[PATCH v2] f2fs: Introduce linear search for dentries

[PATCH v2] f2fs: Introduce linear search for dentries

[Daniel Lee]

This patch addresses an issue where some files in case-insensitive
directories become inaccessible due to changes in how the kernel function,
utf8_casefold(), generates case-folded strings from the commit 5c26d2f1d3f5
("unicode: Don't special case ignorable code points").

F2FS uses these case-folded names to calculate hash values for locating
dentries and stores them on disk. Since utf8_casefold() can produce
different output across kernel versions, stored hash values and newly
calculated hash values may differ. This results in affected files no
longer being found via the hash-based lookup.

To resolve this, the patch introduces a linear search fallback.
If the initial hash-based search fails, F2FS will sequentially scan the
directory entries.

Fixes: 5c26d2f1d3f5 ("unicode: Don't special case ignorable code points")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=219586
Signed-off-by: Daniel Lee <chullee@google.com>
---
v2:
- Only update chash if use_hash is true
 fs/f2fs/dir.c    | 40 +++++++++++++++++++++++++++++-----------
 fs/f2fs/f2fs.h   |  6 ++++--
 fs/f2fs/inline.c |  5 +++--
 3 files changed, 36 insertions(+), 15 deletions(-)

diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index 47a5c806cf16..3e8e5ddf9dbc 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -175,7 +175,8 @@ static unsigned long dir_block_index(unsigned int level,
 static struct f2fs_dir_entry *find_in_block(struct inode *dir,
 				struct page *dentry_page,
 				const struct f2fs_filename *fname,
-				int *max_slots)
+				int *max_slots,
+				bool use_hash)
 {
 	struct f2fs_dentry_block *dentry_blk;
 	struct f2fs_dentry_ptr d;
@@ -183,7 +184,7 @@ static struct f2fs_dir_entry *find_in_block(struct inode *dir,
 	dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
 
 	make_dentry_ptr_block(dir, &d, dentry_blk);
-	return f2fs_find_target_dentry(&d, fname, max_slots);
+	return f2fs_find_target_dentry(&d, fname, max_slots, use_hash);
 }
 
 static inline int f2fs_match_name(const struct inode *dir,
@@ -208,7 +209,8 @@ static inline int f2fs_match_name(const struct inode *dir,
 }
 
 struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
-			const struct f2fs_filename *fname, int *max_slots)
+			const struct f2fs_filename *fname, int *max_slots,
+			bool use_hash)
 {
 	struct f2fs_dir_entry *de;
 	unsigned long bit_pos = 0;
@@ -231,7 +233,7 @@ struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
 			continue;
 		}
 
-		if (de->hash_code == fname->hash) {
+		if (!use_hash || de->hash_code == fname->hash) {
 			res = f2fs_match_name(d->inode, fname,
 					      d->filename[bit_pos],
 					      le16_to_cpu(de->name_len));
@@ -258,11 +260,12 @@ struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
 static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 					unsigned int level,
 					const struct f2fs_filename *fname,
-					struct page **res_page)
+					struct page **res_page,
+					bool use_hash)
 {
 	int s = GET_DENTRY_SLOTS(fname->disk_name.len);
 	unsigned int nbucket, nblock;
-	unsigned int bidx, end_block;
+	unsigned int bidx, end_block, bucket_no;
 	struct page *dentry_page;
 	struct f2fs_dir_entry *de = NULL;
 	pgoff_t next_pgofs;
@@ -272,8 +275,11 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
 	nblock = bucket_blocks(level);
 
+	bucket_no = use_hash ? le32_to_cpu(fname->hash) % nbucket : 0;
+
+start_find_bucket:
 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
-			       le32_to_cpu(fname->hash) % nbucket);
+			       bucket_no);
 	end_block = bidx + nblock;
 
 	while (bidx < end_block) {
@@ -290,7 +296,7 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 			}
 		}
 
-		de = find_in_block(dir, dentry_page, fname, &max_slots);
+		de = find_in_block(dir, dentry_page, fname, &max_slots, use_hash);
 		if (IS_ERR(de)) {
 			*res_page = ERR_CAST(de);
 			de = NULL;
@@ -307,7 +313,10 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 		bidx++;
 	}
 
-	if (!de && room && F2FS_I(dir)->chash != fname->hash) {
+	if (!use_hash && !de && ++bucket_no < nbucket)
+		goto start_find_bucket;
+
+	if (use_hash && !de && room && F2FS_I(dir)->chash != fname->hash) {
 		F2FS_I(dir)->chash = fname->hash;
 		F2FS_I(dir)->clevel = level;
 	}
@@ -323,11 +332,13 @@ struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
 	struct f2fs_dir_entry *de = NULL;
 	unsigned int max_depth;
 	unsigned int level;
+	bool use_hash = true;
 
 	*res_page = NULL;
 
+start_find_entry:
 	if (f2fs_has_inline_dentry(dir)) {
-		de = f2fs_find_in_inline_dir(dir, fname, res_page);
+		de = f2fs_find_in_inline_dir(dir, fname, res_page, use_hash);
 		goto out;
 	}
 
@@ -343,11 +354,18 @@ struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
 	}
 
 	for (level = 0; level < max_depth; level++) {
-		de = find_in_level(dir, level, fname, res_page);
+		de = find_in_level(dir, level, fname, res_page, use_hash);
 		if (de || IS_ERR(*res_page))
 			break;
 	}
+
 out:
+#if IS_ENABLED(CONFIG_UNICODE)
+	if (IS_CASEFOLDED(dir) && !de && use_hash) {
+		use_hash = false;
+		goto start_find_entry;
+	}
+#endif
 	/* This is to increase the speed of f2fs_create */
 	if (!de)
 		F2FS_I(dir)->task = current;
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index f523dd302bf6..1afebb9c4061 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -3588,7 +3588,8 @@ int f2fs_prepare_lookup(struct inode *dir, struct dentry *dentry,
 			struct f2fs_filename *fname);
 void f2fs_free_filename(struct f2fs_filename *fname);
 struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
-			const struct f2fs_filename *fname, int *max_slots);
+			const struct f2fs_filename *fname, int *max_slots,
+			bool use_hash);
 int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
 			unsigned int start_pos, struct fscrypt_str *fstr);
 void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
@@ -4224,7 +4225,8 @@ int f2fs_write_inline_data(struct inode *inode, struct folio *folio);
 int f2fs_recover_inline_data(struct inode *inode, struct page *npage);
 struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
 					const struct f2fs_filename *fname,
-					struct page **res_page);
+					struct page **res_page,
+					bool use_hash);
 int f2fs_make_empty_inline_dir(struct inode *inode, struct inode *parent,
 			struct page *ipage);
 int f2fs_add_inline_entry(struct inode *dir, const struct f2fs_filename *fname,
diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
index cbd2a0d34804..3e3c35d4c98b 100644
--- a/fs/f2fs/inline.c
+++ b/fs/f2fs/inline.c
@@ -352,7 +352,8 @@ int f2fs_recover_inline_data(struct inode *inode, struct page *npage)
 
 struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
 					const struct f2fs_filename *fname,
-					struct page **res_page)
+					struct page **res_page,
+					bool use_hash)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	struct f2fs_dir_entry *de;
@@ -369,7 +370,7 @@ struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
 	inline_dentry = inline_data_addr(dir, ipage);
 
 	make_dentry_ptr_inline(dir, &d, inline_dentry);
-	de = f2fs_find_target_dentry(&d, fname, NULL);
+	de = f2fs_find_target_dentry(&d, fname, NULL, use_hash);
 	unlock_page(ipage);
 	if (IS_ERR(de)) {
 		*res_page = ERR_CAST(de);
-- 
2.47.1.613.gc27f4b7a9f-goog

Re: [PATCH v2] f2fs: Introduce linear search for dentries

[Jaegeuk Kim]

On 12/20, Daniel Lee wrote:
> This patch addresses an issue where some files in case-insensitive
> directories become inaccessible due to changes in how the kernel function,
> utf8_casefold(), generates case-folded strings from the commit 5c26d2f1d3f5
> ("unicode: Don't special case ignorable code points").
> 
> F2FS uses these case-folded names to calculate hash values for locating
> dentries and stores them on disk. Since utf8_casefold() can produce
> different output across kernel versions, stored hash values and newly
> calculated hash values may differ. This results in affected files no
> longer being found via the hash-based lookup.
> 
> To resolve this, the patch introduces a linear search fallback.
> If the initial hash-based search fails, F2FS will sequentially scan the
> directory entries.
> 
> Fixes: 5c26d2f1d3f5 ("unicode: Don't special case ignorable code points")
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=219586
> Signed-off-by: Daniel Lee <chullee@google.com>
> ---
> v2:
> - Only update chash if use_hash is true
>  fs/f2fs/dir.c    | 40 +++++++++++++++++++++++++++++-----------
>  fs/f2fs/f2fs.h   |  6 ++++--
>  fs/f2fs/inline.c |  5 +++--
>  3 files changed, 36 insertions(+), 15 deletions(-)
> 
> diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
> index 47a5c806cf16..3e8e5ddf9dbc 100644
> --- a/fs/f2fs/dir.c
> +++ b/fs/f2fs/dir.c
> @@ -175,7 +175,8 @@ static unsigned long dir_block_index(unsigned int level,
>  static struct f2fs_dir_entry *find_in_block(struct inode *dir,
>  				struct page *dentry_page,
>  				const struct f2fs_filename *fname,
> -				int *max_slots)
> +				int *max_slots,
> +				bool use_hash)
>  {
>  	struct f2fs_dentry_block *dentry_blk;
>  	struct f2fs_dentry_ptr d;
> @@ -183,7 +184,7 @@ static struct f2fs_dir_entry *find_in_block(struct inode *dir,
>  	dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
>  
>  	make_dentry_ptr_block(dir, &d, dentry_blk);
> -	return f2fs_find_target_dentry(&d, fname, max_slots);
> +	return f2fs_find_target_dentry(&d, fname, max_slots, use_hash);
>  }
>  
>  static inline int f2fs_match_name(const struct inode *dir,
> @@ -208,7 +209,8 @@ static inline int f2fs_match_name(const struct inode *dir,
>  }
>  
>  struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
> -			const struct f2fs_filename *fname, int *max_slots)
> +			const struct f2fs_filename *fname, int *max_slots,
> +			bool use_hash)
>  {
>  	struct f2fs_dir_entry *de;
>  	unsigned long bit_pos = 0;
> @@ -231,7 +233,7 @@ struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
>  			continue;
>  		}
>  
> -		if (de->hash_code == fname->hash) {
> +		if (!use_hash || de->hash_code == fname->hash) {
>  			res = f2fs_match_name(d->inode, fname,
>  					      d->filename[bit_pos],
>  					      le16_to_cpu(de->name_len));
> @@ -258,11 +260,12 @@ struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
>  static struct f2fs_dir_entry *find_in_level(struct inode *dir,
>  					unsigned int level,
>  					const struct f2fs_filename *fname,
> -					struct page **res_page)
> +					struct page **res_page,
> +					bool use_hash)
>  {
>  	int s = GET_DENTRY_SLOTS(fname->disk_name.len);
>  	unsigned int nbucket, nblock;
> -	unsigned int bidx, end_block;
> +	unsigned int bidx, end_block, bucket_no;
>  	struct page *dentry_page;
>  	struct f2fs_dir_entry *de = NULL;
>  	pgoff_t next_pgofs;
> @@ -272,8 +275,11 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
>  	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
>  	nblock = bucket_blocks(level);
>  
> +	bucket_no = use_hash ? le32_to_cpu(fname->hash) % nbucket : 0;
> +
> +start_find_bucket:
>  	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
> -			       le32_to_cpu(fname->hash) % nbucket);
> +			       bucket_no);
>  	end_block = bidx + nblock;
>  
>  	while (bidx < end_block) {
> @@ -290,7 +296,7 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
>  			}
>  		}
>  
> -		de = find_in_block(dir, dentry_page, fname, &max_slots);
> +		de = find_in_block(dir, dentry_page, fname, &max_slots, use_hash);
>  		if (IS_ERR(de)) {
>  			*res_page = ERR_CAST(de);
>  			de = NULL;
> @@ -307,7 +313,10 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
>  		bidx++;
>  	}
>  
> -	if (!de && room && F2FS_I(dir)->chash != fname->hash) {
> +	if (!use_hash && !de && ++bucket_no < nbucket)
> +		goto start_find_bucket;
> +
> +	if (use_hash && !de && room && F2FS_I(dir)->chash != fname->hash) {
>  		F2FS_I(dir)->chash = fname->hash;
>  		F2FS_I(dir)->clevel = level;
>  	}
> @@ -323,11 +332,13 @@ struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
>  	struct f2fs_dir_entry *de = NULL;
>  	unsigned int max_depth;
>  	unsigned int level;
> +	bool use_hash = true;
>  
>  	*res_page = NULL;
>  
> +start_find_entry:

This label is only used by #if IS_ENABLED(CONFIG_UNICODE).

>  	if (f2fs_has_inline_dentry(dir)) {
> -		de = f2fs_find_in_inline_dir(dir, fname, res_page);
> +		de = f2fs_find_in_inline_dir(dir, fname, res_page, use_hash);
>  		goto out;
>  	}
>  
> @@ -343,11 +354,18 @@ struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
>  	}
>  
>  	for (level = 0; level < max_depth; level++) {
> -		de = find_in_level(dir, level, fname, res_page);
> +		de = find_in_level(dir, level, fname, res_page, use_hash);
>  		if (de || IS_ERR(*res_page))
>  			break;
>  	}
> +
>  out:
> +#if IS_ENABLED(CONFIG_UNICODE)
> +	if (IS_CASEFOLDED(dir) && !de && use_hash) {
> +		use_hash = false;
> +		goto start_find_entry;
> +	}
> +#endif
>  	/* This is to increase the speed of f2fs_create */
>  	if (!de)
>  		F2FS_I(dir)->task = current;
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index f523dd302bf6..1afebb9c4061 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -3588,7 +3588,8 @@ int f2fs_prepare_lookup(struct inode *dir, struct dentry *dentry,
>  			struct f2fs_filename *fname);
>  void f2fs_free_filename(struct f2fs_filename *fname);
>  struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
> -			const struct f2fs_filename *fname, int *max_slots);
> +			const struct f2fs_filename *fname, int *max_slots,
> +			bool use_hash);
>  int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
>  			unsigned int start_pos, struct fscrypt_str *fstr);
>  void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
> @@ -4224,7 +4225,8 @@ int f2fs_write_inline_data(struct inode *inode, struct folio *folio);
>  int f2fs_recover_inline_data(struct inode *inode, struct page *npage);
>  struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
>  					const struct f2fs_filename *fname,
> -					struct page **res_page);
> +					struct page **res_page,
> +					bool use_hash);
>  int f2fs_make_empty_inline_dir(struct inode *inode, struct inode *parent,
>  			struct page *ipage);
>  int f2fs_add_inline_entry(struct inode *dir, const struct f2fs_filename *fname,
> diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
> index cbd2a0d34804..3e3c35d4c98b 100644
> --- a/fs/f2fs/inline.c
> +++ b/fs/f2fs/inline.c
> @@ -352,7 +352,8 @@ int f2fs_recover_inline_data(struct inode *inode, struct page *npage)
>  
>  struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
>  					const struct f2fs_filename *fname,
> -					struct page **res_page)
> +					struct page **res_page,
> +					bool use_hash)
>  {
>  	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
>  	struct f2fs_dir_entry *de;
> @@ -369,7 +370,7 @@ struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
>  	inline_dentry = inline_data_addr(dir, ipage);
>  
>  	make_dentry_ptr_inline(dir, &d, inline_dentry);
> -	de = f2fs_find_target_dentry(&d, fname, NULL);
> +	de = f2fs_find_target_dentry(&d, fname, NULL, use_hash);
>  	unlock_page(ipage);
>  	if (IS_ERR(de)) {
>  		*res_page = ERR_CAST(de);
> -- 
> 2.47.1.613.gc27f4b7a9f-goog

Re: [PATCH v2] f2fs: Introduce linear search for dentries

[Christoph Hellwig]

On Fri, Dec 20, 2024 at 09:21:36AM -0800, Daniel Lee wrote:
> This patch addresses an issue where some files in case-insensitive
> directories become inaccessible due to changes in how the kernel function,
> utf8_casefold(), generates case-folded strings from the commit 5c26d2f1d3f5
> ("unicode: Don't special case ignorable code points").

That also breaks all other direct hashed and needs to be fixed.

> 
> F2FS uses these case-folded names to calculate hash values for locating
> dentries and stores them on disk. Since utf8_casefold() can produce
> different output across kernel versions, stored hash values and newly
> calculated hash values may differ. This results in affected files no
> longer being found via the hash-based lookup.
> 
> To resolve this, the patch introduces a linear search fallback.
> If the initial hash-based search fails, F2FS will sequentially scan the
> directory entries.

That means you have really bad worst case behavior for negative lookups.
I don' think this is in any way a good idea.

Re: [PATCH v2] f2fs: Introduce linear search for dentries

[Jaegeuk Kim]

On 12/23, Christoph Hellwig wrote:
> On Fri, Dec 20, 2024 at 09:21:36AM -0800, Daniel Lee wrote:
> > This patch addresses an issue where some files in case-insensitive
> > directories become inaccessible due to changes in how the kernel function,
> > utf8_casefold(), generates case-folded strings from the commit 5c26d2f1d3f5
> > ("unicode: Don't special case ignorable code points").
> 
> That also breaks all other direct hashed and needs to be fixed.
> 
> > 
> > F2FS uses these case-folded names to calculate hash values for locating
> > dentries and stores them on disk. Since utf8_casefold() can produce
> > different output across kernel versions, stored hash values and newly
> > calculated hash values may differ. This results in affected files no
> > longer being found via the hash-based lookup.
> > 
> > To resolve this, the patch introduces a linear search fallback.
> > If the initial hash-based search fails, F2FS will sequentially scan the
> > directory entries.
> 
> That means you have really bad worst case behavior for negative lookups.
> I don' think this is in any way a good idea.

No surprise. Please check:
https://lore.kernel.org/lkml/Z1mzu4Eg6CPURra3@google.com/

Re: [PATCH v2] f2fs: Introduce linear search for dentries

[Christoph Hellwig]

On Mon, Dec 23, 2024 at 04:57:35PM +0000, Jaegeuk Kim wrote:
> > That means you have really bad worst case behavior for negative lookups.
> > I don' think this is in any way a good idea.
> 
> No surprise. Please check:
> https://lore.kernel.org/lkml/Z1mzu4Eg6CPURra3@google.com/

Well, I think Linus finally gave in to revert the offending commit
after the usual case insensitivity rants?

Re: [PATCH v2] f2fs: Introduce linear search for dentries

[Jaegeuk Kim]

On 01/03, Christoph Hellwig wrote:
> On Mon, Dec 23, 2024 at 04:57:35PM +0000, Jaegeuk Kim wrote:
> > > That means you have really bad worst case behavior for negative lookups.
> > > I don' think this is in any way a good idea.
> > 
> > No surprise. Please check:
> > https://lore.kernel.org/lkml/Z1mzu4Eg6CPURra3@google.com/
> 
> Well, I think Linus finally gave in to revert the offending commit
> after the usual case insensitivity rants?

In any ways, we need a way to give the files back for customers, which is only
doable by the linear search, whether it sucks or not. :(