[lxc-devel] [PATCH 1/2] Setup devtmpfs and /dev for autodev bind mounts.
Serge Hallyn
serge.hallyn at ubuntu.com
Tue Nov 19 20:53:44 UTC 2013
Quoting Michael H. Warfield (mhw at WittsEnd.com):
> If autodev is not specifically set to 0 or 1, attempts to determine if
> systemd is being utilized and forces autodev=1 to prevent host system
> conflicts and collisions.
>
> If autodev is enabled and the host /dev is mounted with devtmpfs
> or /dev/.lxc is mounted with another file system...
>
> Each container created by a privileged user gets a /dev directory
> mapped off the host /dev here:
>
> /dev/.lxc/${name}.$( hash $lxcpath/$name )
>
> Each container created by a non-privileged user gets a /dev/directory
> mapped off the host /dev here:
>
> /dev/.lxc/user/${name}.$( hash $lxcpath/$name )
>
> The /dev/.lxc/user is mode 1777 to allow unpriv access.
>
> The /dev/.lxc/{containerdev} is bind mounted into the container /dev.
>
> Fallback on failure is to mount tmpfs into the container /dev.
>
> A symlink is created from $lxcpath/$name/rootfs.dev back to the /dev
> relative directory to provid a code consistent reference for updating
> container devs.
Hi Michael,
I haven't run it, but looking over the code in detail, it looks good
to me. I know Stéphane in the past has objected to trying to autodetect
systemd based containers. IMO it'd be fine to expect the templates to
set lxc.autodev, as I worry about building distro-detecting hacks into
core lxc, but unless Stéphane objects I don't care enough in this case
to nack it.
Did you consider this patch good enough to apply, or was this still an
RFC? I'm happy either way.
Acked-by: Serge E. Hallyn <serge.hallyn at ubuntu.com>
> ---
> src/lxc/conf.c | 369 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
> src/lxc/conf.h | 4 +-
> src/lxc/start.c | 2 +-
> 3 files changed, 363 insertions(+), 12 deletions(-)
>
> diff --git a/src/lxc/conf.c b/src/lxc/conf.c
> index a756731..d17cb2a 100644
> --- a/src/lxc/conf.c
> +++ b/src/lxc/conf.c
> @@ -29,6 +29,7 @@
> #include <string.h>
> #include <dirent.h>
> #include <unistd.h>
> +#include <inttypes.h>
> #include <sys/wait.h>
> #include <sys/syscall.h>
> #include <time.h>
> @@ -1164,20 +1165,275 @@ static int setup_rootfs_pivot_root(const char *rootfs, const char *pivotdir)
> return 0;
> }
>
> +
> +/*
> + * Note: This is a verbatum copy of what is in monitor.c. We're just
> + * usint it here to generate a safe subdirectory in /dev/ for the
> + * containers /dev/
> + */
> +
> +/* Note we don't use SHA-1 here as we don't want to depend on HAVE_GNUTLS.
> + * FNV has good anti collision properties and we're not worried
> + * about pre-image resistance or one-way-ness, we're just trying to make
> + * the name unique in the 108 bytes of space we have.
> + */
> +#define FNV1A_64_INIT ((uint64_t)0xcbf29ce484222325ULL)
> +static uint64_t fnv_64a_buf(void *buf, size_t len, uint64_t hval)
> +{
> + unsigned char *bp;
> +
> + for(bp = buf; bp < (unsigned char *)buf + len; bp++)
> + {
> + /* xor the bottom with the current octet */
> + hval ^= (uint64_t)*bp;
> +
> + /* gcc optimised:
> + * multiply by the 64 bit FNV magic prime mod 2^64
> + */
> + hval += (hval << 1) + (hval << 4) + (hval << 5) +
> + (hval << 7) + (hval << 8) + (hval << 40);
> + }
> +
> + return hval;
> +}
> +
> +/*
> + * Check to see if a directory has something mounted on it and,
> + * if it does, return the fstype.
> + *
> + * Code largely based on detect_shared_rootfs below
> + *
> + * Returns: # of matching entries in /proc/self/mounts
> + * if != 0 fstype is filled with the last filesystem value.
> + * if == 0 no matches found, fstype unchanged.
> + *
> + * ToDo: Maybe return the mount options in another parameter...
> + */
> +
> +#define LINELEN 4096
> +#define MAX_FSTYPE_LEN 128
> +int mount_check_fs( const char *dir, char *fstype )
> +{
> + char buf[LINELEN], *p;
> + struct stat s;
> + FILE *f;
> + int found_fs = 0;
> + char *p2;
> +
> + DEBUG("entering mount_check_fs for %s\n", dir);
> +
> + if ( 0 != access(dir, F_OK) || 0 != stat(dir, &s) || 0 == S_ISDIR(s.st_mode) ) {
> + return 0;
> + }
> +
> + process_lock();
> + f = fopen("/proc/self/mounts", "r");
> + process_unlock();
> + if (!f)
> + return 0;
> + while ((p = fgets(buf, LINELEN, f))) {
> + p = index(buf, ' ');
> + if( !p )
> + continue;
> + *p = '\0';
> + p2 = p + 1;
> +
> + p = index(p2, ' ');
> + if( !p )
> + continue;
> + *p = '\0';
> +
> + /* Compare the directory in the entry to desired */
> + if( strcmp( p2, dir ) ) {
> + continue;
> + }
> +
> + p2 = p + 1;
> + p = index( p2, ' ');
> + if( !p )
> + continue;
> + *p = '\0';
> +
> + ++found_fs;
> +
> + if( fstype ) {
> + strncpy( fstype, p2, MAX_FSTYPE_LEN - 1 );
> + fstype [ MAX_FSTYPE_LEN - 1 ] = '\0';
> + }
> + }
> +
> + process_lock();
> + fclose(f);
> + process_unlock();
> +
> + DEBUG("mount_check_fs returning %d last %s\n", found_fs, fstype);
> +
> + return found_fs;
> +}
> +
> +/*
> + * Locate a devtmpfs mount (should be on /dev) and create a container
> + * subdirectory on it which we can then bind mount to the container
> + * /dev instead of mounting a tmpfs there.
> + * If we fail, return NULL.
> + * Else return the pointer to the name buffer with the string to
> + * the devtmpfs subdirectory.
> + */
> +
> +char *mk_devtmpfs(const char *name, char *path, const char *lxcpath)
> +{
> + int ret;
> + struct stat s;
> + char tmp_path[MAXPATHLEN];
> + char fstype[MAX_FSTYPE_LEN];
> + char *base_path = "/dev/.lxc";
> + char *user_path = "/dev/.lxc/user";
> + uint64_t hash;
> +
> + if ( 0 != access(base_path, F_OK) || 0 != stat(base_path, &s) || 0 == S_ISDIR(s.st_mode) ) {
> + /* This is just making /dev/.lxc it better work or we're done */
> + ret = mkdir(base_path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
> + if ( ret ) {
> + SYSERROR( "Unable to create /dev/.lxc for autodev" );
> + return NULL;
> + }
> + }
> +
> + /*
> + * Programmers notes:
> + * We can not do mounts in this area of code that we want
> + * to be visible in the host. Consequently, /dev/.lxc must
> + * be set up earlier if we need a tmpfs mounted there.
> + * That only affects the rare cases where autodev is enabled
> + * for a container and devtmpfs is not mounted on /dev in the
> + * host. In that case, we'll fall back to the old method
> + * of mounting a tmpfs in the container and have no visibility
> + * into the container /dev.
> + */
> + if( ! mount_check_fs( "/dev", fstype )
> + || strcmp( "devtmpfs", fstype ) ) {
> + /* Either /dev was not mounted or was not devtmpfs */
> +
> + if ( ! mount_check_fs( "/dev/.lxc", NULL ) ) {
> + /*
> + * /dev/.lxc is not already mounted
> + * Doing a mount here does no good, since
> + * it's not visible in the host.
> + */
> +
> + ERROR("/dev/.lxc is not setup - taking fallback" );
> + return NULL;
> + }
> + }
> +
> + if ( 0 != access(user_path, F_OK) || 0 != stat(user_path, &s) || 0 == S_ISDIR(s.st_mode) ) {
> + /*
> + * This is making /dev/.lxc/user path for non-priv users.
> + * If this doesn't work, we'll have to fall back in the
> + * case of non-priv users. It's mode 1777 like /tmp.
> + */
> + ret = mkdir(user_path, S_IRWXU | S_IRWXG | S_IRWXO | S_ISVTX);
> + if ( ret ) {
> + /* Issue an error but don't fail yet! */
> + ERROR("Unable to create /dev/.lxc/user");
> + }
> + /* Umask tends to screw us up here */
> + chmod(user_path, S_IRWXU | S_IRWXG | S_IRWXO | S_ISVTX);
> + }
> +
> + /*
> + * Since the container name must be unique within a given
> + * lxcpath, we're going to use a hash of the path
> + * /lxcpath/name as our hash name in /dev/.lxc/
> + */
> +
> + ret = snprintf(tmp_path, MAXPATHLEN, "%s/%s", lxcpath, name);
> + if (ret < 0 || ret >= MAXPATHLEN)
> + return NULL;
> +
> + hash = fnv_64a_buf(tmp_path, ret, FNV1A_64_INIT);
> +
> + ret = snprintf(tmp_path, MAXPATHLEN, "%s/%s.%016" PRIx64, base_path, name, hash);
> + if (ret < 0 || ret >= MAXPATHLEN)
> + return NULL;
> +
> + if ( 0 != access(tmp_path, F_OK) || 0 != stat(tmp_path, &s) || 0 == S_ISDIR(s.st_mode) ) {
> + ret = mkdir(tmp_path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
> + if ( ret ) {
> + /* Something must have failed with the base_path...
> + * Maybe unpriv user. Try user_path now... */
> + INFO("Setup in /dev/.lxc failed. Trying /dev/.lxc/user." );
> +
> + ret = snprintf(tmp_path, MAXPATHLEN, "%s/%s.%016" PRIx64, user_path, name, hash);
> + if (ret < 0 || ret >= MAXPATHLEN)
> + return NULL;
> +
> + if ( 0 != access(tmp_path, F_OK) || 0 != stat(tmp_path, &s) || 0 == S_ISDIR(s.st_mode) ) {
> + ret = mkdir(tmp_path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
> + if ( ret ) {
> + ERROR("Container /dev setup in host /dev failed - taking fallback" );
> + return NULL;
> + }
> + }
> + }
> + }
> +
> + strcpy( path, tmp_path );
> + return path;
> +}
> +
> +
> /*
> * Do we want to add options for max size of /dev and a file to
> * specify which devices to create?
> */
> -static int mount_autodev(char *root)
> +static int mount_autodev(const char *name, char *root, const char *lxcpath)
> {
> int ret;
> + struct stat s;
> char path[MAXPATHLEN];
> + char host_path[MAXPATHLEN];
> + char devtmpfs_path[MAXPATHLEN];
>
> INFO("Mounting /dev under %s\n", root);
> +
> + ret = snprintf(host_path, MAXPATHLEN, "%s/%s/rootfs.dev", lxcpath, name);
> + if (ret < 0 || ret > MAXPATHLEN)
> + return -1;
> +
> ret = snprintf(path, MAXPATHLEN, "%s/dev", root);
> if (ret < 0 || ret > MAXPATHLEN)
> return -1;
> - ret = mount("none", path, "tmpfs", 0, "size=100000");
> +
> + if (mk_devtmpfs( name, devtmpfs_path, lxcpath ) ) {
> + /*
> + * Get rid of old links and directoriess
> + * This could be either a symlink and we remove it,
> + * or an empty directory and we remove it,
> + * or non-existant and we don't care,
> + * or a non-empty directory, and we will then emit an error
> + * but we will not fail out the process.
> + */
> + unlink( host_path );
> + rmdir( host_path );
> + ret = symlink(devtmpfs_path, host_path);
> +
> + if ( ret < 0 ) {
> + SYSERROR("WARNING: Failed to create symlink '%s'->'%s'\n", host_path, devtmpfs_path);
> + }
> + DEBUG("Bind mounting %s to %s", devtmpfs_path , path );
> + ret = mount(devtmpfs_path, path, NULL, MS_BIND, 0 );
> + } else {
> + /* Only mount a tmpfs on here if we don't already a mount */
> + if ( ! mount_check_fs( host_path, NULL ) ) {
> + DEBUG("Mounting tmpfs to %s", host_path );
> + ret = mount("none", path, "tmpfs", 0, "size=100000");
> + } else {
> + /* This allows someone to manually set up a mount */
> + DEBUG("Bind mounting %s to %s", host_path, path );
> + ret = mount(host_path , path, NULL, MS_BIND, 0 );
> + }
> + }
> if (ret) {
> SYSERROR("Failed to mount /dev at %s\n", root);
> return -1;
> @@ -1185,10 +1441,16 @@ static int mount_autodev(char *root)
> ret = snprintf(path, MAXPATHLEN, "%s/dev/pts", root);
> if (ret < 0 || ret >= MAXPATHLEN)
> return -1;
> - ret = mkdir(path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
> - if (ret) {
> - SYSERROR("Failed to create /dev/pts in container");
> - return -1;
> + /*
> + * If we are running on a devtmpfs mapping, dev/pts may already exist.
> + * If not, then create it and exit if that fails...
> + */
> + if ( 0 != access(path, F_OK) || 0 != stat(path, &s) || 0 == S_ISDIR(s.st_mode) ) {
> + ret = mkdir(path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
> + if (ret) {
> + SYSERROR("Failed to create /dev/pts in container");
> + return -1;
> + }
> }
>
> INFO("Mounted /dev under %s\n", root);
> @@ -2369,6 +2631,7 @@ struct lxc_conf *lxc_conf_init(void)
>
> new->loglevel = LXC_LOG_PRIORITY_NOTSET;
> new->personality = -1;
> + new->autodev = -1;
> new->console.log_path = NULL;
> new->console.log_fd = -1;
> new->console.path = NULL;
> @@ -3136,7 +3399,89 @@ int ttys_shift_ids(struct lxc_conf *c)
> return 0;
> }
>
> -int lxc_setup(const char *name, struct lxc_conf *lxc_conf, const char *lxcpath, struct cgroup_process_info *cgroup_info)
> +/*
> + * This routine is called when the configuration does not already specify a value
> + * for autodev (mounting a file system on /dev and populating it in a container).
> + * If a hard override value has not be specified, then we try to apply some
> + * heuristics to determine if we should switch to autodev mode.
> + *
> + * For instance, if the container has an /etc/systemd/system directory then it
> + * is probably running systemd as the init process and it needs the autodev
> + * mount to prevent it from mounting devtmpfs on /dev on it's own causing conflicts
> + * in the host.
> + *
> + * We may also want to enable autodev if the host has devtmpfs mounted on its
> + * /dev as this then enable us to use subdirectories under /dev for the container
> + * /dev directories and we can fake udev devices.
> + */
> +struct start_args {
> + char *const *argv;
> +};
> +
> +#define MAX_SYMLINK_DEPTH 32
> +
> +int check_autodev( const char *rootfs, void *data )
> +{
> + struct start_args *arg = data;
> + int ret;
> + int loop_count = 0;
> + struct stat s;
> + char absrootfs[MAXPATHLEN];
> + char path[MAXPATHLEN];
> + char abs_path[MAXPATHLEN];
> + char *command = "/sbin/init";
> +
> + if (rootfs == NULL || strlen(rootfs) == 0)
> + return -2;
> +
> + if (!realpath(rootfs, absrootfs))
> + return -2;
> +
> + if( arg && arg->argv[0] ) {
> + command = arg->argv[0];
> + DEBUG("Set exec command to %s\n", command );
> + }
> +
> + strncpy( path, command, MAXPATHLEN-1 );
> +
> + if ( 0 != access(path, F_OK) || 0 != stat(path, &s) )
> + return -2;
> +
> + /* Dereference down the symlink merry path testing as we go. */
> + /* If anything references systemd in the path - set autodev! */
> + /* Renormalize to the rootfs before each dereference */
> + /* Relative symlinks should fall out in the wash even with .. */
> + while( 1 ) {
> + if ( strstr( path, "systemd" ) ) {
> + INFO("Container with systemd init detected - enabling autodev!");
> + return 1;
> + }
> +
> + ret = snprintf(abs_path, MAXPATHLEN-1, "%s/%s", absrootfs, path);
> + if (ret < 0 || ret > MAXPATHLEN)
> + return -2;
> +
> + ret = readlink( abs_path, path, MAXPATHLEN-1 );
> +
> + if ( ( ret <= 0 ) || ( ++loop_count > MAX_SYMLINK_DEPTH ) ) {
> + break; /* Break out for other tests */
> + }
> + path[ret] = '\0';
> + }
> +
> + /*
> + * Add future checks here.
> + * Return positive if we should go autodev
> + * Return 0 if we should NOT go autodev
> + * Return negative if we encounter an error or can not determine...
> + */
> +
> + /* All else fails, we don't need autodev */
> + INFO("Autodev not required.");
> + return 0;
> +}
> +
> +int lxc_setup(const char *name, struct lxc_conf *lxc_conf, const char *lxcpath, struct cgroup_process_info *cgroup_info, void *data)
> {
> if (setup_utsname(lxc_conf->utsname)) {
> ERROR("failed to setup the utsname for '%s'", name);
> @@ -3158,8 +3503,12 @@ int lxc_setup(const char *name, struct lxc_conf *lxc_conf, const char *lxcpath,
> return -1;
> }
>
> - if (lxc_conf->autodev) {
> - if (mount_autodev(lxc_conf->rootfs.mount)) {
> + if (lxc_conf->autodev < 0) {
> + lxc_conf->autodev = check_autodev(lxc_conf->rootfs.mount, data);
> + }
> +
> + if (lxc_conf->autodev > 0) {
> + if (mount_autodev(name, lxc_conf->rootfs.mount, lxcpath)) {
> ERROR("failed to mount /dev in the container");
> return -1;
> }
> @@ -3197,7 +3546,7 @@ int lxc_setup(const char *name, struct lxc_conf *lxc_conf, const char *lxcpath,
> return -1;
> }
>
> - if (lxc_conf->autodev) {
> + if (lxc_conf->autodev > 0) {
> if (run_lxc_hooks(name, "autodev", lxc_conf, lxcpath, NULL)) {
> ERROR("failed to run autodev hooks for container '%s'.", name);
> return -1;
> diff --git a/src/lxc/conf.h b/src/lxc/conf.h
> index 940d493..090c5b3 100644
> --- a/src/lxc/conf.h
> +++ b/src/lxc/conf.h
> @@ -356,7 +356,9 @@ extern int lxc_clear_hooks(struct lxc_conf *c, const char *key);
>
> struct cgroup_process_info;
> extern int lxc_setup(const char *name, struct lxc_conf *lxc_conf,
> - const char *lxcpath, struct cgroup_process_info *cgroup_info);
> + const char *lxcpath,
> + struct cgroup_process_info *cgroup_info,
> + void *data);
>
> extern void lxc_rename_phys_nics_on_shutdown(struct lxc_conf *conf);
>
> diff --git a/src/lxc/start.c b/src/lxc/start.c
> index 3b2ba8f..3d2bc86 100644
> --- a/src/lxc/start.c
> +++ b/src/lxc/start.c
> @@ -553,7 +553,7 @@ static int do_start(void *data)
> #endif
>
> /* Setup the container, ip, names, utsname, ... */
> - if (lxc_setup(handler->name, handler->conf, handler->lxcpath, handler->cgroup)) {
> + if (lxc_setup(handler->name, handler->conf, handler->lxcpath, handler->cgroup, handler->data) ){
> ERROR("failed to setup the container");
> goto out_warn_father;
> }
> --
> 1.8.3.1
>
>
> --
> Michael H. Warfield (AI4NB) | (770) 978-7061 | mhw at WittsEnd.com
> /\/\|=mhw=|\/\/ | (678) 463-0932 | http://www.wittsend.com/mhw/
> NIC whois: MHW9 | An optimist believes we live in the best of all
> PGP Key: 0x674627FF | possible worlds. A pessimist is sure of it!
>
> ------------------------------------------------------------------------------
> November Webinars for C, C++, Fortran Developers
> Accelerate application performance with scalable programming models. Explore
> techniques for threading, error checking, porting, and tuning. Get the most
> from the latest Intel processors and coprocessors. See abstracts and register
> http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk
> _______________________________________________
> Lxc-devel mailing list
> Lxc-devel at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/lxc-devel
More information about the lxc-devel
mailing list