Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
On Mon, Oct 26, 2020 at 01:07:15PM -0700, Eric Biggers wrote:
> From: Eric Biggers
>
> Commit 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> names") made the kernel start accepting arbitrarily long algorithm names
> in sockaddr_alg. However, the actual length of the salg_name field
> stayed at the original 64 bytes.
>
> This is broken because the kernel can access indices >= 64 in salg_name,
> which is undefined behavior -- even though the memory that is accessed
> is still located within the sockaddr structure. It would only be
> defined behavior if the array were properly marked as arbitrary-length
> (either by making it a flexible array, which is the recommended way
> these days, or by making it an array of length 0 or 1).
>
> We can't simply change salg_name into a flexible array, since that would
> break source compatibility with userspace programs that embed
> sockaddr_alg into another struct, or (more commonly) declare a
> sockaddr_alg like 'struct sockaddr_alg sa = { .salg_name = "foo" };'.
>
> One solution would be to change salg_name into a flexible array only
> when '#ifdef __KERNEL__'. However, that would keep userspace without an
> easy way to actually use the longer algorithm names.
>
> Instead, add a new structure 'sockaddr_alg_new' that has the flexible
> array field, and expose it to both userspace and the kernel.
> Make the kernel use it correctly in alg_bind().
>
> This addresses the syzbot report
> "UBSAN: array-index-out-of-bounds in alg_bind"
> (https://syzkaller.appspot.com/bug?extid=92ead4eb8e26a26d465e).
>
> Reported-by: syzbot+92ead4eb8e26a26d4...@syzkaller.appspotmail.com
> Fixes: 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> names")
> Cc: # v4.12+
> Signed-off-by: Eric Biggers
> ---
> crypto/af_alg.c | 10 +++---
> include/uapi/linux/if_alg.h | 16
> 2 files changed, 23 insertions(+), 3 deletions(-)
Patch applied. Thanks.
--
Email: Herbert Xu
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
On Mon, Oct 26, 2020 at 06:10:59PM -0500, Gustavo A. R. Silva wrote:
> On Mon, Oct 26, 2020 at 04:21:48PM -0500, Gustavo A. R. Silva wrote:
> > > +/*
> > > + * Linux v4.12 and later removed the 64-byte limit on salg_name[]; it's
> > > now an
> > > + * arbitrary-length field. We had to keep the original struct above for
> > > source
> > > + * compatibility with existing userspace programs, though. Use the new
> > > struct
> > > + * below if support for very long algorithm names is needed. To do this,
> > > + * allocate 'sizeof(struct sockaddr_alg_new) + strlen(algname) + 1'
> > > bytes, and
> > > + * copy algname (including the null terminator) into salg_name.
> > > + */
> > > +struct sockaddr_alg_new {
> > > + __u16 salg_family;
> > > + __u8salg_type[14];
> > > + __u32 salg_feat;
> > > + __u32 salg_mask;
> > > + __u8salg_name[];
> > > +};
> > > +
> >
> > How something like this, instead:
> >
> > struct sockaddr_alg {
> > - __u16 salg_family;
> > - __u8salg_type[14];
> > - __u32 salg_feat;
> > - __u32 salg_mask;
> > - __u8salg_name[64];
> > + union {
> > + struct {
> > + __u16 salg_v1_family;
> > + __u8salg_v1_type[14];
> > + __u32 salg_v1_feat;
> > + __u32 salg_v1_mask;
> > + __u8salg_name[64];
> > + };
> > + struct {
> > + __u16 salg_family;
> > + __u8salg_type[14];
> > + __u32 salg_feat;
> > + __u32 salg_mask;
> > + __u8salg_name_new[];
> > + };
> > + };
> > };
> >
>
> Something similar to the following approach might work:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux.git/commit/?h=testing/uapi/gntalloc=db46c8aba41c436edb0b4ef2941bd7390b0e5d61
>
I suppose so. It's very confusing to see a union like that at first glance,
though. It definitely needs an explanatory comment...
- Eric
Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
On Mon, Oct 26, 2020 at 04:21:48PM -0500, Gustavo A. R. Silva wrote:
> > +/*
> > + * Linux v4.12 and later removed the 64-byte limit on salg_name[]; it's
> > now an
> > + * arbitrary-length field. We had to keep the original struct above for
> > source
> > + * compatibility with existing userspace programs, though. Use the new
> > struct
> > + * below if support for very long algorithm names is needed. To do this,
> > + * allocate 'sizeof(struct sockaddr_alg_new) + strlen(algname) + 1' bytes,
> > and
> > + * copy algname (including the null terminator) into salg_name.
> > + */
> > +struct sockaddr_alg_new {
> > + __u16 salg_family;
> > + __u8salg_type[14];
> > + __u32 salg_feat;
> > + __u32 salg_mask;
> > + __u8salg_name[];
> > +};
> > +
>
> How something like this, instead:
>
> struct sockaddr_alg {
> - __u16 salg_family;
> - __u8salg_type[14];
> - __u32 salg_feat;
> - __u32 salg_mask;
> - __u8salg_name[64];
> + union {
> + struct {
> + __u16 salg_v1_family;
> + __u8salg_v1_type[14];
> + __u32 salg_v1_feat;
> + __u32 salg_v1_mask;
> + __u8salg_name[64];
> + };
> + struct {
> + __u16 salg_family;
> + __u8salg_type[14];
> + __u32 salg_feat;
> + __u32 salg_mask;
> + __u8salg_name_new[];
> + };
> + };
> };
>
Something similar to the following approach might work:
https://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux.git/commit/?h=testing/uapi/gntalloc=db46c8aba41c436edb0b4ef2941bd7390b0e5d61
--
Gustavo
Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
On Mon, Oct 26, 2020 at 10:57 PM Eric Biggers wrote:
> On Mon, Oct 26, 2020 at 10:23:35PM +0100, 'Jann Horn' via syzkaller-bugs
> wrote:
> > On Mon, Oct 26, 2020 at 9:08 PM Eric Biggers wrote:
> > > Commit 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> > > names") made the kernel start accepting arbitrarily long algorithm names
> > > in sockaddr_alg.
> >
> > That's not true; it's still limited by the size of struct
> > sockaddr_storage (128 bytes total for the entire address).
>
> Interesting, so the actual limit is 104 bytes. It seems like the intent of
> that
> commit was to make it unlimited, though...
>
> > If you make it longer, __copy_msghdr_from_user() will silently truncate the
> > size.
>
> That's used for sys_sendmsg(), which AFAICT isn't relevant here. sockaddr_alg
> is used with sys_bind(), which fails with EINVAL if the address is longer than
> sizeof(struct sockaddr_storage).
Ugh, of course you're right, sorry.
> However, since sys_sendmsg() is truncating overly-long addresses, it's
> probably
> the case that sizeof(struct sockaddr_storage) can never be increased in the
> future...
Eh, I think there'd probably be bigger issues with that elsewhere.
> > > This is broken because the kernel can access indices >= 64 in salg_name,
> > > which is undefined behavior -- even though the memory that is accessed
> > > is still located within the sockaddr structure. It would only be
> > > defined behavior if the array were properly marked as arbitrary-length
> > > (either by making it a flexible array, which is the recommended way
> > > these days, or by making it an array of length 0 or 1).
> > >
> > > We can't simply change salg_name into a flexible array, since that would
> > > break source compatibility with userspace programs that embed
> > > sockaddr_alg into another struct, or (more commonly) declare a
> > > sockaddr_alg like 'struct sockaddr_alg sa = { .salg_name = "foo" };'.
> > >
> > > One solution would be to change salg_name into a flexible array only
> > > when '#ifdef __KERNEL__'. However, that would keep userspace without an
> > > easy way to actually use the longer algorithm names.
> > >
> > > Instead, add a new structure 'sockaddr_alg_new' that has the flexible
> > > array field, and expose it to both userspace and the kernel.
> > > Make the kernel use it correctly in alg_bind().
> > [...]
> > > @@ -147,7 +147,7 @@ static int alg_bind(struct socket *sock, struct
> > > sockaddr *uaddr, int addr_len)
> > > const u32 allowed = CRYPTO_ALG_KERN_DRIVER_ONLY;
> > > struct sock *sk = sock->sk;
> > > struct alg_sock *ask = alg_sk(sk);
> > > - struct sockaddr_alg *sa = (void *)uaddr;
> > > + struct sockaddr_alg_new *sa = (void *)uaddr;
> > > const struct af_alg_type *type;
> > > void *private;
> > > int err;
> > > @@ -155,7 +155,11 @@ static int alg_bind(struct socket *sock, struct
> > > sockaddr *uaddr, int addr_len)
> > > if (sock->state == SS_CONNECTED)
> > > return -EINVAL;
> > >
> > > - if (addr_len < sizeof(*sa))
> > > + BUILD_BUG_ON(offsetof(struct sockaddr_alg_new, salg_name) !=
> > > +offsetof(struct sockaddr_alg, salg_name));
> > > + BUILD_BUG_ON(offsetof(struct sockaddr_alg, salg_name) !=
> > > sizeof(*sa));
> > > +
> > > + if (addr_len < sizeof(*sa) + 1)
> > > return -EINVAL;
> > >
> > > /* If caller uses non-allowed flag, return error. */
> > > @@ -163,7 +167,7 @@ static int alg_bind(struct socket *sock, struct
> > > sockaddr *uaddr, int addr_len)
> > > return -EINVAL;
> > >
> > > sa->salg_type[sizeof(sa->salg_type) - 1] = 0;
> > > - sa->salg_name[sizeof(sa->salg_name) + addr_len - sizeof(*sa) - 1]
> > > = 0;
> > > + sa->salg_name[addr_len - sizeof(*sa) - 1] = 0;
> >
> > This looks like an out-of-bounds write in the case `addr_len ==
> > sizeof(struct sockaddr_storage)`.
Sorry, I've been unusually unconcentrated today. Sorry about the
noise, ignore what I said.
Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
On Mon, Oct 26, 2020 at 10:23:35PM +0100, 'Jann Horn' via syzkaller-bugs wrote:
> On Mon, Oct 26, 2020 at 9:08 PM Eric Biggers wrote:
> > Commit 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> > names") made the kernel start accepting arbitrarily long algorithm names
> > in sockaddr_alg.
>
> That's not true; it's still limited by the size of struct
> sockaddr_storage (128 bytes total for the entire address).
Interesting, so the actual limit is 104 bytes. It seems like the intent of that
commit was to make it unlimited, though...
> If you make it longer, __copy_msghdr_from_user() will silently truncate the
> size.
That's used for sys_sendmsg(), which AFAICT isn't relevant here. sockaddr_alg
is used with sys_bind(), which fails with EINVAL if the address is longer than
sizeof(struct sockaddr_storage).
However, since sys_sendmsg() is truncating overly-long addresses, it's probably
the case that sizeof(struct sockaddr_storage) can never be increased in the
future...
>
> > This is broken because the kernel can access indices >= 64 in salg_name,
> > which is undefined behavior -- even though the memory that is accessed
> > is still located within the sockaddr structure. It would only be
> > defined behavior if the array were properly marked as arbitrary-length
> > (either by making it a flexible array, which is the recommended way
> > these days, or by making it an array of length 0 or 1).
> >
> > We can't simply change salg_name into a flexible array, since that would
> > break source compatibility with userspace programs that embed
> > sockaddr_alg into another struct, or (more commonly) declare a
> > sockaddr_alg like 'struct sockaddr_alg sa = { .salg_name = "foo" };'.
> >
> > One solution would be to change salg_name into a flexible array only
> > when '#ifdef __KERNEL__'. However, that would keep userspace without an
> > easy way to actually use the longer algorithm names.
> >
> > Instead, add a new structure 'sockaddr_alg_new' that has the flexible
> > array field, and expose it to both userspace and the kernel.
> > Make the kernel use it correctly in alg_bind().
> [...]
> > @@ -147,7 +147,7 @@ static int alg_bind(struct socket *sock, struct
> > sockaddr *uaddr, int addr_len)
> > const u32 allowed = CRYPTO_ALG_KERN_DRIVER_ONLY;
> > struct sock *sk = sock->sk;
> > struct alg_sock *ask = alg_sk(sk);
> > - struct sockaddr_alg *sa = (void *)uaddr;
> > + struct sockaddr_alg_new *sa = (void *)uaddr;
> > const struct af_alg_type *type;
> > void *private;
> > int err;
> > @@ -155,7 +155,11 @@ static int alg_bind(struct socket *sock, struct
> > sockaddr *uaddr, int addr_len)
> > if (sock->state == SS_CONNECTED)
> > return -EINVAL;
> >
> > - if (addr_len < sizeof(*sa))
> > + BUILD_BUG_ON(offsetof(struct sockaddr_alg_new, salg_name) !=
> > +offsetof(struct sockaddr_alg, salg_name));
> > + BUILD_BUG_ON(offsetof(struct sockaddr_alg, salg_name) !=
> > sizeof(*sa));
> > +
> > + if (addr_len < sizeof(*sa) + 1)
> > return -EINVAL;
> >
> > /* If caller uses non-allowed flag, return error. */
> > @@ -163,7 +167,7 @@ static int alg_bind(struct socket *sock, struct
> > sockaddr *uaddr, int addr_len)
> > return -EINVAL;
> >
> > sa->salg_type[sizeof(sa->salg_type) - 1] = 0;
> > - sa->salg_name[sizeof(sa->salg_name) + addr_len - sizeof(*sa) - 1] =
> > 0;
> > + sa->salg_name[addr_len - sizeof(*sa) - 1] = 0;
>
> This looks like an out-of-bounds write in the case `addr_len ==
> sizeof(struct sockaddr_storage)`.
I think you mean addr_len == sizeof(*sa)? That's what the
'if (addr_len < sizeof(*sa) + 1) return -EINVAL' above is for.
- Eric
Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
On Mon, Oct 26, 2020 at 9:08 PM Eric Biggers wrote:
> Commit 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> names") made the kernel start accepting arbitrarily long algorithm names
> in sockaddr_alg.
That's not true; it's still limited by the size of struct
sockaddr_storage (128 bytes total for the entire address). If you make
it longer, __copy_msghdr_from_user() will silently truncate the size.
> This is broken because the kernel can access indices >= 64 in salg_name,
> which is undefined behavior -- even though the memory that is accessed
> is still located within the sockaddr structure. It would only be
> defined behavior if the array were properly marked as arbitrary-length
> (either by making it a flexible array, which is the recommended way
> these days, or by making it an array of length 0 or 1).
>
> We can't simply change salg_name into a flexible array, since that would
> break source compatibility with userspace programs that embed
> sockaddr_alg into another struct, or (more commonly) declare a
> sockaddr_alg like 'struct sockaddr_alg sa = { .salg_name = "foo" };'.
>
> One solution would be to change salg_name into a flexible array only
> when '#ifdef __KERNEL__'. However, that would keep userspace without an
> easy way to actually use the longer algorithm names.
>
> Instead, add a new structure 'sockaddr_alg_new' that has the flexible
> array field, and expose it to both userspace and the kernel.
> Make the kernel use it correctly in alg_bind().
[...]
> @@ -147,7 +147,7 @@ static int alg_bind(struct socket *sock, struct sockaddr
> *uaddr, int addr_len)
> const u32 allowed = CRYPTO_ALG_KERN_DRIVER_ONLY;
> struct sock *sk = sock->sk;
> struct alg_sock *ask = alg_sk(sk);
> - struct sockaddr_alg *sa = (void *)uaddr;
> + struct sockaddr_alg_new *sa = (void *)uaddr;
> const struct af_alg_type *type;
> void *private;
> int err;
> @@ -155,7 +155,11 @@ static int alg_bind(struct socket *sock, struct sockaddr
> *uaddr, int addr_len)
> if (sock->state == SS_CONNECTED)
> return -EINVAL;
>
> - if (addr_len < sizeof(*sa))
> + BUILD_BUG_ON(offsetof(struct sockaddr_alg_new, salg_name) !=
> +offsetof(struct sockaddr_alg, salg_name));
> + BUILD_BUG_ON(offsetof(struct sockaddr_alg, salg_name) != sizeof(*sa));
> +
> + if (addr_len < sizeof(*sa) + 1)
> return -EINVAL;
>
> /* If caller uses non-allowed flag, return error. */
> @@ -163,7 +167,7 @@ static int alg_bind(struct socket *sock, struct sockaddr
> *uaddr, int addr_len)
> return -EINVAL;
>
> sa->salg_type[sizeof(sa->salg_type) - 1] = 0;
> - sa->salg_name[sizeof(sa->salg_name) + addr_len - sizeof(*sa) - 1] = 0;
> + sa->salg_name[addr_len - sizeof(*sa) - 1] = 0;
This looks like an out-of-bounds write in the case `addr_len ==
sizeof(struct sockaddr_storage)`.
Re: [PATCH] crypto: af_alg - avoid undefined behavior accessing salg_name
Hi,
On Mon, Oct 26, 2020 at 01:07:15PM -0700, Eric Biggers wrote:
> From: Eric Biggers
>
> Commit 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> names") made the kernel start accepting arbitrarily long algorithm names
> in sockaddr_alg. However, the actual length of the salg_name field
> stayed at the original 64 bytes.
>
> This is broken because the kernel can access indices >= 64 in salg_name,
> which is undefined behavior -- even though the memory that is accessed
> is still located within the sockaddr structure. It would only be
> defined behavior if the array were properly marked as arbitrary-length
> (either by making it a flexible array, which is the recommended way
> these days, or by making it an array of length 0 or 1).
>
> We can't simply change salg_name into a flexible array, since that would
> break source compatibility with userspace programs that embed
> sockaddr_alg into another struct, or (more commonly) declare a
> sockaddr_alg like 'struct sockaddr_alg sa = { .salg_name = "foo" };'.
>
> One solution would be to change salg_name into a flexible array only
> when '#ifdef __KERNEL__'. However, that would keep userspace without an
> easy way to actually use the longer algorithm names.
>
> Instead, add a new structure 'sockaddr_alg_new' that has the flexible
> array field, and expose it to both userspace and the kernel.
> Make the kernel use it correctly in alg_bind().
>
> This addresses the syzbot report
> "UBSAN: array-index-out-of-bounds in alg_bind"
> (https://syzkaller.appspot.com/bug?extid=92ead4eb8e26a26d465e).
>
> Reported-by: syzbot+92ead4eb8e26a26d4...@syzkaller.appspotmail.com
> Fixes: 3f69cc60768b ("crypto: af_alg - Allow arbitrarily long algorithm
> names")
> Cc: # v4.12+
> Signed-off-by: Eric Biggers
> ---
> crypto/af_alg.c | 10 +++---
> include/uapi/linux/if_alg.h | 16
> 2 files changed, 23 insertions(+), 3 deletions(-)
>
> diff --git a/crypto/af_alg.c b/crypto/af_alg.c
> index d11db80d24cd1..9acb9d2c4bcf9 100644
> --- a/crypto/af_alg.c
> +++ b/crypto/af_alg.c
> @@ -147,7 +147,7 @@ static int alg_bind(struct socket *sock, struct sockaddr
> *uaddr, int addr_len)
> const u32 allowed = CRYPTO_ALG_KERN_DRIVER_ONLY;
> struct sock *sk = sock->sk;
> struct alg_sock *ask = alg_sk(sk);
> - struct sockaddr_alg *sa = (void *)uaddr;
> + struct sockaddr_alg_new *sa = (void *)uaddr;
> const struct af_alg_type *type;
> void *private;
> int err;
> @@ -155,7 +155,11 @@ static int alg_bind(struct socket *sock, struct sockaddr
> *uaddr, int addr_len)
> if (sock->state == SS_CONNECTED)
> return -EINVAL;
>
> - if (addr_len < sizeof(*sa))
> + BUILD_BUG_ON(offsetof(struct sockaddr_alg_new, salg_name) !=
> + offsetof(struct sockaddr_alg, salg_name));
> + BUILD_BUG_ON(offsetof(struct sockaddr_alg, salg_name) != sizeof(*sa));
> +
> + if (addr_len < sizeof(*sa) + 1)
> return -EINVAL;
>
> /* If caller uses non-allowed flag, return error. */
> @@ -163,7 +167,7 @@ static int alg_bind(struct socket *sock, struct sockaddr
> *uaddr, int addr_len)
> return -EINVAL;
>
> sa->salg_type[sizeof(sa->salg_type) - 1] = 0;
> - sa->salg_name[sizeof(sa->salg_name) + addr_len - sizeof(*sa) - 1] = 0;
> + sa->salg_name[addr_len - sizeof(*sa) - 1] = 0;
>
> type = alg_get_type(sa->salg_type);
> if (PTR_ERR(type) == -ENOENT) {
> diff --git a/include/uapi/linux/if_alg.h b/include/uapi/linux/if_alg.h
> index 60b7c2efd921c..dc52a11ba6d15 100644
> --- a/include/uapi/linux/if_alg.h
> +++ b/include/uapi/linux/if_alg.h
> @@ -24,6 +24,22 @@ struct sockaddr_alg {
> __u8salg_name[64];
> };
>
> +/*
> + * Linux v4.12 and later removed the 64-byte limit on salg_name[]; it's now
> an
> + * arbitrary-length field. We had to keep the original struct above for
> source
> + * compatibility with existing userspace programs, though. Use the new
> struct
> + * below if support for very long algorithm names is needed. To do this,
> + * allocate 'sizeof(struct sockaddr_alg_new) + strlen(algname) + 1' bytes,
> and
> + * copy algname (including the null terminator) into salg_name.
> + */
> +struct sockaddr_alg_new {
> + __u16 salg_family;
> + __u8salg_type[14];
> + __u32 salg_feat;
> + __u32 salg_mask;
> + __u8salg_name[];
> +};
> +
How something like this, instead:
struct sockaddr_alg {
- __u16 salg_family;
- __u8salg_type[14];
- __u32 salg_feat;
- __u32 salg_mask;
- __u8salg_name[64];
+ union {
+ struct {
+ __u16 salg_v1_family;
+ __u8salg_v1_type[14];
+ __u32 salg_v1_feat;
+ __u32 salg_v1_mask;
+ __u8salg_name[64];
+ };
+ struct {
+
