[Drbd-dev] [PATCH] drivers/block/drbd: add NULL test around call to crypto_free_hash

[walter harms]

Am 31.01.2011 18:51, schrieb Julia Lawall:
> crypto_free_hash calls the function crypto_hash_tfm and then
> crypto_free_tfm on the result.  crypto_free_tfm calls crypto_destroy_tfm,
> which tests this result for NULL and then dereferences it.  crypto_hash_tfm
> returns &tfm->base where tfm is its argument.  base is actually the first
> and only field of a crypto_hash-typed structure, so perhaps one can rely on
> it to return NULL for a NULL value of tfm.  But most calls to
> crypto_hash_tfm where the argument might be NULL don't rely on this
> property and test for NULL explicitly.
> 
> The semantic match that finds this problem is as follows:
> (http://coccinelle.lip6.fr/)
> 
> // <smpl>
> @safe@
> position p;
> expression x;
> @@
> 
> if (x) { <+... crypto_free_hash at p(x) ...+> }
> 
> @@
> expression x;
> position p!=safe.p;
> @@
> 
> *x = NULL
> ...
> *crypto_free_hash at p(x)
> // </smpl>
> 
> Signed-off-by: Julia Lawall <julia at diku.dk>
> 
> ---
>  drivers/block/drbd/drbd_nl.c       |   18 ++++++++++++------
>  drivers/block/drbd/drbd_receiver.c |    6 ++++--
>  2 files changed, 16 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
> index 8cbfaa6..aa5fbc0 100644
> --- a/drivers/block/drbd/drbd_nl.c
> +++ b/drivers/block/drbd/drbd_nl.c
> @@ -1482,13 +1482,16 @@ static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
>  		mdev->ee_hash = new_ee_hash;
>  	}
>  
> -	crypto_free_hash(mdev->cram_hmac_tfm);
> +	if (mdev->cram_hmac_tfm)
> +		crypto_free_hash(mdev->cram_hmac_tfm);
>  	mdev->cram_hmac_tfm = tfm;
>  
> -	crypto_free_hash(mdev->integrity_w_tfm);
> +	if (mdev->integrity_w_tfm)
> +		crypto_free_hash(mdev->integrity_w_tfm);
>  	mdev->integrity_w_tfm = integrity_w_tfm;
>  
> -	crypto_free_hash(mdev->integrity_r_tfm);
> +	if (mdev->integrity_r_tfm)
> +		crypto_free_hash(mdev->integrity_r_tfm);
>  	mdev->integrity_r_tfm = integrity_r_tfm;
>  
>  	kfree(mdev->int_dig_out);
> @@ -1509,9 +1512,12 @@ fail:
>  	kfree(int_dig_out);
>  	kfree(int_dig_in);
>  	kfree(int_dig_vv);
> -	crypto_free_hash(tfm);
> -	crypto_free_hash(integrity_w_tfm);
> -	crypto_free_hash(integrity_r_tfm);
> +	if (tfm)
> +		crypto_free_hash(tfm);
> +	if (integrity_w_tfm)
> +		crypto_free_hash(integrity_w_tfm);
> +	if (integrity_r_tfm)
> +		crypto_free_hash(integrity_r_tfm);
>  	kfree(new_tl_hash);
>  	kfree(new_ee_hash);
>  	kfree(new_conf);
> diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
> index 24487d4..3453cc3 100644
> --- a/drivers/block/drbd/drbd_receiver.c
> +++ b/drivers/block/drbd/drbd_receiver.c
> @@ -2871,9 +2871,11 @@ static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packets cmd, unsi
>  disconnect:
>  	/* just for completeness: actually not needed,
>  	 * as this is not reached if csums_tfm was ok. */
> -	crypto_free_hash(csums_tfm);
> +	if (csums_tfm)
> +		crypto_free_hash(csums_tfm);
>  	/* but free the verify_tfm again, if csums_tfm did not work out */
> -	crypto_free_hash(verify_tfm);
> +	if (verify_tfm)
> +		crypto_free_hash(verify_tfm);
>  	drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
>  	return FALSE;
>  }
> 

it looks that it would be more sensibel to change crypto_free_hash() to handle
NULL that would be more consistent with the free() family.

just my 2 cents,
re,
 wh