fs/f2fs/file.c

0001 // SPDX-License-Identifier: GPL-2.0
0002 /*
0003  * fs/f2fs/file.c
0004  *
0005  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
0006  *             http://www.samsung.com/
0007  */
0008 #include <linux/fs.h>
0009 #include <linux/f2fs_fs.h>
0010 #include <linux/stat.h>
0011 #include <linux/buffer_head.h>
0012 #include <linux/writeback.h>
0013 #include <linux/blkdev.h>
0014 #include <linux/falloc.h>
0015 #include <linux/types.h>
0016 #include <linux/compat.h>
0017 #include <linux/uaccess.h>
0018 #include <linux/mount.h>
0019 #include <linux/pagevec.h>
0020 #include <linux/uio.h>
0021 #include <linux/uuid.h>
0022 #include <linux/file.h>
0023 #include <linux/nls.h>
0024 #include <linux/sched/signal.h>
0025 #include <linux/fileattr.h>
0026 #include <linux/fadvise.h>
0027 #include <linux/iomap.h>
0028
0029 #include "f2fs.h"
0030 #include "node.h"
0031 #include "segment.h"
0032 #include "xattr.h"
0033 #include "acl.h"
0034 #include "gc.h"
0035 #include "iostat.h"
0036 #include <trace/events/f2fs.h>
0037 #include <uapi/linux/f2fs.h>
0038
0039 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
0040 {
0041     struct inode *inode = file_inode(vmf->vma->vm_file);
0042     vm_fault_t ret;
0043
0044     ret = filemap_fault(vmf);
0045     if (!ret)
0046         f2fs_update_iostat(F2FS_I_SB(inode), APP_MAPPED_READ_IO,
0047                             F2FS_BLKSIZE);
0048
0049     trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
0050
0051     return ret;
0052 }
0053
0054 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
0055 {
0056     struct page *page = vmf->page;
0057     struct inode *inode = file_inode(vmf->vma->vm_file);
0058     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0059     struct dnode_of_data dn;
0060     bool need_alloc = true;
0061     int err = 0;
0062
0063     if (unlikely(IS_IMMUTABLE(inode)))
0064         return VM_FAULT_SIGBUS;
0065
0066     if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
0067         return VM_FAULT_SIGBUS;
0068
0069     if (unlikely(f2fs_cp_error(sbi))) {
0070         err = -EIO;
0071         goto err;
0072     }
0073
0074     if (!f2fs_is_checkpoint_ready(sbi)) {
0075         err = -ENOSPC;
0076         goto err;
0077     }
0078
0079     err = f2fs_convert_inline_inode(inode);
0080     if (err)
0081         goto err;
0082
0083 #ifdef CONFIG_F2FS_FS_COMPRESSION
0084     if (f2fs_compressed_file(inode)) {
0085         int ret = f2fs_is_compressed_cluster(inode, page->index);
0086
0087         if (ret < 0) {
0088             err = ret;
0089             goto err;
0090         } else if (ret) {
0091             need_alloc = false;
0092         }
0093     }
0094 #endif
0095     /* should do out of any locked page */
0096     if (need_alloc)
0097         f2fs_balance_fs(sbi, true);
0098
0099     sb_start_pagefault(inode->i_sb);
0100
0101     f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
0102
0103     file_update_time(vmf->vma->vm_file);
0104     filemap_invalidate_lock_shared(inode->i_mapping);
0105     lock_page(page);
0106     if (unlikely(page->mapping != inode->i_mapping ||
0107             page_offset(page) > i_size_read(inode) ||
0108             !PageUptodate(page))) {
0109         unlock_page(page);
0110         err = -EFAULT;
0111         goto out_sem;
0112     }
0113
0114     if (need_alloc) {
0115         /* block allocation */
0116         f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
0117         set_new_dnode(&dn, inode, NULL, NULL, 0);
0118         err = f2fs_get_block(&dn, page->index);
0119         f2fs_do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
0120     }
0121
0122 #ifdef CONFIG_F2FS_FS_COMPRESSION
0123     if (!need_alloc) {
0124         set_new_dnode(&dn, inode, NULL, NULL, 0);
0125         err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
0126         f2fs_put_dnode(&dn);
0127     }
0128 #endif
0129     if (err) {
0130         unlock_page(page);
0131         goto out_sem;
0132     }
0133
0134     f2fs_wait_on_page_writeback(page, DATA, false, true);
0135
0136     /* wait for GCed page writeback via META_MAPPING */
0137     f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
0138
0139     /*
0140      * check to see if the page is mapped already (no holes)
0141      */
0142     if (PageMappedToDisk(page))
0143         goto out_sem;
0144
0145     /* page is wholly or partially inside EOF */
0146     if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
0147                         i_size_read(inode)) {
0148         loff_t offset;
0149
0150         offset = i_size_read(inode) & ~PAGE_MASK;
0151         zero_user_segment(page, offset, PAGE_SIZE);
0152     }
0153     set_page_dirty(page);
0154     if (!PageUptodate(page))
0155         SetPageUptodate(page);
0156
0157     f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
0158     f2fs_update_time(sbi, REQ_TIME);
0159
0160     trace_f2fs_vm_page_mkwrite(page, DATA);
0161 out_sem:
0162     filemap_invalidate_unlock_shared(inode->i_mapping);
0163
0164     sb_end_pagefault(inode->i_sb);
0165 err:
0166     return block_page_mkwrite_return(err);
0167 }
0168
0169 static const struct vm_operations_struct f2fs_file_vm_ops = {
0170     .fault      = f2fs_filemap_fault,
0171     .map_pages  = filemap_map_pages,
0172     .page_mkwrite   = f2fs_vm_page_mkwrite,
0173 };
0174
0175 static int get_parent_ino(struct inode *inode, nid_t *pino)
0176 {
0177     struct dentry *dentry;
0178
0179     /*
0180      * Make sure to get the non-deleted alias.  The alias associated with
0181      * the open file descriptor being fsync()'ed may be deleted already.
0182      */
0183     dentry = d_find_alias(inode);
0184     if (!dentry)
0185         return 0;
0186
0187     *pino = parent_ino(dentry);
0188     dput(dentry);
0189     return 1;
0190 }
0191
0192 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
0193 {
0194     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0195     enum cp_reason_type cp_reason = CP_NO_NEEDED;
0196
0197     if (!S_ISREG(inode->i_mode))
0198         cp_reason = CP_NON_REGULAR;
0199     else if (f2fs_compressed_file(inode))
0200         cp_reason = CP_COMPRESSED;
0201     else if (inode->i_nlink != 1)
0202         cp_reason = CP_HARDLINK;
0203     else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
0204         cp_reason = CP_SB_NEED_CP;
0205     else if (file_wrong_pino(inode))
0206         cp_reason = CP_WRONG_PINO;
0207     else if (!f2fs_space_for_roll_forward(sbi))
0208         cp_reason = CP_NO_SPC_ROLL;
0209     else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
0210         cp_reason = CP_NODE_NEED_CP;
0211     else if (test_opt(sbi, FASTBOOT))
0212         cp_reason = CP_FASTBOOT_MODE;
0213     else if (F2FS_OPTION(sbi).active_logs == 2)
0214         cp_reason = CP_SPEC_LOG_NUM;
0215     else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
0216         f2fs_need_dentry_mark(sbi, inode->i_ino) &&
0217         f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
0218                             TRANS_DIR_INO))
0219         cp_reason = CP_RECOVER_DIR;
0220
0221     return cp_reason;
0222 }
0223
0224 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
0225 {
0226     struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
0227     bool ret = false;
0228     /* But we need to avoid that there are some inode updates */
0229     if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
0230         ret = true;
0231     f2fs_put_page(i, 0);
0232     return ret;
0233 }
0234
0235 static void try_to_fix_pino(struct inode *inode)
0236 {
0237     struct f2fs_inode_info *fi = F2FS_I(inode);
0238     nid_t pino;
0239
0240     f2fs_down_write(&fi->i_sem);
0241     if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
0242             get_parent_ino(inode, &pino)) {
0243         f2fs_i_pino_write(inode, pino);
0244         file_got_pino(inode);
0245     }
0246     f2fs_up_write(&fi->i_sem);
0247 }
0248
0249 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
0250                         int datasync, bool atomic)
0251 {
0252     struct inode *inode = file->f_mapping->host;
0253     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0254     nid_t ino = inode->i_ino;
0255     int ret = 0;
0256     enum cp_reason_type cp_reason = 0;
0257     struct writeback_control wbc = {
0258         .sync_mode = WB_SYNC_ALL,
0259         .nr_to_write = LONG_MAX,
0260         .for_reclaim = 0,
0261     };
0262     unsigned int seq_id = 0;
0263
0264     if (unlikely(f2fs_readonly(inode->i_sb)))
0265         return 0;
0266
0267     trace_f2fs_sync_file_enter(inode);
0268
0269     if (S_ISDIR(inode->i_mode))
0270         goto go_write;
0271
0272     /* if fdatasync is triggered, let's do in-place-update */
0273     if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
0274         set_inode_flag(inode, FI_NEED_IPU);
0275     ret = file_write_and_wait_range(file, start, end);
0276     clear_inode_flag(inode, FI_NEED_IPU);
0277
0278     if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) {
0279         trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
0280         return ret;
0281     }
0282
0283     /* if the inode is dirty, let's recover all the time */
0284     if (!f2fs_skip_inode_update(inode, datasync)) {
0285         f2fs_write_inode(inode, NULL);
0286         goto go_write;
0287     }
0288
0289     /*
0290      * if there is no written data, don't waste time to write recovery info.
0291      */
0292     if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
0293             !f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
0294
0295         /* it may call write_inode just prior to fsync */
0296         if (need_inode_page_update(sbi, ino))
0297             goto go_write;
0298
0299         if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
0300                 f2fs_exist_written_data(sbi, ino, UPDATE_INO))
0301             goto flush_out;
0302         goto out;
0303     } else {
0304         /*
0305          * for OPU case, during fsync(), node can be persisted before
0306          * data when lower device doesn't support write barrier, result
0307          * in data corruption after SPO.
0308          * So for strict fsync mode, force to use atomic write sematics
0309          * to keep write order in between data/node and last node to
0310          * avoid potential data corruption.
0311          */
0312         if (F2FS_OPTION(sbi).fsync_mode ==
0313                 FSYNC_MODE_STRICT && !atomic)
0314             atomic = true;
0315     }
0316 go_write:
0317     /*
0318      * Both of fdatasync() and fsync() are able to be recovered from
0319      * sudden-power-off.
0320      */
0321     f2fs_down_read(&F2FS_I(inode)->i_sem);
0322     cp_reason = need_do_checkpoint(inode);
0323     f2fs_up_read(&F2FS_I(inode)->i_sem);
0324
0325     if (cp_reason) {
0326         /* all the dirty node pages should be flushed for POR */
0327         ret = f2fs_sync_fs(inode->i_sb, 1);
0328
0329         /*
0330          * We've secured consistency through sync_fs. Following pino
0331          * will be used only for fsynced inodes after checkpoint.
0332          */
0333         try_to_fix_pino(inode);
0334         clear_inode_flag(inode, FI_APPEND_WRITE);
0335         clear_inode_flag(inode, FI_UPDATE_WRITE);
0336         goto out;
0337     }
0338 sync_nodes:
0339     atomic_inc(&sbi->wb_sync_req[NODE]);
0340     ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
0341     atomic_dec(&sbi->wb_sync_req[NODE]);
0342     if (ret)
0343         goto out;
0344
0345     /* if cp_error was enabled, we should avoid infinite loop */
0346     if (unlikely(f2fs_cp_error(sbi))) {
0347         ret = -EIO;
0348         goto out;
0349     }
0350
0351     if (f2fs_need_inode_block_update(sbi, ino)) {
0352         f2fs_mark_inode_dirty_sync(inode, true);
0353         f2fs_write_inode(inode, NULL);
0354         goto sync_nodes;
0355     }
0356
0357     /*
0358      * If it's atomic_write, it's just fine to keep write ordering. So
0359      * here we don't need to wait for node write completion, since we use
0360      * node chain which serializes node blocks. If one of node writes are
0361      * reordered, we can see simply broken chain, resulting in stopping
0362      * roll-forward recovery. It means we'll recover all or none node blocks
0363      * given fsync mark.
0364      */
0365     if (!atomic) {
0366         ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
0367         if (ret)
0368             goto out;
0369     }
0370
0371     /* once recovery info is written, don't need to tack this */
0372     f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
0373     clear_inode_flag(inode, FI_APPEND_WRITE);
0374 flush_out:
0375     if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) ||
0376         (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi)))
0377         ret = f2fs_issue_flush(sbi, inode->i_ino);
0378     if (!ret) {
0379         f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
0380         clear_inode_flag(inode, FI_UPDATE_WRITE);
0381         f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
0382     }
0383     f2fs_update_time(sbi, REQ_TIME);
0384 out:
0385     trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
0386     return ret;
0387 }
0388
0389 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
0390 {
0391     if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
0392         return -EIO;
0393     return f2fs_do_sync_file(file, start, end, datasync, false);
0394 }
0395
0396 static bool __found_offset(struct address_space *mapping, block_t blkaddr,
0397                 pgoff_t index, int whence)
0398 {
0399     switch (whence) {
0400     case SEEK_DATA:
0401         if (__is_valid_data_blkaddr(blkaddr))
0402             return true;
0403         if (blkaddr == NEW_ADDR &&
0404             xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY))
0405             return true;
0406         break;
0407     case SEEK_HOLE:
0408         if (blkaddr == NULL_ADDR)
0409             return true;
0410         break;
0411     }
0412     return false;
0413 }
0414
0415 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
0416 {
0417     struct inode *inode = file->f_mapping->host;
0418     loff_t maxbytes = inode->i_sb->s_maxbytes;
0419     struct dnode_of_data dn;
0420     pgoff_t pgofs, end_offset;
0421     loff_t data_ofs = offset;
0422     loff_t isize;
0423     int err = 0;
0424
0425     inode_lock(inode);
0426
0427     isize = i_size_read(inode);
0428     if (offset >= isize)
0429         goto fail;
0430
0431     /* handle inline data case */
0432     if (f2fs_has_inline_data(inode)) {
0433         if (whence == SEEK_HOLE) {
0434             data_ofs = isize;
0435             goto found;
0436         } else if (whence == SEEK_DATA) {
0437             data_ofs = offset;
0438             goto found;
0439         }
0440     }
0441
0442     pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
0443
0444     for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
0445         set_new_dnode(&dn, inode, NULL, NULL, 0);
0446         err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
0447         if (err && err != -ENOENT) {
0448             goto fail;
0449         } else if (err == -ENOENT) {
0450             /* direct node does not exists */
0451             if (whence == SEEK_DATA) {
0452                 pgofs = f2fs_get_next_page_offset(&dn, pgofs);
0453                 continue;
0454             } else {
0455                 goto found;
0456             }
0457         }
0458
0459         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
0460
0461         /* find data/hole in dnode block */
0462         for (; dn.ofs_in_node < end_offset;
0463                 dn.ofs_in_node++, pgofs++,
0464                 data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
0465             block_t blkaddr;
0466
0467             blkaddr = f2fs_data_blkaddr(&dn);
0468
0469             if (__is_valid_data_blkaddr(blkaddr) &&
0470                 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
0471                     blkaddr, DATA_GENERIC_ENHANCE)) {
0472                 f2fs_put_dnode(&dn);
0473                 goto fail;
0474             }
0475
0476             if (__found_offset(file->f_mapping, blkaddr,
0477                             pgofs, whence)) {
0478                 f2fs_put_dnode(&dn);
0479                 goto found;
0480             }
0481         }
0482         f2fs_put_dnode(&dn);
0483     }
0484
0485     if (whence == SEEK_DATA)
0486         goto fail;
0487 found:
0488     if (whence == SEEK_HOLE && data_ofs > isize)
0489         data_ofs = isize;
0490     inode_unlock(inode);
0491     return vfs_setpos(file, data_ofs, maxbytes);
0492 fail:
0493     inode_unlock(inode);
0494     return -ENXIO;
0495 }
0496
0497 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
0498 {
0499     struct inode *inode = file->f_mapping->host;
0500     loff_t maxbytes = inode->i_sb->s_maxbytes;
0501
0502     if (f2fs_compressed_file(inode))
0503         maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS;
0504
0505     switch (whence) {
0506     case SEEK_SET:
0507     case SEEK_CUR:
0508     case SEEK_END:
0509         return generic_file_llseek_size(file, offset, whence,
0510                         maxbytes, i_size_read(inode));
0511     case SEEK_DATA:
0512     case SEEK_HOLE:
0513         if (offset < 0)
0514             return -ENXIO;
0515         return f2fs_seek_block(file, offset, whence);
0516     }
0517
0518     return -EINVAL;
0519 }
0520
0521 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
0522 {
0523     struct inode *inode = file_inode(file);
0524
0525     if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
0526         return -EIO;
0527
0528     if (!f2fs_is_compress_backend_ready(inode))
0529         return -EOPNOTSUPP;
0530
0531     file_accessed(file);
0532     vma->vm_ops = &f2fs_file_vm_ops;
0533     set_inode_flag(inode, FI_MMAP_FILE);
0534     return 0;
0535 }
0536
0537 static int f2fs_file_open(struct inode *inode, struct file *filp)
0538 {
0539     int err = fscrypt_file_open(inode, filp);
0540
0541     if (err)
0542         return err;
0543
0544     if (!f2fs_is_compress_backend_ready(inode))
0545         return -EOPNOTSUPP;
0546
0547     err = fsverity_file_open(inode, filp);
0548     if (err)
0549         return err;
0550
0551     filp->f_mode |= FMODE_NOWAIT;
0552
0553     return dquot_file_open(inode, filp);
0554 }
0555
0556 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
0557 {
0558     struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
0559     struct f2fs_node *raw_node;
0560     int nr_free = 0, ofs = dn->ofs_in_node, len = count;
0561     __le32 *addr;
0562     int base = 0;
0563     bool compressed_cluster = false;
0564     int cluster_index = 0, valid_blocks = 0;
0565     int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
0566     bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks);
0567
0568     if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
0569         base = get_extra_isize(dn->inode);
0570
0571     raw_node = F2FS_NODE(dn->node_page);
0572     addr = blkaddr_in_node(raw_node) + base + ofs;
0573
0574     /* Assumption: truncateion starts with cluster */
0575     for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
0576         block_t blkaddr = le32_to_cpu(*addr);
0577
0578         if (f2fs_compressed_file(dn->inode) &&
0579                     !(cluster_index & (cluster_size - 1))) {
0580             if (compressed_cluster)
0581                 f2fs_i_compr_blocks_update(dn->inode,
0582                             valid_blocks, false);
0583             compressed_cluster = (blkaddr == COMPRESS_ADDR);
0584             valid_blocks = 0;
0585         }
0586
0587         if (blkaddr == NULL_ADDR)
0588             continue;
0589
0590         dn->data_blkaddr = NULL_ADDR;
0591         f2fs_set_data_blkaddr(dn);
0592
0593         if (__is_valid_data_blkaddr(blkaddr)) {
0594             if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
0595                     DATA_GENERIC_ENHANCE))
0596                 continue;
0597             if (compressed_cluster)
0598                 valid_blocks++;
0599         }
0600
0601         if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
0602             clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
0603
0604         f2fs_invalidate_blocks(sbi, blkaddr);
0605
0606         if (!released || blkaddr != COMPRESS_ADDR)
0607             nr_free++;
0608     }
0609
0610     if (compressed_cluster)
0611         f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
0612
0613     if (nr_free) {
0614         pgoff_t fofs;
0615         /*
0616          * once we invalidate valid blkaddr in range [ofs, ofs + count],
0617          * we will invalidate all blkaddr in the whole range.
0618          */
0619         fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
0620                             dn->inode) + ofs;
0621         f2fs_update_extent_cache_range(dn, fofs, 0, len);
0622         dec_valid_block_count(sbi, dn->inode, nr_free);
0623     }
0624     dn->ofs_in_node = ofs;
0625
0626     f2fs_update_time(sbi, REQ_TIME);
0627     trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
0628                      dn->ofs_in_node, nr_free);
0629 }
0630
0631 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
0632 {
0633     f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
0634 }
0635
0636 static int truncate_partial_data_page(struct inode *inode, u64 from,
0637                                 bool cache_only)
0638 {
0639     loff_t offset = from & (PAGE_SIZE - 1);
0640     pgoff_t index = from >> PAGE_SHIFT;
0641     struct address_space *mapping = inode->i_mapping;
0642     struct page *page;
0643
0644     if (!offset && !cache_only)
0645         return 0;
0646
0647     if (cache_only) {
0648         page = find_lock_page(mapping, index);
0649         if (page && PageUptodate(page))
0650             goto truncate_out;
0651         f2fs_put_page(page, 1);
0652         return 0;
0653     }
0654
0655     page = f2fs_get_lock_data_page(inode, index, true);
0656     if (IS_ERR(page))
0657         return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
0658 truncate_out:
0659     f2fs_wait_on_page_writeback(page, DATA, true, true);
0660     zero_user(page, offset, PAGE_SIZE - offset);
0661
0662     /* An encrypted inode should have a key and truncate the last page. */
0663     f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
0664     if (!cache_only)
0665         set_page_dirty(page);
0666     f2fs_put_page(page, 1);
0667     return 0;
0668 }
0669
0670 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock)
0671 {
0672     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
0673     struct dnode_of_data dn;
0674     pgoff_t free_from;
0675     int count = 0, err = 0;
0676     struct page *ipage;
0677     bool truncate_page = false;
0678
0679     trace_f2fs_truncate_blocks_enter(inode, from);
0680
0681     free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
0682
0683     if (free_from >= max_file_blocks(inode))
0684         goto free_partial;
0685
0686     if (lock)
0687         f2fs_lock_op(sbi);
0688
0689     ipage = f2fs_get_node_page(sbi, inode->i_ino);
0690     if (IS_ERR(ipage)) {
0691         err = PTR_ERR(ipage);
0692         goto out;
0693     }
0694
0695     if (f2fs_has_inline_data(inode)) {
0696         f2fs_truncate_inline_inode(inode, ipage, from);
0697         f2fs_put_page(ipage, 1);
0698         truncate_page = true;
0699         goto out;
0700     }
0701
0702     set_new_dnode(&dn, inode, ipage, NULL, 0);
0703     err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
0704     if (err) {
0705         if (err == -ENOENT)
0706             goto free_next;
0707         goto out;
0708     }
0709
0710     count = ADDRS_PER_PAGE(dn.node_page, inode);
0711
0712     count -= dn.ofs_in_node;
0713     f2fs_bug_on(sbi, count < 0);
0714
0715     if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
0716         f2fs_truncate_data_blocks_range(&dn, count);
0717         free_from += count;
0718     }
0719
0720     f2fs_put_dnode(&dn);
0721 free_next:
0722     err = f2fs_truncate_inode_blocks(inode, free_from);
0723 out:
0724     if (lock)
0725         f2fs_unlock_op(sbi);
0726 free_partial:
0727     /* lastly zero out the first data page */
0728     if (!err)
0729         err = truncate_partial_data_page(inode, from, truncate_page);
0730
0731     trace_f2fs_truncate_blocks_exit(inode, err);
0732     return err;
0733 }
0734
0735 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
0736 {
0737     u64 free_from = from;
0738     int err;
0739
0740 #ifdef CONFIG_F2FS_FS_COMPRESSION
0741     /*
0742      * for compressed file, only support cluster size
0743      * aligned truncation.
0744      */
0745     if (f2fs_compressed_file(inode))
0746         free_from = round_up(from,
0747                 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT);
0748 #endif
0749
0750     err = f2fs_do_truncate_blocks(inode, free_from, lock);
0751     if (err)
0752         return err;
0753
0754 #ifdef CONFIG_F2FS_FS_COMPRESSION
0755     /*
0756      * For compressed file, after release compress blocks, don't allow write
0757      * direct, but we should allow write direct after truncate to zero.
0758      */
0759     if (f2fs_compressed_file(inode) && !free_from
0760             && is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
0761         clear_inode_flag(inode, FI_COMPRESS_RELEASED);
0762
0763     if (from != free_from) {
0764         err = f2fs_truncate_partial_cluster(inode, from, lock);
0765         if (err)
0766             return err;
0767     }
0768 #endif
0769
0770     return 0;
0771 }
0772
0773 int f2fs_truncate(struct inode *inode)
0774 {
0775     int err;
0776
0777     if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
0778         return -EIO;
0779
0780     if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
0781                 S_ISLNK(inode->i_mode)))
0782         return 0;
0783
0784     trace_f2fs_truncate(inode);
0785
0786     if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
0787         f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
0788         return -EIO;
0789     }
0790
0791     err = f2fs_dquot_initialize(inode);
0792     if (err)
0793         return err;
0794
0795     /* we should check inline_data size */
0796     if (!f2fs_may_inline_data(inode)) {
0797         err = f2fs_convert_inline_inode(inode);
0798         if (err)
0799             return err;
0800     }
0801
0802     err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
0803     if (err)
0804         return err;
0805
0806     inode->i_mtime = inode->i_ctime = current_time(inode);
0807     f2fs_mark_inode_dirty_sync(inode, false);
0808     return 0;
0809 }
0810
0811 int f2fs_getattr(struct user_namespace *mnt_userns, const struct path *path,
0812          struct kstat *stat, u32 request_mask, unsigned int query_flags)
0813 {
0814     struct inode *inode = d_inode(path->dentry);
0815     struct f2fs_inode_info *fi = F2FS_I(inode);
0816     struct f2fs_inode *ri = NULL;
0817     unsigned int flags;
0818
0819     if (f2fs_has_extra_attr(inode) &&
0820             f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
0821             F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
0822         stat->result_mask |= STATX_BTIME;
0823         stat->btime.tv_sec = fi->i_crtime.tv_sec;
0824         stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
0825     }
0826
0827     flags = fi->i_flags;
0828     if (flags & F2FS_COMPR_FL)
0829         stat->attributes |= STATX_ATTR_COMPRESSED;
0830     if (flags & F2FS_APPEND_FL)
0831         stat->attributes |= STATX_ATTR_APPEND;
0832     if (IS_ENCRYPTED(inode))
0833         stat->attributes |= STATX_ATTR_ENCRYPTED;
0834     if (flags & F2FS_IMMUTABLE_FL)
0835         stat->attributes |= STATX_ATTR_IMMUTABLE;
0836     if (flags & F2FS_NODUMP_FL)
0837         stat->attributes |= STATX_ATTR_NODUMP;
0838     if (IS_VERITY(inode))
0839         stat->attributes |= STATX_ATTR_VERITY;
0840
0841     stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
0842                   STATX_ATTR_APPEND |
0843                   STATX_ATTR_ENCRYPTED |
0844                   STATX_ATTR_IMMUTABLE |
0845                   STATX_ATTR_NODUMP |
0846                   STATX_ATTR_VERITY);
0847
0848     generic_fillattr(mnt_userns, inode, stat);
0849
0850     /* we need to show initial sectors used for inline_data/dentries */
0851     if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
0852                     f2fs_has_inline_dentry(inode))
0853         stat->blocks += (stat->size + 511) >> 9;
0854
0855     return 0;
0856 }
0857
0858 #ifdef CONFIG_F2FS_FS_POSIX_ACL
0859 static void __setattr_copy(struct user_namespace *mnt_userns,
0860                struct inode *inode, const struct iattr *attr)
0861 {
0862     unsigned int ia_valid = attr->ia_valid;
0863
0864     i_uid_update(mnt_userns, attr, inode);
0865     i_gid_update(mnt_userns, attr, inode);
0866     if (ia_valid & ATTR_ATIME)
0867         inode->i_atime = attr->ia_atime;
0868     if (ia_valid & ATTR_MTIME)
0869         inode->i_mtime = attr->ia_mtime;
0870     if (ia_valid & ATTR_CTIME)
0871         inode->i_ctime = attr->ia_ctime;
0872     if (ia_valid & ATTR_MODE) {
0873         umode_t mode = attr->ia_mode;
0874         kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
0875
0876         if (!in_group_p(kgid) && !capable_wrt_inode_uidgid(mnt_userns, inode, CAP_FSETID))
0877             mode &= ~S_ISGID;
0878         set_acl_inode(inode, mode);
0879     }
0880 }
0881 #else
0882 #define __setattr_copy setattr_copy
0883 #endif
0884
0885 int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
0886          struct iattr *attr)
0887 {
0888     struct inode *inode = d_inode(dentry);
0889     int err;
0890
0891     if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
0892         return -EIO;
0893
0894     if (unlikely(IS_IMMUTABLE(inode)))
0895         return -EPERM;
0896
0897     if (unlikely(IS_APPEND(inode) &&
0898             (attr->ia_valid & (ATTR_MODE | ATTR_UID |
0899                   ATTR_GID | ATTR_TIMES_SET))))
0900         return -EPERM;
0901
0902     if ((attr->ia_valid & ATTR_SIZE) &&
0903         !f2fs_is_compress_backend_ready(inode))
0904         return -EOPNOTSUPP;
0905
0906     err = setattr_prepare(mnt_userns, dentry, attr);
0907     if (err)
0908         return err;
0909
0910     err = fscrypt_prepare_setattr(dentry, attr);
0911     if (err)
0912         return err;
0913
0914     err = fsverity_prepare_setattr(dentry, attr);
0915     if (err)
0916         return err;
0917
0918     if (is_quota_modification(mnt_userns, inode, attr)) {
0919         err = f2fs_dquot_initialize(inode);
0920         if (err)
0921             return err;
0922     }
0923     if (i_uid_needs_update(mnt_userns, attr, inode) ||
0924         i_gid_needs_update(mnt_userns, attr, inode)) {
0925         f2fs_lock_op(F2FS_I_SB(inode));
0926         err = dquot_transfer(mnt_userns, inode, attr);
0927         if (err) {
0928             set_sbi_flag(F2FS_I_SB(inode),
0929                     SBI_QUOTA_NEED_REPAIR);
0930             f2fs_unlock_op(F2FS_I_SB(inode));
0931             return err;
0932         }
0933         /*
0934          * update uid/gid under lock_op(), so that dquot and inode can
0935          * be updated atomically.
0936          */
0937         i_uid_update(mnt_userns, attr, inode);
0938         i_gid_update(mnt_userns, attr, inode);
0939         f2fs_mark_inode_dirty_sync(inode, true);
0940         f2fs_unlock_op(F2FS_I_SB(inode));
0941     }
0942
0943     if (attr->ia_valid & ATTR_SIZE) {
0944         loff_t old_size = i_size_read(inode);
0945
0946         if (attr->ia_size > MAX_INLINE_DATA(inode)) {
0947             /*
0948              * should convert inline inode before i_size_write to
0949              * keep smaller than inline_data size with inline flag.
0950              */
0951             err = f2fs_convert_inline_inode(inode);
0952             if (err)
0953                 return err;
0954         }
0955
0956         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
0957         filemap_invalidate_lock(inode->i_mapping);
0958
0959         truncate_setsize(inode, attr->ia_size);
0960
0961         if (attr->ia_size <= old_size)
0962             err = f2fs_truncate(inode);
0963         /*
0964          * do not trim all blocks after i_size if target size is
0965          * larger than i_size.
0966          */
0967         filemap_invalidate_unlock(inode->i_mapping);
0968         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
0969         if (err)
0970             return err;
0971
0972         spin_lock(&F2FS_I(inode)->i_size_lock);
0973         inode->i_mtime = inode->i_ctime = current_time(inode);
0974         F2FS_I(inode)->last_disk_size = i_size_read(inode);
0975         spin_unlock(&F2FS_I(inode)->i_size_lock);
0976     }
0977
0978     __setattr_copy(mnt_userns, inode, attr);
0979
0980     if (attr->ia_valid & ATTR_MODE) {
0981         err = posix_acl_chmod(mnt_userns, inode, f2fs_get_inode_mode(inode));
0982
0983         if (is_inode_flag_set(inode, FI_ACL_MODE)) {
0984             if (!err)
0985                 inode->i_mode = F2FS_I(inode)->i_acl_mode;
0986             clear_inode_flag(inode, FI_ACL_MODE);
0987         }
0988     }
0989
0990     /* file size may changed here */
0991     f2fs_mark_inode_dirty_sync(inode, true);
0992
0993     /* inode change will produce dirty node pages flushed by checkpoint */
0994     f2fs_balance_fs(F2FS_I_SB(inode), true);
0995
0996     return err;
0997 }
0998
0999 const struct inode_operations f2fs_file_inode_operations = {
1000     .getattr    = f2fs_getattr,
1001     .setattr    = f2fs_setattr,
1002     .get_acl    = f2fs_get_acl,
1003     .set_acl    = f2fs_set_acl,
1004     .listxattr  = f2fs_listxattr,
1005     .fiemap     = f2fs_fiemap,
1006     .fileattr_get   = f2fs_fileattr_get,
1007     .fileattr_set   = f2fs_fileattr_set,
1008 };
1009
1010 static int fill_zero(struct inode *inode, pgoff_t index,
1011                     loff_t start, loff_t len)
1012 {
1013     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1014     struct page *page;
1015
1016     if (!len)
1017         return 0;
1018
1019     f2fs_balance_fs(sbi, true);
1020
1021     f2fs_lock_op(sbi);
1022     page = f2fs_get_new_data_page(inode, NULL, index, false);
1023     f2fs_unlock_op(sbi);
1024
1025     if (IS_ERR(page))
1026         return PTR_ERR(page);
1027
1028     f2fs_wait_on_page_writeback(page, DATA, true, true);
1029     zero_user(page, start, len);
1030     set_page_dirty(page);
1031     f2fs_put_page(page, 1);
1032     return 0;
1033 }
1034
1035 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
1036 {
1037     int err;
1038
1039     while (pg_start < pg_end) {
1040         struct dnode_of_data dn;
1041         pgoff_t end_offset, count;
1042
1043         set_new_dnode(&dn, inode, NULL, NULL, 0);
1044         err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1045         if (err) {
1046             if (err == -ENOENT) {
1047                 pg_start = f2fs_get_next_page_offset(&dn,
1048                                 pg_start);
1049                 continue;
1050             }
1051             return err;
1052         }
1053
1054         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1055         count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1056
1057         f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1058
1059         f2fs_truncate_data_blocks_range(&dn, count);
1060         f2fs_put_dnode(&dn);
1061
1062         pg_start += count;
1063     }
1064     return 0;
1065 }
1066
1067 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1068 {
1069     pgoff_t pg_start, pg_end;
1070     loff_t off_start, off_end;
1071     int ret;
1072
1073     ret = f2fs_convert_inline_inode(inode);
1074     if (ret)
1075         return ret;
1076
1077     pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1078     pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1079
1080     off_start = offset & (PAGE_SIZE - 1);
1081     off_end = (offset + len) & (PAGE_SIZE - 1);
1082
1083     if (pg_start == pg_end) {
1084         ret = fill_zero(inode, pg_start, off_start,
1085                         off_end - off_start);
1086         if (ret)
1087             return ret;
1088     } else {
1089         if (off_start) {
1090             ret = fill_zero(inode, pg_start++, off_start,
1091                         PAGE_SIZE - off_start);
1092             if (ret)
1093                 return ret;
1094         }
1095         if (off_end) {
1096             ret = fill_zero(inode, pg_end, 0, off_end);
1097             if (ret)
1098                 return ret;
1099         }
1100
1101         if (pg_start < pg_end) {
1102             loff_t blk_start, blk_end;
1103             struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1104
1105             f2fs_balance_fs(sbi, true);
1106
1107             blk_start = (loff_t)pg_start << PAGE_SHIFT;
1108             blk_end = (loff_t)pg_end << PAGE_SHIFT;
1109
1110             f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1111             filemap_invalidate_lock(inode->i_mapping);
1112
1113             truncate_pagecache_range(inode, blk_start, blk_end - 1);
1114
1115             f2fs_lock_op(sbi);
1116             ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1117             f2fs_unlock_op(sbi);
1118
1119             filemap_invalidate_unlock(inode->i_mapping);
1120             f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1121         }
1122     }
1123
1124     return ret;
1125 }
1126
1127 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1128                 int *do_replace, pgoff_t off, pgoff_t len)
1129 {
1130     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1131     struct dnode_of_data dn;
1132     int ret, done, i;
1133
1134 next_dnode:
1135     set_new_dnode(&dn, inode, NULL, NULL, 0);
1136     ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1137     if (ret && ret != -ENOENT) {
1138         return ret;
1139     } else if (ret == -ENOENT) {
1140         if (dn.max_level == 0)
1141             return -ENOENT;
1142         done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1143                         dn.ofs_in_node, len);
1144         blkaddr += done;
1145         do_replace += done;
1146         goto next;
1147     }
1148
1149     done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1150                             dn.ofs_in_node, len);
1151     for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1152         *blkaddr = f2fs_data_blkaddr(&dn);
1153
1154         if (__is_valid_data_blkaddr(*blkaddr) &&
1155             !f2fs_is_valid_blkaddr(sbi, *blkaddr,
1156                     DATA_GENERIC_ENHANCE)) {
1157             f2fs_put_dnode(&dn);
1158             return -EFSCORRUPTED;
1159         }
1160
1161         if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1162
1163             if (f2fs_lfs_mode(sbi)) {
1164                 f2fs_put_dnode(&dn);
1165                 return -EOPNOTSUPP;
1166             }
1167
1168             /* do not invalidate this block address */
1169             f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1170             *do_replace = 1;
1171         }
1172     }
1173     f2fs_put_dnode(&dn);
1174 next:
1175     len -= done;
1176     off += done;
1177     if (len)
1178         goto next_dnode;
1179     return 0;
1180 }
1181
1182 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1183                 int *do_replace, pgoff_t off, int len)
1184 {
1185     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1186     struct dnode_of_data dn;
1187     int ret, i;
1188
1189     for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1190         if (*do_replace == 0)
1191             continue;
1192
1193         set_new_dnode(&dn, inode, NULL, NULL, 0);
1194         ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1195         if (ret) {
1196             dec_valid_block_count(sbi, inode, 1);
1197             f2fs_invalidate_blocks(sbi, *blkaddr);
1198         } else {
1199             f2fs_update_data_blkaddr(&dn, *blkaddr);
1200         }
1201         f2fs_put_dnode(&dn);
1202     }
1203     return 0;
1204 }
1205
1206 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1207             block_t *blkaddr, int *do_replace,
1208             pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1209 {
1210     struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1211     pgoff_t i = 0;
1212     int ret;
1213
1214     while (i < len) {
1215         if (blkaddr[i] == NULL_ADDR && !full) {
1216             i++;
1217             continue;
1218         }
1219
1220         if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1221             struct dnode_of_data dn;
1222             struct node_info ni;
1223             size_t new_size;
1224             pgoff_t ilen;
1225
1226             set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1227             ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1228             if (ret)
1229                 return ret;
1230
1231             ret = f2fs_get_node_info(sbi, dn.nid, &ni, false);
1232             if (ret) {
1233                 f2fs_put_dnode(&dn);
1234                 return ret;
1235             }
1236
1237             ilen = min((pgoff_t)
1238                 ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1239                         dn.ofs_in_node, len - i);
1240             do {
1241                 dn.data_blkaddr = f2fs_data_blkaddr(&dn);
1242                 f2fs_truncate_data_blocks_range(&dn, 1);
1243
1244                 if (do_replace[i]) {
1245                     f2fs_i_blocks_write(src_inode,
1246                             1, false, false);
1247                     f2fs_i_blocks_write(dst_inode,
1248                             1, true, false);
1249                     f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1250                     blkaddr[i], ni.version, true, false);
1251
1252                     do_replace[i] = 0;
1253                 }
1254                 dn.ofs_in_node++;
1255                 i++;
1256                 new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1257                 if (dst_inode->i_size < new_size)
1258                     f2fs_i_size_write(dst_inode, new_size);
1259             } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1260
1261             f2fs_put_dnode(&dn);
1262         } else {
1263             struct page *psrc, *pdst;
1264
1265             psrc = f2fs_get_lock_data_page(src_inode,
1266                             src + i, true);
1267             if (IS_ERR(psrc))
1268                 return PTR_ERR(psrc);
1269             pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1270                                 true);
1271             if (IS_ERR(pdst)) {
1272                 f2fs_put_page(psrc, 1);
1273                 return PTR_ERR(pdst);
1274             }
1275             memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE);
1276             set_page_dirty(pdst);
1277             f2fs_put_page(pdst, 1);
1278             f2fs_put_page(psrc, 1);
1279
1280             ret = f2fs_truncate_hole(src_inode,
1281                         src + i, src + i + 1);
1282             if (ret)
1283                 return ret;
1284             i++;
1285         }
1286     }
1287     return 0;
1288 }
1289
1290 static int __exchange_data_block(struct inode *src_inode,
1291             struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1292             pgoff_t len, bool full)
1293 {
1294     block_t *src_blkaddr;
1295     int *do_replace;
1296     pgoff_t olen;
1297     int ret;
1298
1299     while (len) {
1300         olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1301
1302         src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1303                     array_size(olen, sizeof(block_t)),
1304                     GFP_NOFS);
1305         if (!src_blkaddr)
1306             return -ENOMEM;
1307
1308         do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1309                     array_size(olen, sizeof(int)),
1310                     GFP_NOFS);
1311         if (!do_replace) {
1312             kvfree(src_blkaddr);
1313             return -ENOMEM;
1314         }
1315
1316         ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1317                     do_replace, src, olen);
1318         if (ret)
1319             goto roll_back;
1320
1321         ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1322                     do_replace, src, dst, olen, full);
1323         if (ret)
1324             goto roll_back;
1325
1326         src += olen;
1327         dst += olen;
1328         len -= olen;
1329
1330         kvfree(src_blkaddr);
1331         kvfree(do_replace);
1332     }
1333     return 0;
1334
1335 roll_back:
1336     __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1337     kvfree(src_blkaddr);
1338     kvfree(do_replace);
1339     return ret;
1340 }
1341
1342 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1343 {
1344     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1345     pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1346     pgoff_t start = offset >> PAGE_SHIFT;
1347     pgoff_t end = (offset + len) >> PAGE_SHIFT;
1348     int ret;
1349
1350     f2fs_balance_fs(sbi, true);
1351
1352     /* avoid gc operation during block exchange */
1353     f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1354     filemap_invalidate_lock(inode->i_mapping);
1355
1356     f2fs_lock_op(sbi);
1357     f2fs_drop_extent_tree(inode);
1358     truncate_pagecache(inode, offset);
1359     ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1360     f2fs_unlock_op(sbi);
1361
1362     filemap_invalidate_unlock(inode->i_mapping);
1363     f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1364     return ret;
1365 }
1366
1367 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1368 {
1369     loff_t new_size;
1370     int ret;
1371
1372     if (offset + len >= i_size_read(inode))
1373         return -EINVAL;
1374
1375     /* collapse range should be aligned to block size of f2fs. */
1376     if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1377         return -EINVAL;
1378
1379     ret = f2fs_convert_inline_inode(inode);
1380     if (ret)
1381         return ret;
1382
1383     /* write out all dirty pages from offset */
1384     ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1385     if (ret)
1386         return ret;
1387
1388     ret = f2fs_do_collapse(inode, offset, len);
1389     if (ret)
1390         return ret;
1391
1392     /* write out all moved pages, if possible */
1393     filemap_invalidate_lock(inode->i_mapping);
1394     filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1395     truncate_pagecache(inode, offset);
1396
1397     new_size = i_size_read(inode) - len;
1398     ret = f2fs_truncate_blocks(inode, new_size, true);
1399     filemap_invalidate_unlock(inode->i_mapping);
1400     if (!ret)
1401         f2fs_i_size_write(inode, new_size);
1402     return ret;
1403 }
1404
1405 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1406                                 pgoff_t end)
1407 {
1408     struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1409     pgoff_t index = start;
1410     unsigned int ofs_in_node = dn->ofs_in_node;
1411     blkcnt_t count = 0;
1412     int ret;
1413
1414     for (; index < end; index++, dn->ofs_in_node++) {
1415         if (f2fs_data_blkaddr(dn) == NULL_ADDR)
1416             count++;
1417     }
1418
1419     dn->ofs_in_node = ofs_in_node;
1420     ret = f2fs_reserve_new_blocks(dn, count);
1421     if (ret)
1422         return ret;
1423
1424     dn->ofs_in_node = ofs_in_node;
1425     for (index = start; index < end; index++, dn->ofs_in_node++) {
1426         dn->data_blkaddr = f2fs_data_blkaddr(dn);
1427         /*
1428          * f2fs_reserve_new_blocks will not guarantee entire block
1429          * allocation.
1430          */
1431         if (dn->data_blkaddr == NULL_ADDR) {
1432             ret = -ENOSPC;
1433             break;
1434         }
1435
1436         if (dn->data_blkaddr == NEW_ADDR)
1437             continue;
1438
1439         if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr,
1440                     DATA_GENERIC_ENHANCE)) {
1441             ret = -EFSCORRUPTED;
1442             break;
1443         }
1444
1445         f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1446         dn->data_blkaddr = NEW_ADDR;
1447         f2fs_set_data_blkaddr(dn);
1448     }
1449
1450     f2fs_update_extent_cache_range(dn, start, 0, index - start);
1451
1452     return ret;
1453 }
1454
1455 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1456                                 int mode)
1457 {
1458     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1459     struct address_space *mapping = inode->i_mapping;
1460     pgoff_t index, pg_start, pg_end;
1461     loff_t new_size = i_size_read(inode);
1462     loff_t off_start, off_end;
1463     int ret = 0;
1464
1465     ret = inode_newsize_ok(inode, (len + offset));
1466     if (ret)
1467         return ret;
1468
1469     ret = f2fs_convert_inline_inode(inode);
1470     if (ret)
1471         return ret;
1472
1473     ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1474     if (ret)
1475         return ret;
1476
1477     pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1478     pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1479
1480     off_start = offset & (PAGE_SIZE - 1);
1481     off_end = (offset + len) & (PAGE_SIZE - 1);
1482
1483     if (pg_start == pg_end) {
1484         ret = fill_zero(inode, pg_start, off_start,
1485                         off_end - off_start);
1486         if (ret)
1487             return ret;
1488
1489         new_size = max_t(loff_t, new_size, offset + len);
1490     } else {
1491         if (off_start) {
1492             ret = fill_zero(inode, pg_start++, off_start,
1493                         PAGE_SIZE - off_start);
1494             if (ret)
1495                 return ret;
1496
1497             new_size = max_t(loff_t, new_size,
1498                     (loff_t)pg_start << PAGE_SHIFT);
1499         }
1500
1501         for (index = pg_start; index < pg_end;) {
1502             struct dnode_of_data dn;
1503             unsigned int end_offset;
1504             pgoff_t end;
1505
1506             f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1507             filemap_invalidate_lock(mapping);
1508
1509             truncate_pagecache_range(inode,
1510                 (loff_t)index << PAGE_SHIFT,
1511                 ((loff_t)pg_end << PAGE_SHIFT) - 1);
1512
1513             f2fs_lock_op(sbi);
1514
1515             set_new_dnode(&dn, inode, NULL, NULL, 0);
1516             ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1517             if (ret) {
1518                 f2fs_unlock_op(sbi);
1519                 filemap_invalidate_unlock(mapping);
1520                 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1521                 goto out;
1522             }
1523
1524             end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1525             end = min(pg_end, end_offset - dn.ofs_in_node + index);
1526
1527             ret = f2fs_do_zero_range(&dn, index, end);
1528             f2fs_put_dnode(&dn);
1529
1530             f2fs_unlock_op(sbi);
1531             filemap_invalidate_unlock(mapping);
1532             f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1533
1534             f2fs_balance_fs(sbi, dn.node_changed);
1535
1536             if (ret)
1537                 goto out;
1538
1539             index = end;
1540             new_size = max_t(loff_t, new_size,
1541                     (loff_t)index << PAGE_SHIFT);
1542         }
1543
1544         if (off_end) {
1545             ret = fill_zero(inode, pg_end, 0, off_end);
1546             if (ret)
1547                 goto out;
1548
1549             new_size = max_t(loff_t, new_size, offset + len);
1550         }
1551     }
1552
1553 out:
1554     if (new_size > i_size_read(inode)) {
1555         if (mode & FALLOC_FL_KEEP_SIZE)
1556             file_set_keep_isize(inode);
1557         else
1558             f2fs_i_size_write(inode, new_size);
1559     }
1560     return ret;
1561 }
1562
1563 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1564 {
1565     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1566     struct address_space *mapping = inode->i_mapping;
1567     pgoff_t nr, pg_start, pg_end, delta, idx;
1568     loff_t new_size;
1569     int ret = 0;
1570
1571     new_size = i_size_read(inode) + len;
1572     ret = inode_newsize_ok(inode, new_size);
1573     if (ret)
1574         return ret;
1575
1576     if (offset >= i_size_read(inode))
1577         return -EINVAL;
1578
1579     /* insert range should be aligned to block size of f2fs. */
1580     if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1581         return -EINVAL;
1582
1583     ret = f2fs_convert_inline_inode(inode);
1584     if (ret)
1585         return ret;
1586
1587     f2fs_balance_fs(sbi, true);
1588
1589     filemap_invalidate_lock(mapping);
1590     ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1591     filemap_invalidate_unlock(mapping);
1592     if (ret)
1593         return ret;
1594
1595     /* write out all dirty pages from offset */
1596     ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1597     if (ret)
1598         return ret;
1599
1600     pg_start = offset >> PAGE_SHIFT;
1601     pg_end = (offset + len) >> PAGE_SHIFT;
1602     delta = pg_end - pg_start;
1603     idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1604
1605     /* avoid gc operation during block exchange */
1606     f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1607     filemap_invalidate_lock(mapping);
1608     truncate_pagecache(inode, offset);
1609
1610     while (!ret && idx > pg_start) {
1611         nr = idx - pg_start;
1612         if (nr > delta)
1613             nr = delta;
1614         idx -= nr;
1615
1616         f2fs_lock_op(sbi);
1617         f2fs_drop_extent_tree(inode);
1618
1619         ret = __exchange_data_block(inode, inode, idx,
1620                     idx + delta, nr, false);
1621         f2fs_unlock_op(sbi);
1622     }
1623     filemap_invalidate_unlock(mapping);
1624     f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1625
1626     /* write out all moved pages, if possible */
1627     filemap_invalidate_lock(mapping);
1628     filemap_write_and_wait_range(mapping, offset, LLONG_MAX);
1629     truncate_pagecache(inode, offset);
1630     filemap_invalidate_unlock(mapping);
1631
1632     if (!ret)
1633         f2fs_i_size_write(inode, new_size);
1634     return ret;
1635 }
1636
1637 static int expand_inode_data(struct inode *inode, loff_t offset,
1638                     loff_t len, int mode)
1639 {
1640     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1641     struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1642             .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1643             .m_may_create = true };
1644     struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
1645             .init_gc_type = FG_GC,
1646             .should_migrate_blocks = false,
1647             .err_gc_skipped = true,
1648             .nr_free_secs = 0 };
1649     pgoff_t pg_start, pg_end;
1650     loff_t new_size = i_size_read(inode);
1651     loff_t off_end;
1652     block_t expanded = 0;
1653     int err;
1654
1655     err = inode_newsize_ok(inode, (len + offset));
1656     if (err)
1657         return err;
1658
1659     err = f2fs_convert_inline_inode(inode);
1660     if (err)
1661         return err;
1662
1663     f2fs_balance_fs(sbi, true);
1664
1665     pg_start = ((unsigned long long)offset) >> PAGE_SHIFT;
1666     pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1667     off_end = (offset + len) & (PAGE_SIZE - 1);
1668
1669     map.m_lblk = pg_start;
1670     map.m_len = pg_end - pg_start;
1671     if (off_end)
1672         map.m_len++;
1673
1674     if (!map.m_len)
1675         return 0;
1676
1677     if (f2fs_is_pinned_file(inode)) {
1678         block_t sec_blks = CAP_BLKS_PER_SEC(sbi);
1679         block_t sec_len = roundup(map.m_len, sec_blks);
1680
1681         map.m_len = sec_blks;
1682 next_alloc:
1683         if (has_not_enough_free_secs(sbi, 0,
1684             GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1685             f2fs_down_write(&sbi->gc_lock);
1686             err = f2fs_gc(sbi, &gc_control);
1687             if (err && err != -ENODATA)
1688                 goto out_err;
1689         }
1690
1691         f2fs_down_write(&sbi->pin_sem);
1692
1693         f2fs_lock_op(sbi);
1694         f2fs_allocate_new_section(sbi, CURSEG_COLD_DATA_PINNED, false);
1695         f2fs_unlock_op(sbi);
1696
1697         map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1698         err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1699         file_dont_truncate(inode);
1700
1701         f2fs_up_write(&sbi->pin_sem);
1702
1703         expanded += map.m_len;
1704         sec_len -= map.m_len;
1705         map.m_lblk += map.m_len;
1706         if (!err && sec_len)
1707             goto next_alloc;
1708
1709         map.m_len = expanded;
1710     } else {
1711         err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1712         expanded = map.m_len;
1713     }
1714 out_err:
1715     if (err) {
1716         pgoff_t last_off;
1717
1718         if (!expanded)
1719             return err;
1720
1721         last_off = pg_start + expanded - 1;
1722
1723         /* update new size to the failed position */
1724         new_size = (last_off == pg_end) ? offset + len :
1725                     (loff_t)(last_off + 1) << PAGE_SHIFT;
1726     } else {
1727         new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1728     }
1729
1730     if (new_size > i_size_read(inode)) {
1731         if (mode & FALLOC_FL_KEEP_SIZE)
1732             file_set_keep_isize(inode);
1733         else
1734             f2fs_i_size_write(inode, new_size);
1735     }
1736
1737     return err;
1738 }
1739
1740 static long f2fs_fallocate(struct file *file, int mode,
1741                 loff_t offset, loff_t len)
1742 {
1743     struct inode *inode = file_inode(file);
1744     long ret = 0;
1745
1746     if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1747         return -EIO;
1748     if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1749         return -ENOSPC;
1750     if (!f2fs_is_compress_backend_ready(inode))
1751         return -EOPNOTSUPP;
1752
1753     /* f2fs only support ->fallocate for regular file */
1754     if (!S_ISREG(inode->i_mode))
1755         return -EINVAL;
1756
1757     if (IS_ENCRYPTED(inode) &&
1758         (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1759         return -EOPNOTSUPP;
1760
1761     /*
1762      * Pinned file should not support partial trucation since the block
1763      * can be used by applications.
1764      */
1765     if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) &&
1766         (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1767             FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1768         return -EOPNOTSUPP;
1769
1770     if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1771             FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1772             FALLOC_FL_INSERT_RANGE))
1773         return -EOPNOTSUPP;
1774
1775     inode_lock(inode);
1776
1777     ret = file_modified(file);
1778     if (ret)
1779         goto out;
1780
1781     if (mode & FALLOC_FL_PUNCH_HOLE) {
1782         if (offset >= inode->i_size)
1783             goto out;
1784
1785         ret = punch_hole(inode, offset, len);
1786     } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1787         ret = f2fs_collapse_range(inode, offset, len);
1788     } else if (mode & FALLOC_FL_ZERO_RANGE) {
1789         ret = f2fs_zero_range(inode, offset, len, mode);
1790     } else if (mode & FALLOC_FL_INSERT_RANGE) {
1791         ret = f2fs_insert_range(inode, offset, len);
1792     } else {
1793         ret = expand_inode_data(inode, offset, len, mode);
1794     }
1795
1796     if (!ret) {
1797         inode->i_mtime = inode->i_ctime = current_time(inode);
1798         f2fs_mark_inode_dirty_sync(inode, false);
1799         f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1800     }
1801
1802 out:
1803     inode_unlock(inode);
1804
1805     trace_f2fs_fallocate(inode, mode, offset, len, ret);
1806     return ret;
1807 }
1808
1809 static int f2fs_release_file(struct inode *inode, struct file *filp)
1810 {
1811     /*
1812      * f2fs_relase_file is called at every close calls. So we should
1813      * not drop any inmemory pages by close called by other process.
1814      */
1815     if (!(filp->f_mode & FMODE_WRITE) ||
1816             atomic_read(&inode->i_writecount) != 1)
1817         return 0;
1818
1819     f2fs_abort_atomic_write(inode, true);
1820     return 0;
1821 }
1822
1823 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1824 {
1825     struct inode *inode = file_inode(file);
1826
1827     /*
1828      * If the process doing a transaction is crashed, we should do
1829      * roll-back. Otherwise, other reader/write can see corrupted database
1830      * until all the writers close its file. Since this should be done
1831      * before dropping file lock, it needs to do in ->flush.
1832      */
1833     if (F2FS_I(inode)->atomic_write_task == current)
1834         f2fs_abort_atomic_write(inode, true);
1835     return 0;
1836 }
1837
1838 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1839 {
1840     struct f2fs_inode_info *fi = F2FS_I(inode);
1841     u32 masked_flags = fi->i_flags & mask;
1842
1843     /* mask can be shrunk by flags_valid selector */
1844     iflags &= mask;
1845
1846     /* Is it quota file? Do not allow user to mess with it */
1847     if (IS_NOQUOTA(inode))
1848         return -EPERM;
1849
1850     if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) {
1851         if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1852             return -EOPNOTSUPP;
1853         if (!f2fs_empty_dir(inode))
1854             return -ENOTEMPTY;
1855     }
1856
1857     if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1858         if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1859             return -EOPNOTSUPP;
1860         if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1861             return -EINVAL;
1862     }
1863
1864     if ((iflags ^ masked_flags) & F2FS_COMPR_FL) {
1865         if (masked_flags & F2FS_COMPR_FL) {
1866             if (!f2fs_disable_compressed_file(inode))
1867                 return -EINVAL;
1868         } else {
1869             if (!f2fs_may_compress(inode))
1870                 return -EINVAL;
1871             if (S_ISREG(inode->i_mode) && F2FS_HAS_BLOCKS(inode))
1872                 return -EINVAL;
1873             if (set_compress_context(inode))
1874                 return -EOPNOTSUPP;
1875         }
1876     }
1877
1878     fi->i_flags = iflags | (fi->i_flags & ~mask);
1879     f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1880                     (fi->i_flags & F2FS_NOCOMP_FL));
1881
1882     if (fi->i_flags & F2FS_PROJINHERIT_FL)
1883         set_inode_flag(inode, FI_PROJ_INHERIT);
1884     else
1885         clear_inode_flag(inode, FI_PROJ_INHERIT);
1886
1887     inode->i_ctime = current_time(inode);
1888     f2fs_set_inode_flags(inode);
1889     f2fs_mark_inode_dirty_sync(inode, true);
1890     return 0;
1891 }
1892
1893 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */
1894
1895 /*
1896  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1897  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1898  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1899  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1900  *
1901  * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and
1902  * FS_IOC_FSSETXATTR is done by the VFS.
1903  */
1904
1905 static const struct {
1906     u32 iflag;
1907     u32 fsflag;
1908 } f2fs_fsflags_map[] = {
1909     { F2FS_COMPR_FL,    FS_COMPR_FL },
1910     { F2FS_SYNC_FL,     FS_SYNC_FL },
1911     { F2FS_IMMUTABLE_FL,    FS_IMMUTABLE_FL },
1912     { F2FS_APPEND_FL,   FS_APPEND_FL },
1913     { F2FS_NODUMP_FL,   FS_NODUMP_FL },
1914     { F2FS_NOATIME_FL,  FS_NOATIME_FL },
1915     { F2FS_NOCOMP_FL,   FS_NOCOMP_FL },
1916     { F2FS_INDEX_FL,    FS_INDEX_FL },
1917     { F2FS_DIRSYNC_FL,  FS_DIRSYNC_FL },
1918     { F2FS_PROJINHERIT_FL,  FS_PROJINHERIT_FL },
1919     { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL },
1920 };
1921
1922 #define F2FS_GETTABLE_FS_FL (       \
1923         FS_COMPR_FL |       \
1924         FS_SYNC_FL |        \
1925         FS_IMMUTABLE_FL |   \
1926         FS_APPEND_FL |      \
1927         FS_NODUMP_FL |      \
1928         FS_NOATIME_FL |     \
1929         FS_NOCOMP_FL |      \
1930         FS_INDEX_FL |       \
1931         FS_DIRSYNC_FL |     \
1932         FS_PROJINHERIT_FL | \
1933         FS_ENCRYPT_FL |     \
1934         FS_INLINE_DATA_FL | \
1935         FS_NOCOW_FL |       \
1936         FS_VERITY_FL |      \
1937         FS_CASEFOLD_FL)
1938
1939 #define F2FS_SETTABLE_FS_FL (       \
1940         FS_COMPR_FL |       \
1941         FS_SYNC_FL |        \
1942         FS_IMMUTABLE_FL |   \
1943         FS_APPEND_FL |      \
1944         FS_NODUMP_FL |      \
1945         FS_NOATIME_FL |     \
1946         FS_NOCOMP_FL |      \
1947         FS_DIRSYNC_FL |     \
1948         FS_PROJINHERIT_FL | \
1949         FS_CASEFOLD_FL)
1950
1951 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1952 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1953 {
1954     u32 fsflags = 0;
1955     int i;
1956
1957     for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1958         if (iflags & f2fs_fsflags_map[i].iflag)
1959             fsflags |= f2fs_fsflags_map[i].fsflag;
1960
1961     return fsflags;
1962 }
1963
1964 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1965 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1966 {
1967     u32 iflags = 0;
1968     int i;
1969
1970     for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1971         if (fsflags & f2fs_fsflags_map[i].fsflag)
1972             iflags |= f2fs_fsflags_map[i].iflag;
1973
1974     return iflags;
1975 }
1976
1977 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1978 {
1979     struct inode *inode = file_inode(filp);
1980
1981     return put_user(inode->i_generation, (int __user *)arg);
1982 }
1983
1984 static int f2fs_ioc_start_atomic_write(struct file *filp)
1985 {
1986     struct inode *inode = file_inode(filp);
1987     struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
1988     struct f2fs_inode_info *fi = F2FS_I(inode);
1989     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1990     struct inode *pinode;
1991     int ret;
1992
1993     if (!inode_owner_or_capable(mnt_userns, inode))
1994         return -EACCES;
1995
1996     if (!S_ISREG(inode->i_mode))
1997         return -EINVAL;
1998
1999     if (filp->f_flags & O_DIRECT)
2000         return -EINVAL;
2001
2002     ret = mnt_want_write_file(filp);
2003     if (ret)
2004         return ret;
2005
2006     inode_lock(inode);
2007
2008     if (!f2fs_disable_compressed_file(inode)) {
2009         ret = -EINVAL;
2010         goto out;
2011     }
2012
2013     if (f2fs_is_atomic_file(inode))
2014         goto out;
2015
2016     ret = f2fs_convert_inline_inode(inode);
2017     if (ret)
2018         goto out;
2019
2020     f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
2021
2022     /*
2023      * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2024      * f2fs_is_atomic_file.
2025      */
2026     if (get_dirty_pages(inode))
2027         f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2028               inode->i_ino, get_dirty_pages(inode));
2029     ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2030     if (ret) {
2031         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2032         goto out;
2033     }
2034
2035     /* Create a COW inode for atomic write */
2036     pinode = f2fs_iget(inode->i_sb, fi->i_pino);
2037     if (IS_ERR(pinode)) {
2038         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2039         ret = PTR_ERR(pinode);
2040         goto out;
2041     }
2042
2043     ret = f2fs_get_tmpfile(mnt_userns, pinode, &fi->cow_inode);
2044     iput(pinode);
2045     if (ret) {
2046         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2047         goto out;
2048     }
2049     f2fs_i_size_write(fi->cow_inode, i_size_read(inode));
2050
2051     spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2052     sbi->atomic_files++;
2053     spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2054
2055     set_inode_flag(inode, FI_ATOMIC_FILE);
2056     set_inode_flag(fi->cow_inode, FI_COW_FILE);
2057     clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
2058     f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
2059
2060     f2fs_update_time(sbi, REQ_TIME);
2061     fi->atomic_write_task = current;
2062     stat_update_max_atomic_write(inode);
2063     fi->atomic_write_cnt = 0;
2064 out:
2065     inode_unlock(inode);
2066     mnt_drop_write_file(filp);
2067     return ret;
2068 }
2069
2070 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2071 {
2072     struct inode *inode = file_inode(filp);
2073     struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2074     int ret;
2075
2076     if (!inode_owner_or_capable(mnt_userns, inode))
2077         return -EACCES;
2078
2079     ret = mnt_want_write_file(filp);
2080     if (ret)
2081         return ret;
2082
2083     f2fs_balance_fs(F2FS_I_SB(inode), true);
2084
2085     inode_lock(inode);
2086
2087     if (f2fs_is_atomic_file(inode)) {
2088         ret = f2fs_commit_atomic_write(inode);
2089         if (ret)
2090             goto unlock_out;
2091
2092         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2093         if (!ret)
2094             f2fs_abort_atomic_write(inode, false);
2095     } else {
2096         ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2097     }
2098 unlock_out:
2099     inode_unlock(inode);
2100     mnt_drop_write_file(filp);
2101     return ret;
2102 }
2103
2104 static int f2fs_ioc_abort_atomic_write(struct file *filp)
2105 {
2106     struct inode *inode = file_inode(filp);
2107     struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
2108     int ret;
2109
2110     if (!inode_owner_or_capable(mnt_userns, inode))
2111         return -EACCES;
2112
2113     ret = mnt_want_write_file(filp);
2114     if (ret)
2115         return ret;
2116
2117     inode_lock(inode);
2118
2119     f2fs_abort_atomic_write(inode, true);
2120
2121     inode_unlock(inode);
2122
2123     mnt_drop_write_file(filp);
2124     f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2125     return ret;
2126 }
2127
2128 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2129 {
2130     struct inode *inode = file_inode(filp);
2131     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2132     struct super_block *sb = sbi->sb;
2133     __u32 in;
2134     int ret = 0;
2135
2136     if (!capable(CAP_SYS_ADMIN))
2137         return -EPERM;
2138
2139     if (get_user(in, (__u32 __user *)arg))
2140         return -EFAULT;
2141
2142     if (in != F2FS_GOING_DOWN_FULLSYNC) {
2143         ret = mnt_want_write_file(filp);
2144         if (ret) {
2145             if (ret == -EROFS) {
2146                 ret = 0;
2147                 f2fs_stop_checkpoint(sbi, false);
2148                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2149                 trace_f2fs_shutdown(sbi, in, ret);
2150             }
2151             return ret;
2152         }
2153     }
2154
2155     switch (in) {
2156     case F2FS_GOING_DOWN_FULLSYNC:
2157         ret = freeze_bdev(sb->s_bdev);
2158         if (ret)
2159             goto out;
2160         f2fs_stop_checkpoint(sbi, false);
2161         set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2162         thaw_bdev(sb->s_bdev);
2163         break;
2164     case F2FS_GOING_DOWN_METASYNC:
2165         /* do checkpoint only */
2166         ret = f2fs_sync_fs(sb, 1);
2167         if (ret)
2168             goto out;
2169         f2fs_stop_checkpoint(sbi, false);
2170         set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2171         break;
2172     case F2FS_GOING_DOWN_NOSYNC:
2173         f2fs_stop_checkpoint(sbi, false);
2174         set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2175         break;
2176     case F2FS_GOING_DOWN_METAFLUSH:
2177         f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2178         f2fs_stop_checkpoint(sbi, false);
2179         set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2180         break;
2181     case F2FS_GOING_DOWN_NEED_FSCK:
2182         set_sbi_flag(sbi, SBI_NEED_FSCK);
2183         set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2184         set_sbi_flag(sbi, SBI_IS_DIRTY);
2185         /* do checkpoint only */
2186         ret = f2fs_sync_fs(sb, 1);
2187         goto out;
2188     default:
2189         ret = -EINVAL;
2190         goto out;
2191     }
2192
2193     f2fs_stop_gc_thread(sbi);
2194     f2fs_stop_discard_thread(sbi);
2195
2196     f2fs_drop_discard_cmd(sbi);
2197     clear_opt(sbi, DISCARD);
2198
2199     f2fs_update_time(sbi, REQ_TIME);
2200 out:
2201     if (in != F2FS_GOING_DOWN_FULLSYNC)
2202         mnt_drop_write_file(filp);
2203
2204     trace_f2fs_shutdown(sbi, in, ret);
2205
2206     return ret;
2207 }
2208
2209 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2210 {
2211     struct inode *inode = file_inode(filp);
2212     struct super_block *sb = inode->i_sb;
2213     struct fstrim_range range;
2214     int ret;
2215
2216     if (!capable(CAP_SYS_ADMIN))
2217         return -EPERM;
2218
2219     if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2220         return -EOPNOTSUPP;
2221
2222     if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2223                 sizeof(range)))
2224         return -EFAULT;
2225
2226     ret = mnt_want_write_file(filp);
2227     if (ret)
2228         return ret;
2229
2230     range.minlen = max((unsigned int)range.minlen,
2231                bdev_discard_granularity(sb->s_bdev));
2232     ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2233     mnt_drop_write_file(filp);
2234     if (ret < 0)
2235         return ret;
2236
2237     if (copy_to_user((struct fstrim_range __user *)arg, &range,
2238                 sizeof(range)))
2239         return -EFAULT;
2240     f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2241     return 0;
2242 }
2243
2244 static bool uuid_is_nonzero(__u8 u[16])
2245 {
2246     int i;
2247
2248     for (i = 0; i < 16; i++)
2249         if (u[i])
2250             return true;
2251     return false;
2252 }
2253
2254 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2255 {
2256     struct inode *inode = file_inode(filp);
2257
2258     if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2259         return -EOPNOTSUPP;
2260
2261     f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2262
2263     return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2264 }
2265
2266 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2267 {
2268     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2269         return -EOPNOTSUPP;
2270     return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2271 }
2272
2273 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2274 {
2275     struct inode *inode = file_inode(filp);
2276     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2277     int err;
2278
2279     if (!f2fs_sb_has_encrypt(sbi))
2280         return -EOPNOTSUPP;
2281
2282     err = mnt_want_write_file(filp);
2283     if (err)
2284         return err;
2285
2286     f2fs_down_write(&sbi->sb_lock);
2287
2288     if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2289         goto got_it;
2290
2291     /* update superblock with uuid */
2292     generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2293
2294     err = f2fs_commit_super(sbi, false);
2295     if (err) {
2296         /* undo new data */
2297         memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2298         goto out_err;
2299     }
2300 got_it:
2301     if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2302                                     16))
2303         err = -EFAULT;
2304 out_err:
2305     f2fs_up_write(&sbi->sb_lock);
2306     mnt_drop_write_file(filp);
2307     return err;
2308 }
2309
2310 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2311                          unsigned long arg)
2312 {
2313     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2314         return -EOPNOTSUPP;
2315
2316     return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2317 }
2318
2319 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2320 {
2321     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2322         return -EOPNOTSUPP;
2323
2324     return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2325 }
2326
2327 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2328 {
2329     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2330         return -EOPNOTSUPP;
2331
2332     return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2333 }
2334
2335 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2336                             unsigned long arg)
2337 {
2338     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2339         return -EOPNOTSUPP;
2340
2341     return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2342 }
2343
2344 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2345                           unsigned long arg)
2346 {
2347     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2348         return -EOPNOTSUPP;
2349
2350     return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2351 }
2352
2353 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg)
2354 {
2355     if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2356         return -EOPNOTSUPP;
2357
2358     return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
2359 }
2360
2361 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2362 {
2363     struct inode *inode = file_inode(filp);
2364     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2365     struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO,
2366             .no_bg_gc = false,
2367             .should_migrate_blocks = false,
2368             .nr_free_secs = 0 };
2369     __u32 sync;
2370     int ret;
2371
2372     if (!capable(CAP_SYS_ADMIN))
2373         return -EPERM;
2374
2375     if (get_user(sync, (__u32 __user *)arg))
2376         return -EFAULT;
2377
2378     if (f2fs_readonly(sbi->sb))
2379         return -EROFS;
2380
2381     ret = mnt_want_write_file(filp);
2382     if (ret)
2383         return ret;
2384
2385     if (!sync) {
2386         if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2387             ret = -EBUSY;
2388             goto out;
2389         }
2390     } else {
2391         f2fs_down_write(&sbi->gc_lock);
2392     }
2393
2394     gc_control.init_gc_type = sync ? FG_GC : BG_GC;
2395     gc_control.err_gc_skipped = sync;
2396     ret = f2fs_gc(sbi, &gc_control);
2397 out:
2398     mnt_drop_write_file(filp);
2399     return ret;
2400 }
2401
2402 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range)
2403 {
2404     struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
2405     struct f2fs_gc_control gc_control = {
2406             .init_gc_type = range->sync ? FG_GC : BG_GC,
2407             .no_bg_gc = false,
2408             .should_migrate_blocks = false,
2409             .err_gc_skipped = range->sync,
2410             .nr_free_secs = 0 };
2411     u64 end;
2412     int ret;
2413
2414     if (!capable(CAP_SYS_ADMIN))
2415         return -EPERM;
2416     if (f2fs_readonly(sbi->sb))
2417         return -EROFS;
2418
2419     end = range->start + range->len;
2420     if (end < range->start || range->start < MAIN_BLKADDR(sbi) ||
2421                     end >= MAX_BLKADDR(sbi))
2422         return -EINVAL;
2423
2424     ret = mnt_want_write_file(filp);
2425     if (ret)
2426         return ret;
2427
2428 do_more:
2429     if (!range->sync) {
2430         if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2431             ret = -EBUSY;
2432             goto out;
2433         }
2434     } else {
2435         f2fs_down_write(&sbi->gc_lock);
2436     }
2437
2438     gc_control.victim_segno = GET_SEGNO(sbi, range->start);
2439     ret = f2fs_gc(sbi, &gc_control);
2440     if (ret) {
2441         if (ret == -EBUSY)
2442             ret = -EAGAIN;
2443         goto out;
2444     }
2445     range->start += CAP_BLKS_PER_SEC(sbi);
2446     if (range->start <= end)
2447         goto do_more;
2448 out:
2449     mnt_drop_write_file(filp);
2450     return ret;
2451 }
2452
2453 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2454 {
2455     struct f2fs_gc_range range;
2456
2457     if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2458                             sizeof(range)))
2459         return -EFAULT;
2460     return __f2fs_ioc_gc_range(filp, &range);
2461 }
2462
2463 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2464 {
2465     struct inode *inode = file_inode(filp);
2466     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2467     int ret;
2468
2469     if (!capable(CAP_SYS_ADMIN))
2470         return -EPERM;
2471
2472     if (f2fs_readonly(sbi->sb))
2473         return -EROFS;
2474
2475     if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2476         f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2477         return -EINVAL;
2478     }
2479
2480     ret = mnt_want_write_file(filp);
2481     if (ret)
2482         return ret;
2483
2484     ret = f2fs_sync_fs(sbi->sb, 1);
2485
2486     mnt_drop_write_file(filp);
2487     return ret;
2488 }
2489
2490 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2491                     struct file *filp,
2492                     struct f2fs_defragment *range)
2493 {
2494     struct inode *inode = file_inode(filp);
2495     struct f2fs_map_blocks map = { .m_next_extent = NULL,
2496                     .m_seg_type = NO_CHECK_TYPE,
2497                     .m_may_create = false };
2498     struct extent_info ei = {0, 0, 0};
2499     pgoff_t pg_start, pg_end, next_pgofs;
2500     unsigned int blk_per_seg = sbi->blocks_per_seg;
2501     unsigned int total = 0, sec_num;
2502     block_t blk_end = 0;
2503     bool fragmented = false;
2504     int err;
2505
2506     pg_start = range->start >> PAGE_SHIFT;
2507     pg_end = (range->start + range->len) >> PAGE_SHIFT;
2508
2509     f2fs_balance_fs(sbi, true);
2510
2511     inode_lock(inode);
2512
2513     /* if in-place-update policy is enabled, don't waste time here */
2514     set_inode_flag(inode, FI_OPU_WRITE);
2515     if (f2fs_should_update_inplace(inode, NULL)) {
2516         err = -EINVAL;
2517         goto out;
2518     }
2519
2520     /* writeback all dirty pages in the range */
2521     err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2522                         range->start + range->len - 1);
2523     if (err)
2524         goto out;
2525
2526     /*
2527      * lookup mapping info in extent cache, skip defragmenting if physical
2528      * block addresses are continuous.
2529      */
2530     if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2531         if (ei.fofs + ei.len >= pg_end)
2532             goto out;
2533     }
2534
2535     map.m_lblk = pg_start;
2536     map.m_next_pgofs = &next_pgofs;
2537
2538     /*
2539      * lookup mapping info in dnode page cache, skip defragmenting if all
2540      * physical block addresses are continuous even if there are hole(s)
2541      * in logical blocks.
2542      */
2543     while (map.m_lblk < pg_end) {
2544         map.m_len = pg_end - map.m_lblk;
2545         err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2546         if (err)
2547             goto out;
2548
2549         if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2550             map.m_lblk = next_pgofs;
2551             continue;
2552         }
2553
2554         if (blk_end && blk_end != map.m_pblk)
2555             fragmented = true;
2556
2557         /* record total count of block that we're going to move */
2558         total += map.m_len;
2559
2560         blk_end = map.m_pblk + map.m_len;
2561
2562         map.m_lblk += map.m_len;
2563     }
2564
2565     if (!fragmented) {
2566         total = 0;
2567         goto out;
2568     }
2569
2570     sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi));
2571
2572     /*
2573      * make sure there are enough free section for LFS allocation, this can
2574      * avoid defragment running in SSR mode when free section are allocated
2575      * intensively
2576      */
2577     if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2578         err = -EAGAIN;
2579         goto out;
2580     }
2581
2582     map.m_lblk = pg_start;
2583     map.m_len = pg_end - pg_start;
2584     total = 0;
2585
2586     while (map.m_lblk < pg_end) {
2587         pgoff_t idx;
2588         int cnt = 0;
2589
2590 do_map:
2591         map.m_len = pg_end - map.m_lblk;
2592         err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2593         if (err)
2594             goto clear_out;
2595
2596         if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2597             map.m_lblk = next_pgofs;
2598             goto check;
2599         }
2600
2601         set_inode_flag(inode, FI_SKIP_WRITES);
2602
2603         idx = map.m_lblk;
2604         while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2605             struct page *page;
2606
2607             page = f2fs_get_lock_data_page(inode, idx, true);
2608             if (IS_ERR(page)) {
2609                 err = PTR_ERR(page);
2610                 goto clear_out;
2611             }
2612
2613             set_page_dirty(page);
2614             set_page_private_gcing(page);
2615             f2fs_put_page(page, 1);
2616
2617             idx++;
2618             cnt++;
2619             total++;
2620         }
2621
2622         map.m_lblk = idx;
2623 check:
2624         if (map.m_lblk < pg_end && cnt < blk_per_seg)
2625             goto do_map;
2626
2627         clear_inode_flag(inode, FI_SKIP_WRITES);
2628
2629         err = filemap_fdatawrite(inode->i_mapping);
2630         if (err)
2631             goto out;
2632     }
2633 clear_out:
2634     clear_inode_flag(inode, FI_SKIP_WRITES);
2635 out:
2636     clear_inode_flag(inode, FI_OPU_WRITE);
2637     inode_unlock(inode);
2638     if (!err)
2639         range->len = (u64)total << PAGE_SHIFT;
2640     return err;
2641 }
2642
2643 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2644 {
2645     struct inode *inode = file_inode(filp);
2646     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2647     struct f2fs_defragment range;
2648     int err;
2649
2650     if (!capable(CAP_SYS_ADMIN))
2651         return -EPERM;
2652
2653     if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2654         return -EINVAL;
2655
2656     if (f2fs_readonly(sbi->sb))
2657         return -EROFS;
2658
2659     if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2660                             sizeof(range)))
2661         return -EFAULT;
2662
2663     /* verify alignment of offset & size */
2664     if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2665         return -EINVAL;
2666
2667     if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2668                     max_file_blocks(inode)))
2669         return -EINVAL;
2670
2671     err = mnt_want_write_file(filp);
2672     if (err)
2673         return err;
2674
2675     err = f2fs_defragment_range(sbi, filp, &range);
2676     mnt_drop_write_file(filp);
2677
2678     f2fs_update_time(sbi, REQ_TIME);
2679     if (err < 0)
2680         return err;
2681
2682     if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2683                             sizeof(range)))
2684         return -EFAULT;
2685
2686     return 0;
2687 }
2688
2689 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2690             struct file *file_out, loff_t pos_out, size_t len)
2691 {
2692     struct inode *src = file_inode(file_in);
2693     struct inode *dst = file_inode(file_out);
2694     struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2695     size_t olen = len, dst_max_i_size = 0;
2696     size_t dst_osize;
2697     int ret;
2698
2699     if (file_in->f_path.mnt != file_out->f_path.mnt ||
2700                 src->i_sb != dst->i_sb)
2701         return -EXDEV;
2702
2703     if (unlikely(f2fs_readonly(src->i_sb)))
2704         return -EROFS;
2705
2706     if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2707         return -EINVAL;
2708
2709     if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2710         return -EOPNOTSUPP;
2711
2712     if (pos_out < 0 || pos_in < 0)
2713         return -EINVAL;
2714
2715     if (src == dst) {
2716         if (pos_in == pos_out)
2717             return 0;
2718         if (pos_out > pos_in && pos_out < pos_in + len)
2719             return -EINVAL;
2720     }
2721
2722     inode_lock(src);
2723     if (src != dst) {
2724         ret = -EBUSY;
2725         if (!inode_trylock(dst))
2726             goto out;
2727     }
2728
2729     ret = -EINVAL;
2730     if (pos_in + len > src->i_size || pos_in + len < pos_in)
2731         goto out_unlock;
2732     if (len == 0)
2733         olen = len = src->i_size - pos_in;
2734     if (pos_in + len == src->i_size)
2735         len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2736     if (len == 0) {
2737         ret = 0;
2738         goto out_unlock;
2739     }
2740
2741     dst_osize = dst->i_size;
2742     if (pos_out + olen > dst->i_size)
2743         dst_max_i_size = pos_out + olen;
2744
2745     /* verify the end result is block aligned */
2746     if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2747             !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2748             !IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2749         goto out_unlock;
2750
2751     ret = f2fs_convert_inline_inode(src);
2752     if (ret)
2753         goto out_unlock;
2754
2755     ret = f2fs_convert_inline_inode(dst);
2756     if (ret)
2757         goto out_unlock;
2758
2759     /* write out all dirty pages from offset */
2760     ret = filemap_write_and_wait_range(src->i_mapping,
2761                     pos_in, pos_in + len);
2762     if (ret)
2763         goto out_unlock;
2764
2765     ret = filemap_write_and_wait_range(dst->i_mapping,
2766                     pos_out, pos_out + len);
2767     if (ret)
2768         goto out_unlock;
2769
2770     f2fs_balance_fs(sbi, true);
2771
2772     f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2773     if (src != dst) {
2774         ret = -EBUSY;
2775         if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2776             goto out_src;
2777     }
2778
2779     f2fs_lock_op(sbi);
2780     ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2781                 pos_out >> F2FS_BLKSIZE_BITS,
2782                 len >> F2FS_BLKSIZE_BITS, false);
2783
2784     if (!ret) {
2785         if (dst_max_i_size)
2786             f2fs_i_size_write(dst, dst_max_i_size);
2787         else if (dst_osize != dst->i_size)
2788             f2fs_i_size_write(dst, dst_osize);
2789     }
2790     f2fs_unlock_op(sbi);
2791
2792     if (src != dst)
2793         f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2794 out_src:
2795     f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2796 out_unlock:
2797     if (src != dst)
2798         inode_unlock(dst);
2799 out:
2800     inode_unlock(src);
2801     return ret;
2802 }
2803
2804 static int __f2fs_ioc_move_range(struct file *filp,
2805                 struct f2fs_move_range *range)
2806 {
2807     struct fd dst;
2808     int err;
2809
2810     if (!(filp->f_mode & FMODE_READ) ||
2811             !(filp->f_mode & FMODE_WRITE))
2812         return -EBADF;
2813
2814     dst = fdget(range->dst_fd);
2815     if (!dst.file)
2816         return -EBADF;
2817
2818     if (!(dst.file->f_mode & FMODE_WRITE)) {
2819         err = -EBADF;
2820         goto err_out;
2821     }
2822
2823     err = mnt_want_write_file(filp);
2824     if (err)
2825         goto err_out;
2826
2827     err = f2fs_move_file_range(filp, range->pos_in, dst.file,
2828                     range->pos_out, range->len);
2829
2830     mnt_drop_write_file(filp);
2831 err_out:
2832     fdput(dst);
2833     return err;
2834 }
2835
2836 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2837 {
2838     struct f2fs_move_range range;
2839
2840     if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2841                             sizeof(range)))
2842         return -EFAULT;
2843     return __f2fs_ioc_move_range(filp, &range);
2844 }
2845
2846 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2847 {
2848     struct inode *inode = file_inode(filp);
2849     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2850     struct sit_info *sm = SIT_I(sbi);
2851     unsigned int start_segno = 0, end_segno = 0;
2852     unsigned int dev_start_segno = 0, dev_end_segno = 0;
2853     struct f2fs_flush_device range;
2854     struct f2fs_gc_control gc_control = {
2855             .init_gc_type = FG_GC,
2856             .should_migrate_blocks = true,
2857             .err_gc_skipped = true,
2858             .nr_free_secs = 0 };
2859     int ret;
2860
2861     if (!capable(CAP_SYS_ADMIN))
2862         return -EPERM;
2863
2864     if (f2fs_readonly(sbi->sb))
2865         return -EROFS;
2866
2867     if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2868         return -EINVAL;
2869
2870     if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2871                             sizeof(range)))
2872         return -EFAULT;
2873
2874     if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2875             __is_large_section(sbi)) {
2876         f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2877               range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2878         return -EINVAL;
2879     }
2880
2881     ret = mnt_want_write_file(filp);
2882     if (ret)
2883         return ret;
2884
2885     if (range.dev_num != 0)
2886         dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2887     dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2888
2889     start_segno = sm->last_victim[FLUSH_DEVICE];
2890     if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2891         start_segno = dev_start_segno;
2892     end_segno = min(start_segno + range.segments, dev_end_segno);
2893
2894     while (start_segno < end_segno) {
2895         if (!f2fs_down_write_trylock(&sbi->gc_lock)) {
2896             ret = -EBUSY;
2897             goto out;
2898         }
2899         sm->last_victim[GC_CB] = end_segno + 1;
2900         sm->last_victim[GC_GREEDY] = end_segno + 1;
2901         sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2902
2903         gc_control.victim_segno = start_segno;
2904         ret = f2fs_gc(sbi, &gc_control);
2905         if (ret == -EAGAIN)
2906             ret = 0;
2907         else if (ret < 0)
2908             break;
2909         start_segno++;
2910     }
2911 out:
2912     mnt_drop_write_file(filp);
2913     return ret;
2914 }
2915
2916 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2917 {
2918     struct inode *inode = file_inode(filp);
2919     u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2920
2921     /* Must validate to set it with SQLite behavior in Android. */
2922     sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2923
2924     return put_user(sb_feature, (u32 __user *)arg);
2925 }
2926
2927 #ifdef CONFIG_QUOTA
2928 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2929 {
2930     struct dquot *transfer_to[MAXQUOTAS] = {};
2931     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2932     struct super_block *sb = sbi->sb;
2933     int err = 0;
2934
2935     transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2936     if (!IS_ERR(transfer_to[PRJQUOTA])) {
2937         err = __dquot_transfer(inode, transfer_to);
2938         if (err)
2939             set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2940         dqput(transfer_to[PRJQUOTA]);
2941     }
2942     return err;
2943 }
2944
2945 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2946 {
2947     struct f2fs_inode_info *fi = F2FS_I(inode);
2948     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2949     struct f2fs_inode *ri = NULL;
2950     kprojid_t kprojid;
2951     int err;
2952
2953     if (!f2fs_sb_has_project_quota(sbi)) {
2954         if (projid != F2FS_DEF_PROJID)
2955             return -EOPNOTSUPP;
2956         else
2957             return 0;
2958     }
2959
2960     if (!f2fs_has_extra_attr(inode))
2961         return -EOPNOTSUPP;
2962
2963     kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2964
2965     if (projid_eq(kprojid, fi->i_projid))
2966         return 0;
2967
2968     err = -EPERM;
2969     /* Is it quota file? Do not allow user to mess with it */
2970     if (IS_NOQUOTA(inode))
2971         return err;
2972
2973     if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
2974         return -EOVERFLOW;
2975
2976     err = f2fs_dquot_initialize(inode);
2977     if (err)
2978         return err;
2979
2980     f2fs_lock_op(sbi);
2981     err = f2fs_transfer_project_quota(inode, kprojid);
2982     if (err)
2983         goto out_unlock;
2984
2985     fi->i_projid = kprojid;
2986     inode->i_ctime = current_time(inode);
2987     f2fs_mark_inode_dirty_sync(inode, true);
2988 out_unlock:
2989     f2fs_unlock_op(sbi);
2990     return err;
2991 }
2992 #else
2993 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2994 {
2995     return 0;
2996 }
2997
2998 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid)
2999 {
3000     if (projid != F2FS_DEF_PROJID)
3001         return -EOPNOTSUPP;
3002     return 0;
3003 }
3004 #endif
3005
3006 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
3007 {
3008     struct inode *inode = d_inode(dentry);
3009     struct f2fs_inode_info *fi = F2FS_I(inode);
3010     u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
3011
3012     if (IS_ENCRYPTED(inode))
3013         fsflags |= FS_ENCRYPT_FL;
3014     if (IS_VERITY(inode))
3015         fsflags |= FS_VERITY_FL;
3016     if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
3017         fsflags |= FS_INLINE_DATA_FL;
3018     if (is_inode_flag_set(inode, FI_PIN_FILE))
3019         fsflags |= FS_NOCOW_FL;
3020
3021     fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL);
3022
3023     if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3024         fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3025
3026     return 0;
3027 }
3028
3029 int f2fs_fileattr_set(struct user_namespace *mnt_userns,
3030               struct dentry *dentry, struct fileattr *fa)
3031 {
3032     struct inode *inode = d_inode(dentry);
3033     u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL;
3034     u32 iflags;
3035     int err;
3036
3037     if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
3038         return -EIO;
3039     if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
3040         return -ENOSPC;
3041     if (fsflags & ~F2FS_GETTABLE_FS_FL)
3042         return -EOPNOTSUPP;
3043     fsflags &= F2FS_SETTABLE_FS_FL;
3044     if (!fa->flags_valid)
3045         mask &= FS_COMMON_FL;
3046
3047     iflags = f2fs_fsflags_to_iflags(fsflags);
3048     if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3049         return -EOPNOTSUPP;
3050
3051     err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask));
3052     if (!err)
3053         err = f2fs_ioc_setproject(inode, fa->fsx_projid);
3054
3055     return err;
3056 }
3057
3058 int f2fs_pin_file_control(struct inode *inode, bool inc)
3059 {
3060     struct f2fs_inode_info *fi = F2FS_I(inode);
3061     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3062
3063     /* Use i_gc_failures for normal file as a risk signal. */
3064     if (inc)
3065         f2fs_i_gc_failures_write(inode,
3066                 fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3067
3068     if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3069         f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3070               __func__, inode->i_ino,
3071               fi->i_gc_failures[GC_FAILURE_PIN]);
3072         clear_inode_flag(inode, FI_PIN_FILE);
3073         return -EAGAIN;
3074     }
3075     return 0;
3076 }
3077
3078 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3079 {
3080     struct inode *inode = file_inode(filp);
3081     __u32 pin;
3082     int ret = 0;
3083
3084     if (get_user(pin, (__u32 __user *)arg))
3085         return -EFAULT;
3086
3087     if (!S_ISREG(inode->i_mode))
3088         return -EINVAL;
3089
3090     if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3091         return -EROFS;
3092
3093     ret = mnt_want_write_file(filp);
3094     if (ret)
3095         return ret;
3096
3097     inode_lock(inode);
3098
3099     if (!pin) {
3100         clear_inode_flag(inode, FI_PIN_FILE);
3101         f2fs_i_gc_failures_write(inode, 0);
3102         goto done;
3103     }
3104
3105     if (f2fs_should_update_outplace(inode, NULL)) {
3106         ret = -EINVAL;
3107         goto out;
3108     }
3109
3110     if (f2fs_pin_file_control(inode, false)) {
3111         ret = -EAGAIN;
3112         goto out;
3113     }
3114
3115     ret = f2fs_convert_inline_inode(inode);
3116     if (ret)
3117         goto out;
3118
3119     if (!f2fs_disable_compressed_file(inode)) {
3120         ret = -EOPNOTSUPP;
3121         goto out;
3122     }
3123
3124     set_inode_flag(inode, FI_PIN_FILE);
3125     ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3126 done:
3127     f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3128 out:
3129     inode_unlock(inode);
3130     mnt_drop_write_file(filp);
3131     return ret;
3132 }
3133
3134 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3135 {
3136     struct inode *inode = file_inode(filp);
3137     __u32 pin = 0;
3138
3139     if (is_inode_flag_set(inode, FI_PIN_FILE))
3140         pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3141     return put_user(pin, (u32 __user *)arg);
3142 }
3143
3144 int f2fs_precache_extents(struct inode *inode)
3145 {
3146     struct f2fs_inode_info *fi = F2FS_I(inode);
3147     struct f2fs_map_blocks map;
3148     pgoff_t m_next_extent;
3149     loff_t end;
3150     int err;
3151
3152     if (is_inode_flag_set(inode, FI_NO_EXTENT))
3153         return -EOPNOTSUPP;
3154
3155     map.m_lblk = 0;
3156     map.m_next_pgofs = NULL;
3157     map.m_next_extent = &m_next_extent;
3158     map.m_seg_type = NO_CHECK_TYPE;
3159     map.m_may_create = false;
3160     end = max_file_blocks(inode);
3161
3162     while (map.m_lblk < end) {
3163         map.m_len = end - map.m_lblk;
3164
3165         f2fs_down_write(&fi->i_gc_rwsem[WRITE]);
3166         err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3167         f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
3168         if (err)
3169             return err;
3170
3171         map.m_lblk = m_next_extent;
3172     }
3173
3174     return 0;
3175 }
3176
3177 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3178 {
3179     return f2fs_precache_extents(file_inode(filp));
3180 }
3181
3182 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3183 {
3184     struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3185     __u64 block_count;
3186
3187     if (!capable(CAP_SYS_ADMIN))
3188         return -EPERM;
3189
3190     if (f2fs_readonly(sbi->sb))
3191         return -EROFS;
3192
3193     if (copy_from_user(&block_count, (void __user *)arg,
3194                sizeof(block_count)))
3195         return -EFAULT;
3196
3197     return f2fs_resize_fs(sbi, block_count);
3198 }
3199
3200 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3201 {
3202     struct inode *inode = file_inode(filp);
3203
3204     f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3205
3206     if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3207         f2fs_warn(F2FS_I_SB(inode),
3208               "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem",
3209               inode->i_ino);
3210         return -EOPNOTSUPP;
3211     }
3212
3213     return fsverity_ioctl_enable(filp, (const void __user *)arg);
3214 }
3215
3216 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3217 {
3218     if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3219         return -EOPNOTSUPP;
3220
3221     return fsverity_ioctl_measure(filp, (void __user *)arg);
3222 }
3223
3224 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg)
3225 {
3226     if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3227         return -EOPNOTSUPP;
3228
3229     return fsverity_ioctl_read_metadata(filp, (const void __user *)arg);
3230 }
3231
3232 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg)
3233 {
3234     struct inode *inode = file_inode(filp);
3235     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3236     char *vbuf;
3237     int count;
3238     int err = 0;
3239
3240     vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3241     if (!vbuf)
3242         return -ENOMEM;
3243
3244     f2fs_down_read(&sbi->sb_lock);
3245     count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3246             ARRAY_SIZE(sbi->raw_super->volume_name),
3247             UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3248     f2fs_up_read(&sbi->sb_lock);
3249
3250     if (copy_to_user((char __user *)arg, vbuf,
3251                 min(FSLABEL_MAX, count)))
3252         err = -EFAULT;
3253
3254     kfree(vbuf);
3255     return err;
3256 }
3257
3258 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg)
3259 {
3260     struct inode *inode = file_inode(filp);
3261     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3262     char *vbuf;
3263     int err = 0;
3264
3265     if (!capable(CAP_SYS_ADMIN))
3266         return -EPERM;
3267
3268     vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3269     if (IS_ERR(vbuf))
3270         return PTR_ERR(vbuf);
3271
3272     err = mnt_want_write_file(filp);
3273     if (err)
3274         goto out;
3275
3276     f2fs_down_write(&sbi->sb_lock);
3277
3278     memset(sbi->raw_super->volume_name, 0,
3279             sizeof(sbi->raw_super->volume_name));
3280     utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3281             sbi->raw_super->volume_name,
3282             ARRAY_SIZE(sbi->raw_super->volume_name));
3283
3284     err = f2fs_commit_super(sbi, false);
3285
3286     f2fs_up_write(&sbi->sb_lock);
3287
3288     mnt_drop_write_file(filp);
3289 out:
3290     kfree(vbuf);
3291     return err;
3292 }
3293
3294 static int f2fs_get_compress_blocks(struct file *filp, unsigned long arg)
3295 {
3296     struct inode *inode = file_inode(filp);
3297     __u64 blocks;
3298
3299     if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3300         return -EOPNOTSUPP;
3301
3302     if (!f2fs_compressed_file(inode))
3303         return -EINVAL;
3304
3305     blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks);
3306     return put_user(blocks, (u64 __user *)arg);
3307 }
3308
3309 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3310 {
3311     struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3312     unsigned int released_blocks = 0;
3313     int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3314     block_t blkaddr;
3315     int i;
3316
3317     for (i = 0; i < count; i++) {
3318         blkaddr = data_blkaddr(dn->inode, dn->node_page,
3319                         dn->ofs_in_node + i);
3320
3321         if (!__is_valid_data_blkaddr(blkaddr))
3322             continue;
3323         if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3324                     DATA_GENERIC_ENHANCE)))
3325             return -EFSCORRUPTED;
3326     }
3327
3328     while (count) {
3329         int compr_blocks = 0;
3330
3331         for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3332             blkaddr = f2fs_data_blkaddr(dn);
3333
3334             if (i == 0) {
3335                 if (blkaddr == COMPRESS_ADDR)
3336                     continue;
3337                 dn->ofs_in_node += cluster_size;
3338                 goto next;
3339             }
3340
3341             if (__is_valid_data_blkaddr(blkaddr))
3342                 compr_blocks++;
3343
3344             if (blkaddr != NEW_ADDR)
3345                 continue;
3346
3347             dn->data_blkaddr = NULL_ADDR;
3348             f2fs_set_data_blkaddr(dn);
3349         }
3350
3351         f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false);
3352         dec_valid_block_count(sbi, dn->inode,
3353                     cluster_size - compr_blocks);
3354
3355         released_blocks += cluster_size - compr_blocks;
3356 next:
3357         count -= cluster_size;
3358     }
3359
3360     return released_blocks;
3361 }
3362
3363 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
3364 {
3365     struct inode *inode = file_inode(filp);
3366     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3367     pgoff_t page_idx = 0, last_idx;
3368     unsigned int released_blocks = 0;
3369     int ret;
3370     int writecount;
3371
3372     if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3373         return -EOPNOTSUPP;
3374
3375     if (!f2fs_compressed_file(inode))
3376         return -EINVAL;
3377
3378     if (f2fs_readonly(sbi->sb))
3379         return -EROFS;
3380
3381     ret = mnt_want_write_file(filp);
3382     if (ret)
3383         return ret;
3384
3385     f2fs_balance_fs(F2FS_I_SB(inode), true);
3386
3387     inode_lock(inode);
3388
3389     writecount = atomic_read(&inode->i_writecount);
3390     if ((filp->f_mode & FMODE_WRITE && writecount != 1) ||
3391             (!(filp->f_mode & FMODE_WRITE) && writecount)) {
3392         ret = -EBUSY;
3393         goto out;
3394     }
3395
3396     if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3397         ret = -EINVAL;
3398         goto out;
3399     }
3400
3401     ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3402     if (ret)
3403         goto out;
3404
3405     set_inode_flag(inode, FI_COMPRESS_RELEASED);
3406     inode->i_ctime = current_time(inode);
3407     f2fs_mark_inode_dirty_sync(inode, true);
3408
3409     if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
3410         goto out;
3411
3412     f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3413     filemap_invalidate_lock(inode->i_mapping);
3414
3415     last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3416
3417     while (page_idx < last_idx) {
3418         struct dnode_of_data dn;
3419         pgoff_t end_offset, count;
3420
3421         set_new_dnode(&dn, inode, NULL, NULL, 0);
3422         ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3423         if (ret) {
3424             if (ret == -ENOENT) {
3425                 page_idx = f2fs_get_next_page_offset(&dn,
3426                                 page_idx);
3427                 ret = 0;
3428                 continue;
3429             }
3430             break;
3431         }
3432
3433         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3434         count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3435         count = round_up(count, F2FS_I(inode)->i_cluster_size);
3436
3437         ret = release_compress_blocks(&dn, count);
3438
3439         f2fs_put_dnode(&dn);
3440
3441         if (ret < 0)
3442             break;
3443
3444         page_idx += count;
3445         released_blocks += ret;
3446     }
3447
3448     filemap_invalidate_unlock(inode->i_mapping);
3449     f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3450 out:
3451     inode_unlock(inode);
3452
3453     mnt_drop_write_file(filp);
3454
3455     if (ret >= 0) {
3456         ret = put_user(released_blocks, (u64 __user *)arg);
3457     } else if (released_blocks &&
3458             atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3459         set_sbi_flag(sbi, SBI_NEED_FSCK);
3460         f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3461             "iblocks=%llu, released=%u, compr_blocks=%u, "
3462             "run fsck to fix.",
3463             __func__, inode->i_ino, inode->i_blocks,
3464             released_blocks,
3465             atomic_read(&F2FS_I(inode)->i_compr_blocks));
3466     }
3467
3468     return ret;
3469 }
3470
3471 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
3472 {
3473     struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
3474     unsigned int reserved_blocks = 0;
3475     int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
3476     block_t blkaddr;
3477     int i;
3478
3479     for (i = 0; i < count; i++) {
3480         blkaddr = data_blkaddr(dn->inode, dn->node_page,
3481                         dn->ofs_in_node + i);
3482
3483         if (!__is_valid_data_blkaddr(blkaddr))
3484             continue;
3485         if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr,
3486                     DATA_GENERIC_ENHANCE)))
3487             return -EFSCORRUPTED;
3488     }
3489
3490     while (count) {
3491         int compr_blocks = 0;
3492         blkcnt_t reserved;
3493         int ret;
3494
3495         for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) {
3496             blkaddr = f2fs_data_blkaddr(dn);
3497
3498             if (i == 0) {
3499                 if (blkaddr == COMPRESS_ADDR)
3500                     continue;
3501                 dn->ofs_in_node += cluster_size;
3502                 goto next;
3503             }
3504
3505             if (__is_valid_data_blkaddr(blkaddr)) {
3506                 compr_blocks++;
3507                 continue;
3508             }
3509
3510             dn->data_blkaddr = NEW_ADDR;
3511             f2fs_set_data_blkaddr(dn);
3512         }
3513
3514         reserved = cluster_size - compr_blocks;
3515         ret = inc_valid_block_count(sbi, dn->inode, &reserved);
3516         if (ret)
3517             return ret;
3518
3519         if (reserved != cluster_size - compr_blocks)
3520             return -ENOSPC;
3521
3522         f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
3523
3524         reserved_blocks += reserved;
3525 next:
3526         count -= cluster_size;
3527     }
3528
3529     return reserved_blocks;
3530 }
3531
3532 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
3533 {
3534     struct inode *inode = file_inode(filp);
3535     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3536     pgoff_t page_idx = 0, last_idx;
3537     unsigned int reserved_blocks = 0;
3538     int ret;
3539
3540     if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3541         return -EOPNOTSUPP;
3542
3543     if (!f2fs_compressed_file(inode))
3544         return -EINVAL;
3545
3546     if (f2fs_readonly(sbi->sb))
3547         return -EROFS;
3548
3549     ret = mnt_want_write_file(filp);
3550     if (ret)
3551         return ret;
3552
3553     if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
3554         goto out;
3555
3556     f2fs_balance_fs(F2FS_I_SB(inode), true);
3557
3558     inode_lock(inode);
3559
3560     if (!is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3561         ret = -EINVAL;
3562         goto unlock_inode;
3563     }
3564
3565     f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3566     filemap_invalidate_lock(inode->i_mapping);
3567
3568     last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3569
3570     while (page_idx < last_idx) {
3571         struct dnode_of_data dn;
3572         pgoff_t end_offset, count;
3573
3574         set_new_dnode(&dn, inode, NULL, NULL, 0);
3575         ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE);
3576         if (ret) {
3577             if (ret == -ENOENT) {
3578                 page_idx = f2fs_get_next_page_offset(&dn,
3579                                 page_idx);
3580                 ret = 0;
3581                 continue;
3582             }
3583             break;
3584         }
3585
3586         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3587         count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
3588         count = round_up(count, F2FS_I(inode)->i_cluster_size);
3589
3590         ret = reserve_compress_blocks(&dn, count);
3591
3592         f2fs_put_dnode(&dn);
3593
3594         if (ret < 0)
3595             break;
3596
3597         page_idx += count;
3598         reserved_blocks += ret;
3599     }
3600
3601     filemap_invalidate_unlock(inode->i_mapping);
3602     f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3603
3604     if (ret >= 0) {
3605         clear_inode_flag(inode, FI_COMPRESS_RELEASED);
3606         inode->i_ctime = current_time(inode);
3607         f2fs_mark_inode_dirty_sync(inode, true);
3608     }
3609 unlock_inode:
3610     inode_unlock(inode);
3611 out:
3612     mnt_drop_write_file(filp);
3613
3614     if (ret >= 0) {
3615         ret = put_user(reserved_blocks, (u64 __user *)arg);
3616     } else if (reserved_blocks &&
3617             atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
3618         set_sbi_flag(sbi, SBI_NEED_FSCK);
3619         f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx "
3620             "iblocks=%llu, reserved=%u, compr_blocks=%u, "
3621             "run fsck to fix.",
3622             __func__, inode->i_ino, inode->i_blocks,
3623             reserved_blocks,
3624             atomic_read(&F2FS_I(inode)->i_compr_blocks));
3625     }
3626
3627     return ret;
3628 }
3629
3630 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode,
3631         pgoff_t off, block_t block, block_t len, u32 flags)
3632 {
3633     sector_t sector = SECTOR_FROM_BLOCK(block);
3634     sector_t nr_sects = SECTOR_FROM_BLOCK(len);
3635     int ret = 0;
3636
3637     if (flags & F2FS_TRIM_FILE_DISCARD) {
3638         if (bdev_max_secure_erase_sectors(bdev))
3639             ret = blkdev_issue_secure_erase(bdev, sector, nr_sects,
3640                     GFP_NOFS);
3641         else
3642             ret = blkdev_issue_discard(bdev, sector, nr_sects,
3643                     GFP_NOFS);
3644     }
3645
3646     if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) {
3647         if (IS_ENCRYPTED(inode))
3648             ret = fscrypt_zeroout_range(inode, off, block, len);
3649         else
3650             ret = blkdev_issue_zeroout(bdev, sector, nr_sects,
3651                     GFP_NOFS, 0);
3652     }
3653
3654     return ret;
3655 }
3656
3657 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg)
3658 {
3659     struct inode *inode = file_inode(filp);
3660     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3661     struct address_space *mapping = inode->i_mapping;
3662     struct block_device *prev_bdev = NULL;
3663     struct f2fs_sectrim_range range;
3664     pgoff_t index, pg_end, prev_index = 0;
3665     block_t prev_block = 0, len = 0;
3666     loff_t end_addr;
3667     bool to_end = false;
3668     int ret = 0;
3669
3670     if (!(filp->f_mode & FMODE_WRITE))
3671         return -EBADF;
3672
3673     if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg,
3674                 sizeof(range)))
3675         return -EFAULT;
3676
3677     if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) ||
3678             !S_ISREG(inode->i_mode))
3679         return -EINVAL;
3680
3681     if (((range.flags & F2FS_TRIM_FILE_DISCARD) &&
3682             !f2fs_hw_support_discard(sbi)) ||
3683             ((range.flags & F2FS_TRIM_FILE_ZEROOUT) &&
3684              IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi)))
3685         return -EOPNOTSUPP;
3686
3687     file_start_write(filp);
3688     inode_lock(inode);
3689
3690     if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) ||
3691             range.start >= inode->i_size) {
3692         ret = -EINVAL;
3693         goto err;
3694     }
3695
3696     if (range.len == 0)
3697         goto err;
3698
3699     if (inode->i_size - range.start > range.len) {
3700         end_addr = range.start + range.len;
3701     } else {
3702         end_addr = range.len == (u64)-1 ?
3703             sbi->sb->s_maxbytes : inode->i_size;
3704         to_end = true;
3705     }
3706
3707     if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) ||
3708             (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) {
3709         ret = -EINVAL;
3710         goto err;
3711     }
3712
3713     index = F2FS_BYTES_TO_BLK(range.start);
3714     pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE);
3715
3716     ret = f2fs_convert_inline_inode(inode);
3717     if (ret)
3718         goto err;
3719
3720     f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3721     filemap_invalidate_lock(mapping);
3722
3723     ret = filemap_write_and_wait_range(mapping, range.start,
3724             to_end ? LLONG_MAX : end_addr - 1);
3725     if (ret)
3726         goto out;
3727
3728     truncate_inode_pages_range(mapping, range.start,
3729             to_end ? -1 : end_addr - 1);
3730
3731     while (index < pg_end) {
3732         struct dnode_of_data dn;
3733         pgoff_t end_offset, count;
3734         int i;
3735
3736         set_new_dnode(&dn, inode, NULL, NULL, 0);
3737         ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
3738         if (ret) {
3739             if (ret == -ENOENT) {
3740                 index = f2fs_get_next_page_offset(&dn, index);
3741                 continue;
3742             }
3743             goto out;
3744         }
3745
3746         end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
3747         count = min(end_offset - dn.ofs_in_node, pg_end - index);
3748         for (i = 0; i < count; i++, index++, dn.ofs_in_node++) {
3749             struct block_device *cur_bdev;
3750             block_t blkaddr = f2fs_data_blkaddr(&dn);
3751
3752             if (!__is_valid_data_blkaddr(blkaddr))
3753                 continue;
3754
3755             if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
3756                         DATA_GENERIC_ENHANCE)) {
3757                 ret = -EFSCORRUPTED;
3758                 f2fs_put_dnode(&dn);
3759                 goto out;
3760             }
3761
3762             cur_bdev = f2fs_target_device(sbi, blkaddr, NULL);
3763             if (f2fs_is_multi_device(sbi)) {
3764                 int di = f2fs_target_device_index(sbi, blkaddr);
3765
3766                 blkaddr -= FDEV(di).start_blk;
3767             }
3768
3769             if (len) {
3770                 if (prev_bdev == cur_bdev &&
3771                         index == prev_index + len &&
3772                         blkaddr == prev_block + len) {
3773                     len++;
3774                 } else {
3775                     ret = f2fs_secure_erase(prev_bdev,
3776                         inode, prev_index, prev_block,
3777                         len, range.flags);
3778                     if (ret) {
3779                         f2fs_put_dnode(&dn);
3780                         goto out;
3781                     }
3782
3783                     len = 0;
3784                 }
3785             }
3786
3787             if (!len) {
3788                 prev_bdev = cur_bdev;
3789                 prev_index = index;
3790                 prev_block = blkaddr;
3791                 len = 1;
3792             }
3793         }
3794
3795         f2fs_put_dnode(&dn);
3796
3797         if (fatal_signal_pending(current)) {
3798             ret = -EINTR;
3799             goto out;
3800         }
3801         cond_resched();
3802     }
3803
3804     if (len)
3805         ret = f2fs_secure_erase(prev_bdev, inode, prev_index,
3806                 prev_block, len, range.flags);
3807 out:
3808     filemap_invalidate_unlock(mapping);
3809     f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
3810 err:
3811     inode_unlock(inode);
3812     file_end_write(filp);
3813
3814     return ret;
3815 }
3816
3817 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg)
3818 {
3819     struct inode *inode = file_inode(filp);
3820     struct f2fs_comp_option option;
3821
3822     if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
3823         return -EOPNOTSUPP;
3824
3825     inode_lock_shared(inode);
3826
3827     if (!f2fs_compressed_file(inode)) {
3828         inode_unlock_shared(inode);
3829         return -ENODATA;
3830     }
3831
3832     option.algorithm = F2FS_I(inode)->i_compress_algorithm;
3833     option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size;
3834
3835     inode_unlock_shared(inode);
3836
3837     if (copy_to_user((struct f2fs_comp_option __user *)arg, &option,
3838                 sizeof(option)))
3839         return -EFAULT;
3840
3841     return 0;
3842 }
3843
3844 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg)
3845 {
3846     struct inode *inode = file_inode(filp);
3847     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3848     struct f2fs_comp_option option;
3849     int ret = 0;
3850
3851     if (!f2fs_sb_has_compression(sbi))
3852         return -EOPNOTSUPP;
3853
3854     if (!(filp->f_mode & FMODE_WRITE))
3855         return -EBADF;
3856
3857     if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg,
3858                 sizeof(option)))
3859         return -EFAULT;
3860
3861     if (!f2fs_compressed_file(inode) ||
3862             option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
3863             option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
3864             option.algorithm >= COMPRESS_MAX)
3865         return -EINVAL;
3866
3867     file_start_write(filp);
3868     inode_lock(inode);
3869
3870     if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) {
3871         ret = -EBUSY;
3872         goto out;
3873     }
3874
3875     if (inode->i_size != 0) {
3876         ret = -EFBIG;
3877         goto out;
3878     }
3879
3880     F2FS_I(inode)->i_compress_algorithm = option.algorithm;
3881     F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size;
3882     F2FS_I(inode)->i_cluster_size = 1 << option.log_cluster_size;
3883     f2fs_mark_inode_dirty_sync(inode, true);
3884
3885     if (!f2fs_is_compress_backend_ready(inode))
3886         f2fs_warn(sbi, "compression algorithm is successfully set, "
3887             "but current kernel doesn't support this algorithm.");
3888 out:
3889     inode_unlock(inode);
3890     file_end_write(filp);
3891
3892     return ret;
3893 }
3894
3895 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len)
3896 {
3897     DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx);
3898     struct address_space *mapping = inode->i_mapping;
3899     struct page *page;
3900     pgoff_t redirty_idx = page_idx;
3901     int i, page_len = 0, ret = 0;
3902
3903     page_cache_ra_unbounded(&ractl, len, 0);
3904
3905     for (i = 0; i < len; i++, page_idx++) {
3906         page = read_cache_page(mapping, page_idx, NULL, NULL);
3907         if (IS_ERR(page)) {
3908             ret = PTR_ERR(page);
3909             break;
3910         }
3911         page_len++;
3912     }
3913
3914     for (i = 0; i < page_len; i++, redirty_idx++) {
3915         page = find_lock_page(mapping, redirty_idx);
3916
3917         /* It will never fail, when page has pinned above */
3918         f2fs_bug_on(F2FS_I_SB(inode), !page);
3919
3920         set_page_dirty(page);
3921         f2fs_put_page(page, 1);
3922         f2fs_put_page(page, 0);
3923     }
3924
3925     return ret;
3926 }
3927
3928 static int f2fs_ioc_decompress_file(struct file *filp, unsigned long arg)
3929 {
3930     struct inode *inode = file_inode(filp);
3931     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3932     struct f2fs_inode_info *fi = F2FS_I(inode);
3933     pgoff_t page_idx = 0, last_idx;
3934     unsigned int blk_per_seg = sbi->blocks_per_seg;
3935     int cluster_size = fi->i_cluster_size;
3936     int count, ret;
3937
3938     if (!f2fs_sb_has_compression(sbi) ||
3939             F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
3940         return -EOPNOTSUPP;
3941
3942     if (!(filp->f_mode & FMODE_WRITE))
3943         return -EBADF;
3944
3945     if (!f2fs_compressed_file(inode))
3946         return -EINVAL;
3947
3948     f2fs_balance_fs(F2FS_I_SB(inode), true);
3949
3950     file_start_write(filp);
3951     inode_lock(inode);
3952
3953     if (!f2fs_is_compress_backend_ready(inode)) {
3954         ret = -EOPNOTSUPP;
3955         goto out;
3956     }
3957
3958     if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
3959         ret = -EINVAL;
3960         goto out;
3961     }
3962
3963     ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
3964     if (ret)
3965         goto out;
3966
3967     if (!atomic_read(&fi->i_compr_blocks))
3968         goto out;
3969
3970     last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
3971
3972     count = last_idx - page_idx;
3973     while (count) {
3974         int len = min(cluster_size, count);
3975
3976         ret = redirty_blocks(inode, page_idx, len);
3977         if (ret < 0)
3978             break;
3979
3980         if (get_dirty_pages(inode) >= blk_per_seg)
3981             filemap_fdatawrite(inode->i_mapping);
3982
3983         count -= len;
3984         page_idx += len;
3985     }
3986
3987     if (!ret)
3988         ret = filemap_write_and_wait_range(inode->i_mapping, 0,
3989                             LLONG_MAX);
3990
3991     if (ret)
3992         f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.",
3993               __func__, ret);
3994 out:
3995     inode_unlock(inode);
3996     file_end_write(filp);
3997
3998     return ret;
3999 }
4000
4001 static int f2fs_ioc_compress_file(struct file *filp, unsigned long arg)
4002 {
4003     struct inode *inode = file_inode(filp);
4004     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4005     pgoff_t page_idx = 0, last_idx;
4006     unsigned int blk_per_seg = sbi->blocks_per_seg;
4007     int cluster_size = F2FS_I(inode)->i_cluster_size;
4008     int count, ret;
4009
4010     if (!f2fs_sb_has_compression(sbi) ||
4011             F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER)
4012         return -EOPNOTSUPP;
4013
4014     if (!(filp->f_mode & FMODE_WRITE))
4015         return -EBADF;
4016
4017     if (!f2fs_compressed_file(inode))
4018         return -EINVAL;
4019
4020     f2fs_balance_fs(F2FS_I_SB(inode), true);
4021
4022     file_start_write(filp);
4023     inode_lock(inode);
4024
4025     if (!f2fs_is_compress_backend_ready(inode)) {
4026         ret = -EOPNOTSUPP;
4027         goto out;
4028     }
4029
4030     if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) {
4031         ret = -EINVAL;
4032         goto out;
4033     }
4034
4035     ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
4036     if (ret)
4037         goto out;
4038
4039     set_inode_flag(inode, FI_ENABLE_COMPRESS);
4040
4041     last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
4042
4043     count = last_idx - page_idx;
4044     while (count) {
4045         int len = min(cluster_size, count);
4046
4047         ret = redirty_blocks(inode, page_idx, len);
4048         if (ret < 0)
4049             break;
4050
4051         if (get_dirty_pages(inode) >= blk_per_seg)
4052             filemap_fdatawrite(inode->i_mapping);
4053
4054         count -= len;
4055         page_idx += len;
4056     }
4057
4058     if (!ret)
4059         ret = filemap_write_and_wait_range(inode->i_mapping, 0,
4060                             LLONG_MAX);
4061
4062     clear_inode_flag(inode, FI_ENABLE_COMPRESS);
4063
4064     if (ret)
4065         f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.",
4066               __func__, ret);
4067 out:
4068     inode_unlock(inode);
4069     file_end_write(filp);
4070
4071     return ret;
4072 }
4073
4074 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4075 {
4076     switch (cmd) {
4077     case FS_IOC_GETVERSION:
4078         return f2fs_ioc_getversion(filp, arg);
4079     case F2FS_IOC_START_ATOMIC_WRITE:
4080         return f2fs_ioc_start_atomic_write(filp);
4081     case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4082         return f2fs_ioc_commit_atomic_write(filp);
4083     case F2FS_IOC_ABORT_ATOMIC_WRITE:
4084         return f2fs_ioc_abort_atomic_write(filp);
4085     case F2FS_IOC_START_VOLATILE_WRITE:
4086     case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4087         return -EOPNOTSUPP;
4088     case F2FS_IOC_SHUTDOWN:
4089         return f2fs_ioc_shutdown(filp, arg);
4090     case FITRIM:
4091         return f2fs_ioc_fitrim(filp, arg);
4092     case FS_IOC_SET_ENCRYPTION_POLICY:
4093         return f2fs_ioc_set_encryption_policy(filp, arg);
4094     case FS_IOC_GET_ENCRYPTION_POLICY:
4095         return f2fs_ioc_get_encryption_policy(filp, arg);
4096     case FS_IOC_GET_ENCRYPTION_PWSALT:
4097         return f2fs_ioc_get_encryption_pwsalt(filp, arg);
4098     case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4099         return f2fs_ioc_get_encryption_policy_ex(filp, arg);
4100     case FS_IOC_ADD_ENCRYPTION_KEY:
4101         return f2fs_ioc_add_encryption_key(filp, arg);
4102     case FS_IOC_REMOVE_ENCRYPTION_KEY:
4103         return f2fs_ioc_remove_encryption_key(filp, arg);
4104     case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4105         return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
4106     case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4107         return f2fs_ioc_get_encryption_key_status(filp, arg);
4108     case FS_IOC_GET_ENCRYPTION_NONCE:
4109         return f2fs_ioc_get_encryption_nonce(filp, arg);
4110     case F2FS_IOC_GARBAGE_COLLECT:
4111         return f2fs_ioc_gc(filp, arg);
4112     case F2FS_IOC_GARBAGE_COLLECT_RANGE:
4113         return f2fs_ioc_gc_range(filp, arg);
4114     case F2FS_IOC_WRITE_CHECKPOINT:
4115         return f2fs_ioc_write_checkpoint(filp, arg);
4116     case F2FS_IOC_DEFRAGMENT:
4117         return f2fs_ioc_defragment(filp, arg);
4118     case F2FS_IOC_MOVE_RANGE:
4119         return f2fs_ioc_move_range(filp, arg);
4120     case F2FS_IOC_FLUSH_DEVICE:
4121         return f2fs_ioc_flush_device(filp, arg);
4122     case F2FS_IOC_GET_FEATURES:
4123         return f2fs_ioc_get_features(filp, arg);
4124     case F2FS_IOC_GET_PIN_FILE:
4125         return f2fs_ioc_get_pin_file(filp, arg);
4126     case F2FS_IOC_SET_PIN_FILE:
4127         return f2fs_ioc_set_pin_file(filp, arg);
4128     case F2FS_IOC_PRECACHE_EXTENTS:
4129         return f2fs_ioc_precache_extents(filp, arg);
4130     case F2FS_IOC_RESIZE_FS:
4131         return f2fs_ioc_resize_fs(filp, arg);
4132     case FS_IOC_ENABLE_VERITY:
4133         return f2fs_ioc_enable_verity(filp, arg);
4134     case FS_IOC_MEASURE_VERITY:
4135         return f2fs_ioc_measure_verity(filp, arg);
4136     case FS_IOC_READ_VERITY_METADATA:
4137         return f2fs_ioc_read_verity_metadata(filp, arg);
4138     case FS_IOC_GETFSLABEL:
4139         return f2fs_ioc_getfslabel(filp, arg);
4140     case FS_IOC_SETFSLABEL:
4141         return f2fs_ioc_setfslabel(filp, arg);
4142     case F2FS_IOC_GET_COMPRESS_BLOCKS:
4143         return f2fs_get_compress_blocks(filp, arg);
4144     case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4145         return f2fs_release_compress_blocks(filp, arg);
4146     case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4147         return f2fs_reserve_compress_blocks(filp, arg);
4148     case F2FS_IOC_SEC_TRIM_FILE:
4149         return f2fs_sec_trim_file(filp, arg);
4150     case F2FS_IOC_GET_COMPRESS_OPTION:
4151         return f2fs_ioc_get_compress_option(filp, arg);
4152     case F2FS_IOC_SET_COMPRESS_OPTION:
4153         return f2fs_ioc_set_compress_option(filp, arg);
4154     case F2FS_IOC_DECOMPRESS_FILE:
4155         return f2fs_ioc_decompress_file(filp, arg);
4156     case F2FS_IOC_COMPRESS_FILE:
4157         return f2fs_ioc_compress_file(filp, arg);
4158     default:
4159         return -ENOTTY;
4160     }
4161 }
4162
4163 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
4164 {
4165     if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
4166         return -EIO;
4167     if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
4168         return -ENOSPC;
4169
4170     return __f2fs_ioctl(filp, cmd, arg);
4171 }
4172
4173 /*
4174  * Return %true if the given read or write request should use direct I/O, or
4175  * %false if it should use buffered I/O.
4176  */
4177 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb,
4178                 struct iov_iter *iter)
4179 {
4180     unsigned int align;
4181
4182     if (!(iocb->ki_flags & IOCB_DIRECT))
4183         return false;
4184
4185     if (f2fs_force_buffered_io(inode, iocb, iter))
4186         return false;
4187
4188     /*
4189      * Direct I/O not aligned to the disk's logical_block_size will be
4190      * attempted, but will fail with -EINVAL.
4191      *
4192      * f2fs additionally requires that direct I/O be aligned to the
4193      * filesystem block size, which is often a stricter requirement.
4194      * However, f2fs traditionally falls back to buffered I/O on requests
4195      * that are logical_block_size-aligned but not fs-block aligned.
4196      *
4197      * The below logic implements this behavior.
4198      */
4199     align = iocb->ki_pos | iov_iter_alignment(iter);
4200     if (!IS_ALIGNED(align, i_blocksize(inode)) &&
4201         IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev)))
4202         return false;
4203
4204     return true;
4205 }
4206
4207 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
4208                 unsigned int flags)
4209 {
4210     struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4211
4212     dec_page_count(sbi, F2FS_DIO_READ);
4213     if (error)
4214         return error;
4215     f2fs_update_iostat(sbi, APP_DIRECT_READ_IO, size);
4216     return 0;
4217 }
4218
4219 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = {
4220     .end_io = f2fs_dio_read_end_io,
4221 };
4222
4223 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to)
4224 {
4225     struct file *file = iocb->ki_filp;
4226     struct inode *inode = file_inode(file);
4227     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4228     struct f2fs_inode_info *fi = F2FS_I(inode);
4229     const loff_t pos = iocb->ki_pos;
4230     const size_t count = iov_iter_count(to);
4231     struct iomap_dio *dio;
4232     ssize_t ret;
4233
4234     if (count == 0)
4235         return 0; /* skip atime update */
4236
4237     trace_f2fs_direct_IO_enter(inode, iocb, count, READ);
4238
4239     if (iocb->ki_flags & IOCB_NOWAIT) {
4240         if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4241             ret = -EAGAIN;
4242             goto out;
4243         }
4244     } else {
4245         f2fs_down_read(&fi->i_gc_rwsem[READ]);
4246     }
4247
4248     /*
4249      * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4250      * the higher-level function iomap_dio_rw() in order to ensure that the
4251      * F2FS_DIO_READ counter will be decremented correctly in all cases.
4252      */
4253     inc_page_count(sbi, F2FS_DIO_READ);
4254     dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops,
4255                  &f2fs_iomap_dio_read_ops, 0, NULL, 0);
4256     if (IS_ERR_OR_NULL(dio)) {
4257         ret = PTR_ERR_OR_ZERO(dio);
4258         if (ret != -EIOCBQUEUED)
4259             dec_page_count(sbi, F2FS_DIO_READ);
4260     } else {
4261         ret = iomap_dio_complete(dio);
4262     }
4263
4264     f2fs_up_read(&fi->i_gc_rwsem[READ]);
4265
4266     file_accessed(file);
4267 out:
4268     trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret);
4269     return ret;
4270 }
4271
4272 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
4273 {
4274     struct inode *inode = file_inode(iocb->ki_filp);
4275     const loff_t pos = iocb->ki_pos;
4276     ssize_t ret;
4277
4278     if (!f2fs_is_compress_backend_ready(inode))
4279         return -EOPNOTSUPP;
4280
4281     if (trace_f2fs_dataread_start_enabled()) {
4282         char *p = f2fs_kmalloc(F2FS_I_SB(inode), PATH_MAX, GFP_KERNEL);
4283         char *path;
4284
4285         if (!p)
4286             goto skip_read_trace;
4287
4288         path = dentry_path_raw(file_dentry(iocb->ki_filp), p, PATH_MAX);
4289         if (IS_ERR(path)) {
4290             kfree(p);
4291             goto skip_read_trace;
4292         }
4293
4294         trace_f2fs_dataread_start(inode, pos, iov_iter_count(to),
4295                     current->pid, path, current->comm);
4296         kfree(p);
4297     }
4298 skip_read_trace:
4299     if (f2fs_should_use_dio(inode, iocb, to)) {
4300         ret = f2fs_dio_read_iter(iocb, to);
4301     } else {
4302         ret = filemap_read(iocb, to, 0);
4303         if (ret > 0)
4304             f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_READ_IO, ret);
4305     }
4306     if (trace_f2fs_dataread_end_enabled())
4307         trace_f2fs_dataread_end(inode, pos, ret);
4308     return ret;
4309 }
4310
4311 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from)
4312 {
4313     struct file *file = iocb->ki_filp;
4314     struct inode *inode = file_inode(file);
4315     ssize_t count;
4316     int err;
4317
4318     if (IS_IMMUTABLE(inode))
4319         return -EPERM;
4320
4321     if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED))
4322         return -EPERM;
4323
4324     count = generic_write_checks(iocb, from);
4325     if (count <= 0)
4326         return count;
4327
4328     err = file_modified(file);
4329     if (err)
4330         return err;
4331     return count;
4332 }
4333
4334 /*
4335  * Preallocate blocks for a write request, if it is possible and helpful to do
4336  * so.  Returns a positive number if blocks may have been preallocated, 0 if no
4337  * blocks were preallocated, or a negative errno value if something went
4338  * seriously wrong.  Also sets FI_PREALLOCATED_ALL on the inode if *all* the
4339  * requested blocks (not just some of them) have been allocated.
4340  */
4341 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
4342                    bool dio)
4343 {
4344     struct inode *inode = file_inode(iocb->ki_filp);
4345     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4346     const loff_t pos = iocb->ki_pos;
4347     const size_t count = iov_iter_count(iter);
4348     struct f2fs_map_blocks map = {};
4349     int flag;
4350     int ret;
4351
4352     /* If it will be an out-of-place direct write, don't bother. */
4353     if (dio && f2fs_lfs_mode(sbi))
4354         return 0;
4355     /*
4356      * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
4357      * buffered IO, if DIO meets any holes.
4358      */
4359     if (dio && i_size_read(inode) &&
4360         (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
4361         return 0;
4362
4363     /* No-wait I/O can't allocate blocks. */
4364     if (iocb->ki_flags & IOCB_NOWAIT)
4365         return 0;
4366
4367     /* If it will be a short write, don't bother. */
4368     if (fault_in_iov_iter_readable(iter, count))
4369         return 0;
4370
4371     if (f2fs_has_inline_data(inode)) {
4372         /* If the data will fit inline, don't bother. */
4373         if (pos + count <= MAX_INLINE_DATA(inode))
4374             return 0;
4375         ret = f2fs_convert_inline_inode(inode);
4376         if (ret)
4377             return ret;
4378     }
4379
4380     /* Do not preallocate blocks that will be written partially in 4KB. */
4381     map.m_lblk = F2FS_BLK_ALIGN(pos);
4382     map.m_len = F2FS_BYTES_TO_BLK(pos + count);
4383     if (map.m_len > map.m_lblk)
4384         map.m_len -= map.m_lblk;
4385     else
4386         map.m_len = 0;
4387     map.m_may_create = true;
4388     if (dio) {
4389         map.m_seg_type = f2fs_rw_hint_to_seg_type(inode->i_write_hint);
4390         flag = F2FS_GET_BLOCK_PRE_DIO;
4391     } else {
4392         map.m_seg_type = NO_CHECK_TYPE;
4393         flag = F2FS_GET_BLOCK_PRE_AIO;
4394     }
4395
4396     ret = f2fs_map_blocks(inode, &map, 1, flag);
4397     /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */
4398     if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0))
4399         return ret;
4400     if (ret == 0)
4401         set_inode_flag(inode, FI_PREALLOCATED_ALL);
4402     return map.m_len;
4403 }
4404
4405 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb,
4406                     struct iov_iter *from)
4407 {
4408     struct file *file = iocb->ki_filp;
4409     struct inode *inode = file_inode(file);
4410     ssize_t ret;
4411
4412     if (iocb->ki_flags & IOCB_NOWAIT)
4413         return -EOPNOTSUPP;
4414
4415     current->backing_dev_info = inode_to_bdi(inode);
4416     ret = generic_perform_write(iocb, from);
4417     current->backing_dev_info = NULL;
4418
4419     if (ret > 0) {
4420         iocb->ki_pos += ret;
4421         f2fs_update_iostat(F2FS_I_SB(inode), APP_BUFFERED_IO, ret);
4422     }
4423     return ret;
4424 }
4425
4426 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
4427                  unsigned int flags)
4428 {
4429     struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp));
4430
4431     dec_page_count(sbi, F2FS_DIO_WRITE);
4432     if (error)
4433         return error;
4434     f2fs_update_iostat(sbi, APP_DIRECT_IO, size);
4435     return 0;
4436 }
4437
4438 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = {
4439     .end_io = f2fs_dio_write_end_io,
4440 };
4441
4442 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from,
4443                    bool *may_need_sync)
4444 {
4445     struct file *file = iocb->ki_filp;
4446     struct inode *inode = file_inode(file);
4447     struct f2fs_inode_info *fi = F2FS_I(inode);
4448     struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4449     const bool do_opu = f2fs_lfs_mode(sbi);
4450     const loff_t pos = iocb->ki_pos;
4451     const ssize_t count = iov_iter_count(from);
4452     unsigned int dio_flags;
4453     struct iomap_dio *dio;
4454     ssize_t ret;
4455
4456     trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE);
4457
4458     if (iocb->ki_flags & IOCB_NOWAIT) {
4459         /* f2fs_convert_inline_inode() and block allocation can block */
4460         if (f2fs_has_inline_data(inode) ||
4461             !f2fs_overwrite_io(inode, pos, count)) {
4462             ret = -EAGAIN;
4463             goto out;
4464         }
4465
4466         if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) {
4467             ret = -EAGAIN;
4468             goto out;
4469         }
4470         if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) {
4471             f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4472             ret = -EAGAIN;
4473             goto out;
4474         }
4475     } else {
4476         ret = f2fs_convert_inline_inode(inode);
4477         if (ret)
4478             goto out;
4479
4480         f2fs_down_read(&fi->i_gc_rwsem[WRITE]);
4481         if (do_opu)
4482             f2fs_down_read(&fi->i_gc_rwsem[READ]);
4483     }
4484
4485     /*
4486      * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of
4487      * the higher-level function iomap_dio_rw() in order to ensure that the
4488      * F2FS_DIO_WRITE counter will be decremented correctly in all cases.
4489      */
4490     inc_page_count(sbi, F2FS_DIO_WRITE);
4491     dio_flags = 0;
4492     if (pos + count > inode->i_size)
4493         dio_flags |= IOMAP_DIO_FORCE_WAIT;
4494     dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops,
4495                  &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0);
4496     if (IS_ERR_OR_NULL(dio)) {
4497         ret = PTR_ERR_OR_ZERO(dio);
4498         if (ret == -ENOTBLK)
4499             ret = 0;
4500         if (ret != -EIOCBQUEUED)
4501             dec_page_count(sbi, F2FS_DIO_WRITE);
4502     } else {
4503         ret = iomap_dio_complete(dio);
4504     }
4505
4506     if (do_opu)
4507         f2fs_up_read(&fi->i_gc_rwsem[READ]);
4508     f2fs_up_read(&fi->i_gc_rwsem[WRITE]);
4509
4510     if (ret < 0)
4511         goto out;
4512     if (pos + ret > inode->i_size)
4513         f2fs_i_size_write(inode, pos + ret);
4514     if (!do_opu)
4515         set_inode_flag(inode, FI_UPDATE_WRITE);
4516
4517     if (iov_iter_count(from)) {
4518         ssize_t ret2;
4519         loff_t bufio_start_pos = iocb->ki_pos;
4520
4521         /*
4522          * The direct write was partial, so we need to fall back to a
4523          * buffered write for the remainder.
4524          */
4525
4526         ret2 = f2fs_buffered_write_iter(iocb, from);
4527         if (iov_iter_count(from))
4528             f2fs_write_failed(inode, iocb->ki_pos);
4529         if (ret2 < 0)
4530             goto out;
4531
4532         /*
4533          * Ensure that the pagecache pages are written to disk and
4534          * invalidated to preserve the expected O_DIRECT semantics.
4535          */
4536         if (ret2 > 0) {
4537             loff_t bufio_end_pos = bufio_start_pos + ret2 - 1;
4538
4539             ret += ret2;
4540
4541             ret2 = filemap_write_and_wait_range(file->f_mapping,
4542                                 bufio_start_pos,
4543                                 bufio_end_pos);
4544             if (ret2 < 0)
4545                 goto out;
4546             invalidate_mapping_pages(file->f_mapping,
4547                          bufio_start_pos >> PAGE_SHIFT,
4548                          bufio_end_pos >> PAGE_SHIFT);
4549         }
4550     } else {
4551         /* iomap_dio_rw() already handled the generic_write_sync(). */
4552         *may_need_sync = false;
4553     }
4554 out:
4555     trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret);
4556     return ret;
4557 }
4558
4559 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
4560 {
4561     struct inode *inode = file_inode(iocb->ki_filp);
4562     const loff_t orig_pos = iocb->ki_pos;
4563     const size_t orig_count = iov_iter_count(from);
4564     loff_t target_size;
4565     bool dio;
4566     bool may_need_sync = true;
4567     int preallocated;
4568     ssize_t ret;
4569
4570     if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
4571         ret = -EIO;
4572         goto out;
4573     }
4574
4575     if (!f2fs_is_compress_backend_ready(inode)) {
4576         ret = -EOPNOTSUPP;
4577         goto out;
4578     }
4579
4580     if (iocb->ki_flags & IOCB_NOWAIT) {
4581         if (!inode_trylock(inode)) {
4582             ret = -EAGAIN;
4583             goto out;
4584         }
4585     } else {
4586         inode_lock(inode);
4587     }
4588
4589     ret = f2fs_write_checks(iocb, from);
4590     if (ret <= 0)
4591         goto out_unlock;
4592
4593     /* Determine whether we will do a direct write or a buffered write. */
4594     dio = f2fs_should_use_dio(inode, iocb, from);
4595
4596     /* Possibly preallocate the blocks for the write. */
4597     target_size = iocb->ki_pos + iov_iter_count(from);
4598     preallocated = f2fs_preallocate_blocks(iocb, from, dio);
4599     if (preallocated < 0) {
4600         ret = preallocated;
4601     } else {
4602         if (trace_f2fs_datawrite_start_enabled()) {
4603             char *p = f2fs_kmalloc(F2FS_I_SB(inode),
4604                         PATH_MAX, GFP_KERNEL);
4605             char *path;
4606
4607             if (!p)
4608                 goto skip_write_trace;
4609             path = dentry_path_raw(file_dentry(iocb->ki_filp),
4610                                 p, PATH_MAX);
4611             if (IS_ERR(path)) {
4612                 kfree(p);
4613                 goto skip_write_trace;
4614             }
4615             trace_f2fs_datawrite_start(inode, orig_pos, orig_count,
4616                     current->pid, path, current->comm);
4617             kfree(p);
4618         }
4619 skip_write_trace:
4620         /* Do the actual write. */
4621         ret = dio ?
4622             f2fs_dio_write_iter(iocb, from, &may_need_sync):
4623             f2fs_buffered_write_iter(iocb, from);
4624
4625         if (trace_f2fs_datawrite_end_enabled())
4626             trace_f2fs_datawrite_end(inode, orig_pos, ret);
4627     }
4628
4629     /* Don't leave any preallocated blocks around past i_size. */
4630     if (preallocated && i_size_read(inode) < target_size) {
4631         f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4632         filemap_invalidate_lock(inode->i_mapping);
4633         if (!f2fs_truncate(inode))
4634             file_dont_truncate(inode);
4635         filemap_invalidate_unlock(inode->i_mapping);
4636         f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
4637     } else {
4638         file_dont_truncate(inode);
4639     }
4640
4641     clear_inode_flag(inode, FI_PREALLOCATED_ALL);
4642 out_unlock:
4643     inode_unlock(inode);
4644 out:
4645     trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret);
4646     if (ret > 0 && may_need_sync)
4647         ret = generic_write_sync(iocb, ret);
4648     return ret;
4649 }
4650
4651 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
4652         int advice)
4653 {
4654     struct address_space *mapping;
4655     struct backing_dev_info *bdi;
4656     struct inode *inode = file_inode(filp);
4657     int err;
4658
4659     if (advice == POSIX_FADV_SEQUENTIAL) {
4660         if (S_ISFIFO(inode->i_mode))
4661             return -ESPIPE;
4662
4663         mapping = filp->f_mapping;
4664         if (!mapping || len < 0)
4665             return -EINVAL;
4666
4667         bdi = inode_to_bdi(mapping->host);
4668         filp->f_ra.ra_pages = bdi->ra_pages *
4669             F2FS_I_SB(inode)->seq_file_ra_mul;
4670         spin_lock(&filp->f_lock);
4671         filp->f_mode &= ~FMODE_RANDOM;
4672         spin_unlock(&filp->f_lock);
4673         return 0;
4674     }
4675
4676     err = generic_fadvise(filp, offset, len, advice);
4677     if (!err && advice == POSIX_FADV_DONTNEED &&
4678         test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
4679         f2fs_compressed_file(inode))
4680         f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
4681
4682     return err;
4683 }
4684
4685 #ifdef CONFIG_COMPAT
4686 struct compat_f2fs_gc_range {
4687     u32 sync;
4688     compat_u64 start;
4689     compat_u64 len;
4690 };
4691 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE    _IOW(F2FS_IOCTL_MAGIC, 11,\
4692                         struct compat_f2fs_gc_range)
4693
4694 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg)
4695 {
4696     struct compat_f2fs_gc_range __user *urange;
4697     struct f2fs_gc_range range;
4698     int err;
4699
4700     urange = compat_ptr(arg);
4701     err = get_user(range.sync, &urange->sync);
4702     err |= get_user(range.start, &urange->start);
4703     err |= get_user(range.len, &urange->len);
4704     if (err)
4705         return -EFAULT;
4706
4707     return __f2fs_ioc_gc_range(file, &range);
4708 }
4709
4710 struct compat_f2fs_move_range {
4711     u32 dst_fd;
4712     compat_u64 pos_in;
4713     compat_u64 pos_out;
4714     compat_u64 len;
4715 };
4716 #define F2FS_IOC32_MOVE_RANGE       _IOWR(F2FS_IOCTL_MAGIC, 9,  \
4717                     struct compat_f2fs_move_range)
4718
4719 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg)
4720 {
4721     struct compat_f2fs_move_range __user *urange;
4722     struct f2fs_move_range range;
4723     int err;
4724
4725     urange = compat_ptr(arg);
4726     err = get_user(range.dst_fd, &urange->dst_fd);
4727     err |= get_user(range.pos_in, &urange->pos_in);
4728     err |= get_user(range.pos_out, &urange->pos_out);
4729     err |= get_user(range.len, &urange->len);
4730     if (err)
4731         return -EFAULT;
4732
4733     return __f2fs_ioc_move_range(file, &range);
4734 }
4735
4736 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4737 {
4738     if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
4739         return -EIO;
4740     if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file))))
4741         return -ENOSPC;
4742
4743     switch (cmd) {
4744     case FS_IOC32_GETVERSION:
4745         cmd = FS_IOC_GETVERSION;
4746         break;
4747     case F2FS_IOC32_GARBAGE_COLLECT_RANGE:
4748         return f2fs_compat_ioc_gc_range(file, arg);
4749     case F2FS_IOC32_MOVE_RANGE:
4750         return f2fs_compat_ioc_move_range(file, arg);
4751     case F2FS_IOC_START_ATOMIC_WRITE:
4752     case F2FS_IOC_COMMIT_ATOMIC_WRITE:
4753     case F2FS_IOC_START_VOLATILE_WRITE:
4754     case F2FS_IOC_RELEASE_VOLATILE_WRITE:
4755     case F2FS_IOC_ABORT_ATOMIC_WRITE:
4756     case F2FS_IOC_SHUTDOWN:
4757     case FITRIM:
4758     case FS_IOC_SET_ENCRYPTION_POLICY:
4759     case FS_IOC_GET_ENCRYPTION_PWSALT:
4760     case FS_IOC_GET_ENCRYPTION_POLICY:
4761     case FS_IOC_GET_ENCRYPTION_POLICY_EX:
4762     case FS_IOC_ADD_ENCRYPTION_KEY:
4763     case FS_IOC_REMOVE_ENCRYPTION_KEY:
4764     case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
4765     case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
4766     case FS_IOC_GET_ENCRYPTION_NONCE:
4767     case F2FS_IOC_GARBAGE_COLLECT:
4768     case F2FS_IOC_WRITE_CHECKPOINT:
4769     case F2FS_IOC_DEFRAGMENT:
4770     case F2FS_IOC_FLUSH_DEVICE:
4771     case F2FS_IOC_GET_FEATURES:
4772     case F2FS_IOC_GET_PIN_FILE:
4773     case F2FS_IOC_SET_PIN_FILE:
4774     case F2FS_IOC_PRECACHE_EXTENTS:
4775     case F2FS_IOC_RESIZE_FS:
4776     case FS_IOC_ENABLE_VERITY:
4777     case FS_IOC_MEASURE_VERITY:
4778     case FS_IOC_READ_VERITY_METADATA:
4779     case FS_IOC_GETFSLABEL:
4780     case FS_IOC_SETFSLABEL:
4781     case F2FS_IOC_GET_COMPRESS_BLOCKS:
4782     case F2FS_IOC_RELEASE_COMPRESS_BLOCKS:
4783     case F2FS_IOC_RESERVE_COMPRESS_BLOCKS:
4784     case F2FS_IOC_SEC_TRIM_FILE:
4785     case F2FS_IOC_GET_COMPRESS_OPTION:
4786     case F2FS_IOC_SET_COMPRESS_OPTION:
4787     case F2FS_IOC_DECOMPRESS_FILE:
4788     case F2FS_IOC_COMPRESS_FILE:
4789         break;
4790     default:
4791         return -ENOIOCTLCMD;
4792     }
4793     return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4794 }
4795 #endif
4796
4797 const struct file_operations f2fs_file_operations = {
4798     .llseek     = f2fs_llseek,
4799     .read_iter  = f2fs_file_read_iter,
4800     .write_iter = f2fs_file_write_iter,
4801     .open       = f2fs_file_open,
4802     .release    = f2fs_release_file,
4803     .mmap       = f2fs_file_mmap,
4804     .flush      = f2fs_file_flush,
4805     .fsync      = f2fs_sync_file,
4806     .fallocate  = f2fs_fallocate,
4807     .unlocked_ioctl = f2fs_ioctl,
4808 #ifdef CONFIG_COMPAT
4809     .compat_ioctl   = f2fs_compat_ioctl,
4810 #endif
4811     .splice_read    = generic_file_splice_read,
4812     .splice_write   = iter_file_splice_write,
4813     .fadvise    = f2fs_file_fadvise,
4814 };