2 files changed, 140 insertions, 0 deletions

diff --git a/util-linux/Config.in b/util-linux/Config.in
index dc6d8fdad..fe71dac2c 100644
--- a/util-linux/Config.in
+++ b/util-linux/Config.in
@@ -288,6 +288,28 @@ config CONFIG_PIVOT_ROOT
           of wild and crazy things with your Linux system and is far more
           powerful than 'chroot'.
 
+          Note: This is for initrd in linux 2.4.  Under initramfs (introduced
+          in linux 2.6) use switch_root instead.
+
+config CONFIG_SWITCH_ROOT
+        bool "switch_root"
+        default n
+        help
+          The switch_root utility is used from initramfs to select a new
+          root device.  Under initramfs, you have to use this instead of
+          pivot_root.  (Stop reading here if you don't care why.)
+
+          Booting with initramfs extracts a gzipped cpio archive into rootfs
+          (which is a variant of ramfs/tmpfs).  Because rootfs can't be moved
+          or unmounted*, pivot_root will not work from initramfs.  Instead,
+          switch_root deletes everything out of rootfs (including itself),
+          does a mount --move that overmounts rootfs with the new root, and
+          then execs the specified init program.
+
+          * Because the Linux kernel uses rootfs internally as the starting
+          and ending point for searching through the kernel's doubly linked
+          list of active mount points.  That's why.
+
 config CONFIG_RDATE
         bool "rdate"
         default n
diff --git a/util-linux/switch_root.c b/util-linux/switch_root.c
new file mode 100644
index 000000000..242699560
--- /dev/null
+++ b/util-linux/switch_root.c
@@ -0,0 +1,118 @@
+/* vi:set ts=4:*/
+
+#include <dirent.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <sys/mount.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/vfs.h>
+#include <unistd.h>
+
+#include "busybox.h"
+
+// Make up for header deficiencies.
+
+#ifndef RAMFS_MAGIC
+#define RAMFS_MAGIC             0x858458f6
+#endif
+
+#ifndef TMPFS_MAGIC
+#define TMPFS_MAGIC             0x01021994
+#endif
+
+#ifndef MS_MOVE
+#define MS_MOVE                 8192
+#endif
+
+dev_t rootdev;
+
+// Recursively delete contents of rootfs.
+
+static void delete_contents(char *directory)
+{
+        DIR *dir;
+        struct dirent *d;
+        struct stat st;
+
+        // Don't descend into other filesystems
+        if (stat(directory,&st) || st.st_dev != rootdev) return;
+
+        // Recursively delete the contents of directories.
+        if (S_ISDIR(st.st_mode)) {
+                if((dir = opendir(directory))) {
+                        while ((d = readdir(dir))) {
+                                char *newdir=d->d_name;
+
+                                // Skip . and ..
+                                if(*newdir=='.' && (!newdir[1] || (newdir[1]=='.' && !newdir[2])))
+                                        continue;
+                                
+                                // Recurse to delete contents
+                                newdir = alloca(strlen(directory) + strlen(d->d_name) + 2);
+                                sprintf(newdir, "%s/%s", directory, d->d_name);
+                                delete_contents(newdir);
+                        }
+                        closedir(dir);
+                        
+                        // Directory should now be empty.  Zap it.
+                        printf("rmdir %s\n",directory); // rmdir(directory);
+                }
+                
+        // It wasn't a directory.  Zap it.
+                
+        } else printf("unlink %s\n",directory); //unlink(directory);
+}
+
+int switch_root_main(int argc, char *argv[])
+{
+        char *newroot, *console="/dev/console";
+        struct stat st1, st2;
+        struct statfs stfs;
+
+        // Parse args (-c console)
+
+        bb_opt_complementally="-2";
+        bb_getopt_ulflags(argc,argv,"c:",&console);
+        
+        // Change to new root directory and verify it's a different fs.
+
+        newroot=argv[optind++];
+        
+        if (chdir(newroot) || stat(".", &st1) || stat("/", &st2) ||
+                st1.st_dev == st2.st_dev)
+        {
+                bb_error_msg_and_die("bad newroot %s",newroot);
+        }
+        rootdev=st2.st_dev;
+        
+        // Additional sanity checks: we're about to rm -rf /,  so be REALLY SURE
+        // we mean it.  (I could make this a CONFIG option, but I would get email
+        // from all the people who WILL eat their filesystemss.)
+
+        if (stat("/init", &st1) || !S_ISREG(st1.st_mode) || statfs("/", &stfs) ||
+                (stfs.f_type != RAMFS_MAGIC && stfs.f_type != TMPFS_MAGIC) ||
+                getpid() != 1)
+        {
+                bb_error_msg_and_die("not rootfs");
+        }
+
+        // Zap everything out of rootdev
+        delete_contents("/");
+        
+        // Overmount / with newdir
+
+        if (mount(".", "/", NULL, MS_MOVE, NULL) || chdir("/"))
+                bb_error_msg_and_die("moving root");
+        
+        // Reopen stdin/stdout/stderr to /dev/console
+        close(0);
+        if(open(console, O_RDWR) < 0)
+                bb_error_msg_and_die("Bad console '%s'",console);
+        dup2(0, 1);
+        dup2(0, 2);
+
+        // Exec real init.  (This is why we must be pid 1.)
+        execv(argv[optind],argv+optind+1);
+        bb_error_msg_and_die("Bad init '%s'",argv[optind]);
+}