All of lore.kernel.org
 help / color / mirror / Atom feed
From: Gao Xiang via Linux-erofs <linux-erofs@lists.ozlabs.org>
To: linux-erofs@lists.ozlabs.org, Li Guifu <bluce.lee@aliyun.com>
Subject: [PATCH v2] erofs-utils: update i_nlink stat for directories
Date: Sat,  5 Dec 2020 17:16:37 +0800	[thread overview]
Message-ID: <20201205091637.8944-1-hsiangkao@aol.com> (raw)
In-Reply-To: 20201205091637.8944-1-hsiangkao.ref@aol.com

From: Gao Xiang <hsiangkao@aol.com>

Previously, nlink of directories was treated as 1 for simplicity.

Since st_nlink for dirs is actually not well defined, nlink=1 seems
to pacify `find' (even without -noleaf option) and other utilities.
AFAICT, isofs, romfs and cramfs always set it to 1, Overlayfs sets
it to 1 conditionally, btrfs[1], ceph[2] and FUSE client historically
set it to 1.

The convention under unix is that it's # of subdirs including "."
and "..". This patch tries to follow such convention if possible to
optimize `find' performance since it's not quite hard for local fs.

[1] https://lore.kernel.org/r/20100124003336.GP23006@think
[2] https://lore.kernel.org/r/20180521092729.17470-1-lhenriques@suse.com
Signed-off-by: Gao Xiang <hsiangkao@aol.com>
---
v1: https://lore.kernel.org/r/20201205055732.14276-1-hsiangkao@aol.com
changes since v1:
 - update a DBG_BUGON statement suggestted by Guifu.

 lib/inode.c | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)

diff --git a/lib/inode.c b/lib/inode.c
index 618eb284550f..3d634fc92852 100644
--- a/lib/inode.c
+++ b/lib/inode.c
@@ -25,7 +25,7 @@
 struct erofs_sb_info sbi;
 
 #define S_SHIFT                 12
-static unsigned char erofs_type_by_mode[S_IFMT >> S_SHIFT] = {
+static unsigned char erofs_ftype_by_mode[S_IFMT >> S_SHIFT] = {
 	[S_IFREG >> S_SHIFT]  = EROFS_FT_REG_FILE,
 	[S_IFDIR >> S_SHIFT]  = EROFS_FT_DIR,
 	[S_IFCHR >> S_SHIFT]  = EROFS_FT_CHRDEV,
@@ -35,6 +35,11 @@ static unsigned char erofs_type_by_mode[S_IFMT >> S_SHIFT] = {
 	[S_IFLNK >> S_SHIFT]  = EROFS_FT_SYMLINK,
 };
 
+static unsigned char erofs_mode_to_ftype(umode_t mode)
+{
+	return erofs_ftype_by_mode[(mode & S_IFMT) >> S_SHIFT];
+}
+
 #define NR_INODE_HASHTABLE	16384
 
 struct list_head inode_hashtable[NR_INODE_HASHTABLE];
@@ -156,7 +161,7 @@ static int __allocate_inode_bh_data(struct erofs_inode *inode,
 int erofs_prepare_dir_file(struct erofs_inode *dir)
 {
 	struct erofs_dentry *d;
-	unsigned int d_size;
+	unsigned int d_size, i_nlink;
 	int ret;
 
 	/* dot is pointed to the current dir inode */
@@ -169,16 +174,28 @@ int erofs_prepare_dir_file(struct erofs_inode *dir)
 	d->inode = erofs_igrab(dir->i_parent);
 	d->type = EROFS_FT_DIR;
 
-	/* let's calculate dir size */
+	/* let's calculate dir size and update i_nlink */
 	d_size = 0;
+	i_nlink = 0;
 	list_for_each_entry(d, &dir->i_subdirs, d_child) {
 		int len = strlen(d->name) + sizeof(struct erofs_dirent);
 
 		if (d_size % EROFS_BLKSIZ + len > EROFS_BLKSIZ)
 			d_size = round_up(d_size, EROFS_BLKSIZ);
 		d_size += len;
+
+		i_nlink += (d->type == EROFS_FT_DIR);
 	}
 	dir->i_size = d_size;
+	/*
+	 * if there're too many subdirs as compact form, set nlink=1
+	 * rather than upgrade to use extented form instead.
+	 */
+	if (i_nlink > USHRT_MAX &&
+	    dir->inode_isize == sizeof(struct erofs_inode_compact))
+		dir->i_nlink = 1;
+	else
+		dir->i_nlink = i_nlink;
 
 	/* no compression for all dirs */
 	dir->datalayout = EROFS_INODE_FLAT_INLINE;
@@ -957,6 +974,10 @@ struct erofs_inode *erofs_mkfs_build_tree(struct erofs_inode *dir)
 			ret = PTR_ERR(d);
 			goto err_closedir;
 		}
+
+		/* to count i_nlink for directories */
+		d->type = (dp->d_type == DT_DIR ?
+			EROFS_FT_DIR : EROFS_FT_UNKNOWN);
 	}
 
 	if (errno) {
@@ -978,6 +999,7 @@ struct erofs_inode *erofs_mkfs_build_tree(struct erofs_inode *dir)
 
 	list_for_each_entry(d, &dir->i_subdirs, d_child) {
 		char buf[PATH_MAX];
+		unsigned char ftype;
 
 		if (is_dot_dotdot(d->name)) {
 			erofs_d_invalidate(d);
@@ -1000,7 +1022,10 @@ fail:
 			goto err;
 		}
 
-		d->type = erofs_type_by_mode[d->inode->i_mode >> S_SHIFT];
+		ftype = erofs_mode_to_ftype(d->inode->i_mode);
+		DBG_BUGON(ftype == EROFS_FT_DIR && d->type != ftype);
+		d->type = ftype;
+
 		erofs_d_invalidate(d);
 		erofs_info("add file %s/%s (nid %llu, type %d)",
 			   dir->i_srcpath, d->name, (unsigned long long)d->nid,
-- 
2.24.0


       reply	other threads:[~2020-12-05  9:17 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20201205091637.8944-1-hsiangkao.ref@aol.com>
2020-12-05  9:16 ` Gao Xiang via Linux-erofs [this message]
2020-12-05  9:22   ` [PATCH v2] erofs-utils: update i_nlink stat for directories Li GuiFu via Linux-erofs
2020-12-05  9:34     ` Gao Xiang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201205091637.8944-1-hsiangkao@aol.com \
    --to=linux-erofs@lists.ozlabs.org \
    --cc=bluce.lee@aliyun.com \
    --cc=hsiangkao@aol.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.