2018-08-23 16:03:05

by Nicholas Mc Guire

[permalink] [raw]
Subject: [PATCH] cifs: check kmalloc before use

The kmalloc was not being checked - if it fails issue a warning
and return -ENOMEM to the caller.

Signed-off-by: Nicholas Mc Guire <[email protected]>
Fixes: b8da344b74c8 ("cifs: dynamic allocation of ntlmssp blob")
---

Problem was located with an experimental coccinelle script

Both call-sites will expect non-0 to indicate error so -ENOMEM
should be OK here. Setting buflen to 0 seems to be the expected
behavior on failure although it will not be checked/used at
the call-sites - probably this could be removed here as well as
in the handling of authentication error (preceding this change)
it was just retained for consistency.

Patch was compile tested with: x86_64_defconfig + CONFIG_CIFS=y
(with sparse and smatch warning unrelated to proposed change)

Patch is against 4.18 (localversion-next is next-20180823)

fs/cifs/sess.c | 6 ++++++
1 file changed, 6 insertions(+)

diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c
index 8b0502c..aa23c00 100644
--- a/fs/cifs/sess.c
+++ b/fs/cifs/sess.c
@@ -398,6 +398,12 @@ int build_ntlmssp_auth_blob(unsigned char **pbuffer,
goto setup_ntlmv2_ret;
}
*pbuffer = kmalloc(size_of_ntlmssp_blob(ses), GFP_KERNEL);
+ if (!*pbuffer) {
+ rc = -ENOMEM;
+ cifs_dbg(VFS, "Error %d during NTLMSSP allocation\n", rc);
+ *buflen = 0;
+ goto setup_ntlmv2_ret;
+ }
sec_blob = (AUTHENTICATE_MESSAGE *)*pbuffer;

memcpy(sec_blob->Signature, NTLMSSP_SIGNATURE, 8);
--
2.1.4



2018-08-23 20:08:01

by Steve French

[permalink] [raw]
Subject: Re: [PATCH] cifs: check kmalloc before use

merged into cifs-2.6.git for-next
On Thu, Aug 23, 2018 at 11:02 AM Nicholas Mc Guire <[email protected]> wrote:
>
> The kmalloc was not being checked - if it fails issue a warning
> and return -ENOMEM to the caller.
>
> Signed-off-by: Nicholas Mc Guire <[email protected]>
> Fixes: b8da344b74c8 ("cifs: dynamic allocation of ntlmssp blob")
> ---
>
> Problem was located with an experimental coccinelle script
>
> Both call-sites will expect non-0 to indicate error so -ENOMEM
> should be OK here. Setting buflen to 0 seems to be the expected
> behavior on failure although it will not be checked/used at
> the call-sites - probably this could be removed here as well as
> in the handling of authentication error (preceding this change)
> it was just retained for consistency.
>
> Patch was compile tested with: x86_64_defconfig + CONFIG_CIFS=y
> (with sparse and smatch warning unrelated to proposed change)
>
> Patch is against 4.18 (localversion-next is next-20180823)
>
> fs/cifs/sess.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c
> index 8b0502c..aa23c00 100644
> --- a/fs/cifs/sess.c
> +++ b/fs/cifs/sess.c
> @@ -398,6 +398,12 @@ int build_ntlmssp_auth_blob(unsigned char **pbuffer,
> goto setup_ntlmv2_ret;
> }
> *pbuffer = kmalloc(size_of_ntlmssp_blob(ses), GFP_KERNEL);
> + if (!*pbuffer) {
> + rc = -ENOMEM;
> + cifs_dbg(VFS, "Error %d during NTLMSSP allocation\n", rc);
> + *buflen = 0;
> + goto setup_ntlmv2_ret;
> + }
> sec_blob = (AUTHENTICATE_MESSAGE *)*pbuffer;
>
> memcpy(sec_blob->Signature, NTLMSSP_SIGNATURE, 8);
> --
> 2.1.4
>


--
Thanks,

Steve