Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / init / initramfs.c
diff --git a/kernel/init/initramfs.c b/kernel/init/initramfs.c
new file mode 100644 (file)
index 0000000..ad1bd77
--- /dev/null
@@ -0,0 +1,658 @@
+/*
+ * Many of the syscalls used in this file expect some of the arguments
+ * to be __user pointers not __kernel pointers.  To limit the sparse
+ * noise, turn off sparse checking for this file.
+ */
+#ifdef __CHECKER__
+#undef __CHECKER__
+#warning "Sparse checking disabled for this file"
+#endif
+
+#include <linux/init.h>
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+#include <linux/fcntl.h>
+#include <linux/delay.h>
+#include <linux/string.h>
+#include <linux/dirent.h>
+#include <linux/syscalls.h>
+#include <linux/utime.h>
+
+static ssize_t __init xwrite(int fd, const char *p, size_t count)
+{
+       ssize_t out = 0;
+
+       /* sys_write only can write MAX_RW_COUNT aka 2G-4K bytes at most */
+       while (count) {
+               ssize_t rv = sys_write(fd, p, count);
+
+               if (rv < 0) {
+                       if (rv == -EINTR || rv == -EAGAIN)
+                               continue;
+                       return out ? out : rv;
+               } else if (rv == 0)
+                       break;
+
+               p += rv;
+               out += rv;
+               count -= rv;
+       }
+
+       return out;
+}
+
+static __initdata char *message;
+static void __init error(char *x)
+{
+       if (!message)
+               message = x;
+}
+
+/* link hash */
+
+#define N_ALIGN(len) ((((len) + 1) & ~3) + 2)
+
+static __initdata struct hash {
+       int ino, minor, major;
+       umode_t mode;
+       struct hash *next;
+       char name[N_ALIGN(PATH_MAX)];
+} *head[32];
+
+static inline int hash(int major, int minor, int ino)
+{
+       unsigned long tmp = ino + minor + (major << 3);
+       tmp += tmp >> 5;
+       return tmp & 31;
+}
+
+static char __init *find_link(int major, int minor, int ino,
+                             umode_t mode, char *name)
+{
+       struct hash **p, *q;
+       for (p = head + hash(major, minor, ino); *p; p = &(*p)->next) {
+               if ((*p)->ino != ino)
+                       continue;
+               if ((*p)->minor != minor)
+                       continue;
+               if ((*p)->major != major)
+                       continue;
+               if (((*p)->mode ^ mode) & S_IFMT)
+                       continue;
+               return (*p)->name;
+       }
+       q = kmalloc(sizeof(struct hash), GFP_KERNEL);
+       if (!q)
+               panic("can't allocate link hash entry");
+       q->major = major;
+       q->minor = minor;
+       q->ino = ino;
+       q->mode = mode;
+       strcpy(q->name, name);
+       q->next = NULL;
+       *p = q;
+       return NULL;
+}
+
+static void __init free_hash(void)
+{
+       struct hash **p, *q;
+       for (p = head; p < head + 32; p++) {
+               while (*p) {
+                       q = *p;
+                       *p = q->next;
+                       kfree(q);
+               }
+       }
+}
+
+static long __init do_utime(char *filename, time_t mtime)
+{
+       struct timespec t[2];
+
+       t[0].tv_sec = mtime;
+       t[0].tv_nsec = 0;
+       t[1].tv_sec = mtime;
+       t[1].tv_nsec = 0;
+
+       return do_utimes(AT_FDCWD, filename, t, AT_SYMLINK_NOFOLLOW);
+}
+
+static __initdata LIST_HEAD(dir_list);
+struct dir_entry {
+       struct list_head list;
+       char *name;
+       time_t mtime;
+};
+
+static void __init dir_add(const char *name, time_t mtime)
+{
+       struct dir_entry *de = kmalloc(sizeof(struct dir_entry), GFP_KERNEL);
+       if (!de)
+               panic("can't allocate dir_entry buffer");
+       INIT_LIST_HEAD(&de->list);
+       de->name = kstrdup(name, GFP_KERNEL);
+       de->mtime = mtime;
+       list_add(&de->list, &dir_list);
+}
+
+static void __init dir_utime(void)
+{
+       struct dir_entry *de, *tmp;
+       list_for_each_entry_safe(de, tmp, &dir_list, list) {
+               list_del(&de->list);
+               do_utime(de->name, de->mtime);
+               kfree(de->name);
+               kfree(de);
+       }
+}
+
+static __initdata time_t mtime;
+
+/* cpio header parsing */
+
+static __initdata unsigned long ino, major, minor, nlink;
+static __initdata umode_t mode;
+static __initdata unsigned long body_len, name_len;
+static __initdata uid_t uid;
+static __initdata gid_t gid;
+static __initdata unsigned rdev;
+
+static void __init parse_header(char *s)
+{
+       unsigned long parsed[12];
+       char buf[9];
+       int i;
+
+       buf[8] = '\0';
+       for (i = 0, s += 6; i < 12; i++, s += 8) {
+               memcpy(buf, s, 8);
+               parsed[i] = simple_strtoul(buf, NULL, 16);
+       }
+       ino = parsed[0];
+       mode = parsed[1];
+       uid = parsed[2];
+       gid = parsed[3];
+       nlink = parsed[4];
+       mtime = parsed[5];
+       body_len = parsed[6];
+       major = parsed[7];
+       minor = parsed[8];
+       rdev = new_encode_dev(MKDEV(parsed[9], parsed[10]));
+       name_len = parsed[11];
+}
+
+/* FSM */
+
+static __initdata enum state {
+       Start,
+       Collect,
+       GotHeader,
+       SkipIt,
+       GotName,
+       CopyFile,
+       GotSymlink,
+       Reset
+} state, next_state;
+
+static __initdata char *victim;
+static unsigned long byte_count __initdata;
+static __initdata loff_t this_header, next_header;
+
+static inline void __init eat(unsigned n)
+{
+       victim += n;
+       this_header += n;
+       byte_count -= n;
+}
+
+static __initdata char *vcollected;
+static __initdata char *collected;
+static long remains __initdata;
+static __initdata char *collect;
+
+static void __init read_into(char *buf, unsigned size, enum state next)
+{
+       if (byte_count >= size) {
+               collected = victim;
+               eat(size);
+               state = next;
+       } else {
+               collect = collected = buf;
+               remains = size;
+               next_state = next;
+               state = Collect;
+       }
+}
+
+static __initdata char *header_buf, *symlink_buf, *name_buf;
+
+static int __init do_start(void)
+{
+       read_into(header_buf, 110, GotHeader);
+       return 0;
+}
+
+static int __init do_collect(void)
+{
+       unsigned long n = remains;
+       if (byte_count < n)
+               n = byte_count;
+       memcpy(collect, victim, n);
+       eat(n);
+       collect += n;
+       if ((remains -= n) != 0)
+               return 1;
+       state = next_state;
+       return 0;
+}
+
+static int __init do_header(void)
+{
+       if (memcmp(collected, "070707", 6)==0) {
+               error("incorrect cpio method used: use -H newc option");
+               return 1;
+       }
+       if (memcmp(collected, "070701", 6)) {
+               error("no cpio magic");
+               return 1;
+       }
+       parse_header(collected);
+       next_header = this_header + N_ALIGN(name_len) + body_len;
+       next_header = (next_header + 3) & ~3;
+       state = SkipIt;
+       if (name_len <= 0 || name_len > PATH_MAX)
+               return 0;
+       if (S_ISLNK(mode)) {
+               if (body_len > PATH_MAX)
+                       return 0;
+               collect = collected = symlink_buf;
+               remains = N_ALIGN(name_len) + body_len;
+               next_state = GotSymlink;
+               state = Collect;
+               return 0;
+       }
+       if (S_ISREG(mode) || !body_len)
+               read_into(name_buf, N_ALIGN(name_len), GotName);
+       return 0;
+}
+
+static int __init do_skip(void)
+{
+       if (this_header + byte_count < next_header) {
+               eat(byte_count);
+               return 1;
+       } else {
+               eat(next_header - this_header);
+               state = next_state;
+               return 0;
+       }
+}
+
+static int __init do_reset(void)
+{
+       while (byte_count && *victim == '\0')
+               eat(1);
+       if (byte_count && (this_header & 3))
+               error("broken padding");
+       return 1;
+}
+
+static int __init maybe_link(void)
+{
+       if (nlink >= 2) {
+               char *old = find_link(major, minor, ino, mode, collected);
+               if (old)
+                       return (sys_link(old, collected) < 0) ? -1 : 1;
+       }
+       return 0;
+}
+
+static void __init clean_path(char *path, umode_t fmode)
+{
+       struct stat st;
+
+       if (!sys_newlstat(path, &st) && (st.st_mode ^ fmode) & S_IFMT) {
+               if (S_ISDIR(st.st_mode))
+                       sys_rmdir(path);
+               else
+                       sys_unlink(path);
+       }
+}
+
+static __initdata int wfd;
+
+static int __init do_name(void)
+{
+       state = SkipIt;
+       next_state = Reset;
+       if (strcmp(collected, "TRAILER!!!") == 0) {
+               free_hash();
+               return 0;
+       }
+       clean_path(collected, mode);
+       if (S_ISREG(mode)) {
+               int ml = maybe_link();
+               if (ml >= 0) {
+                       int openflags = O_WRONLY|O_CREAT;
+                       if (ml != 1)
+                               openflags |= O_TRUNC;
+                       wfd = sys_open(collected, openflags, mode);
+
+                       if (wfd >= 0) {
+                               sys_fchown(wfd, uid, gid);
+                               sys_fchmod(wfd, mode);
+                               if (body_len)
+                                       sys_ftruncate(wfd, body_len);
+                               vcollected = kstrdup(collected, GFP_KERNEL);
+                               state = CopyFile;
+                       }
+               }
+       } else if (S_ISDIR(mode)) {
+               sys_mkdir(collected, mode);
+               sys_chown(collected, uid, gid);
+               sys_chmod(collected, mode);
+               dir_add(collected, mtime);
+       } else if (S_ISBLK(mode) || S_ISCHR(mode) ||
+                  S_ISFIFO(mode) || S_ISSOCK(mode)) {
+               if (maybe_link() == 0) {
+                       sys_mknod(collected, mode, rdev);
+                       sys_chown(collected, uid, gid);
+                       sys_chmod(collected, mode);
+                       do_utime(collected, mtime);
+               }
+       }
+       return 0;
+}
+
+static int __init do_copy(void)
+{
+       if (byte_count >= body_len) {
+               if (xwrite(wfd, victim, body_len) != body_len)
+                       error("write error");
+               sys_close(wfd);
+               do_utime(vcollected, mtime);
+               kfree(vcollected);
+               eat(body_len);
+               state = SkipIt;
+               return 0;
+       } else {
+               if (xwrite(wfd, victim, byte_count) != byte_count)
+                       error("write error");
+               body_len -= byte_count;
+               eat(byte_count);
+               return 1;
+       }
+}
+
+static int __init do_symlink(void)
+{
+       collected[N_ALIGN(name_len) + body_len] = '\0';
+       clean_path(collected, 0);
+       sys_symlink(collected + N_ALIGN(name_len), collected);
+       sys_lchown(collected, uid, gid);
+       do_utime(collected, mtime);
+       state = SkipIt;
+       next_state = Reset;
+       return 0;
+}
+
+static __initdata int (*actions[])(void) = {
+       [Start]         = do_start,
+       [Collect]       = do_collect,
+       [GotHeader]     = do_header,
+       [SkipIt]        = do_skip,
+       [GotName]       = do_name,
+       [CopyFile]      = do_copy,
+       [GotSymlink]    = do_symlink,
+       [Reset]         = do_reset,
+};
+
+static long __init write_buffer(char *buf, unsigned long len)
+{
+       byte_count = len;
+       victim = buf;
+
+       while (!actions[state]())
+               ;
+       return len - byte_count;
+}
+
+static long __init flush_buffer(void *bufv, unsigned long len)
+{
+       char *buf = (char *) bufv;
+       long written;
+       long origLen = len;
+       if (message)
+               return -1;
+       while ((written = write_buffer(buf, len)) < len && !message) {
+               char c = buf[written];
+               if (c == '0') {
+                       buf += written;
+                       len -= written;
+                       state = Start;
+               } else if (c == 0) {
+                       buf += written;
+                       len -= written;
+                       state = Reset;
+               } else
+                       error("junk in compressed archive");
+       }
+       return origLen;
+}
+
+static unsigned long my_inptr; /* index of next byte to be processed in inbuf */
+
+#include <linux/decompress/generic.h>
+
+static char * __init unpack_to_rootfs(char *buf, unsigned long len)
+{
+       long written;
+       decompress_fn decompress;
+       const char *compress_name;
+       static __initdata char msg_buf[64];
+
+       header_buf = kmalloc(110, GFP_KERNEL);
+       symlink_buf = kmalloc(PATH_MAX + N_ALIGN(PATH_MAX) + 1, GFP_KERNEL);
+       name_buf = kmalloc(N_ALIGN(PATH_MAX), GFP_KERNEL);
+
+       if (!header_buf || !symlink_buf || !name_buf)
+               panic("can't allocate buffers");
+
+       state = Start;
+       this_header = 0;
+       message = NULL;
+       while (!message && len) {
+               loff_t saved_offset = this_header;
+               if (*buf == '0' && !(this_header & 3)) {
+                       state = Start;
+                       written = write_buffer(buf, len);
+                       buf += written;
+                       len -= written;
+                       continue;
+               }
+               if (!*buf) {
+                       buf++;
+                       len--;
+                       this_header++;
+                       continue;
+               }
+               this_header = 0;
+               decompress = decompress_method(buf, len, &compress_name);
+               pr_debug("Detected %s compressed data\n", compress_name);
+               if (decompress) {
+                       int res = decompress(buf, len, NULL, flush_buffer, NULL,
+                                  &my_inptr, error);
+                       if (res)
+                               error("decompressor failed");
+               } else if (compress_name) {
+                       if (!message) {
+                               snprintf(msg_buf, sizeof msg_buf,
+                                        "compression method %s not configured",
+                                        compress_name);
+                               message = msg_buf;
+                       }
+               } else
+                       error("junk in compressed archive");
+               if (state != Reset)
+                       error("junk in compressed archive");
+               this_header = saved_offset + my_inptr;
+               buf += my_inptr;
+               len -= my_inptr;
+       }
+       dir_utime();
+       kfree(name_buf);
+       kfree(symlink_buf);
+       kfree(header_buf);
+       return message;
+}
+
+static int __initdata do_retain_initrd;
+
+static int __init retain_initrd_param(char *str)
+{
+       if (*str)
+               return 0;
+       do_retain_initrd = 1;
+       return 1;
+}
+__setup("retain_initrd", retain_initrd_param);
+
+extern char __initramfs_start[];
+extern unsigned long __initramfs_size;
+#include <linux/initrd.h>
+#include <linux/kexec.h>
+
+static void __init free_initrd(void)
+{
+#ifdef CONFIG_KEXEC
+       unsigned long crashk_start = (unsigned long)__va(crashk_res.start);
+       unsigned long crashk_end   = (unsigned long)__va(crashk_res.end);
+#endif
+       if (do_retain_initrd)
+               goto skip;
+
+#ifdef CONFIG_KEXEC
+       /*
+        * If the initrd region is overlapped with crashkernel reserved region,
+        * free only memory that is not part of crashkernel region.
+        */
+       if (initrd_start < crashk_end && initrd_end > crashk_start) {
+               /*
+                * Initialize initrd memory region since the kexec boot does
+                * not do.
+                */
+               memset((void *)initrd_start, 0, initrd_end - initrd_start);
+               if (initrd_start < crashk_start)
+                       free_initrd_mem(initrd_start, crashk_start);
+               if (initrd_end > crashk_end)
+                       free_initrd_mem(crashk_end, initrd_end);
+       } else
+#endif
+               free_initrd_mem(initrd_start, initrd_end);
+skip:
+       initrd_start = 0;
+       initrd_end = 0;
+}
+
+#ifdef CONFIG_BLK_DEV_RAM
+#define BUF_SIZE 1024
+static void __init clean_rootfs(void)
+{
+       int fd;
+       void *buf;
+       struct linux_dirent64 *dirp;
+       int num;
+
+       fd = sys_open("/", O_RDONLY, 0);
+       WARN_ON(fd < 0);
+       if (fd < 0)
+               return;
+       buf = kzalloc(BUF_SIZE, GFP_KERNEL);
+       WARN_ON(!buf);
+       if (!buf) {
+               sys_close(fd);
+               return;
+       }
+
+       dirp = buf;
+       num = sys_getdents64(fd, dirp, BUF_SIZE);
+       while (num > 0) {
+               while (num > 0) {
+                       struct stat st;
+                       int ret;
+
+                       ret = sys_newlstat(dirp->d_name, &st);
+                       WARN_ON_ONCE(ret);
+                       if (!ret) {
+                               if (S_ISDIR(st.st_mode))
+                                       sys_rmdir(dirp->d_name);
+                               else
+                                       sys_unlink(dirp->d_name);
+                       }
+
+                       num -= dirp->d_reclen;
+                       dirp = (void *)dirp + dirp->d_reclen;
+               }
+               dirp = buf;
+               memset(buf, 0, BUF_SIZE);
+               num = sys_getdents64(fd, dirp, BUF_SIZE);
+       }
+
+       sys_close(fd);
+       kfree(buf);
+}
+#endif
+
+static int __init populate_rootfs(void)
+{
+       char *err = unpack_to_rootfs(__initramfs_start, __initramfs_size);
+       if (err)
+               panic("%s", err); /* Failed to decompress INTERNAL initramfs */
+       if (initrd_start) {
+#ifdef CONFIG_BLK_DEV_RAM
+               int fd;
+               printk(KERN_INFO "Trying to unpack rootfs image as initramfs...\n");
+               err = unpack_to_rootfs((char *)initrd_start,
+                       initrd_end - initrd_start);
+               if (!err) {
+                       free_initrd();
+                       goto done;
+               } else {
+                       clean_rootfs();
+                       unpack_to_rootfs(__initramfs_start, __initramfs_size);
+               }
+               printk(KERN_INFO "rootfs image is not initramfs (%s)"
+                               "; looks like an initrd\n", err);
+               fd = sys_open("/initrd.image",
+                             O_WRONLY|O_CREAT, 0700);
+               if (fd >= 0) {
+                       ssize_t written = xwrite(fd, (char *)initrd_start,
+                                               initrd_end - initrd_start);
+
+                       if (written != initrd_end - initrd_start)
+                               pr_err("/initrd.image: incomplete write (%zd != %ld)\n",
+                                      written, initrd_end - initrd_start);
+
+                       sys_close(fd);
+                       free_initrd();
+               }
+       done:
+#else
+               printk(KERN_INFO "Unpacking initramfs...\n");
+               err = unpack_to_rootfs((char *)initrd_start,
+                       initrd_end - initrd_start);
+               if (err)
+                       printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);
+               free_initrd();
+#endif
+               /*
+                * Try loading default modules from initramfs.  This gives
+                * us a chance to load before device_initcalls.
+                */
+               load_default_modules();
+       }
+       return 0;
+}
+rootfs_initcall(populate_rootfs);