Re: [edk2-devel] [PATCH v7 0/5] Use RngLib instead of TimerLib for OpensslLib

["Michael D Kinney" <michael.d.kinney@intel.com>]

Hi Matt,

BaseRngLibTimerLib
===================
Thank you for updating BaseRngLibTimerLib to use GetPerformanceCounterProperties().
StartValue and EndValue are OPTIONAL, so the function DecentDelay() can be simplified
to remove the StartValue and EndValue local variables and get the rate of the counter
using the following:

  // Get the counter properties
  CounterHz = GetPerformanceCounterProperties (NULL, NULL);

When you compute the min delay, I see the formula will generate a value of 0 when
the rate of the performance counter is greater than 1.5MHz.  MicroSecondDelay()
may return immediately if MicroSeconds is 0.  Is this your intended behavior?
Or did you want to make sure the min value is 1 such as:

  MinumumDelayInMicroSeconds = MAX (1500000 / CounterHz, 1);

CounterHz is also type UINT64 so this is a 64-bit divide operation that must
use the BaseLib function DivU64x64Remainder() for 32-bit builds.

  MinumumDelayInMicroSeconds = MAX (DivU64x64Remainder (1500000, CounterHz, NULL), 1);

The function DecentDelay() may interact with HW to get the performance counter
rate and then do the divide operation.  For the RngLib APIs that need the delay,
I recommend you call DecentDelay() to get the MinumumDelayInMicroSeconds into
a local variable and then use that value for calls to MicroSecondDelay() in the
RngLib APIs.

The comments in the RngLib APIs that describe the length of the delays in uS/mS
need to be updated because the length of the delay is computed.  Update with
a more generic comment to perform a minimum delay to guarantee a different
performance counter value. 

The UNI file header and strings need to be updated to match INF/C files.


DxeRngLib
==========
1) Please add a UNI file for this lib.

Best regards,

Mike

> -----Original Message-----
> From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Matthew Carlson
> Sent: Thursday, August 13, 2020 12:45 PM
> To: devel@edk2.groups.io
> Cc: Ard Biesheuvel <ard.biesheuvel@arm.com>; Anthony Perard <anthony.perard@citrix.com>; Yao, Jiewen
> <jiewen.yao@intel.com>; Wang, Jian J <jian.j.wang@intel.com>; Julien Grall <julien@xen.org>; Justen, Jordan L
> <jordan.l.justen@intel.com>; Laszlo Ersek <lersek@redhat.com>; Gao, Liming <liming.gao@intel.com>; Leif Lindholm
> <leif@nuviainc.com>; Kinney, Michael D <michael.d.kinney@intel.com>; Lu, XiaoyuX <xiaoyux.lu@intel.com>; Liu, Zhiguang
> <zhiguang.liu@intel.com>; Sean Brogan <sean.brogan@microsoft.com>; Matthew Carlson <matthewfcarlson@gmail.com>
> Subject: [edk2-devel] [PATCH v7 0/5] Use RngLib instead of TimerLib for OpensslLib
> 
> From: Matthew Carlson <macarl@microsoft.com>
> 
> Hello all,
> 
> This patch contains a fix for Bugzilla 1871.
> There's been a good bit of community discussion around the topic,
> so below follows a general overview of the discussion and what this patch does.
> 
> This is the seventh iteration of this patch series, focused on code style and a
> few functions being renamed to comply with style.
> 
> Back in Devel message#40590 (https://edk2.groups.io/g/devel/message/40590)
> around the patch series that updates OpenSSL to 1.1.1b, a comment was made
> that suggested that platforms be in charge of the entropy/randomness that
> is provided to OpenSSL as currently the entropry source seems to be a
> hand-rolled random number generator that uses the PerformanceCounter from
> TimerLib. This causes OpenSSL to depend on TimerLib, which is often platform
> specific. In addition to being a potentially weaker source of randomness,
> this also poses a challenge to compile BaseCryptLibOnProtocol with a platform-
> agnostic version of TimerLib that works universally.
> 
> The solution here is to allow platform to specify their source of entropy in
> addition to providing two new RngLibs: one that uses the TimerLib as well as
> one that uses RngProtocol to provide randomness. Then the decision to use
> RDRAND or other entropy sources is up to the platform. Mixing various entropy
> sources is the onus of the platform. It has been suggested on Devel#40590 and
> BZ#1871 that there should be mixing of the PerformanceCounter and RDRAND using
> something similar to the yarrow alogirthm that FreeBSD uses for example. This
> patch series doesn't offer an RngLib that offers that sort of mixing as the
> ultimate source of random is defined by the platform.
> 
> This patch series offers three benefits:
> 1. Dependency reduction: Removes the need for a platform specific timer
> library.  We publish a single binary used on numerous platforms for
> crypto and the introduced timer lib dependency caused issues because we
> could not fulfill our platform needs with one library instance.
> 
> 2. Code maintenance: Removing this additional code and leveraging an existing
> library within Edk2 means less code to maintain.
> 
> 3. Platform defined quality: A platform can choose which instance to use and
> the implications of that instance.
> 
> This patch series seeks to address five seperate issues.
>   1) Use RngLib interface to generate random entropy in rand_pool
>   2) Remove dependency on TimerLib in OpensslLib
>   3) Add a new version of RngLib implemented by TimerLib
>   4) Add a new version of RngLib implemented by EFI_RNG_PROTOCOL
>   5) Add RngLib to platforms in EDK2 such as ArmVirtPkg and OvmfPkg
> 
> Since this changes the dependencies of OpenSSL, this has the potential of being
> a breaking change for platforms in edk2-platforms. The easiest solution is just
> to use the RngLib that uses the TimerLib as this closely mimics the behavior of
> OpenSSL prior to this patch series. There is also a null version of RngLib for
> CI environments that need this change
> (https://edk2.groups.io/g/devel/message/50432). Though it should be pointed out
> that in CI environments, the null version of BaseCryptLib or OpenSSL should be
> used.
> 
> In addition, it has been suggested that
> 1) Add AsmRdSeed to BaseLib.
> 2) Update BaseRngLib to use AsmRdSeed() for the random number,
> if RdSeed is supported (CPUID BIT18)
> 
> However, this is largely out of scope for this particular patch series and
> will likely need to be in a follow-up series later.
> 
> It is my understanding that the OpenSSL code uses the values provided as a
> randomness pool rather than a seed or random numbers itself, so the
> requirements for randomness are not quite as stringent as other applications.
> 
> For the ArmVirtPkg and OvmfPkg platforms, the patch series here just adds in
> the TimerLib based RngLib as that is similar to the functionality of before.
> It is added as a common library so any custom RngLib defined in the DSC
> should take precedence over the TimerLibRngLib.
> 
> Ref: https://github.com/tianocore/edk2/pull/845
> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=1871
> 
> Cc: Ard Biesheuvel <ard.biesheuvel@arm.com>
> Cc: Anthony Perard <anthony.perard@citrix.com>
> Cc: Jiewen Yao <jiewen.yao@intel.com>
> Cc: Jian J Wang <jian.j.wang@intel.com>
> Cc: Julien Grall <julien@xen.org>
> Cc: Jordan Justen <jordan.l.justen@intel.com>
> Cc: Laszlo Ersek <lersek@redhat.com>
> Cc: Liming Gao <liming.gao@intel.com>
> Cc: Leif Lindholm <leif@nuviainc.com>
> Cc: Michael D Kinney <michael.d.kinney@intel.com>
> Cc: Xiaoyu Lu <xiaoyux.lu@intel.com>
> Cc: Zhiguang Liu <zhiguang.liu@intel.com>
> Cc: Sean Brogan <sean.brogan@microsoft.com>
> 
> Signed-off-by: Matthew Carlson <matthewfcarlson@gmail.com>
> 
> 
> Matthew Carlson (5):
>   MdePkg: TimerRngLib: Added RngLib that uses TimerLib
>   MdePkg: BaseRngLibDxe: Add RngLib that uses RngDxe
>   OvmfPkg: Add RngLib based on TimerLib for Crypto
>   ArmVirtPkg: Add RngLib based on TimerLib for CryptoPkg
>   CryptoPkg: OpensslLib: Use RngLib to generate entropy in rand_pool
> 
>  CryptoPkg/Library/OpensslLib/rand_pool.c                 | 265 +++++---------------
>  CryptoPkg/Library/OpensslLib/rand_pool_noise.c           |  29 ---
>  CryptoPkg/Library/OpensslLib/rand_pool_noise_tsc.c       |  43 ----
>  MdePkg/Library/BaseRngLibTimerLib/RngLibTimer.c          | 188 ++++++++++++++
>  MdePkg/Library/DxeRngLib/DxeRngLib.c                     | 206 +++++++++++++++
>  ArmVirtPkg/ArmVirt.dsc.inc                               |   1 +
>  CryptoPkg/CryptoPkg.dsc                                  |   1 +
>  CryptoPkg/Library/OpensslLib/OpensslLib.inf              |  15 +-
>  CryptoPkg/Library/OpensslLib/OpensslLibCrypto.inf        |  15 +-
>  CryptoPkg/Library/OpensslLib/rand_pool_noise.h           |  29 ---
>  MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf |  36 +++
>  MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.uni |  15 ++
>  MdePkg/Library/DxeRngLib/DxeRngLib.inf                   |  37 +++
>  MdePkg/MdePkg.dsc                                        |   5 +-
>  OvmfPkg/Bhyve/BhyvePkgX64.dsc                            |   1 +
>  OvmfPkg/OvmfPkgIa32.dsc                                  |   1 +
>  OvmfPkg/OvmfPkgIa32X64.dsc                               |   1 +
>  OvmfPkg/OvmfPkgX64.dsc                                   |   1 +
>  OvmfPkg/OvmfXen.dsc                                      |   1 +
>  19 files changed, 555 insertions(+), 335 deletions(-)
>  delete mode 100644 CryptoPkg/Library/OpensslLib/rand_pool_noise.c
>  delete mode 100644 CryptoPkg/Library/OpensslLib/rand_pool_noise_tsc.c
>  create mode 100644 MdePkg/Library/BaseRngLibTimerLib/RngLibTimer.c
>  create mode 100644 MdePkg/Library/DxeRngLib/DxeRngLib.c
>  delete mode 100644 CryptoPkg/Library/OpensslLib/rand_pool_noise.h
>  create mode 100644 MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf
>  create mode 100644 MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.uni
>  create mode 100644 MdePkg/Library/DxeRngLib/DxeRngLib.inf
> 
> --
> 2.27.0.windows.1
> 
> 
>