proxmox-mirrors/mirror_lxc
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_lxc synced 2025-10-31 09:54:42 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

cgroups: don't mount under init's cgroup

Browse Source 
1. deeper hierarchy has steep performance costs
2. init may be under /init, but containers should be under /lxc
3. in a nested container we like to bind-mount $cgroup_path/$c/$c.real
   into $cgroup_path - but task 1's cgroup is $c/$c.real, so a nested
   container would be in $c/$c.real/lxc, which would become
   /$c/$c.real/$c/$c.real/lxc when expanded
4. this pulls quite a bit of code (of mine) which is always nice

Signed-off-by: Serge Hallyn <serge.hallyn@ubuntu.com>
This commit is contained in:
Serge Hallyn 2013-03-12 21:34:26 -05:00
parent fc7de5616f
commit 23622a2ac8
2 changed files with 23 additions and 159 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
hooks/mountcgroups
View File

@ -34,6 +34,8 @@ fi
mount -n -t tmpfs tmpfs ${d2}
# XXX TODO - we'll need to account for other cgroup groups beside 'lxc',
# i.e. 'build' or 'users/joe'.
for dir in `/bin/ls $d`; do
mkdir -p "${d}/${dir}/lxc/${c}/${c}.real"
echo 1 > "${d}/${dir}/lxc/${c}/${c}.real/tasks"

180
src/lxc/cgroup.c
View File

@ -88,76 +88,7 @@ static char *mount_has_subsystem(const struct mntent *mntent)
}
/*
* get_init_cgroup: get the cgroup init is in.
* @subsystem: the exact cgroup subsystem to look up (I.e. "freezer")
* @mntent: a mntent (from getmntent) whose mntopts contains the subsystem to
* look up.
* @dsg: preallocated buffer of at least size MAXPATHLEN in which the path will
* be copied.
* @prependslash: if 1, the path will have a '/' prepended for easy of use by
* the caller.
*
* subsystem and mntent can both be NULL, in which case we return
* the first entry in /proc/1/cgroup.
*
* Returns a pointer to the answer (which is just the passed-in @dsg), which
* may be "".
*/
static char *get_init_cgroup(const char *subsystem, struct mntent *mntent,
char *dsg, int prependslash)
{
FILE *f;
char *c, *c2;
char line[MAXPATHLEN];
int ret;
*dsg = '\0';
f = fopen("/proc/1/cgroup", "r");
if (!f)
return dsg;
while (fgets(line, MAXPATHLEN, f)) {
c = index(line, ':');
if (!c)
continue;
c++;
c2 = index(c, ':');
if (!c2)
continue;
*c2 = '\0';
c2++;
if (!subsystem && !mntent)
goto good;
if (subsystem && strcmp(c, subsystem) != 0)
continue;
if (mntent && !hasmntopt(mntent, c))
continue;
good:
DEBUG("get_init_cgroup: found init cgroup for subsys %s at %s\n",
subsystem, c2);
ret = snprintf(dsg, MAXPATHLEN, "%s%s", prependslash ? "/" : "", c2);
if (ret < 0 || ret >= MAXPATHLEN) {
WARN("init cgroup path name was too long.");
goto found;
}
if (ret < 1)
goto found;
c = &dsg[ret-1];
if (*c == '\n')
*c = '\0';
goto found;
}
found:
fclose(f);
return dsg;
}
/*
* Determine mountpoint for a cgroup subsystem, plus the cgroup path in that
* subsytem of the container init.
* Determine mountpoint for a cgroup subsystem.
* @subsystem: cgroup subsystem (i.e. freezer). If this is NULL, the first
* cgroup mountpoint with any subsystems is used.
* @mnt: a passed-in buffer of at least size MAXPATHLEN into which the path
@ -168,7 +99,6 @@ found:
static int get_cgroup_mount(const char *subsystem, char *mnt)
{
struct mntent *mntent;
char initcgroup[MAXPATHLEN], *init;
FILE *file = NULL;
int ret, err = -1;
@ -190,9 +120,7 @@ static int get_cgroup_mount(const char *subsystem, char *mnt)
continue;
}
init = get_init_cgroup(NULL, mntent, initcgroup, 1);
ret = snprintf(mnt, MAXPATHLEN, "%s%s", mntent->mnt_dir,
init);
ret = snprintf(mnt, MAXPATHLEN, "%s", mntent->mnt_dir);
if (ret < 0 || ret >= MAXPATHLEN)
goto fail;
@ -504,59 +432,6 @@ int lxc_cgroup_nrtasks(const char *cgpath)
return count;
}
/*
* Set of helper functions to make sure that, when we create a new
* cpuset cgroup, its cpus and mems files inherit the values in the
* parent cgroup
*/
static long get_value(const char *dir, const char *file)
{
FILE *f;
char path[MAXPATHLEN];
int ret, retv;
retv = snprintf(path, MAXPATHLEN, "%s/%s", dir, file);
if (retv < 0 || retv >= MAXPATHLEN)
return 0;
f = fopen(path, "r");
ret = fscanf(f, "%d", &retv);
fclose(f);
if (ret != 1)
return 0;
return retv;
}
static void set_value(const char *dir, const char *file, long v)
{
FILE *f;
char path[MAXPATHLEN];
int retv;
retv = snprintf(path, MAXPATHLEN, "%s/%s", dir, file);
if (retv < 0 || retv >= MAXPATHLEN)
return;
f = fopen(path, "w");
fprintf(f, "%ld\n", v);
fclose(f);
}
static void setup_cpuset(const char *path)
{
/* copy parent values for mems_allowed and cpus_allowed */
char *parentpath = strdup(path);
char *p;
long v;
if ((p = rindex(parentpath, '/')) == NULL)
goto out;
v = get_value(parentpath, "cpuset.mems");
set_value(path, "cpuset.mems", v);
v = get_value(parentpath, "cpuset.cpus");
set_value(path, "cpuset.cpus", v);
out:
free(parentpath);
}
/*
* If first creating the /sys/fs/cgroup/$subsys/lxc container, then
* try to set clone_children to 1. Some kernels don't support
@ -565,13 +440,13 @@ out:
* hooks/mountcgroup) check if cpuset is in the subsystems, and if so
* manually copy over mems and cpus.
*/
static void set_clone_children(const char *mntdir, const char *init)
static void set_clone_children(const char *mntdir)
{
char path[MAXPATHLEN];
FILE *fout;
int ret;
ret = snprintf(path, MAXPATHLEN, "%s%s/cgroup.clone_children", mntdir, init);
ret = snprintf(path, MAXPATHLEN, "%s/cgroup.clone_children", mntdir);
INFO("writing to %s\n", path);
if (ret < 0 || ret > MAXPATHLEN)
return;
@ -597,7 +472,7 @@ static int create_lxcgroups(const char *lxcgroup)
FILE *file = NULL;
struct mntent *mntent;
int ret, retv = -1;
char path[MAXPATHLEN], initpath[MAXPATHLEN];
char path[MAXPATHLEN];
file = setmntent(MTAB, "r");
if (!file) {
@ -606,7 +481,6 @@ static int create_lxcgroups(const char *lxcgroup)
}
while ((mntent = getmntent(file))) {
char *init = get_init_cgroup(NULL, mntent, initpath, 1);
if (strcmp(mntent->mnt_type, "cgroup"))
continue;
@ -615,21 +489,19 @@ static int create_lxcgroups(const char *lxcgroup)
/*
* TODO - handle case where lxcgroup has subdirs? (i.e. build/l1)
* May not be worthwhile - remember cgroup depth has perf penalties
* */
ret = snprintf(path, MAXPATHLEN, "%s%s/%s",
mntent->mnt_dir, init, lxcgroup ? lxcgroup : "lxc");
* We probably only want to support that for /users/joe
*/
ret = snprintf(path, MAXPATHLEN, "%s/%s",
mntent->mnt_dir, lxcgroup ? lxcgroup : "lxc");
if (ret < 0 || ret >= MAXPATHLEN)
goto fail;
if (access(path, F_OK)) {
set_clone_children(mntent->mnt_dir, init);
set_clone_children(mntent->mnt_dir);
ret = mkdir(path, 0755);
if (ret == -1 && errno != EEXIST) {
SYSERROR("failed to create '%s' directory", path);
goto fail;
}
} else if (hasmntopt(mntent, "cpuset")) {
setup_cpuset(path);
}
}
@ -656,12 +528,9 @@ fail:
* free that eventually, however the lxc monitor will keep that string so as to
* return it in response to a LXC_COMMAND_CGROUP query.
*
* Note the path is relative to cgroup mounts plus the caller's init task's
* cgroup. I.e. if init is in cgroup /init and the freezer subsystem is at
* /sys/fs/cgroup/freezer, and this fn returns '/lxc/r1', then the freezer
* cgroup's full path will be /sys/fs/cgroup/freezer/init/lxc/r1/. Note also
* that this should cleanly account for init being in different cgroups for
* different subsystems.
* Note the path is relative to cgroup mounts. I.e. if the freezer subsystem
* is at /sys/fs/cgroup/freezer, and this fn returns '/lxc/r1', then the
* freezer cgroup's full path will be /sys/fs/cgroup/freezer/lxc/r1/.
*
* XXX This should probably be locked globally
*
@ -670,7 +539,7 @@ fail:
char *lxc_cgroup_path_create(const char *lxcgroup, const char *name)
{
int i = 0, ret;
char *retpath, path[MAXPATHLEN], initpath[MAXPATHLEN], *init;
char *retpath, path[MAXPATHLEN];
char tail[12];
FILE *file = NULL;
struct mntent *mntent;
@ -697,10 +566,8 @@ again:
if (!mount_has_subsystem(mntent))
continue;
init = get_init_cgroup(NULL, mntent, initpath, 1);
/* find unused mnt_dir + init_cgroup + lxcgroup + name + -$i */
ret = snprintf(path, MAXPATHLEN, "%s%s/%s/%s%s",
mntent->mnt_dir, init,
/* find unused mnt_dir + lxcgroup + name + -$i */
ret = snprintf(path, MAXPATHLEN, "%s/%s/%s%s", mntent->mnt_dir,
lxcgroup ? lxcgroup : "lxc", name, tail);
if (ret < 0 || ret >= MAXPATHLEN)
goto fail;
@ -712,9 +579,6 @@ again:
goto fail;
}
if (hasmntopt(mntent, "cpuset")) {
setup_cpuset(path);
}
}
endmntent(file);
@ -741,7 +605,7 @@ fail:
int lxc_cgroup_enter(const char *cgpath, pid_t pid)
{
char path[MAXPATHLEN], initpath[MAXPATHLEN], *init;
char path[MAXPATHLEN];
FILE *file = NULL, *fout;
struct mntent *mntent;
int ret, retv = -1;
@ -757,9 +621,8 @@ int lxc_cgroup_enter(const char *cgpath, pid_t pid)
continue;
if (!mount_has_subsystem(mntent))
continue;
init = get_init_cgroup(NULL, mntent, initpath, 1);
ret = snprintf(path, MAXPATHLEN, "%s%s/%s/tasks",
mntent->mnt_dir, init, cgpath);
ret = snprintf(path, MAXPATHLEN, "%s/%s/tasks",
mntent->mnt_dir, cgpath);
if (ret < 0 || ret >= MAXPATHLEN) {
ERROR("entering cgroup");
goto out;
@ -826,12 +689,11 @@ int recursive_rmdir(char *dirname)
static int lxc_one_cgroup_destroy(struct mntent *mntent, const char *cgpath)
{
char cgname[MAXPATHLEN], initcgroup[MAXPATHLEN];
char cgname[MAXPATHLEN];
char *cgmnt = mntent->mnt_dir;
int rc;
rc = snprintf(cgname, MAXPATHLEN, "%s%s/%s", cgmnt,
get_init_cgroup(NULL, mntent, initcgroup, 1), cgpath);
rc = snprintf(cgname, MAXPATHLEN, "%s/%s", cgmnt, cgpath);
if (rc < 0 || rc >= MAXPATHLEN) {
ERROR("name too long");
return -1;