-/*
+/*
* Copyright (C) 2000 Jeff Dike (jdike@karaya.com)
* Licensed under the GPL
*/
* James McMechan
*/
-#define MAJOR_NR UBD_MAJOR
#define UBD_SHIFT 4
-#include "linux/config.h"
+#include "linux/kernel.h"
#include "linux/module.h"
#include "linux/blkdev.h"
+#include "linux/ata.h"
#include "linux/hdreg.h"
#include "linux/init.h"
-#include "linux/devfs_fs_kernel.h"
#include "linux/cdrom.h"
#include "linux/proc_fs.h"
+#include "linux/seq_file.h"
#include "linux/ctype.h"
#include "linux/capability.h"
#include "linux/mm.h"
#include "linux/blkpg.h"
#include "linux/genhd.h"
#include "linux/spinlock.h"
+#include "linux/platform_device.h"
+#include "linux/scatterlist.h"
#include "asm/segment.h"
#include "asm/uaccess.h"
#include "asm/irq.h"
#include "asm/types.h"
#include "asm/tlbflush.h"
-#include "user_util.h"
#include "mem_user.h"
#include "kern_util.h"
#include "kern.h"
#include "irq_user.h"
#include "irq_kern.h"
#include "ubd_user.h"
-#include "2_5compat.h"
#include "os.h"
#include "mem.h"
#include "mem_kern.h"
#include "cow.h"
-enum ubd_req { UBD_READ, UBD_WRITE, UBD_MMAP };
+enum ubd_req { UBD_READ, UBD_WRITE };
struct io_thread_req {
+ struct request *req;
enum ubd_req op;
int fds[2];
unsigned long offsets[2];
unsigned long sector_mask;
unsigned long long cow_offset;
unsigned long bitmap_words[2];
- int map_fd;
- unsigned long long map_offset;
int error;
};
-extern int open_ubd_file(char *file, struct openflags *openflags,
- char **backing_file_out, int *bitmap_offset_out,
- unsigned long *bitmap_len_out, int *data_offset_out,
- int *create_cow_out);
-extern int create_cow_file(char *cow_file, char *backing_file,
- struct openflags flags, int sectorsize,
- int alignment, int *bitmap_offset_out,
- unsigned long *bitmap_len_out,
- int *data_offset_out);
-extern int read_cow_bitmap(int fd, void *buf, int offset, int len);
-extern void do_io(struct io_thread_req *req);
-
static inline int ubd_test_bit(__u64 bit, unsigned char *data)
{
__u64 n;
bits = sizeof(data[0]) * 8;
n = bit / bits;
off = bit % bits;
- return((data[n] & (1 << off)) != 0);
+ return (data[n] & (1 << off)) != 0;
}
static inline void ubd_set_bit(__u64 bit, unsigned char *data)
#define DRIVER_NAME "uml-blkdev"
-static DEFINE_SPINLOCK(ubd_io_lock);
-static DEFINE_SPINLOCK(ubd_lock);
-
-static void (*do_ubd)(void);
+static DEFINE_MUTEX(ubd_lock);
-static int ubd_open(struct inode * inode, struct file * filp);
-static int ubd_release(struct inode * inode, struct file * file);
-static int ubd_ioctl(struct inode * inode, struct file * file,
+static int ubd_open(struct block_device *bdev, fmode_t mode);
+static int ubd_release(struct gendisk *disk, fmode_t mode);
+static int ubd_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg);
+static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo);
-#define MAX_DEV (8)
+#define MAX_DEV (16)
-/* Changed in early boot */
-static int ubd_do_mmap = 0;
-#define UBD_MMAP_BLOCK_SIZE PAGE_SIZE
-
-static struct block_device_operations ubd_blops = {
+static const struct block_device_operations ubd_blops = {
.owner = THIS_MODULE,
.open = ubd_open,
.release = ubd_release,
.ioctl = ubd_ioctl,
+ .getgeo = ubd_getgeo,
};
-/* Protected by the queue_lock */
-static request_queue_t *ubd_queue;
-
/* Protected by ubd_lock */
-static int fake_major = MAJOR_NR;
-
+static int fake_major = UBD_MAJOR;
static struct gendisk *ubd_gendisk[MAX_DEV];
static struct gendisk *fake_gendisk[MAX_DEV];
-
+
#ifdef CONFIG_BLK_DEV_UBD_SYNC
#define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 1, .c = 0, \
.cl = 1 })
#define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 0, .c = 0, \
.cl = 1 })
#endif
-
-/* Not protected - changed only in ubd_setup_common and then only to
- * to enable O_SYNC.
- */
static struct openflags global_openflags = OPEN_FLAGS;
struct cow {
- /* This is the backing file, actually */
+ /* backing file name */
char *file;
+ /* backing file fd */
int fd;
unsigned long *bitmap;
unsigned long bitmap_len;
int bitmap_offset;
- int data_offset;
+ int data_offset;
};
+#define MAX_SG 64
+
struct ubd {
+ struct list_head restart;
+ /* name (and fd, below) of the file opened for writing, either the
+ * backing or the cow file. */
char *file;
int count;
int fd;
__u64 size;
struct openflags boot_openflags;
struct openflags openflags;
- int no_cow;
+ unsigned shared:1;
+ unsigned no_cow:1;
struct cow cow;
struct platform_device pdev;
-
- int map_writes;
- int map_reads;
- int nomap_writes;
- int nomap_reads;
- int write_maps;
+ struct request_queue *queue;
+ spinlock_t lock;
+ struct scatterlist sg[MAX_SG];
+ struct request *request;
+ int start_sg, end_sg;
};
#define DEFAULT_COW { \
.file = NULL, \
- .fd = -1, \
- .bitmap = NULL, \
+ .fd = -1, \
+ .bitmap = NULL, \
.bitmap_offset = 0, \
- .data_offset = 0, \
+ .data_offset = 0, \
}
#define DEFAULT_UBD { \
.size = -1, \
.boot_openflags = OPEN_FLAGS, \
.openflags = OPEN_FLAGS, \
- .no_cow = 0, \
- .cow = DEFAULT_COW, \
- .map_writes = 0, \
- .map_reads = 0, \
- .nomap_writes = 0, \
- .nomap_reads = 0, \
- .write_maps = 0, \
-}
-
-struct ubd ubd_dev[MAX_DEV] = { [ 0 ... MAX_DEV - 1 ] = DEFAULT_UBD };
-
-static int ubd0_init(void)
-{
- struct ubd *dev = &ubd_dev[0];
-
- if(dev->file == NULL)
- dev->file = "root_fs";
- return(0);
+ .no_cow = 0, \
+ .shared = 0, \
+ .cow = DEFAULT_COW, \
+ .lock = SPIN_LOCK_UNLOCKED, \
+ .request = NULL, \
+ .start_sg = 0, \
+ .end_sg = 0, \
}
-__initcall(ubd0_init);
+/* Protected by ubd_lock */
+static struct ubd ubd_devs[MAX_DEV] = { [0 ... MAX_DEV - 1] = DEFAULT_UBD };
/* Only changed by fake_ide_setup which is a setup */
static int fake_ide = 0;
proc_ide = proc_mkdir("ide0", proc_ide_root);
}
-static int proc_ide_read_media(char *page, char **start, off_t off, int count,
- int *eof, void *data)
+static int fake_ide_media_proc_show(struct seq_file *m, void *v)
{
- int len;
-
- strcpy(page, "disk\n");
- len = strlen("disk\n");
- len -= off;
- if (len < count){
- *eof = 1;
- if (len <= 0) return 0;
- }
- else len = count;
- *start = page + off;
- return len;
+ seq_puts(m, "disk\n");
+ return 0;
}
-static void make_ide_entries(char *dev_name)
+static int fake_ide_media_proc_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, fake_ide_media_proc_show, NULL);
+}
+
+static const struct file_operations fake_ide_media_proc_fops = {
+ .owner = THIS_MODULE,
+ .open = fake_ide_media_proc_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+static void make_ide_entries(const char *dev_name)
{
struct proc_dir_entry *dir, *ent;
char name[64];
dir = proc_mkdir(dev_name, proc_ide);
if(!dir) return;
- ent = create_proc_entry("media", S_IFREG|S_IRUGO, dir);
+ ent = proc_create("media", S_IRUGO, dir, &fake_ide_media_proc_fops);
if(!ent) return;
- ent->nlink = 1;
- ent->data = NULL;
- ent->read_proc = proc_ide_read_media;
- ent->write_proc = NULL;
- sprintf(name,"ide0/%s", dev_name);
+ snprintf(name, sizeof(name), "ide0/%s", dev_name);
proc_symlink(dev_name, proc_ide_root, name);
}
static int fake_ide_setup(char *str)
{
fake_ide = 1;
- return(1);
+ return 1;
}
__setup("fake_ide", fake_ide_setup);
if(isdigit(*str)) {
n = simple_strtoul(str, &end, 0);
if(end == str)
- return(-1);
+ return -1;
*ptr = end;
}
- else if (('a' <= *str) && (*str <= 'h')) {
+ else if (('a' <= *str) && (*str <= 'z')) {
n = *str - 'a';
str++;
*ptr = str;
}
- return(n);
+ return n;
}
-static int ubd_setup_common(char *str, int *index_out)
+/* If *index_out == -1 at exit, the passed option was a general one;
+ * otherwise, the str pointer is used (and owned) inside ubd_devs array, so it
+ * should not be freed on exit.
+ */
+static int ubd_setup_common(char *str, int *index_out, char **error_out)
{
- struct ubd *dev;
+ struct ubd *ubd_dev;
struct openflags flags = global_openflags;
char *backing_file;
- int n, err, i;
+ int n, err = 0, i;
if(index_out) *index_out = -1;
n = *str;
int major;
str++;
- if(!strcmp(str, "mmap")){
- CHOOSE_MODE(printk("mmap not supported by the ubd "
- "driver in tt mode\n"),
- ubd_do_mmap = 1);
- return(0);
- }
-
if(!strcmp(str, "sync")){
global_openflags = of_sync(global_openflags);
- return(0);
+ goto out1;
}
+
+ err = -EINVAL;
major = simple_strtoul(str, &end, 0);
if((*end != '\0') || (end == str)){
- printk(KERN_ERR
- "ubd_setup : didn't parse major number\n");
- return(1);
+ *error_out = "Didn't parse major number";
+ goto out1;
+ }
+
+ mutex_lock(&ubd_lock);
+ if (fake_major != UBD_MAJOR) {
+ *error_out = "Can't assign a fake major twice";
+ goto out1;
}
- err = 1;
- spin_lock(&ubd_lock);
- if(fake_major != MAJOR_NR){
- printk(KERN_ERR "Can't assign a fake major twice\n");
- goto out1;
- }
-
- fake_major = major;
+ fake_major = major;
printk(KERN_INFO "Setting extra ubd major number to %d\n",
major);
- err = 0;
- out1:
- spin_unlock(&ubd_lock);
- return(err);
+ err = 0;
+ out1:
+ mutex_unlock(&ubd_lock);
+ return err;
}
n = parse_unit(&str);
if(n < 0){
- printk(KERN_ERR "ubd_setup : couldn't parse unit number "
- "'%s'\n", str);
- return(1);
+ *error_out = "Couldn't parse device number";
+ return -EINVAL;
}
if(n >= MAX_DEV){
- printk(KERN_ERR "ubd_setup : index %d out of range "
- "(%d devices, from 0 to %d)\n", n, MAX_DEV, MAX_DEV - 1);
- return(1);
+ *error_out = "Device number out of range";
+ return 1;
}
- err = 1;
- spin_lock(&ubd_lock);
+ err = -EBUSY;
+ mutex_lock(&ubd_lock);
- dev = &ubd_dev[n];
- if(dev->file != NULL){
- printk(KERN_ERR "ubd_setup : device already configured\n");
+ ubd_dev = &ubd_devs[n];
+ if(ubd_dev->file != NULL){
+ *error_out = "Device is already configured";
goto out;
}
if (index_out)
*index_out = n;
- for (i = 0; i < 4; i++) {
+ err = -EINVAL;
+ for (i = 0; i < sizeof("rscd="); i++) {
switch (*str) {
case 'r':
flags.w = 0;
flags.s = 1;
break;
case 'd':
- dev->no_cow = 1;
+ ubd_dev->no_cow = 1;
+ break;
+ case 'c':
+ ubd_dev->shared = 1;
break;
case '=':
str++;
goto break_loop;
default:
- printk(KERN_ERR "ubd_setup : Expected '=' or flag letter (r,s or d)\n");
+ *error_out = "Expected '=' or flag letter "
+ "(r, s, c, or d)";
goto out;
}
str++;
}
- if (*str == '=')
- printk(KERN_ERR "ubd_setup : Too many flags specified\n");
- else
- printk(KERN_ERR "ubd_setup : Expected '='\n");
+ if (*str == '=')
+ *error_out = "Too many flags specified";
+ else
+ *error_out = "Missing '='";
goto out;
break_loop:
- err = 0;
backing_file = strchr(str, ',');
- if (!backing_file) {
+ if (backing_file == NULL)
backing_file = strchr(str, ':');
- }
- if(backing_file){
- if(dev->no_cow)
- printk(KERN_ERR "Can't specify both 'd' and a "
- "cow file\n");
+ if(backing_file != NULL){
+ if(ubd_dev->no_cow){
+ *error_out = "Can't specify both 'd' and a cow file";
+ goto out;
+ }
else {
*backing_file = '\0';
backing_file++;
}
}
- dev->file = str;
- dev->cow.file = backing_file;
- dev->boot_openflags = flags;
+ err = 0;
+ ubd_dev->file = str;
+ ubd_dev->cow.file = backing_file;
+ ubd_dev->boot_openflags = flags;
out:
- spin_unlock(&ubd_lock);
- return(err);
+ mutex_unlock(&ubd_lock);
+ return err;
}
static int ubd_setup(char *str)
{
- ubd_setup_common(str, NULL);
- return(1);
+ char *error;
+ int err;
+
+ err = ubd_setup_common(str, NULL, &error);
+ if(err)
+ printk(KERN_ERR "Failed to initialize device with \"%s\" : "
+ "%s\n", str, error);
+ return 1;
}
__setup("ubd", ubd_setup);
" use either a ':' or a ',': the first one allows writing things like;\n"
" ubd0=~/Uml/root_cow:~/Uml/root_backing_file\n"
" while with a ',' the shell would not expand the 2nd '~'.\n"
-" When using only one filename, UML will detect whether to thread it like\n"
+" When using only one filename, UML will detect whether to treat it like\n"
" a COW file or a backing file. To override this detection, add the 'd'\n"
" flag:\n"
" ubd0d=BackingFile\n"
" machine by running 'dd' on the device. <n> must be in the range\n"
" 0 to 7. Appending an 'r' to the number will cause that device\n"
" to be mounted read-only. For example ubd1r=./ext_fs. Appending\n"
-" an 's' will cause data to be written to disk on the host immediately.\n\n"
+" an 's' will cause data to be written to disk on the host immediately.\n"
+" 'c' will cause the device to be treated as being shared between multiple\n"
+" UMLs and file locking will be turned off - this is appropriate for a\n"
+" cluster filesystem and inappropriate at almost all other times.\n\n"
);
static int udb_setup(char *str)
{
printk("udb%s specified on command line is almost certainly a ubd -> "
"udb TYPO\n", str);
- return(1);
+ return 1;
}
__setup("udb", udb_setup);
__uml_help(udb_setup,
"udb\n"
-" This option is here solely to catch ubd -> udb typos, which can be\n\n"
-" to impossible to catch visually unless you specifically look for\n\n"
-" them. The only result of any option starting with 'udb' is an error\n\n"
+" This option is here solely to catch ubd -> udb typos, which can be\n"
+" to impossible to catch visually unless you specifically look for\n"
+" them. The only result of any option starting with 'udb' is an error\n"
" in the boot output.\n\n"
);
-static int fakehd_set = 0;
-static int fakehd(char *str)
+static void do_ubd_request(struct request_queue * q);
+
+/* Only changed by ubd_init, which is an initcall. */
+static int thread_fd = -1;
+static LIST_HEAD(restart);
+
+/* XXX - move this inside ubd_intr. */
+/* Called without dev->lock held, and only in interrupt context. */
+static void ubd_handler(void)
{
- printk(KERN_INFO "fakehd : Changing ubd name to \"hd\".\n");
- fakehd_set = 1;
- return 1;
-}
+ struct io_thread_req *req;
+ struct ubd *ubd;
+ struct list_head *list, *next_ele;
+ unsigned long flags;
+ int n;
-__setup("fakehd", fakehd);
-__uml_help(fakehd,
-"fakehd\n"
-" Change the ubd device name to \"hd\".\n\n"
-);
+ while(1){
+ n = os_read_file(thread_fd, &req,
+ sizeof(struct io_thread_req *));
+ if(n != sizeof(req)){
+ if(n == -EAGAIN)
+ break;
+ printk(KERN_ERR "spurious interrupt in ubd_handler, "
+ "err = %d\n", -n);
+ return;
+ }
-static void do_ubd_request(request_queue_t * q);
+ blk_end_request(req->req, 0, req->length);
+ kfree(req);
+ }
+ reactivate_fd(thread_fd, UBD_IRQ);
+
+ list_for_each_safe(list, next_ele, &restart){
+ ubd = container_of(list, struct ubd, restart);
+ list_del_init(&ubd->restart);
+ spin_lock_irqsave(&ubd->lock, flags);
+ do_ubd_request(ubd->queue);
+ spin_unlock_irqrestore(&ubd->lock, flags);
+ }
+}
+
+static irqreturn_t ubd_intr(int irq, void *dev)
+{
+ ubd_handler();
+ return IRQ_HANDLED;
+}
/* Only changed by ubd_init, which is an initcall. */
-int thread_fd = -1;
+static int io_pid = -1;
-/* Changed by ubd_handler, which is serialized because interrupts only
- * happen on CPU 0.
- */
-int intr_count = 0;
+static void kill_io_thread(void)
+{
+ if(io_pid != -1)
+ os_kill_process(io_pid, 1);
+}
-/* call ubd_finish if you need to serialize */
-static void __ubd_finish(struct request *req, int error)
+__uml_exitcall(kill_io_thread);
+
+static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out)
{
- int nsect;
+ char *file;
- if(error){
- end_request(req, 0);
- return;
- }
- nsect = req->current_nr_sectors;
- req->sector += nsect;
- req->buffer += nsect << 9;
- req->errors = 0;
- req->nr_sectors -= nsect;
- req->current_nr_sectors = 0;
- end_request(req, 1);
+ file = ubd_dev->cow.file ? ubd_dev->cow.file : ubd_dev->file;
+ return os_file_size(file, size_out);
}
-static inline void ubd_finish(struct request *req, int error)
+static int read_cow_bitmap(int fd, void *buf, int offset, int len)
{
- spin_lock(&ubd_io_lock);
- __ubd_finish(req, error);
- spin_unlock(&ubd_io_lock);
+ int err;
+
+ err = os_seek_file(fd, offset);
+ if (err < 0)
+ return err;
+
+ err = os_read_file(fd, buf, len);
+ if (err < 0)
+ return err;
+
+ return 0;
}
-/* Called without ubd_io_lock held */
-static void ubd_handler(void)
+static int backing_file_mismatch(char *file, __u64 size, time_t mtime)
{
- struct io_thread_req req;
- struct request *rq = elv_next_request(ubd_queue);
- int n, err;
-
- do_ubd = NULL;
- intr_count++;
- n = os_read_file(thread_fd, &req, sizeof(req));
- if(n != sizeof(req)){
- printk(KERN_ERR "Pid %d - spurious interrupt in ubd_handler, "
- "err = %d\n", os_getpid(), -n);
- spin_lock(&ubd_io_lock);
- end_request(rq, 0);
- spin_unlock(&ubd_io_lock);
- return;
+ unsigned long modtime;
+ unsigned long long actual;
+ int err;
+
+ err = os_file_modtime(file, &modtime);
+ if (err < 0) {
+ printk(KERN_ERR "Failed to get modification time of backing "
+ "file \"%s\", err = %d\n", file, -err);
+ return err;
}
-
- if((req.op != UBD_MMAP) &&
- ((req.offset != ((__u64) (rq->sector)) << 9) ||
- (req.length != (rq->current_nr_sectors) << 9)))
- panic("I/O op mismatch");
-
- if(req.map_fd != -1){
- err = physmem_subst_mapping(req.buffer, req.map_fd,
- req.map_offset, 1);
- if(err)
- printk("ubd_handler - physmem_subst_mapping failed, "
- "err = %d\n", -err);
+
+ err = os_file_size(file, &actual);
+ if (err < 0) {
+ printk(KERN_ERR "Failed to get size of backing file \"%s\", "
+ "err = %d\n", file, -err);
+ return err;
}
- ubd_finish(rq, req.error);
- reactivate_fd(thread_fd, UBD_IRQ);
- do_ubd_request(ubd_queue);
+ if (actual != size) {
+ /*__u64 can be a long on AMD64 and with %lu GCC complains; so
+ * the typecast.*/
+ printk(KERN_ERR "Size mismatch (%llu vs %llu) of COW header "
+ "vs backing file\n", (unsigned long long) size, actual);
+ return -EINVAL;
+ }
+ if (modtime != mtime) {
+ printk(KERN_ERR "mtime mismatch (%ld vs %ld) of COW header vs "
+ "backing file\n", mtime, modtime);
+ return -EINVAL;
+ }
+ return 0;
}
-static irqreturn_t ubd_intr(int irq, void *dev, struct pt_regs *unused)
+static int path_requires_switch(char *from_cmdline, char *from_cow, char *cow)
{
- ubd_handler();
- return(IRQ_HANDLED);
-}
+ struct uml_stat buf1, buf2;
+ int err;
-/* Only changed by ubd_init, which is an initcall. */
-static int io_pid = -1;
+ if (from_cmdline == NULL)
+ return 0;
+ if (!strcmp(from_cmdline, from_cow))
+ return 0;
-void kill_io_thread(void)
-{
- if(io_pid != -1)
- os_kill_process(io_pid, 1);
+ err = os_stat_file(from_cmdline, &buf1);
+ if (err < 0) {
+ printk(KERN_ERR "Couldn't stat '%s', err = %d\n", from_cmdline,
+ -err);
+ return 0;
+ }
+ err = os_stat_file(from_cow, &buf2);
+ if (err < 0) {
+ printk(KERN_ERR "Couldn't stat '%s', err = %d\n", from_cow,
+ -err);
+ return 1;
+ }
+ if ((buf1.ust_dev == buf2.ust_dev) && (buf1.ust_ino == buf2.ust_ino))
+ return 0;
+
+ printk(KERN_ERR "Backing file mismatch - \"%s\" requested, "
+ "\"%s\" specified in COW header of \"%s\"\n",
+ from_cmdline, from_cow, cow);
+ return 1;
}
-__uml_exitcall(kill_io_thread);
+static int open_ubd_file(char *file, struct openflags *openflags, int shared,
+ char **backing_file_out, int *bitmap_offset_out,
+ unsigned long *bitmap_len_out, int *data_offset_out,
+ int *create_cow_out)
+{
+ time_t mtime;
+ unsigned long long size;
+ __u32 version, align;
+ char *backing_file;
+ int fd, err, sectorsize, asked_switch, mode = 0644;
+
+ fd = os_open_file(file, *openflags, mode);
+ if (fd < 0) {
+ if ((fd == -ENOENT) && (create_cow_out != NULL))
+ *create_cow_out = 1;
+ if (!openflags->w ||
+ ((fd != -EROFS) && (fd != -EACCES)))
+ return fd;
+ openflags->w = 0;
+ fd = os_open_file(file, *openflags, mode);
+ if (fd < 0)
+ return fd;
+ }
+
+ if (shared)
+ printk(KERN_INFO "Not locking \"%s\" on the host\n", file);
+ else {
+ err = os_lock_file(fd, openflags->w);
+ if (err < 0) {
+ printk(KERN_ERR "Failed to lock '%s', err = %d\n",
+ file, -err);
+ goto out_close;
+ }
+ }
+
+ /* Successful return case! */
+ if (backing_file_out == NULL)
+ return fd;
+
+ err = read_cow_header(file_reader, &fd, &version, &backing_file, &mtime,
+ &size, §orsize, &align, bitmap_offset_out);
+ if (err && (*backing_file_out != NULL)) {
+ printk(KERN_ERR "Failed to read COW header from COW file "
+ "\"%s\", errno = %d\n", file, -err);
+ goto out_close;
+ }
+ if (err)
+ return fd;
+
+ asked_switch = path_requires_switch(*backing_file_out, backing_file,
+ file);
+
+ /* Allow switching only if no mismatch. */
+ if (asked_switch && !backing_file_mismatch(*backing_file_out, size,
+ mtime)) {
+ printk(KERN_ERR "Switching backing file to '%s'\n",
+ *backing_file_out);
+ err = write_cow_header(file, fd, *backing_file_out,
+ sectorsize, align, &size);
+ if (err) {
+ printk(KERN_ERR "Switch failed, errno = %d\n", -err);
+ goto out_close;
+ }
+ } else {
+ *backing_file_out = backing_file;
+ err = backing_file_mismatch(*backing_file_out, size, mtime);
+ if (err)
+ goto out_close;
+ }
+
+ cow_sizes(version, size, sectorsize, align, *bitmap_offset_out,
+ bitmap_len_out, data_offset_out);
+
+ return fd;
+ out_close:
+ os_close_file(fd);
+ return err;
+}
-static int ubd_file_size(struct ubd *dev, __u64 *size_out)
+static int create_cow_file(char *cow_file, char *backing_file,
+ struct openflags flags,
+ int sectorsize, int alignment, int *bitmap_offset_out,
+ unsigned long *bitmap_len_out, int *data_offset_out)
{
- char *file;
+ int err, fd;
- file = dev->cow.file ? dev->cow.file : dev->file;
- return(os_file_size(file, size_out));
+ flags.c = 1;
+ fd = open_ubd_file(cow_file, &flags, 0, NULL, NULL, NULL, NULL, NULL);
+ if (fd < 0) {
+ err = fd;
+ printk(KERN_ERR "Open of COW file '%s' failed, errno = %d\n",
+ cow_file, -err);
+ goto out;
+ }
+
+ err = init_cow_file(fd, cow_file, backing_file, sectorsize, alignment,
+ bitmap_offset_out, bitmap_len_out,
+ data_offset_out);
+ if (!err)
+ return fd;
+ os_close_file(fd);
+ out:
+ return err;
}
-static void ubd_close(struct ubd *dev)
+static void ubd_close_dev(struct ubd *ubd_dev)
{
- if(ubd_do_mmap)
- physmem_forget_descriptor(dev->fd);
- os_close_file(dev->fd);
- if(dev->cow.file == NULL)
+ os_close_file(ubd_dev->fd);
+ if(ubd_dev->cow.file == NULL)
return;
- if(ubd_do_mmap)
- physmem_forget_descriptor(dev->cow.fd);
- os_close_file(dev->cow.fd);
- vfree(dev->cow.bitmap);
- dev->cow.bitmap = NULL;
+ os_close_file(ubd_dev->cow.fd);
+ vfree(ubd_dev->cow.bitmap);
+ ubd_dev->cow.bitmap = NULL;
}
-static int ubd_open_dev(struct ubd *dev)
+static int ubd_open_dev(struct ubd *ubd_dev)
{
struct openflags flags;
char **back_ptr;
int err, create_cow, *create_ptr;
+ int fd;
- dev->openflags = dev->boot_openflags;
+ ubd_dev->openflags = ubd_dev->boot_openflags;
create_cow = 0;
- create_ptr = (dev->cow.file != NULL) ? &create_cow : NULL;
- back_ptr = dev->no_cow ? NULL : &dev->cow.file;
- dev->fd = open_ubd_file(dev->file, &dev->openflags, back_ptr,
- &dev->cow.bitmap_offset, &dev->cow.bitmap_len,
- &dev->cow.data_offset, create_ptr);
-
- if((dev->fd == -ENOENT) && create_cow){
- dev->fd = create_cow_file(dev->file, dev->cow.file,
- dev->openflags, 1 << 9, PAGE_SIZE,
- &dev->cow.bitmap_offset,
- &dev->cow.bitmap_len,
- &dev->cow.data_offset);
- if(dev->fd >= 0){
+ create_ptr = (ubd_dev->cow.file != NULL) ? &create_cow : NULL;
+ back_ptr = ubd_dev->no_cow ? NULL : &ubd_dev->cow.file;
+
+ fd = open_ubd_file(ubd_dev->file, &ubd_dev->openflags, ubd_dev->shared,
+ back_ptr, &ubd_dev->cow.bitmap_offset,
+ &ubd_dev->cow.bitmap_len, &ubd_dev->cow.data_offset,
+ create_ptr);
+
+ if((fd == -ENOENT) && create_cow){
+ fd = create_cow_file(ubd_dev->file, ubd_dev->cow.file,
+ ubd_dev->openflags, 1 << 9, PAGE_SIZE,
+ &ubd_dev->cow.bitmap_offset,
+ &ubd_dev->cow.bitmap_len,
+ &ubd_dev->cow.data_offset);
+ if(fd >= 0){
printk(KERN_INFO "Creating \"%s\" as COW file for "
- "\"%s\"\n", dev->file, dev->cow.file);
+ "\"%s\"\n", ubd_dev->file, ubd_dev->cow.file);
}
}
- if(dev->fd < 0){
- printk("Failed to open '%s', errno = %d\n", dev->file,
- -dev->fd);
- return(dev->fd);
+ if(fd < 0){
+ printk("Failed to open '%s', errno = %d\n", ubd_dev->file,
+ -fd);
+ return fd;
}
+ ubd_dev->fd = fd;
+
+ if(ubd_dev->cow.file != NULL){
+ blk_queue_max_sectors(ubd_dev->queue, 8 * sizeof(long));
- if(dev->cow.file != NULL){
err = -ENOMEM;
- dev->cow.bitmap = (void *) vmalloc(dev->cow.bitmap_len);
- if(dev->cow.bitmap == NULL){
+ ubd_dev->cow.bitmap = vmalloc(ubd_dev->cow.bitmap_len);
+ if(ubd_dev->cow.bitmap == NULL){
printk(KERN_ERR "Failed to vmalloc COW bitmap\n");
goto error;
}
flush_tlb_kernel_vm();
- err = read_cow_bitmap(dev->fd, dev->cow.bitmap,
- dev->cow.bitmap_offset,
- dev->cow.bitmap_len);
+ err = read_cow_bitmap(ubd_dev->fd, ubd_dev->cow.bitmap,
+ ubd_dev->cow.bitmap_offset,
+ ubd_dev->cow.bitmap_len);
if(err < 0)
goto error;
- flags = dev->openflags;
+ flags = ubd_dev->openflags;
flags.w = 0;
- err = open_ubd_file(dev->cow.file, &flags, NULL, NULL, NULL,
- NULL, NULL);
+ err = open_ubd_file(ubd_dev->cow.file, &flags, ubd_dev->shared, NULL,
+ NULL, NULL, NULL, NULL);
if(err < 0) goto error;
- dev->cow.fd = err;
+ ubd_dev->cow.fd = err;
}
- return(0);
+ return 0;
error:
- os_close_file(dev->fd);
- return(err);
+ os_close_file(ubd_dev->fd);
+ return err;
}
-static int ubd_new_disk(int major, u64 size, int unit,
- struct gendisk **disk_out)
-
+static void ubd_device_release(struct device *dev)
+{
+ struct ubd *ubd_dev = dev_get_drvdata(dev);
+
+ blk_cleanup_queue(ubd_dev->queue);
+ *ubd_dev = ((struct ubd) DEFAULT_UBD);
+}
+
+static int ubd_disk_register(int major, u64 size, int unit,
+ struct gendisk **disk_out)
{
struct gendisk *disk;
- char from[sizeof("ubd/nnnnn\0")], to[sizeof("discnnnnn/disc\0")];
- int err;
disk = alloc_disk(1 << UBD_SHIFT);
if(disk == NULL)
- return(-ENOMEM);
+ return -ENOMEM;
disk->major = major;
disk->first_minor = unit << UBD_SHIFT;
disk->fops = &ubd_blops;
set_capacity(disk, size / 512);
- if(major == MAJOR_NR){
+ if (major == UBD_MAJOR)
sprintf(disk->disk_name, "ubd%c", 'a' + unit);
- sprintf(disk->devfs_name, "ubd/disc%d", unit);
- sprintf(from, "ubd/%d", unit);
- sprintf(to, "disc%d/disc", unit);
- err = devfs_mk_symlink(from, to);
- if(err)
- printk("ubd_new_disk failed to make link from %s to "
- "%s, error = %d\n", from, to, err);
- }
- else {
+ else
sprintf(disk->disk_name, "ubd_fake%d", unit);
- sprintf(disk->devfs_name, "ubd_fake/disc%d", unit);
- }
/* sysfs register (not for ide fake devices) */
- if (major == MAJOR_NR) {
- ubd_dev[unit].pdev.id = unit;
- ubd_dev[unit].pdev.name = DRIVER_NAME;
- platform_device_register(&ubd_dev[unit].pdev);
- disk->driverfs_dev = &ubd_dev[unit].pdev.dev;
+ if (major == UBD_MAJOR) {
+ ubd_devs[unit].pdev.id = unit;
+ ubd_devs[unit].pdev.name = DRIVER_NAME;
+ ubd_devs[unit].pdev.dev.release = ubd_device_release;
+ dev_set_drvdata(&ubd_devs[unit].pdev.dev, &ubd_devs[unit]);
+ platform_device_register(&ubd_devs[unit].pdev);
+ disk->driverfs_dev = &ubd_devs[unit].pdev.dev;
}
- disk->private_data = &ubd_dev[unit];
- disk->queue = ubd_queue;
+ disk->private_data = &ubd_devs[unit];
+ disk->queue = ubd_devs[unit].queue;
add_disk(disk);
*disk_out = disk;
#define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9))
-static int ubd_add(int n)
+static int ubd_add(int n, char **error_out)
{
- struct ubd *dev = &ubd_dev[n];
- int err;
+ struct ubd *ubd_dev = &ubd_devs[n];
+ int err = 0;
+
+ if(ubd_dev->file == NULL)
+ goto out;
+
+ err = ubd_file_size(ubd_dev, &ubd_dev->size);
+ if(err < 0){
+ *error_out = "Couldn't determine size of device's file";
+ goto out;
+ }
- if(dev->file == NULL)
- return(-ENODEV);
+ ubd_dev->size = ROUND_BLOCK(ubd_dev->size);
- if (ubd_open_dev(dev))
- return(-ENODEV);
+ INIT_LIST_HEAD(&ubd_dev->restart);
+ sg_init_table(ubd_dev->sg, MAX_SG);
- err = ubd_file_size(dev, &dev->size);
- if(err < 0)
- return(err);
+ err = -ENOMEM;
+ ubd_dev->queue = blk_init_queue(do_ubd_request, &ubd_dev->lock);
+ if (ubd_dev->queue == NULL) {
+ *error_out = "Failed to initialize device queue";
+ goto out;
+ }
+ ubd_dev->queue->queuedata = ubd_dev;
- dev->size = ROUND_BLOCK(dev->size);
+ blk_queue_max_hw_segments(ubd_dev->queue, MAX_SG);
+ err = ubd_disk_register(UBD_MAJOR, ubd_dev->size, n, &ubd_gendisk[n]);
+ if(err){
+ *error_out = "Failed to register device";
+ goto out_cleanup;
+ }
- err = ubd_new_disk(MAJOR_NR, dev->size, n, &ubd_gendisk[n]);
- if(err)
- return(err);
-
- if(fake_major != MAJOR_NR)
- ubd_new_disk(fake_major, dev->size, n,
- &fake_gendisk[n]);
+ if (fake_major != UBD_MAJOR)
+ ubd_disk_register(fake_major, ubd_dev->size, n,
+ &fake_gendisk[n]);
- /* perhaps this should also be under the "if (fake_major)" above */
- /* using the fake_disk->disk_name and also the fakehd_set name */
+ /*
+ * Perhaps this should also be under the "if (fake_major)" above
+ * using the fake_disk->disk_name
+ */
if (fake_ide)
make_ide_entries(ubd_gendisk[n]->disk_name);
- ubd_close(dev);
- return 0;
+ err = 0;
+out:
+ return err;
+
+out_cleanup:
+ blk_cleanup_queue(ubd_dev->queue);
+ goto out;
}
-static int ubd_config(char *str)
+static int ubd_config(char *str, char **error_out)
{
- int n, err;
+ int n, ret;
- str = uml_strdup(str);
- if(str == NULL){
- printk(KERN_ERR "ubd_config failed to strdup string\n");
- return(1);
+ /* This string is possibly broken up and stored, so it's only
+ * freed if ubd_setup_common fails, or if only general options
+ * were set.
+ */
+ str = kstrdup(str, GFP_KERNEL);
+ if (str == NULL) {
+ *error_out = "Failed to allocate memory";
+ return -ENOMEM;
}
- err = ubd_setup_common(str, &n);
- if(err){
- kfree(str);
- return(-1);
+
+ ret = ubd_setup_common(str, &n, error_out);
+ if (ret)
+ goto err_free;
+
+ if (n == -1) {
+ ret = 0;
+ goto err_free;
}
- if(n == -1) return(0);
- spin_lock(&ubd_lock);
- err = ubd_add(n);
- if(err)
- ubd_dev[n].file = NULL;
- spin_unlock(&ubd_lock);
+ mutex_lock(&ubd_lock);
+ ret = ubd_add(n, error_out);
+ if (ret)
+ ubd_devs[n].file = NULL;
+ mutex_unlock(&ubd_lock);
- return(err);
+out:
+ return ret;
+
+err_free:
+ kfree(str);
+ goto out;
}
static int ubd_get_config(char *name, char *str, int size, char **error_out)
{
- struct ubd *dev;
+ struct ubd *ubd_dev;
int n, len = 0;
n = parse_unit(&name);
if((n >= MAX_DEV) || (n < 0)){
*error_out = "ubd_get_config : device number out of range";
- return(-1);
+ return -1;
}
- dev = &ubd_dev[n];
- spin_lock(&ubd_lock);
+ ubd_dev = &ubd_devs[n];
+ mutex_lock(&ubd_lock);
- if(dev->file == NULL){
+ if(ubd_dev->file == NULL){
CONFIG_CHUNK(str, size, len, "", 1);
goto out;
}
- CONFIG_CHUNK(str, size, len, dev->file, 0);
+ CONFIG_CHUNK(str, size, len, ubd_dev->file, 0);
- if(dev->cow.file != NULL){
+ if(ubd_dev->cow.file != NULL){
CONFIG_CHUNK(str, size, len, ",", 0);
- CONFIG_CHUNK(str, size, len, dev->cow.file, 1);
+ CONFIG_CHUNK(str, size, len, ubd_dev->cow.file, 1);
}
else CONFIG_CHUNK(str, size, len, "", 1);
out:
- spin_unlock(&ubd_lock);
- return(len);
+ mutex_unlock(&ubd_lock);
+ return len;
}
-static int ubd_remove(char *str)
+static int ubd_id(char **str, int *start_out, int *end_out)
{
- struct ubd *dev;
- int n, err = -ENODEV;
+ int n;
- n = parse_unit(&str);
+ n = parse_unit(str);
+ *start_out = 0;
+ *end_out = MAX_DEV - 1;
+ return n;
+}
- if((n < 0) || (n >= MAX_DEV))
- return(err);
+static int ubd_remove(int n, char **error_out)
+{
+ struct gendisk *disk = ubd_gendisk[n];
+ struct ubd *ubd_dev;
+ int err = -ENODEV;
- dev = &ubd_dev[n];
- if(dev->count > 0)
- return(-EBUSY); /* you cannot remove a open disk */
+ mutex_lock(&ubd_lock);
- err = 0;
- spin_lock(&ubd_lock);
+ ubd_dev = &ubd_devs[n];
- if(ubd_gendisk[n] == NULL)
+ if(ubd_dev->file == NULL)
+ goto out;
+
+ /* you cannot remove a open disk */
+ err = -EBUSY;
+ if(ubd_dev->count > 0)
goto out;
- del_gendisk(ubd_gendisk[n]);
- put_disk(ubd_gendisk[n]);
ubd_gendisk[n] = NULL;
+ if(disk != NULL){
+ del_gendisk(disk);
+ put_disk(disk);
+ }
if(fake_gendisk[n] != NULL){
del_gendisk(fake_gendisk[n]);
fake_gendisk[n] = NULL;
}
- platform_device_unregister(&dev->pdev);
- *dev = ((struct ubd) DEFAULT_UBD);
err = 0;
- out:
- spin_unlock(&ubd_lock);
- return(err);
+ platform_device_unregister(&ubd_dev->pdev);
+out:
+ mutex_unlock(&ubd_lock);
+ return err;
}
+/* All these are called by mconsole in process context and without
+ * ubd-specific locks. The structure itself is const except for .list.
+ */
static struct mc_device ubd_mc = {
+ .list = LIST_HEAD_INIT(ubd_mc.list),
.name = "ubd",
.config = ubd_config,
- .get_config = ubd_get_config,
+ .get_config = ubd_get_config,
+ .id = ubd_id,
.remove = ubd_remove,
};
-static int ubd_mc_init(void)
+static int __init ubd_mc_init(void)
{
mconsole_register_dev(&ubd_mc);
return 0;
__initcall(ubd_mc_init);
-static struct device_driver ubd_driver = {
- .name = DRIVER_NAME,
- .bus = &platform_bus_type,
+static int __init ubd0_init(void)
+{
+ struct ubd *ubd_dev = &ubd_devs[0];
+
+ mutex_lock(&ubd_lock);
+ if(ubd_dev->file == NULL)
+ ubd_dev->file = "root_fs";
+ mutex_unlock(&ubd_lock);
+
+ return 0;
+}
+
+__initcall(ubd0_init);
+
+/* Used in ubd_init, which is an initcall */
+static struct platform_driver ubd_driver = {
+ .driver = {
+ .name = DRIVER_NAME,
+ },
};
-int ubd_init(void)
+static int __init ubd_init(void)
{
- int i;
+ char *error;
+ int i, err;
- devfs_mk_dir("ubd");
- if (register_blkdev(MAJOR_NR, "ubd"))
+ if (register_blkdev(UBD_MAJOR, "ubd"))
return -1;
- ubd_queue = blk_init_queue(do_ubd_request, &ubd_io_lock);
- if (!ubd_queue) {
- unregister_blkdev(MAJOR_NR, "ubd");
- return -1;
- }
-
- if (fake_major != MAJOR_NR) {
+ if (fake_major != UBD_MAJOR) {
char name[sizeof("ubd_nnn\0")];
snprintf(name, sizeof(name), "ubd_%d", fake_major);
- devfs_mk_dir(name);
if (register_blkdev(fake_major, "ubd"))
return -1;
}
- driver_register(&ubd_driver);
- for (i = 0; i < MAX_DEV; i++)
- ubd_add(i);
+ platform_driver_register(&ubd_driver);
+ mutex_lock(&ubd_lock);
+ for (i = 0; i < MAX_DEV; i++){
+ err = ubd_add(i, &error);
+ if(err)
+ printk(KERN_ERR "Failed to initialize ubd device %d :"
+ "%s\n", i, error);
+ }
+ mutex_unlock(&ubd_lock);
return 0;
}
late_initcall(ubd_init);
-int ubd_driver_init(void){
+static int __init ubd_driver_init(void){
unsigned long stack;
int err;
* enough. So use anyway the io thread. */
}
stack = alloc_stack(0, 0);
- io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *),
+ io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *),
&thread_fd);
if(io_pid < 0){
- printk(KERN_ERR
+ printk(KERN_ERR
"ubd : Failed to start I/O thread (errno = %d) - "
"falling back to synchronous I/O\n", -io_pid);
io_pid = -1;
- return(0);
+ return 0;
}
- err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr,
- SA_INTERRUPT, "ubd", ubd_dev);
+ err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr,
+ IRQF_DISABLED, "ubd", ubd_devs);
if(err != 0)
printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err);
- return(err);
+ return 0;
}
device_initcall(ubd_driver_init);
-static int ubd_open(struct inode *inode, struct file *filp)
+static int ubd_open(struct block_device *bdev, fmode_t mode)
{
- struct gendisk *disk = inode->i_bdev->bd_disk;
- struct ubd *dev = disk->private_data;
+ struct gendisk *disk = bdev->bd_disk;
+ struct ubd *ubd_dev = disk->private_data;
int err = 0;
- if(dev->count == 0){
- err = ubd_open_dev(dev);
+ if(ubd_dev->count == 0){
+ err = ubd_open_dev(ubd_dev);
if(err){
printk(KERN_ERR "%s: Can't open \"%s\": errno = %d\n",
- disk->disk_name, dev->file, -err);
+ disk->disk_name, ubd_dev->file, -err);
goto out;
}
}
- dev->count++;
- set_disk_ro(disk, !dev->openflags.w);
+ ubd_dev->count++;
+ set_disk_ro(disk, !ubd_dev->openflags.w);
/* This should no more be needed. And it didn't work anyway to exclude
* read-write remounting of filesystems.*/
- /*if((filp->f_mode & FMODE_WRITE) && !dev->openflags.w){
- if(--dev->count == 0) ubd_close(dev);
+ /*if((mode & FMODE_WRITE) && !ubd_dev->openflags.w){
+ if(--ubd_dev->count == 0) ubd_close_dev(ubd_dev);
err = -EROFS;
}*/
out:
- return(err);
+ return err;
}
-static int ubd_release(struct inode * inode, struct file * file)
+static int ubd_release(struct gendisk *disk, fmode_t mode)
{
- struct gendisk *disk = inode->i_bdev->bd_disk;
- struct ubd *dev = disk->private_data;
+ struct ubd *ubd_dev = disk->private_data;
- if(--dev->count == 0)
- ubd_close(dev);
- return(0);
+ if(--ubd_dev->count == 0)
+ ubd_close_dev(ubd_dev);
+ return 0;
}
static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask,
* by one word. Thanks to Lynn Kerby for the fix and James McMechan
* for the original diagnosis.
*/
- if(*cow_offset == ((bitmap_len + sizeof(unsigned long) - 1) /
- sizeof(unsigned long) - 1))
+ if (*cow_offset == (DIV_ROUND_UP(bitmap_len,
+ sizeof(unsigned long)) - 1))
(*cow_offset)--;
bitmap_words[0] = bitmap[*cow_offset];
if(req->op == UBD_READ) {
for(i = 0; i < req->length >> 9; i++){
if(ubd_test_bit(sector + i, (unsigned char *) bitmap))
- ubd_set_bit(i, (unsigned char *)
+ ubd_set_bit(i, (unsigned char *)
&req->sector_mask);
- }
+ }
}
else cowify_bitmap(req->offset, req->length, &req->sector_mask,
&req->cow_offset, bitmap, bitmap_offset,
req->bitmap_words, bitmap_len);
}
-static int mmap_fd(struct request *req, struct ubd *dev, __u64 offset)
-{
- __u64 sector;
- unsigned char *bitmap;
- int bit, i;
-
- /* mmap must have been requested on the command line */
- if(!ubd_do_mmap)
- return(-1);
-
- /* The buffer must be page aligned */
- if(((unsigned long) req->buffer % UBD_MMAP_BLOCK_SIZE) != 0)
- return(-1);
-
- /* The request must be a page long */
- if((req->current_nr_sectors << 9) != PAGE_SIZE)
- return(-1);
-
- if(dev->cow.file == NULL)
- return(dev->fd);
-
- sector = offset >> 9;
- bitmap = (unsigned char *) dev->cow.bitmap;
- bit = ubd_test_bit(sector, bitmap);
-
- for(i = 1; i < req->current_nr_sectors; i++){
- if(ubd_test_bit(sector + i, bitmap) != bit)
- return(-1);
- }
-
- if(bit || (rq_data_dir(req) == WRITE))
- offset += dev->cow.data_offset;
-
- /* The data on disk must be page aligned */
- if((offset % UBD_MMAP_BLOCK_SIZE) != 0)
- return(-1);
-
- return(bit ? dev->fd : dev->cow.fd);
-}
-
-static int prepare_mmap_request(struct ubd *dev, int fd, __u64 offset,
- struct request *req,
- struct io_thread_req *io_req)
-{
- int err;
-
- if(rq_data_dir(req) == WRITE){
- /* Writes are almost no-ops since the new data is already in the
- * host page cache
- */
- dev->map_writes++;
- if(dev->cow.file != NULL)
- cowify_bitmap(io_req->offset, io_req->length,
- &io_req->sector_mask, &io_req->cow_offset,
- dev->cow.bitmap, dev->cow.bitmap_offset,
- io_req->bitmap_words,
- dev->cow.bitmap_len);
- }
- else {
- int w;
-
- if((dev->cow.file != NULL) && (fd == dev->cow.fd))
- w = 0;
- else w = dev->openflags.w;
-
- if((dev->cow.file != NULL) && (fd == dev->fd))
- offset += dev->cow.data_offset;
-
- err = physmem_subst_mapping(req->buffer, fd, offset, w);
- if(err){
- printk("physmem_subst_mapping failed, err = %d\n",
- -err);
- return(1);
- }
- dev->map_reads++;
- }
- io_req->op = UBD_MMAP;
- io_req->buffer = req->buffer;
- return(0);
-}
-
-/* Called with ubd_io_lock held */
-static int prepare_request(struct request *req, struct io_thread_req *io_req)
+/* Called with dev->lock held */
+static void prepare_request(struct request *req, struct io_thread_req *io_req,
+ unsigned long long offset, int page_offset,
+ int len, struct page *page)
{
struct gendisk *disk = req->rq_disk;
- struct ubd *dev = disk->private_data;
- __u64 offset;
- int len, fd;
-
- if(req->rq_status == RQ_INACTIVE) return(1);
-
- /* This should be impossible now */
- if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
- printk("Write attempted on readonly ubd device %s\n",
- disk->disk_name);
- end_request(req, 0);
- return(1);
- }
+ struct ubd *ubd_dev = disk->private_data;
- offset = ((__u64) req->sector) << 9;
- len = req->current_nr_sectors << 9;
-
- io_req->fds[0] = (dev->cow.file != NULL) ? dev->cow.fd : dev->fd;
- io_req->fds[1] = dev->fd;
- io_req->map_fd = -1;
+ io_req->req = req;
+ io_req->fds[0] = (ubd_dev->cow.file != NULL) ? ubd_dev->cow.fd :
+ ubd_dev->fd;
+ io_req->fds[1] = ubd_dev->fd;
io_req->cow_offset = -1;
io_req->offset = offset;
io_req->length = len;
io_req->error = 0;
io_req->sector_mask = 0;
- fd = mmap_fd(req, dev, io_req->offset);
- if(fd > 0){
- /* If mmapping is otherwise OK, but the first access to the
- * page is a write, then it's not mapped in yet. So we have
- * to write the data to disk first, then we can map the disk
- * page in and continue normally from there.
- */
- if((rq_data_dir(req) == WRITE) && !is_remapped(req->buffer)){
- io_req->map_fd = dev->fd;
- io_req->map_offset = io_req->offset +
- dev->cow.data_offset;
- dev->write_maps++;
- }
- else return(prepare_mmap_request(dev, fd, io_req->offset, req,
- io_req));
- }
-
- if(rq_data_dir(req) == READ)
- dev->nomap_reads++;
- else dev->nomap_writes++;
-
io_req->op = (rq_data_dir(req) == READ) ? UBD_READ : UBD_WRITE;
io_req->offsets[0] = 0;
- io_req->offsets[1] = dev->cow.data_offset;
- io_req->buffer = req->buffer;
+ io_req->offsets[1] = ubd_dev->cow.data_offset;
+ io_req->buffer = page_address(page) + page_offset;
io_req->sectorsize = 1 << 9;
- if(dev->cow.file != NULL)
- cowify_req(io_req, dev->cow.bitmap, dev->cow.bitmap_offset,
- dev->cow.bitmap_len);
+ if(ubd_dev->cow.file != NULL)
+ cowify_req(io_req, ubd_dev->cow.bitmap,
+ ubd_dev->cow.bitmap_offset, ubd_dev->cow.bitmap_len);
- return(0);
}
-/* Called with ubd_io_lock held */
-static void do_ubd_request(request_queue_t *q)
+/* Called with dev->lock held */
+static void do_ubd_request(struct request_queue *q)
{
- struct io_thread_req io_req;
+ struct io_thread_req *io_req;
struct request *req;
- int err, n;
-
- if(thread_fd == -1){
- while((req = elv_next_request(q)) != NULL){
- err = prepare_request(req, &io_req);
- if(!err){
- do_io(&io_req);
- __ubd_finish(req, io_req.error);
- }
+ sector_t sector;
+ int n;
+
+ while(1){
+ struct ubd *dev = q->queuedata;
+ if(dev->end_sg == 0){
+ struct request *req = blk_fetch_request(q);
+ if(req == NULL)
+ return;
+
+ dev->request = req;
+ dev->start_sg = 0;
+ dev->end_sg = blk_rq_map_sg(q, req, dev->sg);
}
- }
- else {
- if(do_ubd || (req = elv_next_request(q)) == NULL)
- return;
- err = prepare_request(req, &io_req);
- if(!err){
- do_ubd = ubd_handler;
- n = os_write_file(thread_fd, (char *) &io_req,
- sizeof(io_req));
- if(n != sizeof(io_req))
- printk("write to io thread failed, "
- "errno = %d\n", -n);
+
+ req = dev->request;
+ sector = blk_rq_pos(req);
+ while(dev->start_sg < dev->end_sg){
+ struct scatterlist *sg = &dev->sg[dev->start_sg];
+
+ io_req = kmalloc(sizeof(struct io_thread_req),
+ GFP_ATOMIC);
+ if(io_req == NULL){
+ if(list_empty(&dev->restart))
+ list_add(&dev->restart, &restart);
+ return;
+ }
+ prepare_request(req, io_req,
+ (unsigned long long)sector << 9,
+ sg->offset, sg->length, sg_page(sg));
+
+ sector += sg->length >> 9;
+ n = os_write_file(thread_fd, &io_req,
+ sizeof(struct io_thread_req *));
+ if(n != sizeof(struct io_thread_req *)){
+ if(n != -EAGAIN)
+ printk("write to io thread failed, "
+ "errno = %d\n", -n);
+ else if(list_empty(&dev->restart))
+ list_add(&dev->restart, &restart);
+ kfree(io_req);
+ return;
+ }
+
+ dev->start_sg++;
}
+ dev->end_sg = 0;
+ dev->request = NULL;
}
}
-static int ubd_ioctl(struct inode * inode, struct file * file,
+static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
+{
+ struct ubd *ubd_dev = bdev->bd_disk->private_data;
+
+ geo->heads = 128;
+ geo->sectors = 32;
+ geo->cylinders = ubd_dev->size / (128 * 32 * 512);
+ return 0;
+}
+
+static int ubd_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg)
{
- struct hd_geometry __user *loc = (struct hd_geometry __user *) arg;
- struct ubd *dev = inode->i_bdev->bd_disk->private_data;
- struct hd_driveid ubd_id = {
- .cyls = 0,
- .heads = 128,
- .sectors = 32,
- };
+ struct ubd *ubd_dev = bdev->bd_disk->private_data;
+ u16 ubd_id[ATA_ID_WORDS];
switch (cmd) {
- struct hd_geometry g;
struct cdrom_volctrl volume;
- case HDIO_GETGEO:
- if(!loc) return(-EINVAL);
- g.heads = 128;
- g.sectors = 32;
- g.cylinders = dev->size / (128 * 32 * 512);
- g.start = get_start_sect(inode->i_bdev);
- return(copy_to_user(loc, &g, sizeof(g)) ? -EFAULT : 0);
-
case HDIO_GET_IDENTITY:
- ubd_id.cyls = dev->size / (128 * 32 * 512);
+ memset(&ubd_id, 0, ATA_ID_WORDS * 2);
+ ubd_id[ATA_ID_CYLS] = ubd_dev->size / (128 * 32 * 512);
+ ubd_id[ATA_ID_HEADS] = 128;
+ ubd_id[ATA_ID_SECTORS] = 32;
if(copy_to_user((char __user *) arg, (char *) &ubd_id,
sizeof(ubd_id)))
- return(-EFAULT);
- return(0);
-
+ return -EFAULT;
+ return 0;
+
case CDROMVOLREAD:
if(copy_from_user(&volume, (char __user *) arg, sizeof(volume)))
- return(-EFAULT);
+ return -EFAULT;
volume.channel0 = 255;
volume.channel1 = 255;
volume.channel2 = 255;
volume.channel3 = 255;
if(copy_to_user((char __user *) arg, &volume, sizeof(volume)))
- return(-EFAULT);
- return(0);
+ return -EFAULT;
+ return 0;
}
- return(-EINVAL);
-}
-
-static int ubd_check_remapped(int fd, unsigned long address, int is_write,
- __u64 offset)
-{
- __u64 bitmap_offset;
- unsigned long new_bitmap[2];
- int i, err, n;
-
- /* If it's not a write access, we can't do anything about it */
- if(!is_write)
- return(0);
-
- /* We have a write */
- for(i = 0; i < sizeof(ubd_dev) / sizeof(ubd_dev[0]); i++){
- struct ubd *dev = &ubd_dev[i];
-
- if((dev->fd != fd) && (dev->cow.fd != fd))
- continue;
-
- /* It's a write to a ubd device */
-
- /* This should be impossible now */
- if(!dev->openflags.w){
- /* It's a write access on a read-only device - probably
- * shouldn't happen. If the kernel is trying to change
- * something with no intention of writing it back out,
- * then this message will clue us in that this needs
- * fixing
- */
- printk("Write access to mapped page from readonly ubd "
- "device %d\n", i);
- return(0);
- }
-
- /* It's a write to a writeable ubd device - it must be COWed
- * because, otherwise, the page would have been mapped in
- * writeable
- */
-
- if(!dev->cow.file)
- panic("Write fault on writeable non-COW ubd device %d",
- i);
-
- /* It should also be an access to the backing file since the
- * COW pages should be mapped in read-write
- */
-
- if(fd == dev->fd)
- panic("Write fault on a backing page of ubd "
- "device %d\n", i);
-
- /* So, we do the write, copying the backing data to the COW
- * file...
- */
-
- err = os_seek_file(dev->fd, offset + dev->cow.data_offset);
- if(err < 0)
- panic("Couldn't seek to %lld in COW file of ubd "
- "device %d, err = %d",
- offset + dev->cow.data_offset, i, -err);
-
- n = os_write_file(dev->fd, (void *) address, PAGE_SIZE);
- if(n != PAGE_SIZE)
- panic("Couldn't copy data to COW file of ubd "
- "device %d, err = %d", i, -n);
-
- /* ... updating the COW bitmap... */
-
- cowify_bitmap(offset, PAGE_SIZE, NULL, &bitmap_offset,
- dev->cow.bitmap, dev->cow.bitmap_offset,
- new_bitmap, dev->cow.bitmap_len);
-
- err = os_seek_file(dev->fd, bitmap_offset);
- if(err < 0)
- panic("Couldn't seek to %lld in COW file of ubd "
- "device %d, err = %d", bitmap_offset, i, -err);
-
- n = os_write_file(dev->fd, new_bitmap, sizeof(new_bitmap));
- if(n != sizeof(new_bitmap))
- panic("Couldn't update bitmap of ubd device %d, "
- "err = %d", i, -n);
-
- /* Maybe we can map the COW page in, and maybe we can't. If
- * it is a pre-V3 COW file, we can't, since the alignment will
- * be wrong. If it is a V3 or later COW file which has been
- * moved to a system with a larger page size, then maybe we
- * can't, depending on the exact location of the page.
- */
-
- offset += dev->cow.data_offset;
-
- /* Remove the remapping, putting the original anonymous page
- * back. If the COW file can be mapped in, that is done.
- * Otherwise, the COW page is read in.
- */
-
- if(!physmem_remove_mapping((void *) address))
- panic("Address 0x%lx not remapped by ubd device %d",
- address, i);
- if((offset % UBD_MMAP_BLOCK_SIZE) == 0)
- physmem_subst_mapping((void *) address, dev->fd,
- offset, 1);
- else {
- err = os_seek_file(dev->fd, offset);
- if(err < 0)
- panic("Couldn't seek to %lld in COW file of "
- "ubd device %d, err = %d", offset, i,
- -err);
-
- n = os_read_file(dev->fd, (void *) address, PAGE_SIZE);
- if(n != PAGE_SIZE)
- panic("Failed to read page from offset %llx of "
- "COW file of ubd device %d, err = %d",
- offset, i, -n);
- }
-
- return(1);
- }
-
- /* It's not a write on a ubd device */
- return(0);
-}
-
-static struct remapper ubd_remapper = {
- .list = LIST_HEAD_INIT(ubd_remapper.list),
- .proc = ubd_check_remapped,
-};
-
-static int ubd_remapper_setup(void)
-{
- if(ubd_do_mmap)
- register_remapper(&ubd_remapper);
-
- return(0);
-}
-
-__initcall(ubd_remapper_setup);
-
-static int same_backing_files(char *from_cmdline, char *from_cow, char *cow)
-{
- struct uml_stat buf1, buf2;
- int err;
-
- if(from_cmdline == NULL) return(1);
- if(!strcmp(from_cmdline, from_cow)) return(1);
-
- err = os_stat_file(from_cmdline, &buf1);
- if(err < 0){
- printk("Couldn't stat '%s', err = %d\n", from_cmdline, -err);
- return(1);
- }
- err = os_stat_file(from_cow, &buf2);
- if(err < 0){
- printk("Couldn't stat '%s', err = %d\n", from_cow, -err);
- return(1);
- }
- if((buf1.ust_dev == buf2.ust_dev) && (buf1.ust_ino == buf2.ust_ino))
- return(1);
-
- printk("Backing file mismatch - \"%s\" requested,\n"
- "\"%s\" specified in COW header of \"%s\"\n",
- from_cmdline, from_cow, cow);
- return(0);
-}
-
-static int backing_file_mismatch(char *file, __u64 size, time_t mtime)
-{
- unsigned long modtime;
- long long actual;
- int err;
-
- err = os_file_modtime(file, &modtime);
- if(err < 0){
- printk("Failed to get modification time of backing file "
- "\"%s\", err = %d\n", file, -err);
- return(err);
- }
-
- err = os_file_size(file, &actual);
- if(err < 0){
- printk("Failed to get size of backing file \"%s\", "
- "err = %d\n", file, -err);
- return(err);
- }
-
- if(actual != size){
- /*__u64 can be a long on AMD64 and with %lu GCC complains; so
- * the typecast.*/
- printk("Size mismatch (%llu vs %llu) of COW header vs backing "
- "file\n", (unsigned long long) size, actual);
- return(-EINVAL);
- }
- if(modtime != mtime){
- printk("mtime mismatch (%ld vs %ld) of COW header vs backing "
- "file\n", mtime, modtime);
- return(-EINVAL);
- }
- return(0);
-}
-
-int read_cow_bitmap(int fd, void *buf, int offset, int len)
-{
- int err;
-
- err = os_seek_file(fd, offset);
- if(err < 0)
- return(err);
-
- err = os_read_file(fd, buf, len);
- if(err < 0)
- return(err);
-
- return(0);
-}
-
-int open_ubd_file(char *file, struct openflags *openflags,
- char **backing_file_out, int *bitmap_offset_out,
- unsigned long *bitmap_len_out, int *data_offset_out,
- int *create_cow_out)
-{
- time_t mtime;
- unsigned long long size;
- __u32 version, align;
- char *backing_file;
- int fd, err, sectorsize, same, mode = 0644;
-
- fd = os_open_file(file, *openflags, mode);
- if(fd < 0){
- if((fd == -ENOENT) && (create_cow_out != NULL))
- *create_cow_out = 1;
- if(!openflags->w ||
- ((fd != -EROFS) && (fd != -EACCES))) return(fd);
- openflags->w = 0;
- fd = os_open_file(file, *openflags, mode);
- if(fd < 0)
- return(fd);
- }
-
- err = os_lock_file(fd, openflags->w);
- if(err < 0){
- printk("Failed to lock '%s', err = %d\n", file, -err);
- goto out_close;
- }
-
- if(backing_file_out == NULL) return(fd);
-
- err = read_cow_header(file_reader, &fd, &version, &backing_file, &mtime,
- &size, §orsize, &align, bitmap_offset_out);
- if(err && (*backing_file_out != NULL)){
- printk("Failed to read COW header from COW file \"%s\", "
- "errno = %d\n", file, -err);
- goto out_close;
- }
- if(err) return(fd);
-
- if(backing_file_out == NULL) return(fd);
-
- same = same_backing_files(*backing_file_out, backing_file, file);
-
- if(!same && !backing_file_mismatch(*backing_file_out, size, mtime)){
- printk("Switching backing file to '%s'\n", *backing_file_out);
- err = write_cow_header(file, fd, *backing_file_out,
- sectorsize, align, &size);
- if(err){
- printk("Switch failed, errno = %d\n", -err);
- return(err);
- }
- }
- else {
- *backing_file_out = backing_file;
- err = backing_file_mismatch(*backing_file_out, size, mtime);
- if(err) goto out_close;
- }
-
- cow_sizes(version, size, sectorsize, align, *bitmap_offset_out,
- bitmap_len_out, data_offset_out);
-
- return(fd);
- out_close:
- os_close_file(fd);
- return(err);
-}
-
-int create_cow_file(char *cow_file, char *backing_file, struct openflags flags,
- int sectorsize, int alignment, int *bitmap_offset_out,
- unsigned long *bitmap_len_out, int *data_offset_out)
-{
- int err, fd;
-
- flags.c = 1;
- fd = open_ubd_file(cow_file, &flags, NULL, NULL, NULL, NULL, NULL);
- if(fd < 0){
- err = fd;
- printk("Open of COW file '%s' failed, errno = %d\n", cow_file,
- -err);
- goto out;
- }
-
- err = init_cow_file(fd, cow_file, backing_file, sectorsize, alignment,
- bitmap_offset_out, bitmap_len_out,
- data_offset_out);
- if(!err)
- return(fd);
- os_close_file(fd);
- out:
- return(err);
+ return -EINVAL;
}
static int update_bitmap(struct io_thread_req *req)
int n;
if(req->cow_offset == -1)
- return(0);
+ return 0;
n = os_seek_file(req->fds[1], req->cow_offset);
if(n < 0){
printk("do_io - bitmap lseek failed : err = %d\n", -n);
- return(1);
+ return 1;
}
n = os_write_file(req->fds[1], &req->bitmap_words,
- sizeof(req->bitmap_words));
+ sizeof(req->bitmap_words));
if(n != sizeof(req->bitmap_words)){
printk("do_io - bitmap update failed, err = %d fd = %d\n", -n,
req->fds[1]);
- return(1);
+ return 1;
}
- return(0);
+ return 0;
}
-void do_io(struct io_thread_req *req)
+static void do_io(struct io_thread_req *req)
{
char *buf;
unsigned long len;
int err;
__u64 off;
- if(req->op == UBD_MMAP){
- /* Touch the page to force the host to do any necessary IO to
- * get it into memory
- */
- n = *((volatile int *) req->buffer);
- req->error = update_bitmap(req);
- return;
- }
-
nsectors = req->length / req->sectorsize;
start = 0;
do {
*/
int kernel_fd = -1;
-/* Only changed by the io thread */
-int io_count = 0;
+/* Only changed by the io thread. XXX: currently unused. */
+static int io_count = 0;
int io_thread(void *arg)
{
- struct io_thread_req req;
+ struct io_thread_req *req;
int n;
ignore_sigwinch_sig();
while(1){
- n = os_read_file(kernel_fd, &req, sizeof(req));
- if(n != sizeof(req)){
+ n = os_read_file(kernel_fd, &req,
+ sizeof(struct io_thread_req *));
+ if(n != sizeof(struct io_thread_req *)){
if(n < 0)
printk("io_thread - read failed, fd = %d, "
"err = %d\n", kernel_fd, -n);
continue;
}
io_count++;
- do_io(&req);
- n = os_write_file(kernel_fd, &req, sizeof(req));
- if(n != sizeof(req))
+ do_io(req);
+ n = os_write_file(kernel_fd, &req,
+ sizeof(struct io_thread_req *));
+ if(n != sizeof(struct io_thread_req *))
printk("io_thread - write failed, fd = %d, err = %d\n",
kernel_fd, -n);
}
-}
-/*
- * Overrides for Emacs so that we follow Linus's tabbing style.
- * Emacs will notice this stuff at the end of the file and automatically
- * adjust the settings for this buffer only. This must remain at the end
- * of the file.
- * ---------------------------------------------------------------------------
- * Local variables:
- * c-file-style: "linux"
- * End:
- */
+ return 0;
+}