Re: [PATCH v2] cryptodev: fix NULL pointer dereference

On 01/08/2017 10:35, Thomas Monjalon wrote:

01/08/2017 10:13, Sergio Gonzalez Monroy:

quoted

On 31/07/2017 20:33, Thomas Monjalon wrote:

quoted

31/07/2017 11:18, Pablo de Lara:

quoted

When register a crypto driver, a cryptodev driver
structure was being allocated, using malloc.
Since this call may fail, it is safer to allocate
this memory statically in each PMD, so driver registration
will never fail.

Coverity issue: 158645

Fixes: 7a364faef185 ("cryptodev: remove crypto device type enumeration")

Signed-off-by: Pablo de Lara <redacted>
---

Changes in v2:

- Allocate statically the cryptodev driver structure,
    instead of using malloc, that can potentially fail.

   drivers/crypto/aesni_gcm/aesni_gcm_pmd.c    |  5 ++++-
   drivers/crypto/aesni_mb/rte_aesni_mb_pmd.c  |  6 +++++-
   drivers/crypto/armv8/rte_armv8_pmd.c        |  9 ++++++---
   drivers/crypto/dpaa2_sec/dpaa2_sec_dpseci.c |  5 ++++-
   drivers/crypto/kasumi/rte_kasumi_pmd.c      |  5 ++++-
   drivers/crypto/null/null_crypto_pmd.c       |  5 ++++-
   drivers/crypto/openssl/rte_openssl_pmd.c    |  5 ++++-
   drivers/crypto/qat/rte_qat_cryptodev.c      |  7 +++++--
   drivers/crypto/scheduler/scheduler_pmd.c    |  5 ++++-
   drivers/crypto/snow3g/rte_snow3g_pmd.c      |  5 ++++-
   drivers/crypto/zuc/rte_zuc_pmd.c            |  5 ++++-
   lib/librte_cryptodev/rte_cryptodev.c        | 18 +++++------------
   lib/librte_cryptodev/rte_cryptodev.h        | 20 -------------------
   lib/librte_cryptodev/rte_cryptodev_pmd.h    | 30 +++++++++++++++++++++++++++++
   14 files changed, 83 insertions(+), 47 deletions(-)

This is a big change for a small/unlikely issue.
The main benefit of this patch is an allocation cleanup.
I think it is better to wait 17.11 cycle to integrate it.

We initially thought of exit given that it is a constructor and if you
fail to allocate memory at this stage, things are likely not going to
work out anyway.

You don't know how the application wants to manage it.

IMHO setting an internal variable indicating an error in constructors 
and then reporting the problem during EAL init seems overly complex.
I think the proposed change is a cleaner solution.

quoted

The patch is an API change, do we really want to break again (we are
breaking in this release) next release?

Good question. Any opinions?

Merge the patch unless there are already outstanding and/or planned 
changes for the next release that are going to break ABI/API?


Thanks,
Sergio