NuKVM/libvirt
3
0
Fork 0
mirror of https://github.com/libvirt/libvirt.git synced 2025-02-15 01:43:40 -06:00
Code Issues Packages Projects Releases Wiki Activity

Fix QEMU save/restore with block devices

Browse Source 
The save process was relying on use of the shell >> append
operator to ensure the save data was placed after the libvirt
header + XML. This doesn't work for block devices though.
Replace this code with use of 'dd' and its 'seek' parameter.
This means that we need to pad the header + XML out to a
multiple of dd block size (in this case we choose 512).

The qemuMonitorMigateToCommand() monitor API is used for both
save/coredump, and migration via UNIX socket. We can't simply
switch this to use 'dd' since this causes problems with the
migration usage. Thus, create a dedicated qemuMonitorMigateToFile
which can accept an filename + offset, and remove the filename
from the current qemuMonitorMigateToCommand() API

* src/qemu/qemu_driver.c: Switch to qemuMonitorMigateToFile
  for save and core dump
* src/qemu/qemu_monitor.c, src/qemu/qemu_monitor.h,
  src/qemu/qemu_monitor_json.c, src/qemu/qemu_monitor_json.h,
  src/qemu/qemu_monitor_text.c, src/qemu/qemu_monitor_text.h: Create
  a new qemuMonitorMigateToFile, separate from the existing
  qemuMonitorMigateToCommand to allow handling file offsets
This commit is contained in:
Daniel P. Berrange 2010-04-21 14:06:37 +01:00
parent ae42979a74
commit 93e0b3c8d6
7 changed files with 235 additions and 99 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

196
src/qemu/qemu_driver.c
View File

@ -4715,6 +4715,7 @@ static int qemudDomainSaveFlag(virDomainPtr dom, const char *path,
qemuDomainObjPrivatePtr priv;
struct stat sb;
int is_reg = 0;
unsigned long long offset;
memset(&header, 0, sizeof(header));
memcpy(header.magic, QEMUD_SAVE_MAGIC, sizeof(header.magic));
@ -4788,104 +4789,137 @@ static int qemudDomainSaveFlag(virDomainPtr dom, const char *path,
hdata.path = path;
hdata.xml = xml;
hdata.header = &header;
offset = sizeof(header) + header.xml_len;
/* Due to way we append QEMU state on our header with dd,
* we need to ensure there's a 512 byte boundary. Unfortunately
* we don't have an explicit offset in the header, so we fake
* it by padding the XML string with NULLs */
if (offset % QEMU_MONITOR_MIGRATE_TO_FILE_BS) {
unsigned long long pad =
QEMU_MONITOR_MIGRATE_TO_FILE_BS -
(offset % QEMU_MONITOR_MIGRATE_TO_FILE_BS);
if (VIR_REALLOC_N(xml, header.xml_len + pad) < 0) {
virReportOOMError();
goto endjob;
}
memset(xml + header.xml_len, 0, pad);
offset += pad;
header.xml_len += pad;
}
/* Write header to file, followed by XML */
/* First try creating the file as root */
if (is_reg &&
(rc = virFileOperation(path, O_CREAT|O_TRUNC|O_WRONLY,
S_IRUSR|S_IWUSR,
getuid(), getgid(),
qemudDomainSaveFileOpHook, &hdata,
0)) != 0) {
/* If we failed as root, and the error was permission-denied
(EACCES), assume it's on a network-connected share where
root access is restricted (eg, root-squashed NFS). If the
qemu user (driver->user) is non-root, just set a flag to
bypass security driver shenanigans, and retry the operation
after doing setuid to qemu user */
if ((rc != EACCES) ||
driver->user == getuid()) {
virReportSystemError(rc, _("Failed to create domain save file '%s'"),
path);
if (!is_reg) {
int fd = open(path, O_WRONLY | O_TRUNC);
if (fd < 0) {
virReportSystemError(errno, _("unable to open %s"), path);
goto endjob;
}
if ((rc = qemudDomainSaveFileOpHook(fd, &hdata)) != 0) {
close(fd);
goto endjob;
}
if (close(fd) < 0) {
virReportSystemError(errno, _("unable to close %s"), path);
goto endjob;
}
} else {
if ((rc = virFileOperation(path, O_CREAT|O_TRUNC|O_WRONLY,
S_IRUSR|S_IWUSR,
getuid(), getgid(),
qemudDomainSaveFileOpHook, &hdata,
0)) != 0) {
/* If we failed as root, and the error was permission-denied
(EACCES), assume it's on a network-connected share where
root access is restricted (eg, root-squashed NFS). If the
qemu user (driver->user) is non-root, just set a flag to
bypass security driver shenanigans, and retry the operation
after doing setuid to qemu user */
if ((rc != EACCES) ||
driver->user == getuid()) {
virReportSystemError(rc, _("Failed to create domain save file '%s'"),
path);
goto endjob;
}
#ifdef __linux__
/* On Linux we can also verify the FS-type of the directory. */
char *dirpath, *p;
struct statfs st;
int statfs_ret;
/* On Linux we can also verify the FS-type of the directory. */
char *dirpath, *p;
struct statfs st;
int statfs_ret;
if ((dirpath = strdup(path)) == NULL) {
virReportOOMError();
goto endjob;
}
if ((dirpath = strdup(path)) == NULL) {
virReportOOMError();
goto endjob;
}
do {
// Try less and less of the path until we get to a
// directory we can stat. Even if we don't have 'x'
// permission on any directory in the path on the NFS
// server (assuming it's NFS), we will be able to stat the
// mount point, and that will properly tell us if the
// fstype is NFS.
do {
// Try less and less of the path until we get to a
// directory we can stat. Even if we don't have 'x'
// permission on any directory in the path on the NFS
// server (assuming it's NFS), we will be able to stat the
// mount point, and that will properly tell us if the
// fstype is NFS.
if ((p = strrchr(dirpath, '/')) == NULL) {
qemuReportError(VIR_ERR_INVALID_ARG,
_("Invalid relative path '%s' for domain save file"),
path);
if ((p = strrchr(dirpath, '/')) == NULL) {
qemuReportError(VIR_ERR_INVALID_ARG,
_("Invalid relative path '%s' for domain save file"),
path);
VIR_FREE(dirpath);
goto endjob;
}
if (p == dirpath)
*(p+1) = '\0';
else
*p = '\0';
statfs_ret = statfs(dirpath, &st);
} while ((statfs_ret == -1) && (p != dirpath));
if (statfs_ret == -1) {
virReportSystemError(errno,
_("Failed to create domain save file '%s'"
" statfs of all elements of path failed."),
path);
VIR_FREE(dirpath);
goto endjob;
}
if (p == dirpath)
*(p+1) = '\0';
else
*p = '\0';
statfs_ret = statfs(dirpath, &st);
} while ((statfs_ret == -1) && (p != dirpath));
if (statfs_ret == -1) {
virReportSystemError(errno,
_("Failed to create domain save file '%s'"
" statfs of all elements of path failed."),
path);
if (st.f_type != NFS_SUPER_MAGIC) {
virReportSystemError(rc,
_("Failed to create domain save file '%s'"
" (fstype of '%s' is 0x%X"),
path, dirpath, (unsigned int) st.f_type);
VIR_FREE(dirpath);
goto endjob;
}
VIR_FREE(dirpath);
goto endjob;
}
if (st.f_type != NFS_SUPER_MAGIC) {
virReportSystemError(rc,
_("Failed to create domain save file '%s'"
" (fstype of '%s' is 0x%X"),
path, dirpath, (unsigned int) st.f_type);
VIR_FREE(dirpath);
goto endjob;
}
VIR_FREE(dirpath);
#endif
/* Retry creating the file as driver->user */
/* Retry creating the file as driver->user */
if ((rc = virFileOperation(path, O_CREAT|O_TRUNC|O_WRONLY,
S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP,
driver->user, driver->group,
qemudDomainSaveFileOpHook, &hdata,
VIR_FILE_OP_AS_UID)) != 0) {
virReportSystemError(rc, _("Error from child process creating '%s'"),
if ((rc = virFileOperation(path, O_CREAT|O_TRUNC|O_WRONLY,
S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP,
driver->user, driver->group,
qemudDomainSaveFileOpHook, &hdata,
VIR_FILE_OP_AS_UID)) != 0) {
virReportSystemError(rc, _("Error from child process creating '%s'"),
path);
goto endjob;
goto endjob;
}
/* Since we had to setuid to create the file, and the fstype
is NFS, we assume it's a root-squashing NFS share, and that
the security driver stuff would have failed anyway */
bypassSecurityDriver = 1;
}
/* Since we had to setuid to create the file, and the fstype
is NFS, we assume it's a root-squashing NFS share, and that
the security driver stuff would have failed anyway */
bypassSecurityDriver = 1;
}
@ -4898,7 +4932,7 @@ static int qemudDomainSaveFlag(virDomainPtr dom, const char *path,
if (header.compressed == QEMUD_SAVE_FORMAT_RAW) {
const char *args[] = { "cat", NULL };
qemuDomainObjEnterMonitorWithDriver(driver, vm);
rc = qemuMonitorMigrateToCommand(priv->mon, 1, args, path);
rc = qemuMonitorMigrateToFile(priv->mon, 1, args, path, offset);
qemuDomainObjExitMonitorWithDriver(driver, vm);
} else {
const char *prog = qemudSaveCompressionTypeToString(header.compressed);
@ -4908,7 +4942,7 @@ static int qemudDomainSaveFlag(virDomainPtr dom, const char *path,
NULL
};
qemuDomainObjEnterMonitorWithDriver(driver, vm);
rc = qemuMonitorMigrateToCommand(priv->mon, 1, args, path);
rc = qemuMonitorMigrateToFile(priv->mon, 1, args, path, offset);
qemuDomainObjExitMonitorWithDriver(driver, vm);
}
@ -5203,7 +5237,7 @@ static int qemudDomainCoreDump(virDomainPtr dom,
}
qemuDomainObjEnterMonitorWithDriver(driver, vm);
ret = qemuMonitorMigrateToCommand(priv->mon, 1, args, path);
ret = qemuMonitorMigrateToFile(priv->mon, 1, args, path, 0);
qemuDomainObjExitMonitorWithDriver(driver, vm);
if (ret < 0)
@ -9896,7 +9930,7 @@ static int doTunnelMigrate(virDomainPtr dom,
internalret = qemuMonitorMigrateToUnix(priv->mon, 1, unixfile);
else if (qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_EXEC) {
const char *args[] = { "nc", "-U", unixfile, NULL };
internalret = qemuMonitorMigrateToCommand(priv->mon, 1, args, "/dev/null");
internalret = qemuMonitorMigrateToCommand(priv->mon, 1, args);
} else {
internalret = -1;
}

35
src/qemu/qemu_monitor.c
View File

@ -1178,17 +1178,40 @@ int qemuMonitorMigrateToHost(qemuMonitorPtr mon,
int qemuMonitorMigrateToCommand(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target)
const char * const *argv)
{
int ret;
DEBUG("mon=%p, fd=%d argv=%p target=%s",
mon, mon->fd, argv, target);
DEBUG("mon=%p, fd=%d argv=%p",
mon, mon->fd, argv);
if (mon->json)
ret = qemuMonitorJSONMigrateToCommand(mon, background, argv, target);
ret = qemuMonitorJSONMigrateToCommand(mon, background, argv);
else
ret = qemuMonitorTextMigrateToCommand(mon, background, argv, target);
ret = qemuMonitorTextMigrateToCommand(mon, background, argv);
return ret;
}
int qemuMonitorMigrateToFile(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target,
unsigned long long offset)
{
int ret;
DEBUG("mon=%p, fd=%d argv=%p target=%s offset=%llu",
mon, mon->fd, argv, target, offset);
if (offset % QEMU_MONITOR_MIGRATE_TO_FILE_BS) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("file offset must be a multiple of %llu"),
QEMU_MONITOR_MIGRATE_TO_FILE_BS);
return -1;
}
if (mon->json)
ret = qemuMonitorJSONMigrateToFile(mon, background, argv, target, offset);
else
ret = qemuMonitorTextMigrateToFile(mon, background, argv, target, offset);
return ret;
}

11
src/qemu/qemu_monitor.h
View File

@ -246,8 +246,15 @@ int qemuMonitorMigrateToHost(qemuMonitorPtr mon,
int qemuMonitorMigrateToCommand(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target);
const char * const *argv);
# define QEMU_MONITOR_MIGRATE_TO_FILE_BS 512llu
int qemuMonitorMigrateToFile(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target,
unsigned long long offset);
int qemuMonitorMigrateToUnix(qemuMonitorPtr mon,
int background,

37
src/qemu/qemu_monitor_json.c
View File

@ -1483,8 +1483,36 @@ int qemuMonitorJSONMigrateToHost(qemuMonitorPtr mon,
int qemuMonitorJSONMigrateToCommand(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target)
const char * const *argv)
{
char *argstr;
char *dest = NULL;
int ret = -1;
argstr = virArgvToString(argv);
if (!argstr) {
virReportOOMError();
goto cleanup;
}
if (virAsprintf(&dest, "exec:%s", argstr) < 0) {
virReportOOMError();
goto cleanup;
}
ret = qemuMonitorJSONMigrate(mon, background, dest);
cleanup:
VIR_FREE(argstr);
VIR_FREE(dest);
return ret;
}
int qemuMonitorJSONMigrateToFile(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target,
unsigned long long offset)
{
char *argstr;
char *dest = NULL;
@ -1504,7 +1532,10 @@ int qemuMonitorJSONMigrateToCommand(qemuMonitorPtr mon,
goto cleanup;
}
if (virAsprintf(&dest, "exec:%s >>%s 2>/dev/null", argstr, safe_target) < 0) {
if (virAsprintf(&dest, "exec:%s | dd of=%s bs=%llu seek=%llu",
argstr, safe_target,
QEMU_MONITOR_MIGRATE_TO_FILE_BS,
offset / QEMU_MONITOR_MIGRATE_TO_FILE_BS) < 0) {
virReportOOMError();
goto cleanup;
}

9
src/qemu/qemu_monitor_json.h
View File

@ -100,8 +100,13 @@ int qemuMonitorJSONMigrateToHost(qemuMonitorPtr mon,
int qemuMonitorJSONMigrateToCommand(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target);
const char * const *argv);
int qemuMonitorJSONMigrateToFile(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target,
unsigned long long offset);
int qemuMonitorJSONMigrateToUnix(qemuMonitorPtr mon,
int background,

37
src/qemu/qemu_monitor_text.c
View File

@ -1202,8 +1202,36 @@ int qemuMonitorTextMigrateToHost(qemuMonitorPtr mon,
int qemuMonitorTextMigrateToCommand(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target)
const char * const *argv)
{
char *argstr;
char *dest = NULL;
int ret = -1;
argstr = virArgvToString(argv);
if (!argstr) {
virReportOOMError();
goto cleanup;
}
if (virAsprintf(&dest, "exec:%s", argstr) < 0) {
virReportOOMError();
goto cleanup;
}
ret = qemuMonitorTextMigrate(mon, background, dest);
cleanup:
VIR_FREE(argstr);
VIR_FREE(dest);
return ret;
}
int qemuMonitorTextMigrateToFile(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target,
unsigned long long offset)
{
char *argstr;
char *dest = NULL;
@ -1223,7 +1251,10 @@ int qemuMonitorTextMigrateToCommand(qemuMonitorPtr mon,
goto cleanup;
}
if (virAsprintf(&dest, "exec:%s >>%s 2>/dev/null", argstr, safe_target) < 0) {
if (virAsprintf(&dest, "exec:%s | dd of=%s bs=%llu seek=%llu",
argstr, safe_target,
QEMU_MONITOR_MIGRATE_TO_FILE_BS,
offset / QEMU_MONITOR_MIGRATE_TO_FILE_BS) < 0) {
virReportOOMError();
goto cleanup;
}

9
src/qemu/qemu_monitor_text.h
View File

@ -99,8 +99,13 @@ int qemuMonitorTextMigrateToHost(qemuMonitorPtr mon,
int qemuMonitorTextMigrateToCommand(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target);
const char * const *argv);
int qemuMonitorTextMigrateToFile(qemuMonitorPtr mon,
int background,
const char * const *argv,
const char *target,
unsigned long long offset);
int qemuMonitorTextMigrateToUnix(qemuMonitorPtr mon,
int background,