Re: [dm-devel] [PATCH 1/3 v4] crypto: introduce the flag CRYPTO_ALG_ALLOCATES_MEMORY

2020-06-30 Thread Mikulas Patocka



On Tue, 30 Jun 2020, Eric Biggers wrote:

> On Tue, Jun 30, 2020 at 01:01:16PM -0400, Mikulas Patocka wrote:
> > > diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c
> > > index 7240e8bbdebb..643f7f1cc91c 100644
> > > --- a/crypto/pcrypt.c
> > > +++ b/crypto/pcrypt.c
> > > @@ -232,12 +232,15 @@ static int pcrypt_create_aead(struct 
> > > crypto_template *tmpl, struct rtattr **tb,
> > >   struct crypto_attr_type *algt;
> > >   struct aead_instance *inst;
> > >   struct aead_alg *alg;
> > > + u32 mask;
> > >   int err;
> > >  
> > >   algt = crypto_get_attr_type(tb);
> > >   if (IS_ERR(algt))
> > >   return PTR_ERR(algt);
> > >  
> > > + mask = crypto_alg_inherited_mask(algt->type, algt->mask);
> > > +
> > >   inst = kzalloc(sizeof(*inst) + sizeof(*ctx), GFP_KERNEL);
> > >   if (!inst)
> > >   return -ENOMEM;
> > > @@ -254,7 +257,7 @@ static int pcrypt_create_aead(struct crypto_template 
> > > *tmpl, struct rtattr **tb,
> > >   goto err_free_inst;
> > >  
> > >   err = crypto_grab_aead(>spawn, aead_crypto_instance(inst),
> > > -crypto_attr_alg_name(tb[1]), 0, 0);
> > > +crypto_attr_alg_name(tb[1]), 0, mask);
> > >   if (err)
> > >   goto err_free_inst;
> > >  
> > 
> > I added "mask" there - but there is still a "mask" argument that is 
> > unused - is it a bug to have two "mask" variables?
> 
> Right, I didn't see that algt->type and algt->mask are already being passed 
> to 
> pcrypt_create_aead().  It's redundant because pcrypt_create_aead() has access 
> to
> those via crypto_get_attr_type() anyway.
> 
> How about just removing those two arguments for now?
> 
> - Eric

Yes.

Mikulas



Re: [dm-devel] [PATCH 1/3 v4] crypto: introduce the flag CRYPTO_ALG_ALLOCATES_MEMORY

2020-06-30 Thread Eric Biggers
On Tue, Jun 30, 2020 at 01:01:16PM -0400, Mikulas Patocka wrote:
> > diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c
> > index 7240e8bbdebb..643f7f1cc91c 100644
> > --- a/crypto/pcrypt.c
> > +++ b/crypto/pcrypt.c
> > @@ -232,12 +232,15 @@ static int pcrypt_create_aead(struct crypto_template 
> > *tmpl, struct rtattr **tb,
> > struct crypto_attr_type *algt;
> > struct aead_instance *inst;
> > struct aead_alg *alg;
> > +   u32 mask;
> > int err;
> >  
> > algt = crypto_get_attr_type(tb);
> > if (IS_ERR(algt))
> > return PTR_ERR(algt);
> >  
> > +   mask = crypto_alg_inherited_mask(algt->type, algt->mask);
> > +
> > inst = kzalloc(sizeof(*inst) + sizeof(*ctx), GFP_KERNEL);
> > if (!inst)
> > return -ENOMEM;
> > @@ -254,7 +257,7 @@ static int pcrypt_create_aead(struct crypto_template 
> > *tmpl, struct rtattr **tb,
> > goto err_free_inst;
> >  
> > err = crypto_grab_aead(>spawn, aead_crypto_instance(inst),
> > -  crypto_attr_alg_name(tb[1]), 0, 0);
> > +  crypto_attr_alg_name(tb[1]), 0, mask);
> > if (err)
> > goto err_free_inst;
> >  
> 
> I added "mask" there - but there is still a "mask" argument that is 
> unused - is it a bug to have two "mask" variables?

Right, I didn't see that algt->type and algt->mask are already being passed to 
pcrypt_create_aead().  It's redundant because pcrypt_create_aead() has access to
those via crypto_get_attr_type() anyway.

How about just removing those two arguments for now?

- Eric


Re: [dm-devel] [PATCH 1/3 v4] crypto: introduce the flag CRYPTO_ALG_ALLOCATES_MEMORY

2020-06-30 Thread Mikulas Patocka



On Tue, 30 Jun 2020, Eric Biggers wrote:

> On Tue, Jun 30, 2020 at 09:56:22AM -0400, Mikulas Patocka wrote:
> > Index: linux-2.6/crypto/cryptd.c
> > ===
> > --- linux-2.6.orig/crypto/cryptd.c  2020-06-29 16:03:07.346417000 +0200
> > +++ linux-2.6/crypto/cryptd.c   2020-06-30 15:49:04.206417000 +0200
> > @@ -202,6 +202,7 @@ static inline void cryptd_check_internal
> >  
> > *type |= algt->type & CRYPTO_ALG_INTERNAL;
> > *mask |= algt->mask & CRYPTO_ALG_INTERNAL;
> > +   *mask |= algt->mask & CRYPTO_ALG_INHERITED_FLAGS;
> >  }
> 
> This needs to use the correct logic (crypto_alg_inherited_mask) to decide 
> which
> inherited flags to set in 'mask'.

I added "*mask |= crypto_alg_inherited_mask(algt->type, algt->mask);" 
there.


> > --- linux-2.6.orig/crypto/adiantum.c2020-06-29 16:03:07.346417000 
> > +0200
> > +++ linux-2.6/crypto/adiantum.c 2020-06-29 16:03:07.346417000 +0200
> > @@ -507,7 +507,7 @@ static int adiantum_create(struct crypto
> > if ((algt->type ^ CRYPTO_ALG_TYPE_SKCIPHER) & algt->mask)
> > return -EINVAL;
> >  
> > -   mask = crypto_requires_sync(algt->type, algt->mask);
> > +   mask = crypto_alg_inherited_mask(algt->type, algt->mask);
> >  
> > inst = kzalloc(sizeof(*inst) + sizeof(*ictx), GFP_KERNEL);
> > if (!inst)
> 
> This is still missing setting the flags correctly on the template instance 
> being
> constructed.  The flags need to be inherited from all "inner" algorithms:
> 
>   inst->alg.base.cra_flags = (streamcipher_alg->base.cra_flags |
>   blockcipher_alg->cra_flags |
>   hash_alg->base.cra_flags) &
>  CRYPTO_ALG_INHERITED_FLAGS;
> 
> If we don't do that, the template instance may allocate memory but not have
> CRYPTO_ALG_ALLOCATES_MEMORY set.

OK


> > Index: linux-2.6/crypto/pcrypt.c
> > ===
> > --- linux-2.6.orig/crypto/pcrypt.c  2020-06-29 16:03:07.346417000 +0200
> > +++ linux-2.6/crypto/pcrypt.c   2020-06-30 15:47:32.776417000 +0200
> > @@ -263,7 +263,9 @@ static int pcrypt_create_aead(struct cry
> > if (err)
> > goto err_free_inst;
> >  
> > -   inst->alg.base.cra_flags = CRYPTO_ALG_ASYNC;
> > +   inst->alg.base.cra_flags =
> > +   CRYPTO_ALG_ASYNC |
> > +   (alg->base.cra_flags & CRYPTO_ALG_INHERITED_FLAGS);
> >  
> > inst->alg.ivsize = crypto_aead_alg_ivsize(alg);
> > inst->alg.maxauthsize = crypto_aead_alg_maxauthsize(alg);

What's wrong with this?


> And this is still missing setting the mask:
> 
> diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c
> index 7240e8bbdebb..643f7f1cc91c 100644
> --- a/crypto/pcrypt.c
> +++ b/crypto/pcrypt.c
> @@ -232,12 +232,15 @@ static int pcrypt_create_aead(struct crypto_template 
> *tmpl, struct rtattr **tb,
>   struct crypto_attr_type *algt;
>   struct aead_instance *inst;
>   struct aead_alg *alg;
> + u32 mask;
>   int err;
>  
>   algt = crypto_get_attr_type(tb);
>   if (IS_ERR(algt))
>   return PTR_ERR(algt);
>  
> + mask = crypto_alg_inherited_mask(algt->type, algt->mask);
> +
>   inst = kzalloc(sizeof(*inst) + sizeof(*ctx), GFP_KERNEL);
>   if (!inst)
>   return -ENOMEM;
> @@ -254,7 +257,7 @@ static int pcrypt_create_aead(struct crypto_template 
> *tmpl, struct rtattr **tb,
>   goto err_free_inst;
>  
>   err = crypto_grab_aead(>spawn, aead_crypto_instance(inst),
> -crypto_attr_alg_name(tb[1]), 0, 0);
> +crypto_attr_alg_name(tb[1]), 0, mask);
>   if (err)
>   goto err_free_inst;
>  

I added "mask" there - but there is still a "mask" argument that is 
unused - is it a bug to have two "mask" variables?


> Can you please think logically about what you're trying to accomplish?

I am not an expert in crypto code.

> In particular, if someone requests an algorithm that doesn't allocate memory
> (which is indicated by type=0, mask=CRYPTO_ALG_ALLOCATES_MEMORY), that request
> needs to be honored.
> 
> - Eric

Mikulas



Re: [dm-devel] [PATCH 1/3 v4] crypto: introduce the flag CRYPTO_ALG_ALLOCATES_MEMORY

2020-06-30 Thread Eric Biggers
On Tue, Jun 30, 2020 at 09:56:22AM -0400, Mikulas Patocka wrote:
> Index: linux-2.6/crypto/cryptd.c
> ===
> --- linux-2.6.orig/crypto/cryptd.c2020-06-29 16:03:07.346417000 +0200
> +++ linux-2.6/crypto/cryptd.c 2020-06-30 15:49:04.206417000 +0200
> @@ -202,6 +202,7 @@ static inline void cryptd_check_internal
>  
>   *type |= algt->type & CRYPTO_ALG_INTERNAL;
>   *mask |= algt->mask & CRYPTO_ALG_INTERNAL;
> + *mask |= algt->mask & CRYPTO_ALG_INHERITED_FLAGS;
>  }

This needs to use the correct logic (crypto_alg_inherited_mask) to decide which
inherited flags to set in 'mask'.

> --- linux-2.6.orig/crypto/adiantum.c  2020-06-29 16:03:07.346417000 +0200
> +++ linux-2.6/crypto/adiantum.c   2020-06-29 16:03:07.346417000 +0200
> @@ -507,7 +507,7 @@ static int adiantum_create(struct crypto
>   if ((algt->type ^ CRYPTO_ALG_TYPE_SKCIPHER) & algt->mask)
>   return -EINVAL;
>  
> - mask = crypto_requires_sync(algt->type, algt->mask);
> + mask = crypto_alg_inherited_mask(algt->type, algt->mask);
>  
>   inst = kzalloc(sizeof(*inst) + sizeof(*ictx), GFP_KERNEL);
>   if (!inst)

This is still missing setting the flags correctly on the template instance being
constructed.  The flags need to be inherited from all "inner" algorithms:

inst->alg.base.cra_flags = (streamcipher_alg->base.cra_flags |
blockcipher_alg->cra_flags |
hash_alg->base.cra_flags) &
   CRYPTO_ALG_INHERITED_FLAGS;

If we don't do that, the template instance may allocate memory but not have
CRYPTO_ALG_ALLOCATES_MEMORY set.

> Index: linux-2.6/crypto/pcrypt.c
> ===
> --- linux-2.6.orig/crypto/pcrypt.c2020-06-29 16:03:07.346417000 +0200
> +++ linux-2.6/crypto/pcrypt.c 2020-06-30 15:47:32.776417000 +0200
> @@ -263,7 +263,9 @@ static int pcrypt_create_aead(struct cry
>   if (err)
>   goto err_free_inst;
>  
> - inst->alg.base.cra_flags = CRYPTO_ALG_ASYNC;
> + inst->alg.base.cra_flags =
> + CRYPTO_ALG_ASYNC |
> + (alg->base.cra_flags & CRYPTO_ALG_INHERITED_FLAGS);
>  
>   inst->alg.ivsize = crypto_aead_alg_ivsize(alg);
>   inst->alg.maxauthsize = crypto_aead_alg_maxauthsize(alg);

And this is still missing setting the mask:

diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c
index 7240e8bbdebb..643f7f1cc91c 100644
--- a/crypto/pcrypt.c
+++ b/crypto/pcrypt.c
@@ -232,12 +232,15 @@ static int pcrypt_create_aead(struct crypto_template 
*tmpl, struct rtattr **tb,
struct crypto_attr_type *algt;
struct aead_instance *inst;
struct aead_alg *alg;
+   u32 mask;
int err;
 
algt = crypto_get_attr_type(tb);
if (IS_ERR(algt))
return PTR_ERR(algt);
 
+   mask = crypto_alg_inherited_mask(algt->type, algt->mask);
+
inst = kzalloc(sizeof(*inst) + sizeof(*ctx), GFP_KERNEL);
if (!inst)
return -ENOMEM;
@@ -254,7 +257,7 @@ static int pcrypt_create_aead(struct crypto_template *tmpl, 
struct rtattr **tb,
goto err_free_inst;
 
err = crypto_grab_aead(>spawn, aead_crypto_instance(inst),
-  crypto_attr_alg_name(tb[1]), 0, 0);
+  crypto_attr_alg_name(tb[1]), 0, mask);
if (err)
goto err_free_inst;
 

Can you please think logically about what you're trying to accomplish?

In particular, if someone requests an algorithm that doesn't allocate memory
(which is indicated by type=0, mask=CRYPTO_ALG_ALLOCATES_MEMORY), that request
needs to be honored.

- Eric