-/*
+/*
* Copyright (C) 2000 Jeff Dike (jdike@karaya.com)
* Licensed under the GPL
*/
#define MAJOR_NR UBD_MAJOR
#define UBD_SHIFT 4
-#include "linux/config.h"
#include "linux/module.h"
#include "linux/blkdev.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/ctype.h"
#include "linux/blkpg.h"
#include "linux/genhd.h"
#include "linux/spinlock.h"
+#include "linux/platform_device.h"
#include "asm/segment.h"
#include "asm/uaccess.h"
#include "asm/irq.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 };
+
+struct io_thread_req {
+ enum ubd_req op;
+ int fds[2];
+ unsigned long offsets[2];
+ unsigned long long offset;
+ unsigned long length;
+ char *buffer;
+ int sectorsize;
+ unsigned long sector_mask;
+ unsigned long long cow_offset;
+ unsigned long bitmap_words[2];
+ int error;
+};
+
+extern 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);
+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;
+ int bits, off;
+
+ bits = sizeof(data[0]) * 8;
+ n = bit / bits;
+ off = bit % bits;
+ return((data[n] & (1 << off)) != 0);
+}
+
+static inline void ubd_set_bit(__u64 bit, unsigned char *data)
+{
+ __u64 n;
+ int bits, off;
-static spinlock_t ubd_io_lock = SPIN_LOCK_UNLOCKED;
-static spinlock_t ubd_lock = SPIN_LOCK_UNLOCKED;
+ bits = sizeof(data[0]) * 8;
+ n = bit / bits;
+ off = bit % bits;
+ data[n] |= (1 << off);
+}
+/*End stuff from ubd_user.h*/
+
+#define DRIVER_NAME "uml-blkdev"
+
+/* Can be taken in interrupt context, and is passed to the block layer to lock
+ * the request queue. Kernel side code knows that. */
+static DEFINE_SPINLOCK(ubd_io_lock);
+
+static DEFINE_MUTEX(ubd_lock);
-static void (*do_ubd)(void);
+/* XXX - this made sense in 2.4 days, now it's only used as a boolean, and
+ * probably it doesn't make sense even for that. */
+static int do_ubd;
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,
unsigned int cmd, unsigned long arg);
+static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo);
-#define MAX_DEV (8)
-
-/* Changed in early boot */
-static int ubd_do_mmap = 0;
-#define UBD_MMAP_BLOCK_SIZE PAGE_SIZE
+#define MAX_DEV (16)
static 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 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 })
static struct openflags global_openflags = OPEN_FLAGS;
struct cow {
+ /* backing file name */
char *file;
+ /* backing file fd */
int fd;
unsigned long *bitmap;
unsigned long bitmap_len;
};
struct ubd {
+ /* 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;
-
- int map_writes;
- int map_reads;
- int nomap_writes;
- int nomap_reads;
- int write_maps;
+ struct platform_device pdev;
};
#define DEFAULT_COW { \
.boot_openflags = OPEN_FLAGS, \
.openflags = OPEN_FLAGS, \
.no_cow = 0, \
+ .shared = 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);
-}
-
-__initcall(ubd0_init);
+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;
static void make_proc_ide(void)
{
- proc_ide_root = proc_mkdir("ide", 0);
+ proc_ide_root = proc_mkdir("ide", NULL);
proc_ide = proc_mkdir("ide0", proc_ide_root);
}
return(-1);
*ptr = end;
}
- else if (('a' <= *str) && (*str <= 'h')) {
+ else if (('a' <= *str) && (*str <= 'z')) {
n = *str - 'a';
str++;
*ptr = str;
return(n);
}
+/* 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)
{
- struct ubd *dev;
+ struct ubd *ubd_dev;
struct openflags flags = global_openflags;
char *backing_file;
- int n, err;
+ int n, err, 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.s = 1;
+ global_openflags = of_sync(global_openflags);
return(0);
}
major = simple_strtoul(str, &end, 0);
if((*end != '\0') || (end == str)){
- printk(KERN_ERR
+ printk(KERN_ERR
"ubd_setup : didn't parse major number\n");
return(1);
}
err = 1;
- spin_lock(&ubd_lock);
+ mutex_lock(&ubd_lock);
if(fake_major != MAJOR_NR){
printk(KERN_ERR "Can't assign a fake major twice\n");
goto out1;
}
-
+
fake_major = major;
printk(KERN_INFO "Setting extra ubd major number to %d\n",
major);
err = 0;
out1:
- spin_unlock(&ubd_lock);
+ mutex_unlock(&ubd_lock);
return(err);
}
}
if(n >= MAX_DEV){
printk(KERN_ERR "ubd_setup : index %d out of range "
- "(%d devices)\n", n, MAX_DEV);
+ "(%d devices, from 0 to %d)\n", n, MAX_DEV, MAX_DEV - 1);
return(1);
}
err = 1;
- spin_lock(&ubd_lock);
+ mutex_lock(&ubd_lock);
- dev = &ubd_dev[n];
- if(dev->file != NULL){
+ ubd_dev = &ubd_devs[n];
+ if(ubd_dev->file != NULL){
printk(KERN_ERR "ubd_setup : device already configured\n");
- goto out2;
+ goto out;
}
- if(index_out) *index_out = n;
-
- if (*str == 'r'){
- flags.w = 0;
- str++;
- }
- if (*str == 's'){
- flags.s = 1;
- str++;
- }
- if (*str == 'd'){
- dev->no_cow = 1;
+ if (index_out)
+ *index_out = n;
+
+ for (i = 0; i < sizeof("rscd="); i++) {
+ switch (*str) {
+ case 'r':
+ flags.w = 0;
+ break;
+ case 's':
+ flags.s = 1;
+ break;
+ case 'd':
+ 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, c, or d)\n");
+ goto out;
+ }
str++;
}
- if(*str++ != '='){
+ if (*str == '=')
+ printk(KERN_ERR "ubd_setup : Too many flags specified\n");
+ else
printk(KERN_ERR "ubd_setup : Expected '='\n");
- goto out2;
- }
+ goto out;
+break_loop:
err = 0;
backing_file = strchr(str, ',');
+
+ if (!backing_file) {
+ backing_file = strchr(str, ':');
+ }
+
if(backing_file){
- if(dev->no_cow)
+ if(ubd_dev->no_cow)
printk(KERN_ERR "Can't specify both 'd' and a "
"cow file\n");
else {
backing_file++;
}
}
- dev->file = str;
- dev->cow.file = backing_file;
- dev->boot_openflags = flags;
- out2:
- spin_unlock(&ubd_lock);
+ ubd_dev->file = str;
+ ubd_dev->cow.file = backing_file;
+ ubd_dev->boot_openflags = flags;
+out:
+ mutex_unlock(&ubd_lock);
return(err);
}
__setup("ubd", ubd_setup);
__uml_help(ubd_setup,
-"ubd<n>=<filename>\n"
+"ubd<n><flags>=<filename>[(:|,)<filename2>]\n"
" This is used to associate a device with a file in the underlying\n"
-" filesystem. Usually, there is a filesystem in the file, but \n"
+" filesystem. When specifying two filenames, the first one is the\n"
+" COW name and the second is the backing file name. As separator you can\n"
+" 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"
+" a COW file or a backing file. To override this detection, add the 'd'\n"
+" flag:\n"
+" ubd0d=BackingFile\n"
+" Usually, there is a filesystem in the file, but \n"
" that's not required. Swap devices containing swap files can be\n"
" specified like this. Also, a file which doesn't contain a\n"
" filesystem can have its contents read in the virtual \n"
-" machine by running dd on the device. n must be in the range\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' (has to be _after_ 'r', if there is one) will cause data\n"
-" 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\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);
+}
+
+__setup("udb", udb_setup);
+__uml_help(udb_setup,
+"udb\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;
/* Changed by ubd_handler, which is serialized because interrupts only
* happen on CPU 0.
+ * XXX: currently unused.
*/
-int intr_count = 0;
+static int intr_count = 0;
-static void ubd_finish(struct request *req, int error)
+/* call ubd_finish if you need to serialize */
+static void __ubd_finish(struct request *req, int error)
{
int nsect;
if(error){
- spin_lock(&ubd_io_lock);
end_request(req, 0);
- spin_unlock(&ubd_io_lock);
return;
}
nsect = req->current_nr_sectors;
req->errors = 0;
req->nr_sectors -= nsect;
req->current_nr_sectors = 0;
- spin_lock(&ubd_io_lock);
end_request(req, 1);
+}
+
+/* Callable only from interrupt context - otherwise you need to do
+ * spin_lock_irq()/spin_lock_irqsave() */
+static inline void ubd_finish(struct request *req, int error)
+{
+ spin_lock(&ubd_io_lock);
+ __ubd_finish(req, error);
spin_unlock(&ubd_io_lock);
}
+/* XXX - move this inside ubd_intr. */
+/* Called without ubd_io_lock held, and only in interrupt context. */
static void ubd_handler(void)
{
struct io_thread_req req;
struct request *rq = elv_next_request(ubd_queue);
- int n, err;
+ int n;
- do_ubd = NULL;
+ do_ubd = 0;
intr_count++;
- n = read_ubd_fs(thread_fd, &req, sizeof(req));
+ 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_unlock(&ubd_io_lock);
return;
}
-
- 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);
- }
ubd_finish(rq, req.error);
reactivate_fd(thread_fd, UBD_IRQ);
+ spin_lock(&ubd_io_lock);
do_ubd_request(ubd_queue);
+ spin_unlock(&ubd_io_lock);
}
-static irqreturn_t ubd_intr(int irq, void *dev, struct pt_regs *unused)
+static irqreturn_t ubd_intr(int irq, void *dev)
{
ubd_handler();
return(IRQ_HANDLED);
void kill_io_thread(void)
{
- if(io_pid != -1)
+ if(io_pid != -1)
os_kill_process(io_pid, 1);
}
__uml_exitcall(kill_io_thread);
-static int ubd_file_size(struct ubd *dev, __u64 *size_out)
+static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out)
{
char *file;
- file = dev->cow.file ? dev->cow.file : dev->file;
+ file = ubd_dev->cow.file ? ubd_dev->cow.file : ubd_dev->file;
return(os_file_size(file, size_out));
}
-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) 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(dev->cow.file != NULL){
+ if(ubd_dev->cow.file != NULL){
err = -ENOMEM;
- dev->cow.bitmap = (void *) vmalloc(dev->cow.bitmap_len);
- if(dev->cow.bitmap == NULL){
+ ubd_dev->cow.bitmap = (void *) 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);
error:
- os_close_file(dev->fd);
+ os_close_file(ubd_dev->fd);
return(err);
}
-static int ubd_new_disk(int major, u64 size, int unit,
+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)
disk->first_minor = unit << UBD_SHIFT;
disk->fops = &ubd_blops;
set_capacity(disk, size / 512);
- if(major == MAJOR_NR){
+ if(major == MAJOR_NR)
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_devs[unit].pdev.id = unit;
+ ubd_devs[unit].pdev.name = DRIVER_NAME;
+ platform_device_register(&ubd_devs[unit].pdev);
+ disk->driverfs_dev = &ubd_devs[unit].pdev.dev;
}
- disk->private_data = &ubd_dev[unit];
+ disk->private_data = &ubd_devs[unit];
disk->queue = ubd_queue;
add_disk(disk);
return 0;
}
+#define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9))
+
static int ubd_add(int n)
{
- struct ubd *dev = &ubd_dev[n];
+ struct ubd *ubd_dev = &ubd_devs[n];
int err;
- if(dev->file == NULL)
- return(-ENODEV);
-
- if (ubd_open_dev(dev))
- return(-ENODEV);
+ err = -ENODEV;
+ if(ubd_dev->file == NULL)
+ goto out;
- err = ubd_file_size(dev, &dev->size);
+ err = ubd_file_size(ubd_dev, &ubd_dev->size);
if(err < 0)
- return(err);
+ goto out;
+
+ ubd_dev->size = ROUND_BLOCK(ubd_dev->size);
+
+ err = ubd_disk_register(MAJOR_NR, ubd_dev->size, n, &ubd_gendisk[n]);
+ if(err)
+ goto out;
- 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,
+ ubd_disk_register(fake_major, ubd_dev->size, n,
&fake_gendisk[n]);
/* perhaps this should also be under the "if (fake_major)" above */
if (fake_ide)
make_ide_entries(ubd_gendisk[n]->disk_name);
- ubd_close(dev);
- return 0;
+ err = 0;
+out:
+ return err;
}
static int ubd_config(char *str)
{
- int n, err;
+ int n, ret;
- str = uml_strdup(str);
- if(str == NULL){
+ str = kstrdup(str, GFP_KERNEL);
+ if (str == NULL) {
printk(KERN_ERR "ubd_config failed to strdup string\n");
- return(1);
+ ret = 1;
+ goto out;
}
- err = ubd_setup_common(str, &n);
- if(err){
- kfree(str);
- return(-1);
+ ret = ubd_setup_common(str, &n);
+ if (ret) {
+ ret = -1;
+ 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);
+ 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;
- char *end;
+ struct ubd *ubd_dev;
int n, len = 0;
- n = simple_strtoul(name, &end, 0);
- if((*end != '\0') || (end == name)){
- *error_out = "ubd_get_config : didn't parse device number";
- return(-1);
- }
-
+ n = parse_unit(&name);
if((n >= MAX_DEV) || (n < 0)){
*error_out = "ubd_get_config : device number out of range";
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);
+ 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)
+{
+ struct ubd *ubd_dev;
+ int err = -ENODEV;
- dev = &ubd_dev[n];
- if(dev->count > 0)
- return(-EBUSY); /* you cannot remove a open disk */
-
- err = 0;
- spin_lock(&ubd_lock);
+ mutex_lock(&ubd_lock);
if(ubd_gendisk[n] == NULL)
goto out;
+ ubd_dev = &ubd_devs[n];
+
+ 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;
fake_gendisk[n] = NULL;
}
- *dev = ((struct ubd) DEFAULT_UBD);
+ platform_device_unregister(&ubd_dev->pdev);
+ *ubd_dev = ((struct ubd) DEFAULT_UBD);
err = 0;
- out:
- spin_unlock(&ubd_lock);
- return(err);
+out:
+ mutex_unlock(&ubd_lock);
+ return err;
}
+/* All these are called by mconsole in process context and without ubd-specific locks. */
static struct mc_device ubd_mc = {
.name = "ubd",
.config = ubd_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);
-int ubd_init(void)
+static int __init ubd0_init(void)
+{
+ struct ubd *ubd_dev = &ubd_devs[0];
+
+ if(ubd_dev->file == NULL)
+ ubd_dev->file = "root_fs";
+ return(0);
+}
+
+__initcall(ubd0_init);
+
+static struct platform_driver ubd_driver = {
+ .driver = {
+ .name = DRIVER_NAME,
+ },
+};
+
+static int __init ubd_init(void)
{
int i;
- devfs_mk_dir("ubd");
if (register_blkdev(MAJOR_NR, "ubd"))
return -1;
return -1;
}
- elevator_init(ubd_queue, &elevator_noop);
-
if (fake_major != MAJOR_NR) {
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;
}
- for (i = 0; i < MAX_DEV; i++)
+ platform_driver_register(&ubd_driver);
+ for (i = 0; i < MAX_DEV; i++)
ubd_add(i);
return 0;
}
late_initcall(ubd_init);
-int ubd_driver_init(void){
+static int __init ubd_driver_init(void){
unsigned long stack;
int err;
+ /* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/
if(global_openflags.s){
- printk(KERN_INFO "ubd : Synchronous mode\n");
- return(0);
+ printk(KERN_INFO "ubd: Synchronous mode\n");
+ /* Letting ubd=sync be like using ubd#s= instead of ubd#= is
+ * 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){
- io_pid = -1;
- 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);
}
- err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr,
- SA_INTERRUPT, "ubd", ubd_dev);
- if(err != 0)
+ 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)
{
struct gendisk *disk = inode->i_bdev->bd_disk;
- struct ubd *dev = disk->private_data;
+ 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++;
- if((filp->f_mode & FMODE_WRITE) && !dev->openflags.w){
- if(--dev->count == 0) ubd_close(dev);
+ 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) && !ubd_dev->openflags.w){
+ if(--ubd_dev->count == 0) ubd_close_dev(ubd_dev);
err = -EROFS;
- }
+ }*/
out:
return(err);
}
static int ubd_release(struct inode * inode, struct file * file)
{
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);
+ 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,
- __u64 *cow_offset, unsigned long *bitmap,
+static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask,
+ __u64 *cow_offset, unsigned long *bitmap,
__u64 bitmap_offset, unsigned long *bitmap_words,
__u64 bitmap_len)
{
* 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) /
+ if(*cow_offset == ((bitmap_len + sizeof(unsigned long) - 1) /
sizeof(unsigned long) - 1))
(*cow_offset)--;
*cow_offset += bitmap_offset;
}
-static void cowify_req(struct io_thread_req *req, unsigned long *bitmap,
+static void cowify_req(struct io_thread_req *req, unsigned long *bitmap,
__u64 bitmap_offset, __u64 bitmap_len)
{
__u64 sector = req->offset >> 9;
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->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)
{
struct gendisk *disk = req->rq_disk;
- struct ubd *dev = disk->private_data;
+ struct ubd *ubd_dev = disk->private_data;
__u64 offset;
- int len, fd;
-
- if(req->rq_status == RQ_INACTIVE) return(1);
+ int len;
- if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
- printk("Write attempted on readonly ubd device %s\n",
+ /* This should be impossible now */
+ if((rq_data_dir(req) == WRITE) && !ubd_dev->openflags.w){
+ printk("Write attempted on readonly ubd device %s\n",
disk->disk_name);
- spin_lock(&ubd_io_lock);
end_request(req, 0);
- spin_unlock(&ubd_io_lock);
return(1);
}
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->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, dev->fd, io_req->offset + dev->cow.data_offset)){
- 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->offsets[1] = ubd_dev->cow.data_offset;
io_req->buffer = req->buffer;
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)
{
struct io_thread_req io_req;
int err, n;
if(thread_fd == -1){
- while(!elv_queue_empty(q)){
- req = elv_next_request(q);
+ 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);
+ __ubd_finish(req, io_req.error);
}
}
}
else {
- if(do_ubd || elv_queue_empty(q))
+ if(do_ubd || (req = elv_next_request(q)) == NULL)
return;
- req = elv_next_request(q);
err = prepare_request(req, &io_req);
if(!err){
- do_ubd = ubd_handler;
- n = write_ubd_fs(thread_fd, (char *) &io_req,
+ do_ubd = 1;
+ n = os_write_file(thread_fd, (char *) &io_req,
sizeof(io_req));
if(n != sizeof(io_req))
printk("write to io thread failed, "
}
}
+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 inode * inode, struct file * file,
unsigned int cmd, unsigned long arg)
{
- struct hd_geometry *loc = (struct hd_geometry *) arg;
- struct ubd *dev = inode->i_bdev->bd_disk->private_data;
- int err;
+ struct ubd *ubd_dev = inode->i_bdev->bd_disk->private_data;
struct hd_driveid ubd_id = {
.cyls = 0,
.heads = 128,
};
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_SET_UNMASKINTR:
- if(!capable(CAP_SYS_ADMIN)) return(-EACCES);
- if((arg > 1) || (inode->i_bdev->bd_contains != inode->i_bdev))
- return(-EINVAL);
- return(0);
-
- case HDIO_GET_UNMASKINTR:
- if(!arg) return(-EINVAL);
- err = verify_area(VERIFY_WRITE, (long *) arg, sizeof(long));
- if(err)
- return(err);
- return(0);
-
- case HDIO_GET_MULTCOUNT:
- if(!arg) return(-EINVAL);
- err = verify_area(VERIFY_WRITE, (long *) arg, sizeof(long));
- if(err)
- return(err);
- return(0);
-
- case HDIO_SET_MULTCOUNT:
- if(!capable(CAP_SYS_ADMIN)) return(-EACCES);
- if(inode->i_bdev->bd_contains != inode->i_bdev)
- return(-EINVAL);
- return(0);
-
case HDIO_GET_IDENTITY:
- ubd_id.cyls = dev->size / (128 * 32 * 512);
- if(copy_to_user((char *) arg, (char *) &ubd_id,
+ ubd_id.cyls = ubd_dev->size / (128 * 32 * 512);
+ if(copy_to_user((char __user *) arg, (char *) &ubd_id,
sizeof(ubd_id)))
return(-EFAULT);
return(0);
case CDROMVOLREAD:
- if(copy_from_user(&volume, (char *) arg, sizeof(volume)))
+ if(copy_from_user(&volume, (char __user *) arg, sizeof(volume)))
return(-EFAULT);
volume.channel0 = 255;
volume.channel1 = 255;
volume.channel2 = 255;
volume.channel3 = 255;
- if(copy_to_user((char *) arg, &volume, sizeof(volume)))
+ if(copy_to_user((char __user *) arg, &volume, sizeof(volume)))
return(-EFAULT);
return(0);
}
return(-EINVAL);
}
-static int ubd_check_remapped(int fd, unsigned long address, int is_write,
- __u64 offset, int is_user)
+static int path_requires_switch(char *from_cmdline, char *from_cow, char *cow)
{
- __u64 bitmap_offset;
- unsigned long new_bitmap[2];
- int i, err, n;
+ struct uml_stat buf1, buf2;
+ int err;
- /* This can only fix kernelspace faults */
- if(is_user)
- return(0);
+ if(from_cmdline == NULL)
+ return 0;
+ if(!strcmp(from_cmdline, from_cow))
+ return 0;
- /* ubd-mmap is only enabled in skas mode */
- if(CHOOSE_MODE(1, 0))
- return(0);
+ err = os_stat_file(from_cmdline, &buf1);
+ if(err < 0){
+ printk("Couldn't stat '%s', err = %d\n", from_cmdline, -err);
+ return 0;
+ }
+ 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 0;
- /* If it's not a write access, we can't do anything about it */
- if(!is_write)
- return(0);
+ printk("Backing file mismatch - \"%s\" requested,\n"
+ "\"%s\" specified in COW header of \"%s\"\n",
+ from_cmdline, from_cow, cow);
+ return 1;
+}
- /* We have a write */
- for(i = 0; i < sizeof(ubd_dev) / sizeof(ubd_dev[0]); i++){
- struct ubd *dev = &ubd_dev[i];
+static int backing_file_mismatch(char *file, __u64 size, time_t mtime)
+{
+ unsigned long modtime;
+ unsigned long long actual;
+ int err;
- if((dev->fd != fd) && (dev->cow.fd != fd))
- continue;
+ 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);
+ }
- /* It's a write to a ubd device */
-
- 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);
- }
+ 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);
+ }
- /* It's a write to a writeable ubd device - it must be COWed
- * because, otherwise, the page would have been mapped in
- * writeable
- */
+ 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);
+}
- if(!dev->cow.file)
- panic("Write fault on writeable non-COW ubd device %d",
- i);
+int read_cow_bitmap(int fd, void *buf, int offset, int len)
+{
+ int err;
- /* It should also be an access to the backing file since the
- * COW pages should be mapped in read-write
- */
+ err = os_seek_file(fd, offset);
+ if(err < 0)
+ return(err);
- if(fd == dev->fd)
- panic("Write fault on a backing page of ubd "
- "device %d\n", i);
+ err = os_read_file(fd, buf, len);
+ if(err < 0)
+ return(err);
- /* So, we do the write, copying the backing data to the COW
- * file...
- */
+ return(0);
+}
- 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);
+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("Not locking \"%s\" on the host\n", file);
+ else {
+ err = os_lock_file(fd, openflags->w);
+ if(err < 0){
+ printk("Failed to lock '%s', err = %d\n", file, -err);
+ goto out_close;
+ }
+ }
- 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);
+ /* Successful return case! */
+ if(backing_file_out == NULL)
+ return(fd);
- /* ... updating the COW bitmap... */
+ 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);
+
+ 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("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);
+ goto out_close;
+ }
+ } else {
+ *backing_file_out = backing_file;
+ err = backing_file_mismatch(*backing_file_out, size, mtime);
+ if (err)
+ goto out_close;
+ }
- cowify_bitmap(offset, PAGE_SIZE, NULL, &bitmap_offset,
- dev->cow.bitmap, dev->cow.bitmap_offset,
- new_bitmap, dev->cow.bitmap_len);
+ cow_sizes(version, size, sectorsize, align, *bitmap_offset_out,
+ bitmap_len_out, data_offset_out);
- 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);
+ return fd;
+ out_close:
+ os_close_file(fd);
+ return 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);
- }
+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, 0, 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);
+}
+static int update_bitmap(struct io_thread_req *req)
+{
+ int n;
+
+ if(req->cow_offset == -1)
+ 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);
+ }
+
+ n = os_write_file(req->fds[1], &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);
}
- /* 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)
+void do_io(struct io_thread_req *req)
{
- if(ubd_do_mmap)
- register_remapper(&ubd_remapper);
+ char *buf;
+ unsigned long len;
+ int n, nsectors, start, end, bit;
+ int err;
+ __u64 off;
+
+ nsectors = req->length / req->sectorsize;
+ start = 0;
+ do {
+ bit = ubd_test_bit(start, (unsigned char *) &req->sector_mask);
+ end = start;
+ while((end < nsectors) &&
+ (ubd_test_bit(end, (unsigned char *)
+ &req->sector_mask) == bit))
+ end++;
+
+ off = req->offset + req->offsets[bit] +
+ start * req->sectorsize;
+ len = (end - start) * req->sectorsize;
+ buf = &req->buffer[start * req->sectorsize];
+
+ err = os_seek_file(req->fds[bit], off);
+ if(err < 0){
+ printk("do_io - lseek failed : err = %d\n", -err);
+ req->error = 1;
+ return;
+ }
+ if(req->op == UBD_READ){
+ n = 0;
+ do {
+ buf = &buf[n];
+ len -= n;
+ n = os_read_file(req->fds[bit], buf, len);
+ if (n < 0) {
+ printk("do_io - read failed, err = %d "
+ "fd = %d\n", -n, req->fds[bit]);
+ req->error = 1;
+ return;
+ }
+ } while((n < len) && (n != 0));
+ if (n < len) memset(&buf[n], 0, len - n);
+ } else {
+ n = os_write_file(req->fds[bit], buf, len);
+ if(n != len){
+ printk("do_io - write failed err = %d "
+ "fd = %d\n", -n, req->fds[bit]);
+ req->error = 1;
+ return;
+ }
+ }
- return(0);
-}
+ start = end;
+ } while(start < nsectors);
-__initcall(ubd_remapper_setup);
+ req->error = update_bitmap(req);
+}
-/*
- * 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:
+/* Changed in start_io_thread, which is serialized by being called only
+ * from ubd_init, which is an initcall.
*/
+int kernel_fd = -1;
+
+/* Only changed by the io thread. XXX: currently unused. */
+static int io_count = 0;
+
+int io_thread(void *arg)
+{
+ 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)){
+ if(n < 0)
+ printk("io_thread - read failed, fd = %d, "
+ "err = %d\n", kernel_fd, -n);
+ else {
+ printk("io_thread - short read, fd = %d, "
+ "length = %d\n", kernel_fd, n);
+ }
+ continue;
+ }
+ io_count++;
+ do_io(&req);
+ n = os_write_file(kernel_fd, &req, sizeof(req));
+ if(n != sizeof(req))
+ printk("io_thread - write failed, fd = %d, err = %d\n",
+ kernel_fd, -n);
+ }
+
+ return 0;
+}