Skip to contentSkip to conversation

Prep patches for porting overlayfs to new mount api

PATCH In review Amir Goldstein 3 participants 16 messages Jun 17, 2023 – Jun 20, 2023

Subsystems (2)
lore
Share & tools
lore

Amir Goldstein

author maintainer
#0 lore raw

v2 cover Prep patches for porting overlayfs to new mount api

Miklos,

Following some more cleanup patches that make Christian's new mount api patches smaller and easier to review.

I had rebased Christain's patches over these cleanups and pushed the result to github branch fs-overlayfs-mount_api [1].

The v1 prep patches had a bug with xino option parsing that resulted in some tests being skipped (not failing) and I had only noticed the skipped test after posting v1.

The v2 prep patches + new mount api patches have passed all the tests with no new tests skipped.

In addition to running the tests with the default kernel config, I also ran the tests with the following non-default configs (individually):

1) CONFIG_OVERLAY_FS_REDIRECT_DIR=y
2) CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW=n
3) CONFIG_OVERLAY_FS_XINO_AUTO=y

Thanks.
Amir.

Changes since v1:
- Fix xino opt name table vs. enum order
- Add cleanup patch for xino
- Add cleanup of share_whiteout state
- Add cleanup patch for ovl_get_root()

[1] https://github.com/amir73il/linux/commits/fs-overlayfs-mount_api

Amir Goldstein (5):
ovl: negate the ofs->share_whiteout boolean
ovl: clarify ovl_get_root() semantics
ovl: pass ovl_fs to xino helpers
ovl: store enum redirect_mode in config instead of a string
ovl: factor out ovl_parse_options() helper

 Documentation/filesystems/overlayfs.rst |   9 +-
 fs/overlayfs/dir.c                      |   6 +-
 fs/overlayfs/inode.c                    |  18 +-
 fs/overlayfs/namei.c                    |   6 +-
 fs/overlayfs/overlayfs.h                |  63 ++--
 fs/overlayfs/ovl_entry.h                |   8 +-
 fs/overlayfs/readdir.c                  |  19 +-
 fs/overlayfs/super.c                    | 364 +++++++++++++-----------
 fs/overlayfs/util.c                     |   7 -
 9 files changed, 274 insertions(+), 226 deletions(-)
Signature · 1 line
2.34.1
author maintainer
#1 lore raw

PATCH v2 1/5 ovl: negate the ofs->share_whiteout boolean

The default common case is that whiteout sharing is enabled. Change to storing the negated no_shared_whiteout state, so we will not need to initialize it.

This is the first step towards removing all config and feature initializations out of ovl_fill_super().

---
 fs/overlayfs/dir.c       | 4 ++--
 fs/overlayfs/ovl_entry.h | 4 ++--
 fs/overlayfs/super.c     | 3 ---
 3 files changed, 4 insertions(+), 7 deletions(-)
fs/overlayfs/dir.c +2 -2
index 92bdcedfaaec..0da45727099b 100644
@@ -83,7 +83,7 @@ static struct dentry *ovl_whiteout(struct ovl_fs *ofs)
 		ofs->whiteout = whiteout;
 	}
 
-	if (ofs->share_whiteout) {
+	if (!ofs->no_shared_whiteout) {
 		whiteout = ovl_lookup_temp(ofs, workdir);
 		if (IS_ERR(whiteout))
 			goto out;
@@ -95,7 +95,7 @@ static struct dentry *ovl_whiteout(struct ovl_fs *ofs)
 		if (err != -EMLINK) {
 			pr_warn("Failed to link whiteout - disabling whiteout inode sharing(nlink=%u, err=%i)\n",
 				ofs->whiteout->d_inode->i_nlink, err);
-			ofs->share_whiteout = false;
+			ofs->no_shared_whiteout = true;
 		}
 		dput(whiteout);
 	}
fs/overlayfs/ovl_entry.h +2 -2
index e5207c4bf5b8..40232b056be8 100644
@@ -86,7 +86,6 @@ struct ovl_fs {
 	/* Did we take the inuse lock? */
 	bool upperdir_locked;
 	bool workdir_locked;
-	bool share_whiteout;
 	/* Traps in ovl inode cache */
 	struct inode *workbasedir_trap;
 	struct inode *workdir_trap;
@@ -95,8 +94,9 @@ struct ovl_fs {
 	int xino_mode;
 	/* For allocation of non-persistent inode numbers */
 	atomic_long_t last_ino;
-	/* Whiteout dentry cache */
+	/* Shared whiteout cache */
 	struct dentry *whiteout;
+	bool no_shared_whiteout;
 	/* r/o snapshot of upperdir sb's only taken on volatile mounts */
 	errseq_t errseq;
 };
fs/overlayfs/super.c +0 -3
index 14a2ebdc8126..ee9adb413d0e 100644
@@ -1954,9 +1954,6 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 	if (!cred)
 		goto out_err;
 
-	/* Is there a reason anyone would want not to share whiteouts? */
-	ofs->share_whiteout = true;
-
 	ofs->config.index = ovl_index_def;
 	ofs->config.uuid = true;
 	ofs->config.nfs_export = ovl_nfs_export_def;
Signature · 1 line
2.34.1
Reviewed-by on v2 1/5
#2 lore raw

Re: [PATCH v2 1/5] ovl: negate the ofs->share_whiteout boolean

Quoted 9 lines · Amir Goldstein

The default common case is that whiteout sharing is enabled. Change to storing the negated no_shared_whiteout state, so we will not need to initialize it.

This is the first step towards removing all config and feature initializations out of ovl_fill_super().

---

Looks good to me,

author maintainer
#3 lore raw

PATCH v2 2/5 ovl: clarify ovl_get_root() semantics

In reply to Amir Goldstein

Change the semantics to take a reference on upperdentry instead of transferrig the reference.

This is needed for upcoming port to new mount api.

---
 fs/overlayfs/super.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
fs/overlayfs/super.c +3 -1
index ee9adb413d0e..280f2aa2f356 100644
@@ -1922,6 +1922,8 @@ static struct dentry *ovl_get_root(struct super_block *sb,
 	ovl_set_upperdata(d_inode(root));
 	ovl_inode_init(d_inode(root), &oip, ino, fsid);
 	ovl_dentry_init_flags(root, upperdentry, oe, DCACHE_OP_WEAK_REVALIDATE);
+	/* root keeps a reference of upperdentry */
+	dget(upperdentry);
 
 	return root;
 }
@@ -2100,7 +2102,7 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
 	if (!root_dentry)
 		goto out_free_oe;
 
-	mntput(upperpath.mnt);
+	path_put(&upperpath);
 	kfree(splitlower);
 
 	sb->s_root = root_dentry;
Signature · 1 line
2.34.1
author maintainer
#5 lore raw

PATCH v2 3/5 ovl: pass ovl_fs to xino helpers

In reply to Amir Goldstein

Internal ovl methods should use ovl_fs and not sb as much as possible.

Use a constant_table to translate from enum xino mode to string in preperation for new mount api option parsing.

---
 fs/overlayfs/inode.c     | 18 ++++++++++--------
 fs/overlayfs/overlayfs.h | 16 ++++++++--------
 fs/overlayfs/readdir.c   | 19 +++++++++++--------
 fs/overlayfs/super.c     | 21 ++++++++++++++-------
 4 files changed, 43 insertions(+), 31 deletions(-)
fs/overlayfs/inode.c +10 -8
index bf47a0a94d1e..a63e57447be9 100644
@@ -97,8 +97,9 @@ int ovl_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
 
 static void ovl_map_dev_ino(struct dentry *dentry, struct kstat *stat, int fsid)
 {
-	bool samefs = ovl_same_fs(dentry->d_sb);
-	unsigned int xinobits = ovl_xino_bits(dentry->d_sb);
+	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
+	bool samefs = ovl_same_fs(ofs);
+	unsigned int xinobits = ovl_xino_bits(ofs);
 	unsigned int xinoshift = 64 - xinobits;
 
 	if (samefs) {
@@ -123,7 +124,7 @@ static void ovl_map_dev_ino(struct dentry *dentry, struct kstat *stat, int fsid)
 			stat->ino |= ((u64)fsid) << (xinoshift + 1);
 			stat->dev = dentry->d_sb->s_dev;
 			return;
-		} else if (ovl_xino_warn(dentry->d_sb)) {
+		} else if (ovl_xino_warn(ofs)) {
 			pr_warn_ratelimited("inode number too big (%pd2, ino=%llu, xinobits=%d)\n",
 					    dentry, stat->ino, xinobits);
 		}
@@ -149,7 +150,7 @@ static void ovl_map_dev_ino(struct dentry *dentry, struct kstat *stat, int fsid)
 		 * is unique per underlying fs, so we use the unique anonymous
 		 * bdev assigned to the underlying fs.
 		 */
-		stat->dev = OVL_FS(dentry->d_sb)->fs[fsid].pseudo_dev;
+		stat->dev = ofs->fs[fsid].pseudo_dev;
 	}
 }
 
@@ -186,7 +187,7 @@ int ovl_getattr(struct mnt_idmap *idmap, const struct path *path,
 	 * If lower filesystem supports NFS file handles, this also guaranties
 	 * persistent st_ino across mount cycle.
 	 */
-	if (!is_dir || ovl_same_dev(dentry->d_sb)) {
+	if (!is_dir || ovl_same_dev(OVL_FS(dentry->d_sb))) {
 		if (!OVL_TYPE_UPPER(type)) {
 			fsid = ovl_layer_lower(dentry)->fsid;
 		} else if (OVL_TYPE_ORIGIN(type)) {
@@ -961,7 +962,7 @@ static inline void ovl_lockdep_annotate_inode_mutex_key(struct inode *inode)
 
 static void ovl_next_ino(struct inode *inode)
 {
-	struct ovl_fs *ofs = inode->i_sb->s_fs_info;
+	struct ovl_fs *ofs = OVL_FS(inode->i_sb);
 
 	inode->i_ino = atomic_long_inc_return(&ofs->last_ino);
 	if (unlikely(!inode->i_ino))
@@ -970,7 +971,8 @@ static void ovl_next_ino(struct inode *inode)
 
 static void ovl_map_ino(struct inode *inode, unsigned long ino, int fsid)
 {
-	int xinobits = ovl_xino_bits(inode->i_sb);
+	struct ovl_fs *ofs = OVL_FS(inode->i_sb);
+	int xinobits = ovl_xino_bits(ofs);
 	unsigned int xinoshift = 64 - xinobits;
 
 	/*
@@ -981,7 +983,7 @@ static void ovl_map_ino(struct inode *inode, unsigned long ino, int fsid)
 	 * with d_ino also causes nfsd readdirplus to fail.
 	 */
 	inode->i_ino = ino;
-	if (ovl_same_fs(inode->i_sb)) {
+	if (ovl_same_fs(ofs)) {
 		return;
 	} else if (xinobits && likely(!(ino >> xinoshift))) {
 		inode->i_ino |= (unsigned long)fsid << (xinoshift + 1);
fs/overlayfs/overlayfs.h +8 -8
index fcac4e2c56ab..05e9acfe1590 100644
@@ -494,26 +494,26 @@ static inline bool ovl_is_impuredir(struct super_block *sb,
  * d_ino consistent with st_ino.
  * With xino=on, we do the same effort but we warn if we failed.
  */
-static inline bool ovl_xino_warn(struct super_block *sb)
+static inline bool ovl_xino_warn(struct ovl_fs *ofs)
 {
-	return OVL_FS(sb)->config.xino == OVL_XINO_ON;
+	return ofs->config.xino == OVL_XINO_ON;
 }
 
 /* All layers on same fs? */
-static inline bool ovl_same_fs(struct super_block *sb)
+static inline bool ovl_same_fs(struct ovl_fs *ofs)
 {
-	return OVL_FS(sb)->xino_mode == 0;
+	return ofs->xino_mode == 0;
 }
 
 /* All overlay inodes have same st_dev? */
-static inline bool ovl_same_dev(struct super_block *sb)
+static inline bool ovl_same_dev(struct ovl_fs *ofs)
 {
-	return OVL_FS(sb)->xino_mode >= 0;
+	return ofs->xino_mode >= 0;
 }
 
-static inline unsigned int ovl_xino_bits(struct super_block *sb)
+static inline unsigned int ovl_xino_bits(struct ovl_fs *ofs)
 {
-	return ovl_same_dev(sb) ? OVL_FS(sb)->xino_mode : 0;
+	return ovl_same_dev(ofs) ? ofs->xino_mode : 0;
 }
 
 static inline void ovl_inode_lock(struct inode *inode)
fs/overlayfs/readdir.c +11 -8
index b6952b21a7ee..ee5c4736480f 100644
@@ -118,7 +118,7 @@ static bool ovl_calc_d_ino(struct ovl_readdir_data *rdd,
 		return false;
 
 	/* Always recalc d_ino when remapping lower inode numbers */
-	if (ovl_xino_bits(rdd->dentry->d_sb))
+	if (ovl_xino_bits(OVL_FS(rdd->dentry->d_sb)))
 		return true;
 
 	/* Always recalc d_ino for parent */
@@ -460,13 +460,14 @@ static int ovl_cache_update_ino(const struct path *path, struct ovl_cache_entry
 
 {
 	struct dentry *dir = path->dentry;
+	struct ovl_fs *ofs = OVL_FS(dir->d_sb);
 	struct dentry *this = NULL;
 	enum ovl_path_type type;
 	u64 ino = p->real_ino;
-	int xinobits = ovl_xino_bits(dir->d_sb);
+	int xinobits = ovl_xino_bits(ofs);
 	int err = 0;
 
-	if (!ovl_same_dev(dir->d_sb))
+	if (!ovl_same_dev(ofs))
 		goto out;
 
 	if (p->name[0] == '.') {
@@ -515,7 +516,7 @@ static int ovl_cache_update_ino(const struct path *path, struct ovl_cache_entry
 		ino = ovl_remap_lower_ino(ino, xinobits,
 					  ovl_layer_lower(this)->fsid,
 					  p->name, p->len,
-					  ovl_xino_warn(dir->d_sb));
+					  ovl_xino_warn(ofs));
 	}
 
 out:
@@ -694,12 +695,13 @@ static int ovl_iterate_real(struct file *file, struct dir_context *ctx)
 	int err;
 	struct ovl_dir_file *od = file->private_data;
 	struct dentry *dir = file->f_path.dentry;
+	struct ovl_fs *ofs = OVL_FS(dir->d_sb);
 	const struct ovl_layer *lower_layer = ovl_layer_lower(dir);
 	struct ovl_readdir_translate rdt = {
 		.ctx.actor = ovl_fill_real,
 		.orig_ctx = ctx,
-		.xinobits = ovl_xino_bits(dir->d_sb),
-		.xinowarn = ovl_xino_warn(dir->d_sb),
+		.xinobits = ovl_xino_bits(ofs),
+		.xinowarn = ovl_xino_warn(ofs),
 	};
 
 	if (rdt.xinobits && lower_layer)
@@ -735,6 +737,7 @@ static int ovl_iterate(struct file *file, struct dir_context *ctx)
 {
 	struct ovl_dir_file *od = file->private_data;
 	struct dentry *dentry = file->f_path.dentry;
+	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
 	struct ovl_cache_entry *p;
 	const struct cred *old_cred;
 	int err;
@@ -749,8 +752,8 @@ static int ovl_iterate(struct file *file, struct dir_context *ctx)
 		 * dir is impure then need to adjust d_ino for copied up
 		 * entries.
 		 */
-		if (ovl_xino_bits(dentry->d_sb) ||
-		    (ovl_same_fs(dentry->d_sb) &&
+		if (ovl_xino_bits(ofs) ||
+		    (ovl_same_fs(ofs) &&
 		     (ovl_is_impure_dir(file) ||
 		      OVL_TYPE_MERGE(ovl_path_type(dentry->d_parent))))) {
 			err = ovl_iterate_real(file, ctx);
fs/overlayfs/super.c +14 -7
index 280f2aa2f356..5bcb26528408 100644
@@ -16,6 +16,7 @@
 #include <linux/posix_acl_xattr.h>
 #include <linux/exportfs.h>
 #include <linux/file.h>
+#include <linux/fs_parser.h>
 #include "overlayfs.h"
 
 MODULE_AUTHOR("Miklos Szeredi <[email protected]>");
@@ -334,12 +335,18 @@ static const char *ovl_redirect_mode_def(void)
 	return ovl_redirect_dir_def ? "on" : "off";
 }
 
-static const char * const ovl_xino_str[] = {
-	"off",
-	"auto",
-	"on",
+static const struct constant_table ovl_parameter_xino[] = {
+	{ "off",	OVL_XINO_OFF  },
+	{ "auto",	OVL_XINO_AUTO },
+	{ "on",		OVL_XINO_ON   },
+	{}
 };
 
+static const char *ovl_xino_mode(struct ovl_config *config)
+{
+	return ovl_parameter_xino[config->xino].name;
+}
+
 static inline int ovl_xino_def(void)
 {
 	return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF;
@@ -374,8 +381,8 @@ static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
 	if (ofs->config.nfs_export != ovl_nfs_export_def)
 		seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ?
 						"on" : "off");
-	if (ofs->config.xino != ovl_xino_def() && !ovl_same_fs(sb))
-		seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]);
+	if (ofs->config.xino != ovl_xino_def() && !ovl_same_fs(ofs))
+		seq_printf(m, ",xino=%s", ovl_xino_mode(&ofs->config));
 	if (ofs->config.metacopy != ovl_metacopy_def)
 		seq_printf(m, ",metacopy=%s",
 			   ofs->config.metacopy ? "on" : "off");
@@ -1566,7 +1573,7 @@ static int ovl_get_fsid(struct ovl_fs *ofs, const struct path *path)
 			pr_warn("%s uuid detected in lower fs '%pd2', falling back to xino=%s,index=off,nfs_export=off.\n",
 				uuid_is_null(&sb->s_uuid) ? "null" :
 							    "conflicting",
-				path->dentry, ovl_xino_str[ofs->config.xino]);
+				path->dentry, ovl_xino_mode(&ofs->config));
 		}
 	}
 
Signature · 1 line
2.34.1
author maintainer
#7 lore raw

PATCH v2 4/5 ovl: store enum redirect_mode in config instead of a string

In reply to Amir Goldstein

Do all the logic to set the mode during mount options parsing and do not keep the option string around.

Use a constant_table to translate from enum redirect mode to string in preperation for new mount api option parsing.

The mount option "off" is translated to either "follow" or "nofollow", depending on the "redirect_always_follow" build/module config, so in effect, there are only three possible redirect modes.

This results in a minor change to the string that is displayed in show_options() - when redirect_dir is enabled by default and the user mounts with the option "redirect_dir=off", instead of displaying the mode "redirect_dir=off" in show_options(), the displayed mode will be either "redirect_dir=follow" or "redirect_dir=nofollow", depending on the value of "redirect_always_follow" build/module config.

The displayed mode reflects the effective mode, so mounting overlayfs again with the dispalyed redirect_dir option will result with the same effective and displayed mode.

---
 Documentation/filesystems/overlayfs.rst |   9 +-
 fs/overlayfs/dir.c                      |   2 +-
 fs/overlayfs/namei.c                    |   6 +-
 fs/overlayfs/overlayfs.h                |  39 ++++---
 fs/overlayfs/ovl_entry.h                |   4 +-
 fs/overlayfs/super.c                    | 129 +++++++++++++-----------
 fs/overlayfs/util.c                     |   7 --
 7 files changed, 107 insertions(+), 89 deletions(-)
Documentation/filesystems/overlayfs.rst +4 -5
index 4f36b8919f7c..eb7d2c88ddec 100644
@@ -231,12 +231,11 @@ Mount options:
     Redirects are enabled.
 - "redirect_dir=follow":
     Redirects are not created, but followed.
-- "redirect_dir=off":
-    Redirects are not created and only followed if "redirect_always_follow"
-    feature is enabled in the kernel/module config.
 - "redirect_dir=nofollow":
-    Redirects are not created and not followed (equivalent to "redirect_dir=off"
-    if "redirect_always_follow" feature is not enabled).
+    Redirects are not created and not followed.
+- "redirect_dir=off":
+    If "redirect_always_follow" is enabled in the kernel/module config,
+    this "off" traslates to "follow", otherwise it translates to "nofollow".
 
 When the NFS export feature is enabled, every copied up directory is
 indexed by the file handle of the lower inode and a file handle of the
fs/overlayfs/dir.c +1 -1
index 0da45727099b..033fc0458a3d 100644
@@ -952,7 +952,7 @@ static bool ovl_type_merge_or_lower(struct dentry *dentry)
 
 static bool ovl_can_move(struct dentry *dentry)
 {
-	return ovl_redirect_dir(dentry->d_sb) ||
+	return ovl_redirect_dir(OVL_FS(dentry->d_sb)) ||
 		!d_is_dir(dentry) || !ovl_type_merge_or_lower(dentry);
 }
 
fs/overlayfs/namei.c +3 -3
index 292b8a948f1a..288e3c0ee0e6 100644
@@ -961,7 +961,7 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 		.is_dir = false,
 		.opaque = false,
 		.stop = false,
-		.last = ofs->config.redirect_follow ? false : !ovl_numlower(poe),
+		.last = ovl_redirect_follow(ofs) ? false : !ovl_numlower(poe),
 		.redirect = NULL,
 		.metacopy = false,
 	};
@@ -1022,7 +1022,7 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 	for (i = 0; !d.stop && i < ovl_numlower(poe); i++) {
 		struct ovl_path lower = ovl_lowerstack(poe)[i];
 
-		if (!ofs->config.redirect_follow)
+		if (!ovl_redirect_follow(ofs))
 			d.last = i == ovl_numlower(poe) - 1;
 		else if (d.is_dir || !ofs->numdatalayer)
 			d.last = lower.layer->idx == ovl_numlower(roe);
@@ -1102,7 +1102,7 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
 		 * this attack vector when not necessary.
 		 */
 		err = -EPERM;
-		if (d.redirect && !ofs->config.redirect_follow) {
+		if (d.redirect && !ovl_redirect_follow(ofs)) {
 			pr_warn_ratelimited("refusing to follow redirect for (%pd2)\n",
 					    dentry);
 			goto out_put;
fs/overlayfs/overlayfs.h +27 -12
index 05e9acfe1590..80c10228bd64 100644
@@ -57,6 +57,13 @@ enum ovl_entry_flag {
 	OVL_E_CONNECTED,
 };
 
+enum {
+	OVL_REDIRECT_OFF,	/* "off" mode is never used. In effect	*/
+	OVL_REDIRECT_FOLLOW,	/* ...it translates to either "follow"	*/
+	OVL_REDIRECT_NOFOLLOW,	/* ...or "nofollow".			*/
+	OVL_REDIRECT_ON,
+};
+
 enum {
 	OVL_XINO_OFF,
 	OVL_XINO_AUTO,
@@ -352,17 +359,6 @@ static inline bool ovl_open_flags_need_copy_up(int flags)
 	return ((OPEN_FMODE(flags) & FMODE_WRITE) || (flags & O_TRUNC));
 }
 
-static inline bool ovl_allow_offline_changes(struct ovl_fs *ofs)
-{
-	/*
-	 * To avoid regressions in existing setups with overlay lower offline
-	 * changes, we allow lower changes only if none of the new features
-	 * are used.
-	 */
-	return (!ofs->config.index && !ofs->config.metacopy &&
-		!ofs->config.redirect_dir && ofs->config.xino != OVL_XINO_ON);
-}
-
 
 /* util.c */
 int ovl_want_write(struct dentry *dentry);
@@ -421,7 +417,6 @@ bool ovl_dentry_needs_data_copy_up(struct dentry *dentry, int flags);
 bool ovl_dentry_needs_data_copy_up_locked(struct dentry *dentry, int flags);
 bool ovl_has_upperdata(struct inode *inode);
 void ovl_set_upperdata(struct inode *inode);
-bool ovl_redirect_dir(struct super_block *sb);
 const char *ovl_dentry_get_redirect(struct dentry *dentry);
 void ovl_dentry_set_redirect(struct dentry *dentry, const char *redirect);
 void ovl_inode_update(struct inode *inode, struct dentry *upperdentry);
@@ -489,6 +484,16 @@ static inline bool ovl_is_impuredir(struct super_block *sb,
 	return ovl_path_check_dir_xattr(ofs, &upperpath, OVL_XATTR_IMPURE);
 }
 
+static inline bool ovl_redirect_follow(struct ovl_fs *ofs)
+{
+	return ofs->config.redirect_mode != OVL_REDIRECT_NOFOLLOW;
+}
+
+static inline bool ovl_redirect_dir(struct ovl_fs *ofs)
+{
+	return ofs->config.redirect_mode == OVL_REDIRECT_ON;
+}
+
 /*
  * With xino=auto, we do best effort to keep all inodes on same st_dev and
  * d_ino consistent with st_ino.
@@ -499,6 +504,16 @@ static inline bool ovl_xino_warn(struct ovl_fs *ofs)
 	return ofs->config.xino == OVL_XINO_ON;
 }
 
+/*
+ * To avoid regressions in existing setups with overlay lower offline changes,
+ * we allow lower changes only if none of the new features are used.
+ */
+static inline bool ovl_allow_offline_changes(struct ovl_fs *ofs)
+{
+	return (!ofs->config.index && !ofs->config.metacopy &&
+		!ovl_redirect_dir(ofs) && !ovl_xino_warn(ofs));
+}
+
 /* All layers on same fs? */
 static inline bool ovl_same_fs(struct ovl_fs *ofs)
 {
fs/overlayfs/ovl_entry.h +1 -3
index 40232b056be8..b4eb0bd5d0b6 100644
@@ -10,9 +10,7 @@ struct ovl_config {
 	char *upperdir;
 	char *workdir;
 	bool default_permissions;
-	bool redirect_dir;
-	bool redirect_follow;
-	const char *redirect_mode;
+	int redirect_mode;
 	bool index;
 	bool uuid;
 	bool nfs_export;
fs/overlayfs/super.c +71 -58

231 lines not on this page · raw message

fs/overlayfs/util.c +0 -7
index 939e4d586ec2..7ef9e13c404a 100644
@@ -506,13 +506,6 @@ bool ovl_dentry_needs_data_copy_up(struct dentry *dentry, int flags)
 	return !ovl_has_upperdata(d_inode(dentry));
 }
 
-bool ovl_redirect_dir(struct super_block *sb)
-{
-	struct ovl_fs *ofs = sb->s_fs_info;
-
-	return ofs->config.redirect_dir && !ofs->noxattr;
-}
-
 const char *ovl_dentry_get_redirect(struct dentry *dentry)
 {
 	return OVL_I(d_inode(dentry))->redirect;
Signature · 1 line
2.34.1
maintainer
#8 lore raw

Re: [PATCH v2 4/5] ovl: store enum redirect_mode in config instead of a string

fs/overlayfs/super.c static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs, · line 1342

Quoted 9 lines · Amir Goldstein
fs/overlayfs/super.c +3 -2
@@ -1332,10 +1337,17 @@ static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs,
        if (err) {
                pr_warn("failed to set xattr on upper\n");
                ofs->noxattr = true;
-               if (ofs->config.index || ofs->config.metacopy) {
-                       ofs->config.index = false;
+               if (ovl_redirect_follow(ofs)) {
+                       ofs->config.redirect_mode = OVL_REDIRECT_NOFOLLOW;
+                       pr_warn("...falling back to redirect_dir=nofollow.\n");

So if there's no xattr support, then there won't be any redirects to
follow. Is this just asserting the fact?

Should this be a separate patch?

Thanks,
Miklos

maintainer
#9 lore raw

Re: [PATCH v2 4/5] ovl: store enum redirect_mode in config instead of a string

Quoted 17 lines · Miklos Szeredi

fs/overlayfs/super.c static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs, · line 1342

Quoted 9 lines · Amir Goldstein
fs/overlayfs/super.c +3 -2
@@ -1332,10 +1337,17 @@ static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs,
        if (err) {
                pr_warn("failed to set xattr on upper\n");
                ofs->noxattr = true;
-               if (ofs->config.index || ofs->config.metacopy) {
-                       ofs->config.index = false;
+               if (ovl_redirect_follow(ofs)) {
+                       ofs->config.redirect_mode = OVL_REDIRECT_NOFOLLOW;
+                       pr_warn("...falling back to redirect_dir=nofollow.\n");

So if there's no xattr support, then there won't be any redirects to
follow. Is this just asserting the fact?

Quoted 1 line

Should this be a separate patch?

Oh, I see you are changing the ovl_redirect_dir() logic. Ack.

Thanks,
Miklos

Reviewed-by on v2 4/5
#10 lore raw

Re: [PATCH v2 4/5] ovl: store enum redirect_mode in config instead of a string

In reply to Amir Goldstein
Quoted 17 lines · Amir Goldstein

Do all the logic to set the mode during mount options parsing and do not keep the option string around.

Use a constant_table to translate from enum redirect mode to string in preperation for new mount api option parsing.

The mount option "off" is translated to either "follow" or "nofollow", depending on the "redirect_always_follow" build/module config, so in effect, there are only three possible redirect modes.

This results in a minor change to the string that is displayed in show_options() - when redirect_dir is enabled by default and the user mounts with the option "redirect_dir=off", instead of displaying the mode "redirect_dir=off" in show_options(), the displayed mode will be either "redirect_dir=follow" or "redirect_dir=nofollow", depending on the value of "redirect_always_follow" build/module config.

Quoted 6 lines

The displayed mode reflects the effective mode, so mounting overlayfs again with the dispalyed redirect_dir option will result with the same effective and displayed mode.

---

Looks good to me,

author maintainer
#11 lore raw

PATCH v2 5/5 ovl: factor out ovl_parse_options() helper

In reply to Amir Goldstein

For parsing a single mount option.

---
 fs/overlayfs/overlayfs.h |   8 ++
 fs/overlayfs/super.c     | 243 ++++++++++++++++++++-------------------
 2 files changed, 135 insertions(+), 116 deletions(-)
fs/overlayfs/overlayfs.h +8 -0
index 80c10228bd64..30227ccc758d 100644
@@ -70,6 +70,14 @@ enum {
 	OVL_XINO_ON,
 };
 
+/* The set of options that user requested explicitly via mount options */
+struct ovl_opt_set {
+	bool metacopy;
+	bool redirect;
+	bool nfs_export;
+	bool index;
+};
+
 /*
  * The tuple (fh,uuid) is a universal unique identifier for a copy up origin,
  * where:
fs/overlayfs/super.c +127 -116

320 lines not on this page · raw message

Signature · 1 line
2.34.1
#13 lore raw
In reply to Amir Goldstein
Quoted 15 lines · Amir Goldstein

Miklos,

Following some more cleanup patches that make Christian's new mount api patches smaller and easier to review.

I had rebased Christain's patches over these cleanups and pushed the result to github branch fs-overlayfs-mount_api [1].

The v1 prep patches had a bug with xino option parsing that resulted in some tests being skipped (not failing) and I had only noticed the skipped test after posting v1.

The v2 prep patches + new mount api patches have passed all the tests with no new tests skipped.

Quoted 6 lines

In addition to running the tests with the default kernel config, I also ran the tests with the following non-default configs (individually):

1) CONFIG_OVERLAY_FS_REDIRECT_DIR=y
2) CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW=n
3) CONFIG_OVERLAY_FS_XINO_AUTO=y

Thanks for splitting some work into preparatory patches. I'm not sure how worthwhile this actually is given they aren't marked as backports for LTS releases so the overall delta ould still the same between LTSes and mainline but it might make bisection easier.

author maintainer
#14 lore raw
Quoted 20 lines · Christian Brauner
Quoted 18 lines · Amir Goldstein

Miklos,

Following some more cleanup patches that make Christian's new mount api patches smaller and easier to review.

I had rebased Christain's patches over these cleanups and pushed the result to github branch fs-overlayfs-mount_api [1].

The v1 prep patches had a bug with xino option parsing that resulted in some tests being skipped (not failing) and I had only noticed the skipped test after posting v1.

The v2 prep patches + new mount api patches have passed all the tests with no new tests skipped.

In addition to running the tests with the default kernel config, I also ran the tests with the following non-default configs (individually):

Quoted 8 lines
Quoted 3 lines

1) CONFIG_OVERLAY_FS_REDIRECT_DIR=y
2) CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW=n
3) CONFIG_OVERLAY_FS_XINO_AUTO=y

Thanks for splitting some work into preparatory patches. I'm not sure how worthwhile this actually is given they aren't marked as backports for LTS releases so the overall delta ould still the same between LTSes and mainline but it might make bisection easier.

Yeh, bisection, review, all the usual reasons for keep unrelated changes split. I am not usually fanatic about splitting hairs on this, but the mount api porting patch was already a big change that was hard for me to review and it grew all those extra additions like redirect_mode which were good changes, but not related, so I did this to make my own (and others) review of your patches easier.

I am glad if we are all happy with the end result.

Thanks,
Amir.

#15 lore raw
Quoted 30 lines · Amir Goldstein
Quoted 20 lines · Christian Brauner
Quoted 18 lines · Amir Goldstein

Miklos,

Following some more cleanup patches that make Christian's new mount api patches smaller and easier to review.

I had rebased Christain's patches over these cleanups and pushed the result to github branch fs-overlayfs-mount_api [1].

The v1 prep patches had a bug with xino option parsing that resulted in some tests being skipped (not failing) and I had only noticed the skipped test after posting v1.

The v2 prep patches + new mount api patches have passed all the tests with no new tests skipped.

In addition to running the tests with the default kernel config, I also ran the tests with the following non-default configs (individually):

Quoted 8 lines
Quoted 3 lines

1) CONFIG_OVERLAY_FS_REDIRECT_DIR=y
2) CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW=n
3) CONFIG_OVERLAY_FS_XINO_AUTO=y

Thanks for splitting some work into preparatory patches. I'm not sure how worthwhile this actually is given they aren't marked as backports for LTS releases so the overall delta ould still the same between LTSes and mainline but it might make bisection easier.

Quoted 8 lines

Yeh, bisection, review, all the usual reasons for keep unrelated changes split. I am not usually fanatic about splitting hairs on this, but the mount api porting patch was already a big change that was hard for me to review and it grew all those extra additions like redirect_mode which were good changes, but not related, so I did this to make my own (and others) review of your patches easier.

I am glad if we are all happy with the end result.

Yeah, I appreciate the work and wasn't trying to critique it. I just wouldn't have bothered because the mount api port in itself is cumbersome enough. :)