| Message ID | 3394653.LODBE1uz62@tachyon.chronox.de (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Herbert Xu |
| Headers | show |
On Thu, Apr 02, 2015 at 07:05:27PM +0200, Stephan Mueller wrote: > > diff --git a/crypto/algapi.c b/crypto/algapi.c > index 83b04e0..215c604 100644 > --- a/crypto/algapi.c > +++ b/crypto/algapi.c > @@ -545,6 +545,10 @@ unlock: > goto err; > > crypto_wait_for_test(larval); > + > + /* Remove instance if test failed */ > + if (!(inst->alg.cra_flags & CRYPTO_ALG_TESTED)) > + crypto_unregister_instance(inst); Unfortunately I don't think this is safe because the moment you release the mutex the instance can be freed. So you'll need to hold a reference to it. Cheers,
Am Freitag, 3. April 2015, 17:58:28 schrieb Herbert Xu: Hi Herbert, > On Thu, Apr 02, 2015 at 07:05:27PM +0200, Stephan Mueller wrote: > > diff --git a/crypto/algapi.c b/crypto/algapi.c > > index 83b04e0..215c604 100644 > > --- a/crypto/algapi.c > > +++ b/crypto/algapi.c > > > > @@ -545,6 +545,10 @@ unlock: > > goto err; > > > > crypto_wait_for_test(larval); > > > > + > > + /* Remove instance if test failed */ > > + if (!(inst->alg.cra_flags & CRYPTO_ALG_TESTED)) > > + crypto_unregister_instance(inst); > > Unfortunately I don't think this is safe because the moment you > release the mutex the instance can be freed. So you'll need to > hold a reference to it. > > Cheers, Wpuldn't crypto_del_alg suffer from the same issue? I see that the cra_refcnt is checked. But I guess there would be the same kind of race?
On Fri, Apr 03, 2015 at 05:17:13PM +0200, Stephan Mueller wrote: > > Wpuldn't crypto_del_alg suffer from the same issue? I see that the cra_refcnt > is checked. But I guess there would be the same kind of race? You're quite right. It too needs to take a ref count on the algorithm in crypto_alg_match before releasing the mutex. Cheers,
Am Freitag, 3. April 2015, 23:52:46 schrieb Herbert Xu: Hi Herbert, > On Fri, Apr 03, 2015 at 05:17:13PM +0200, Stephan Mueller wrote: > > Wpuldn't crypto_del_alg suffer from the same issue? I see that the > > cra_refcnt is checked. But I guess there would be the same kind of race? > > You're quite right. It too needs to take a ref count on the > algorithm in crypto_alg_match before releasing the mutex. > What about moving the current crypto_unregister_instance into __crypto_unregister_instance and creating a new crypto_unregister_instance that takes the ref count and once it got it, it calls __crypto_unregister_instance. If it did not get the ref count, it simply returns because the inst apparently was already removed. Only the crypto_unregister_instance is EXPORT_SYMBOL whereas the __crypto_unregister_instance is static. > Cheers,
On Fri, Apr 03, 2015 at 09:29:44PM +0200, Stephan Mueller wrote: > > What about moving the current crypto_unregister_instance into > __crypto_unregister_instance and creating a new crypto_unregister_instance > that takes the ref count and once it got it, it calls > __crypto_unregister_instance. If it did not get the ref count, it simply > returns because the inst apparently was already removed. > > Only the crypto_unregister_instance is EXPORT_SYMBOL whereas the > __crypto_unregister_instance is static. That's not going to work because you need to grab the reference count before you release the mutex. Let me fix up crypto_user first and then you can use it for your case too (or at least borrow the idea). Cheers,
diff --git a/crypto/algapi.c b/crypto/algapi.c index 83b04e0..215c604 100644 --- a/crypto/algapi.c +++ b/crypto/algapi.c @@ -545,6 +545,10 @@ unlock: goto err; crypto_wait_for_test(larval); + + /* Remove instance if test failed */ + if (!(inst->alg.cra_flags & CRYPTO_ALG_TESTED)) + crypto_unregister_instance(inst); err = 0; err:
A cipher instance is added to the list of instances unconditionally regardless of whether the associated test failed. However, a failed test implies that during another lookup, the cipher instance will be added to the list again as it will not be found by the lookup code. That means that the list can be filled up with instances whose tests failed. Note: tests only fail in reality in FIPS mode when a cipher is not marked as fips_allowed=1. This can be seen with cmac(des3_ede) that does not have a fips_allowed=1. When allocating the cipher, the allocation fails with -ENOENT due to the missing fips_allowed=1 flag (which causes the testmgr to return EINVAL). Yet, the instance of cmac(des3_ede) is shown in /proc/crypto. Allocating the cipher again fails again, but a 2nd instance is listed in /proc/crypto. The patch simply de-registers the instance when the testing failed. Signed-off-by: Stephan Mueller <smueller@chronox.de> --- crypto/algapi.c | 4 ++++ 1 file changed, 4 insertions(+)