From mboxrd@z Thu Jan  1 00:00:00 1970
From: Hin-Tak Leung <htl10@users.sourceforge.net>
Subject: Re: [PATCH] hfs: add error checking for hfs_find_init()
Date: Sat, 30 Mar 2013 14:01:03 +0000 (GMT)
Message-ID: <1364652063.24328.YahooMailClassic@web172301.mail.ir2.yahoo.com>
References: <0D203BE4-D965-42FE-ADA7-831070860222@dubeyko.com>
Reply-To: htl10@users.sourceforge.net
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: Al Viro <viro@zeniv.linux.org.uk>,
	Artem Bityutskiy <artem.bityutskiy@linux.intel.com>,
	Christoph Hellwig <hch@lst.de>, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, ldv-project@linuxtesting.org
To: Alexey Khoroshilov <khoroshilov@ispras.ru>,
	Vyacheslav Dubeyko <slava@dubeyko.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <0D203BE4-D965-42FE-ADA7-831070860222@dubeyko.com>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

--- On Sat, 30/3/13, Vyacheslav Dubeyko <slava@dubeyko.com> wrote:

> From: Vyacheslav Dubeyko <slava@dubeyko.com>
> Subject: Re: [PATCH] hfs: add error checking for hfs_find_init()
> To: "Alexey Khoroshilov" <khoroshilov@ispras.ru>
> Cc: "Al Viro" <viro@zeniv.linux.org.uk>, "Artem Bityutskiy" <artem.bi=
tyutskiy@linux.intel.com>, "Christoph Hellwig" <hch@lst.de>, linux-fsde=
vel@vger.kernel.org, linux-kernel@vger.kernel.org, ldv-project@linuxtes=
ting.org, "Hin-Tak Leung" <htl10@users.sourceforge.net>
> Date: Saturday, 30 March, 2013, 11:35
> Hi Alexey,
>=20
> On Mar 30, 2013, at 12:44 AM, Alexey Khoroshilov wrote:
>=20
> > hfs_find_init() may fail with ENOMEM, but there are
> places,
> > where the returned value is not checked. The
> consequences can be
> > very unpleasant, e.g. kfree uninitialized pointer and
> > inappropriate mutex unlocking.
> >=20
> > The patch adds checks for errors in hfs_find_init().
> >=20
>=20
> Thank you for your efforts. I have several remarks. Please,
> see below.

Argh, interesting. I wonder if that is related to how I can get the hfs=
plus driver all confused just running 'du' on one large directory in on=
e of my disks repeatedly. I'll be interested to trying the hfsplus vers=
ion out. Perhaps I would suggest/add a few dprint/printk's so that ther=
e is a sign in dmesg when the error condition is triggered.

Hin-Tak
=20
> > Found by Linux Driver Verification project
> (linuxtesting.org).
> >=20
> > Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
> > ---
> > fs/hfs/catalog.c |=A0=A0=A012 +++++++++---
> > fs/hfs/dir.c=A0 =A0=A0=A0|=A0 =A0 8
> ++++++--
> > fs/hfs/extent.c=A0 |=A0=A0=A048
> +++++++++++++++++++++++++++++++++---------------
> > fs/hfs/hfs_fs.h=A0 |=A0 =A0 2 +-
> > fs/hfs/inode.c=A0=A0=A0|=A0=A0=A011
> +++++++++--
> > fs/hfs/super.c=A0=A0=A0|=A0 =A0 4 +++-
> > 6 files changed, 61 insertions(+), 24 deletions(-)
> >=20
> > diff --git a/fs/hfs/catalog.c b/fs/hfs/catalog.c
> > index 424b033..9569b39 100644
> > --- a/fs/hfs/catalog.c
> > +++ b/fs/hfs/catalog.c
> > @@ -92,7 +92,9 @@ int hfs_cat_create(u32 cnid, struct
> inode *dir, struct qstr *str, struct inode *
> > =A0=A0=A0 =A0=A0=A0 return -ENOSPC;
> >=20
> > =A0=A0=A0 sb =3D dir->i_sb;
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 err =3D
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 if (err)
> > +=A0=A0=A0 =A0=A0=A0 return err;
> >=20
> > =A0=A0=A0 hfs_cat_build_key(sb, fd.search_key,
> cnid, NULL);
> > =A0=A0=A0 entry_size =3D
> hfs_cat_build_thread(sb, &entry,
> S_ISDIR(inode->i_mode) ?
> > @@ -214,7 +216,9 @@ int hfs_cat_delete(u32 cnid, struct
> inode *dir, struct qstr *str)
> >=20
> > =A0=A0=A0 dprint(DBG_CAT_MOD, "delete_cat:
> %s,%u\n", str ? str->name : NULL, cnid);
> > =A0=A0=A0 sb =3D dir->i_sb;
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0 return res;
> >=20
> > =A0=A0=A0 hfs_cat_build_key(sb, fd.search_key,
> dir->i_ino, str);
> > =A0=A0=A0 res =3D hfs_brec_find(&fd);
> > @@ -281,7 +285,9 @@ int hfs_cat_move(u32 cnid, struct
> inode *src_dir, struct qstr *src_name,
> > =A0=A0=A0 dprint(DBG_CAT_MOD, "rename_cat: %u
> - %lu,%s - %lu,%s\n", cnid, src_dir->i_ino,
> src_name->name,
> > =A0=A0=A0 =A0=A0=A0
> dst_dir->i_ino, dst_name->name);
> > =A0=A0=A0 sb =3D src_dir->i_sb;
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->cat_tree, &src_fd);
> > +=A0=A0=A0 err =3D
> hfs_find_init(HFS_SB(sb)->cat_tree, &src_fd);
> > +=A0=A0=A0 if (err)
> > +=A0=A0=A0 =A0=A0=A0 return err;
> > =A0=A0=A0 dst_fd =3D src_fd;
> >=20
> > =A0=A0=A0 /* find the old dir entry and read
> the data */
> > diff --git a/fs/hfs/dir.c b/fs/hfs/dir.c
> > index 5f7f1ab..e1c8048 100644
> > --- a/fs/hfs/dir.c
> > +++ b/fs/hfs/dir.c
> > @@ -25,7 +25,9 @@ static struct dentry
> *hfs_lookup(struct inode *dir, struct dentry *dentry,
> > =A0=A0=A0 struct inode *inode =3D NULL;
> > =A0=A0=A0 int res;
> >=20
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(dir->i_sb)->cat_tree, &fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(dir->i_sb)->cat_tree, &fd);
> > +=A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0 return
> ERR_PTR(res);
> > =A0=A0=A0 hfs_cat_build_key(dir->i_sb,
> fd.search_key, dir->i_ino, &dentry->d_name);
> > =A0=A0=A0 res =3D hfs_brec_read(&fd,
> &rec, sizeof(rec));
> > =A0=A0=A0 if (res) {
> > @@ -63,7 +65,9 @@ static int hfs_readdir(struct file
> *filp, void *dirent, filldir_t filldir)
> > =A0=A0=A0 if (filp->f_pos >=3D
> inode->i_size)
> > =A0=A0=A0 =A0=A0=A0 return 0;
> >=20
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 err =3D
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 if (err)
> > +=A0=A0=A0 =A0=A0=A0 return err;
> > =A0=A0=A0 hfs_cat_build_key(sb, fd.search_key,
> inode->i_ino, NULL);
> > =A0=A0=A0 err =3D hfs_brec_find(&fd);
> > =A0=A0=A0 if (err)
> > diff --git a/fs/hfs/extent.c b/fs/hfs/extent.c
> > index a67955a..813447b 100644
> > --- a/fs/hfs/extent.c
> > +++ b/fs/hfs/extent.c
> > @@ -107,7 +107,7 @@ static u16 hfs_ext_lastblock(struct
> hfs_extent *ext)
> > =A0=A0=A0 return be16_to_cpu(ext->block) +
> be16_to_cpu(ext->count);
> > }
> >=20
> > -static void __hfs_ext_write_extent(struct inode
> *inode, struct hfs_find_data *fd)
> > +static int __hfs_ext_write_extent(struct inode *inode,
> struct hfs_find_data *fd)
> > {
> > =A0=A0=A0 int res;
> >=20
> > @@ -116,26 +116,31 @@ static void
> __hfs_ext_write_extent(struct inode *inode, struct
> hfs_find_data *fd
> > =A0=A0=A0 res =3D hfs_brec_find(fd);
> > =A0=A0=A0 if (HFS_I(inode)->flags &
> HFS_FLG_EXT_NEW) {
> > =A0=A0=A0 =A0=A0=A0 if (res !=3D
> -ENOENT)
> > -=A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 return;
> > +=A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 return res;
> > =A0=A0=A0 =A0=A0=A0
> hfs_brec_insert(fd, HFS_I(inode)->cached_extents,
> sizeof(hfs_extent_rec));
> > =A0=A0=A0 =A0=A0=A0
> HFS_I(inode)->flags &=3D
> ~(HFS_FLG_EXT_DIRTY|HFS_FLG_EXT_NEW);
> > =A0=A0=A0 } else {
> > =A0=A0=A0 =A0=A0=A0 if (res)
> > -=A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 return;
> > +=A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 return res;
> > =A0=A0=A0 =A0=A0=A0
> hfs_bnode_write(fd->bnode,
> HFS_I(inode)->cached_extents, fd->entryoffset,
> fd->entrylength);
> > =A0=A0=A0 =A0=A0=A0
> HFS_I(inode)->flags &=3D ~HFS_FLG_EXT_DIRTY;
> > =A0=A0=A0 }
> > +=A0=A0=A0 return 0;
> > }
> >=20
>=20
> As I see, this fix makes sense and for hfsplus also. Please,
> make it and for hfsplus.
>=20
> > -void hfs_ext_write_extent(struct inode *inode)
> > +int hfs_ext_write_extent(struct inode *inode)
> > {
> > =A0=A0=A0 struct hfs_find_data fd;
> > +=A0=A0=A0 int res =3D 0;
> >=20
> > =A0=A0=A0 if (HFS_I(inode)->flags &
> HFS_FLG_EXT_DIRTY) {
> > -=A0=A0=A0 =A0=A0=A0
> hfs_find_init(HFS_SB(inode->i_sb)->ext_tree,
> &fd);
> > -=A0=A0=A0 =A0=A0=A0
> __hfs_ext_write_extent(inode, &fd);
> > +=A0=A0=A0 =A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(inode->i_sb)->ext_tree,
> &fd);
> > +=A0=A0=A0 =A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 return res;
> > +=A0=A0=A0 =A0=A0=A0 res =3D
> __hfs_ext_write_extent(inode, &fd);
> > =A0=A0=A0 =A0=A0=A0
> hfs_find_exit(&fd);
> > =A0=A0=A0 }
> > +=A0=A0=A0 return res;
> > }
> >=20
> > static inline int __hfs_ext_read_extent(struct
> hfs_find_data *fd, struct hfs_extent *extent,
> > @@ -161,8 +166,11 @@ static inline int
> __hfs_ext_cache_extent(struct hfs_find_data *fd, struct
> inode
> > {
> > =A0=A0=A0 int res;
> >=20
> > -=A0=A0=A0 if (HFS_I(inode)->flags &
> HFS_FLG_EXT_DIRTY)
> > -=A0=A0=A0 =A0=A0=A0
> __hfs_ext_write_extent(inode, fd);
> > +=A0=A0=A0 if (HFS_I(inode)->flags &
> HFS_FLG_EXT_DIRTY) {
> > +=A0=A0=A0 =A0=A0=A0 res =3D
> __hfs_ext_write_extent(inode, fd);
> > +=A0=A0=A0 =A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 return res;
> > +=A0=A0=A0 }
> >=20
>=20
> Ditto for hfsplus.
>=20
> Thanks,
> Vyacheslav Dubeyko.
>=20
> > =A0=A0=A0 res =3D __hfs_ext_read_extent(fd,
> HFS_I(inode)->cached_extents, inode->i_ino,
> > =A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 =A0=A0=A0 =A0 =A0 block,
> HFS_IS_RSRC(inode) ? HFS_FK_RSRC : HFS_FK_DATA);
> > @@ -185,9 +193,11 @@ static int
> hfs_ext_read_extent(struct inode *inode, u16 block)
> > =A0=A0=A0 =A0 =A0 block <
> HFS_I(inode)->cached_start +
> HFS_I(inode)->cached_blocks)
> > =A0=A0=A0 =A0=A0=A0 return 0;
> >=20
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(inode->i_sb)->ext_tree,
> &fd);
> > -=A0=A0=A0 res =3D
> __hfs_ext_cache_extent(&fd, inode, block);
> > -=A0=A0=A0 hfs_find_exit(&fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(inode->i_sb)->ext_tree,
> &fd);
> > +=A0=A0=A0 if (!res) {
> > +=A0=A0=A0 =A0=A0=A0 res =3D
> __hfs_ext_cache_extent(&fd, inode, block);
> > +=A0=A0=A0 =A0=A0=A0
> hfs_find_exit(&fd);
> > +=A0=A0=A0 }
> > =A0=A0=A0 return res;
> > }
> >=20
> > @@ -298,7 +308,9 @@ int hfs_free_fork(struct
> super_block *sb, struct hfs_cat_file *file, int type)
> > =A0=A0=A0 if (total_blocks =3D=3D blocks)
> > =A0=A0=A0 =A0=A0=A0 return 0;
> >=20
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->ext_tree, &fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(sb)->ext_tree, &fd);
> > +=A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0 return res;
> > =A0=A0=A0 do {
> > =A0=A0=A0 =A0=A0=A0 res =3D
> __hfs_ext_read_extent(&fd, extent, cnid, total_blocks,
> type);
> > =A0=A0=A0 =A0=A0=A0 if (res)
> > @@ -438,7 +450,9 @@ out:
> >=20
> > insert_extent:
> > =A0=A0=A0 dprint(DBG_EXTENT, "insert new
> extent\n");
> > -=A0=A0=A0 hfs_ext_write_extent(inode);
> > +=A0=A0=A0 res =3D hfs_ext_write_extent(inode);
> > +=A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0 goto out;
> >=20
> > =A0=A0=A0
> memset(HFS_I(inode)->cached_extents, 0,
> sizeof(hfs_extent_rec));
> > =A0=A0=A0
> HFS_I(inode)->cached_extents[0].block =3D
> cpu_to_be16(start);
> > @@ -466,7 +480,6 @@ void hfs_file_truncate(struct inode
> *inode)
> > =A0=A0=A0 =A0=A0=A0 struct
> address_space *mapping =3D inode->i_mapping;
> > =A0=A0=A0 =A0=A0=A0 void *fsdata;
> > =A0=A0=A0 =A0=A0=A0 struct page
> *page;
> > -=A0=A0=A0 =A0=A0=A0 int res;
> >=20
> > =A0=A0=A0 =A0=A0=A0 /* XXX: Can use
> generic_cont_expand? */
> > =A0=A0=A0 =A0=A0=A0 size =3D
> inode->i_size - 1;
> > @@ -488,7 +501,12 @@ void hfs_file_truncate(struct
> inode *inode)
> > =A0=A0=A0 =A0=A0=A0 goto out;
> >=20
> > =A0=A0=A0
> mutex_lock(&HFS_I(inode)->extents_lock);
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->ext_tree, &fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(sb)->ext_tree, &fd);
> > +=A0=A0=A0 if (res) {
> > +=A0=A0=A0 =A0=A0=A0
> mutex_unlock(&HFS_I(inode)->extents_lock);
> > +=A0=A0=A0 =A0=A0=A0 /* XXX: We lack
> error handling of hfs_file_truncate() */
> > +=A0=A0=A0 =A0=A0=A0 return;
> > +=A0=A0=A0 }
> > =A0=A0=A0 while (1) {
> > =A0=A0=A0 =A0=A0=A0 if (alloc_cnt =3D=3D
> HFS_I(inode)->first_blocks) {
> > =A0=A0=A0 =A0=A0=A0
> =A0=A0=A0 hfs_free_extents(sb,
> HFS_I(inode)->first_extents,
> > diff --git a/fs/hfs/hfs_fs.h b/fs/hfs/hfs_fs.h
> > index 693df9f..67817af 100644
> > --- a/fs/hfs/hfs_fs.h
> > +++ b/fs/hfs/hfs_fs.h
> > @@ -174,7 +174,7 @@ extern const struct
> inode_operations hfs_dir_inode_operations;
> > /* extent.c */
> > extern int hfs_ext_keycmp(const btree_key *, const
> btree_key *);
> > extern int hfs_free_fork(struct super_block *, struct
> hfs_cat_file *, int);
> > -extern void hfs_ext_write_extent(struct inode *);
> > +extern int hfs_ext_write_extent(struct inode *);
> > extern int hfs_extend_file(struct inode *);
> > extern void hfs_file_truncate(struct inode *);
> >=20
> > diff --git a/fs/hfs/inode.c b/fs/hfs/inode.c
> > index 3031dfd..8ff4b5a 100644
> > --- a/fs/hfs/inode.c
> > +++ b/fs/hfs/inode.c
> > @@ -416,9 +416,12 @@ int hfs_write_inode(struct inode
> *inode, struct writeback_control *wbc)
> > =A0=A0=A0 struct inode *main_inode =3D inode;
> > =A0=A0=A0 struct hfs_find_data fd;
> > =A0=A0=A0 hfs_cat_rec rec;
> > +=A0=A0=A0 int res;
> >=20
> > =A0=A0=A0 dprint(DBG_INODE, "hfs_write_inode:
> %lu\n", inode->i_ino);
> > -=A0=A0=A0 hfs_ext_write_extent(inode);
> > +=A0=A0=A0 res =3D hfs_ext_write_extent(inode);
> > +=A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0 return res;
> >=20
> > =A0=A0=A0 if (inode->i_ino <
> HFS_FIRSTUSER_CNID) {
> > =A0=A0=A0 =A0=A0=A0 switch
> (inode->i_ino) {
> > @@ -515,7 +518,11 @@ static struct dentry
> *hfs_file_lookup(struct inode *dir, struct dentry *dentry,
> > =A0=A0=A0 if (!inode)
> > =A0=A0=A0 =A0=A0=A0 return
> ERR_PTR(-ENOMEM);
> >=20
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(dir->i_sb)->cat_tree, &fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(dir->i_sb)->cat_tree, &fd);
> > +=A0=A0=A0 if (res) {
> > +=A0=A0=A0 =A0=A0=A0 iput(inode);
> > +=A0=A0=A0 =A0=A0=A0 return
> ERR_PTR(res);
> > +=A0=A0=A0 }
> > =A0=A0=A0 fd.search_key->cat =3D
> HFS_I(dir)->cat_key;
> > =A0=A0=A0 res =3D hfs_brec_read(&fd,
> &rec, sizeof(rec));
> > =A0=A0=A0 if (!res) {
> > diff --git a/fs/hfs/super.c b/fs/hfs/super.c
> > index bbaaa8a..719760b 100644
> > --- a/fs/hfs/super.c
> > +++ b/fs/hfs/super.c
> > @@ -418,7 +418,9 @@ static int hfs_fill_super(struct
> super_block *sb, void *data, int silent)
> > =A0=A0=A0 }
> >=20
> > =A0=A0=A0 /* try to get the root inode */
> > -=A0=A0=A0
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 res =3D
> hfs_find_init(HFS_SB(sb)->cat_tree, &fd);
> > +=A0=A0=A0 if (res)
> > +=A0=A0=A0 =A0=A0=A0 goto
> bail_no_root;
> > =A0=A0=A0 res =3D hfs_cat_find_brec(sb,
> HFS_ROOT_CNID, &fd);
> > =A0=A0=A0 if (!res) {
> > =A0=A0=A0 =A0=A0=A0 if
> (fd.entrylength > sizeof(rec) || fd.entrylength < 0)
> {
> > --=20
> > 1.7.9.5
> >=20
> > --
> > To unsubscribe from this list: send the line
> "unsubscribe linux-kernel" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at=A0 http://vger.kernel.org/majordomo-info.htm=
l
> > Please read the FAQ at=A0 http://www.tux.org/lkml/
>=20
>=20