These patches fix potential bugs associated with link target handling by
NUL-terminating names read from disk.
This is version 3 of these patches. It fixes the bug, pointed out by Al Viro
and Evgeniy Dushistov, that i_size was not being validated. In order to
facilitate this it introduces a helper function for terminating the link name,
as suggested by Al.
diffstat:
fs/ext2/inode.c | 7 +++++--
fs/ext3/inode.c | 7 +++++--
fs/ext4/inode.c | 7 +++++--
fs/freevxfs/vxfs_inode.c | 4 +++-
fs/namei.c | 7 +++++--
fs/sysv/inode.c | 6 +++++-
fs/ufs/inode.c | 8 ++++++--
include/linux/namei.h | 5 +++++
8 files changed, 39 insertions(+), 12 deletions(-)
A number of filesystems were potentially triggering kernel bugs due to
corrupted symlink names on disk. This helper helps safely terminate the
names.
Cc: Al Viro <[email protected]>
Cc: Andrew Morton <[email protected]>
Signed-off-by: Duane Griffin <[email protected]>
---
include/linux/namei.h | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 99eb803..231ce28 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -94,4 +94,9 @@ static inline char *nd_get_link(struct nameidata *nd)
return nd->saved_names[nd->depth];
}
+static inline void nd_terminate_link(void *name, unsigned len, unsigned maxlen)
+{
+ ((char *) name)[min(len, maxlen)] = '\0';
+}
+
#endif /* _LINUX_NAMEI_H */
--
1.6.0.4
On-disk data corruption could cause a page link to have its i_size set
to PAGE_SIZE (or a multiple thereof) and its contents all non-NUL.
NUL-terminate the link name to ensure this doesn't cause further
problems for the kernel.
Cc: Al Viro <[email protected]>
Cc: Andrew Morton <[email protected]>
Signed-off-by: Duane Griffin <[email protected]>
---
fs/namei.c | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/fs/namei.c b/fs/namei.c
index af3783f..2416922 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2750,13 +2750,16 @@ int vfs_follow_link(struct nameidata *nd, const char *link)
/* get the link contents into pagecache */
static char *page_getlink(struct dentry * dentry, struct page **ppage)
{
- struct page * page;
+ char *kaddr;
+ struct page *page;
struct address_space *mapping = dentry->d_inode->i_mapping;
page = read_mapping_page(mapping, 0, NULL);
if (IS_ERR(page))
return (char*)page;
*ppage = page;
- return kmap(page);
+ kaddr = kmap(page);
+ nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1);
+ return kaddr;
}
int page_readlink(struct dentry *dentry, char __user *buffer, int buflen)
--
1.6.0.4
Ensure fast symlink targets are NUL-terminated, even if corrupted
on-disk.
Cc: Andrew Morton <[email protected]>
Signed-off-by: Duane Griffin <[email protected]>
---
fs/ext2/inode.c | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
index 7658b33..c902504 100644
--- a/fs/ext2/inode.c
+++ b/fs/ext2/inode.c
@@ -32,6 +32,7 @@
#include <linux/buffer_head.h>
#include <linux/mpage.h>
#include <linux/fiemap.h>
+#include <linux/namei.h>
#include "ext2.h"
#include "acl.h"
#include "xip.h"
@@ -1286,9 +1287,11 @@ struct inode *ext2_iget (struct super_block *sb, unsigned long ino)
else
inode->i_mapping->a_ops = &ext2_aops;
} else if (S_ISLNK(inode->i_mode)) {
- if (ext2_inode_is_fast_symlink(inode))
+ if (ext2_inode_is_fast_symlink(inode)) {
inode->i_op = &ext2_fast_symlink_inode_operations;
- else {
+ nd_terminate_link(ei->i_data, inode->i_size,
+ sizeof(ei->i_data));
+ } else {
inode->i_op = &ext2_symlink_inode_operations;
if (test_opt(inode->i_sb, NOBH))
inode->i_mapping->a_ops = &ext2_nobh_aops;
--
1.6.0.4
Ensure fast symlink targets are NUL-terminated, even if corrupted
on-disk.
Cc: Andrew Morton <[email protected]>
Cc: Stephen Tweedie <[email protected]>
Cc: [email protected]
Signed-off-by: Duane Griffin <[email protected]>
---
fs/ext3/inode.c | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
index f8424ad..24aacdf 100644
--- a/fs/ext3/inode.c
+++ b/fs/ext3/inode.c
@@ -37,6 +37,7 @@
#include <linux/uio.h>
#include <linux/bio.h>
#include <linux/fiemap.h>
+#include <linux/namei.h>
#include "xattr.h"
#include "acl.h"
@@ -2817,9 +2818,11 @@ struct inode *ext3_iget(struct super_block *sb, unsigned long ino)
inode->i_op = &ext3_dir_inode_operations;
inode->i_fop = &ext3_dir_operations;
} else if (S_ISLNK(inode->i_mode)) {
- if (ext3_inode_is_fast_symlink(inode))
+ if (ext3_inode_is_fast_symlink(inode)) {
inode->i_op = &ext3_fast_symlink_inode_operations;
- else {
+ nd_terminate_link(ei->i_data, inode->i_size,
+ sizeof(ei->i_data));
+ } else {
inode->i_op = &ext3_symlink_inode_operations;
ext3_set_aops(inode);
}
--
1.6.0.4
Ensure fast symlink targets are NUL-terminated, even if corrupted
on-disk.
Cc: Christoph Hellwig <[email protected]>
Cc: Al Viro <[email protected]>
Signed-off-by: Duane Griffin <[email protected]>
---
fs/sysv/inode.c | 6 +++++-
1 files changed, 5 insertions(+), 1 deletions(-)
diff --git a/fs/sysv/inode.c b/fs/sysv/inode.c
index df0d435..c93189d 100644
--- a/fs/sysv/inode.c
+++ b/fs/sysv/inode.c
@@ -27,6 +27,7 @@
#include <linux/init.h>
#include <linux/buffer_head.h>
#include <linux/vfs.h>
+#include <linux/namei.h>
#include <asm/byteorder.h>
#include "sysv.h"
@@ -163,8 +164,11 @@ void sysv_set_inode(struct inode *inode, dev_t rdev)
if (inode->i_blocks) {
inode->i_op = &sysv_symlink_inode_operations;
inode->i_mapping->a_ops = &sysv_aops;
- } else
+ } else {
inode->i_op = &sysv_fast_symlink_inode_operations;
+ nd_terminate_link(SYSV_I(inode)->i_data, inode->i_size,
+ sizeof(SYSV_I(inode)->i_data));
+ }
} else
init_special_inode(inode, inode->i_mode, rdev);
}
--
1.6.0.4
Ensure fast symlink targets are NUL-terminated, even if corrupted
on-disk.
Cc: Christoph Hellwig <[email protected]>
Signed-off-by: Duane Griffin <[email protected]>
---
fs/freevxfs/vxfs_inode.c | 4 +++-
1 files changed, 3 insertions(+), 1 deletions(-)
diff --git a/fs/freevxfs/vxfs_inode.c b/fs/freevxfs/vxfs_inode.c
index 9f3f2ce..03a6ea5 100644
--- a/fs/freevxfs/vxfs_inode.c
+++ b/fs/freevxfs/vxfs_inode.c
@@ -325,8 +325,10 @@ vxfs_iget(struct super_block *sbp, ino_t ino)
if (!VXFS_ISIMMED(vip)) {
ip->i_op = &page_symlink_inode_operations;
ip->i_mapping->a_ops = &vxfs_aops;
- } else
+ } else {
ip->i_op = &vxfs_immed_symlink_iops;
+ vip->vii_immed.vi_immed[ip->i_size] = '\0';
+ }
} else
init_special_inode(ip, ip->i_mode, old_decode_dev(vip->vii_rdev));
--
1.6.0.4
Ensure fast symlink targets are NUL-terminated, even if corrupted
on-disk.
Cc: Evgeniy Dushistov <[email protected]>
Signed-off-by: Duane Griffin <[email protected]>
---
fs/ufs/inode.c | 8 ++++++--
1 files changed, 6 insertions(+), 2 deletions(-)
diff --git a/fs/ufs/inode.c b/fs/ufs/inode.c
index 39f8778..aca4b58 100644
--- a/fs/ufs/inode.c
+++ b/fs/ufs/inode.c
@@ -36,6 +36,7 @@
#include <linux/mm.h>
#include <linux/smp_lock.h>
#include <linux/buffer_head.h>
+#include <linux/namei.h>
#include "ufs_fs.h"
#include "ufs.h"
@@ -606,9 +607,12 @@ static void ufs_set_inode_ops(struct inode *inode)
inode->i_fop = &ufs_dir_operations;
inode->i_mapping->a_ops = &ufs_aops;
} else if (S_ISLNK(inode->i_mode)) {
- if (!inode->i_blocks)
+ if (!inode->i_blocks) {
inode->i_op = &ufs_fast_symlink_inode_operations;
- else {
+ nd_terminate_link(UFS_I(inode)->i_u1.i_symlink,
+ inode->i_size,
+ sizeof(UFS_I(inode)->i_u1.i_symlink));
+ } else {
inode->i_op = &page_symlink_inode_operations;
inode->i_mapping->a_ops = &ufs_aops;
}
--
1.6.0.4
Ensure fast symlink targets are NUL-terminated, even if corrupted
on-disk.
Cc: Andrew Morton <[email protected]>
Cc: Theodore Ts'o <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Duane Griffin <[email protected]>
---
fs/ext4/inode.c | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index ca88060..f0e6bfb 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -34,6 +34,7 @@
#include <linux/writeback.h>
#include <linux/pagevec.h>
#include <linux/mpage.h>
+#include <linux/namei.h>
#include <linux/uio.h>
#include <linux/bio.h>
#include "ext4_jbd2.h"
@@ -4200,9 +4201,11 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
inode->i_op = &ext4_dir_inode_operations;
inode->i_fop = &ext4_dir_operations;
} else if (S_ISLNK(inode->i_mode)) {
- if (ext4_inode_is_fast_symlink(inode))
+ if (ext4_inode_is_fast_symlink(inode)) {
inode->i_op = &ext4_fast_symlink_inode_operations;
- else {
+ nd_terminate_link(ei->i_data, inode->i_size,
+ sizeof(ei->i_data));
+ } else {
inode->i_op = &ext4_symlink_inode_operations;
ext4_set_aops(inode);
}
--
1.6.0.4
On Tue, Dec 16, 2008 at 03:51:53PM +0000, Duane Griffin wrote:
> A number of filesystems were potentially triggering kernel bugs due to
> corrupted symlink names on disk. This helper helps safely terminate the
> names.
s/unsigned/size_t/g and into the queue that goes...
On Tue, 2008-12-16 at 15:51 +0000, Duane Griffin wrote:
> These patches fix potential bugs associated with link target handling by
> NUL-terminating names read from disk.
>
> This is version 3 of these patches. It fixes the bug, pointed out by Al Viro
> and Evgeniy Dushistov, that i_size was not being validated. In order to
> facilitate this it introduces a helper function for terminating the link name,
> as suggested by Al.
Duane,
I see you left out jfs, so I'm assuming you're letting me handle that
one. I've push my patch upstream, so it should show up in the -next
tree.
Shaggy
> diffstat:
> fs/ext2/inode.c | 7 +++++--
> fs/ext3/inode.c | 7 +++++--
> fs/ext4/inode.c | 7 +++++--
> fs/freevxfs/vxfs_inode.c | 4 +++-
> fs/namei.c | 7 +++++--
> fs/sysv/inode.c | 6 +++++-
> fs/ufs/inode.c | 8 ++++++--
> include/linux/namei.h | 5 +++++
> 8 files changed, 39 insertions(+), 12 deletions(-)
--
David Kleikamp
IBM Linux Technology Center
2008/12/16 Dave Kleikamp <[email protected]>:
> Duane,
> I see you left out jfs, so I'm assuming you're letting me handle that
> one. I've push my patch upstream, so it should show up in the -next
> tree.
That's right, since jfs doesn't need the helper and your patch was a
full and complete fix. I should have clarified whether that was how
you wanted it handled, sorry.
Which reminds me, I need to follow up about the original befs and 9p
patches which didn't make it to the list and haven't been ACKed or
NAKed yet...
> Shaggy
Cheers,
Duane.
--
"I never could learn to drink that blood and call it wine" - Bob Dylan
On Tue, Dec 16, 2008 at 03:51:58PM +0000, Duane Griffin wrote:
> Ensure fast symlink targets are NUL-terminated, even if corrupted
> on-disk.
>
> Cc: Evgeniy Dushistov <[email protected]>
> Signed-off-by: Duane Griffin <[email protected]>
> ---
> fs/ufs/inode.c | 8 ++++++--
> 1 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/fs/ufs/inode.c b/fs/ufs/inode.c
> index 39f8778..aca4b58 100644
> --- a/fs/ufs/inode.c
> +++ b/fs/ufs/inode.c
> @@ -36,6 +36,7 @@
> #include <linux/mm.h>
> #include <linux/smp_lock.h>
> #include <linux/buffer_head.h>
> +#include <linux/namei.h>
>
> #include "ufs_fs.h"
> #include "ufs.h"
> @@ -606,9 +607,12 @@ static void ufs_set_inode_ops(struct inode *inode)
> inode->i_fop = &ufs_dir_operations;
> inode->i_mapping->a_ops = &ufs_aops;
> } else if (S_ISLNK(inode->i_mode)) {
> - if (!inode->i_blocks)
> + if (!inode->i_blocks) {
> inode->i_op = &ufs_fast_symlink_inode_operations;
> - else {
> + nd_terminate_link(UFS_I(inode)->i_u1.i_symlink,
> + inode->i_size,
> + sizeof(UFS_I(inode)->i_u1.i_symlink));
> + } else {
> inode->i_op = &page_symlink_inode_operations;
> inode->i_mapping->a_ops = &ufs_aops;
> }
> --
> 1.6.0.4
There is different types of ufs, one used 64 bit for "pointers to
blocks", another 32 bit,
so sizeof(UFS_I(inode)->i_u1.i_symlink))
is not right choice every time,
in ufs2 it should be
sizeof(UFS_I(inode)->i_u1.u2_i_data) which 2 times bigger,
also there is hint for *BSD ufs
fs/ufs/ufs_fs.h:
__fs32 fs_maxsymlinklen;/* max length of an internal symlink */
which may be used if ufs type ufs1 or ufs2
--
/Evgeniy
On Tue, Dec 16, 2008 at 11:18:50PM +0000, Duane Griffin wrote:
> On Tue, Dec 16, 2008 at 10:40:55PM +0300, Evgeniy Dushistov wrote:
> > There is different types of ufs, one used 64 bit for "pointers to
> > blocks", another 32 bit,
> > so sizeof(UFS_I(inode)->i_u1.i_symlink))
> > is not right choice every time,
> > in ufs2 it should be
> > sizeof(UFS_I(inode)->i_u1.u2_i_data) which 2 times bigger,
> >
> > also there is hint for *BSD ufs
> >
> > fs/ufs/ufs_fs.h:
> > __fs32 fs_maxsymlinklen;/* max length of an internal symlink */
> >
> > which may be used if ufs type ufs1 or ufs2
>
> Hmm, I see. However it looks like ufs1_read_inode and ufs2_read_inode
> both copy the same, ((UFS_NDADDR + UFS_NINDIR) * 4), amount of inline
> symlink data. They also both copy it to ufs_inode_info->i_u1.i_symlink
> (not that that matters, I suppose). Perhaps I'm being obtuse, but it
> looks like inline ufs2 symlinks between 60 and 120 characters long are
> being truncated to 60 characters, no?
>
> There also doesn't seem to be any validation of (f)s_maxsymlinklen being
> done. Unless I'm mistaken ufs_symlink could end up overwriting random
> memory if it contains a large bogus value.
>
> Does that all sound correct? If so would you like me to whip up a couple
> of patches to fix it? I'll respin the NUL-termination patch on top of
> those, if so.
>
Yes, it looks like there is typo in ufs2 variant of copying symlink names.
Typical value of superblock's maxsymlinklen field for ufs2 is 120.
Patches to fix this are welcome.
--
/Evgeniy
On Tue, Dec 16, 2008 at 10:40:55PM +0300, Evgeniy Dushistov wrote:
> There is different types of ufs, one used 64 bit for "pointers to
> blocks", another 32 bit,
> so sizeof(UFS_I(inode)->i_u1.i_symlink))
> is not right choice every time,
> in ufs2 it should be
> sizeof(UFS_I(inode)->i_u1.u2_i_data) which 2 times bigger,
>
> also there is hint for *BSD ufs
>
> fs/ufs/ufs_fs.h:
> __fs32 fs_maxsymlinklen;/* max length of an internal symlink */
>
> which may be used if ufs type ufs1 or ufs2
Hmm, I see. However it looks like ufs1_read_inode and ufs2_read_inode
both copy the same, ((UFS_NDADDR + UFS_NINDIR) * 4), amount of inline
symlink data. They also both copy it to ufs_inode_info->i_u1.i_symlink
(not that that matters, I suppose). Perhaps I'm being obtuse, but it
looks like inline ufs2 symlinks between 60 and 120 characters long are
being truncated to 60 characters, no?
There also doesn't seem to be any validation of (f)s_maxsymlinklen being
done. Unless I'm mistaken ufs_symlink could end up overwriting random
memory if it contains a large bogus value.
Does that all sound correct? If so would you like me to whip up a couple
of patches to fix it? I'll respin the NUL-termination patch on top of
those, if so.
> /Evgeniy
Cheers,
Duane.
--
"I never could learn to drink that blood and call it wine" - Bob Dylan
On Dec 16, 2008 15:51 +0000, Duane Griffin wrote:
> A number of filesystems were potentially triggering kernel bugs due to
> corrupted symlink names on disk. This helper helps safely terminate the
> names.
>
> +static inline void nd_terminate_link(void *name,unsigned len,unsigned maxlen)
> +{
> + ((char *) name)[min(len, maxlen)] = '\0';
> +}
> @@ -4200,9 +4201,11 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
> } else if (S_ISLNK(inode->i_mode)) {
> + if (ext4_inode_is_fast_symlink(inode)) {
> inode->i_op = &ext4_fast_symlink_inode_operations;
> + nd_terminate_link(ei->i_data, inode->i_size,
> + sizeof(ei->i_data));
> + } else {
> inode->i_op = &ext4_symlink_inode_operations;
> ext4_set_aops(inode);
> }
With sizeof(ei->i_data) = 15 * 4 = 60 bytes, this will set ei->i_data[60]
as NUL, which is writing 1 byte beyond the end of the array.
Note that in ext[234]_symlink() the check for fast symlinks is:
l = strlen(symname)+1;
if (l > sizeof (EXT3_I(inode)->i_data)) {
inode->i_op = &ext3_symlink_inode_operations;
} else {
inode->i_op = &ext3_fast_symlink_inode_operations;
inode->i_size = l-1;
}
so in fact the fast symlinks should always have space for a trailing NUL
character, and "sizeof(ei->i_data) - 1" is the right maxlen to use for
ext[234].
That might not be true for other filesystems, in which case you would
need to add a "padding" field after the symlink name in memory to hold
the trailing NUL.
Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.
2008/12/16 Andreas Dilger <[email protected]>:
> On Dec 16, 2008 15:51 +0000, Duane Griffin wrote:
>> A number of filesystems were potentially triggering kernel bugs due to
>> corrupted symlink names on disk. This helper helps safely terminate the
>> names.
>>
>> +static inline void nd_terminate_link(void *name,unsigned len,unsigned maxlen)
>> +{
>> + ((char *) name)[min(len, maxlen)] = '\0';
>> +}
>
>> @@ -4200,9 +4201,11 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
>> } else if (S_ISLNK(inode->i_mode)) {
>> + if (ext4_inode_is_fast_symlink(inode)) {
>> inode->i_op = &ext4_fast_symlink_inode_operations;
>> + nd_terminate_link(ei->i_data, inode->i_size,
>> + sizeof(ei->i_data));
>> + } else {
>> inode->i_op = &ext4_symlink_inode_operations;
>> ext4_set_aops(inode);
>> }
>
> With sizeof(ei->i_data) = 15 * 4 = 60 bytes, this will set ei->i_data[60]
> as NUL, which is writing 1 byte beyond the end of the array.
Argh, you are correct, of course.
> Note that in ext[234]_symlink() the check for fast symlinks is:
>
> l = strlen(symname)+1;
> if (l > sizeof (EXT3_I(inode)->i_data)) {
> inode->i_op = &ext3_symlink_inode_operations;
> } else {
> inode->i_op = &ext3_fast_symlink_inode_operations;
> inode->i_size = l-1;
> }
>
> so in fact the fast symlinks should always have space for a trailing NUL
> character, and "sizeof(ei->i_data) - 1" is the right maxlen to use for
> ext[234].
>
> That might not be true for other filesystems, in which case you would
> need to add a "padding" field after the symlink name in memory to hold
> the trailing NUL.
It is true for sysv, too. The ufs code also, but that has other
issues, anyway. The generic page symlink and freevxfs patches should
be fine, though.
Cheers,
Duane.
--
"I never could learn to drink that blood and call it wine" - Bob Dylan