1 files changed, 133 insertions, 139 deletions

diff --git a/fs/erofs/inode.c b/fs/erofs/inode.c
index 0eb0e6f933c3..a0ae0b4f7b01 100644
--- a/fs/erofs/inode.c
+++ b/fs/erofs/inode.c
@@ -5,35 +5,49 @@
  * Copyright (C) 2021, Alibaba Cloud
  */
 #include "xattr.h"
-
 #include <trace/events/erofs.h>
 
-static void *erofs_read_inode(struct erofs_buf *buf,
-			      struct inode *inode, unsigned int *ofs)
+static int erofs_fill_symlink(struct inode *inode, void *kaddr,
+			      unsigned int m_pofs)
+{
+	struct erofs_inode *vi = EROFS_I(inode);
+	loff_t off;
+
+	m_pofs += vi->xattr_isize;
+	/* check if it cannot be handled with fast symlink scheme */
+	if (vi->datalayout != EROFS_INODE_FLAT_INLINE ||
+	    check_add_overflow(m_pofs, inode->i_size, &off) ||
+	    off > i_blocksize(inode))
+		return 0;
+
+	inode->i_link = kmemdup_nul(kaddr + m_pofs, inode->i_size, GFP_KERNEL);
+	return inode->i_link ? 0 : -ENOMEM;
+}
+
+static int erofs_read_inode(struct inode *inode)
 {
 	struct super_block *sb = inode->i_sb;
+	erofs_blk_t blkaddr = erofs_blknr(sb, erofs_iloc(inode));
+	unsigned int ofs = erofs_blkoff(sb, erofs_iloc(inode));
+	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
+	erofs_blk_t addrmask = BIT_ULL(48) - 1;
 	struct erofs_inode *vi = EROFS_I(inode);
-	const erofs_off_t inode_loc = erofs_iloc(inode);
-	erofs_blk_t blkaddr, nblks = 0;
-	void *kaddr;
+	struct erofs_inode_extended *die, copied;
 	struct erofs_inode_compact *dic;
-	struct erofs_inode_extended *die, *copied = NULL;
-	union erofs_inode_i_u iu;
 	unsigned int ifmt;
-	int err;
-
-	blkaddr = erofs_blknr(sb, inode_loc);
-	*ofs = erofs_blkoff(sb, inode_loc);
+	void *ptr;
+	int err = 0;
 
-	kaddr = erofs_read_metabuf(buf, sb, blkaddr, EROFS_KMAP);
-	if (IS_ERR(kaddr)) {
-		erofs_err(sb, "failed to get inode (nid: %llu) page, err %ld",
-			  vi->nid, PTR_ERR(kaddr));
-		return kaddr;
+	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr), true);
+	if (IS_ERR(ptr)) {
+		err = PTR_ERR(ptr);
+		erofs_err(sb, "failed to get inode (nid: %llu) page, err %d",
+			  vi->nid, err);
+		goto err_out;
 	}
 
-	dic = kaddr + *ofs;
+	dic = ptr + ofs;
 	ifmt = le16_to_cpu(dic->i_format);
 	if (ifmt & ~EROFS_I_ALL) {
 		erofs_err(sb, "unsupported i_format %u of nid %llu",
@@ -54,57 +68,58 @@ static void *erofs_read_inode(struct erofs_buf *buf,
 	case EROFS_INODE_LAYOUT_EXTENDED:
 		vi->inode_isize = sizeof(struct erofs_inode_extended);
 		/* check if the extended inode acrosses block boundary */
-		if (*ofs + vi->inode_isize <= sb->s_blocksize) {
-			*ofs += vi->inode_isize;
+		if (ofs + vi->inode_isize <= sb->s_blocksize) {
+			ofs += vi->inode_isize;
 			die = (struct erofs_inode_extended *)dic;
+			copied.i_u = die->i_u;
+			copied.i_nb = die->i_nb;
 		} else {
-			const unsigned int gotten = sb->s_blocksize - *ofs;
-
-			copied = kmalloc(vi->inode_isize, GFP_KERNEL);
-			if (!copied) {
-				err = -ENOMEM;
+			const unsigned int gotten = sb->s_blocksize - ofs;
+
+			memcpy(&copied, dic, gotten);
+			ptr = erofs_read_metabuf(&buf, sb,
+					erofs_pos(sb, blkaddr + 1), true);
+			if (IS_ERR(ptr)) {
+				err = PTR_ERR(ptr);
+				erofs_err(sb, "failed to get inode payload block (nid: %llu), err %d",
+					  vi->nid, err);
 				goto err_out;
 			}
-			memcpy(copied, dic, gotten);
-			kaddr = erofs_read_metabuf(buf, sb, blkaddr + 1,
-						   EROFS_KMAP);
-			if (IS_ERR(kaddr)) {
-				erofs_err(sb, "failed to get inode payload block (nid: %llu), err %ld",
-					  vi->nid, PTR_ERR(kaddr));
-				kfree(copied);
-				return kaddr;
-			}
-			*ofs = vi->inode_isize - gotten;
-			memcpy((u8 *)copied + gotten, kaddr, *ofs);
-			die = copied;
+			ofs = vi->inode_isize - gotten;
+			memcpy((u8 *)&copied + gotten, ptr, ofs);
+			die = &copied;
 		}
 		vi->xattr_isize = erofs_xattr_ibody_size(die->i_xattr_icount);
 
 		inode->i_mode = le16_to_cpu(die->i_mode);
-		iu = die->i_u;
 		i_uid_write(inode, le32_to_cpu(die->i_uid));
 		i_gid_write(inode, le32_to_cpu(die->i_gid));
 		set_nlink(inode, le32_to_cpu(die->i_nlink));
-		/* each extended inode has its own timestamp */
-		inode_set_ctime(inode, le64_to_cpu(die->i_mtime),
+		inode_set_mtime(inode, le64_to_cpu(die->i_mtime),
 				le32_to_cpu(die->i_mtime_nsec));
 
 		inode->i_size = le64_to_cpu(die->i_size);
-		kfree(copied);
-		copied = NULL;
 		break;
 	case EROFS_INODE_LAYOUT_COMPACT:
 		vi->inode_isize = sizeof(struct erofs_inode_compact);
-		*ofs += vi->inode_isize;
+		ofs += vi->inode_isize;
 		vi->xattr_isize = erofs_xattr_ibody_size(dic->i_xattr_icount);
 
 		inode->i_mode = le16_to_cpu(dic->i_mode);
-		iu = dic->i_u;
+		copied.i_u = dic->i_u;
 		i_uid_write(inode, le16_to_cpu(dic->i_uid));
 		i_gid_write(inode, le16_to_cpu(dic->i_gid));
-		set_nlink(inode, le16_to_cpu(dic->i_nlink));
-		/* use build time for compact inodes */
-		inode_set_ctime(inode, sbi->build_time, sbi->build_time_nsec);
+		if (!S_ISDIR(inode->i_mode) &&
+		    ((ifmt >> EROFS_I_NLINK_1_BIT) & 1)) {
+			set_nlink(inode, 1);
+			copied.i_nb = dic->i_nb;
+		} else {
+			set_nlink(inode, le16_to_cpu(dic->i_nb.nlink));
+			copied.i_nb.startblk_hi = 0;
+			addrmask = BIT_ULL(32) - 1;
+		}
+		inode_set_mtime(inode, sbi->epoch + le32_to_cpu(dic->i_mtime),
+				sbi->fixed_nsec);
 
 		inode->i_size = le32_to_cpu(dic->i_size);
 		break;
@@ -115,15 +130,32 @@ static void *erofs_read_inode(struct erofs_buf *buf,
 		goto err_out;
 	}
 
+	if (unlikely(inode->i_size < 0)) {
+		erofs_err(sb, "negative i_size @ nid %llu", vi->nid);
+		err = -EFSCORRUPTED;
+		goto err_out;
+	}
 	switch (inode->i_mode & S_IFMT) {
-	case S_IFREG:
 	case S_IFDIR:
+		vi->dot_omitted = (ifmt >> EROFS_I_DOT_OMITTED_BIT) & 1;
+		fallthrough;
+	case S_IFREG:
 	case S_IFLNK:
-		vi->raw_blkaddr = le32_to_cpu(iu.raw_blkaddr);
+		vi->startblk = le32_to_cpu(copied.i_u.startblk_lo) |
+			((u64)le16_to_cpu(copied.i_nb.startblk_hi) << 32);
+		if (vi->datalayout == EROFS_INODE_FLAT_PLAIN &&
+		    !((vi->startblk ^ EROFS_NULL_ADDR) & addrmask))
+			vi->startblk = EROFS_NULL_ADDR;
+
+		if(S_ISLNK(inode->i_mode)) {
+			err = erofs_fill_symlink(inode, ptr, ofs);
+			if (err)
+				goto err_out;
+		}
 		break;
 	case S_IFCHR:
 	case S_IFBLK:
-		inode->i_rdev = new_decode_dev(le32_to_cpu(iu.rdev));
+		inode->i_rdev = new_decode_dev(le32_to_cpu(copied.i_u.rdev));
 		break;
 	case S_IFIFO:
 	case S_IFSOCK:
@@ -136,12 +168,15 @@ static void *erofs_read_inode(struct erofs_buf *buf,
 		goto err_out;
 	}
 
-	/* total blocks for compressed files */
-	if (erofs_inode_is_data_compressed(vi->datalayout)) {
-		nblks = le32_to_cpu(iu.compressed_blocks);
-	} else if (vi->datalayout == EROFS_INODE_CHUNK_BASED) {
+	if (erofs_inode_is_data_compressed(vi->datalayout))
+		inode->i_blocks = le32_to_cpu(copied.i_u.blocks_lo) <<
+					(sb->s_blocksize_bits - 9);
+	else
+		inode->i_blocks = round_up(inode->i_size, sb->s_blocksize) >> 9;
+
+	if (vi->datalayout == EROFS_INODE_CHUNK_BASED) {
 		/* fill chunked inode summary info */
-		vi->chunkformat = le16_to_cpu(iu.c.format);
+		vi->chunkformat = le16_to_cpu(copied.i_u.c.format);
 		if (vi->chunkformat & ~EROFS_CHUNK_FORMAT_ALL) {
 			erofs_err(sb, "unsupported chunk format %x of nid %llu",
 				  vi->chunkformat, vi->nid);
@@ -151,81 +186,29 @@ static void *erofs_read_inode(struct erofs_buf *buf,
 		vi->chunkbits = sb->s_blocksize_bits +
 			(vi->chunkformat & EROFS_CHUNK_FORMAT_BLKBITS_MASK);
 	}
-	inode_set_mtime_to_ts(inode,
-			      inode_set_atime_to_ts(inode, inode_get_ctime(inode)));
+	inode_set_atime_to_ts(inode,
+			      inode_set_ctime_to_ts(inode, inode_get_mtime(inode)));
 
 	inode->i_flags &= ~S_DAX;
 	if (test_opt(&sbi->opt, DAX_ALWAYS) && S_ISREG(inode->i_mode) &&
 	    (vi->datalayout == EROFS_INODE_FLAT_PLAIN ||
 	     vi->datalayout == EROFS_INODE_CHUNK_BASED))
 		inode->i_flags |= S_DAX;
-
-	if (!nblks)
-		/* measure inode.i_blocks as generic filesystems */
-		inode->i_blocks = round_up(inode->i_size, sb->s_blocksize) >> 9;
-	else
-		inode->i_blocks = nblks << (sb->s_blocksize_bits - 9);
-	return kaddr;
-
 err_out:
-	DBG_BUGON(1);
-	kfree(copied);
-	erofs_put_metabuf(buf);
-	return ERR_PTR(err);
-}
-
-static int erofs_fill_symlink(struct inode *inode, void *kaddr,
-			      unsigned int m_pofs)
-{
-	struct erofs_inode *vi = EROFS_I(inode);
-	unsigned int bsz = i_blocksize(inode);
-	char *lnk;
-
-	/* if it cannot be handled with fast symlink scheme */
-	if (vi->datalayout != EROFS_INODE_FLAT_INLINE ||
-	    inode->i_size >= bsz || inode->i_size < 0) {
-		inode->i_op = &erofs_symlink_iops;
-		return 0;
-	}
-
-	lnk = kmalloc(inode->i_size + 1, GFP_KERNEL);
-	if (!lnk)
-		return -ENOMEM;
-
-	m_pofs += vi->xattr_isize;
-	/* inline symlink data shouldn't cross block boundary */
-	if (m_pofs + inode->i_size > bsz) {
-		kfree(lnk);
-		erofs_err(inode->i_sb,
-			  "inline data cross block boundary @ nid %llu",
-			  vi->nid);
-		DBG_BUGON(1);
-		return -EFSCORRUPTED;
-	}
-	memcpy(lnk, kaddr + m_pofs, inode->i_size);
-	lnk[inode->i_size] = '\0';
-
-	inode->i_link = lnk;
-	inode->i_op = &erofs_fast_symlink_iops;
-	return 0;
+	erofs_put_metabuf(&buf);
+	return err;
 }
 
 static int erofs_fill_inode(struct inode *inode)
 {
 	struct erofs_inode *vi = EROFS_I(inode);
-	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
-	void *kaddr;
-	unsigned int ofs;
-	int err = 0;
+	int err;
 
 	trace_erofs_fill_inode(inode);
+	err = erofs_read_inode(inode);
+	if (err)
+		return err;
 
-	/* read inode base data from disk */
-	kaddr = erofs_read_inode(&buf, inode, &ofs);
-	if (IS_ERR(kaddr))
-		return PTR_ERR(kaddr);
-
-	/* setup the new inode */
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFREG:
 		inode->i_op = &erofs_generic_iops;
@@ -240,44 +223,40 @@ static int erofs_fill_inode(struct inode *inode)
 		inode_nohighmem(inode);
 		break;
 	case S_IFLNK:
-		err = erofs_fill_symlink(inode, kaddr, ofs);
-		if (err)
-			goto out_unlock;
+		if (inode->i_link)
+			inode->i_op = &erofs_fast_symlink_iops;
+		else
+			inode->i_op = &erofs_symlink_iops;
 		inode_nohighmem(inode);
 		break;
-	case S_IFCHR:
-	case S_IFBLK:
-	case S_IFIFO:
-	case S_IFSOCK:
+	default:
 		inode->i_op = &erofs_generic_iops;
 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
-		goto out_unlock;
-	default:
-		err = -EFSCORRUPTED;
-		goto out_unlock;
+		return 0;
 	}
 
+	mapping_set_large_folios(inode->i_mapping);
 	if (erofs_inode_is_data_compressed(vi->datalayout)) {
 #ifdef CONFIG_EROFS_FS_ZIP
 		DO_ONCE_LITE_IF(inode->i_blkbits != PAGE_SHIFT,
 			  erofs_info, inode->i_sb,
 			  "EXPERIMENTAL EROFS subpage compressed block support in use. Use at your own risk!");
 		inode->i_mapping->a_ops = &z_erofs_aops;
-		err = 0;
-		goto out_unlock;
-#endif
+#else
 		err = -EOPNOTSUPP;
-		goto out_unlock;
-	}
-	inode->i_mapping->a_ops = &erofs_raw_access_aops;
-	mapping_set_large_folios(inode->i_mapping);
+#endif
+	} else {
+		inode->i_mapping->a_ops = &erofs_aops;
 #ifdef CONFIG_EROFS_FS_ONDEMAND
-	if (erofs_is_fscache_mode(inode->i_sb))
-		inode->i_mapping->a_ops = &erofs_fscache_access_aops;
+		if (erofs_is_fscache_mode(inode->i_sb))
+			inode->i_mapping->a_ops = &erofs_fscache_access_aops;
 #endif
+#ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
+		if (erofs_is_fileio_mode(EROFS_SB(inode->i_sb)))
+			inode->i_mapping->a_ops = &erofs_fileio_aops;
+#endif
+	}
 
-out_unlock:
-	erofs_put_metabuf(&buf);
 	return err;
 }
 
@@ -334,14 +313,29 @@ int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
 		  unsigned int query_flags)
 {
 	struct inode *const inode = d_inode(path->dentry);
+	struct block_device *bdev = inode->i_sb->s_bdev;
+	bool compressed =
+		erofs_inode_is_data_compressed(EROFS_I(inode)->datalayout);
 
-	if (erofs_inode_is_data_compressed(EROFS_I(inode)->datalayout))
+	if (compressed)
 		stat->attributes |= STATX_ATTR_COMPRESSED;
-
 	stat->attributes |= STATX_ATTR_IMMUTABLE;
 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 				  STATX_ATTR_IMMUTABLE);
 
+	/*
+	 * Return the DIO alignment restrictions if requested.
+	 *
+	 * In EROFS, STATX_DIOALIGN is only supported in bdev-based mode
+	 * and uncompressed inodes, otherwise we report no DIO support.
+	 */
+	if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) {
+		stat->result_mask |= STATX_DIOALIGN;
+		if (bdev && !compressed) {
+			stat->dio_mem_align = bdev_dma_alignment(bdev) + 1;
+			stat->dio_offset_align = bdev_logical_block_size(bdev);
+		}
+	}
 	generic_fillattr(idmap, request_mask, inode, stat);
 	return 0;
 }