Search
j0ke.net Open Build Service
>
Projects
>
virtualization
:
Qemu
>
qemu
> qemu-0.9.0-migration.patch
Sign Up
|
Log In
Username
Password
Cancel
Overview
Repositories
Revisions
Requests
Users
Advanced
Attributes
Meta
File qemu-0.9.0-migration.patch of Package qemu
2007-03-13 Gwenole Beauchesne <gbeauchesne@mandriva.com> * Merge Anthony Liguori's QEMU Live Migration (combined) patch + changes from the KVM tree. ================================================================================ --- qemu-0.9.0/Makefile.target +++ qemu-0.9.0/Makefile.target @@ -320,7 +320,7 @@ # must use static linking to avoid leaving stuff in virtual address space VL_OBJS=vl.o osdep.o readline.o monitor.o pci.o console.o loader.o isa_mmio.o -VL_OBJS+=cutils.o +VL_OBJS+=cutils.o migration.o VL_OBJS+=block.o block-raw.o VL_OBJS+=block-cow.o block-qcow.o aes.o block-vmdk.o block-cloop.o block-dmg.o block-bochs.o block-vpc.o block-vvfat.o block-qcow2.o ifdef CONFIG_WIN32 --- qemu-0.9.0/audio/wavaudio.c +++ qemu-0.9.0/audio/wavaudio.c @@ -151,7 +151,7 @@ le_store (hdr + 28, hw->info.freq << (bits16 + stereo), 4); le_store (hdr + 32, 1 << (bits16 + stereo), 2); - wav->f = qemu_fopen (conf.wav_path, "wb"); + wav->f = qemu_fopen_file (conf.wav_path, "wb"); if (!wav->f) { dolog ("Failed to open wave file `%s'\nReason: %s\n", conf.wav_path, strerror (errno)); --- qemu-0.9.0/audio/wavcapture.c +++ qemu-0.9.0/audio/wavcapture.c @@ -132,7 +132,7 @@ le_store (hdr + 28, freq << shift, 4); le_store (hdr + 32, 1 << shift, 2); - wav->f = qemu_fopen (path, "wb"); + wav->f = qemu_fopen_file (path, "wb"); if (!wav->f) { term_printf ("Failed to open wave file `%s'\nReason: %s\n", path, strerror (errno)); --- qemu-0.9.0/cpu-all.h +++ qemu-0.9.0/cpu-all.h @@ -892,6 +892,7 @@ #define VGA_DIRTY_FLAG 0x01 #define CODE_DIRTY_FLAG 0x02 +#define MIGRATION_DIRTY_FLAG 0x08 /* read dirty bit (return 0 or 1) */ static inline int cpu_physical_memory_is_dirty(ram_addr_t addr) @@ -914,6 +915,10 @@ int dirty_flags); void cpu_tlb_update_dirty(CPUState *env); +int cpu_physical_memory_set_dirty_tracking(int enable); + +int cpu_physical_memory_get_dirty_tracking(void); + void dump_exec_info(FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...)); --- qemu-0.9.0/cutils.c +++ qemu-0.9.0/cutils.c @@ -81,3 +81,43 @@ *ptr = p; return 1; } + +int hex2bin(char ch) +{ + if (ch >= '0' && ch <= '9') + return ch - '0'; + else if (ch >= 'A' && ch <= 'Z') + return 10 + ch - 'A'; + else if (ch >= 'a' && ch <= 'z') + return 10 + ch - 'a'; + + return -1; +} + +char *urldecode(const char *ptr) +{ + char *ret; + int i; + + ret = qemu_mallocz(strlen(ptr) + 1); + if (ret == NULL) + return NULL; + + for (i = 0; *ptr; ptr++, i++) { + switch (*ptr) { + case '%': + if (ptr[1] == 0 || ptr[2] == 0) + break; + ret[i] = hex2bin(ptr[1]) << 4 | hex2bin(ptr[2]); + ptr += 2; + break; + default: + ret[i] = *ptr; + break; + } + } + ret[i] = 0; + + return ret; +} + --- qemu-0.9.0/exec.c +++ qemu-0.9.0/exec.c @@ -82,6 +82,7 @@ int phys_ram_fd; uint8_t *phys_ram_base; uint8_t *phys_ram_dirty; +static int in_migration; CPUState *first_cpu; /* current CPU in the current thread. It is only valid inside @@ -1420,6 +1421,19 @@ #endif } +int cpu_physical_memory_set_dirty_tracking(int enable) +{ + int r=0; + + in_migration = enable; + return r; +} + +int cpu_physical_memory_get_dirty_tracking(void) +{ + return in_migration; +} + static inline void tlb_update_dirty(CPUTLBEntry *tlb_entry) { ram_addr_t ram_addr; @@ -2287,6 +2301,14 @@ return tswap16(val); } +#ifdef __GNUC__ +#define likely(x) __builtin_expect(!!(x), 1) +#define unlikely(x) __builtin_expect(!!(x), 0) +#else +#define likely(x) x +#define unlikely(x) x +#endif + /* warning: addr must be aligned. The ram page is not masked as dirty and the code inside is not invalidated. It is useful if the dirty bits are used to track modified PTEs */ @@ -2308,9 +2330,21 @@ io_index = (pd >> IO_MEM_SHIFT) & (IO_MEM_NB_ENTRIES - 1); io_mem_write[io_index][2](io_mem_opaque[io_index], addr, val); } else { - ptr = phys_ram_base + (pd & TARGET_PAGE_MASK) + - (addr & ~TARGET_PAGE_MASK); + unsigned long addr1; + addr1 = (pd & TARGET_PAGE_MASK) + (addr & ~TARGET_PAGE_MASK); + + ptr = phys_ram_base + addr1; stl_p(ptr, val); + + if (unlikely(in_migration)) { + if (!cpu_physical_memory_is_dirty(addr1)) { + /* invalidate code */ + tb_invalidate_phys_page_range(addr1, addr1 + 4, 0); + /* set dirty bit */ + phys_ram_dirty[addr1 >> TARGET_PAGE_BITS] |= + (0xff & ~CODE_DIRTY_FLAG); + } + } } } --- qemu-0.9.0/hw/usb-uhci.c +++ qemu-0.9.0/hw/usb-uhci.c @@ -144,6 +144,58 @@ } } +static void uhci_save(QEMUFile *f, void *opaque) +{ + UHCIState *s = opaque; + uint8_t num_ports = NB_PORTS; + int i; + + pci_device_save(&s->dev, f); + + qemu_put_8s(f, &num_ports); + for (i = 0; i < num_ports; ++i) + qemu_put_be16s(f, &s->ports[i].ctrl); + qemu_put_be16s(f, &s->cmd); + qemu_put_be16s(f, &s->status); + qemu_put_be16s(f, &s->intr); + qemu_put_be16s(f, &s->frnum); + qemu_put_be32s(f, &s->fl_base_addr); + qemu_put_8s(f, &s->sof_timing); + qemu_put_8s(f, &s->status2); + qemu_put_timer(f, s->frame_timer); +} + +static int uhci_load(QEMUFile* f,void* opaque,int version_id) +{ + UHCIState *s = opaque; + uint8_t num_ports; + int i, ret; + + if (version_id > 1) + return -EINVAL; + + ret = pci_device_load(&s->dev, f); + if (ret < 0) + return ret; + + qemu_get_8s(f, &num_ports); + if (num_ports != NB_PORTS) + return -EINVAL; + + for (i = 0; i < num_ports; ++i) + qemu_get_be16s(f, &s->ports[i].ctrl); + qemu_get_be16s(f, &s->cmd); + qemu_get_be16s(f, &s->status); + qemu_get_be16s(f, &s->intr); + qemu_get_be16s(f, &s->frnum); + qemu_get_be32s(f, &s->fl_base_addr); + qemu_get_8s(f, &s->sof_timing); + qemu_get_8s(f, &s->status2); + qemu_get_timer(f, s->frame_timer); + + return 0; +} + static void uhci_ioport_writeb(void *opaque, uint32_t addr, uint32_t val) { UHCIState *s = opaque; @@ -793,4 +845,6 @@ to rely on this. */ pci_register_io_region(&s->dev, 4, 0x20, PCI_ADDRESS_SPACE_IO, uhci_map); + + register_savevm("uhci", 0, 1, uhci_save, uhci_load, s); } --- qemu-0.9.0/migration.c +++ qemu-0.9.0/migration.c @@ -0,0 +1,753 @@ +/* + * QEMU migration support + * + * Copyright (C) 2007 Anthony Liguori <anthony@codemonkey.ws> + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + */ + +#include "vl.h" +#include "qemu_socket.h" + +#include <sys/wait.h> + +#define MIN_FINALIZE_SIZE (200 << 10) + +typedef struct MigrationState +{ + int fd; + int throttle_count; + int bps; + int updated_pages; + int last_updated_pages; + int iteration; + int n_buffer; + int throttled; + int *has_error; + char buffer[TARGET_PAGE_SIZE + 4]; + target_ulong addr; + QEMUTimer *timer; + void *opaque; + int detach; + int (*release)(void *opaque); +} MigrationState; + +static uint32_t max_throttle = (32 << 20); +static MigrationState *current_migration; + +//#define MIGRATION_VERIFY +#ifdef MIGRATION_VERIFY +static int save_verify_memory(QEMUFile *f, void *opaque); +static int load_verify_memory(QEMUFile *f, void *opaque, int version_id); +#endif /* MIGRATION_VERIFY */ + +/* QEMUFile migration implementation */ + +static void migrate_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size) +{ + MigrationState *s = opaque; + int offset = 0; + + if (*s->has_error) + return; + + while (offset < size) { + ssize_t len; + + len = write(s->fd, buf + offset, size - offset); + if (len == -1) { + if (errno == EAGAIN || errno == EINTR) + continue; + term_printf("migration: write failed (%s)\n", strerror(errno)); + *s->has_error = 10; + break; + } else if (len == 0) { + term_printf("migration: other side closed connection\n"); + *s->has_error = 11; + break; + } + + offset += len; + } +} + +static void migrate_close(void *opaque) +{ + MigrationState *s = opaque; + + if (s->release && s->release(s->opaque)) + *s->has_error = 12; + + qemu_free(s); + current_migration = NULL; +} + +/* Outgoing migration routines */ + +static void migrate_finish(MigrationState *s) +{ + QEMUFile *f; + int ret = 0; + int *has_error = s->has_error; + + fcntl(s->fd, F_SETFL, 0); + + if (! *has_error) { + f = qemu_fopen(s, migrate_put_buffer, NULL, migrate_close); + qemu_aio_flush(); + vm_stop(0); + qemu_put_be32(f, 1); + ret = qemu_live_savevm_state(f); +#ifdef MIGRATION_VERIFY + save_verify_memory(f, NULL); +#endif /* MIGRATION_VERIFY */ + qemu_fclose(f); + } + if (ret != 0 || *has_error) { + term_printf("Migration failed! ret=%d error=%d\n", ret, *has_error); + vm_start(); + } + if (!s->detach) + monitor_resume(); + qemu_free(has_error); + cpu_physical_memory_set_dirty_tracking(0); +} + +static int migrate_write_buffer(MigrationState *s) +{ + if (*s->has_error) + return 0; + + if (s->n_buffer != sizeof(s->buffer)) { + ssize_t len; + again: + len = write(s->fd, s->buffer + s->n_buffer, sizeof(s->buffer) - s->n_buffer); + if (len == -1) { + if (errno == EINTR) + goto again; + if (errno == EAGAIN) + return 1; + *s->has_error = 13; + return 0; + } + if (len == 0) { + *s->has_error = 14; + return 0; + } + + s->throttle_count += len; + s->n_buffer += len; + if (s->n_buffer != sizeof(s->buffer)) + goto again; + } + + if (s->throttle_count > max_throttle) { + s->throttled = 1; + qemu_set_fd_handler2(s->fd, NULL, NULL, NULL, NULL); + return 1; + } + + return 0; +} + +static int migrate_check_convergence(MigrationState *s) +{ + target_ulong addr; + int dirty_count = 0; + + for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) { + if (cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG)) + dirty_count++; + } + + return ((dirty_count * TARGET_PAGE_SIZE) < MIN_FINALIZE_SIZE); +} + +static void migrate_write(void *opaque) +{ + MigrationState *s = opaque; + + if (migrate_write_buffer(s)) + return; + + if (migrate_check_convergence(s) || *s->has_error) { + qemu_del_timer(s->timer); + qemu_free_timer(s->timer); + qemu_set_fd_handler2(s->fd, NULL, NULL, NULL, NULL); + migrate_finish(s); + return; + } + + while (s->addr < phys_ram_size) { + if (cpu_physical_memory_get_dirty(s->addr, MIGRATION_DIRTY_FLAG)) { + uint32_t value = cpu_to_be32(s->addr); + + memcpy(s->buffer, &value, 4); + memcpy(s->buffer + 4, phys_ram_base + s->addr, TARGET_PAGE_SIZE); + s->n_buffer = 0; + + cpu_physical_memory_reset_dirty(s->addr, s->addr + TARGET_PAGE_SIZE, MIGRATION_DIRTY_FLAG); + + s->addr += TARGET_PAGE_SIZE; + + s->updated_pages++; + + if (migrate_write_buffer(s)) + return; + } else + s->addr += TARGET_PAGE_SIZE; + } + + s->last_updated_pages = s->updated_pages; + s->updated_pages = 0; + s->addr = 0; + s->iteration++; +} + +static void migrate_reset_throttle(void *opaque) +{ + MigrationState *s = opaque; + + s->bps = s->throttle_count; + + if (s->throttled) { + s->throttled = 0; + qemu_set_fd_handler2(s->fd, NULL, NULL, migrate_write, s); + } + s->throttle_count = 0; + qemu_mod_timer(s->timer, qemu_get_clock(rt_clock) + 1000); +} + +static int start_migration(MigrationState *s) +{ + uint32_t value = cpu_to_be32(phys_ram_size); + target_phys_addr_t addr; + size_t offset = 0; + + while (offset != 4) { + ssize_t len = write(s->fd, ((char *)&value) + offset, 4 - offset); + if (len == -1 && errno == EINTR) + continue; + + if (len < 1) + return -EIO; + + offset += len; + } + + fcntl(s->fd, F_SETFL, O_NONBLOCK); + + for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) { + if (!cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG)) + cpu_physical_memory_set_dirty(addr); + } + + if (cpu_physical_memory_set_dirty_tracking(1)) { + *s->has_error = 16; + return -1; + } + + s->addr = 0; + s->iteration = 0; + s->updated_pages = 0; + s->last_updated_pages = 0; + s->n_buffer = sizeof(s->buffer); + s->timer = qemu_new_timer(rt_clock, migrate_reset_throttle, s); + + qemu_mod_timer(s->timer, qemu_get_clock(rt_clock)); + qemu_set_fd_handler2(s->fd, NULL, NULL, migrate_write, s); + return 0; +} + +static MigrationState *migration_init_fd(int detach, int fd) +{ + MigrationState *s; + + s = qemu_mallocz(sizeof(MigrationState)); + if (s == NULL) { + term_printf("Allocation error\n"); + return NULL; + } + + s->fd = fd; + s->has_error = qemu_mallocz(sizeof(int)); + if (s->has_error == NULL) { + term_printf("malloc failed (for has_error)\n"); + return NULL; + } + s->detach = detach; + + current_migration = s; + + if (start_migration(s) == -1) { + term_printf("Could not start migration\n"); + return NULL; + } + + if (!detach) + monitor_suspend(); + + return s; +} + +typedef struct MigrationCmdState +{ + int fd; + pid_t pid; +} MigrationCmdState; + +static int cmd_release(void *opaque) +{ + MigrationCmdState *c = opaque; + int status, ret; + + close(c->fd); + +again: + ret = waitpid(c->pid, &status, 0); + if (ret == -1 && errno == EINTR) + goto again; + + if (ret == -1) { + term_printf("migration: waitpid failed (%s)\n", strerror(errno)); + return -1; + } + /* FIXME: check and uncomment + * if (WIFEXITED(status)) + * status = WEXITSTATUS(status); + */ + return status; +} + +static MigrationState *migration_init_cmd(int detach, const char *command, char **argv) +{ + int fds[2]; + pid_t pid; + int i; + MigrationState *s; + + if (pipe(fds) == -1) { + term_printf("pipe() (%s)\n", strerror(errno)); + return NULL; + } + + pid = fork(); + if (pid == -1) { + close(fds[0]); + close(fds[1]); + term_printf("fork error (%s)\n", strerror(errno)); + return NULL; + } + if (pid == 0) { + close(fds[1]); + dup2(fds[0], STDIN_FILENO); + execvp(command, argv); + exit(1); + } else + close(fds[0]); + + for (i = 0; argv[i]; i++) + qemu_free(argv[i]); + qemu_free(argv); + + s = migration_init_fd(detach, fds[1]); + if (s) { + MigrationCmdState *c = qemu_mallocz(sizeof(*c)); + c->pid = pid; + c->fd = fds[1]; + s->release = cmd_release; + s->opaque = c; + } + + return s; +} + +static MigrationState *migration_init_exec(int detach, const char *command) +{ + char **argv = NULL; + + argv = qemu_mallocz(sizeof(char *) * 4); + argv[0] = strdup("sh"); + argv[1] = strdup("-c"); + argv[2] = strdup(command); + argv[3] = NULL; + + return migration_init_cmd(detach, "/bin/sh", argv); +} + +static MigrationState *migration_init_ssh(int detach, const char *host) +{ + int qemu_argc, daemonize = 0, argc, i; + char **qemu_argv, **argv; + const char *incoming = NULL; + + qemu_get_launch_info(&qemu_argc, &qemu_argv, &daemonize, &incoming); + + argc = 3 + qemu_argc; + if (!daemonize) + argc++; + if (!incoming) + argc+=2; + + argv = qemu_mallocz(sizeof(char *) * (argc + 1)); + argv[0] = strdup("ssh"); + argv[1] = strdup("-XC"); + argv[2] = strdup(host); + + for (i = 0; i < qemu_argc; i++) + argv[3 + i] = strdup(qemu_argv[i]); + + if (!daemonize) + argv[3 + i++] = strdup("-daemonize"); + if (!incoming) { + argv[3 + i++] = strdup("-incoming"); + argv[3 + i++] = strdup("stdio"); + } + + argv[3 + i] = NULL; + + return migration_init_cmd(detach, "ssh", argv); +} + +static int tcp_release(void *opaque) +{ + MigrationState *s = opaque; + uint8_t status = 0; + ssize_t len; + +again: + len = read(s->fd, &status, 1); + if (len == -1 && errno == EINTR) + goto again; + + close(s->fd); + + return (len != 1 || status != 0); +} + +static MigrationState *migration_init_tcp(int detach, const char *host) +{ + int fd; + struct sockaddr_in addr; + MigrationState *s; + + fd = socket(PF_INET, SOCK_STREAM, 0); + if (fd == -1) { + term_printf("socket() failed %s\n", strerror(errno)); + return NULL; + } + + addr.sin_family = AF_INET; + if (parse_host_port(&addr, host) == -1) { + term_printf("parse_host_port() FAILED for %s\n", host); + close(fd); + return NULL; + } + +again: + if (connect(fd, (struct sockaddr *)&addr, sizeof(addr)) == -1) { + if (errno == EINTR) + goto again; + term_printf("connect() failed %s\n", strerror(errno)); + close(fd); + return NULL; + } + + s = migration_init_fd(detach, fd); + if (s) { + s->opaque = s; + s->release = tcp_release; + } + return s; +} + +/* Incoming migration */ + +static int migrate_incoming_fd(int fd) +{ + int ret; + QEMUFile *f = qemu_fopen_fd(fd); + uint32_t addr; + extern void qemu_announce_self(void); + + if (qemu_get_be32(f) != phys_ram_size) + return 101; + + do { + int l; + addr = qemu_get_be32(f); + if (addr == 1) + break; + l = qemu_get_buffer(f, phys_ram_base + addr, TARGET_PAGE_SIZE); + if (l != TARGET_PAGE_SIZE) + return 102; + } while (1); + + + qemu_aio_flush(); + vm_stop(0); + ret = qemu_live_loadvm_state(f); +#ifdef MIGRATION_VERIFY + if (ret==0) ret=load_verify_memory(f, NULL, 1); +#endif /* MIGRATION_VERIFY */ + qemu_fclose(f); + + return ret; +} + +static int migrate_incoming_tcp(const char *host) +{ + struct sockaddr_in addr; + socklen_t addrlen = sizeof(addr); + int fd, sfd; + ssize_t len; + uint8_t status = 0; + int reuse = 1; + int rc; + + addr.sin_family = AF_INET; + if (parse_host_port(&addr, host) == -1) { + fprintf(stderr, "parse_host_port() failed for %s\n", host); + rc = 201; + goto error; + } + + fd = socket(PF_INET, SOCK_STREAM, 0); + if (fd == -1) { + perror("socket failed"); + rc = 202; + goto error; + } + + if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &reuse, sizeof(reuse)) == -1) { + perror("setsockopt() failed"); + rc = 203; + goto error_socket; + } + + if (bind(fd, (struct sockaddr *)&addr, sizeof(addr)) == -1) { + perror("bind() failed"); + rc = 204; + goto error_socket; + } + + if (listen(fd, 1) == -1) { + perror("listen() failed"); + rc = 205; + goto error_socket; + } + +again: + sfd = accept(fd, (struct sockaddr *)&addr, &addrlen); + if (sfd == -1) { + if (errno == EINTR) + goto again; + perror("accept() failed"); + rc = 206; + goto error_socket; + } + + rc = migrate_incoming_fd(sfd); + if (rc != 0) { + rc = 207; + fprintf(stderr, "migrate_incoming_fd failed (rc=%d)\n", rc); + goto error_accept; + } + +again1: + len = write(sfd, &status, 1); + if (len == -1 && errno == EAGAIN) + goto again1; + if (len != 1) { + rc = 208; + goto error_accept; + + } + +error_accept: + close(sfd); +error_socket: + close(fd); +error: + return rc; +} + +int migrate_incoming(const char *device) +{ + const char *ptr; + int ret = 0; + + if (strcmp(device, "stdio") == 0) + ret = migrate_incoming_fd(STDIN_FILENO); + else if (strstart(device, "tcp://", &ptr)) { + char *host, *end; + host = strdup(ptr); + end = strchr(host, '/'); + if (end) *end = 0; + ret = migrate_incoming_tcp(host); + qemu_free(host); + } else { + errno = EINVAL; + ret = -1; + } + + return ret; +} + +/* Migration monitor command */ + +/* TODO: + 1) audit all error paths +*/ + +void do_migrate(int detach, const char *uri) +{ + const char *ptr; + + if (strstart(uri, "exec:", &ptr)) { + char *command = urldecode(ptr); + migration_init_exec(detach, command); + free(command); + } else if (strstart(uri, "ssh://", &ptr)) { + char *host, *end; + + host = strdup(ptr); + end = strchr(host, '/'); + if (end) *end = 0; + migration_init_ssh(detach, host); + qemu_free(host); + } else if (strstart(uri, "tcp://", &ptr)) { + char *host, *end; + + host = strdup(ptr); + end = strchr(host, '/'); + if (end) *end = 0; + + if (migration_init_tcp(detach, host) == NULL) + term_printf("migration failed (migration_init_tcp for %s failed)\n", host); + free(host); + } else { + term_printf("Unknown migration protocol '%s'\n", uri); + return; + } +} + +void do_migrate_set_speed(const char *value) +{ + double d; + char *ptr; + + d = strtod(value, &ptr); + switch (*ptr) { + case 'G': case 'g': + d *= 1024; + case 'M': case 'm': + d *= 1024; + case 'K': case 'k': + d *= 1024; + default: + break; + } + + max_throttle = (uint32_t)d; +} + +void do_info_migration(void) +{ + MigrationState *s = current_migration; + + if (s) { + term_printf("Migration active\n"); + if (s->bps < (1 << 20)) + term_printf("Transfer rate %3.1f kb/s\n", + (double)s->bps / 1024); + else + term_printf("Transfer rate %3.1f mb/s\n", + (double)s->bps / (1024 * 1024)); + term_printf("Iteration %d\n", s->iteration); + term_printf("Transferred %d/%d pages\n", s->updated_pages, phys_ram_size >> TARGET_PAGE_BITS); + if (s->iteration) + term_printf("Last iteration found %d dirty pages\n", s->last_updated_pages); + } else + term_printf("Migration inactive\n"); + + term_printf("Maximum migration speed is "); + if (max_throttle < (1 << 20)) + term_printf("%3.1f kb/s\n", (double)max_throttle / 1024); + else + term_printf("%3.1f mb/s\n", (double)max_throttle / (1024 * 1024)); +} + +void do_migrate_cancel(void) +{ + MigrationState *s = current_migration; + + if (s) + *s->has_error = 20; +} + + + +#ifdef MIGRATION_VERIFY +unsigned int calc_page_checksum(target_ulong addr) +{ + unsigned int sum=0; + unsigned int *p = (unsigned int *)(phys_ram_base + addr); + unsigned int *q = p + (TARGET_PAGE_SIZE / sizeof(unsigned int)); + + for ( /*initialized already */ ; p<q ; p++) + sum += *p; + return sum; +} + + +static int save_verify_memory(QEMUFile *f, void *opaque) +{ + unsigned int addr; + unsigned int sum; + + for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) { + sum = calc_page_checksum(addr); + qemu_put_be32(f, addr); + qemu_put_be32(f, sum); + } + return 0; +} + +static int load_verify_memory(QEMUFile *f, void *opaque, int version_id) +{ + unsigned int addr, raddr; + unsigned int sum, rsum; + int num_errors = 0; + + for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) { + sum = calc_page_checksum(addr); + raddr = qemu_get_be32(f); + rsum = qemu_get_be32(f); + if ((raddr != addr) || (rsum != sum)) { + term_printf("checksum mismatch: src:0x%x 0x%x , dst:0x%x 0x%x\n", + raddr, rsum, addr, sum); + num_errors++; + } + } + printf("memory_verify: num_errors=%d\n", num_errors); + term_printf("memory_verify: num_errors=%d\n", num_errors); + return 0/* num_errors */; +} +#endif /* MIGRATION_VERIFY */ --- qemu-0.9.0/monitor.c +++ qemu-0.9.0/monitor.c @@ -1252,7 +1252,13 @@ { "stopcapture", "i", do_stop_capture, "capture index", "stop capture" }, { "memsave", "lis", do_memory_save, - "addr size file", "save to disk virtual memory dump starting at 'addr' of size 'size'", }, + "addr size file", "save to disk virtual memory dump starting at 'addr' of size 'size'" }, + { "migrate", "-ds", do_migrate, + "[-d] command", "migrate the VM using command (use -d to not wait for command to complete)" }, + { "migrate_cancel", "", do_migrate_cancel, + "", "cancel the current VM migration" }, + { "migrate_set_speed", "s", do_migrate_set_speed, + "value", "set maximum speed (in bytes) for migrations" }, { NULL, NULL, }, }; @@ -1299,6 +1305,8 @@ "", "show which guest mouse is receiving events" }, { "vnc", "", do_info_vnc, "", "show the vnc server status"}, + { "migration", "", do_info_migration, + "", "show migration information" }, { NULL, NULL, }, }; @@ -2428,12 +2436,26 @@ readline_handle_byte(buf[i]); } +static int monitor_suspended; + +void monitor_suspend(void) +{ + monitor_suspended = 1; +} + +void monitor_resume(void) +{ + monitor_suspended = 0; + monitor_start_input(); +} + static void monitor_start_input(void); static void monitor_handle_command1(void *opaque, const char *cmdline) { monitor_handle_command(cmdline); - monitor_start_input(); + if (!monitor_suspended) + monitor_start_input(); } static void monitor_start_input(void) --- qemu-0.9.0/qemu_socket.h +++ qemu-0.9.0/qemu_socket.h @@ -28,4 +28,6 @@ void socket_set_nonblock(int fd); +int parse_host_port(struct sockaddr_in *saddr, const char *str); + #endif /* QEMU_SOCKET_H */ --- qemu-0.9.0/vl.c +++ qemu-0.9.0/vl.c @@ -169,6 +169,7 @@ int fd_bootchk = 1; int no_reboot = 0; int daemonize = 0; +const char *incoming; const char *option_rom[MAX_OPTION_ROMS]; int nb_option_roms; int semihosting_enabled = 0; @@ -2301,7 +2302,6 @@ } } -int parse_host_port(struct sockaddr_in *saddr, const char *str); #ifndef _WIN32 static int parse_unix_path(struct sockaddr_un *uaddr, const char *str); #endif @@ -4281,17 +4281,55 @@ } #endif +#define SELF_ANNOUNCE_ROUNDS 5 +#define ETH_P_EXPERIMENTAL 0x01F1 /* just a number */ +//#define ETH_P_EXPERIMENTAL 0x0012 /* make it the size of the packet */ +#define EXPERIMENTAL_MAGIC 0xf1f23f4f + +static int announce_self_create(uint8_t *buf, + uint8_t *mac_addr) +{ + uint32_t magic = EXPERIMENTAL_MAGIC; + uint16_t proto = htons(ETH_P_EXPERIMENTAL); + + /* FIXME: should we send a different packet (arp/rarp/ping)? */ + + memset(buf, 0xff, 6); /* h_dst */ + memcpy(buf + 6, mac_addr, 6); /* h_src */ + memcpy(buf + 12, &proto, 2); /* h_proto */ + memcpy(buf + 14, &magic, 4); /* magic */ + + return 18; /* len */ +} + +static void qemu_announce_self(void) +{ + int i, j, len; + VLANState *vlan; + VLANClientState *vc; + uint8_t buf[256]; + + for (i = 0; i < nb_nics; i++) { + len = announce_self_create(buf, nd_table[i].macaddr); + vlan = nd_table[i].vlan; + for(vc = vlan->first_client; vc != NULL; vc = vc->next) { + for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++) + vc->fd_read(vc->opaque, buf, len); + } + } +} + /***********************************************************/ /* savevm/loadvm support */ #define IO_BUF_SIZE 32768 struct QEMUFile { - FILE *outfile; - BlockDriverState *bs; - int is_file; - int is_writable; - int64_t base_offset; + QEMUFilePutBufferFunc *put_buffer; + QEMUFileGetBufferFunc *get_buffer; + QEMUFileCloseFunc *close; + void *opaque; + int64_t buf_offset; /* start of buffer when writing, end of buffer when reading */ int buf_index; @@ -4299,58 +4337,143 @@ uint8_t buf[IO_BUF_SIZE]; }; -QEMUFile *qemu_fopen(const char *filename, const char *mode) +typedef struct QEMUFileFD { - QEMUFile *f; + int fd; +} QEMUFileFD; - f = qemu_mallocz(sizeof(QEMUFile)); - if (!f) - return NULL; - if (!strcmp(mode, "wb")) { - f->is_writable = 1; - } else if (!strcmp(mode, "rb")) { - f->is_writable = 0; - } else { - goto fail; +static int fd_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size) +{ + QEMUFileFD *s = opaque; + int offset = 0; + ssize_t len; + +again: + len = read(s->fd, buf + offset, size - offset); + if (len == -1) { + if (errno == EINTR || errno == EAGAIN) + goto again; } - f->outfile = fopen(filename, mode); - if (!f->outfile) + + return len; +} + +QEMUFile *qemu_fopen_fd(int fd) +{ + QEMUFileFD *s = qemu_mallocz(sizeof(QEMUFileFD)); + s->fd = fd; + return qemu_fopen(s, NULL, fd_get_buffer, qemu_free); +} + +typedef struct QEMUFileUnix +{ + FILE *outfile; +} QEMUFileUnix; + +static void file_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size) +{ + QEMUFileUnix *s = opaque; + fseek(s->outfile, pos, SEEK_SET); + fwrite(buf, 1, size, s->outfile); +} + +static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size) +{ + QEMUFileUnix *s = opaque; + fseek(s->outfile, pos, SEEK_SET); + return fread(buf, 1, size, s->outfile); +} + +static void file_close(void *opaque) +{ + QEMUFileUnix *s = opaque; + fclose(s->outfile); + qemu_free(s); +} + +QEMUFile *qemu_fopen_file(const char *filename, const char *mode) +{ + QEMUFileUnix *s; + + s = qemu_mallocz(sizeof(QEMUFileUnix)); + if (!s) + return NULL; + + s->outfile = fopen(filename, mode); + if (!s->outfile) goto fail; - f->is_file = 1; - return f; - fail: - if (f->outfile) - fclose(f->outfile); - qemu_free(f); + + if (!strcmp(mode, "wb")) + return qemu_fopen(s, file_put_buffer, NULL, file_close); + else if (!strcmp(mode, "rb")) + return qemu_fopen(s, NULL, file_get_buffer, file_close); + +fail: + if (s->outfile) + fclose(s->outfile); + qemu_free(s); return NULL; } +typedef struct QEMUFileBdrv +{ + BlockDriverState *bs; + int64_t base_offset; +} QEMUFileBdrv; + +static void bdrv_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size) +{ + QEMUFileBdrv *s = opaque; + bdrv_pwrite(s->bs, s->base_offset + pos, buf, size); +} + +static int bdrv_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size) +{ + QEMUFileBdrv *s = opaque; + return bdrv_pread(s->bs, s->base_offset + pos, buf, size); +} + QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable) { + QEMUFileBdrv *s; + + s = qemu_mallocz(sizeof(QEMUFileBdrv)); + if (!s) + return NULL; + + s->bs = bs; + s->base_offset = offset; + + if (is_writable) + return qemu_fopen(s, bdrv_put_buffer, NULL, qemu_free); + + return qemu_fopen(s, NULL, bdrv_get_buffer, qemu_free); +} + +QEMUFile *qemu_fopen(void *opaque, QEMUFilePutBufferFunc *put_buffer, + QEMUFileGetBufferFunc *get_buffer, QEMUFileCloseFunc *close) +{ QEMUFile *f; f = qemu_mallocz(sizeof(QEMUFile)); if (!f) - return NULL; - f->is_file = 0; - f->bs = bs; - f->is_writable = is_writable; - f->base_offset = offset; + return NULL; + + f->opaque = opaque; + f->put_buffer = put_buffer; + f->get_buffer = get_buffer; + f->close = close; + return f; } void qemu_fflush(QEMUFile *f) { - if (!f->is_writable) + if (!f->put_buffer) return; + if (f->buf_index > 0) { - if (f->is_file) { - fseek(f->outfile, f->buf_offset, SEEK_SET); - fwrite(f->buf, 1, f->buf_index, f->outfile); - } else { - bdrv_pwrite(f->bs, f->base_offset + f->buf_offset, - f->buf, f->buf_index); - } + f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index); f->buf_offset += f->buf_index; f->buf_index = 0; } @@ -4360,19 +4483,13 @@ { int len; - if (f->is_writable) + if (!f->get_buffer) return; - if (f->is_file) { - fseek(f->outfile, f->buf_offset, SEEK_SET); - len = fread(f->buf, 1, IO_BUF_SIZE, f->outfile); - if (len < 0) - len = 0; - } else { - len = bdrv_pread(f->bs, f->base_offset + f->buf_offset, - f->buf, IO_BUF_SIZE); - if (len < 0) - len = 0; - } + + len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE); + if (len < 0) + len = 0; + f->buf_index = 0; f->buf_size = len; f->buf_offset += len; @@ -4380,11 +4497,9 @@ void qemu_fclose(QEMUFile *f) { - if (f->is_writable) - qemu_fflush(f); - if (f->is_file) { - fclose(f->outfile); - } + qemu_fflush(f); + if (f->close) + f->close(f->opaque); qemu_free(f); } @@ -4459,7 +4574,7 @@ /* SEEK_END not supported */ return -1; } - if (f->is_writable) { + if (f->put_buffer) { qemu_fflush(f); f->buf_offset = pos; } else { @@ -4659,8 +4774,78 @@ } /* always seek to exact end of record */ qemu_fseek(f, cur_pos + record_len, SEEK_SET); + } + ret = 0; + the_end: + return ret; +} + +int qemu_live_savevm_state(QEMUFile *f) +{ + SaveStateEntry *se; + int len, ret; + + qemu_put_be32(f, QEMU_VM_FILE_MAGIC); + qemu_put_be32(f, QEMU_VM_FILE_VERSION); + + for(se = first_se; se != NULL; se = se->next) { + len = strlen(se->idstr); + + qemu_put_byte(f, len); + qemu_put_buffer(f, se->idstr, len); + qemu_put_be32(f, se->instance_id); + qemu_put_be32(f, se->version_id); + + se->save_state(f, se->opaque); + } + + qemu_put_byte(f, 0); + + ret = 0; + return ret; +} + +int qemu_live_loadvm_state(QEMUFile *f) +{ + SaveStateEntry *se; + int len, ret, instance_id, version_id; + unsigned int v; + char idstr[256]; + + v = qemu_get_be32(f); + if (v != QEMU_VM_FILE_MAGIC) + goto fail; + v = qemu_get_be32(f); + if (v != QEMU_VM_FILE_VERSION) { + fail: + ret = -1; + goto the_end; + } + + for(;;) { + len = qemu_get_byte(f); + if (len == 0) + break; + qemu_get_buffer(f, idstr, len); + idstr[len] = '\0'; + instance_id = qemu_get_be32(f); + version_id = qemu_get_be32(f); + se = find_se(idstr, instance_id); + if (!se) { + fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n", + instance_id, idstr); + } else { + ret = se->load_state(f, se->opaque, version_id); + if (ret < 0) { + fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n", + instance_id, idstr); + } + } } ret = 0; + + qemu_announce_self(); + the_end: return ret; } @@ -4980,7 +5165,12 @@ uint16_t fptag, fpus, fpuc, fpregs_format; uint32_t hflags; int i; - + +#ifdef USE_KVM + if (kvm_allowed) + kvm_save_registers(env); +#endif + for(i = 0; i < CPU_NB_REGS; i++) qemu_put_betls(f, &env->regs[i]); qemu_put_betls(f, &env->eip); @@ -5065,6 +5255,16 @@ qemu_put_be64s(f, &env->kernelgsbase); #endif qemu_put_be32s(f, &env->smbase); + +#ifdef USE_KVM + if (kvm_allowed) { + for (i = 0; i < NR_IRQ_WORDS ; i++) { + qemu_put_betls(f, &env->kvm_interrupt_bitmap[i]); + } + qemu_put_be64s(f, &env->tsc); + } +#endif + } #ifdef USE_X86LDOUBLE @@ -5484,7 +5684,20 @@ inflateEnd(&s->zstream); } -static void ram_save(QEMUFile *f, void *opaque) +static void ram_save_live(QEMUFile *f, void *opaque) +{ + target_ulong addr; + + for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) { + if (cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG)) { + qemu_put_be32(f, addr); + qemu_put_buffer(f, phys_ram_base + addr, TARGET_PAGE_SIZE); + } + } + qemu_put_be32(f, 1); +} + +static void ram_save_static(QEMUFile *f, void *opaque) { int i; RamCompressState s1, *s = &s1; @@ -5528,16 +5741,39 @@ ram_compress_close(s); } -static int ram_load(QEMUFile *f, void *opaque, int version_id) +static void ram_save(QEMUFile *f, void *opaque) +{ + int in_migration = cpu_physical_memory_get_dirty_tracking(); + + qemu_put_byte(f, in_migration); + + if (in_migration) + ram_save_live(f, opaque); + else + ram_save_static(f, opaque); +} + +static int ram_load_live(QEMUFile *f, void *opaque) +{ + target_ulong addr; + + do { + addr = qemu_get_be32(f); + if (addr == 1) + break; + + qemu_get_buffer(f, phys_ram_base + addr, TARGET_PAGE_SIZE); + } while (1); + + return 0; +} + +static int ram_load_static(QEMUFile *f, void *opaque) { RamDecompressState s1, *s = &s1; uint8_t buf[10]; int i; - if (version_id == 1) - return ram_load_v1(f, opaque); - if (version_id != 2) - return -EINVAL; if (qemu_get_be32(f) != phys_ram_size) return -EINVAL; if (ram_decompress_open(s, f) < 0) @@ -5583,6 +5819,30 @@ return 0; } +static int ram_load(QEMUFile *f, void *opaque, int version_id) +{ + int ret; + + switch (version_id) { + case 1: + ret = ram_load_v1(f, opaque); + break; + case 3: + if (qemu_get_byte(f)) { + ret = ram_load_live(f, opaque); + break; + } + case 2: + ret = ram_load_static(f, opaque); + break; + default: + ret = -EINVAL; + break; + } + + return ret; +} + /***********************************************************/ /* bottom halves (can be seen as timers which expire ASAP) */ @@ -6205,7 +6465,8 @@ QEMU_OPTION_no_reboot, QEMU_OPTION_daemonize, QEMU_OPTION_option_rom, - QEMU_OPTION_semihosting + QEMU_OPTION_semihosting, + QEMU_OPTION_incoming, }; typedef struct QEMUOption { @@ -6272,6 +6533,7 @@ { "serial", 1, QEMU_OPTION_serial }, { "parallel", 1, QEMU_OPTION_parallel }, { "loadvm", HAS_ARG, QEMU_OPTION_loadvm }, + { "incoming", 1, QEMU_OPTION_incoming }, { "full-screen", 0, QEMU_OPTION_full_screen }, #ifdef CONFIG_SDL { "no-quit", 0, QEMU_OPTION_no_quit }, @@ -6496,6 +6758,17 @@ #define MAX_NET_CLIENTS 32 +static int saved_argc; +static char **saved_argv; + +void qemu_get_launch_info(int *argc, char ***argv, int *opt_daemonize, const char **opt_incoming) +{ + *argc = saved_argc; + *argv = saved_argv; + *opt_daemonize = daemonize; + *opt_incoming = incoming; +} + int main(int argc, char **argv) { #ifdef CONFIG_GDBSTUB @@ -6524,6 +6797,9 @@ int usb_devices_index; int fds[2]; + saved_argc = argc; + saved_argv = argv; + LIST_INIT (&vm_change_state_head); #ifndef _WIN32 { @@ -6890,6 +7166,9 @@ case QEMU_OPTION_loadvm: loadvm = optarg; break; + case QEMU_OPTION_incoming: + incoming = optarg; + break; case QEMU_OPTION_full_screen: full_screen = 1; break; @@ -6963,11 +7242,6 @@ } #ifndef _WIN32 - if (daemonize && !nographic && vnc_display == NULL) { - fprintf(stderr, "Can only daemonize if using -nographic or -vnc\n"); - daemonize = 0; - } - if (daemonize) { pid_t pid; @@ -7002,7 +7276,6 @@ exit(1); umask(027); - chdir("/"); signal(SIGTSTP, SIG_IGN); signal(SIGTTOU, SIG_IGN); @@ -7146,7 +7419,7 @@ } register_savevm("timer", 0, 2, timer_save, timer_load, NULL); - register_savevm("ram", 0, 2, ram_save, ram_load, NULL); + register_savevm("ram", 0, 3, ram_save, ram_load, NULL); init_ioports(); @@ -7228,8 +7501,19 @@ } } else #endif - if (loadvm) - do_loadvm(loadvm); + if (loadvm) { + do_loadvm(loadvm); + } + + if (incoming) { + int rc; + + rc = migrate_incoming(incoming); + if (rc != 0) { + fprintf(stderr, "Migration failed rc=%d\n", rc); + exit(rc); + } + } { /* XXX: simplify init */ @@ -7252,6 +7536,7 @@ if (len != 1) exit(1); + chdir("/"); fd = open("/dev/null", O_RDWR); if (fd == -1) exit(1); --- qemu-0.9.0/vl.h +++ qemu-0.9.0/vl.h @@ -105,8 +105,12 @@ char *pstrcat(char *buf, int buf_size, const char *s); int strstart(const char *str, const char *val, const char **ptr); int stristart(const char *str, const char *val, const char **ptr); +int hex2bin(char ch); +char *urldecode(const char *ptr); /* vl.c */ +void qemu_get_launch_info(int *argc, char ***argv, int *opt_daemonize, const char **opt_incoming); + uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c); void hw_error(const char *fmt, ...); @@ -434,7 +438,14 @@ typedef struct QEMUFile QEMUFile; -QEMUFile *qemu_fopen(const char *filename, const char *mode); +typedef void (QEMUFilePutBufferFunc)(void *opaque, const uint8_t *buf, int64_t pos, int size); +typedef int (QEMUFileGetBufferFunc)(void *opaque, uint8_t *buf, int64_t pos, int size); +typedef void (QEMUFileCloseFunc)(void *opaque); + +QEMUFile *qemu_fopen(void *opaque, QEMUFilePutBufferFunc *put_buffer, + QEMUFileGetBufferFunc *get_buffer, QEMUFileCloseFunc *close); +QEMUFile *qemu_fopen_file(const char *filename, const char *mode); +QEMUFile *qemu_fopen_fd(int fd); void qemu_fflush(QEMUFile *f); void qemu_fclose(QEMUFile *f); void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size); @@ -523,6 +534,9 @@ void do_delvm(const char *name); void do_info_snapshots(void); +int qemu_live_savevm_state(QEMUFile *f); +int qemu_live_loadvm_state(QEMUFile *f); + /* bottom halves */ typedef void QEMUBHFunc(void *opaque); @@ -1373,6 +1387,13 @@ #endif /* defined(QEMU_TOOL) */ +/* migration.c */ +void do_info_migration(void); +void do_migrate(int detach, const char *uri); +void do_migrate_cancel(void); +void do_migrate_set_speed(const char *value); +int migrate_incoming(const char *device); + /* monitor.c */ void monitor_init(CharDriverState *hd, int show_banner); void term_puts(const char *str); @@ -1383,6 +1404,8 @@ void term_print_help(void); void monitor_readline(const char *prompt, int is_password, char *buf, int buf_size); +void monitor_suspend(void); +void monitor_resume(void); /* readline.c */ typedef void ReadLineFunc(void *opaque, const char *str);
