
|
wp->flags = flags;
|
|
|
|
|
2008-11-18 21:56:59 +01:00
|
|
|
/* keep all GDB-injected watchpoints in front */
|
2013-08-26 18:23:18 +02:00
|
|
|
if (flags & BP_GDB) {
|
|
|
|
QTAILQ_INSERT_HEAD(&cpu->watchpoints, wp, entry);
|
|
|
|
} else {
|
|
|
|
QTAILQ_INSERT_TAIL(&cpu->watchpoints, wp, entry);
|
|
|
|
}
|
2007-03-17 00:58:11 +01:00
|
|
|
|
2013-09-04 01:29:02 +02:00
|
|
|
tlb_flush_page(cpu, addr);
|
2008-11-18 21:07:32 +01:00
|
|
|
|
|
|
|
if (watchpoint)
|
|
|
|
*watchpoint = wp;
|
|
|
|
return 0;
|
2007-03-17 00:58:11 +01:00
|
|
|
}
|
|
|
|
|
2008-11-18 21:07:32 +01:00
|
|
|
/* Remove a specific watchpoint. */
|
2013-09-02 16:57:02 +02:00
|
|
|
int cpu_watchpoint_remove(CPUState *cpu, vaddr addr, vaddr len,
|
2008-11-18 21:07:32 +01:00
|
|
|
int flags)
|
2007-03-17 00:58:11 +01:00
|
|
|
{
|
2008-11-18 21:07:32 +01:00
|
|
|
CPUWatchpoint *wp;
|
2007-03-17 00:58:11 +01:00
|
|
|
|
2013-08-26 18:23:18 +02:00
|
|
|
QTAILQ_FOREACH(wp, &cpu->watchpoints, entry) {
|
2014-09-12 15:06:48 +02:00
|
|
|
if (addr == wp->vaddr && len == wp->len
|
2008-11-18 21:37:55 +01:00
|
|
|
&& flags == (wp->flags & ~BP_WATCHPOINT_HIT)) {
|
2013-09-02 16:57:02 +02:00
|
|
|
cpu_watchpoint_remove_by_ref(cpu, wp);
|
2007-03-17 00:58:11 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
2008-11-18 21:07:32 +01:00
|
|
|
return -ENOENT;
|
2007-03-17 00:58:11 +01:00
|
|
|
}
|
|
|
|
|
2008-11-18 21:07:32 +01:00
|
|
|
/* Remove a specific watchpoint by reference. */
|
2013-09-02 16:57:02 +02:00
|
|
|
void cpu_watchpoint_remove_by_ref(CPUState *cpu, CPUWatchpoint *watchpoint)
|
2008-11-18 21:07:32 +01:00
|
|
|
{
|
2013-08-26 18:23:18 +02:00
|
|
|
QTAILQ_REMOVE(&cpu->watchpoints, watchpoint, entry);
|
2008-05-17 20:58:29 +02:00
|
|
|
|
2013-09-04 01:29:02 +02:00
|
|
|
tlb_flush_page(cpu, watchpoint->vaddr);
|
2008-11-18 21:07:32 +01:00
|
|
|
|
2011-08-21 05:09:37 +02:00
|
|
|
g_free(watchpoint);
|
2008-11-18 21:07:32 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Remove all matching watchpoints. */
|
2013-09-02 16:57:02 +02:00
|
|
|
void cpu_watchpoint_remove_all(CPUState *cpu, int mask)
|
2008-11-18 21:07:32 +01:00
|
|
|
{
|
2008-11-25 23:13:57 +01:00
|
|
|
CPUWatchpoint *wp, *next;
|
2008-11-18 21:07:32 +01:00
|
|
|
|
2013-08-26 18:23:18 +02:00
|
|
|
QTAILQ_FOREACH_SAFE(wp, &cpu->watchpoints, entry, next) {
|
2013-09-02 16:57:02 +02:00
|
|
|
if (wp->flags & mask) {
|
|
|
|
cpu_watchpoint_remove_by_ref(cpu, wp);
|
|
|
|
}
|
2008-11-25 23:13:57 +01:00
|
|
|
}
|
2008-05-17 20:58:29 +02:00
|
|
|
}
|
2014-09-12 15:06:48 +02:00
|
|
|
|
|
|
|
/* Return true if this watchpoint address matches the specified
|
|
|
|
* access (ie the address range covered by the watchpoint overlaps
|
|
|
|
* partially or completely with the address range covered by the
|
|
|
|
* access).
|
|
|
|
*/
|
|
|
|
static inline bool cpu_watchpoint_address_matches(CPUWatchpoint *wp,
|
|
|
|
vaddr addr,
|
|
|
|
vaddr len)
|
|
|
|
{
|
|
|
|
/* We know the lengths are non-zero, but a little caution is
|
|
|
|
* required to avoid errors in the case where the range ends
|
|
|
|
* exactly at the top of the address space and so addr + len
|
|
|
|
* wraps round to zero.
|
|
|
|
*/
|
|
|
|
vaddr wpend = wp->vaddr + wp->len - 1;
|
|
|
|
vaddr addrend = addr + len - 1;
|
|
|
|
|
|
|
|
return !(addr > wpend || wp->vaddr > addrend);
|
|
|
|
}
|
|
|
|
|
2010-03-01 04:31:14 +01:00
|
|
|
#endif
|
2008-05-17 20:58:29 +02:00
|
|
|
|
2008-11-18 21:07:32 +01:00
|
|
|
/* Add a breakpoint. */
|
2013-09-02 17:26:20 +02:00
|
|
|
int cpu_breakpoint_insert(CPUState *cpu, vaddr pc, int flags,
|
2008-11-18 21:07:32 +01:00
|
|
|
CPUBreakpoint **breakpoint)
|
2003-07-26 14:06:08 +02:00
|
|
|
{
|
2008-11-25 23:13:57 +01:00
|
|
|
CPUBreakpoint *bp;
|
2007-09-17 10:09:54 +02:00
|
|
|
|
2011-08-21 05:09:37 +02:00
|
|
|
bp = g_malloc(sizeof(*bp));
|
2003-07-26 14:06:08 +02:00
|
|
|
|
2008-11-18 21:07:32 +01:00
|
|
|
bp->pc = pc;
|
|
|
|
bp->flags = flags;
|
|
|
|
|
2008-11-18 21:56:59 +01:00
|
|
|
/* keep all GDB-injected breakpoints in front */
|
2013-06-29 18:55:54 +02:00
|
|
|
if (flags & BP_GDB) {
|
2013-08-26 21:22:53 +02:00
|
|
|
QTAILQ_INSERT_HEAD(&cpu->breakpoints, bp, entry);
|
2013-06-29 18:55:54 +02:00
|
|
|
} else {
|
2013-08-26 21:22:53 +02:00
|
|
|
QTAILQ_INSERT_TAIL(&cpu->breakpoints, bp, entry);
|
2013-06-29 18:55:54 +02:00
|
|
|
}
|
2007-09-17 10:09:54 +02:00
|
|
|
|
2013-08-26 21:22:53 +02:00
|
|
|
breakpoint_invalidate(cpu, pc);
|
2008-11-18 21:07:32 +01:00
|
|
|
|
2013-06-29 18:55:54 +02:00
|
|
|
if (breakpoint) {
|
2008-11-18 21:07:32 +01:00
|
|
|
*breakpoint = bp;
|
2013-06-29 18:55:54 +02:00
|
|
|
}
|
2003-07-26 14:06:08 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-11-18 21:07:32 +01:00
|
|
|
/* Remove a specific breakpoint. */
|
2013-09-02 17:26:20 +02:00
|
|
|
int cpu_breakpoint_remove(CPUState *cpu, vaddr pc, int flags)
|
2008-11-18 21:07:32 +01:00
|
|
|
{
|
|
|
|
CPUBreakpoint *bp;
|
|
|
|
|
2013-08-26 21:22:53 +02:00
|
|
|
QTAILQ_FOREACH(bp, &cpu->breakpoints, entry) {
|
2008-11-18 21:07:32 +01:00
|
|
|
if (bp->pc == pc && bp->flags == flags) {
|
2013-09-02 17:26:20 +02:00
|
|
|
cpu_breakpoint_remove_by_ref(cpu, bp);
|
2008-11-18 21:07:32 +01:00
|
|
|
return 0;
|
|
|
|
}
|
2008-05-17 20:58:29 +02:00
|
|
|
}
|
2008-11-18 21:07:32 +01:00
|
|
|
return -ENOENT;
|
2008-05-17 20:58:29 +02:00
|
|
|
}
|
|
|
|
|
2008-11-18 21:07:32 +01:00
|
|
|
/* Remove a specific breakpoint by reference. */
|
2013-09-02 17:26:20 +02:00
|
|
|
void cpu_breakpoint_remove_by_ref(CPUState *cpu, CPUBreakpoint *breakpoint)
|
2003-07-26 14:06:08 +02:00
|
|
|
{
|
2013-08-26 21:22:53 +02:00
|
|
|
QTAILQ_REMOVE(&cpu->breakpoints, breakpoint, entry);
|
|
|
|
|
|
|
|
breakpoint_invalidate(cpu, breakpoint->pc);
|
2008-11-18 21:07:32 +01:00
|
|
|
|
2011-08-21 05:09:37 +02:00
|
|
|
g_free(breakpoint);
|
2008-11-18 21:07:32 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Remove all matching breakpoints. */
|
2013-09-02 17:26:20 +02:00
|
|
|
void cpu_breakpoint_remove_all(CPUState *cpu, int mask)
|
2008-11-18 21:07:32 +01:00
|
|
|
{
|
2008-11-25 23:13:57 +01:00
|
|
|
CPUBreakpoint *bp, *next;
|
2008-11-18 21:07:32 +01:00
|
|
|
|
2013-08-26 21:22:53 +02:00
|
|
|
QTAILQ_FOREACH_SAFE(bp, &cpu->breakpoints, entry, next) {
|
2013-09-02 17:26:20 +02:00
|
|
|
if (bp->flags & mask) {
|
|
|
|
cpu_breakpoint_remove_by_ref(cpu, bp);
|
|
|
|
}
|
2008-11-25 23:13:57 +01:00
|
|
|
}
|
2003-07-26 14:06:08 +02:00
|
|
|
}
|
|
|
|
|
2003-07-29 22:50:33 +02:00
|
|
|
/* enable or disable single step mode. EXCP_DEBUG is returned by the
|
|
|
|
CPU loop after each instruction */
|
2013-06-24 18:41:06 +02:00
|
|
|
void cpu_single_step(CPUState *cpu, int enabled)
|
2003-07-29 22:50:33 +02:00
|
|
|
{
|
2013-06-21 20:20:45 +02:00
|
|
|
if (cpu->singlestep_enabled != enabled) {
|
|
|
|
cpu->singlestep_enabled = enabled;
|
|
|
|
if (kvm_enabled()) {
|
2013-07-25 20:50:21 +02:00
|
|
|
kvm_update_guest_debug(cpu, 0);
|
2013-06-21 20:20:45 +02:00
|
|
|
} else {
|
2009-05-03 13:15:06 +02:00
|
|
|
/* must flush all the translated code to avoid inconsistencies */
|
2009-03-12 21:12:48 +01:00
|
|
|
/* XXX: only flush what is necessary */
|
2015-06-24 04:31:15 +02:00
|
|
|
tb_flush(cpu);
|
2009-03-12 21:12:48 +01:00
|
|
|
}
|
2003-07-29 22:50:33 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-09-03 17:38:47 +02:00
|
|
|
void cpu_abort(CPUState *cpu, const char *fmt, ...)
|
2003-06-21 15:11:07 +02:00
|
|
|
{
|
|
|
|
va_list ap;
|
2007-11-23 17:53:59 +01:00
|
|
|
va_list ap2;
|
2003-06-21 15:11:07 +02:00
|
|
|
|
|
|
|
va_start(ap, fmt);
|
2007-11-23 17:53:59 +01:00
|
|
|
va_copy(ap2, ap);
|
2003-06-21 15:11:07 +02:00
|
|
|
fprintf(stderr, "qemu: fatal: ");
|
|
|
|
vfprintf(stderr, fmt, ap);
|
|
|
|
fprintf(stderr, "\n");
|
2013-05-27 01:33:50 +02:00
|
|
|
cpu_dump_state(cpu, stderr, fprintf, CPU_DUMP_FPU | CPU_DUMP_CCOP);
|
2015-11-13 13:16:27 +01:00
|
|
|
if (qemu_log_separate()) {
|
2016-09-23 00:17:10 +02:00
|
|
|
qemu_log_lock();
|
2009-01-15 23:34:14 +01:00
|
|
|
qemu_log("qemu: fatal: ");
|
|
|
|
qemu_log_vprintf(fmt, ap2);
|
|
|
|
qemu_log("\n");
|
2013-06-16 07:28:50 +02:00
|
|
|
log_cpu_state(cpu, CPU_DUMP_FPU | CPU_DUMP_CCOP);
|
2009-01-15 23:35:09 +01:00
|
|
|
qemu_log_flush();
|
2016-09-23 00:17:10 +02:00
|
|
|
qemu_log_unlock();
|
2009-01-15 23:34:14 +01:00
|
|
|
qemu_log_close();
|
2007-06-10 16:07:13 +02:00
|
|
|
}
|
2007-11-23 17:53:59 +01:00
|
|
|
va_end(ap2);
|
2007-09-29 14:18:20 +02:00
|
|
|
va_end(ap);
|
2015-09-17 18:25:07 +02:00
|
|
|
replay_finish();
|
2010-01-25 13:30:49 +01:00
|
|
|
#if defined(CONFIG_USER_ONLY)
|
|
|
|
{
|
|
|
|
struct sigaction act;
|
|
|
|
sigfillset(&act.sa_mask);
|
|
|
|
act.sa_handler = SIG_DFL;
|
2018-05-15 20:27:00 +02:00
|
|
|
act.sa_flags = 0;
|
2010-01-25 13:30:49 +01:00
|
|
|
sigaction(SIGABRT, &act, NULL);
|
|
|
|
}
|
|
|
|
#endif
|
2003-06-21 15:11:07 +02:00
|
|
|
abort();
|
|
|
|
}
|
|
|
|
|
2004-01-04 16:48:17 +01:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2013-09-05 20:41:35 +02:00
|
|
|
/* Called from RCU critical section */
|
2013-09-09 17:49:45 +02:00
|
|
|
static RAMBlock *qemu_get_ram_block(ram_addr_t addr)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
|
2013-09-09 17:58:40 +02:00
|
|
|
block = atomic_rcu_read(&ram_list.mru_block);
|
2014-12-15 21:55:32 +01:00
|
|
|
if (block && addr - block->offset < block->max_length) {
|
2015-10-22 13:51:30 +02:00
|
|
|
return block;
|
2013-09-09 17:49:45 +02:00
|
|
|
}
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2014-12-15 21:55:32 +01:00
|
|
|
if (addr - block->offset < block->max_length) {
|
2013-09-09 17:49:45 +02:00
|
|
|
goto found;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fprintf(stderr, "Bad ram offset %" PRIx64 "\n", (uint64_t)addr);
|
|
|
|
abort();
|
|
|
|
|
|
|
|
found:
|
2013-09-09 17:58:40 +02:00
|
|
|
/* 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.
|
|
|
|
*/
|
2013-09-09 17:49:45 +02:00
|
|
|
ram_list.mru_block = block;
|
|
|
|
return block;
|
|
|
|
}
|
|
|
|
|
2013-10-10 11:49:53 +02:00
|
|
|
static void tlb_reset_dirty_range_all(ram_addr_t start, ram_addr_t length)
|
2012-05-22 00:42:40 +02:00
|
|
|
{
|
2015-09-11 07:39:41 +02:00
|
|
|
CPUState *cpu;
|
2013-09-09 17:49:45 +02:00
|
|
|
ram_addr_t start1;
|
2013-10-10 11:49:53 +02:00
|
|
|
RAMBlock *block;
|
|
|
|
ram_addr_t end;
|
|
|
|
|
2018-06-22 19:45:31 +02:00
|
|
|
assert(tcg_enabled());
|
2013-10-10 11:49:53 +02:00
|
|
|
end = TARGET_PAGE_ALIGN(start + length);
|
|
|
|
start &= TARGET_PAGE_MASK;
|
2012-05-22 00:42:40 +02:00
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_lock();
|
2013-09-09 17:49:45 +02:00
|
|
|
block = qemu_get_ram_block(start);
|
|
|
|
assert(block == qemu_get_ram_block(end - 1));
|
2014-11-12 10:44:41 +01:00
|
|
|
start1 = (uintptr_t)ramblock_ptr(block, start - block->offset);
|
2015-09-11 07:39:41 +02:00
|
|
|
CPU_FOREACH(cpu) {
|
|
|
|
tlb_reset_dirty(cpu, start1, length);
|
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2012-05-22 00:42:40 +02:00
|
|
|
}
|
|
|
|
|
2009-04-11 16:47:08 +02:00
|
|
|
/* Note: start and end must be within the same ram block. */
|
2014-12-02 12:23:18 +01:00
|
|
|
bool cpu_physical_memory_test_and_clear_dirty(ram_addr_t start,
|
|
|
|
ram_addr_t length,
|
|
|
|
unsigned client)
|
2004-02-06 20:46:14 +01:00
|
|
|
{
|
2016-01-25 14:33:20 +01:00
|
|
|
DirtyMemoryBlocks *blocks;
|
2014-12-02 12:23:18 +01:00
|
|
|
unsigned long end, page;
|
2016-01-25 14:33:20 +01:00
|
|
|
bool dirty = false;
|
2014-12-02 12:23:18 +01:00
|
|
|
|
|
|
|
if (length == 0) {
|
|
|
|
return false;
|
|
|
|
}
|
2005-08-21 21:12:28 +02:00
|
|
|
|
2014-12-02 12:23:18 +01:00
|
|
|
end = TARGET_PAGE_ALIGN(start + length) >> TARGET_PAGE_BITS;
|
|
|
|
page = start >> TARGET_PAGE_BITS;
|
2016-01-25 14:33:20 +01:00
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
|
|
|
|
blocks = atomic_rcu_read(&ram_list.dirty_memory[client]);
|
|
|
|
|
|
|
|
while (page < end) {
|
|
|
|
unsigned long idx = page / DIRTY_MEMORY_BLOCK_SIZE;
|
|
|
|
unsigned long offset = page % DIRTY_MEMORY_BLOCK_SIZE;
|
|
|
|
unsigned long num = MIN(end - page, DIRTY_MEMORY_BLOCK_SIZE - offset);
|
|
|
|
|
|
|
|
dirty |= bitmap_test_and_clear_atomic(blocks->blocks[idx],
|
|
|
|
offset, num);
|
|
|
|
page += num;
|
|
|
|
}
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
2014-12-02 12:23:18 +01:00
|
|
|
|
|
|
|
if (dirty && tcg_enabled()) {
|
2013-10-10 11:49:53 +02:00
|
|
|
tlb_reset_dirty_range_all(start, length);
|
2009-04-11 16:47:08 +02:00
|
|
|
}
|
2014-12-02 12:23:18 +01:00
|
|
|
|
|
|
|
return dirty;
|
2004-02-06 20:46:14 +01:00
|
|
|
}
|
|
|
|
|
2017-04-21 11:16:25 +02:00
|
|
|
DirtyBitmapSnapshot *cpu_physical_memory_snapshot_and_clear_dirty
|
|
|
|
(ram_addr_t start, ram_addr_t length, unsigned client)
|
|
|
|
{
|
|
|
|
DirtyMemoryBlocks *blocks;
|
|
|
|
unsigned long align = 1UL << (TARGET_PAGE_BITS + BITS_PER_LEVEL);
|
|
|
|
ram_addr_t first = QEMU_ALIGN_DOWN(start, align);
|
|
|
|
ram_addr_t last = QEMU_ALIGN_UP(start + length, align);
|
|
|
|
DirtyBitmapSnapshot *snap;
|
|
|
|
unsigned long page, end, dest;
|
|
|
|
|
|
|
|
snap = g_malloc0(sizeof(*snap) +
|
|
|
|
((last - first) >> (TARGET_PAGE_BITS + 3)));
|
|
|
|
snap->start = first;
|
|
|
|
snap->end = last;
|
|
|
|
|
|
|
|
page = first >> TARGET_PAGE_BITS;
|
|
|
|
end = last >> TARGET_PAGE_BITS;
|
|
|
|
dest = 0;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
|
|
|
|
blocks = atomic_rcu_read(&ram_list.dirty_memory[client]);
|
|
|
|
|
|
|
|
while (page < end) {
|
|
|
|
unsigned long idx = page / DIRTY_MEMORY_BLOCK_SIZE;
|
|
|
|
unsigned long offset = page % DIRTY_MEMORY_BLOCK_SIZE;
|
|
|
|
unsigned long num = MIN(end - page, DIRTY_MEMORY_BLOCK_SIZE - offset);
|
|
|
|
|
|
|
|
assert(QEMU_IS_ALIGNED(offset, (1 << BITS_PER_LEVEL)));
|
|
|
|
assert(QEMU_IS_ALIGNED(num, (1 << BITS_PER_LEVEL)));
|
|
|
|
offset >>= BITS_PER_LEVEL;
|
|
|
|
|
|
|
|
bitmap_copy_and_clear_atomic(snap->dirty + dest,
|
|
|
|
blocks->blocks[idx] + offset,
|
|
|
|
num);
|
|
|
|
page += num;
|
|
|
|
dest += num >> BITS_PER_LEVEL;
|
|
|
|
}
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
if (tcg_enabled()) {
|
|
|
|
tlb_reset_dirty_range_all(start, length);
|
|
|
|
}
|
|
|
|
|
|
|
|
return snap;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool cpu_physical_memory_snapshot_get_dirty(DirtyBitmapSnapshot *snap,
|
|
|
|
ram_addr_t start,
|
|
|
|
ram_addr_t length)
|
|
|
|
{
|
|
|
|
unsigned long page, end;
|
|
|
|
|
|
|
|
assert(start >= snap->start);
|
|
|
|
assert(start + length <= snap->end);
|
|
|
|
|
|
|
|
end = TARGET_PAGE_ALIGN(start + length - snap->start) >> TARGET_PAGE_BITS;
|
|
|
|
page = (start - snap->start) >> TARGET_PAGE_BITS;
|
|
|
|
|
|
|
|
while (page < end) {
|
|
|
|
if (test_bit(page, snap->dirty)) {
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
page++;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2015-01-21 12:09:14 +01:00
|
|
|
/* Called from RCU critical section */
|
2013-09-03 13:32:01 +02:00
|
|
|
hwaddr memory_region_section_get_iotlb(CPUState *cpu,
|
2013-05-24 12:59:37 +02:00
|
|
|
MemoryRegionSection *section,
|
|
|
|
target_ulong vaddr,
|
|
|
|
hwaddr paddr, hwaddr xlat,
|
|
|
|
int prot,
|
|
|
|
target_ulong *address)
|
2012-04-21 15:08:33 +02:00
|
|
|
{
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr iotlb;
|
2012-04-21 15:08:33 +02:00
|
|
|
CPUWatchpoint *wp;
|
|
|
|
|
2012-04-14 16:56:48 +02:00
|
|
|
if (memory_region_is_ram(section->mr)) {
|
2012-04-21 15:08:33 +02:00
|
|
|
/* Normal RAM. */
|
2016-03-01 10:44:50 +01:00
|
|
|
iotlb = memory_region_get_ram_addr(section->mr) + xlat;
|
2012-04-21 15:08:33 +02:00
|
|
|
if (!section->readonly) {
|
2013-05-29 11:09:17 +02:00
|
|
|
iotlb |= PHYS_SECTION_NOTDIRTY;
|
2012-04-21 15:08:33 +02:00
|
|
|
} else {
|
2013-05-29 11:09:17 +02:00
|
|
|
iotlb |= PHYS_SECTION_ROM;
|
2012-04-21 15:08:33 +02:00
|
|
|
}
|
|
|
|
} else {
|
2015-07-20 13:27:16 +02:00
|
|
|
AddressSpaceDispatch *d;
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
d = flatview_to_dispatch(section->fv);
|
2015-07-20 13:27:16 +02:00
|
|
|
iotlb = section - d->map.sections;
|
2013-05-24 12:59:37 +02:00
|
|
|
iotlb += xlat;
|
2012-04-21 15:08:33 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Make accesses to pages with watchpoints go via the
|
|
|
|
watchpoint trap routines. */
|
2013-08-26 18:23:18 +02:00
|
|
|
QTAILQ_FOREACH(wp, &cpu->watchpoints, entry) {
|
2014-09-12 15:06:48 +02:00
|
|
|
if (cpu_watchpoint_address_matches(wp, vaddr, TARGET_PAGE_SIZE)) {
|
2012-04-21 15:08:33 +02:00
|
|
|
/* Avoid trapping reads of pages with a write breakpoint. */
|
|
|
|
if ((prot & PAGE_WRITE) || (wp->flags & BP_MEM_READ)) {
|
2013-05-29 11:09:17 +02:00
|
|
|
iotlb = PHYS_SECTION_WATCH + paddr;
|
2012-04-21 15:08:33 +02:00
|
|
|
*address |= TLB_MMIO;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return iotlb;
|
|
|
|
}
|
2004-01-04 19:06:42 +01:00
|
|
|
#endif /* defined(CONFIG_USER_ONLY) */
|
|
|
|
|
2008-06-08 03:09:01 +02:00
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
2008-12-01 19:59:50 +01:00
|
|
|
|
2009-10-01 23:12:16 +02:00
|
|
|
static int subpage_register (subpage_t *mmio, uint32_t start, uint32_t end,
|
2012-02-12 17:32:55 +01:00
|
|
|
uint16_t section);
|
2017-09-21 10:50:58 +02:00
|
|
|
static subpage_t *subpage_init(FlatView *fv, hwaddr base);
|
2012-02-09 16:34:32 +01:00
|
|
|
|
2017-12-13 15:37:37 +01:00
|
|
|
static void *(*phys_mem_alloc)(size_t size, uint64_t *align, bool shared) =
|
2014-10-31 17:38:37 +01:00
|
|
|
qemu_anon_ram_alloc;
|
2013-07-31 15:11:08 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Set a custom physical guest memory alloator.
|
|
|
|
* Accelerators with unusual needs may need this. Hopefully, we can
|
|
|
|
* get rid of it eventually.
|
|
|
|
*/
|
2017-12-13 15:37:37 +01:00
|
|
|
void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align, bool shared))
|
2013-07-31 15:11:08 +02:00
|
|
|
{
|
|
|
|
phys_mem_alloc = alloc;
|
|
|
|
}
|
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
static uint16_t phys_section_add(PhysPageMap *map,
|
|
|
|
MemoryRegionSection *section)
|
2012-02-12 17:32:55 +01:00
|
|
|
{
|
2013-05-07 11:30:23 +02:00
|
|
|
/* The physical section number is ORed with a page-aligned
|
|
|
|
* pointer to produce the iotlb entries. Thus it should
|
|
|
|
* never overflow into the page-aligned value.
|
|
|
|
*/
|
2013-12-01 13:02:23 +01:00
|
|
|
assert(map->sections_nb < TARGET_PAGE_SIZE);
|
2013-05-07 11:30:23 +02:00
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
if (map->sections_nb == map->sections_nb_alloc) {
|
|
|
|
map->sections_nb_alloc = MAX(map->sections_nb_alloc * 2, 16);
|
|
|
|
map->sections = g_renew(MemoryRegionSection, map->sections,
|
|
|
|
map->sections_nb_alloc);
|
2012-02-12 17:32:55 +01:00
|
|
|
}
|
2013-12-01 13:02:23 +01:00
|
|
|
map->sections[map->sections_nb] = *section;
|
2013-05-06 10:46:11 +02:00
|
|
|
memory_region_ref(section->mr);
|
2013-12-01 13:02:23 +01:00
|
|
|
return map->sections_nb++;
|
2012-02-12 17:32:55 +01:00
|
|
|
}
|
|
|
|
|
2013-06-25 09:30:48 +02:00
|
|
|
static void phys_section_destroy(MemoryRegion *mr)
|
|
|
|
{
|
2015-11-30 23:11:04 +01:00
|
|
|
bool have_sub_page = mr->subpage;
|
|
|
|
|
2013-05-06 10:46:11 +02:00
|
|
|
memory_region_unref(mr);
|
|
|
|
|
2015-11-30 23:11:04 +01:00
|
|
|
if (have_sub_page) {
|
2013-06-25 09:30:48 +02:00
|
|
|
subpage_t *subpage = container_of(mr, subpage_t, iomem);
|
2014-06-06 08:15:52 +02:00
|
|
|
object_unref(OBJECT(&subpage->iomem));
|
2013-06-25 09:30:48 +02:00
|
|
|
g_free(subpage);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-05-29 12:30:26 +02:00
|
|
|
static void phys_sections_free(PhysPageMap *map)
|
2012-02-12 17:32:55 +01:00
|
|
|
{
|
2013-05-29 12:09:47 +02:00
|
|
|
while (map->sections_nb > 0) {
|
|
|
|
MemoryRegionSection *section = &map->sections[--map->sections_nb];
|
2013-06-25 09:30:48 +02:00
|
|
|
phys_section_destroy(section->mr);
|
|
|
|
}
|
2013-05-29 12:09:47 +02:00
|
|
|
g_free(map->sections);
|
|
|
|
g_free(map->nodes);
|
2012-02-12 17:32:55 +01:00
|
|
|
}
|
|
|
|
|
2017-09-21 10:50:59 +02:00
|
|
|
static void register_subpage(FlatView *fv, MemoryRegionSection *section)
|
2012-02-13 16:14:32 +01:00
|
|
|
{
|
2017-09-21 10:50:59 +02:00
|
|
|
AddressSpaceDispatch *d = flatview_to_dispatch(fv);
|
2012-02-13 16:14:32 +01:00
|
|
|
subpage_t *subpage;
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr base = section->offset_within_address_space
|
2012-02-13 16:14:32 +01:00
|
|
|
& TARGET_PAGE_MASK;
|
2017-05-15 10:50:57 +02:00
|
|
|
MemoryRegionSection *existing = phys_page_find(d, base);
|
2012-02-13 16:14:32 +01:00
|
|
|
MemoryRegionSection subsection = {
|
|
|
|
.offset_within_address_space = base,
|
2013-05-27 10:08:27 +02:00
|
|
|
.size = int128_make64(TARGET_PAGE_SIZE),
|
2012-02-13 16:14:32 +01:00
|
|
|
};
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr start, end;
|
2012-02-13 16:14:32 +01:00
|
|
|
|
2012-03-08 15:16:34 +01:00
|
|
|
assert(existing->mr->subpage || existing->mr == &io_mem_unassigned);
|
2012-02-13 16:14:32 +01:00
|
|
|
|
2012-03-08 15:16:34 +01:00
|
|
|
if (!(existing->mr->subpage)) {
|
2017-09-21 10:50:58 +02:00
|
|
|
subpage = subpage_init(fv, base);
|
|
|
|
subsection.fv = fv;
|
2012-02-13 16:14:32 +01:00
|
|
|
subsection.mr = &subpage->iomem;
|
2012-10-03 16:22:53 +02:00
|
|
|
phys_page_set(d, base >> TARGET_PAGE_BITS, 1,
|
2013-12-01 13:02:23 +01:00
|
|
|
phys_section_add(&d->map, &subsection));
|
2012-02-13 16:14:32 +01:00
|
|
|
} else {
|
2012-03-08 15:16:34 +01:00
|
|
|
subpage = container_of(existing->mr, subpage_t, iomem);
|
2012-02-13 16:14:32 +01:00
|
|
|
}
|
|
|
|
start = section->offset_within_address_space & ~TARGET_PAGE_MASK;
|
2013-05-27 10:08:27 +02:00
|
|
|
end = start + int128_get64(section->size) - 1;
|
2013-12-01 13:02:23 +01:00
|
|
|
subpage_register(subpage, start, end,
|
|
|
|
phys_section_add(&d->map, section));
|
2012-02-13 16:14:32 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2017-09-21 10:50:59 +02:00
|
|
|
static void register_multipage(FlatView *fv,
|
2013-05-27 10:08:27 +02:00
|
|
|
MemoryRegionSection *section)
|
2003-08-10 23:47:01 +02:00
|
|
|
{
|
2017-09-21 10:50:59 +02:00
|
|
|
AddressSpaceDispatch *d = flatview_to_dispatch(fv);
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr start_addr = section->offset_within_address_space;
|
2013-12-01 13:02:23 +01:00
|
|
|
uint16_t section_index = phys_section_add(&d->map, section);
|
2013-05-27 10:08:27 +02:00
|
|
|
uint64_t num_pages = int128_get64(int128_rshift(section->size,
|
|
|
|
TARGET_PAGE_BITS));
|
2012-01-02 11:17:03 +01:00
|
|
|
|
2013-05-27 10:47:10 +02:00
|
|
|
assert(num_pages);
|
|
|
|
phys_page_set(d, start_addr >> TARGET_PAGE_BITS, num_pages, section_index);
|
2003-08-10 23:47:01 +02:00
|
|
|
}
|
|
|
|
|
2017-09-21 10:51:00 +02:00
|
|
|
void flatview_add_to_dispatch(FlatView *fv, MemoryRegionSection *section)
|
2012-02-13 16:14:32 +01:00
|
|
|
{
|
2013-05-27 13:18:01 +02:00
|
|
|
MemoryRegionSection now = *section, remain = *section;
|
2013-05-27 10:08:27 +02:00
|
|
|
Int128 page_size = int128_make64(TARGET_PAGE_SIZE);
|
2012-02-13 16:14:32 +01:00
|
|
|
|
2013-05-27 10:47:10 +02:00
|
|
|
if (now.offset_within_address_space & ~TARGET_PAGE_MASK) {
|
|
|
|
uint64_t left = TARGET_PAGE_ALIGN(now.offset_within_address_space)
|
|
|
|
- now.offset_within_address_space;
|
|
|
|
|
2013-05-27 10:08:27 +02:00
|
|
|
now.size = int128_min(int128_make64(left), now.size);
|
2017-09-21 10:50:59 +02:00
|
|
|
register_subpage(fv, &now);
|
2013-05-27 10:47:10 +02:00
|
|
|
} else {
|
2013-05-27 10:08:27 +02:00
|
|
|
now.size = int128_zero();
|
2013-05-27 10:47:10 +02:00
|
|
|
}
|
2013-05-27 10:08:27 +02:00
|
|
|
while (int128_ne(remain.size, now.size)) {
|
|
|
|
remain.size = int128_sub(remain.size, now.size);
|
|
|
|
remain.offset_within_address_space += int128_get64(now.size);
|
|
|
|
remain.offset_within_region += int128_get64(now.size);
|
2012-07-26 00:45:04 +02:00
|
|
|
now = remain;
|
2013-05-27 10:08:27 +02:00
|
|
|
if (int128_lt(remain.size, page_size)) {
|
2017-09-21 10:50:59 +02:00
|
|
|
register_subpage(fv, &now);
|
2013-08-29 12:21:16 +02:00
|
|
|
} else if (remain.offset_within_address_space & ~TARGET_PAGE_MASK) {
|
2013-05-27 10:08:27 +02:00
|
|
|
now.size = page_size;
|
2017-09-21 10:50:59 +02:00
|
|
|
register_subpage(fv, &now);
|
2012-07-26 00:45:04 +02:00
|
|
|
} else {
|
2013-05-27 10:08:27 +02:00
|
|
|
now.size = int128_and(now.size, int128_neg(page_size));
|
2017-09-21 10:50:59 +02:00
|
|
|
register_multipage(fv, &now);
|
2012-07-26 00:45:04 +02:00
|
|
|
}
|
2012-02-13 16:14:32 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-01-26 12:21:16 +01:00
|
|
|
void qemu_flush_coalesced_mmio_buffer(void)
|
|
|
|
{
|
|
|
|
if (kvm_enabled())
|
|
|
|
kvm_flush_coalesced_mmio_buffer();
|
|
|
|
}
|
|
|
|
|
2011-08-17 09:01:33 +02:00
|
|
|
void qemu_mutex_lock_ramlist(void)
|
|
|
|
{
|
|
|
|
qemu_mutex_lock(&ram_list.mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
void qemu_mutex_unlock_ramlist(void)
|
|
|
|
{
|
|
|
|
qemu_mutex_unlock(&ram_list.mutex);
|
|
|
|
}
|
|
|
|
|
2017-05-12 06:17:41 +02:00
|
|
|
void ram_block_dump(Monitor *mon)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
char *psize;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
monitor_printf(mon, "%24s %8s %18s %18s %18s\n",
|
|
|
|
"Block Name", "PSize", "Offset", "Used", "Total");
|
|
|
|
RAMBLOCK_FOREACH(block) {
|
|
|
|
psize = size_to_str(block->page_size);
|
|
|
|
monitor_printf(mon, "%24s %8s 0x%016" PRIx64 " 0x%016" PRIx64
|
|
|
|
" 0x%016" PRIx64 "\n", block->idstr, psize,
|
|
|
|
(uint64_t)block->offset,
|
|
|
|
(uint64_t)block->used_length,
|
|
|
|
(uint64_t)block->max_length);
|
|
|
|
g_free(psize);
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
2017-03-02 03:36:11 +01:00
|
|
|
#ifdef __linux__
|
|
|
|
/*
|
|
|
|
* FIXME TOCTTOU: this iterates over memory backends' mem-path, which
|
|
|
|
* may or may not name the same files / on the same filesystem now as
|
|
|
|
* when we actually open and map them. Iterate over the file
|
|
|
|
* descriptors instead, and use qemu_fd_getpagesize().
|
|
|
|
*/
|
|
|
|
static int find_max_supported_pagesize(Object *obj, void *opaque)
|
|
|
|
{
|
|
|
|
long *hpsize_min = opaque;
|
|
|
|
|
|
|
|
if (object_dynamic_cast(obj, TYPE_MEMORY_BACKEND)) {
|
2018-04-03 07:05:45 +02:00
|
|
|
long hpsize = host_memory_backend_pagesize(MEMORY_BACKEND(obj));
|
|
|
|
|
2018-04-03 06:55:11 +02:00
|
|
|
if (hpsize < *hpsize_min) {
|
|
|
|
*hpsize_min = hpsize;
|
2017-03-02 03:36:11 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
long qemu_getrampagesize(void)
|
|
|
|
{
|
|
|
|
long hpsize = LONG_MAX;
|
|
|
|
long mainrampagesize;
|
|
|
|
Object *memdev_root;
|
|
|
|
|
2018-04-03 06:55:11 +02:00
|
|
|
mainrampagesize = qemu_mempath_getpagesize(mem_path);
|
2017-03-02 03:36:11 +01:00
|
|
|
|
|
|
|
/* it's possible we have memory-backend objects with
|
|
|
|
* hugepage-backed RAM. these may get mapped into system
|
|
|
|
* address space via -numa parameters or memory hotplug
|
|
|
|
* hooks. we want to take these into account, but we
|
|
|
|
* also want to make sure these supported hugepage
|
|
|
|
* sizes are applicable across the entire range of memory
|
|
|
|
* we may boot from, so we take the min across all
|
|
|
|
* backends, and assume normal pages in cases where a
|
|
|
|
* backend isn't backed by hugepages.
|
|
|
|
*/
|
|
|
|
memdev_root = object_resolve_path("/objects", NULL);
|
|
|
|
if (memdev_root) {
|
|
|
|
object_child_foreach(memdev_root, find_max_supported_pagesize, &hpsize);
|
|
|
|
}
|
|
|
|
if (hpsize == LONG_MAX) {
|
|
|
|
/* No additional memory regions found ==> Report main RAM page size */
|
|
|
|
return mainrampagesize;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If NUMA is disabled or the NUMA nodes are not backed with a
|
|
|
|
* memory-backend, then there is at least one node using "normal" RAM,
|
|
|
|
* so if its page size is smaller we have got to report that size instead.
|
|
|
|
*/
|
|
|
|
if (hpsize > mainrampagesize &&
|
|
|
|
(nb_numa_nodes == 0 || numa_info[0].node_memdev == NULL)) {
|
|
|
|
static bool warned;
|
|
|
|
if (!warned) {
|
|
|
|
error_report("Huge page support disabled (n/a for main memory).");
|
|
|
|
warned = true;
|
|
|
|
}
|
|
|
|
return mainrampagesize;
|
|
|
|
}
|
|
|
|
|
|
|
|
return hpsize;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
long qemu_getrampagesize(void)
|
|
|
|
{
|
|
|
|
return getpagesize();
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2013-07-31 15:11:10 +02:00
|
|
|
#ifdef __linux__
|
2016-10-27 06:22:58 +02:00
|
|
|
static int64_t get_file_size(int fd)
|
|
|
|
{
|
|
|
|
int64_t size = lseek(fd, 0, SEEK_END);
|
|
|
|
if (size < 0) {
|
|
|
|
return -errno;
|
|
|
|
}
|
|
|
|
return size;
|
|
|
|
}
|
|
|
|
|
2017-06-02 16:12:22 +02:00
|
|
|
static int file_ram_open(const char *path,
|
|
|
|
const char *region_name,
|
|
|
|
bool *created,
|
|
|
|
Error **errp)
|
2010-03-02 00:25:08 +01:00
|
|
|
{
|
|
|
|
char *filename;
|
2013-03-04 19:54:25 +01:00
|
|
|
char *sanitized_name;
|
|
|
|
char *c;
|
2016-03-17 15:53:13 +01:00
|
|
|
int fd = -1;
|
2010-03-02 00:25:08 +01:00
|
|
|
|
2017-06-02 16:12:22 +02:00
|
|
|
*created = false;
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
for (;;) {
|
|
|
|
fd = open(path, O_RDWR);
|
|
|
|
if (fd >= 0) {
|
|
|
|
/* @path names an existing file, use it */
|
|
|
|
break;
|
2015-10-28 10:54:07 +01:00
|
|
|
}
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
if (errno == ENOENT) {
|
|
|
|
/* @path names a file that doesn't exist, create it */
|
|
|
|
fd = open(path, O_RDWR | O_CREAT | O_EXCL, 0644);
|
|
|
|
if (fd >= 0) {
|
2017-06-02 16:12:22 +02:00
|
|
|
*created = true;
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
} else if (errno == EISDIR) {
|
|
|
|
/* @path names a directory, create a file there */
|
|
|
|
/* Make name safe to use with mkstemp by replacing '/' with '_'. */
|
2017-06-02 16:12:22 +02:00
|
|
|
sanitized_name = g_strdup(region_name);
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
for (c = sanitized_name; *c != '\0'; c++) {
|
|
|
|
if (*c == '/') {
|
|
|
|
*c = '_';
|
|
|
|
}
|
|
|
|
}
|
2013-03-04 19:54:25 +01:00
|
|
|
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
filename = g_strdup_printf("%s/qemu_back_mem.%s.XXXXXX", path,
|
|
|
|
sanitized_name);
|
|
|
|
g_free(sanitized_name);
|
2015-10-28 10:54:07 +01:00
|
|
|
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
fd = mkstemp(filename);
|
|
|
|
if (fd >= 0) {
|
|
|
|
unlink(filename);
|
|
|
|
g_free(filename);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
g_free(filename);
|
2015-10-28 10:54:07 +01:00
|
|
|
}
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
if (errno != EEXIST && errno != EINTR) {
|
|
|
|
error_setg_errno(errp, errno,
|
|
|
|
"can't open backing store %s for guest RAM",
|
|
|
|
path);
|
2017-06-02 16:12:22 +02:00
|
|
|
return -1;
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Try again on EINTR and EEXIST. The latter happens when
|
|
|
|
* something else creates the file between our two open().
|
|
|
|
*/
|
2015-10-28 10:54:07 +01:00
|
|
|
}
|
2010-03-02 00:25:08 +01:00
|
|
|
|
2017-06-02 16:12:22 +02:00
|
|
|
return fd;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *file_ram_alloc(RAMBlock *block,
|
|
|
|
ram_addr_t memory,
|
|
|
|
int fd,
|
|
|
|
bool truncate,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
void *area;
|
|
|
|
|
2016-09-29 21:09:37 +02:00
|
|
|
block->page_size = qemu_fd_getpagesize(fd);
|
hostmem-file: add "align" option
When mmap(2) the backend files, QEMU uses the host page size
(getpagesize(2)) by default as the alignment of mapping address.
However, some backends may require alignments different than the page
size. For example, mmap a device DAX (e.g., /dev/dax0.0) on Linux
kernel 4.13 to an address, which is 4K-aligned but not 2M-aligned,
fails with a kernel message like
[617494.969768] dax dax0.0: qemu-system-x86: dax_mmap: fail, unaligned vma (0x7fa37c579000 - 0x7fa43c579000, 0x1fffff)
Because there is no common approach to get such alignment requirement,
we add the 'align' option to 'memory-backend-file', so that users or
management utils, which have enough knowledge about the backend, can
specify a proper alignment via this option.
Signed-off-by: Haozhong Zhang <haozhong.zhang@intel.com>
Message-Id: <20171211072806.2812-2-haozhong.zhang@intel.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
[ehabkost: fixed typo, fixed error_setg() format string]
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
2017-12-11 08:28:04 +01:00
|
|
|
if (block->mr->align % block->page_size) {
|
|
|
|
error_setg(errp, "alignment 0x%" PRIx64
|
|
|
|
" must be multiples of page size 0x%zx",
|
|
|
|
block->mr->align, block->page_size);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
block->mr->align = MAX(block->page_size, block->mr->align);
|
exec.c: workaround regression caused by alignment change in d2f39ad
Commit d2f39ad "exec.c: Ensure right alignment also for file backed ram"
added an additional alignment requirement on the size of backend file
besides the previous page size. On x86, the alignment is changed from
4KB in QEMU 2.6 to 2MB in QEMU 2.7.
This change breaks certain usages in QEMU 2.7 on x86, e.g.
-object memory-backend-file,id=mem1,mem-path=/tmp/,size=$SZ
-device pc-dimm,id=dimm1,memdev=mem1
where $SZ is multiple of 4KB but not 2MB (e.g. 1023M). QEMU 2.7
reports the following error message and aborts:
qemu-system-x86_64: -device pc-dimm,memdev=mem1,id=nv1: backend memory size must be multiple of 0x200000
The same regression may also happen in other platforms as indicated by
Igor Mammedov. This change is however necessary for s390 according to
the commit message of d2f39ad, so we workaround the regression by taking
the change only on s390.
Signed-off-by: Haozhong Zhang <haozhong.zhang@intel.com>
Reported-by: "Xu, Anthony" <anthony.xu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-10-24 14:49:37 +02:00
|
|
|
#if defined(__s390x__)
|
|
|
|
if (kvm_enabled()) {
|
|
|
|
block->mr->align = MAX(block->mr->align, QEMU_VMALLOC_ALIGN);
|
|
|
|
}
|
|
|
|
#endif
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
|
2016-09-29 21:09:37 +02:00
|
|
|
if (memory < block->page_size) {
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
error_setg(errp, "memory size 0x" RAM_ADDR_FMT " must be equal to "
|
2016-09-29 21:09:37 +02:00
|
|
|
"or larger than page size 0x%zx",
|
|
|
|
memory, block->page_size);
|
2017-06-02 16:12:22 +02:00
|
|
|
return NULL;
|
2016-11-02 02:05:51 +01:00
|
|
|
}
|
|
|
|
|
2016-09-29 21:09:37 +02:00
|
|
|
memory = ROUND_UP(memory, block->page_size);
|
2010-03-02 00:25:08 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* ftruncate is not supported by hugetlbfs in older
|
|
|
|
* hosts, so don't bother bailing out on errors.
|
|
|
|
* If anything goes wrong with it under other filesystems,
|
|
|
|
* mmap will fail.
|
2016-10-27 06:22:58 +02:00
|
|
|
*
|
|
|
|
* Do not truncate the non-empty backend file to avoid corrupting
|
|
|
|
* the existing data in the file. Disabling shrinking is not
|
|
|
|
* enough. For example, the current vNVDIMM implementation stores
|
|
|
|
* the guest NVDIMM labels at the end of the backend file. If the
|
|
|
|
* backend file is later extended, QEMU will not be able to find
|
|
|
|
* those labels. Therefore, extending the non-empty backend file
|
|
|
|
* is disabled as well.
|
2010-03-02 00:25:08 +01:00
|
|
|
*/
|
2017-06-02 16:12:22 +02:00
|
|
|
if (truncate && ftruncate(fd, memory)) {
|
2010-08-18 06:30:13 +02:00
|
|
|
perror("ftruncate");
|
2014-05-14 11:43:20 +02:00
|
|
|
}
|
2010-03-02 00:25:08 +01:00
|
|
|
|
2016-04-25 13:55:38 +02:00
|
|
|
area = qemu_ram_mmap(fd, memory, block->mr->align,
|
|
|
|
block->flags & RAM_SHARED);
|
2010-03-02 00:25:08 +01:00
|
|
|
if (area == MAP_FAILED) {
|
2014-05-14 11:43:20 +02:00
|
|
|
error_setg_errno(errp, errno,
|
exec: Fix memory allocation when memory path names new file
Commit 8d31d6b extended file_ram_alloc() to accept file names in
addition to directory names. Even though it passes O_CREAT to open(),
it actually works only for existing files. Reproducer adapted from
the commit's qemu-doc.texi update:
$ qemu-system-x86_64 -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1
qemu-system-x86_64: -object memory-backend-file,size=2M,mem-path=/dev/hugepages/my-shmem-file,id=mb1: failed to get page size of file /dev/hugepages/my-shmem-file: No such file or directory
This is because we first get the page size for @path, then open the
actual file. Unwise even before the flawed commit, because the
directory could change in between, invalidating the page size.
Unlikely to bite in practice.
Rearrange the code to create the file (if necessary) before getting
its page size. Carefully avoid TOCTTOU conditions with a method
suggested by Paolo Bonzini.
While there, replace "hugepages" by "guest RAM" in error messages,
because host memory backends can be used for purposes other than huge
pages, e.g. /dev/shm/ shared memory. Help text of -mem-path agrees.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1457378754-21649-2-git-send-email-armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-07 20:25:13 +01:00
|
|
|
"unable to map backing store for guest RAM");
|
2017-06-02 16:12:22 +02:00
|
|
|
return NULL;
|
2010-03-02 00:25:08 +01:00
|
|
|
}
|
2013-10-28 21:51:46 +01:00
|
|
|
|
|
|
|
if (mem_prealloc) {
|
2017-02-24 04:31:43 +01:00
|
|
|
os_mem_prealloc(fd, area, memory, smp_cpus, errp);
|
2016-07-20 11:54:03 +02:00
|
|
|
if (errp && *errp) {
|
2017-06-02 16:12:22 +02:00
|
|
|
qemu_ram_munmap(area, memory);
|
|
|
|
return NULL;
|
2016-07-20 11:54:03 +02:00
|
|
|
}
|
2013-10-28 21:51:46 +01:00
|
|
|
}
|
|
|
|
|
2010-07-02 19:13:17 +02:00
|
|
|
block->fd = fd;
|
2010-03-02 00:25:08 +01:00
|
|
|
return area;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2018-01-05 18:01:37 +01:00
|
|
|
/* Allocate space within the ram_addr_t space that governs the
|
|
|
|
* dirty bitmaps.
|
|
|
|
* Called with the ramlist lock held.
|
|
|
|
*/
|
2010-06-25 19:08:38 +02:00
|
|
|
static ram_addr_t find_ram_offset(ram_addr_t size)
|
2010-07-02 19:13:17 +02:00
|
|
|
{
|
|
|
|
RAMBlock *block, *next_block;
|
2011-10-31 15:54:09 +01:00
|
|
|
ram_addr_t offset = RAM_ADDR_MAX, mingap = RAM_ADDR_MAX;
|
2010-07-02 19:13:17 +02:00
|
|
|
|
2013-03-11 10:20:21 +01:00
|
|
|
assert(size != 0); /* it would hand out same offset multiple times */
|
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
if (QLIST_EMPTY_RCU(&ram_list.blocks)) {
|
2010-07-02 19:13:17 +02:00
|
|
|
return 0;
|
2015-01-21 13:45:24 +01:00
|
|
|
}
|
2010-07-02 19:13:17 +02:00
|
|
|
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2018-01-05 18:01:37 +01:00
|
|
|
ram_addr_t candidate, next = RAM_ADDR_MAX;
|
2010-07-02 19:13:17 +02:00
|
|
|
|
2018-01-05 18:01:38 +01:00
|
|
|
/* Align blocks to start on a 'long' in the bitmap
|
|
|
|
* which makes the bitmap sync'ing take the fast path.
|
|
|
|
*/
|
2018-01-05 18:01:37 +01:00
|
|
|
candidate = block->offset + block->max_length;
|
2018-01-05 18:01:38 +01:00
|
|
|
candidate = ROUND_UP(candidate, BITS_PER_LONG << TARGET_PAGE_BITS);
|
2010-07-02 19:13:17 +02:00
|
|
|
|
2018-01-05 18:01:37 +01:00
|
|
|
/* Search for the closest following block
|
|
|
|
* and find the gap.
|
|
|
|
*/
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(next_block) {
|
2018-01-05 18:01:37 +01:00
|
|
|
if (next_block->offset >= candidate) {
|
2010-07-02 19:13:17 +02:00
|
|
|
next = MIN(next, next_block->offset);
|
|
|
|
}
|
|
|
|
}
|
2018-01-05 18:01:37 +01:00
|
|
|
|
|
|
|
/* If it fits remember our place and remember the size
|
|
|
|
* of gap, but keep going so that we might find a smaller
|
|
|
|
* gap to fill so avoiding fragmentation.
|
|
|
|
*/
|
|
|
|
if (next - candidate >= size && next - candidate < mingap) {
|
|
|
|
offset = candidate;
|
|
|
|
mingap = next - candidate;
|
2010-07-02 19:13:17 +02:00
|
|
|
}
|
2018-01-05 18:01:37 +01:00
|
|
|
|
|
|
|
trace_find_ram_offset_loop(size, candidate, offset, next, mingap);
|
2010-07-02 19:13:17 +02:00
|
|
|
}
|
2011-10-31 15:54:09 +01:00
|
|
|
|
|
|
|
if (offset == RAM_ADDR_MAX) {
|
|
|
|
fprintf(stderr, "Failed to find gap of requested size: %" PRIu64 "\n",
|
|
|
|
(uint64_t)size);
|
|
|
|
abort();
|
|
|
|
}
|
|
|
|
|
2018-01-05 18:01:37 +01:00
|
|
|
trace_find_ram_offset(size, offset);
|
|
|
|
|
2010-07-02 19:13:17 +02:00
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
2018-06-20 22:27:36 +02:00
|
|
|
static unsigned long last_ram_page(void)
|
2010-06-25 19:08:38 +02:00
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
ram_addr_t last = 0;
|
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_lock();
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2014-11-12 13:27:41 +01:00
|
|
|
last = MAX(last, block->offset + block->max_length);
|
2015-01-21 13:45:24 +01:00
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2017-03-21 17:44:30 +01:00
|
|
|
return last >> TARGET_PAGE_BITS;
|
2010-06-25 19:08:38 +02:00
|
|
|
}
|
|
|
|
|
2012-08-02 21:44:16 +02:00
|
|
|
static void qemu_ram_setup_dump(void *addr, ram_addr_t size)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* Use MADV_DONTDUMP, if user doesn't want the guest memory in the core */
|
2015-02-04 16:43:54 +01:00
|
|
|
if (!machine_dump_guest_core(current_machine)) {
|
2012-08-02 21:44:16 +02:00
|
|
|
ret = qemu_madvise(addr, size, QEMU_MADV_DONTDUMP);
|
|
|
|
if (ret) {
|
|
|
|
perror("qemu_madvise");
|
|
|
|
fprintf(stderr, "madvise doesn't support MADV_DONTDUMP, "
|
|
|
|
"but dump_guest_core=off specified\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:10:32 +01:00
|
|
|
const char *qemu_ram_get_idstr(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
return rb->idstr;
|
|
|
|
}
|
|
|
|
|
2017-03-07 19:36:36 +01:00
|
|
|
bool qemu_ram_is_shared(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
return rb->flags & RAM_SHARED;
|
|
|
|
}
|
|
|
|
|
2018-03-12 18:20:58 +01:00
|
|
|
/* Note: Only set at the start of postcopy */
|
|
|
|
bool qemu_ram_is_uf_zeroable(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
return rb->flags & RAM_UF_ZEROPAGE;
|
|
|
|
}
|
|
|
|
|
|
|
|
void qemu_ram_set_uf_zeroable(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
rb->flags |= RAM_UF_ZEROPAGE;
|
|
|
|
}
|
|
|
|
|
2018-05-14 08:57:00 +02:00
|
|
|
bool qemu_ram_is_migratable(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
return rb->flags & RAM_MIGRATABLE;
|
|
|
|
}
|
|
|
|
|
|
|
|
void qemu_ram_set_migratable(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
rb->flags |= RAM_MIGRATABLE;
|
|
|
|
}
|
|
|
|
|
|
|
|
void qemu_ram_unset_migratable(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
rb->flags &= ~RAM_MIGRATABLE;
|
|
|
|
}
|
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
/* Called with iothread lock held. */
|
2016-05-10 04:04:59 +02:00
|
|
|
void qemu_ram_set_idstr(RAMBlock *new_block, const char *name, DeviceState *dev)
|
2014-04-02 09:13:26 +02:00
|
|
|
{
|
2016-05-10 04:04:59 +02:00
|
|
|
RAMBlock *block;
|
2014-04-02 09:13:26 +02:00
|
|
|
|
2011-12-20 14:59:12 +01:00
|
|
|
assert(new_block);
|
|
|
|
assert(!new_block->idstr[0]);
|
2010-07-27 02:10:57 +02:00
|
|
|
|
2012-02-03 19:28:43 +01:00
|
|
|
if (dev) {
|
|
|
|
char *id = qdev_get_dev_path(dev);
|
2010-07-27 02:10:57 +02:00
|
|
|
if (id) {
|
|
|
|
snprintf(new_block->idstr, sizeof(new_block->idstr), "%s/", id);
|
2011-08-21 05:09:37 +02:00
|
|
|
g_free(id);
|
2010-07-27 02:10:57 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
pstrcat(new_block->idstr, sizeof(new_block->idstr), name);
|
|
|
|
|
2016-05-10 04:05:00 +02:00
|
|
|
rcu_read_lock();
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2016-05-10 04:04:59 +02:00
|
|
|
if (block != new_block &&
|
|
|
|
!strcmp(block->idstr, new_block->idstr)) {
|
2010-07-27 02:10:57 +02:00
|
|
|
fprintf(stderr, "RAMBlock \"%s\" already registered, abort!\n",
|
|
|
|
new_block->idstr);
|
|
|
|
abort();
|
|
|
|
}
|
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2011-12-20 14:59:12 +01:00
|
|
|
}
|
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
/* Called with iothread lock held. */
|
2016-05-10 04:04:59 +02:00
|
|
|
void qemu_ram_unset_idstr(RAMBlock *block)
|
2014-04-02 09:13:26 +02:00
|
|
|
{
|
2013-09-05 20:41:35 +02:00
|
|
|
/* FIXME: arch_init.c assumes that this is not called throughout
|
|
|
|
* migration. Ignore the problem since hot-unplug during migration
|
|
|
|
* does not work anyway.
|
|
|
|
*/
|
2014-04-02 09:13:26 +02:00
|
|
|
if (block) {
|
|
|
|
memset(block->idstr, 0, sizeof(block->idstr));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-09-29 21:09:37 +02:00
|
|
|
size_t qemu_ram_pagesize(RAMBlock *rb)
|
|
|
|
{
|
|
|
|
return rb->page_size;
|
|
|
|
}
|
|
|
|
|
2017-02-24 19:28:34 +01:00
|
|
|
/* Returns the largest size of page in use */
|
|
|
|
size_t qemu_ram_pagesize_largest(void)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
size_t largest = 0;
|
|
|
|
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2017-02-24 19:28:34 +01:00
|
|
|
largest = MAX(largest, qemu_ram_pagesize(block));
|
|
|
|
}
|
|
|
|
|
|
|
|
return largest;
|
|
|
|
}
|
|
|
|
|
2012-09-05 21:50:16 +02:00
|
|
|
static int memory_try_enable_merging(void *addr, size_t len)
|
|
|
|
{
|
2015-02-04 16:43:55 +01:00
|
|
|
if (!machine_mem_merge(current_machine)) {
|
2012-09-05 21:50:16 +02:00
|
|
|
/* disabled by the user */
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return qemu_madvise(addr, len, QEMU_MADV_MERGEABLE);
|
|
|
|
}
|
|
|
|
|
2014-11-12 13:27:41 +01:00
|
|
|
/* Only legal before guest might have detected the memory size: e.g. on
|
|
|
|
* incoming migration, or right after reset.
|
|
|
|
*
|
|
|
|
* As memory core doesn't know how is memory accessed, it is up to
|
|
|
|
* resize callback to update device state and/or add assertions to detect
|
|
|
|
* misuse, if necessary.
|
|
|
|
*/
|
2016-05-10 04:04:59 +02:00
|
|
|
int qemu_ram_resize(RAMBlock *block, ram_addr_t newsize, Error **errp)
|
2014-11-12 13:27:41 +01:00
|
|
|
{
|
|
|
|
assert(block);
|
|
|
|
|
2015-11-05 19:11:16 +01:00
|
|
|
newsize = HOST_PAGE_ALIGN(newsize);
|
2015-02-17 10:15:30 +01:00
|
|
|
|
2014-11-12 13:27:41 +01:00
|
|
|
if (block->used_length == newsize) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!(block->flags & RAM_RESIZEABLE)) {
|
|
|
|
error_setg_errno(errp, EINVAL,
|
|
|
|
"Length mismatch: %s: 0x" RAM_ADDR_FMT
|
|
|
|
" in != 0x" RAM_ADDR_FMT, block->idstr,
|
|
|
|
newsize, block->used_length);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (block->max_length < newsize) {
|
|
|
|
error_setg_errno(errp, EINVAL,
|
|
|
|
"Length too large: %s: 0x" RAM_ADDR_FMT
|
|
|
|
" > 0x" RAM_ADDR_FMT, block->idstr,
|
|
|
|
newsize, block->max_length);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
cpu_physical_memory_clear_dirty_range(block->offset, block->used_length);
|
|
|
|
block->used_length = newsize;
|
2015-03-23 11:56:01 +01:00
|
|
|
cpu_physical_memory_set_dirty_range(block->offset, block->used_length,
|
|
|
|
DIRTY_CLIENTS_ALL);
|
2014-11-12 13:27:41 +01:00
|
|
|
memory_region_set_size(block->mr, newsize);
|
|
|
|
if (block->resized) {
|
|
|
|
block->resized(block->idstr, newsize, block->host);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2016-01-25 14:33:20 +01:00
|
|
|
/* Called with ram_list.mutex held */
|
|
|
|
static void dirty_memory_extend(ram_addr_t old_ram_size,
|
|
|
|
ram_addr_t new_ram_size)
|
|
|
|
{
|
|
|
|
ram_addr_t old_num_blocks = DIV_ROUND_UP(old_ram_size,
|
|
|
|
DIRTY_MEMORY_BLOCK_SIZE);
|
|
|
|
ram_addr_t new_num_blocks = DIV_ROUND_UP(new_ram_size,
|
|
|
|
DIRTY_MEMORY_BLOCK_SIZE);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* Only need to extend if block count increased */
|
|
|
|
if (new_num_blocks <= old_num_blocks) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < DIRTY_MEMORY_NUM; i++) {
|
|
|
|
DirtyMemoryBlocks *old_blocks;
|
|
|
|
DirtyMemoryBlocks *new_blocks;
|
|
|
|
int j;
|
|
|
|
|
|
|
|
old_blocks = atomic_rcu_read(&ram_list.dirty_memory[i]);
|
|
|
|
new_blocks = g_malloc(sizeof(*new_blocks) +
|
|
|
|
sizeof(new_blocks->blocks[0]) * new_num_blocks);
|
|
|
|
|
|
|
|
if (old_num_blocks) {
|
|
|
|
memcpy(new_blocks->blocks, old_blocks->blocks,
|
|
|
|
old_num_blocks * sizeof(old_blocks->blocks[0]));
|
|
|
|
}
|
|
|
|
|
|
|
|
for (j = old_num_blocks; j < new_num_blocks; j++) {
|
|
|
|
new_blocks->blocks[j] = bitmap_new(DIRTY_MEMORY_BLOCK_SIZE);
|
|
|
|
}
|
|
|
|
|
|
|
|
atomic_rcu_set(&ram_list.dirty_memory[i], new_blocks);
|
|
|
|
|
|
|
|
if (old_blocks) {
|
|
|
|
g_free_rcu(old_blocks, rcu);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-12-13 15:37:37 +01:00
|
|
|
static void ram_block_add(RAMBlock *new_block, Error **errp, bool shared)
|
2011-12-20 14:59:12 +01:00
|
|
|
{
|
2014-05-14 11:43:18 +02:00
|
|
|
RAMBlock *block;
|
2015-01-21 13:45:24 +01:00
|
|
|
RAMBlock *last_block = NULL;
|
2013-10-08 13:52:02 +02:00
|
|
|
ram_addr_t old_ram_size, new_ram_size;
|
2016-01-14 16:09:39 +01:00
|
|
|
Error *err = NULL;
|
2013-10-08 13:52:02 +02:00
|
|
|
|
2017-03-21 17:44:30 +01:00
|
|
|
old_ram_size = last_ram_page();
|
2011-12-20 14:59:12 +01:00
|
|
|
|
2011-08-17 09:01:33 +02:00
|
|
|
qemu_mutex_lock_ramlist();
|
2014-12-15 21:55:32 +01:00
|
|
|
new_block->offset = find_ram_offset(new_block->max_length);
|
2014-05-14 11:43:18 +02:00
|
|
|
|
|
|
|
if (!new_block->host) {
|
|
|
|
if (xen_enabled()) {
|
2014-12-15 21:55:32 +01:00
|
|
|
xen_ram_alloc(new_block->offset, new_block->max_length,
|
2016-01-14 16:09:39 +01:00
|
|
|
new_block->mr, &err);
|
|
|
|
if (err) {
|
|
|
|
error_propagate(errp, err);
|
|
|
|
qemu_mutex_unlock_ramlist();
|
2016-03-09 18:14:01 +01:00
|
|
|
return;
|
2016-01-14 16:09:39 +01:00
|
|
|
}
|
2014-05-14 11:43:18 +02:00
|
|
|
} else {
|
2014-12-15 21:55:32 +01:00
|
|
|
new_block->host = phys_mem_alloc(new_block->max_length,
|
2017-12-13 15:37:37 +01:00
|
|
|
&new_block->mr->align, shared);
|
2013-07-31 15:11:11 +02:00
|
|
|
if (!new_block->host) {
|
2014-09-09 07:27:54 +02:00
|
|
|
error_setg_errno(errp, errno,
|
|
|
|
"cannot set up guest memory '%s'",
|
|
|
|
memory_region_name(new_block->mr));
|
|
|
|
qemu_mutex_unlock_ramlist();
|
2016-03-09 18:14:01 +01:00
|
|
|
return;
|
2013-07-31 15:11:11 +02:00
|
|
|
}
|
2014-12-15 21:55:32 +01:00
|
|
|
memory_try_enable_merging(new_block->host, new_block->max_length);
|
2010-08-18 08:41:49 +02:00
|
|
|
}
|
2010-03-02 00:25:08 +01:00
|
|
|
}
|
2009-04-11 19:15:54 +02:00
|
|
|
|
2015-07-02 14:18:06 +02:00
|
|
|
new_ram_size = MAX(old_ram_size,
|
|
|
|
(new_block->offset + new_block->max_length) >> TARGET_PAGE_BITS);
|
|
|
|
if (new_ram_size > old_ram_size) {
|
2016-01-25 14:33:20 +01:00
|
|
|
dirty_memory_extend(old_ram_size, new_ram_size);
|
2015-07-02 14:18:06 +02:00
|
|
|
}
|
2015-01-21 13:45:24 +01:00
|
|
|
/* 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.
|
|
|
|
*/
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2015-01-21 13:45:24 +01:00
|
|
|
last_block = block;
|
2014-12-15 21:55:32 +01:00
|
|
|
if (block->max_length < new_block->max_length) {
|
2012-11-14 16:00:51 +01:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (block) {
|
2013-09-05 20:41:35 +02:00
|
|
|
QLIST_INSERT_BEFORE_RCU(block, new_block, next);
|
2015-01-21 13:45:24 +01:00
|
|
|
} else if (last_block) {
|
2013-09-05 20:41:35 +02:00
|
|
|
QLIST_INSERT_AFTER_RCU(last_block, new_block, next);
|
2015-01-21 13:45:24 +01:00
|
|
|
} else { /* list is empty */
|
2013-09-05 20:41:35 +02:00
|
|
|
QLIST_INSERT_HEAD_RCU(&ram_list.blocks, new_block, next);
|
2012-11-14 16:00:51 +01:00
|
|
|
}
|
2012-11-14 15:45:02 +01:00
|
|
|
ram_list.mru_block = NULL;
|
2009-04-11 19:15:54 +02:00
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
/* Write list before version */
|
|
|
|
smp_wmb();
|
2011-08-18 20:41:17 +02:00
|
|
|
ram_list.version++;
|
2011-08-17 09:01:33 +02:00
|
|
|
qemu_mutex_unlock_ramlist();
|
2011-08-18 20:41:17 +02:00
|
|
|
|
2014-12-15 21:55:32 +01:00
|
|
|
cpu_physical_memory_set_dirty_range(new_block->offset,
|
2015-03-23 11:56:01 +01:00
|
|
|
new_block->used_length,
|
|
|
|
DIRTY_CLIENTS_ALL);
|
2009-04-11 19:15:54 +02:00
|
|
|
|
2015-01-21 16:18:35 +01:00
|
|
|
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);
|
2016-09-12 08:34:56 +02:00
|
|
|
/* MADV_DONTFORK is also needed by KVM in absence of synchronous MMU */
|
2015-01-21 16:18:35 +01:00
|
|
|
qemu_madvise(new_block->host, new_block->max_length, QEMU_MADV_DONTFORK);
|
2016-12-20 17:31:36 +01:00
|
|
|
ram_block_notify_add(new_block->host, new_block->max_length);
|
2014-05-14 11:43:18 +02:00
|
|
|
}
|
2009-04-11 19:15:54 +02:00
|
|
|
}
|
2007-02-09 00:08:38 +01:00
|
|
|
|
2014-05-14 11:43:19 +02:00
|
|
|
#ifdef __linux__
|
2017-06-02 16:12:23 +02:00
|
|
|
RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr,
|
|
|
|
bool share, int fd,
|
|
|
|
Error **errp)
|
2014-05-14 11:43:18 +02:00
|
|
|
{
|
|
|
|
RAMBlock *new_block;
|
2014-09-09 07:27:54 +02:00
|
|
|
Error *local_err = NULL;
|
2017-06-02 16:12:22 +02:00
|
|
|
int64_t file_size;
|
2014-05-14 11:43:18 +02:00
|
|
|
|
|
|
|
if (xen_enabled()) {
|
2014-05-14 11:43:20 +02:00
|
|
|
error_setg(errp, "-mem-path not supported with Xen");
|
2016-03-01 07:18:18 +01:00
|
|
|
return NULL;
|
2014-05-14 11:43:18 +02:00
|
|
|
}
|
|
|
|
|
2017-06-02 16:12:21 +02:00
|
|
|
if (kvm_enabled() && !kvm_has_sync_mmu()) {
|
|
|
|
error_setg(errp,
|
|
|
|
"host lacks kvm mmu notifiers, -mem-path unsupported");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-05-14 11:43:18 +02:00
|
|
|
if (phys_mem_alloc != qemu_anon_ram_alloc) {
|
|
|
|
/*
|
|
|
|
* file_ram_alloc() needs to allocate just like
|
|
|
|
* phys_mem_alloc, but we haven't bothered to provide
|
|
|
|
* a hook there.
|
|
|
|
*/
|
2014-05-14 11:43:20 +02:00
|
|
|
error_setg(errp,
|
|
|
|
"-mem-path not supported with this accelerator");
|
2016-03-01 07:18:18 +01:00
|
|
|
return NULL;
|
2014-05-14 11:43:18 +02:00
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:16 +01:00
|
|
|
size = HOST_PAGE_ALIGN(size);
|
2017-06-02 16:12:22 +02:00
|
|
|
file_size = get_file_size(fd);
|
|
|
|
if (file_size > 0 && file_size < size) {
|
|
|
|
error_setg(errp, "backing store %s size 0x%" PRIx64
|
|
|
|
" does not match 'size' option 0x" RAM_ADDR_FMT,
|
|
|
|
mem_path, file_size, size);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-05-14 11:43:18 +02:00
|
|
|
new_block = g_malloc0(sizeof(*new_block));
|
|
|
|
new_block->mr = mr;
|
2014-12-15 21:55:32 +01:00
|
|
|
new_block->used_length = size;
|
|
|
|
new_block->max_length = size;
|
2014-06-10 13:15:24 +02:00
|
|
|
new_block->flags = share ? RAM_SHARED : 0;
|
2017-06-02 16:12:22 +02:00
|
|
|
new_block->host = file_ram_alloc(new_block, size, fd, !file_size, errp);
|
2014-05-14 11:43:20 +02:00
|
|
|
if (!new_block->host) {
|
|
|
|
g_free(new_block);
|
2016-03-01 07:18:18 +01:00
|
|
|
return NULL;
|
2014-05-14 11:43:20 +02:00
|
|
|
}
|
|
|
|
|
2017-12-13 15:37:37 +01:00
|
|
|
ram_block_add(new_block, &local_err, share);
|
2014-09-09 07:27:54 +02:00
|
|
|
if (local_err) {
|
|
|
|
g_free(new_block);
|
|
|
|
error_propagate(errp, local_err);
|
2016-03-01 07:18:18 +01:00
|
|
|
return NULL;
|
2014-09-09 07:27:54 +02:00
|
|
|
}
|
2016-03-01 07:18:18 +01:00
|
|
|
return new_block;
|
2017-06-02 16:12:23 +02:00
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr,
|
|
|
|
bool share, const char *mem_path,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
int fd;
|
|
|
|
bool created;
|
|
|
|
RAMBlock *block;
|
|
|
|
|
|
|
|
fd = file_ram_open(mem_path, memory_region_name(mr), &created, errp);
|
|
|
|
if (fd < 0) {
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
block = qemu_ram_alloc_from_fd(size, mr, share, fd, errp);
|
|
|
|
if (!block) {
|
|
|
|
if (created) {
|
|
|
|
unlink(mem_path);
|
|
|
|
}
|
|
|
|
close(fd);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return block;
|
2014-05-14 11:43:18 +02:00
|
|
|
}
|
2014-05-14 11:43:19 +02:00
|
|
|
#endif
|
2014-05-14 11:43:18 +02:00
|
|
|
|
2014-11-12 13:27:41 +01:00
|
|
|
static
|
2016-03-01 07:18:18 +01:00
|
|
|
RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
|
|
|
|
void (*resized)(const char*,
|
|
|
|
uint64_t length,
|
|
|
|
void *host),
|
2017-12-13 15:37:37 +01:00
|
|
|
void *host, bool resizeable, bool share,
|
2016-03-01 07:18:18 +01:00
|
|
|
MemoryRegion *mr, Error **errp)
|
2014-05-14 11:43:18 +02:00
|
|
|
{
|
|
|
|
RAMBlock *new_block;
|
2014-09-09 07:27:54 +02:00
|
|
|
Error *local_err = NULL;
|
2014-05-14 11:43:18 +02:00
|
|
|
|
2015-11-05 19:11:16 +01:00
|
|
|
size = HOST_PAGE_ALIGN(size);
|
|
|
|
max_size = HOST_PAGE_ALIGN(max_size);
|
2014-05-14 11:43:18 +02:00
|
|
|
new_block = g_malloc0(sizeof(*new_block));
|
|
|
|
new_block->mr = mr;
|
2014-11-12 13:27:41 +01:00
|
|
|
new_block->resized = resized;
|
2014-12-15 21:55:32 +01:00
|
|
|
new_block->used_length = size;
|
|
|
|
new_block->max_length = max_size;
|
2014-11-12 13:27:41 +01:00
|
|
|
assert(max_size >= size);
|
2014-05-14 11:43:18 +02:00
|
|
|
new_block->fd = -1;
|
2016-09-29 21:09:37 +02:00
|
|
|
new_block->page_size = getpagesize();
|
2014-05-14 11:43:18 +02:00
|
|
|
new_block->host = host;
|
|
|
|
if (host) {
|
2014-05-14 11:43:22 +02:00
|
|
|
new_block->flags |= RAM_PREALLOC;
|
2014-05-14 11:43:18 +02:00
|
|
|
}
|
2014-11-12 13:27:41 +01:00
|
|
|
if (resizeable) {
|
|
|
|
new_block->flags |= RAM_RESIZEABLE;
|
|
|
|
}
|
2017-12-13 15:37:37 +01:00
|
|
|
ram_block_add(new_block, &local_err, share);
|
2014-09-09 07:27:54 +02:00
|
|
|
if (local_err) {
|
|
|
|
g_free(new_block);
|
|
|
|
error_propagate(errp, local_err);
|
2016-03-01 07:18:18 +01:00
|
|
|
return NULL;
|
2014-09-09 07:27:54 +02:00
|
|
|
}
|
2016-03-01 07:18:18 +01:00
|
|
|
return new_block;
|
2014-05-14 11:43:18 +02:00
|
|
|
}
|
|
|
|
|
2016-03-01 07:18:18 +01:00
|
|
|
RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
|
2014-11-12 13:27:41 +01:00
|
|
|
MemoryRegion *mr, Error **errp)
|
|
|
|
{
|
2017-12-13 15:37:37 +01:00
|
|
|
return qemu_ram_alloc_internal(size, size, NULL, host, false,
|
|
|
|
false, mr, errp);
|
2014-11-12 13:27:41 +01:00
|
|
|
}
|
|
|
|
|
2017-12-13 15:37:37 +01:00
|
|
|
RAMBlock *qemu_ram_alloc(ram_addr_t size, bool share,
|
|
|
|
MemoryRegion *mr, Error **errp)
|
2010-08-18 08:41:49 +02:00
|
|
|
{
|
2017-12-13 15:37:37 +01:00
|
|
|
return qemu_ram_alloc_internal(size, size, NULL, NULL, false,
|
|
|
|
share, mr, errp);
|
2014-11-12 13:27:41 +01:00
|
|
|
}
|
|
|
|
|
2016-03-01 07:18:18 +01:00
|
|
|
RAMBlock *qemu_ram_alloc_resizeable(ram_addr_t size, ram_addr_t maxsz,
|
2014-11-12 13:27:41 +01:00
|
|
|
void (*resized)(const char*,
|
|
|
|
uint64_t length,
|
|
|
|
void *host),
|
|
|
|
MemoryRegion *mr, Error **errp)
|
|
|
|
{
|
2017-12-13 15:37:37 +01:00
|
|
|
return qemu_ram_alloc_internal(size, maxsz, resized, NULL, true,
|
|
|
|
false, mr, errp);
|
2010-08-18 08:41:49 +02:00
|
|
|
}
|
|
|
|
|
2013-09-09 17:58:40 +02:00
|
|
|
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) {
|
2015-11-06 23:11:21 +01:00
|
|
|
qemu_ram_munmap(block->host, block->max_length);
|
2013-09-09 17:58:40 +02:00
|
|
|
close(block->fd);
|
|
|
|
#endif
|
|
|
|
} else {
|
|
|
|
qemu_anon_ram_free(block->host, block->max_length);
|
|
|
|
}
|
|
|
|
g_free(block);
|
|
|
|
}
|
|
|
|
|
2016-03-01 07:18:22 +01:00
|
|
|
void qemu_ram_free(RAMBlock *block)
|
2007-02-09 00:08:38 +01:00
|
|
|
{
|
2016-03-29 13:20:51 +02:00
|
|
|
if (!block) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2016-12-20 17:31:36 +01:00
|
|
|
if (block->host) {
|
|
|
|
ram_block_notify_remove(block->host, block->max_length);
|
|
|
|
}
|
|
|
|
|
2011-08-17 09:01:33 +02:00
|
|
|
qemu_mutex_lock_ramlist();
|
2016-03-01 07:18:22 +01:00
|
|
|
QLIST_REMOVE_RCU(block, next);
|
|
|
|
ram_list.mru_block = NULL;
|
|
|
|
/* Write list before version */
|
|
|
|
smp_wmb();
|
|
|
|
ram_list.version++;
|
|
|
|
call_rcu(block, reclaim_ramblock, rcu);
|
2011-08-17 09:01:33 +02:00
|
|
|
qemu_mutex_unlock_ramlist();
|
2007-02-09 00:08:38 +01:00
|
|
|
}
|
|
|
|
|
2011-03-02 08:56:19 +01:00
|
|
|
#ifndef _WIN32
|
|
|
|
void qemu_ram_remap(ram_addr_t addr, ram_addr_t length)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
ram_addr_t offset;
|
|
|
|
int flags;
|
|
|
|
void *area, *vaddr;
|
|
|
|
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2011-03-02 08:56:19 +01:00
|
|
|
offset = addr - block->offset;
|
2014-12-15 21:55:32 +01:00
|
|
|
if (offset < block->max_length) {
|
2014-11-12 10:44:41 +01:00
|
|
|
vaddr = ramblock_ptr(block, offset);
|
2014-05-14 11:43:22 +02:00
|
|
|
if (block->flags & RAM_PREALLOC) {
|
2011-03-02 08:56:19 +01:00
|
|
|
;
|
2013-07-31 15:11:05 +02:00
|
|
|
} else if (xen_enabled()) {
|
|
|
|
abort();
|
2011-03-02 08:56:19 +01:00
|
|
|
} else {
|
|
|
|
flags = MAP_FIXED;
|
2013-07-31 15:11:07 +02:00
|
|
|
if (block->fd >= 0) {
|
2014-06-10 13:15:24 +02:00
|
|
|
flags |= (block->flags & RAM_SHARED ?
|
|
|
|
MAP_SHARED : MAP_PRIVATE);
|
2013-07-31 15:11:07 +02:00
|
|
|
area = mmap(vaddr, length, PROT_READ | PROT_WRITE,
|
|
|
|
flags, block->fd, offset);
|
2011-03-02 08:56:19 +01:00
|
|
|
} else {
|
2013-07-31 15:11:09 +02:00
|
|
|
/*
|
|
|
|
* Remap needs to match alloc. Accelerators that
|
|
|
|
* set phys_mem_alloc never remap. If they did,
|
|
|
|
* we'd need a remap hook here.
|
|
|
|
*/
|
|
|
|
assert(phys_mem_alloc == qemu_anon_ram_alloc);
|
|
|
|
|
2011-03-02 08:56:19 +01:00
|
|
|
flags |= MAP_PRIVATE | MAP_ANONYMOUS;
|
|
|
|
area = mmap(vaddr, length, PROT_READ | PROT_WRITE,
|
|
|
|
flags, -1, 0);
|
|
|
|
}
|
|
|
|
if (area != vaddr) {
|
tcg: Replace fprintf(stderr, "*\n" with error_report()
Replace a large number of the fprintf(stderr, "*\n" calls with
error_report(). The functions were renamed with these commands and then
compiler issues where manually fixed.
find ./* -type f -exec sed -i \
'N;N;N;N;N;N;N;N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N;N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N;N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
find ./* -type f -exec sed -i \
'N; {s|fprintf(stderr, "\(.*\)\\n"\(.*\));|error_report("\1"\2);|Ig}' \
{} +
Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Stefan Weil <sw@weilnetz.de>
Conversions that aren't followed by exit() dropped, because they might
be inappropriate.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20180203084315.20497-14-armbru@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
2018-02-03 09:43:14 +01:00
|
|
|
error_report("Could not remap addr: "
|
|
|
|
RAM_ADDR_FMT "@" RAM_ADDR_FMT "",
|
|
|
|
length, addr);
|
2011-03-02 08:56:19 +01:00
|
|
|
exit(1);
|
|
|
|
}
|
2012-09-05 21:50:16 +02:00
|
|
|
memory_try_enable_merging(vaddr, length);
|
2012-08-02 21:44:16 +02:00
|
|
|
qemu_ram_setup_dump(vaddr, length);
|
2011-03-02 08:56:19 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif /* !_WIN32 */
|
|
|
|
|
2013-05-06 14:36:15 +02:00
|
|
|
/* Return a host pointer to ram allocated with qemu_ram_alloc.
|
2013-09-05 20:41:35 +02:00
|
|
|
* 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.
|
2013-09-05 20:41:35 +02:00
|
|
|
*
|
2015-12-16 10:30:47 +01:00
|
|
|
* Called within RCU critical section.
|
2013-05-06 14:36:15 +02:00
|
|
|
*/
|
2016-02-22 11:02:12 +01:00
|
|
|
void *qemu_map_ram_ptr(RAMBlock *ram_block, ram_addr_t addr)
|
2013-05-06 14:36:15 +02:00
|
|
|
{
|
2016-02-20 03:35:20 +01:00
|
|
|
RAMBlock *block = ram_block;
|
|
|
|
|
|
|
|
if (block == NULL) {
|
|
|
|
block = qemu_get_ram_block(addr);
|
2016-02-22 11:02:12 +01:00
|
|
|
addr -= block->offset;
|
2016-02-20 03:35:20 +01:00
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
|
|
|
|
if (xen_enabled() && block->host == NULL) {
|
2012-11-14 15:45:02 +01:00
|
|
|
/* 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) {
|
xen/mapcache: store dma information in revmapcache entries for debugging
The Xen mapcache is able to create long term mappings, they are called
"locked" mappings. The third parameter of the xen_map_cache call
specifies if a mapping is a "locked" mapping.
>From the QEMU point of view there are two kinds of long term mappings:
[a] device memory mappings, such as option roms and video memory
[b] dma mappings, created by dma_memory_map & friends
After certain operations, ballooning a VM in particular, Xen asks QEMU
kindly to destroy all mappings. However, certainly [a] mappings are
present and cannot be removed. That's not a problem as they are not
affected by balloonning. The *real* problem is that if there are any
mappings of type [b], any outstanding dma operations could fail. This is
a known shortcoming. In other words, when Xen asks QEMU to destroy all
mappings, it is an error if any [b] mappings exist.
However today we have no way of distinguishing [a] from [b]. Because of
that, we cannot even print a decent warning.
This patch introduces a new "dma" bool field to MapCacheRev entires, to
remember if a given mapping is for dma or is a long term device memory
mapping. When xen_invalidate_map_cache is called, we print a warning if
any [b] mappings exist. We ignore [a] mappings.
Mappings created by qemu_map_ram_ptr are assumed to be [a], while
mappings created by address_space_map->qemu_ram_ptr_length are assumed
to be [b].
The goal of the patch is to make debugging and system understanding
easier.
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
2017-05-03 23:00:35 +02:00
|
|
|
return xen_map_cache(addr, 0, 0, false);
|
2012-11-14 15:45:02 +01:00
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
|
xen/mapcache: store dma information in revmapcache entries for debugging
The Xen mapcache is able to create long term mappings, they are called
"locked" mappings. The third parameter of the xen_map_cache call
specifies if a mapping is a "locked" mapping.
>From the QEMU point of view there are two kinds of long term mappings:
[a] device memory mappings, such as option roms and video memory
[b] dma mappings, created by dma_memory_map & friends
After certain operations, ballooning a VM in particular, Xen asks QEMU
kindly to destroy all mappings. However, certainly [a] mappings are
present and cannot be removed. That's not a problem as they are not
affected by balloonning. The *real* problem is that if there are any
mappings of type [b], any outstanding dma operations could fail. This is
a known shortcoming. In other words, when Xen asks QEMU to destroy all
mappings, it is an error if any [b] mappings exist.
However today we have no way of distinguishing [a] from [b]. Because of
that, we cannot even print a decent warning.
This patch introduces a new "dma" bool field to MapCacheRev entires, to
remember if a given mapping is for dma or is a long term device memory
mapping. When xen_invalidate_map_cache is called, we print a warning if
any [b] mappings exist. We ignore [a] mappings.
Mappings created by qemu_map_ram_ptr are assumed to be [a], while
mappings created by address_space_map->qemu_ram_ptr_length are assumed
to be [b].
The goal of the patch is to make debugging and system understanding
easier.
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
2017-05-03 23:00:35 +02:00
|
|
|
block->host = xen_map_cache(block->offset, block->max_length, 1, false);
|
2012-11-14 15:45:02 +01:00
|
|
|
}
|
2016-02-22 11:02:12 +01:00
|
|
|
return ramblock_ptr(block, addr);
|
2009-04-10 00:21:07 +02:00
|
|
|
}
|
|
|
|
|
2016-02-22 11:02:12 +01:00
|
|
|
/* Return a host pointer to guest's ram. Similar to qemu_map_ram_ptr
|
2013-09-05 20:41:35 +02:00
|
|
|
* but takes a size argument.
|
2013-09-05 20:41:35 +02:00
|
|
|
*
|
2015-12-16 10:31:26 +01:00
|
|
|
* Called within RCU critical section.
|
2013-09-05 20:41:35 +02:00
|
|
|
*/
|
2016-02-20 03:35:20 +01:00
|
|
|
static void *qemu_ram_ptr_length(RAMBlock *ram_block, ram_addr_t addr,
|
exec: Add lock parameter to qemu_ram_ptr_length
Commit 04bf2526ce87f21b32c9acba1c5518708c243ad0 (exec: use
qemu_ram_ptr_length to access guest ram) start using qemu_ram_ptr_length
instead of qemu_map_ram_ptr, but when used with Xen, the behavior of
both function is different. They both call xen_map_cache, but one with
"lock", meaning the mapping of guest memory is never released
implicitly, and the second one without, which means, mapping can be
release later, when needed.
In the context of address_space_{read,write}_continue, the ptr to those
mapping should not be locked because it is used immediatly and never
used again.
The lock parameter make it explicit in which context qemu_ram_ptr_length
is called.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20170726165326.10327-1-anthony.perard@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-26 18:53:26 +02:00
|
|
|
hwaddr *size, bool lock)
|
2011-05-19 19:35:45 +02:00
|
|
|
{
|
2016-02-20 03:35:20 +01:00
|
|
|
RAMBlock *block = ram_block;
|
2011-06-27 19:26:06 +02:00
|
|
|
if (*size == 0) {
|
|
|
|
return NULL;
|
|
|
|
}
|
2015-12-16 10:31:26 +01:00
|
|
|
|
2016-02-20 03:35:20 +01:00
|
|
|
if (block == NULL) {
|
|
|
|
block = qemu_get_ram_block(addr);
|
2016-02-22 11:02:12 +01:00
|
|
|
addr -= block->offset;
|
2016-02-20 03:35:20 +01:00
|
|
|
}
|
2016-02-22 11:02:12 +01:00
|
|
|
*size = MIN(*size, block->max_length - addr);
|
2015-12-16 10:31:26 +01:00
|
|
|
|
|
|
|
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 the requested area.
|
|
|
|
*/
|
|
|
|
if (block->offset == 0) {
|
exec: Add lock parameter to qemu_ram_ptr_length
Commit 04bf2526ce87f21b32c9acba1c5518708c243ad0 (exec: use
qemu_ram_ptr_length to access guest ram) start using qemu_ram_ptr_length
instead of qemu_map_ram_ptr, but when used with Xen, the behavior of
both function is different. They both call xen_map_cache, but one with
"lock", meaning the mapping of guest memory is never released
implicitly, and the second one without, which means, mapping can be
release later, when needed.
In the context of address_space_{read,write}_continue, the ptr to those
mapping should not be locked because it is used immediatly and never
used again.
The lock parameter make it explicit in which context qemu_ram_ptr_length
is called.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20170726165326.10327-1-anthony.perard@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-26 18:53:26 +02:00
|
|
|
return xen_map_cache(addr, *size, lock, lock);
|
2011-05-19 19:35:45 +02:00
|
|
|
}
|
|
|
|
|
exec: Add lock parameter to qemu_ram_ptr_length
Commit 04bf2526ce87f21b32c9acba1c5518708c243ad0 (exec: use
qemu_ram_ptr_length to access guest ram) start using qemu_ram_ptr_length
instead of qemu_map_ram_ptr, but when used with Xen, the behavior of
both function is different. They both call xen_map_cache, but one with
"lock", meaning the mapping of guest memory is never released
implicitly, and the second one without, which means, mapping can be
release later, when needed.
In the context of address_space_{read,write}_continue, the ptr to those
mapping should not be locked because it is used immediatly and never
used again.
The lock parameter make it explicit in which context qemu_ram_ptr_length
is called.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20170726165326.10327-1-anthony.perard@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-26 18:53:26 +02:00
|
|
|
block->host = xen_map_cache(block->offset, block->max_length, 1, lock);
|
2011-05-19 19:35:45 +02:00
|
|
|
}
|
2015-12-16 10:31:26 +01:00
|
|
|
|
2016-02-22 11:02:12 +01:00
|
|
|
return ramblock_ptr(block, addr);
|
2011-05-19 19:35:45 +02:00
|
|
|
}
|
|
|
|
|
2018-03-12 18:20:57 +01:00
|
|
|
/* Return the offset of a hostpointer within a ramblock */
|
|
|
|
ram_addr_t qemu_ram_block_host_offset(RAMBlock *rb, void *host)
|
|
|
|
{
|
|
|
|
ram_addr_t res = (uint8_t *)host - (uint8_t *)rb->host;
|
|
|
|
assert((uintptr_t)host >= (uintptr_t)rb->host);
|
|
|
|
assert(res < rb->max_length);
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:10:32 +01:00
|
|
|
/*
|
|
|
|
* Translates a host ptr back to a RAMBlock, a ram_addr and an offset
|
|
|
|
* in that RAMBlock.
|
|
|
|
*
|
|
|
|
* ptr: Host pointer to look up
|
|
|
|
* round_offset: If true round the result offset down to a page boundary
|
|
|
|
* *ram_addr: set to result ram_addr
|
|
|
|
* *offset: set to result offset within the RAMBlock
|
|
|
|
*
|
|
|
|
* Returns: RAMBlock (or NULL if not found)
|
2013-09-05 20:41:35 +02:00
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
2015-11-05 19:10:32 +01:00
|
|
|
RAMBlock *qemu_ram_block_from_host(void *ptr, bool round_offset,
|
|
|
|
ram_addr_t *offset)
|
2009-04-11 16:47:08 +02:00
|
|
|
{
|
2009-04-11 19:15:54 +02:00
|
|
|
RAMBlock *block;
|
|
|
|
uint8_t *host = ptr;
|
|
|
|
|
2011-06-21 22:59:09 +02:00
|
|
|
if (xen_enabled()) {
|
2016-05-26 10:07:50 +02:00
|
|
|
ram_addr_t ram_addr;
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_lock();
|
2016-05-26 10:07:50 +02:00
|
|
|
ram_addr = xen_ram_addr_from_mapcache(ptr);
|
|
|
|
block = qemu_get_ram_block(ram_addr);
|
2015-11-05 19:10:32 +01:00
|
|
|
if (block) {
|
2016-06-09 17:56:17 +02:00
|
|
|
*offset = ram_addr - block->offset;
|
2015-11-05 19:10:32 +01:00
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2015-11-05 19:10:32 +01:00
|
|
|
return block;
|
2011-05-19 19:35:46 +02:00
|
|
|
}
|
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
block = atomic_rcu_read(&ram_list.mru_block);
|
2014-12-15 21:55:32 +01:00
|
|
|
if (block && block->host && host - block->host < block->max_length) {
|
2013-05-06 14:28:39 +02:00
|
|
|
goto found;
|
|
|
|
}
|
|
|
|
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2010-08-31 17:41:25 +02:00
|
|
|
/* This case append when the block is not mapped. */
|
|
|
|
if (block->host == NULL) {
|
|
|
|
continue;
|
|
|
|
}
|
2014-12-15 21:55:32 +01:00
|
|
|
if (host - block->host < block->max_length) {
|
2013-05-06 14:28:39 +02:00
|
|
|
goto found;
|
2010-06-11 19:11:42 +02:00
|
|
|
}
|
2009-04-11 19:15:54 +02:00
|
|
|
}
|
2010-08-31 17:41:25 +02:00
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2013-05-06 14:36:15 +02:00
|
|
|
return NULL;
|
2013-05-06 14:28:39 +02:00
|
|
|
|
|
|
|
found:
|
2015-11-05 19:10:32 +01:00
|
|
|
*offset = (host - block->host);
|
|
|
|
if (round_offset) {
|
|
|
|
*offset &= TARGET_PAGE_MASK;
|
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2015-11-05 19:10:32 +01:00
|
|
|
return block;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:10:33 +01:00
|
|
|
/*
|
|
|
|
* Finds the named RAMBlock
|
|
|
|
*
|
|
|
|
* name: The name of RAMBlock to find
|
|
|
|
*
|
|
|
|
* Returns: RAMBlock (or NULL if not found)
|
|
|
|
*/
|
|
|
|
RAMBlock *qemu_ram_block_by_name(const char *name)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2015-11-05 19:10:33 +01:00
|
|
|
if (!strcmp(name, block->idstr)) {
|
|
|
|
return block;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:10:32 +01:00
|
|
|
/* Some of the softmmu routines need to translate from a host pointer
|
|
|
|
(typically a TLB entry) back to a ram offset. */
|
2016-03-25 12:55:08 +01:00
|
|
|
ram_addr_t qemu_ram_addr_from_host(void *ptr)
|
2015-11-05 19:10:32 +01:00
|
|
|
{
|
|
|
|
RAMBlock *block;
|
2016-05-26 10:07:50 +02:00
|
|
|
ram_addr_t offset;
|
2015-11-05 19:10:32 +01:00
|
|
|
|
2016-05-26 10:07:50 +02:00
|
|
|
block = qemu_ram_block_from_host(ptr, false, &offset);
|
2015-11-05 19:10:32 +01:00
|
|
|
if (!block) {
|
2016-03-25 12:55:08 +01:00
|
|
|
return RAM_ADDR_INVALID;
|
2015-11-05 19:10:32 +01:00
|
|
|
}
|
|
|
|
|
2016-03-25 12:55:08 +01:00
|
|
|
return block->offset + offset;
|
2010-10-11 20:31:19 +02:00
|
|
|
}
|
2010-06-11 19:11:42 +02:00
|
|
|
|
2017-11-20 19:08:27 +01:00
|
|
|
/* Called within RCU critical section. */
|
|
|
|
void memory_notdirty_write_prepare(NotDirtyInfo *ndi,
|
|
|
|
CPUState *cpu,
|
|
|
|
vaddr mem_vaddr,
|
|
|
|
ram_addr_t ram_addr,
|
|
|
|
unsigned size)
|
|
|
|
{
|
|
|
|
ndi->cpu = cpu;
|
|
|
|
ndi->ram_addr = ram_addr;
|
|
|
|
ndi->mem_vaddr = mem_vaddr;
|
|
|
|
ndi->size = size;
|
2017-08-05 05:46:31 +02:00
|
|
|
ndi->pages = NULL;
|
2016-10-27 17:10:16 +02:00
|
|
|
|
2017-07-03 17:50:40 +02:00
|
|
|
assert(tcg_enabled());
|
2013-10-08 12:44:04 +02:00
|
|
|
if (!cpu_physical_memory_get_dirty_flag(ram_addr, DIRTY_MEMORY_CODE)) {
|
2017-08-05 05:46:31 +02:00
|
|
|
ndi->pages = page_collection_lock(ram_addr, ram_addr + size);
|
|
|
|
tb_invalidate_phys_page_fast(ndi->pages, ram_addr, size);
|
2005-08-21 11:26:42 +02:00
|
|
|
}
|
2017-11-20 19:08:27 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Called within RCU critical section. */
|
|
|
|
void memory_notdirty_write_complete(NotDirtyInfo *ndi)
|
|
|
|
{
|
2017-08-05 05:46:31 +02:00
|
|
|
if (ndi->pages) {
|
2018-06-22 19:45:31 +02:00
|
|
|
assert(tcg_enabled());
|
2017-08-05 05:46:31 +02:00
|
|
|
page_collection_unlock(ndi->pages);
|
|
|
|
ndi->pages = NULL;
|
2017-11-20 19:08:27 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Set both VGA and migration bits for simplicity and to remove
|
|
|
|
* the notdirty callback faster.
|
|
|
|
*/
|
|
|
|
cpu_physical_memory_set_dirty_range(ndi->ram_addr, ndi->size,
|
|
|
|
DIRTY_CLIENTS_NOCODE);
|
|
|
|
/* we remove the notdirty callback only if the code has been
|
|
|
|
flushed */
|
|
|
|
if (!cpu_physical_memory_is_clean(ndi->ram_addr)) {
|
|
|
|
tlb_set_dirty(ndi->cpu, ndi->mem_vaddr);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called within RCU critical section. */
|
|
|
|
static void notdirty_mem_write(void *opaque, hwaddr ram_addr,
|
|
|
|
uint64_t val, unsigned size)
|
|
|
|
{
|
|
|
|
NotDirtyInfo ndi;
|
|
|
|
|
|
|
|
memory_notdirty_write_prepare(&ndi, current_cpu, current_cpu->mem_io_vaddr,
|
|
|
|
ram_addr, size);
|
|
|
|
|
2018-06-15 15:57:14 +02:00
|
|
|
stn_p(qemu_map_ram_ptr(NULL, ram_addr), size, val);
|
2017-11-20 19:08:27 +01:00
|
|
|
memory_notdirty_write_complete(&ndi);
|
2004-01-04 19:06:42 +01:00
|
|
|
}
|
|
|
|
|
2013-05-24 14:48:38 +02:00
|
|
|
static bool notdirty_mem_accepts(void *opaque, hwaddr addr,
|
2018-05-31 15:50:52 +02:00
|
|
|
unsigned size, bool is_write,
|
|
|
|
MemTxAttrs attrs)
|
2013-05-24 14:48:38 +02:00
|
|
|
{
|
|
|
|
return is_write;
|
|
|
|
}
|
|
|
|
|
2012-01-01 23:32:15 +01:00
|
|
|
static const MemoryRegionOps notdirty_mem_ops = {
|
|
|
|
.write = notdirty_mem_write,
|
2013-05-24 14:48:38 +02:00
|
|
|
.valid.accepts = notdirty_mem_accepts,
|
2012-01-01 23:32:15 +01:00
|
|
|
.endianness = DEVICE_NATIVE_ENDIAN,
|
2017-10-13 20:19:13 +02:00
|
|
|
.valid = {
|
|
|
|
.min_access_size = 1,
|
|
|
|
.max_access_size = 8,
|
|
|
|
.unaligned = false,
|
|
|
|
},
|
|
|
|
.impl = {
|
|
|
|
.min_access_size = 1,
|
|
|
|
.max_access_size = 8,
|
|
|
|
.unaligned = false,
|
|
|
|
},
|
2004-02-06 20:46:14 +01:00
|
|
|
};
|
|
|
|
|
2008-06-09 02:20:13 +02:00
|
|
|
/* Generate a debug exception if a watchpoint has been hit. */
|
2015-04-26 17:49:24 +02:00
|
|
|
static void check_watchpoint(int offset, int len, MemTxAttrs attrs, int flags)
|
2008-06-09 02:20:13 +02:00
|
|
|
{
|
2013-08-26 03:41:01 +02:00
|
|
|
CPUState *cpu = current_cpu;
|
2016-02-11 12:17:32 +01:00
|
|
|
CPUClass *cc = CPU_GET_CLASS(cpu);
|
2008-06-09 02:20:13 +02:00
|
|
|
target_ulong vaddr;
|
2008-11-18 21:07:32 +01:00
|
|
|
CPUWatchpoint *wp;
|
2008-06-09 02:20:13 +02:00
|
|
|
|
2017-07-03 17:50:40 +02:00
|
|
|
assert(tcg_enabled());
|
2013-08-26 18:23:18 +02:00
|
|
|
if (cpu->watchpoint_hit) {
|
2008-11-18 21:24:06 +01:00
|
|
|
/* We re-entered the check after replacing the TB. Now raise
|
|
|
|
* the debug interrupt so that is will trigger after the
|
|
|
|
* current instruction. */
|
2013-08-26 03:41:01 +02:00
|
|
|
cpu_interrupt(cpu, CPU_INTERRUPT_DEBUG);
|
2008-11-18 21:24:06 +01:00
|
|
|
return;
|
|
|
|
}
|
2013-08-26 03:41:01 +02:00
|
|
|
vaddr = (cpu->mem_io_vaddr & TARGET_PAGE_MASK) + offset;
|
2017-02-07 19:29:59 +01:00
|
|
|
vaddr = cc->adjust_watchpoint_address(cpu, vaddr, len);
|
2013-08-26 18:23:18 +02:00
|
|
|
QTAILQ_FOREACH(wp, &cpu->watchpoints, entry) {
|
2014-09-12 15:06:48 +02:00
|
|
|
if (cpu_watchpoint_address_matches(wp, vaddr, len)
|
|
|
|
&& (wp->flags & flags)) {
|
2014-09-12 15:06:48 +02:00
|
|
|
if (flags == BP_MEM_READ) {
|
|
|
|
wp->flags |= BP_WATCHPOINT_HIT_READ;
|
|
|
|
} else {
|
|
|
|
wp->flags |= BP_WATCHPOINT_HIT_WRITE;
|
|
|
|
}
|
|
|
|
wp->hitaddr = vaddr;
|
2015-04-26 17:49:24 +02:00
|
|
|
wp->hitattrs = attrs;
|
2013-08-26 18:23:18 +02:00
|
|
|
if (!cpu->watchpoint_hit) {
|
2016-02-11 12:17:32 +01:00
|
|
|
if (wp->flags & BP_CPU &&
|
|
|
|
!cc->debug_check_watchpoint(cpu, wp)) {
|
|
|
|
wp->flags &= ~BP_WATCHPOINT_HIT;
|
|
|
|
continue;
|
|
|
|
}
|
2013-08-26 18:23:18 +02:00
|
|
|
cpu->watchpoint_hit = wp;
|
2016-10-27 17:10:06 +02:00
|
|
|
|
2017-08-05 05:46:31 +02:00
|
|
|
mmap_lock();
|
2013-09-01 17:12:23 +02:00
|
|
|
tb_check_watchpoint(cpu);
|
2008-11-18 21:37:55 +01:00
|
|
|
if (wp->flags & BP_STOP_BEFORE_ACCESS) {
|
2013-08-26 08:31:06 +02:00
|
|
|
cpu->exception_index = EXCP_DEBUG;
|
2017-08-05 05:46:31 +02:00
|
|
|
mmap_unlock();
|
2013-08-27 17:52:12 +02:00
|
|
|
cpu_loop_exit(cpu);
|
2008-11-18 21:37:55 +01:00
|
|
|
} else {
|
2017-10-13 19:50:02 +02:00
|
|
|
/* Force execution of one insn next time. */
|
|
|
|
cpu->cflags_next_tb = 1 | curr_cflags();
|
2017-08-05 05:46:31 +02:00
|
|
|
mmap_unlock();
|
2016-05-17 16:18:04 +02:00
|
|
|
cpu_loop_exit_noexc(cpu);
|
2008-11-18 21:37:55 +01:00
|
|
|
}
|
2008-11-18 21:24:06 +01:00
|
|
|
}
|
2008-11-18 21:37:55 +01:00
|
|
|
} else {
|
|
|
|
wp->flags &= ~BP_WATCHPOINT_HIT;
|
2008-06-09 02:20:13 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-03-17 00:58:11 +01:00
|
|
|
/* Watchpoint access routines. Watchpoints are inserted using TLB tricks,
|
|
|
|
so these check for a hit then pass through to the normal out-of-line
|
|
|
|
phys routines. */
|
2015-04-26 17:49:24 +02:00
|
|
|
static MemTxResult watch_mem_read(void *opaque, hwaddr addr, uint64_t *pdata,
|
|
|
|
unsigned size, MemTxAttrs attrs)
|
2007-03-17 00:58:11 +01:00
|
|
|
{
|
2015-04-26 17:49:24 +02:00
|
|
|
MemTxResult res;
|
|
|
|
uint64_t data;
|
2016-01-21 15:15:06 +01:00
|
|
|
int asidx = cpu_asidx_from_attrs(current_cpu, attrs);
|
|
|
|
AddressSpace *as = current_cpu->cpu_ases[asidx].as;
|
2015-04-26 17:49:24 +02:00
|
|
|
|
|
|
|
check_watchpoint(addr & ~TARGET_PAGE_MASK, size, attrs, BP_MEM_READ);
|
2012-01-02 11:47:48 +01:00
|
|
|
switch (size) {
|
2015-04-26 17:49:24 +02:00
|
|
|
case 1:
|
2016-01-21 15:15:06 +01:00
|
|
|
data = address_space_ldub(as, addr, attrs, &res);
|
2015-04-26 17:49:24 +02:00
|
|
|
break;
|
|
|
|
case 2:
|
2016-01-21 15:15:06 +01:00
|
|
|
data = address_space_lduw(as, addr, attrs, &res);
|
2015-04-26 17:49:24 +02:00
|
|
|
break;
|
|
|
|
case 4:
|
2016-01-21 15:15:06 +01:00
|
|
|
data = address_space_ldl(as, addr, attrs, &res);
|
2015-04-26 17:49:24 +02:00
|
|
|
break;
|
2017-10-17 14:16:05 +02:00
|
|
|
case 8:
|
|
|
|
data = address_space_ldq(as, addr, attrs, &res);
|
|
|
|
break;
|
2012-01-02 11:47:48 +01:00
|
|
|
default: abort();
|
|
|
|
}
|
2015-04-26 17:49:24 +02:00
|
|
|
*pdata = data;
|
|
|
|
return res;
|
2007-03-17 00:58:11 +01:00
|
|
|
}
|
|
|
|
|
2015-04-26 17:49:24 +02:00
|
|
|
static MemTxResult watch_mem_write(void *opaque, hwaddr addr,
|
|
|
|
uint64_t val, unsigned size,
|
|
|
|
MemTxAttrs attrs)
|
2007-03-17 00:58:11 +01:00
|
|
|
{
|
2015-04-26 17:49:24 +02:00
|
|
|
MemTxResult res;
|
2016-01-21 15:15:06 +01:00
|
|
|
int asidx = cpu_asidx_from_attrs(current_cpu, attrs);
|
|
|
|
AddressSpace *as = current_cpu->cpu_ases[asidx].as;
|
2015-04-26 17:49:24 +02:00
|
|
|
|
|
|
|
check_watchpoint(addr & ~TARGET_PAGE_MASK, size, attrs, BP_MEM_WRITE);
|
2012-01-02 11:47:48 +01:00
|
|
|
switch (size) {
|
2012-01-28 21:01:40 +01:00
|
|
|
case 1:
|
2016-01-21 15:15:06 +01:00
|
|
|
address_space_stb(as, addr, val, attrs, &res);
|
2012-01-28 21:01:40 +01:00
|
|
|
break;
|
|
|
|
case 2:
|
2016-01-21 15:15:06 +01:00
|
|
|
address_space_stw(as, addr, val, attrs, &res);
|
2012-01-28 21:01:40 +01:00
|
|
|
break;
|
|
|
|
case 4:
|
2016-01-21 15:15:06 +01:00
|
|
|
address_space_stl(as, addr, val, attrs, &res);
|
2012-01-28 21:01:40 +01:00
|
|
|
break;
|
2017-10-17 14:16:05 +02:00
|
|
|
case 8:
|
|
|
|
address_space_stq(as, addr, val, attrs, &res);
|
|
|
|
break;
|
2012-01-02 11:47:48 +01:00
|
|
|
default: abort();
|
|
|
|
}
|
2015-04-26 17:49:24 +02:00
|
|
|
return res;
|
2007-03-17 00:58:11 +01:00
|
|
|
}
|
|
|
|
|
2012-01-02 11:47:48 +01:00
|
|
|
static const MemoryRegionOps watch_mem_ops = {
|
2015-04-26 17:49:24 +02:00
|
|
|
.read_with_attrs = watch_mem_read,
|
|
|
|
.write_with_attrs = watch_mem_write,
|
2012-01-02 11:47:48 +01:00
|
|
|
.endianness = DEVICE_NATIVE_ENDIAN,
|
2017-10-17 14:16:05 +02:00
|
|
|
.valid = {
|
|
|
|
.min_access_size = 1,
|
|
|
|
.max_access_size = 8,
|
|
|
|
.unaligned = false,
|
|
|
|
},
|
|
|
|
.impl = {
|
|
|
|
.min_access_size = 1,
|
|
|
|
.max_access_size = 8,
|
|
|
|
.unaligned = false,
|
|
|
|
},
|
2007-03-17 00:58:11 +01:00
|
|
|
};
|
|
|
|
|
2018-03-05 00:19:49 +01:00
|
|
|
static MemTxResult flatview_read(FlatView *fv, hwaddr addr,
|
|
|
|
MemTxAttrs attrs, uint8_t *buf, int len);
|
2017-09-21 10:50:58 +02:00
|
|
|
static MemTxResult flatview_write(FlatView *fv, hwaddr addr, MemTxAttrs attrs,
|
|
|
|
const uint8_t *buf, int len);
|
|
|
|
static bool flatview_access_valid(FlatView *fv, hwaddr addr, int len,
|
2018-05-31 15:50:52 +02:00
|
|
|
bool is_write, MemTxAttrs attrs);
|
2017-09-21 10:50:58 +02:00
|
|
|
|
2015-04-26 17:49:24 +02:00
|
|
|
static MemTxResult subpage_read(void *opaque, hwaddr addr, uint64_t *data,
|
|
|
|
unsigned len, MemTxAttrs attrs)
|
2007-05-26 19:36:03 +02:00
|
|
|
{
|
2013-05-26 21:55:37 +02:00
|
|
|
subpage_t *subpage = opaque;
|
2014-12-22 13:11:39 +01:00
|
|
|
uint8_t buf[8];
|
2015-04-26 17:49:24 +02:00
|
|
|
MemTxResult res;
|
2013-05-24 16:10:39 +02:00
|
|
|
|
2007-05-26 19:36:03 +02:00
|
|
|
#if defined(DEBUG_SUBPAGE)
|
2013-09-27 03:25:38 +02:00
|
|
|
printf("%s: subpage %p len %u addr " TARGET_FMT_plx "\n", __func__,
|
2013-05-26 21:55:37 +02:00
|
|
|
subpage, len, addr);
|
2007-05-26 19:36:03 +02:00
|
|
|
#endif
|
2017-09-21 10:50:58 +02:00
|
|
|
res = flatview_read(subpage->fv, addr + subpage->base, attrs, buf, len);
|
2015-04-26 17:49:24 +02:00
|
|
|
if (res) {
|
|
|
|
return res;
|
2015-04-26 17:49:24 +02:00
|
|
|
}
|
2018-06-15 15:57:14 +02:00
|
|
|
*data = ldn_p(buf, len);
|
|
|
|
return MEMTX_OK;
|
2007-05-26 19:36:03 +02:00
|
|
|
}
|
|
|
|
|
2015-04-26 17:49:24 +02:00
|
|
|
static MemTxResult subpage_write(void *opaque, hwaddr addr,
|
|
|
|
uint64_t value, unsigned len, MemTxAttrs attrs)
|
2007-05-26 19:36:03 +02:00
|
|
|
{
|
2013-05-26 21:55:37 +02:00
|
|
|
subpage_t *subpage = opaque;
|
2014-12-22 13:11:39 +01:00
|
|
|
uint8_t buf[8];
|
2013-05-26 21:55:37 +02:00
|
|
|
|
2007-05-26 19:36:03 +02:00
|
|
|
#if defined(DEBUG_SUBPAGE)
|
2013-09-27 03:25:38 +02:00
|
|
|
printf("%s: subpage %p len %u addr " TARGET_FMT_plx
|
2013-05-26 21:55:37 +02:00
|
|
|
" value %"PRIx64"\n",
|
|
|
|
__func__, subpage, len, addr, value);
|
2007-05-26 19:36:03 +02:00
|
|
|
#endif
|
2018-06-15 15:57:14 +02:00
|
|
|
stn_p(buf, len, value);
|
2017-09-21 10:50:58 +02:00
|
|
|
return flatview_write(subpage->fv, addr + subpage->base, attrs, buf, len);
|
2007-05-26 19:36:03 +02:00
|
|
|
}
|
|
|
|
|
2013-05-24 14:02:39 +02:00
|
|
|
static bool subpage_accepts(void *opaque, hwaddr addr,
|
2018-05-31 15:50:52 +02:00
|
|
|
unsigned len, bool is_write,
|
|
|
|
MemTxAttrs attrs)
|
2013-05-24 14:02:39 +02:00
|
|
|
{
|
2013-05-26 21:55:37 +02:00
|
|
|
subpage_t *subpage = opaque;
|
2013-05-24 14:02:39 +02:00
|
|
|
#if defined(DEBUG_SUBPAGE)
|
2013-09-27 03:25:38 +02:00
|
|
|
printf("%s: subpage %p %c len %u addr " TARGET_FMT_plx "\n",
|
2013-05-26 21:55:37 +02:00
|
|
|
__func__, subpage, is_write ? 'w' : 'r', len, addr);
|
2013-05-24 14:02:39 +02:00
|
|
|
#endif
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
return flatview_access_valid(subpage->fv, addr + subpage->base,
|
2018-05-31 15:50:52 +02:00
|
|
|
len, is_write, attrs);
|
2013-05-24 14:02:39 +02:00
|
|
|
}
|
|
|
|
|
2012-01-02 11:32:48 +01:00
|
|
|
static const MemoryRegionOps subpage_ops = {
|
2015-04-26 17:49:24 +02:00
|
|
|
.read_with_attrs = subpage_read,
|
|
|
|
.write_with_attrs = subpage_write,
|
2014-12-22 13:11:39 +01:00
|
|
|
.impl.min_access_size = 1,
|
|
|
|
.impl.max_access_size = 8,
|
|
|
|
.valid.min_access_size = 1,
|
|
|
|
.valid.max_access_size = 8,
|
2013-05-24 14:02:39 +02:00
|
|
|
.valid.accepts = subpage_accepts,
|
2012-01-02 11:32:48 +01:00
|
|
|
.endianness = DEVICE_NATIVE_ENDIAN,
|
2007-05-26 19:36:03 +02:00
|
|
|
};
|
|
|
|
|
2009-10-01 23:12:16 +02:00
|
|
|
static int subpage_register (subpage_t *mmio, uint32_t start, uint32_t end,
|
2012-02-12 17:32:55 +01:00
|
|
|
uint16_t section)
|
2007-05-26 19:36:03 +02:00
|
|
|
{
|
|
|
|
int idx, eidx;
|
|
|
|
|
|
|
|
if (start >= TARGET_PAGE_SIZE || end >= TARGET_PAGE_SIZE)
|
|
|
|
return -1;
|
|
|
|
idx = SUBPAGE_IDX(start);
|
|
|
|
eidx = SUBPAGE_IDX(end);
|
|
|
|
#if defined(DEBUG_SUBPAGE)
|
2013-09-27 03:25:38 +02:00
|
|
|
printf("%s: %p start %08x end %08x idx %08x eidx %08x section %d\n",
|
|
|
|
__func__, mmio, start, end, idx, eidx, section);
|
2007-05-26 19:36:03 +02:00
|
|
|
#endif
|
|
|
|
for (; idx <= eidx; idx++) {
|
2012-02-12 17:32:55 +01:00
|
|
|
mmio->sub_section[idx] = section;
|
2007-05-26 19:36:03 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
static subpage_t *subpage_init(FlatView *fv, hwaddr base)
|
2007-05-26 19:36:03 +02:00
|
|
|
{
|
2009-10-01 23:12:16 +02:00
|
|
|
subpage_t *mmio;
|
2007-05-26 19:36:03 +02:00
|
|
|
|
2016-10-24 17:26:49 +02:00
|
|
|
mmio = g_malloc0(sizeof(subpage_t) + TARGET_PAGE_SIZE * sizeof(uint16_t));
|
2017-09-21 10:50:58 +02:00
|
|
|
mmio->fv = fv;
|
2009-02-05 23:06:18 +01:00
|
|
|
mmio->base = base;
|
2013-06-06 11:41:28 +02:00
|
|
|
memory_region_init_io(&mmio->iomem, NULL, &subpage_ops, mmio,
|
2014-06-06 08:15:52 +02:00
|
|
|
NULL, TARGET_PAGE_SIZE);
|
2012-01-02 12:20:11 +01:00
|
|
|
mmio->iomem.subpage = true;
|
2007-05-26 19:36:03 +02:00
|
|
|
#if defined(DEBUG_SUBPAGE)
|
2013-09-27 03:25:38 +02:00
|
|
|
printf("%s: %p base " TARGET_FMT_plx " len %08x\n", __func__,
|
|
|
|
mmio, base, TARGET_PAGE_SIZE);
|
2007-05-26 19:36:03 +02:00
|
|
|
#endif
|
2013-05-29 11:09:17 +02:00
|
|
|
subpage_register(mmio, 0, TARGET_PAGE_SIZE-1, PHYS_SECTION_UNASSIGNED);
|
2007-05-26 19:36:03 +02:00
|
|
|
|
|
|
|
return mmio;
|
|
|
|
}
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
static uint16_t dummy_section(PhysPageMap *map, FlatView *fv, MemoryRegion *mr)
|
2012-02-12 17:32:55 +01:00
|
|
|
{
|
2017-09-21 10:50:58 +02:00
|
|
|
assert(fv);
|
2012-02-12 17:32:55 +01:00
|
|
|
MemoryRegionSection section = {
|
2017-09-21 10:50:58 +02:00
|
|
|
.fv = fv,
|
2012-02-12 17:32:55 +01:00
|
|
|
.mr = mr,
|
|
|
|
.offset_within_address_space = 0,
|
|
|
|
.offset_within_region = 0,
|
2013-05-27 10:08:27 +02:00
|
|
|
.size = int128_2_64(),
|
2012-02-12 17:32:55 +01:00
|
|
|
};
|
|
|
|
|
2013-12-01 13:02:23 +01:00
|
|
|
return phys_section_add(map, §ion);
|
2012-02-12 17:32:55 +01:00
|
|
|
}
|
|
|
|
|
2017-12-13 18:52:28 +01:00
|
|
|
static void readonly_mem_write(void *opaque, hwaddr addr,
|
|
|
|
uint64_t val, unsigned size)
|
|
|
|
{
|
|
|
|
/* Ignore any write to ROM. */
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool readonly_mem_accepts(void *opaque, hwaddr addr,
|
2018-05-31 15:50:52 +02:00
|
|
|
unsigned size, bool is_write,
|
|
|
|
MemTxAttrs attrs)
|
2017-12-13 18:52:28 +01:00
|
|
|
{
|
|
|
|
return is_write;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* This will only be used for writes, because reads are special cased
|
|
|
|
* to directly access the underlying host ram.
|
|
|
|
*/
|
|
|
|
static const MemoryRegionOps readonly_mem_ops = {
|
|
|
|
.write = readonly_mem_write,
|
|
|
|
.valid.accepts = readonly_mem_accepts,
|
|
|
|
.endianness = DEVICE_NATIVE_ENDIAN,
|
|
|
|
.valid = {
|
|
|
|
.min_access_size = 1,
|
|
|
|
.max_access_size = 8,
|
|
|
|
.unaligned = false,
|
|
|
|
},
|
|
|
|
.impl = {
|
|
|
|
.min_access_size = 1,
|
|
|
|
.max_access_size = 8,
|
|
|
|
.unaligned = false,
|
|
|
|
},
|
|
|
|
};
|
|
|
|
|
2018-06-15 15:57:14 +02:00
|
|
|
MemoryRegionSection *iotlb_to_section(CPUState *cpu,
|
|
|
|
hwaddr index, MemTxAttrs attrs)
|
2012-03-08 16:06:55 +01:00
|
|
|
{
|
2016-01-21 15:15:05 +01:00
|
|
|
int asidx = cpu_asidx_from_attrs(cpu, attrs);
|
|
|
|
CPUAddressSpace *cpuas = &cpu->cpu_ases[asidx];
|
2015-10-01 16:29:50 +02:00
|
|
|
AddressSpaceDispatch *d = atomic_rcu_read(&cpuas->memory_dispatch);
|
2015-01-21 12:09:14 +01:00
|
|
|
MemoryRegionSection *sections = d->map.sections;
|
2013-08-16 08:26:30 +02:00
|
|
|
|
2018-06-15 15:57:14 +02:00
|
|
|
return §ions[index & ~TARGET_PAGE_MASK];
|
2012-03-08 16:06:55 +01:00
|
|
|
}
|
|
|
|
|
2009-06-14 10:38:52 +02:00
|
|
|
static void io_mem_init(void)
|
|
|
|
{
|
2017-12-13 18:52:28 +01:00
|
|
|
memory_region_init_io(&io_mem_rom, NULL, &readonly_mem_ops,
|
|
|
|
NULL, NULL, UINT64_MAX);
|
2013-06-06 11:41:28 +02:00
|
|
|
memory_region_init_io(&io_mem_unassigned, NULL, &unassigned_mem_ops, NULL,
|
2014-06-13 10:48:06 +02:00
|
|
|
NULL, UINT64_MAX);
|
tcg: drop global lock during TCG code execution
This finally allows TCG to benefit from the iothread introduction: Drop
the global mutex while running pure TCG CPU code. Reacquire the lock
when entering MMIO or PIO emulation, or when leaving the TCG loop.
We have to revert a few optimization for the current TCG threading
model, namely kicking the TCG thread in qemu_mutex_lock_iothread and not
kicking it in qemu_cpu_kick. We also need to disable RAM block
reordering until we have a more efficient locking mechanism at hand.
Still, a Linux x86 UP guest and my Musicpal ARM model boot fine here.
These numbers demonstrate where we gain something:
20338 jan 20 0 331m 75m 6904 R 99 0.9 0:50.95 qemu-system-arm
20337 jan 20 0 331m 75m 6904 S 20 0.9 0:26.50 qemu-system-arm
The guest CPU was fully loaded, but the iothread could still run mostly
independent on a second core. Without the patch we don't get beyond
32206 jan 20 0 330m 73m 7036 R 82 0.9 1:06.00 qemu-system-arm
32204 jan 20 0 330m 73m 7036 S 21 0.9 0:17.03 qemu-system-arm
We don't benefit significantly, though, when the guest is not fully
loading a host CPU.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <1439220437-23957-10-git-send-email-fred.konrad@greensocs.com>
[FK: Rebase, fix qemu_devices_reset deadlock, rm address_space_* mutex]
Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
[EGC: fixed iothread lock for cpu-exec IRQ handling]
Signed-off-by: Emilio G. Cota <cota@braap.org>
[AJB: -smp single-threaded fix, clean commit msg, BQL fixes]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Pranith Kumar <bobby.prani@gmail.com>
[PM: target-arm changes]
Acked-by: Peter Maydell <peter.maydell@linaro.org>
2017-02-23 19:29:11 +01:00
|
|
|
|
|
|
|
/* io_mem_notdirty calls tb_invalidate_phys_page_fast,
|
|
|
|
* which can be called without the iothread mutex.
|
|
|
|
*/
|
2013-06-06 11:41:28 +02:00
|
|
|
memory_region_init_io(&io_mem_notdirty, NULL, ¬dirty_mem_ops, NULL,
|
2014-06-13 10:48:06 +02:00
|
|
|
NULL, UINT64_MAX);
|
tcg: drop global lock during TCG code execution
This finally allows TCG to benefit from the iothread introduction: Drop
the global mutex while running pure TCG CPU code. Reacquire the lock
when entering MMIO or PIO emulation, or when leaving the TCG loop.
We have to revert a few optimization for the current TCG threading
model, namely kicking the TCG thread in qemu_mutex_lock_iothread and not
kicking it in qemu_cpu_kick. We also need to disable RAM block
reordering until we have a more efficient locking mechanism at hand.
Still, a Linux x86 UP guest and my Musicpal ARM model boot fine here.
These numbers demonstrate where we gain something:
20338 jan 20 0 331m 75m 6904 R 99 0.9 0:50.95 qemu-system-arm
20337 jan 20 0 331m 75m 6904 S 20 0.9 0:26.50 qemu-system-arm
The guest CPU was fully loaded, but the iothread could still run mostly
independent on a second core. Without the patch we don't get beyond
32206 jan 20 0 330m 73m 7036 R 82 0.9 1:06.00 qemu-system-arm
32204 jan 20 0 330m 73m 7036 S 21 0.9 0:17.03 qemu-system-arm
We don't benefit significantly, though, when the guest is not fully
loading a host CPU.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <1439220437-23957-10-git-send-email-fred.konrad@greensocs.com>
[FK: Rebase, fix qemu_devices_reset deadlock, rm address_space_* mutex]
Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
[EGC: fixed iothread lock for cpu-exec IRQ handling]
Signed-off-by: Emilio G. Cota <cota@braap.org>
[AJB: -smp single-threaded fix, clean commit msg, BQL fixes]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Pranith Kumar <bobby.prani@gmail.com>
[PM: target-arm changes]
Acked-by: Peter Maydell <peter.maydell@linaro.org>
2017-02-23 19:29:11 +01:00
|
|
|
memory_region_clear_global_locking(&io_mem_notdirty);
|
|
|
|
|
2013-06-06 11:41:28 +02:00
|
|
|
memory_region_init_io(&io_mem_watch, NULL, &watch_mem_ops, NULL,
|
2014-06-13 10:48:06 +02:00
|
|
|
NULL, UINT64_MAX);
|
2009-06-14 10:38:52 +02:00
|
|
|
}
|
|
|
|
|
2017-09-21 10:51:00 +02:00
|
|
|
AddressSpaceDispatch *address_space_dispatch_new(FlatView *fv)
|
2013-05-29 12:13:54 +02:00
|
|
|
{
|
2013-12-01 13:02:23 +01:00
|
|
|
AddressSpaceDispatch *d = g_new0(AddressSpaceDispatch, 1);
|
|
|
|
uint16_t n;
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
n = dummy_section(&d->map, fv, &io_mem_unassigned);
|
2013-12-01 13:02:23 +01:00
|
|
|
assert(n == PHYS_SECTION_UNASSIGNED);
|
2017-09-21 10:50:58 +02:00
|
|
|
n = dummy_section(&d->map, fv, &io_mem_notdirty);
|
2013-12-01 13:02:23 +01:00
|
|
|
assert(n == PHYS_SECTION_NOTDIRTY);
|
2017-09-21 10:50:58 +02:00
|
|
|
n = dummy_section(&d->map, fv, &io_mem_rom);
|
2013-12-01 13:02:23 +01:00
|
|
|
assert(n == PHYS_SECTION_ROM);
|
2017-09-21 10:50:58 +02:00
|
|
|
n = dummy_section(&d->map, fv, &io_mem_watch);
|
2013-12-01 13:02:23 +01:00
|
|
|
assert(n == PHYS_SECTION_WATCH);
|
2013-05-29 12:13:54 +02:00
|
|
|
|
2013-11-11 13:42:43 +01:00
|
|
|
d->phys_map = (PhysPageEntry) { .ptr = PHYS_MAP_NODE_NIL, .skip = 1 };
|
2017-09-21 10:50:56 +02:00
|
|
|
|
|
|
|
return d;
|
2013-05-29 12:13:54 +02:00
|
|
|
}
|
|
|
|
|
2017-09-21 10:50:56 +02:00
|
|
|
void address_space_dispatch_free(AddressSpaceDispatch *d)
|
2015-01-21 12:09:14 +01:00
|
|
|
{
|
|
|
|
phys_sections_free(&d->map);
|
|
|
|
g_free(d);
|
|
|
|
}
|
|
|
|
|
2012-10-02 18:54:45 +02:00
|
|
|
static void tcg_commit(MemoryListener *listener)
|
2012-02-08 20:36:02 +01:00
|
|
|
{
|
2015-10-01 16:29:50 +02:00
|
|
|
CPUAddressSpace *cpuas;
|
|
|
|
AddressSpaceDispatch *d;
|
2012-02-12 20:23:17 +01:00
|
|
|
|
2018-06-22 19:45:31 +02:00
|
|
|
assert(tcg_enabled());
|
2012-02-12 20:23:17 +01:00
|
|
|
/* since each CPU stores ram addresses in its TLB cache, we must
|
|
|
|
reset the modified entries */
|
2015-10-01 16:29:50 +02:00
|
|
|
cpuas = container_of(listener, CPUAddressSpace, tcg_as_listener);
|
|
|
|
cpu_reloading_memory_map();
|
|
|
|
/* The CPU and TLB are protected by the iothread lock.
|
|
|
|
* We reload the dispatch pointer now because cpu_reloading_memory_map()
|
|
|
|
* may have split the RCU critical section.
|
|
|
|
*/
|
2017-09-21 10:50:56 +02:00
|
|
|
d = address_space_to_dispatch(cpuas->as);
|
2016-10-21 17:34:18 +02:00
|
|
|
atomic_rcu_set(&cpuas->memory_dispatch, d);
|
2016-11-14 15:17:28 +01:00
|
|
|
tlb_flush(cpuas->cpu);
|
2012-02-08 20:36:02 +01:00
|
|
|
}
|
|
|
|
|
2011-07-26 13:26:14 +02:00
|
|
|
static void memory_map_init(void)
|
|
|
|
{
|
2011-08-21 05:09:37 +02:00
|
|
|
system_memory = g_malloc(sizeof(*system_memory));
|
2013-11-07 17:14:36 +01:00
|
|
|
|
2013-11-07 17:14:37 +01:00
|
|
|
memory_region_init(system_memory, NULL, "system", UINT64_MAX);
|
2013-04-29 18:25:51 +02:00
|
|
|
address_space_init(&address_space_memory, system_memory, "memory");
|
2011-08-08 15:09:03 +02:00
|
|
|
|
2011-08-21 05:09:37 +02:00
|
|
|
system_io = g_malloc(sizeof(*system_io));
|
2013-09-02 18:43:30 +02:00
|
|
|
memory_region_init_io(system_io, NULL, &unassigned_io_ops, NULL, "io",
|
|
|
|
65536);
|
2013-04-29 18:25:51 +02:00
|
|
|
address_space_init(&address_space_io, system_io, "I/O");
|
2011-07-26 13:26:14 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
MemoryRegion *get_system_memory(void)
|
|
|
|
{
|
|
|
|
return system_memory;
|
|
|
|
}
|
|
|
|
|
2011-08-08 15:09:03 +02:00
|
|
|
MemoryRegion *get_system_io(void)
|
|
|
|
{
|
|
|
|
return system_io;
|
|
|
|
}
|
|
|
|
|
2008-06-08 03:09:01 +02:00
|
|
|
#endif /* !defined(CONFIG_USER_ONLY) */
|
|
|
|
|
2004-01-24 16:23:36 +01:00
|
|
|
/* physical memory access (slow version, mainly for debug) */
|
|
|
|
#if defined(CONFIG_USER_ONLY)
|
2013-06-29 19:40:58 +02:00
|
|
|
int cpu_memory_rw_debug(CPUState *cpu, target_ulong addr,
|
2010-03-01 01:08:59 +01:00
|
|
|
uint8_t *buf, int len, int is_write)
|
2004-01-24 16:23:36 +01:00
|
|
|
{
|
|
|
|
int l, flags;
|
|
|
|
target_ulong page;
|
2006-03-25 20:31:22 +01:00
|
|
|
void * p;
|
2004-01-24 16:23:36 +01:00
|
|
|
|
|
|
|
while (len > 0) {
|
|
|
|
page = addr & TARGET_PAGE_MASK;
|
|
|
|
l = (page + TARGET_PAGE_SIZE) - addr;
|
|
|
|
if (l > len)
|
|
|
|
l = len;
|
|
|
|
flags = page_get_flags(page);
|
|
|
|
if (!(flags & PAGE_VALID))
|
2010-03-01 01:08:59 +01:00
|
|
|
return -1;
|
2004-01-24 16:23:36 +01:00
|
|
|
if (is_write) {
|
|
|
|
if (!(flags & PAGE_WRITE))
|
2010-03-01 01:08:59 +01:00
|
|
|
return -1;
|
2007-11-11 15:26:47 +01:00
|
|
|
/* XXX: this code should not depend on lock_user */
|
2008-04-28 01:53:45 +02:00
|
|
|
if (!(p = lock_user(VERIFY_WRITE, addr, l, 0)))
|
2010-03-01 01:08:59 +01:00
|
|
|
return -1;
|
2008-04-28 01:53:45 +02:00
|
|
|
memcpy(p, buf, l);
|
|
|
|
unlock_user(p, addr, l);
|
2004-01-24 16:23:36 +01:00
|
|
|
} else {
|
|
|
|
if (!(flags & PAGE_READ))
|
2010-03-01 01:08:59 +01:00
|
|
|
return -1;
|
2007-11-11 15:26:47 +01:00
|
|
|
/* XXX: this code should not depend on lock_user */
|
2008-04-28 01:53:45 +02:00
|
|
|
if (!(p = lock_user(VERIFY_READ, addr, l, 1)))
|
2010-03-01 01:08:59 +01:00
|
|
|
return -1;
|
2008-04-28 01:53:45 +02:00
|
|
|
memcpy(buf, p, l);
|
2008-04-28 10:54:59 +02:00
|
|
|
unlock_user(p, addr, 0);
|
2004-01-24 16:23:36 +01:00
|
|
|
}
|
|
|
|
len -= l;
|
|
|
|
buf += l;
|
|
|
|
addr += l;
|
|
|
|
}
|
2010-03-01 01:08:59 +01:00
|
|
|
return 0;
|
2004-01-24 16:23:36 +01:00
|
|
|
}
|
2005-01-28 23:37:22 +01:00
|
|
|
|
2004-01-24 16:23:36 +01:00
|
|
|
#else
|
2012-10-03 15:49:05 +02:00
|
|
|
|
2015-03-23 11:45:53 +01:00
|
|
|
static void invalidate_and_set_dirty(MemoryRegion *mr, hwaddr addr,
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr length)
|
2012-10-03 15:49:05 +02:00
|
|
|
{
|
2015-03-25 15:21:39 +01:00
|
|
|
uint8_t dirty_log_mask = memory_region_get_dirty_log_mask(mr);
|
2016-02-22 11:02:12 +01:00
|
|
|
addr += memory_region_get_ram_addr(mr);
|
|
|
|
|
2015-03-25 15:21:39 +01:00
|
|
|
/* No early return if dirty_log_mask is or becomes 0, because
|
|
|
|
* cpu_physical_memory_set_dirty_range will still call
|
|
|
|
* xen_modified_memory.
|
|
|
|
*/
|
|
|
|
if (dirty_log_mask) {
|
|
|
|
dirty_log_mask =
|
|
|
|
cpu_physical_memory_range_includes_clean(addr, length, dirty_log_mask);
|
|
|
|
}
|
|
|
|
if (dirty_log_mask & (1 << DIRTY_MEMORY_CODE)) {
|
2017-07-03 17:50:40 +02:00
|
|
|
assert(tcg_enabled());
|
2017-08-05 05:46:31 +02:00
|
|
|
mmap_lock();
|
2015-03-25 15:21:39 +01:00
|
|
|
tb_invalidate_phys_range(addr, addr + length);
|
2017-08-05 05:46:31 +02:00
|
|
|
mmap_unlock();
|
2015-03-25 15:21:39 +01:00
|
|
|
dirty_log_mask &= ~(1 << DIRTY_MEMORY_CODE);
|
2012-10-03 15:49:05 +02:00
|
|
|
}
|
2015-03-25 15:21:39 +01:00
|
|
|
cpu_physical_memory_set_dirty_range(addr, length, dirty_log_mask);
|
2012-10-03 15:49:05 +02:00
|
|
|
}
|
|
|
|
|
2013-07-08 23:55:59 +02:00
|
|
|
static int memory_access_size(MemoryRegion *mr, unsigned l, hwaddr addr)
|
2013-05-24 11:59:43 +02:00
|
|
|
{
|
2013-07-17 13:17:41 +02:00
|
|
|
unsigned access_size_max = mr->ops->valid.max_access_size;
|
2013-07-08 23:55:59 +02:00
|
|
|
|
|
|
|
/* Regions are assumed to support 1-4 byte accesses unless
|
|
|
|
otherwise specified. */
|
|
|
|
if (access_size_max == 0) {
|
|
|
|
access_size_max = 4;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Bound the maximum access by the alignment of the address. */
|
|
|
|
if (!mr->ops->impl.unaligned) {
|
|
|
|
unsigned align_size_max = addr & -addr;
|
|
|
|
if (align_size_max != 0 && align_size_max < access_size_max) {
|
|
|
|
access_size_max = align_size_max;
|
|
|
|
}
|
2013-05-24 11:59:43 +02:00
|
|
|
}
|
2013-07-08 23:55:59 +02:00
|
|
|
|
|
|
|
/* Don't attempt accesses larger than the maximum. */
|
|
|
|
if (l > access_size_max) {
|
|
|
|
l = access_size_max;
|
2013-05-24 11:59:43 +02:00
|
|
|
}
|
2015-07-24 14:33:10 +02:00
|
|
|
l = pow2floor(l);
|
2013-07-08 23:55:59 +02:00
|
|
|
|
|
|
|
return l;
|
2013-05-24 11:59:43 +02:00
|
|
|
}
|
|
|
|
|
2015-06-18 18:47:22 +02:00
|
|
|
static bool prepare_mmio_access(MemoryRegion *mr)
|
2015-06-18 18:47:21 +02:00
|
|
|
{
|
2015-06-18 18:47:22 +02:00
|
|
|
bool unlocked = !qemu_mutex_iothread_locked();
|
|
|
|
bool release_lock = false;
|
|
|
|
|
|
|
|
if (unlocked && mr->global_locking) {
|
|
|
|
qemu_mutex_lock_iothread();
|
|
|
|
unlocked = false;
|
|
|
|
release_lock = true;
|
|
|
|
}
|
2015-06-18 18:47:21 +02:00
|
|
|
if (mr->flush_coalesced_mmio) {
|
2015-06-18 18:47:22 +02:00
|
|
|
if (unlocked) {
|
|
|
|
qemu_mutex_lock_iothread();
|
|
|
|
}
|
2015-06-18 18:47:21 +02:00
|
|
|
qemu_flush_coalesced_mmio_buffer();
|
2015-06-18 18:47:22 +02:00
|
|
|
if (unlocked) {
|
|
|
|
qemu_mutex_unlock_iothread();
|
|
|
|
}
|
2015-06-18 18:47:21 +02:00
|
|
|
}
|
2015-06-18 18:47:22 +02:00
|
|
|
|
|
|
|
return release_lock;
|
2015-06-18 18:47:21 +02:00
|
|
|
}
|
|
|
|
|
2015-12-09 10:18:57 +01:00
|
|
|
/* Called within RCU critical section. */
|
2017-09-21 10:50:58 +02:00
|
|
|
static MemTxResult flatview_write_continue(FlatView *fv, hwaddr addr,
|
|
|
|
MemTxAttrs attrs,
|
|
|
|
const uint8_t *buf,
|
|
|
|
int len, hwaddr addr1,
|
|
|
|
hwaddr l, MemoryRegion *mr)
|
2004-01-24 16:23:36 +01:00
|
|
|
{
|
|
|
|
uint8_t *ptr;
|
2013-05-24 16:10:39 +02:00
|
|
|
uint64_t val;
|
2015-04-26 17:49:23 +02:00
|
|
|
MemTxResult result = MEMTX_OK;
|
2015-06-18 18:47:22 +02:00
|
|
|
bool release_lock = false;
|
2007-09-17 10:09:54 +02:00
|
|
|
|
2015-12-09 10:18:57 +01:00
|
|
|
for (;;) {
|
2015-12-09 10:06:31 +01:00
|
|
|
if (!memory_access_is_direct(mr, true)) {
|
|
|
|
release_lock |= prepare_mmio_access(mr);
|
|
|
|
l = memory_access_size(mr, l, addr1);
|
|
|
|
/* XXX: could force current_cpu to NULL to avoid
|
|
|
|
potential bugs */
|
2018-06-15 15:57:14 +02:00
|
|
|
val = ldn_p(buf, l);
|
|
|
|
result |= memory_region_dispatch_write(mr, addr1, val, l, attrs);
|
2004-01-24 16:23:36 +01:00
|
|
|
} else {
|
2015-12-09 10:06:31 +01:00
|
|
|
/* RAM case */
|
exec: Add lock parameter to qemu_ram_ptr_length
Commit 04bf2526ce87f21b32c9acba1c5518708c243ad0 (exec: use
qemu_ram_ptr_length to access guest ram) start using qemu_ram_ptr_length
instead of qemu_map_ram_ptr, but when used with Xen, the behavior of
both function is different. They both call xen_map_cache, but one with
"lock", meaning the mapping of guest memory is never released
implicitly, and the second one without, which means, mapping can be
release later, when needed.
In the context of address_space_{read,write}_continue, the ptr to those
mapping should not be locked because it is used immediatly and never
used again.
The lock parameter make it explicit in which context qemu_ram_ptr_length
is called.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20170726165326.10327-1-anthony.perard@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-26 18:53:26 +02:00
|
|
|
ptr = qemu_ram_ptr_length(mr->ram_block, addr1, &l, false);
|
2015-12-09 10:06:31 +01:00
|
|
|
memcpy(ptr, buf, l);
|
|
|
|
invalidate_and_set_dirty(mr, addr1, l);
|
2004-01-24 16:23:36 +01:00
|
|
|
}
|
2015-06-18 18:47:22 +02:00
|
|
|
|
|
|
|
if (release_lock) {
|
|
|
|
qemu_mutex_unlock_iothread();
|
|
|
|
release_lock = false;
|
|
|
|
}
|
|
|
|
|
2004-01-24 16:23:36 +01:00
|
|
|
len -= l;
|
|
|
|
buf += l;
|
|
|
|
addr += l;
|
2015-12-09 10:18:57 +01:00
|
|
|
|
|
|
|
if (!len) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = flatview_translate(fv, addr, &addr1, &l, true, attrs);
|
2004-01-24 16:23:36 +01:00
|
|
|
}
|
2013-05-21 09:56:55 +02:00
|
|
|
|
2015-04-26 17:49:23 +02:00
|
|
|
return result;
|
2004-01-24 16:23:36 +01:00
|
|
|
}
|
2005-01-28 23:37:22 +01:00
|
|
|
|
2018-03-05 09:23:56 +01:00
|
|
|
/* Called from RCU critical section. */
|
2017-09-21 10:50:58 +02:00
|
|
|
static MemTxResult flatview_write(FlatView *fv, hwaddr addr, MemTxAttrs attrs,
|
|
|
|
const uint8_t *buf, int len)
|
2012-10-03 16:22:53 +02:00
|
|
|
{
|
2015-12-09 10:06:31 +01:00
|
|
|
hwaddr l;
|
|
|
|
hwaddr addr1;
|
|
|
|
MemoryRegion *mr;
|
|
|
|
MemTxResult result = MEMTX_OK;
|
|
|
|
|
2018-03-05 09:23:56 +01:00
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = flatview_translate(fv, addr, &addr1, &l, true, attrs);
|
2018-03-05 09:23:56 +01:00
|
|
|
result = flatview_write_continue(fv, addr, attrs, buf, len,
|
|
|
|
addr1, l, mr);
|
2015-12-09 10:18:57 +01:00
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called within RCU critical section. */
|
2017-09-21 10:50:58 +02:00
|
|
|
MemTxResult flatview_read_continue(FlatView *fv, hwaddr addr,
|
|
|
|
MemTxAttrs attrs, uint8_t *buf,
|
|
|
|
int len, hwaddr addr1, hwaddr l,
|
|
|
|
MemoryRegion *mr)
|
2015-12-09 10:18:57 +01:00
|
|
|
{
|
|
|
|
uint8_t *ptr;
|
|
|
|
uint64_t val;
|
|
|
|
MemTxResult result = MEMTX_OK;
|
|
|
|
bool release_lock = false;
|
2015-12-09 10:06:31 +01:00
|
|
|
|
2015-12-09 10:18:57 +01:00
|
|
|
for (;;) {
|
2015-12-09 10:06:31 +01:00
|
|
|
if (!memory_access_is_direct(mr, false)) {
|
|
|
|
/* I/O case */
|
|
|
|
release_lock |= prepare_mmio_access(mr);
|
|
|
|
l = memory_access_size(mr, l, addr1);
|
2018-06-15 15:57:14 +02:00
|
|
|
result |= memory_region_dispatch_read(mr, addr1, &val, l, attrs);
|
|
|
|
stn_p(buf, l, val);
|
2015-12-09 10:06:31 +01:00
|
|
|
} else {
|
|
|
|
/* RAM case */
|
exec: Add lock parameter to qemu_ram_ptr_length
Commit 04bf2526ce87f21b32c9acba1c5518708c243ad0 (exec: use
qemu_ram_ptr_length to access guest ram) start using qemu_ram_ptr_length
instead of qemu_map_ram_ptr, but when used with Xen, the behavior of
both function is different. They both call xen_map_cache, but one with
"lock", meaning the mapping of guest memory is never released
implicitly, and the second one without, which means, mapping can be
release later, when needed.
In the context of address_space_{read,write}_continue, the ptr to those
mapping should not be locked because it is used immediatly and never
used again.
The lock parameter make it explicit in which context qemu_ram_ptr_length
is called.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20170726165326.10327-1-anthony.perard@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-26 18:53:26 +02:00
|
|
|
ptr = qemu_ram_ptr_length(mr->ram_block, addr1, &l, false);
|
2015-12-09 10:06:31 +01:00
|
|
|
memcpy(buf, ptr, l);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (release_lock) {
|
|
|
|
qemu_mutex_unlock_iothread();
|
|
|
|
release_lock = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
len -= l;
|
|
|
|
buf += l;
|
|
|
|
addr += l;
|
2015-12-09 10:18:57 +01:00
|
|
|
|
|
|
|
if (!len) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = flatview_translate(fv, addr, &addr1, &l, false, attrs);
|
2015-12-09 10:18:57 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2018-03-05 00:19:49 +01:00
|
|
|
/* Called from RCU critical section. */
|
|
|
|
static MemTxResult flatview_read(FlatView *fv, hwaddr addr,
|
|
|
|
MemTxAttrs attrs, uint8_t *buf, int len)
|
2015-12-09 10:18:57 +01:00
|
|
|
{
|
|
|
|
hwaddr l;
|
|
|
|
hwaddr addr1;
|
|
|
|
MemoryRegion *mr;
|
2015-12-09 10:06:31 +01:00
|
|
|
|
2018-03-05 00:19:49 +01:00
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = flatview_translate(fv, addr, &addr1, &l, false, attrs);
|
2018-03-05 00:19:49 +01:00
|
|
|
return flatview_read_continue(fv, addr, attrs, buf, len,
|
|
|
|
addr1, l, mr);
|
2012-10-03 16:22:53 +02:00
|
|
|
}
|
|
|
|
|
2018-03-05 00:19:49 +01:00
|
|
|
MemTxResult address_space_read_full(AddressSpace *as, hwaddr addr,
|
|
|
|
MemTxAttrs attrs, uint8_t *buf, int len)
|
|
|
|
{
|
|
|
|
MemTxResult result = MEMTX_OK;
|
|
|
|
FlatView *fv;
|
|
|
|
|
|
|
|
if (len > 0) {
|
|
|
|
rcu_read_lock();
|
|
|
|
fv = address_space_to_flatview(as);
|
|
|
|
result = flatview_read(fv, addr, attrs, buf, len);
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2018-03-05 09:23:56 +01:00
|
|
|
MemTxResult address_space_write(AddressSpace *as, hwaddr addr,
|
|
|
|
MemTxAttrs attrs,
|
|
|
|
const uint8_t *buf, int len)
|
|
|
|
{
|
|
|
|
MemTxResult result = MEMTX_OK;
|
|
|
|
FlatView *fv;
|
|
|
|
|
|
|
|
if (len > 0) {
|
|
|
|
rcu_read_lock();
|
|
|
|
fv = address_space_to_flatview(as);
|
|
|
|
result = flatview_write(fv, addr, attrs, buf, len);
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2018-03-05 09:29:04 +01:00
|
|
|
MemTxResult address_space_rw(AddressSpace *as, hwaddr addr, MemTxAttrs attrs,
|
|
|
|
uint8_t *buf, int len, bool is_write)
|
|
|
|
{
|
|
|
|
if (is_write) {
|
|
|
|
return address_space_write(as, addr, attrs, buf, len);
|
|
|
|
} else {
|
|
|
|
return address_space_read_full(as, addr, attrs, buf, len);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-10-23 12:30:10 +02:00
|
|
|
void cpu_physical_memory_rw(hwaddr addr, uint8_t *buf,
|
2012-10-03 16:22:53 +02:00
|
|
|
int len, int is_write)
|
|
|
|
{
|
2015-04-26 17:49:24 +02:00
|
|
|
address_space_rw(&address_space_memory, addr, MEMTXATTRS_UNSPECIFIED,
|
|
|
|
buf, len, is_write);
|
2012-10-03 16:22:53 +02:00
|
|
|
}
|
|
|
|
|
2013-12-11 14:17:44 +01:00
|
|
|
enum write_rom_type {
|
|
|
|
WRITE_DATA,
|
|
|
|
FLUSH_CACHE,
|
|
|
|
};
|
|
|
|
|
2013-12-13 07:28:52 +01:00
|
|
|
static inline void cpu_physical_memory_write_rom_internal(AddressSpace *as,
|
2013-12-11 14:17:44 +01:00
|
|
|
hwaddr addr, const uint8_t *buf, int len, enum write_rom_type type)
|
2006-04-23 19:14:48 +02:00
|
|
|
{
|
2013-05-24 12:59:37 +02:00
|
|
|
hwaddr l;
|
2006-04-23 19:14:48 +02:00
|
|
|
uint8_t *ptr;
|
2013-05-24 12:59:37 +02:00
|
|
|
hwaddr addr1;
|
2013-05-29 12:42:00 +02:00
|
|
|
MemoryRegion *mr;
|
2007-09-17 10:09:54 +02:00
|
|
|
|
2015-03-18 14:21:43 +01:00
|
|
|
rcu_read_lock();
|
2006-04-23 19:14:48 +02:00
|
|
|
while (len > 0) {
|
2013-05-24 12:59:37 +02:00
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = address_space_translate(as, addr, &addr1, &l, true,
|
|
|
|
MEMTXATTRS_UNSPECIFIED);
|
2007-09-17 10:09:54 +02:00
|
|
|
|
2013-05-29 12:42:00 +02:00
|
|
|
if (!(memory_region_is_ram(mr) ||
|
|
|
|
memory_region_is_romd(mr))) {
|
exec: skip MMIO regions correctly in cpu_physical_memory_write_rom_internal
Loading the BIOS in the mac99 machine is interesting, because there is a
PROM in the middle of the BIOS region (from 16K to 32K). Before memory
region accesses were clamped, when QEMU was asked to load a BIOS from
0xfff00000 to 0xffffffff it would put even those 16K from the BIOS file
into the region. This is weird because those 16K were not actually
visible between 0xfff04000 and 0xfff07fff. However, it worked.
After clamping was added, this also worked. In this case, the
cpu_physical_memory_write_rom_internal function split the write in
three parts: the first 16K were copied, the PROM area (second 16K) were
ignored, then the rest was copied.
Problems then started with commit 965eb2f (exec: do not clamp accesses
to MMIO regions, 2015-06-17). Clamping accesses is not done for MMIO
regions because they can overlap wildly, and MMIO registers can be
expected to perform full-width accesses based only on their address
(with no respect for adjacent registers that could decode to completely
different MemoryRegions). However, this lack of clamping also applied
to the PROM area! cpu_physical_memory_write_rom_internal thus failed
to copy the third range above, i.e. only copied the first 16K of the BIOS.
In effect, address_space_translate is expecting _something else_ to do
the clamping for MMIO regions if the incoming length is large. This
"something else" is memory_access_size in the case of address_space_rw,
so use the same logic in cpu_physical_memory_write_rom_internal.
Reported-by: Alexander Graf <agraf@redhat.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Tested-by: Laurent Vivier <lvivier@redhat.com>
Fixes: 965eb2f
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2015-07-04 00:24:51 +02:00
|
|
|
l = memory_access_size(mr, l, addr1);
|
2006-04-23 19:14:48 +02:00
|
|
|
} else {
|
|
|
|
/* ROM/RAM case */
|
2016-02-22 11:02:12 +01:00
|
|
|
ptr = qemu_map_ram_ptr(mr->ram_block, addr1);
|
2013-12-11 14:17:44 +01:00
|
|
|
switch (type) {
|
|
|
|
case WRITE_DATA:
|
|
|
|
memcpy(ptr, buf, l);
|
2015-03-23 11:45:53 +01:00
|
|
|
invalidate_and_set_dirty(mr, addr1, l);
|
2013-12-11 14:17:44 +01:00
|
|
|
break;
|
|
|
|
case FLUSH_CACHE:
|
|
|
|
flush_icache_range((uintptr_t)ptr, (uintptr_t)ptr + l);
|
|
|
|
break;
|
|
|
|
}
|
2006-04-23 19:14:48 +02:00
|
|
|
}
|
|
|
|
len -= l;
|
|
|
|
buf += l;
|
|
|
|
addr += l;
|
|
|
|
}
|
2015-03-18 14:21:43 +01:00
|
|
|
rcu_read_unlock();
|
2006-04-23 19:14:48 +02:00
|
|
|
}
|
|
|
|
|
2013-12-11 14:17:44 +01:00
|
|
|
/* used for ROM loading : can write in RAM and ROM */
|
2013-12-13 07:28:52 +01:00
|
|
|
void cpu_physical_memory_write_rom(AddressSpace *as, hwaddr addr,
|
2013-12-11 14:17:44 +01:00
|
|
|
const uint8_t *buf, int len)
|
|
|
|
{
|
2013-12-13 07:28:52 +01:00
|
|
|
cpu_physical_memory_write_rom_internal(as, addr, buf, len, WRITE_DATA);
|
2013-12-11 14:17:44 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void cpu_flush_icache_range(hwaddr start, int len)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* This function should do the same thing as an icache flush that was
|
|
|
|
* triggered from within the guest. For TCG we are always cache coherent,
|
|
|
|
* so there is no need to flush anything. For KVM / Xen we need to flush
|
|
|
|
* the host's instruction cache at least.
|
|
|
|
*/
|
|
|
|
if (tcg_enabled()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-12-13 07:28:52 +01:00
|
|
|
cpu_physical_memory_write_rom_internal(&address_space_memory,
|
|
|
|
start, NULL, len, FLUSH_CACHE);
|
2013-12-11 14:17:44 +01:00
|
|
|
}
|
|
|
|
|
2009-01-22 17:59:11 +01:00
|
|
|
typedef struct {
|
2013-06-28 17:33:29 +02:00
|
|
|
MemoryRegion *mr;
|
2009-01-22 17:59:11 +01:00
|
|
|
void *buffer;
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr addr;
|
|
|
|
hwaddr len;
|
2015-03-16 10:03:33 +01:00
|
|
|
bool in_use;
|
2009-01-22 17:59:11 +01:00
|
|
|
} BounceBuffer;
|
|
|
|
|
|
|
|
static BounceBuffer bounce;
|
|
|
|
|
2009-01-22 17:59:16 +01:00
|
|
|
typedef struct MapClient {
|
2015-03-16 10:03:37 +01:00
|
|
|
QEMUBH *bh;
|
2009-09-12 09:36:22 +02:00
|
|
|
QLIST_ENTRY(MapClient) link;
|
2009-01-22 17:59:16 +01:00
|
|
|
} MapClient;
|
|
|
|
|
2015-03-16 10:03:35 +01:00
|
|
|
QemuMutex map_client_list_lock;
|
2009-09-12 09:36:22 +02:00
|
|
|
static QLIST_HEAD(map_client_list, MapClient) map_client_list
|
|
|
|
= QLIST_HEAD_INITIALIZER(map_client_list);
|
2009-01-22 17:59:16 +01:00
|
|
|
|
2015-03-16 10:03:37 +01:00
|
|
|
static void cpu_unregister_map_client_do(MapClient *client)
|
|
|
|
{
|
|
|
|
QLIST_REMOVE(client, link);
|
|
|
|
g_free(client);
|
|
|
|
}
|
|
|
|
|
2015-03-16 10:03:36 +01:00
|
|
|
static void cpu_notify_map_clients_locked(void)
|
|
|
|
{
|
|
|
|
MapClient *client;
|
|
|
|
|
|
|
|
while (!QLIST_EMPTY(&map_client_list)) {
|
|
|
|
client = QLIST_FIRST(&map_client_list);
|
2015-03-16 10:03:37 +01:00
|
|
|
qemu_bh_schedule(client->bh);
|
|
|
|
cpu_unregister_map_client_do(client);
|
2015-03-16 10:03:36 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-03-16 10:03:37 +01:00
|
|
|
void cpu_register_map_client(QEMUBH *bh)
|
2009-01-22 17:59:16 +01:00
|
|
|
{
|
2011-08-21 05:09:37 +02:00
|
|
|
MapClient *client = g_malloc(sizeof(*client));
|
2009-01-22 17:59:16 +01:00
|
|
|
|
2015-03-16 10:03:35 +01:00
|
|
|
qemu_mutex_lock(&map_client_list_lock);
|
2015-03-16 10:03:37 +01:00
|
|
|
client->bh = bh;
|
2009-09-12 09:36:22 +02:00
|
|
|
QLIST_INSERT_HEAD(&map_client_list, client, link);
|
2015-03-16 10:03:36 +01:00
|
|
|
if (!atomic_read(&bounce.in_use)) {
|
|
|
|
cpu_notify_map_clients_locked();
|
|
|
|
}
|
2015-03-16 10:03:35 +01:00
|
|
|
qemu_mutex_unlock(&map_client_list_lock);
|
2009-01-22 17:59:16 +01:00
|
|
|
}
|
|
|
|
|
2015-03-16 10:03:35 +01:00
|
|
|
void cpu_exec_init_all(void)
|
2009-01-22 17:59:16 +01:00
|
|
|
{
|
2015-03-16 10:03:35 +01:00
|
|
|
qemu_mutex_init(&ram_list.mutex);
|
2016-10-24 17:26:49 +02:00
|
|
|
/* The data structures we set up here depend on knowing the page size,
|
|
|
|
* so no more changes can be made after this point.
|
|
|
|
* In an ideal world, nothing we did before we had finished the
|
|
|
|
* machine setup would care about the target page size, and we could
|
|
|
|
* do this much later, rather than requiring board models to state
|
|
|
|
* up front what their requirements are.
|
|
|
|
*/
|
|
|
|
finalize_target_page_bits();
|
2015-03-16 10:03:35 +01:00
|
|
|
io_mem_init();
|
2015-11-02 09:23:52 +01:00
|
|
|
memory_map_init();
|
2015-03-16 10:03:35 +01:00
|
|
|
qemu_mutex_init(&map_client_list_lock);
|
2009-01-22 17:59:16 +01:00
|
|
|
}
|
|
|
|
|
2015-03-16 10:03:37 +01:00
|
|
|
void cpu_unregister_map_client(QEMUBH *bh)
|
2009-01-22 17:59:16 +01:00
|
|
|
{
|
|
|
|
MapClient *client;
|
|
|
|
|
2015-03-16 10:03:37 +01:00
|
|
|
qemu_mutex_lock(&map_client_list_lock);
|
|
|
|
QLIST_FOREACH(client, &map_client_list, link) {
|
|
|
|
if (client->bh == bh) {
|
|
|
|
cpu_unregister_map_client_do(client);
|
|
|
|
break;
|
|
|
|
}
|
2009-01-22 17:59:16 +01:00
|
|
|
}
|
2015-03-16 10:03:37 +01:00
|
|
|
qemu_mutex_unlock(&map_client_list_lock);
|
2009-01-22 17:59:16 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void cpu_notify_map_clients(void)
|
|
|
|
{
|
2015-03-16 10:03:35 +01:00
|
|
|
qemu_mutex_lock(&map_client_list_lock);
|
2015-03-16 10:03:36 +01:00
|
|
|
cpu_notify_map_clients_locked();
|
2015-03-16 10:03:35 +01:00
|
|
|
qemu_mutex_unlock(&map_client_list_lock);
|
2009-01-22 17:59:16 +01:00
|
|
|
}
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
static bool flatview_access_valid(FlatView *fv, hwaddr addr, int len,
|
2018-05-31 15:50:52 +02:00
|
|
|
bool is_write, MemTxAttrs attrs)
|
2013-04-11 15:40:59 +02:00
|
|
|
{
|
2013-05-29 12:42:00 +02:00
|
|
|
MemoryRegion *mr;
|
2013-04-11 15:40:59 +02:00
|
|
|
hwaddr l, xlat;
|
|
|
|
|
|
|
|
while (len > 0) {
|
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = flatview_translate(fv, addr, &xlat, &l, is_write, attrs);
|
2013-05-29 12:42:00 +02:00
|
|
|
if (!memory_access_is_direct(mr, is_write)) {
|
|
|
|
l = memory_access_size(mr, l, addr);
|
2018-05-31 15:50:52 +02:00
|
|
|
if (!memory_region_access_valid(mr, xlat, l, is_write, attrs)) {
|
2013-04-11 15:40:59 +02:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
len -= l;
|
|
|
|
addr += l;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2017-09-21 10:50:58 +02:00
|
|
|
bool address_space_access_valid(AddressSpace *as, hwaddr addr,
|
2018-05-31 15:50:52 +02:00
|
|
|
int len, bool is_write,
|
|
|
|
MemTxAttrs attrs)
|
2017-09-21 10:50:58 +02:00
|
|
|
{
|
2018-03-05 00:23:26 +01:00
|
|
|
FlatView *fv;
|
|
|
|
bool result;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
fv = address_space_to_flatview(as);
|
2018-05-31 15:50:52 +02:00
|
|
|
result = flatview_access_valid(fv, addr, len, is_write, attrs);
|
2018-03-05 00:23:26 +01:00
|
|
|
rcu_read_unlock();
|
|
|
|
return result;
|
2017-09-21 10:50:58 +02:00
|
|
|
}
|
|
|
|
|
2016-11-22 12:04:31 +01:00
|
|
|
static hwaddr
|
2017-09-21 10:50:58 +02:00
|
|
|
flatview_extend_translation(FlatView *fv, hwaddr addr,
|
2018-05-31 15:50:52 +02:00
|
|
|
hwaddr target_len,
|
|
|
|
MemoryRegion *mr, hwaddr base, hwaddr len,
|
|
|
|
bool is_write, MemTxAttrs attrs)
|
2016-11-22 12:04:31 +01:00
|
|
|
{
|
|
|
|
hwaddr done = 0;
|
|
|
|
hwaddr xlat;
|
|
|
|
MemoryRegion *this_mr;
|
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
target_len -= len;
|
|
|
|
addr += len;
|
|
|
|
done += len;
|
|
|
|
if (target_len == 0) {
|
|
|
|
return done;
|
|
|
|
}
|
|
|
|
|
|
|
|
len = target_len;
|
2017-09-21 10:50:58 +02:00
|
|
|
this_mr = flatview_translate(fv, addr, &xlat,
|
2018-05-31 15:50:52 +02:00
|
|
|
&len, is_write, attrs);
|
2016-11-22 12:04:31 +01:00
|
|
|
if (this_mr != mr || xlat != base + done) {
|
|
|
|
return done;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-01-22 17:59:11 +01:00
|
|
|
/* Map a physical memory region into a host virtual address.
|
|
|
|
* May map a subset of the requested range, given by and returned in *plen.
|
|
|
|
* May return NULL if resources needed to perform the mapping are exhausted.
|
|
|
|
* Use only for reads OR writes - not for read-modify-write operations.
|
2009-01-22 17:59:16 +01:00
|
|
|
* Use cpu_register_map_client() to know when retrying the map operation is
|
|
|
|
* likely to succeed.
|
2009-01-22 17:59:11 +01:00
|
|
|
*/
|
2012-10-03 16:22:53 +02:00
|
|
|
void *address_space_map(AddressSpace *as,
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr addr,
|
|
|
|
hwaddr *plen,
|
2018-05-31 15:50:52 +02:00
|
|
|
bool is_write,
|
|
|
|
MemTxAttrs attrs)
|
2009-01-22 17:59:11 +01:00
|
|
|
{
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr len = *plen;
|
2016-11-22 12:04:31 +01:00
|
|
|
hwaddr l, xlat;
|
|
|
|
MemoryRegion *mr;
|
2015-12-16 10:31:26 +01:00
|
|
|
void *ptr;
|
2018-03-05 00:23:26 +01:00
|
|
|
FlatView *fv;
|
2009-01-22 17:59:11 +01:00
|
|
|
|
2013-06-28 17:29:27 +02:00
|
|
|
if (len == 0) {
|
|
|
|
return NULL;
|
|
|
|
}
|
2011-05-19 19:35:45 +02:00
|
|
|
|
2013-06-28 17:29:27 +02:00
|
|
|
l = len;
|
2015-03-18 14:21:43 +01:00
|
|
|
rcu_read_lock();
|
2018-03-05 00:23:26 +01:00
|
|
|
fv = address_space_to_flatview(as);
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = flatview_translate(fv, addr, &xlat, &l, is_write, attrs);
|
2015-03-18 14:21:43 +01:00
|
|
|
|
2013-06-28 17:29:27 +02:00
|
|
|
if (!memory_access_is_direct(mr, is_write)) {
|
2015-03-16 10:03:33 +01:00
|
|
|
if (atomic_xchg(&bounce.in_use, true)) {
|
2015-03-18 14:21:43 +01:00
|
|
|
rcu_read_unlock();
|
2013-06-28 17:29:27 +02:00
|
|
|
return NULL;
|
2009-01-22 17:59:11 +01:00
|
|
|
}
|
2013-07-22 14:30:23 +02:00
|
|
|
/* Avoid unbounded allocations */
|
|
|
|
l = MIN(l, TARGET_PAGE_SIZE);
|
|
|
|
bounce.buffer = qemu_memalign(TARGET_PAGE_SIZE, l);
|
2013-06-28 17:29:27 +02:00
|
|
|
bounce.addr = addr;
|
|
|
|
bounce.len = l;
|
2013-06-28 17:33:29 +02:00
|
|
|
|
|
|
|
memory_region_ref(mr);
|
|
|
|
bounce.mr = mr;
|
2013-06-28 17:29:27 +02:00
|
|
|
if (!is_write) {
|
2017-09-21 10:50:58 +02:00
|
|
|
flatview_read(fv, addr, MEMTXATTRS_UNSPECIFIED,
|
2015-04-26 17:49:24 +02:00
|
|
|
bounce.buffer, l);
|
2011-06-27 19:26:06 +02:00
|
|
|
}
|
2009-01-22 17:59:11 +01:00
|
|
|
|
2015-03-18 14:21:43 +01:00
|
|
|
rcu_read_unlock();
|
2013-06-28 17:29:27 +02:00
|
|
|
*plen = l;
|
|
|
|
return bounce.buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2013-06-28 17:33:29 +02:00
|
|
|
memory_region_ref(mr);
|
2017-09-21 10:50:58 +02:00
|
|
|
*plen = flatview_extend_translation(fv, addr, len, mr, xlat,
|
2018-05-31 15:50:52 +02:00
|
|
|
l, is_write, attrs);
|
exec: Add lock parameter to qemu_ram_ptr_length
Commit 04bf2526ce87f21b32c9acba1c5518708c243ad0 (exec: use
qemu_ram_ptr_length to access guest ram) start using qemu_ram_ptr_length
instead of qemu_map_ram_ptr, but when used with Xen, the behavior of
both function is different. They both call xen_map_cache, but one with
"lock", meaning the mapping of guest memory is never released
implicitly, and the second one without, which means, mapping can be
release later, when needed.
In the context of address_space_{read,write}_continue, the ptr to those
mapping should not be locked because it is used immediatly and never
used again.
The lock parameter make it explicit in which context qemu_ram_ptr_length
is called.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20170726165326.10327-1-anthony.perard@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-26 18:53:26 +02:00
|
|
|
ptr = qemu_ram_ptr_length(mr->ram_block, xlat, plen, true);
|
2015-12-16 10:31:26 +01:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return ptr;
|
2009-01-22 17:59:11 +01:00
|
|
|
}
|
|
|
|
|
2012-10-03 16:22:53 +02:00
|
|
|
/* Unmaps a memory region previously mapped by address_space_map().
|
2009-01-22 17:59:11 +01:00
|
|
|
* Will also mark the memory as dirty if is_write == 1. access_len gives
|
|
|
|
* the amount of memory that was actually read or written by the caller.
|
|
|
|
*/
|
2012-10-23 12:30:10 +02:00
|
|
|
void address_space_unmap(AddressSpace *as, void *buffer, hwaddr len,
|
|
|
|
int is_write, hwaddr access_len)
|
2009-01-22 17:59:11 +01:00
|
|
|
{
|
|
|
|
if (buffer != bounce.buffer) {
|
2013-06-28 17:33:29 +02:00
|
|
|
MemoryRegion *mr;
|
|
|
|
ram_addr_t addr1;
|
|
|
|
|
2016-03-25 12:55:08 +01:00
|
|
|
mr = memory_region_from_host(buffer, &addr1);
|
2013-06-28 17:33:29 +02:00
|
|
|
assert(mr != NULL);
|
2009-01-22 17:59:11 +01:00
|
|
|
if (is_write) {
|
2015-03-23 11:45:53 +01:00
|
|
|
invalidate_and_set_dirty(mr, addr1, access_len);
|
2009-01-22 17:59:11 +01:00
|
|
|
}
|
2011-06-21 22:59:09 +02:00
|
|
|
if (xen_enabled()) {
|
2011-06-21 22:59:08 +02:00
|
|
|
xen_invalidate_map_cache_entry(buffer);
|
2010-09-16 14:57:49 +02:00
|
|
|
}
|
2013-06-28 17:33:29 +02:00
|
|
|
memory_region_unref(mr);
|
2009-01-22 17:59:11 +01:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (is_write) {
|
2015-04-26 17:49:24 +02:00
|
|
|
address_space_write(as, bounce.addr, MEMTXATTRS_UNSPECIFIED,
|
|
|
|
bounce.buffer, access_len);
|
2009-01-22 17:59:11 +01:00
|
|
|
}
|
2010-01-24 22:23:56 +01:00
|
|
|
qemu_vfree(bounce.buffer);
|
2009-01-22 17:59:11 +01:00
|
|
|
bounce.buffer = NULL;
|
2013-06-28 17:33:29 +02:00
|
|
|
memory_region_unref(bounce.mr);
|
2015-03-16 10:03:33 +01:00
|
|
|
atomic_mb_set(&bounce.in_use, false);
|
2009-01-22 17:59:16 +01:00
|
|
|
cpu_notify_map_clients();
|
2009-01-22 17:59:11 +01:00
|
|
|
}
|
2006-04-23 19:14:48 +02:00
|
|
|
|
2012-10-23 12:30:10 +02:00
|
|
|
void *cpu_physical_memory_map(hwaddr addr,
|
|
|
|
hwaddr *plen,
|
2012-10-03 16:22:53 +02:00
|
|
|
int is_write)
|
|
|
|
{
|
2018-05-31 15:50:52 +02:00
|
|
|
return address_space_map(&address_space_memory, addr, plen, is_write,
|
|
|
|
MEMTXATTRS_UNSPECIFIED);
|
2012-10-03 16:22:53 +02:00
|
|
|
}
|
|
|
|
|
2012-10-23 12:30:10 +02:00
|
|
|
void cpu_physical_memory_unmap(void *buffer, hwaddr len,
|
|
|
|
int is_write, hwaddr access_len)
|
2012-10-03 16:22:53 +02:00
|
|
|
{
|
|
|
|
return address_space_unmap(&address_space_memory, buffer, len, is_write, access_len);
|
|
|
|
}
|
|
|
|
|
2016-11-22 11:34:02 +01:00
|
|
|
#define ARG1_DECL AddressSpace *as
|
|
|
|
#define ARG1 as
|
|
|
|
#define SUFFIX
|
|
|
|
#define TRANSLATE(...) address_space_translate(as, __VA_ARGS__)
|
|
|
|
#define IS_DIRECT(mr, is_write) memory_access_is_direct(mr, is_write)
|
|
|
|
#define MAP_RAM(mr, ofs) qemu_map_ram_ptr((mr)->ram_block, ofs)
|
|
|
|
#define INVALIDATE(mr, ofs, len) invalidate_and_set_dirty(mr, ofs, len)
|
|
|
|
#define RCU_READ_LOCK(...) rcu_read_lock()
|
|
|
|
#define RCU_READ_UNLOCK(...) rcu_read_unlock()
|
|
|
|
#include "memory_ldst.inc.c"
|
2011-07-06 09:09:23 +02:00
|
|
|
|
2016-11-22 12:04:52 +01:00
|
|
|
int64_t address_space_cache_init(MemoryRegionCache *cache,
|
|
|
|
AddressSpace *as,
|
|
|
|
hwaddr addr,
|
|
|
|
hwaddr len,
|
|
|
|
bool is_write)
|
|
|
|
{
|
2018-03-18 18:26:36 +01:00
|
|
|
AddressSpaceDispatch *d;
|
|
|
|
hwaddr l;
|
|
|
|
MemoryRegion *mr;
|
|
|
|
|
|
|
|
assert(len > 0);
|
|
|
|
|
|
|
|
l = len;
|
|
|
|
cache->fv = address_space_get_flatview(as);
|
|
|
|
d = flatview_to_dispatch(cache->fv);
|
|
|
|
cache->mrs = *address_space_translate_internal(d, addr, &cache->xlat, &l, true);
|
|
|
|
|
|
|
|
mr = cache->mrs.mr;
|
|
|
|
memory_region_ref(mr);
|
|
|
|
if (memory_access_is_direct(mr, is_write)) {
|
2018-05-31 15:50:52 +02:00
|
|
|
/* We don't care about the memory attributes here as we're only
|
|
|
|
* doing this if we found actual RAM, which behaves the same
|
|
|
|
* regardless of attributes; so UNSPECIFIED is fine.
|
|
|
|
*/
|
2018-03-18 18:26:36 +01:00
|
|
|
l = flatview_extend_translation(cache->fv, addr, len, mr,
|
2018-05-31 15:50:52 +02:00
|
|
|
cache->xlat, l, is_write,
|
|
|
|
MEMTXATTRS_UNSPECIFIED);
|
2018-03-18 18:26:36 +01:00
|
|
|
cache->ptr = qemu_ram_ptr_length(mr->ram_block, cache->xlat, &l, true);
|
|
|
|
} else {
|
|
|
|
cache->ptr = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
cache->len = l;
|
|
|
|
cache->is_write = is_write;
|
|
|
|
return l;
|
2016-11-22 12:04:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void address_space_cache_invalidate(MemoryRegionCache *cache,
|
|
|
|
hwaddr addr,
|
|
|
|
hwaddr access_len)
|
|
|
|
{
|
2018-03-18 18:26:36 +01:00
|
|
|
assert(cache->is_write);
|
|
|
|
if (likely(cache->ptr)) {
|
|
|
|
invalidate_and_set_dirty(cache->mrs.mr, addr + cache->xlat, access_len);
|
|
|
|
}
|
2016-11-22 12:04:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void address_space_cache_destroy(MemoryRegionCache *cache)
|
|
|
|
{
|
2018-03-18 18:26:36 +01:00
|
|
|
if (!cache->mrs.mr) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (xen_enabled()) {
|
|
|
|
xen_invalidate_map_cache_entry(cache->ptr);
|
|
|
|
}
|
|
|
|
memory_region_unref(cache->mrs.mr);
|
|
|
|
flatview_unref(cache->fv);
|
|
|
|
cache->mrs.mr = NULL;
|
|
|
|
cache->fv = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from RCU critical section. This function has the same
|
|
|
|
* semantics as address_space_translate, but it only works on a
|
|
|
|
* predefined range of a MemoryRegion that was mapped with
|
|
|
|
* address_space_cache_init.
|
|
|
|
*/
|
|
|
|
static inline MemoryRegion *address_space_translate_cached(
|
|
|
|
MemoryRegionCache *cache, hwaddr addr, hwaddr *xlat,
|
2018-05-31 15:50:52 +02:00
|
|
|
hwaddr *plen, bool is_write, MemTxAttrs attrs)
|
2018-03-18 18:26:36 +01:00
|
|
|
{
|
|
|
|
MemoryRegionSection section;
|
|
|
|
MemoryRegion *mr;
|
|
|
|
IOMMUMemoryRegion *iommu_mr;
|
|
|
|
AddressSpace *target_as;
|
|
|
|
|
|
|
|
assert(!cache->ptr);
|
|
|
|
*xlat = addr + cache->xlat;
|
|
|
|
|
|
|
|
mr = cache->mrs.mr;
|
|
|
|
iommu_mr = memory_region_get_iommu(mr);
|
|
|
|
if (!iommu_mr) {
|
|
|
|
/* MMIO region. */
|
|
|
|
return mr;
|
|
|
|
}
|
|
|
|
|
|
|
|
section = address_space_translate_iommu(iommu_mr, xlat, plen,
|
|
|
|
NULL, is_write, true,
|
2018-05-31 15:50:53 +02:00
|
|
|
&target_as, attrs);
|
2018-03-18 18:26:36 +01:00
|
|
|
return section.mr;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from RCU critical section. address_space_read_cached uses this
|
|
|
|
* out of line function when the target is an MMIO or IOMMU region.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
address_space_read_cached_slow(MemoryRegionCache *cache, hwaddr addr,
|
|
|
|
void *buf, int len)
|
|
|
|
{
|
|
|
|
hwaddr addr1, l;
|
|
|
|
MemoryRegion *mr;
|
|
|
|
|
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = address_space_translate_cached(cache, addr, &addr1, &l, false,
|
|
|
|
MEMTXATTRS_UNSPECIFIED);
|
2018-03-18 18:26:36 +01:00
|
|
|
flatview_read_continue(cache->fv,
|
|
|
|
addr, MEMTXATTRS_UNSPECIFIED, buf, len,
|
|
|
|
addr1, l, mr);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from RCU critical section. address_space_write_cached uses this
|
|
|
|
* out of line function when the target is an MMIO or IOMMU region.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
address_space_write_cached_slow(MemoryRegionCache *cache, hwaddr addr,
|
|
|
|
const void *buf, int len)
|
|
|
|
{
|
|
|
|
hwaddr addr1, l;
|
|
|
|
MemoryRegion *mr;
|
|
|
|
|
|
|
|
l = len;
|
2018-05-31 15:50:52 +02:00
|
|
|
mr = address_space_translate_cached(cache, addr, &addr1, &l, true,
|
|
|
|
MEMTXATTRS_UNSPECIFIED);
|
2018-03-18 18:26:36 +01:00
|
|
|
flatview_write_continue(cache->fv,
|
|
|
|
addr, MEMTXATTRS_UNSPECIFIED, buf, len,
|
|
|
|
addr1, l, mr);
|
2016-11-22 12:04:52 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
#define ARG1_DECL MemoryRegionCache *cache
|
|
|
|
#define ARG1 cache
|
2018-03-18 18:26:36 +01:00
|
|
|
#define SUFFIX _cached_slow
|
|
|
|
#define TRANSLATE(...) address_space_translate_cached(cache, __VA_ARGS__)
|
|
|
|
#define IS_DIRECT(mr, is_write) memory_access_is_direct(mr, is_write)
|
|
|
|
#define MAP_RAM(mr, ofs) (cache->ptr + (ofs - cache->xlat))
|
2017-04-03 13:41:28 +02:00
|
|
|
#define INVALIDATE(mr, ofs, len) invalidate_and_set_dirty(mr, ofs, len)
|
2018-03-18 18:26:36 +01:00
|
|
|
#define RCU_READ_LOCK() ((void)0)
|
|
|
|
#define RCU_READ_UNLOCK() ((void)0)
|
2016-11-22 12:04:52 +01:00
|
|
|
#include "memory_ldst.inc.c"
|
|
|
|
|
2009-03-28 18:51:36 +01:00
|
|
|
/* virtual memory access for debug (includes writing to ROM) */
|
2013-06-29 19:40:58 +02:00
|
|
|
int cpu_memory_rw_debug(CPUState *cpu, target_ulong addr,
|
2004-02-26 00:24:04 +01:00
|
|
|
uint8_t *buf, int len, int is_write)
|
2004-01-24 16:23:36 +01:00
|
|
|
{
|
|
|
|
int l;
|
2012-10-23 12:30:10 +02:00
|
|
|
hwaddr phys_addr;
|
2007-04-07 13:21:28 +02:00
|
|
|
target_ulong page;
|
2004-01-24 16:23:36 +01:00
|
|
|
|
2017-03-07 15:19:08 +01:00
|
|
|
cpu_synchronize_state(cpu);
|
2004-01-24 16:23:36 +01:00
|
|
|
while (len > 0) {
|
2016-01-21 15:15:06 +01:00
|
|
|
int asidx;
|
|
|
|
MemTxAttrs attrs;
|
|
|
|
|
2004-01-24 16:23:36 +01:00
|
|
|
page = addr & TARGET_PAGE_MASK;
|
2016-01-21 15:15:06 +01:00
|
|
|
phys_addr = cpu_get_phys_page_attrs_debug(cpu, page, &attrs);
|
|
|
|
asidx = cpu_asidx_from_attrs(cpu, attrs);
|
2004-01-24 16:23:36 +01:00
|
|
|
/* if no physical page mapped, return an error */
|
|
|
|
if (phys_addr == -1)
|
|
|
|
return -1;
|
|
|
|
l = (page + TARGET_PAGE_SIZE) - addr;
|
|
|
|
if (l > len)
|
|
|
|
l = len;
|
2009-03-28 18:51:36 +01:00
|
|
|
phys_addr += (addr & ~TARGET_PAGE_MASK);
|
2013-12-13 07:31:02 +01:00
|
|
|
if (is_write) {
|
2016-01-21 15:15:06 +01:00
|
|
|
cpu_physical_memory_write_rom(cpu->cpu_ases[asidx].as,
|
|
|
|
phys_addr, buf, l);
|
2013-12-13 07:31:02 +01:00
|
|
|
} else {
|
2016-01-21 15:15:06 +01:00
|
|
|
address_space_rw(cpu->cpu_ases[asidx].as, phys_addr,
|
|
|
|
MEMTXATTRS_UNSPECIFIED,
|
2015-04-26 17:49:24 +02:00
|
|
|
buf, l, 0);
|
2013-12-13 07:31:02 +01:00
|
|
|
}
|
2004-01-24 16:23:36 +01:00
|
|
|
len -= l;
|
|
|
|
buf += l;
|
|
|
|
addr += l;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
2015-11-05 19:10:29 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Allows code that needs to deal with migration bitmaps etc to still be built
|
|
|
|
* target independent.
|
|
|
|
*/
|
2017-03-21 09:09:14 +01:00
|
|
|
size_t qemu_target_page_size(void)
|
2015-11-05 19:10:29 +01:00
|
|
|
{
|
2017-03-21 09:09:14 +01:00
|
|
|
return TARGET_PAGE_SIZE;
|
2015-11-05 19:10:29 +01:00
|
|
|
}
|
|
|
|
|
2017-04-24 21:03:48 +02:00
|
|
|
int qemu_target_page_bits(void)
|
|
|
|
{
|
|
|
|
return TARGET_PAGE_BITS;
|
|
|
|
}
|
|
|
|
|
|
|
|
int qemu_target_page_bits_min(void)
|
|
|
|
{
|
|
|
|
return TARGET_PAGE_BITS_MIN;
|
|
|
|
}
|
2010-03-01 01:08:59 +01:00
|
|
|
#endif
|
2004-01-24 16:23:36 +01:00
|
|
|
|
2013-01-06 19:30:17 +01:00
|
|
|
/*
|
|
|
|
* A helper function for the _utterly broken_ virtio device model to find out if
|
|
|
|
* it's running on a big endian machine. Don't do this at home kids!
|
|
|
|
*/
|
2014-06-24 19:26:29 +02:00
|
|
|
bool target_words_bigendian(void);
|
|
|
|
bool target_words_bigendian(void)
|
2013-01-06 19:30:17 +01:00
|
|
|
{
|
|
|
|
#if defined(TARGET_WORDS_BIGENDIAN)
|
|
|
|
return true;
|
|
|
|
#else
|
|
|
|
return false;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2012-05-07 06:04:18 +02:00
|
|
|
#ifndef CONFIG_USER_ONLY
|
2012-10-23 12:30:10 +02:00
|
|
|
bool cpu_physical_memory_is_io(hwaddr phys_addr)
|
2012-05-07 06:04:18 +02:00
|
|
|
{
|
2013-05-29 12:42:00 +02:00
|
|
|
MemoryRegion*mr;
|
2013-05-24 12:59:37 +02:00
|
|
|
hwaddr l = 1;
|
2015-03-18 14:21:43 +01:00
|
|
|
bool res;
|
2012-05-07 06:04:18 +02:00
|
|
|
|
2015-03-18 14:21:43 +01:00
|
|
|
rcu_read_lock();
|
2013-05-29 12:42:00 +02:00
|
|
|
mr = address_space_translate(&address_space_memory,
|
2018-05-31 15:50:52 +02:00
|
|
|
phys_addr, &phys_addr, &l, false,
|
|
|
|
MEMTXATTRS_UNSPECIFIED);
|
2012-05-07 06:04:18 +02:00
|
|
|
|
2015-03-18 14:21:43 +01:00
|
|
|
res = !(memory_region_is_ram(mr) || memory_region_is_romd(mr));
|
|
|
|
rcu_read_unlock();
|
|
|
|
return res;
|
2012-05-07 06:04:18 +02:00
|
|
|
}
|
2013-06-26 03:35:34 +02:00
|
|
|
|
2015-05-21 14:24:13 +02:00
|
|
|
int qemu_ram_foreach_block(RAMBlockIterFunc func, void *opaque)
|
2013-06-26 03:35:34 +02:00
|
|
|
{
|
|
|
|
RAMBlock *block;
|
2015-05-21 14:24:13 +02:00
|
|
|
int ret = 0;
|
2013-06-26 03:35:34 +02:00
|
|
|
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_lock();
|
2017-05-12 06:17:39 +02:00
|
|
|
RAMBLOCK_FOREACH(block) {
|
2015-05-21 14:24:13 +02:00
|
|
|
ret = func(block->idstr, block->host, block->offset,
|
|
|
|
block->used_length, opaque);
|
|
|
|
if (ret) {
|
|
|
|
break;
|
|
|
|
}
|
2013-06-26 03:35:34 +02:00
|
|
|
}
|
2013-09-05 20:41:35 +02:00
|
|
|
rcu_read_unlock();
|
2015-05-21 14:24:13 +02:00
|
|
|
return ret;
|
2013-06-26 03:35:34 +02:00
|
|
|
}
|
2017-02-24 19:28:32 +01:00
|
|
|
|
2018-05-14 08:57:00 +02:00
|
|
|
int qemu_ram_foreach_migratable_block(RAMBlockIterFunc func, void *opaque)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
RAMBLOCK_FOREACH(block) {
|
|
|
|
if (!qemu_ram_is_migratable(block)) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
ret = func(block->idstr, block->host, block->offset,
|
|
|
|
block->used_length, opaque);
|
|
|
|
if (ret) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-02-24 19:28:32 +01:00
|
|
|
/*
|
|
|
|
* Unmap pages of memory from start to start+length such that
|
|
|
|
* they a) read as 0, b) Trigger whatever fault mechanism
|
|
|
|
* the OS provides for postcopy.
|
|
|
|
* The pages must be unmapped by the end of the function.
|
|
|
|
* Returns: 0 on success, none-0 on failure
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
int ram_block_discard_range(RAMBlock *rb, uint64_t start, size_t length)
|
|
|
|
{
|
|
|
|
int ret = -1;
|
|
|
|
|
|
|
|
uint8_t *host_startaddr = rb->host + start;
|
|
|
|
|
|
|
|
if ((uintptr_t)host_startaddr & (rb->page_size - 1)) {
|
|
|
|
error_report("ram_block_discard_range: Unaligned start address: %p",
|
|
|
|
host_startaddr);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((start + length) <= rb->used_length) {
|
2018-03-12 18:20:56 +01:00
|
|
|
bool need_madvise, need_fallocate;
|
2017-02-24 19:28:32 +01:00
|
|
|
uint8_t *host_endaddr = host_startaddr + length;
|
|
|
|
if ((uintptr_t)host_endaddr & (rb->page_size - 1)) {
|
|
|
|
error_report("ram_block_discard_range: Unaligned end address: %p",
|
|
|
|
host_endaddr);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
errno = ENOTSUP; /* If we are missing MADVISE etc */
|
|
|
|
|
2018-03-12 18:20:56 +01:00
|
|
|
/* The logic here is messy;
|
|
|
|
* madvise DONTNEED fails for hugepages
|
|
|
|
* fallocate works on hugepages and shmem
|
|
|
|
*/
|
|
|
|
need_madvise = (rb->page_size == qemu_host_page_size);
|
|
|
|
need_fallocate = rb->fd != -1;
|
|
|
|
if (need_fallocate) {
|
|
|
|
/* For a file, this causes the area of the file to be zero'd
|
|
|
|
* if read, and for hugetlbfs also causes it to be unmapped
|
|
|
|
* so a userfault will trigger.
|
2017-02-24 19:28:33 +01:00
|
|
|
*/
|
|
|
|
#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
|
|
|
|
ret = fallocate(rb->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
|
|
|
|
start, length);
|
2018-03-12 18:20:56 +01:00
|
|
|
if (ret) {
|
|
|
|
ret = -errno;
|
|
|
|
error_report("ram_block_discard_range: Failed to fallocate "
|
|
|
|
"%s:%" PRIx64 " +%zx (%d)",
|
|
|
|
rb->idstr, start, length, ret);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
ret = -ENOSYS;
|
|
|
|
error_report("ram_block_discard_range: fallocate not available/file"
|
|
|
|
"%s:%" PRIx64 " +%zx (%d)",
|
|
|
|
rb->idstr, start, length, ret);
|
|
|
|
goto err;
|
2017-02-24 19:28:33 +01:00
|
|
|
#endif
|
|
|
|
}
|
2018-03-12 18:20:56 +01:00
|
|
|
if (need_madvise) {
|
|
|
|
/* For normal RAM this causes it to be unmapped,
|
|
|
|
* for shared memory it causes the local mapping to disappear
|
|
|
|
* and to fall back on the file contents (which we just
|
|
|
|
* fallocate'd away).
|
|
|
|
*/
|
|
|
|
#if defined(CONFIG_MADVISE)
|
|
|
|
ret = madvise(host_startaddr, length, MADV_DONTNEED);
|
|
|
|
if (ret) {
|
|
|
|
ret = -errno;
|
|
|
|
error_report("ram_block_discard_range: Failed to discard range "
|
|
|
|
"%s:%" PRIx64 " +%zx (%d)",
|
|
|
|
rb->idstr, start, length, ret);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
ret = -ENOSYS;
|
|
|
|
error_report("ram_block_discard_range: MADVISE not available"
|
2017-02-24 19:28:32 +01:00
|
|
|
"%s:%" PRIx64 " +%zx (%d)",
|
|
|
|
rb->idstr, start, length, ret);
|
2018-03-12 18:20:56 +01:00
|
|
|
goto err;
|
|
|
|
#endif
|
2017-02-24 19:28:32 +01:00
|
|
|
}
|
2018-03-12 18:20:56 +01:00
|
|
|
trace_ram_block_discard_range(rb->idstr, host_startaddr, length,
|
|
|
|
need_madvise, need_fallocate, ret);
|
2017-02-24 19:28:32 +01:00
|
|
|
} else {
|
|
|
|
error_report("ram_block_discard_range: Overrun block '%s' (%" PRIu64
|
|
|
|
"/%zx/" RAM_ADDR_FMT")",
|
|
|
|
rb->idstr, start, length, rb->used_length);
|
|
|
|
}
|
|
|
|
|
|
|
|
err:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-06-27 21:53:38 +02:00
|
|
|
#endif
|
2017-07-03 12:12:13 +02:00
|
|
|
|
|
|
|
void page_size_init(void)
|
|
|
|
{
|
|
|
|
/* NOTE: we can always suppose that qemu_host_page_size >=
|
|
|
|
TARGET_PAGE_SIZE */
|
|
|
|
if (qemu_host_page_size == 0) {
|
|
|
|
qemu_host_page_size = qemu_real_host_page_size;
|
|
|
|
}
|
|
|
|
if (qemu_host_page_size < TARGET_PAGE_SIZE) {
|
|
|
|
qemu_host_page_size = TARGET_PAGE_SIZE;
|
|
|
|
}
|
|
|
|
qemu_host_page_mask = -(intptr_t)qemu_host_page_size;
|
|
|
|
}
|
2017-09-21 10:51:06 +02:00
|
|
|
|
|
|
|
#if !defined(CONFIG_USER_ONLY)
|
|
|
|
|
|
|
|
static void mtree_print_phys_entries(fprintf_function mon, void *f,
|
|
|
|
int start, int end, int skip, int ptr)
|
|
|
|
{
|
|
|
|
if (start == end - 1) {
|
|
|
|
mon(f, "\t%3d ", start);
|
|
|
|
} else {
|
|
|
|
mon(f, "\t%3d..%-3d ", start, end - 1);
|
|
|
|
}
|
|
|
|
mon(f, " skip=%d ", skip);
|
|
|
|
if (ptr == PHYS_MAP_NODE_NIL) {
|
|
|
|
mon(f, " ptr=NIL");
|
|
|
|
} else if (!skip) {
|
|
|
|
mon(f, " ptr=#%d", ptr);
|
|
|
|
} else {
|
|
|
|
mon(f, " ptr=[%d]", ptr);
|
|
|
|
}
|
|
|
|
mon(f, "\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
#define MR_SIZE(size) (int128_nz(size) ? (hwaddr)int128_get64( \
|
|
|
|
int128_sub((size), int128_one())) : 0)
|
|
|
|
|
|
|
|
void mtree_print_dispatch(fprintf_function mon, void *f,
|
|
|
|
AddressSpaceDispatch *d, MemoryRegion *root)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
mon(f, " Dispatch\n");
|
|
|
|
mon(f, " Physical sections\n");
|
|
|
|
|
|
|
|
for (i = 0; i < d->map.sections_nb; ++i) {
|
|
|
|
MemoryRegionSection *s = d->map.sections + i;
|
|
|
|
const char *names[] = { " [unassigned]", " [not dirty]",
|
|
|
|
" [ROM]", " [watch]" };
|
|
|
|
|
|
|
|
mon(f, " #%d @" TARGET_FMT_plx ".." TARGET_FMT_plx " %s%s%s%s%s",
|
|
|
|
i,
|
|
|
|
s->offset_within_address_space,
|
|
|
|
s->offset_within_address_space + MR_SIZE(s->mr->size),
|
|
|
|
s->mr->name ? s->mr->name : "(noname)",
|
|
|
|
i < ARRAY_SIZE(names) ? names[i] : "",
|
|
|
|
s->mr == root ? " [ROOT]" : "",
|
|
|
|
s == d->mru_section ? " [MRU]" : "",
|
|
|
|
s->mr->is_iommu ? " [iommu]" : "");
|
|
|
|
|
|
|
|
if (s->mr->alias) {
|
|
|
|
mon(f, " alias=%s", s->mr->alias->name ?
|
|
|
|
s->mr->alias->name : "noname");
|
|
|
|
}
|
|
|
|
mon(f, "\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
mon(f, " Nodes (%d bits per level, %d levels) ptr=[%d] skip=%d\n",
|
|
|
|
P_L2_BITS, P_L2_LEVELS, d->phys_map.ptr, d->phys_map.skip);
|
|
|
|
for (i = 0; i < d->map.nodes_nb; ++i) {
|
|
|
|
int j, jprev;
|
|
|
|
PhysPageEntry prev;
|
|
|
|
Node *n = d->map.nodes + i;
|
|
|
|
|
|
|
|
mon(f, " [%d]\n", i);
|
|
|
|
|
|
|
|
for (j = 0, jprev = 0, prev = *n[0]; j < ARRAY_SIZE(*n); ++j) {
|
|
|
|
PhysPageEntry *pe = *n + j;
|
|
|
|
|
|
|
|
if (pe->ptr == prev.ptr && pe->skip == prev.skip) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
mtree_print_phys_entries(mon, f, jprev, j, prev.skip, prev.ptr);
|
|
|
|
|
|
|
|
jprev = j;
|
|
|
|
prev = *pe;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (jprev != ARRAY_SIZE(*n)) {
|
|
|
|
mtree_print_phys_entries(mon, f, jprev, j, prev.skip, prev.ptr);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|