module/os/linux/zfs/zpl_file_range.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or https://opensource.org/licenses/CDDL-1.0.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2023, Klara Inc.
  23  */
  24
  25 #ifdef CONFIG_COMPAT
  26 #include <linux/compat.h>
  27 #endif
  28 #include <linux/fs.h>
  29 #include <sys/file.h>
  30 #include <sys/zfs_znode.h>
  31 #include <sys/zfs_vnops.h>
  32 #include <sys/zfeature.h>
  33
  34 /*
  35  * Clone part of a file via block cloning.
  36  *
  37  * Note that we are not required to update file offsets; the kernel will take
  38  * care of that depending on how it was called.
  39  */
  40 static ssize_t
  41 __zpl_clone_file_range(struct file *src_file, loff_t src_off,
  42     struct file *dst_file, loff_t dst_off, size_t len)
  43 {
  44         struct inode *src_i = file_inode(src_file);
  45         struct inode *dst_i = file_inode(dst_file);
  46         uint64_t src_off_o = (uint64_t)src_off;
  47         uint64_t dst_off_o = (uint64_t)dst_off;
  48         uint64_t len_o = (uint64_t)len;
  49         cred_t *cr = CRED();
  50         fstrans_cookie_t cookie;
  51         int err;
  52
  53         if (!spa_feature_is_enabled(
  54             dmu_objset_spa(ITOZSB(dst_i)->z_os), SPA_FEATURE_BLOCK_CLONING))
  55                 return (-EOPNOTSUPP);
  56
  57         if (src_i != dst_i)
  58                 spl_inode_lock_shared(src_i);
  59         spl_inode_lock(dst_i);
  60
  61         crhold(cr);
  62         cookie = spl_fstrans_mark();
  63
  64         err = -zfs_clone_range(ITOZ(src_i), &src_off_o, ITOZ(dst_i),
  65             &dst_off_o, &len_o, cr);
  66
  67         spl_fstrans_unmark(cookie);
  68         crfree(cr);
  69
  70         spl_inode_unlock(dst_i);
  71         if (src_i != dst_i)
  72                 spl_inode_unlock_shared(src_i);
  73
  74         if (err < 0)
  75                 return (err);
  76
  77         return ((ssize_t)len_o);
  78 }
  79
  80 #if defined(HAVE_VFS_COPY_FILE_RANGE) || \
  81     defined(HAVE_VFS_FILE_OPERATIONS_EXTEND)
  82 /*
  83  * Entry point for copy_file_range(). Copy len bytes from src_off in src_file
  84  * to dst_off in dst_file. We are permitted to do this however we like, so we
  85  * try to just clone the blocks, and if we can't support it, fall back to the
  86  * kernel's generic byte copy function.
  87  */
  88 ssize_t
  89 zpl_copy_file_range(struct file *src_file, loff_t src_off,
  90     struct file *dst_file, loff_t dst_off, size_t len, unsigned int flags)
  91 {
  92         ssize_t ret;
  93
  94         if (flags != 0)
  95                 return (-EINVAL);
  96
  97         /* Try to do it via zfs_clone_range() */
  98         ret = __zpl_clone_file_range(src_file, src_off,
  99             dst_file, dst_off, len);
 100
 101 #ifdef HAVE_VFS_GENERIC_COPY_FILE_RANGE
 102         /*
 103          * Since Linux 5.3 the filesystem driver is responsible for executing
 104          * an appropriate fallback, and a generic fallback function is provided.
 105          */
 106         if (ret == -EOPNOTSUPP || ret == -EXDEV)
 107                 ret = generic_copy_file_range(src_file, src_off, dst_file,
 108                     dst_off, len, flags);
 109 #else
 110         /*
 111          * Before Linux 5.3 the filesystem has to return -EOPNOTSUPP to signal
 112          * to the kernel that it should fallback to a content copy.
 113          */
 114         if (ret == -EXDEV)
 115                 ret = -EOPNOTSUPP;
 116 #endif /* HAVE_VFS_GENERIC_COPY_FILE_RANGE */
 117
 118         return (ret);
 119 }
 120 #endif /* HAVE_VFS_COPY_FILE_RANGE || HAVE_VFS_FILE_OPERATIONS_EXTEND */
 121
 122 #ifdef HAVE_VFS_REMAP_FILE_RANGE
 123 /*
 124  * Entry point for FICLONE/FICLONERANGE/FIDEDUPERANGE.
 125  *
 126  * FICLONE and FICLONERANGE are basically the same as copy_file_range(), except
 127  * that they must clone - they cannot fall back to copying. FICLONE is exactly
 128  * FICLONERANGE, for the entire file. We don't need to try to tell them apart;
 129  * the kernel will sort that out for us.
 130  *
 131  * FIDEDUPERANGE is for turning a non-clone into a clone, that is, compare the
 132  * range in both files and if they're the same, arrange for them to be backed
 133  * by the same storage.
 134  */
 135 loff_t
 136 zpl_remap_file_range(struct file *src_file, loff_t src_off,
 137     struct file *dst_file, loff_t dst_off, loff_t len, unsigned int flags)
 138 {
 139         if (flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_CAN_SHORTEN))
 140                 return (-EINVAL);
 141
 142         /*
 143          * REMAP_FILE_CAN_SHORTEN lets us know we can clone less than the given
 144          * range if we want. Its designed for filesystems that make data past
 145          * EOF available, and don't want it to be visible in both files. ZFS
 146          * doesn't do that, so we just turn the flag off.
 147          */
 148         flags &= ~REMAP_FILE_CAN_SHORTEN;
 149
 150         if (flags & REMAP_FILE_DEDUP)
 151                 /* No support for dedup yet */
 152                 return (-EOPNOTSUPP);
 153
 154         /* Zero length means to clone everything to the end of the file */
 155         if (len == 0)
 156                 len = i_size_read(file_inode(src_file)) - src_off;
 157
 158         return (__zpl_clone_file_range(src_file, src_off,
 159             dst_file, dst_off, len));
 160 }
 161 #endif /* HAVE_VFS_REMAP_FILE_RANGE */
 162
 163 #if defined(HAVE_VFS_CLONE_FILE_RANGE) || \
 164     defined(HAVE_VFS_FILE_OPERATIONS_EXTEND)
 165 /*
 166  * Entry point for FICLONE and FICLONERANGE, before Linux 4.20.
 167  */
 168 int
 169 zpl_clone_file_range(struct file *src_file, loff_t src_off,
 170     struct file *dst_file, loff_t dst_off, uint64_t len)
 171 {
 172         /* Zero length means to clone everything to the end of the file */
 173         if (len == 0)
 174                 len = i_size_read(file_inode(src_file)) - src_off;
 175
 176         return (__zpl_clone_file_range(src_file, src_off,
 177             dst_file, dst_off, len));
 178 }
 179 #endif /* HAVE_VFS_CLONE_FILE_RANGE || HAVE_VFS_FILE_OPERATIONS_EXTEND */
 180
 181 #ifdef HAVE_VFS_DEDUPE_FILE_RANGE
 182 /*
 183  * Entry point for FIDEDUPERANGE, before Linux 4.20.
 184  */
 185 int
 186 zpl_dedupe_file_range(struct file *src_file, loff_t src_off,
 187     struct file *dst_file, loff_t dst_off, uint64_t len)
 188 {
 189         /* No support for dedup yet */
 190         return (-EOPNOTSUPP);
 191 }
 192 #endif /* HAVE_VFS_DEDUPE_FILE_RANGE */
 193
 194 /* Entry point for FICLONE, before Linux 4.5. */
 195 long
 196 zpl_ioctl_ficlone(struct file *dst_file, void *arg)
 197 {
 198         unsigned long sfd = (unsigned long)arg;
 199
 200         struct file *src_file = fget(sfd);
 201         if (src_file == NULL)
 202                 return (-EBADF);
 203
 204         if (dst_file->f_op != src_file->f_op)
 205                 return (-EXDEV);
 206
 207         size_t len = i_size_read(file_inode(src_file));
 208
 209         ssize_t ret =
 210             __zpl_clone_file_range(src_file, 0, dst_file, 0, len);
 211
 212         fput(src_file);
 213
 214         if (ret < 0) {
 215                 if (ret == -EOPNOTSUPP)
 216                         return (-ENOTTY);
 217                 return (ret);
 218         }
 219
 220         if (ret != len)
 221                 return (-EINVAL);
 222
 223         return (0);
 224 }
 225
 226 /* Entry point for FICLONERANGE, before Linux 4.5. */
 227 long
 228 zpl_ioctl_ficlonerange(struct file *dst_file, void __user *arg)
 229 {
 230         zfs_ioc_compat_file_clone_range_t fcr;
 231
 232         if (copy_from_user(&fcr, arg, sizeof (fcr)))
 233                 return (-EFAULT);
 234
 235         struct file *src_file = fget(fcr.fcr_src_fd);
 236         if (src_file == NULL)
 237                 return (-EBADF);
 238
 239         if (dst_file->f_op != src_file->f_op)
 240                 return (-EXDEV);
 241
 242         size_t len = fcr.fcr_src_length;
 243         if (len == 0)
 244                 len = i_size_read(file_inode(src_file)) - fcr.fcr_src_offset;
 245
 246         ssize_t ret = __zpl_clone_file_range(src_file, fcr.fcr_src_offset,
 247             dst_file, fcr.fcr_dest_offset, len);
 248
 249         fput(src_file);
 250
 251         if (ret < 0) {
 252                 if (ret == -EOPNOTSUPP)
 253                         return (-ENOTTY);
 254                 return (ret);
 255         }
 256
 257         if (ret != len)
 258                 return (-EINVAL);
 259
 260         return (0);
 261 }
 262
 263 /* Entry point for FIDEDUPERANGE, before Linux 4.5. */
 264 long
 265 zpl_ioctl_fideduperange(struct file *filp, void *arg)
 266 {
 267         (void) arg;
 268
 269         /* No support for dedup yet */
 270         return (-ENOTTY);
 271 }