#include <nih/alloc.h>
#include <nih/string.h>
+#include <nih/alloc.h>
#include "cgmanager.h"
};
#define LXCFS_DATA ((struct lxcfs_state *) fuse_get_context()->private_data)
+enum {
+ LXC_TYPE_CGDIR,
+ LXC_TYPE_CGFILE,
+ LXC_TYPE_PROC_MEMINFO,
+ LXC_TYPE_PROC_CPUINFO,
+ LXC_TYPE_PROC_UPTIME,
+ LXC_TYPE_PROC_STAT,
+ LXC_TYPE_PROC_DISKSTATS,
+};
+
+struct file_info {
+ char *controller;
+ char *cgroup;
+ char *file;
+ int type;
+ char *buf; // unused as of yet
+ int buflen;
+};
+
+static char *must_copy_string(const char *str)
+{
+ if (!str)
+ return NULL;
+ return NIH_MUST( nih_strdup(NULL, str) );
+}
+
+/*
+ * TODO - return value should denote whether child exited with failure
+ * so callers can return errors. Esp read/write of tasks and cgroup.procs
+ */
static int wait_for_pid(pid_t pid)
{
int status, ret;
struct fuse_context *fc = fuse_get_context();
nih_local struct cgm_keys **list = NULL;
const char *cgroup;
+ struct file_info *dir_info;
nih_local char *controller = NULL;
- nih_local char *nextcg = NULL;
if (!fc)
return -EIO;
- if (strcmp(path, "/cgroup") == 0)
- return 0;
-
- // return list of keys for the controller, and list of child cgroups
- controller = pick_controller_from_path(fc, path);
- if (!controller)
- return -EIO;
+ if (strcmp(path, "/cgroup") == 0) {
+ cgroup = NULL;
+ controller = NULL;
+ } else {
+ // return list of keys for the controller, and list of child cgroups
+ controller = pick_controller_from_path(fc, path);
+ if (!controller)
+ return -EIO;
- cgroup = find_cgroup_in_path(path);
- if (!cgroup) {
- /* this is just /cgroup/controller, return its contents */
- cgroup = "/";
+ cgroup = find_cgroup_in_path(path);
+ if (!cgroup) {
+ /* this is just /cgroup/controller, return its contents */
+ cgroup = "/";
+ }
}
if (!fc_may_access(fc, controller, cgroup, NULL, O_RDONLY))
return -EACCES;
+
+ /* we'll free this at cg_releasedir */
+ dir_info = NIH_MUST( nih_alloc(NULL, sizeof(*dir_info)) );
+ dir_info->controller = must_copy_string(controller);
+ dir_info->cgroup = must_copy_string(cgroup);
+ dir_info->type = LXC_TYPE_CGDIR;
+ dir_info->buf = NULL;
+ dir_info->file = NULL;
+ dir_info->buflen = 0;
+
+ fi->fh = (unsigned long)dir_info;
return 0;
}
static int cg_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset,
struct fuse_file_info *fi)
{
+ struct file_info *d = (struct file_info *)fi->fh;
+ nih_local struct cgm_keys **list = NULL;
+ int i;
+ nih_local char *nextcg = NULL;
struct fuse_context *fc = fuse_get_context();
- if (!fc)
+ if (d->type != LXC_TYPE_CGDIR) {
+ fprintf(stderr, "Internal error: file cache info used in readdir\n");
return -EIO;
-
- if (strcmp(path, "/cgroup") == 0) {
- // get list of controllers
+ }
+ if (!d->cgroup && !d->controller) {
+ // ls /var/lib/lxcfs/cgroup - just show list of controllers
char **list = LXCFS_DATA ? LXCFS_DATA->subsystems : NULL;
int i;
return 0;
}
- // return list of keys for the controller, and list of child cgroups
- nih_local struct cgm_keys **list = NULL;
- const char *cgroup;
- nih_local char *controller = NULL;
- int i;
- nih_local char *nextcg = NULL;
-
- controller = pick_controller_from_path(fc, path);
- if (!controller)
- return -EIO;
-
- cgroup = find_cgroup_in_path(path);
- if (!cgroup) {
- /* this is just /cgroup/controller, return its contents */
- cgroup = "/";
- }
-
- if (!fc_may_access(fc, controller, cgroup, NULL, O_RDONLY))
- return -EACCES;
-
- if (!cgm_list_keys(controller, cgroup, &list))
+ if (!cgm_list_keys(d->controller, d->cgroup, &list))
// not a valid cgroup
return -EINVAL;
- if (!caller_is_in_ancestor(fc->pid, controller, cgroup, &nextcg)) {
+ if (!caller_is_in_ancestor(fc->pid, d->controller, d->cgroup, &nextcg)) {
if (nextcg) {
int ret;
ret = filler(buf, nextcg, NULL, 0);
}
// now get the list of child cgroups
- nih_local char **clist;
+ nih_local char **clist = NULL;
- if (!cgm_list_children(controller, cgroup, &clist))
+ if (!cgm_list_children(d->controller, d->cgroup, &clist))
return 0;
for (i = 0; clist[i]; i++) {
if (filler(buf, clist[i], NULL, 0) != 0) {
return 0;
}
+static void do_release_file_info(struct file_info *f)
+{
+ if (f->controller)
+ nih_free(f->controller);
+ if (f->cgroup)
+ nih_free(f->cgroup);
+ if (f->file)
+ nih_free(f->file);
+ free(f->buf);
+ nih_free(f);
+}
+
static int cg_releasedir(const char *path, struct fuse_file_info *fi)
{
+ struct file_info *d = (struct file_info *)fi->fh;
+
+ do_release_file_info(d);
return 0;
}
-/*
- * TODO - cache info here for read/write, release in cg_release.
- */
static int cg_open(const char *path, struct fuse_file_info *fi)
{
nih_local char *controller = NULL;
char *fpath = NULL, *path1, *path2;
nih_local char * cgdir = NULL;
nih_local struct cgm_keys *k = NULL;
+ struct file_info *file_info;
struct fuse_context *fc = fuse_get_context();
if (!fc)
path2 = fpath;
}
- if ((k = get_cgroup_key(controller, path1, path2)) != NULL) {
- if (!fc_may_access(fc, controller, path1, path2, fi->flags))
- // should never get here
- return -EACCES;
+ k = get_cgroup_key(controller, path1, path2);
+ if (!k)
+ return -EINVAL;
- return 0;
- }
+ if (!fc_may_access(fc, controller, path1, path2, fi->flags))
+ // should never get here
+ return -EACCES;
- return -EINVAL;
+ /* we'll free this at cg_release */
+ file_info = NIH_MUST( nih_alloc(NULL, sizeof(*file_info)) );
+ file_info->controller = must_copy_string(controller);
+ file_info->cgroup = must_copy_string(path1);
+ file_info->file = must_copy_string(path2);
+ file_info->type = LXC_TYPE_CGFILE;
+ file_info->buf = NULL;
+ file_info->buflen = 0;
+
+ fi->fh = (unsigned long)file_info;
+ return 0;
+}
+
+static int cg_release(const char *path, struct fuse_file_info *fi)
+{
+ struct file_info *f = (struct file_info *)fi->fh;
+
+ do_release_file_info(f);
+ return 0;
}
static int msgrecv(int sockfd, void *buf, size_t len)
tv.tv_sec = 2;
tv.tv_usec = 0;
- if (select(sockfd+1, &rfds, NULL, NULL, &tv) < 0)
+ if (select(sockfd+1, &rfds, NULL, NULL, &tv) <= 0)
return -1;
return recv(sockfd, buf, len, MSG_DONTWAIT);
}
-static bool send_creds(int sock, struct ucred *cred, char v)
+#define SEND_CREDS_OK 0
+#define SEND_CREDS_NOTSK 1
+#define SEND_CREDS_FAIL 2
+static int send_creds(int sock, struct ucred *cred, char v, bool pingfirst)
{
struct msghdr msg = { 0 };
struct iovec iov;
char buf[1];
buf[0] = 'p';
- if (msgrecv(sock, buf, 1) != 1) {
- printf("%s: Error getting reply from server over socketpair",
- __func__);
- return false;
+ if (pingfirst) {
+ if (msgrecv(sock, buf, 1) != 1) {
+ fprintf(stderr, "%s: Error getting reply from server over socketpair\n",
+ __func__);
+ return SEND_CREDS_FAIL;
+ }
}
msg.msg_control = cmsgbuf;
msg.msg_iovlen = 1;
if (sendmsg(sock, &msg, 0) < 0) {
- printf("%s: failed at sendmsg: %s", __func__,
+ fprintf(stderr, "%s: failed at sendmsg: %s\n", __func__,
strerror(errno));
if (errno == 3)
- return true;
- return false;
+ return SEND_CREDS_NOTSK;
+ return SEND_CREDS_FAIL;
}
- return true;
+ return SEND_CREDS_OK;
}
static bool recv_creds(int sock, struct ucred *cred, char *v)
char buf[1];
int ret;
int optval = 1;
+ struct timeval tv;
+ fd_set rfds;
*v = '1';
cred->gid = -1;
if (setsockopt(sock, SOL_SOCKET, SO_PASSCRED, &optval, sizeof(optval)) == -1) {
- printf("Failed to set passcred: %s", strerror(errno));
+ fprintf(stderr, "Failed to set passcred: %s\n", strerror(errno));
return false;
}
buf[0] = '1';
if (write(sock, buf, 1) != 1) {
- printf("Failed to start write on scm fd: %s", strerror(errno));
+ fprintf(stderr, "Failed to start write on scm fd: %s\n", strerror(errno));
return false;
}
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
- // retry logic is not ideal, especially as we are not
- // threaded. Sleep at most 1 second waiting for the client
- // to send us the scm_cred
- ret = recvmsg(sock, &msg, 0);
+ FD_ZERO(&rfds);
+ FD_SET(sock, &rfds);
+ tv.tv_sec = 2;
+ tv.tv_usec = 0;
+ if (select(sock+1, &rfds, NULL, NULL, &tv) <= 0) {
+ fprintf(stderr, "Failed to select for scm_cred: %s\n",
+ strerror(errno));
+ return false;
+ }
+ ret = recvmsg(sock, &msg, MSG_DONTWAIT);
if (ret < 0) {
- printf("Failed to receive scm_cred: %s",
+ fprintf(stderr, "Failed to receive scm_cred: %s\n",
strerror(errno));
return false;
}
/*
- * pidreader - reads pids from a ucred over a socket, then writes the
- * int value back over the socket
+ * pid_to_ns - reads pids from a ucred over a socket, then writes the
+ * int value back over the socket. This shifts the pid from the
+ * sender's pidns into tpid's pidns.
*/
-static void pidreader(int sock, pid_t tpid)
+static void pid_to_ns(int sock, pid_t tpid)
{
char v = '0';
struct ucred cred;
while (recv_creds(sock, &cred, &v)) {
if (v == '1')
exit(0);
- printf("CCC: child received %d\n", cred.pid);
if (write(sock, &cred.pid, sizeof(pid_t)) != sizeof(pid_t))
exit(1);
}
}
/*
- * pidreader_wrapper: when you setns into a pidns, you yourself remain
+ * pid_to_ns_wrapper: when you setns into a pidns, you yourself remain
* in your old pidns. Only children which you fork will be in the target
- * pidns. So the pidreader_wrapper does the setns, then forks a child to
+ * pidns. So the pid_to_ns_wrapper does the setns, then forks a child to
* actually convert pids
*/
-static void pidreader_wrapper(int sock, pid_t tpid)
+static void pid_to_ns_wrapper(int sock, pid_t tpid)
{
- int newnsfd = -1;
+ int newnsfd = -1, ret, cpipe[2];
char fnam[100];
pid_t cpid;
+ struct timeval tv;
+ fd_set s;
+ char v;
sprintf(fnam, "/proc/%d/ns/pid", tpid);
newnsfd = open(fnam, O_RDONLY);
exit(1);
close(newnsfd);
- cpid = fork();
+ if (pipe(cpipe) < 0)
+ exit(1);
+loop:
+ cpid = fork();
if (cpid < 0)
exit(1);
- if (!cpid)
- pidreader(sock, tpid);
+
+ if (!cpid) {
+ char b = '1';
+ close(cpipe[0]);
+ if (write(cpipe[1], &b, sizeof(char)) < 0) {
+ fprintf(stderr, "%s (child): erorr on write: %s\n",
+ __func__, strerror(errno));
+ }
+ close(cpipe[1]);
+ pid_to_ns(sock, tpid);
+ }
+ // give the child 1 second to be done forking and
+ // write it's ack
+ FD_ZERO(&s);
+ FD_SET(cpipe[0], &s);
+ tv.tv_sec = 1;
+ tv.tv_usec = 0;
+ ret = select(cpipe[0]+1, &s, NULL, NULL, &tv);
+ if (ret <= 0)
+ goto again;
+ ret = read(cpipe[0], &v, 1);
+ if (ret != sizeof(char) || v != '1') {
+ goto again;
+ }
+
if (!wait_for_pid(cpid))
exit(1);
exit(0);
+
+again:
+ kill(cpid, SIGKILL);
+ wait_for_pid(cpid);
+ goto loop;
}
/*
goto out;
if (!cpid) // child
- pidreader_wrapper(sock[1], tpid);
+ pid_to_ns_wrapper(sock[1], tpid);
char *ptr = tmpdata;
cred.uid = 0;
cred.gid = 0;
while (sscanf(ptr, "%d\n", &qpid) == 1) {
cred.pid = qpid;
- printf("AAA: sending %d\n", qpid);
- if (!send_creds(sock[0], &cred, v))
+ ret = send_creds(sock[0], &cred, v, true);
+
+ if (ret == SEND_CREDS_NOTSK)
+ goto next;
+ if (ret == SEND_CREDS_FAIL)
goto out;
// read converted results
FD_ZERO(&s);
FD_SET(sock[0], &s);
- tv.tv_sec = 1;
+ tv.tv_sec = 2;
tv.tv_usec = 0;
ret = select(sock[0]+1, &s, NULL, NULL, &tv);
if (ret <= 0) {
- kill(cpid, SIGTERM);
+ fprintf(stderr, "%s: select error waiting for pid from child: %s\n",
+ __func__, strerror(errno));
goto out;
}
if (read(sock[0], &qpid, sizeof(qpid)) != sizeof(qpid)) {
- kill(cpid, SIGTERM);
- perror("read");
+ fprintf(stderr, "%s: error reading pid from child: %s\n",
+ __func__, strerror(errno));
goto out;
}
- printf("BBB: read %d\n", qpid);
NIH_MUST( nih_strcat_sprintf(d, NULL, "%d\n", qpid) );
+next:
ptr = strchr(ptr, '\n');
if (!ptr)
break;
cred.pid = getpid();
v = '1';
- if (!send_creds(sock[0], &cred, v)) {
+ if (send_creds(sock[0], &cred, v, true) != SEND_CREDS_OK) {
// failed to ask child to exit
- kill(cpid, SIGTERM);
+ fprintf(stderr, "%s: failed to ask child to exit: %s\n",
+ __func__, strerror(errno));
goto out;
}
static int cg_read(const char *path, char *buf, size_t size, off_t offset,
struct fuse_file_info *fi)
{
- nih_local char *controller = NULL;
- const char *cgroup;
- char *fpath = NULL, *path1, *path2;
struct fuse_context *fc = fuse_get_context();
- nih_local char * cgdir = NULL;
+ struct file_info *f = (struct file_info *)fi->fh;
nih_local struct cgm_keys *k = NULL;
+ if (f->type != LXC_TYPE_CGFILE) {
+ fprintf(stderr, "Internal error: directory cache info used in cg_read\n");
+ return -EIO;
+ }
+
if (offset)
return -EIO;
if (!fc)
return -EIO;
- controller = pick_controller_from_path(fc, path);
- if (!controller)
- return -EINVAL;
- cgroup = find_cgroup_in_path(path);
- if (!cgroup)
+ if (!f->controller)
return -EINVAL;
- get_cgdir_and_path(cgroup, &cgdir, &fpath);
- if (!fpath) {
- path1 = "/";
- path2 = cgdir;
- } else {
- path1 = cgdir;
- path2 = fpath;
- }
-
- if ((k = get_cgroup_key(controller, path1, path2)) != NULL) {
+ if ((k = get_cgroup_key(f->controller, f->cgroup, f->file)) != NULL) {
nih_local char *data = NULL;
- int s, ret;
+ int s;
+ bool r;
- if (!fc_may_access(fc, controller, path1, path2, O_RDONLY))
+ if (!fc_may_access(fc, f->controller, f->cgroup, f->file, O_RDONLY))
// should never get here
return -EACCES;
- printf("XXX path2 is .%s.\n", path2);
- if (strcmp(path2, "tasks") == 0 ||
- strcmp(path2, "/tasks") == 0 ||
- strcmp(path2, "/cgroup.procs") == 0 ||
- strcmp(path2, "cgroup.procs") == 0)
+ if (strcmp(f->file, "tasks") == 0 ||
+ strcmp(f->file, "/tasks") == 0 ||
+ strcmp(f->file, "/cgroup.procs") == 0 ||
+ strcmp(f->file, "cgroup.procs") == 0)
// special case - we have to translate the pids
- ret = do_read_pids(fc->pid, controller, path1, path2, &data);
+ r = do_read_pids(fc->pid, f->controller, f->cgroup, f->file, &data);
else
- ret = cgm_get_value(controller, path1, path2, &data);
+ r = cgm_get_value(f->controller, f->cgroup, f->file, &data);
- if (ret == 0)
+ if (!r)
return -EINVAL;
+ if (!data)
+ return 0;
s = strlen(data);
if (s > size)
s = size;
return -EINVAL;
}
+static void pid_from_ns(int sock, pid_t tpid)
+{
+ pid_t vpid;
+ struct ucred cred;
+ char v;
+ struct timeval tv;
+ fd_set s;
+ int ret;
+
+ cred.uid = 0;
+ cred.gid = 0;
+ while (1) {
+ FD_ZERO(&s);
+ FD_SET(sock, &s);
+ tv.tv_sec = 2;
+ tv.tv_usec = 0;
+ ret = select(sock+1, &s, NULL, NULL, &tv);
+ if (ret <= 0) {
+ fprintf(stderr, "%s: bad select before read from parent: %s\n",
+ __func__, strerror(errno));
+ exit(1);
+ }
+ if ((ret = read(sock, &vpid, sizeof(pid_t))) != sizeof(pid_t)) {
+ fprintf(stderr, "%s: bad read from parent: %s\n",
+ __func__, strerror(errno));
+ exit(1);
+ }
+ if (vpid == -1) // done
+ break;
+ v = '0';
+ cred.pid = vpid;
+ if (send_creds(sock, &cred, v, true) != SEND_CREDS_OK) {
+ v = '1';
+ cred.pid = getpid();
+ if (send_creds(sock, &cred, v, false) != SEND_CREDS_OK)
+ exit(1);
+ }
+ }
+ exit(0);
+}
+
+static void pid_from_ns_wrapper(int sock, pid_t tpid)
+{
+ int newnsfd = -1, ret, cpipe[2];
+ char fnam[100];
+ pid_t cpid;
+ fd_set s;
+ struct timeval tv;
+ char v;
+
+ sprintf(fnam, "/proc/%d/ns/pid", tpid);
+ newnsfd = open(fnam, O_RDONLY);
+ if (newnsfd < 0)
+ exit(1);
+ if (setns(newnsfd, 0) < 0)
+ exit(1);
+ close(newnsfd);
+
+ if (pipe(cpipe) < 0)
+ exit(1);
+
+loop:
+ cpid = fork();
+
+ if (cpid < 0)
+ exit(1);
+
+ if (!cpid) {
+ char b = '1';
+ close(cpipe[0]);
+ if (write(cpipe[1], &b, sizeof(char)) < 0) {
+ fprintf(stderr, "%s (child): erorr on write: %s\n",
+ __func__, strerror(errno));
+ }
+ close(cpipe[1]);
+ pid_from_ns(sock, tpid);
+ }
+
+ // give the child 1 second to be done forking and
+ // write it's ack
+ FD_ZERO(&s);
+ FD_SET(cpipe[0], &s);
+ tv.tv_sec = 1;
+ tv.tv_usec = 0;
+ ret = select(cpipe[0]+1, &s, NULL, NULL, &tv);
+ if (ret <= 0)
+ goto again;
+ ret = read(cpipe[0], &v, 1);
+ if (ret != sizeof(char) || v != '1') {
+ goto again;
+ }
+
+ if (!wait_for_pid(cpid))
+ exit(1);
+ exit(0);
+
+again:
+ kill(cpid, SIGKILL);
+ wait_for_pid(cpid);
+ goto loop;
+}
+
+static bool do_write_pids(pid_t tpid, const char *contrl, const char *cg, const char *file, const char *buf)
+{
+ int sock[2] = {-1, -1};
+ pid_t qpid, cpid = -1;
+ bool answer = false, fail = false;
+
+ /*
+ * write the pids to a socket, have helper in writer's pidns
+ * call movepid for us
+ */
+ if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sock) < 0) {
+ perror("socketpair");
+ exit(1);
+ }
+
+ cpid = fork();
+ if (cpid == -1)
+ goto out;
+
+ if (!cpid) // child
+ pid_from_ns_wrapper(sock[1], tpid);
+
+ const char *ptr = buf;
+ while (sscanf(ptr, "%d", &qpid) == 1) {
+ struct ucred cred;
+ char v;
+
+ if (write(sock[0], &qpid, sizeof(qpid)) != sizeof(qpid)) {
+ fprintf(stderr, "%s: error writing pid to child: %s\n",
+ __func__, strerror(errno));
+ goto out;
+ }
+
+ if (recv_creds(sock[0], &cred, &v)) {
+ if (v == '0') {
+ if (!cgm_move_pid(contrl, cg, cred.pid))
+ fail = true;
+ }
+ }
+
+ ptr = strchr(ptr, '\n');
+ if (!ptr)
+ break;
+ ptr++;
+ }
+
+ /* All good, write the value */
+ qpid = -1;
+ if (write(sock[0], &qpid ,sizeof(qpid)) != sizeof(qpid))
+ fprintf(stderr, "Warning: failed to ask child to exit\n");
+
+ if (!fail)
+ answer = true;
+
+out:
+ if (cpid != -1)
+ wait_for_pid(cpid);
+ if (sock[0] != -1) {
+ close(sock[0]);
+ close(sock[1]);
+ }
+ return answer;
+}
+
int cg_write(const char *path, const char *buf, size_t size, off_t offset,
struct fuse_file_info *fi)
{
- nih_local char *controller = NULL;
- const char *cgroup;
- char *fpath = NULL, *path1, *path2;
struct fuse_context *fc = fuse_get_context();
- nih_local char * cgdir = NULL;
+ nih_local char *localbuf = NULL;
nih_local struct cgm_keys *k = NULL;
+ struct file_info *f = (struct file_info *)fi->fh;
+
+ if (f->type != LXC_TYPE_CGFILE) {
+ fprintf(stderr, "Internal error: directory cache info used in cg_write\n");
+ return -EIO;
+ }
if (offset)
return -EINVAL;
if (!fc)
return -EIO;
- controller = pick_controller_from_path(fc, path);
- if (!controller)
- return -EINVAL;
- cgroup = find_cgroup_in_path(path);
- if (!cgroup)
- return -EINVAL;
+ localbuf = NIH_MUST( nih_alloc(NULL, size+1) );
+ localbuf[size] = '\0';
+ memcpy(localbuf, buf, size);
- get_cgdir_and_path(cgroup, &cgdir, &fpath);
- if (!fpath) {
- path1 = "/";
- path2 = cgdir;
- } else {
- path1 = cgdir;
- path2 = fpath;
- }
+ if ((k = get_cgroup_key(f->controller, f->cgroup, f->file)) != NULL) {
+ bool r;
- if ((k = get_cgroup_key(controller, path1, path2)) != NULL) {
- if (!fc_may_access(fc, controller, path1, path2, O_WRONLY))
+ if (!fc_may_access(fc, f->controller, f->cgroup, f->file, O_WRONLY))
return -EACCES;
- if (!cgm_set_value(controller, path1, path2, buf))
+ if (strcmp(f->file, "tasks") == 0 ||
+ strcmp(f->file, "/tasks") == 0 ||
+ strcmp(f->file, "/cgroup.procs") == 0 ||
+ strcmp(f->file, "cgroup.procs") == 0)
+ // special case - we have to translate the pids
+ r = do_write_pids(fc->pid, f->controller, f->cgroup, f->file, localbuf);
+ else
+ r = cgm_set_value(f->controller, f->cgroup, f->file, localbuf);
+
+ if (!r)
return -EINVAL;
return size;
}
}
+static void get_blkio_io_value(char *str, unsigned major, unsigned minor, char *iotype, unsigned long *v)
+{
+ char *eol;
+ char key[32];
+
+ memset(key, 0, 32);
+ snprintf(key, 32, "%u:%u %s", major, minor, iotype);
+
+ size_t len = strlen(key);
+ *v = 0;
+
+ while (*str) {
+ if (startswith(str, key)) {
+ sscanf(str + len, "%lu", v);
+ return;
+ }
+ eol = strchr(str, '\n');
+ if (!eol)
+ return;
+ str = eol+1;
+ }
+}
+
static char *get_pid_cgroup(pid_t pid, const char *contrl)
{
nih_local char *fnam = NULL;
total_len += l;
}
+ fclose(f);
+ free(line);
return total_len;
}
}
}
+ fclose(f);
+ free(line);
return total_len;
}
nih_local char *cpuset = NULL;
char *line = NULL;
size_t linelen = 0, total_len = 0;
- int curcpu = 0;
+ int curcpu = -1; /* cpu numbering starts at 0 */
FILE *f;
if (offset)
while (getline(&line, &linelen, f) != -1) {
size_t l;
int cpu;
+ char cpu_char[10]; /* That's a lot of cores */
char *c;
- if (sscanf(line, "cpu%d", &cpu) != 1) {
- /* not a ^cpu line, just print it */
+ if (sscanf(line, "cpu%9[^ ]", cpu_char) != 1) {
+ /* not a ^cpuN line containing a number N, just print it */
l = snprintf(buf, size, "%s", line);
buf += l;
size -= l;
total_len += l;
continue;
}
+
+ if (sscanf(cpu_char, "%d", &cpu) != 1)
+ continue;
if (!cpu_in_cpuset(cpu, cpuset))
continue;
curcpu ++;
total_len += l;
}
+ fclose(f);
+ free(line);
return total_len;
}
static long int get_pid1_time(pid_t pid)
{
char fnam[100];
- int fd;
+ int fd, cpipe[2], ret;
struct stat sb;
- int ret;
- pid_t npid;
+ pid_t cpid;
+ struct timeval tv;
+ fd_set s;
+ char v;
if (unshare(CLONE_NEWNS))
return 0;
+ if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL)) {
+ perror("rslave mount failed");
+ return 0;
+ }
+
sprintf(fnam, "/proc/%d/ns/pid", pid);
fd = open(fnam, O_RDONLY);
if (fd < 0) {
return 0;
}
close(fd);
- npid = fork();
- if (npid < 0)
- return 0;
-
- if (npid) {
- // child will do the writing for us
- wait_for_pid(npid);
- exit(0);
- }
- umount2("/proc", MNT_DETACH);
+ if (pipe(cpipe) < 0)
+ exit(1);
- if (mount("proc", "/proc", "proc", 0, NULL)) {
- perror("get_pid1_time mount");
+loop:
+ cpid = fork();
+ if (cpid < 0)
return 0;
+
+ if (!cpid) {
+ char b = '1';
+ close(cpipe[0]);
+ if (write(cpipe[1], &b, sizeof(char)) < 0) {
+ fprintf(stderr, "%s (child): erorr on write: %s\n",
+ __func__, strerror(errno));
+ }
+ close(cpipe[1]);
+ umount2("/proc", MNT_DETACH);
+ if (mount("proc", "/proc", "proc", 0, NULL)) {
+ perror("get_pid1_time mount");
+ return 0;
+ }
+ ret = lstat("/proc/1", &sb);
+ if (ret) {
+ perror("get_pid1_time lstat");
+ return 0;
+ }
+ return time(NULL) - sb.st_ctime;
}
- ret = lstat("/proc/1", &sb);
- if (ret) {
- perror("get_pid1_time lstat");
- return 0;
+
+ // give the child 1 second to be done forking and
+ // write it's ack
+ FD_ZERO(&s);
+ FD_SET(cpipe[0], &s);
+ tv.tv_sec = 1;
+ tv.tv_usec = 0;
+ ret = select(cpipe[0]+1, &s, NULL, NULL, &tv);
+ if (ret <= 0)
+ goto again;
+ ret = read(cpipe[0], &v, 1);
+ if (ret != sizeof(char) || v != '1') {
+ goto again;
}
- return time(NULL) - sb.st_ctime;
+
+ wait_for_pid(cpid);
+ exit(0);
+
+again:
+ kill(cpid, SIGKILL);
+ wait_for_pid(cpid);
+ goto loop;
}
static long int getreaperage(pid_t qpid)
tv.tv_sec = 1;
tv.tv_usec = 0;
ret = select(mypipe[0]+1, &s, NULL, NULL, &tv);
- if (ret == -1) {
+ if (ret <= 0) {
perror("select");
goto out;
}
if (!ret) {
- printf("timed out\n");
+ fprintf(stderr, "timed out\n");
goto out;
}
if (read(mypipe[0], &mtime, sizeof(mtime)) != sizeof(mtime)) {
{
FILE *f = fopen("/proc/uptime", "r");
long int age, idle;
+ int ret;
if (!f)
return 0;
- if (fscanf(f, "%ld %ld", &age, &idle) != 2)
+ ret = fscanf(f, "%ld %ld", &age, &idle);
+ fclose(f);
+ if (ret != 2)
return 0;
return idle;
}
return snprintf(buf, size, "%ld %ld\n", reaperage, idletime);
}
+static int proc_diskstats_read(char *buf, size_t size, off_t offset,
+ struct fuse_file_info *fi)
+{
+ char dev_name[72];
+ struct fuse_context *fc = fuse_get_context();
+ nih_local char *cg = get_pid_cgroup(fc->pid, "blkio");
+ nih_local char *io_serviced_str = NULL, *io_merged_str = NULL, *io_service_bytes_str = NULL,
+ *io_wait_time_str = NULL, *io_service_time_str = NULL;
+ unsigned long read = 0, write = 0;
+ unsigned long read_merged = 0, write_merged = 0;
+ unsigned long read_sectors = 0, write_sectors = 0;
+ unsigned long read_ticks = 0, write_ticks = 0;
+ unsigned long ios_pgr = 0, tot_ticks = 0, rq_ticks = 0;
+ unsigned long rd_svctm = 0, wr_svctm = 0, rd_wait = 0, wr_wait = 0;
+ char *line = NULL;
+ size_t linelen = 0, total_len = 0;
+ unsigned int major = 0, minor = 0;
+ int i = 0;
+ FILE *f;
+
+ if (offset)
+ return -EINVAL;
+
+ if (!cg)
+ return 0;
+
+ if (!cgm_get_value("blkio", cg, "blkio.io_serviced", &io_serviced_str))
+ return 0;
+ if (!cgm_get_value("blkio", cg, "blkio.io_merged", &io_merged_str))
+ return 0;
+ if (!cgm_get_value("blkio", cg, "blkio.io_service_bytes", &io_service_bytes_str))
+ return 0;
+ if (!cgm_get_value("blkio", cg, "blkio.io_wait_time", &io_wait_time_str))
+ return 0;
+ if (!cgm_get_value("blkio", cg, "blkio.io_service_time", &io_service_time_str))
+ return 0;
+
+
+ f = fopen("/proc/diskstats", "r");
+ if (!f)
+ return 0;
+
+ while (getline(&line, &linelen, f) != -1) {
+ size_t l;
+ char *printme, lbuf[256];
+
+ i = sscanf(line, "%u %u %s", &major, &minor, dev_name);
+ if(i == 3){
+ get_blkio_io_value(io_serviced_str, major, minor, "Read", &read);
+ get_blkio_io_value(io_serviced_str, major, minor, "Write", &write);
+ get_blkio_io_value(io_merged_str, major, minor, "Read", &read_merged);
+ get_blkio_io_value(io_merged_str, major, minor, "Write", &write_merged);
+ get_blkio_io_value(io_service_bytes_str, major, minor, "Read", &read_sectors);
+ read_sectors = read_sectors/512;
+ get_blkio_io_value(io_service_bytes_str, major, minor, "Write", &write_sectors);
+ write_sectors = write_sectors/512;
+
+ get_blkio_io_value(io_service_time_str, major, minor, "Read", &rd_svctm);
+ rd_svctm = rd_svctm/1000000;
+ get_blkio_io_value(io_wait_time_str, major, minor, "Read", &rd_wait);
+ rd_wait = rd_wait/1000000;
+ read_ticks = rd_svctm + rd_wait;
+
+ get_blkio_io_value(io_service_time_str, major, minor, "Write", &wr_svctm);
+ wr_svctm = wr_svctm/1000000;
+ get_blkio_io_value(io_wait_time_str, major, minor, "Write", &wr_wait);
+ wr_wait = wr_wait/1000000;
+ write_ticks = wr_svctm + wr_wait;
+
+ get_blkio_io_value(io_service_time_str, major, minor, "Total", &tot_ticks);
+ tot_ticks = tot_ticks/1000000;
+ }else{
+ continue;
+ }
+
+ memset(lbuf, 0, 256);
+ if (read || write || read_merged || write_merged || read_sectors || write_sectors || read_ticks || write_ticks) {
+ snprintf(lbuf, 256, "%u %u %s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
+ major, minor, dev_name, read, read_merged, read_sectors, read_ticks,
+ write, write_merged, write_sectors, write_ticks, ios_pgr, tot_ticks, rq_ticks);
+ printme = lbuf;
+ } else
+ continue;
+
+ l = snprintf(buf, size, "%s", printme);
+ buf += l;
+ size -= l;
+ total_len += l;
+ }
+
+ fclose(f);
+ free(line);
+ return total_len;
+}
+
static off_t get_procfile_size(const char *which)
{
FILE *f = fopen(which, "r");
while ((sz = getline(&line, &len, f)) != -1)
answer += sz;
fclose (f);
+ free(line);
return answer;
}
if (strcmp(path, "/proc/meminfo") == 0 ||
strcmp(path, "/proc/cpuinfo") == 0 ||
strcmp(path, "/proc/uptime") == 0 ||
- strcmp(path, "/proc/stat") == 0) {
-
+ strcmp(path, "/proc/stat") == 0 ||
+ strcmp(path, "/proc/diskstats") == 0) {
sb->st_size = get_procfile_size(path);
sb->st_mode = S_IFREG | 00444;
sb->st_nlink = 1;
if (filler(buf, "cpuinfo", NULL, 0) != 0 ||
filler(buf, "meminfo", NULL, 0) != 0 ||
filler(buf, "stat", NULL, 0) != 0 ||
- filler(buf, "uptime", NULL, 0) != 0)
+ filler(buf, "uptime", NULL, 0) != 0 ||
+ filler(buf, "diskstats", NULL, 0) != 0)
return -EINVAL;
return 0;
}
if (strcmp(path, "/proc/meminfo") == 0 ||
strcmp(path, "/proc/cpuinfo") == 0 ||
strcmp(path, "/proc/uptime") == 0 ||
- strcmp(path, "/proc/stat") == 0)
+ strcmp(path, "/proc/stat") == 0 ||
+ strcmp(path, "/proc/diskstats") == 0)
return 0;
return -ENOENT;
}
return proc_uptime_read(buf, size, offset, fi);
if (strcmp(path, "/proc/stat") == 0)
return proc_stat_read(buf, size, offset, fi);
+ if (strcmp(path, "/proc/diskstats") == 0)
+ return proc_diskstats_read(buf, size, offset, fi);
return -EINVAL;
}
static int lxcfs_release(const char *path, struct fuse_file_info *fi)
{
- return 0;
+ if (strncmp(path, "/cgroup", 7) == 0)
+ return cg_release(path, fi);
+#if 0
+ if (strncmp(path, "/proc", 5) == 0)
+ return proc_close(path, fi);
+#endif
+
+ return -EINVAL;
}
static int lxcfs_fsync(const char *path, int datasync, struct fuse_file_info *fi)
projects / mirror_lxcfs.git / blobdiff