public inbox for devel@edk2.groups.io
 help / color / mirror / Atom feed
From: "Ard Biesheuvel" <ard.biesheuvel@arm.com>
To: matthewfcarlson@gmail.com, devel@edk2.groups.io
Cc: Michael D Kinney <michael.d.kinney@intel.com>,
	Liming Gao <liming.gao@intel.com>,
	Zhiguang Liu <zhiguang.liu@intel.com>
Subject: Re: [PATCH v6 1/5] MdePkg: TimerRngLib: Added RngLib that uses TimerLib
Date: Thu, 13 Aug 2020 14:09:28 +0200	[thread overview]
Message-ID: <d2b05ea0-fe26-a6e5-60db-ef52ca1bc69d@arm.com> (raw)
In-Reply-To: <20200812224338.287-2-matthewfcarlson@gmail.com>

On 8/13/20 12:43 AM, matthewfcarlson@gmail.com wrote:
> From: Matthew Carlson <macarl@microsoft.com>
> 
> Added a new RngLib that provides random numbers from the TimerLib
> using the performance counter. This is meant to be used for OpenSSL
> to replicate past behavior. This should not be used in production as
> a real source of entropy.
> 
> 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: Michael D Kinney <michael.d.kinney@intel.com>
> Cc: Liming Gao <liming.gao@intel.com>
> Cc: Zhiguang Liu <zhiguang.liu@intel.com>
> Signed-off-by: Matthew Carlson <matthewfcarlson@gmail.com>
> ---
>   MdePkg/Library/BaseRngLibTimerLib/RngLibTimer.c          | 187 ++++++++++++++++++++
>   MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf |  40 +++++
>   MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.uni |  17 ++
>   MdePkg/MdePkg.dsc                                        |   3 +-
>   4 files changed, 246 insertions(+), 1 deletion(-)
> 
> diff --git a/MdePkg/Library/BaseRngLibTimerLib/RngLibTimer.c b/MdePkg/Library/BaseRngLibTimerLib/RngLibTimer.c
> new file mode 100644
> index 000000000000..915382fb9278
> --- /dev/null
> +++ b/MdePkg/Library/BaseRngLibTimerLib/RngLibTimer.c
> @@ -0,0 +1,187 @@
> +/** @file
> +  BaseRng Library that uses the TimerLib to provide reasonably random numbers.
> +  Do not use this on a production system.
> +
> +  Copyright (c) Microsoft Corporation.
> +  SPDX-License-Identifier: BSD-2-Clause-Patent
> +**/
> +
> +#include <Base.h>
> +#include <Library/BaseLib.h>
> +#include <Library/DebugLib.h>
> +#include <Library/TimerLib.h>
> +
> +/**
> + * Using the TimerLib GetPerformanceCounterProperties() we delay
> + * for enough time for the PerformanceCounter to increment.
> + * Depending on your system
> + *
> + * If the return value from GetPerformanceCounterProperties (TimerLib)
> + * is zero, this function will not delay and attempt to assert.
> + */

Make this STATIC please

> +VOID
> +EFIAPI
> +DecentDelay(

space before (

> +  VOID
> +  )
> +{
> +  UINT64 StartValue;
> +  UINT64 EndValue;
> +  UINT64 CounterHz;
> +  UINT64 MinumumDelayInMicroSeconds;

newline here

> +  // Get the counter properties
> +  CounterHz = GetPerformanceCounterProperties(&StartValue, &EndValue);

space before (

> +  // Make sure we won't divide by zero
> +  if (CounterHz == 0) {
> +    ASSERT(FALSE); // Assert so the developer knows something is wrong

This will print

ASSERT (FALSE)

into the DEBUG log, whereas

ASSERT (CounterHz != 0)

will appear if you assert on the actual value, which is much more useful.

> +    return;
> +  }
> +  // Calculate the minimum delay based on 1.5 microseconds divided by the hertz.
> +  // We calculate the length of a cycle (1/CounterHz) and multiply it by 1.5 microseconds
> +  // This ensures that the performance counter has increased by at least one
> +  MinumumDelayInMicroSeconds = 1500000 / CounterHz;
> +
> +  MicroSecondDelay(MinumumDelayInMicroSeconds);

Space before (

> +}
> +
> +
> +/**
> +  Generates a 16-bit random number.
> +
> +  if Rand is NULL, then ASSERT().
> +
> +  @param[out] Rand     Buffer pointer to store the 16-bit random value.
> +
> +  @retval TRUE         Random number generated successfully.
> +  @retval FALSE        Failed to generate the random number.
> +
> +**/
> +BOOLEAN
> +EFIAPI
> +GetRandomNumber16 (
> +  OUT     UINT16                    *Rand
> +  )
> +{
> +  UINT32  Index;
> +  UINT8* RandPtr;

Please align the variable names vertically, and put the * on the right 
hand side.

> +
> +  ASSERT (Rand != NULL);
> +
> +  if (NULL == Rand) {

No yoda style comparisons please

> +    return FALSE;
> +  }
> +
> +  RandPtr = (UINT8 *) Rand;

Please drop the space after )

> +  // Get 2 bytes of random ish data
> +  // This should take around 10us
> +  for (Index = 0; Index < 2; Index ++) {
> +    *RandPtr = (UINT8) (GetPerformanceCounter () & 0xFF);

Same here

> +    DecentDelay (); // delay to give chance for performance counter to catch up

So the delay is intended to ensure that the perf counter assumes its 
prior value + 1? How does that help? I'd argue that this actually makes 
the value more predictable than simply taking 16 bits worth of 
performance counter bits.

Or was this code simply adopted from CryptoPkg in some way?

> +    RandPtr++;
> +  }
> +  return TRUE;
> +}
> +
> +/**
> +  Generates a 32-bit random number.
> +
> +  if Rand is NULL, then ASSERT().
> +
> +  @param[out] Rand     Buffer pointer to store the 32-bit random value.
> +
> +  @retval TRUE         Random number generated successfully.
> +  @retval FALSE        Failed to generate the random number.
> +
> +**/
> +BOOLEAN
> +EFIAPI
> +GetRandomNumber32 (
> +  OUT     UINT32                    *Rand
> +  )
> +{
> +  UINT32  Index;
> +  UINT8* RandPtr;
> +
> +  ASSERT (Rand != NULL);
> +
> +  if (NULL == Rand) {
> +    return FALSE;
> +  }
> +
> +  RandPtr = (UINT8 *) Rand;
> +  // Get 4 bytes of random ish data
> +  // This should take around 20ms
> +  for (Index = 0; Index < 4; Index ++) {
> +    *RandPtr = (UINT8) (GetPerformanceCounter () & 0xFF);
> +    DecentDelay (); // delay to give chance for performance counter to catch up
> +    RandPtr++;
> +  }
> +  return TRUE;
> +}
> +
> +/**
> +  Generates a 64-bit random number.
> +
> +  if Rand is NULL, then ASSERT().
> +
> +  @param[out] Rand     Buffer pointer to store the 64-bit random value.
> +
> +  @retval TRUE         Random number generated successfully.
> +  @retval FALSE        Failed to generate the random number.
> +
> +**/
> +BOOLEAN
> +EFIAPI
> +GetRandomNumber64 (
> +  OUT     UINT64                    *Rand
> +  )
> +{
> +  UINT32  Index;
> +  UINT8* RandPtr;
> +
> +  ASSERT (Rand != NULL);
> +
> +  if (NULL == Rand) {
> +    return FALSE;
> +  }
> +
> +  RandPtr = (UINT8 *) Rand;
> +  // Get 8 bytes of random ish data
> +  // This should take around 40ms
> +  for (Index = 0; Index < 8; Index ++) {
> +    *RandPtr = (UINT8) (GetPerformanceCounter () & 0xFF);
> +    DecentDelay (); // delay to give chance for performance counter to catch up
> +    RandPtr++;
> +  }
> +
> +  return TRUE;
> +}
> +
> +/**
> +  Generates a 128-bit random number.
> +
> +  if Rand is NULL, then ASSERT().
> +
> +  @param[out] Rand     Buffer pointer to store the 128-bit random value.
> +
> +  @retval TRUE         Random number generated successfully.
> +  @retval FALSE        Failed to generate the random number.
> +
> +**/
> +BOOLEAN
> +EFIAPI
> +GetRandomNumber128 (
> +  OUT     UINT64                    *Rand
> +  )
> +{
> +  ASSERT (Rand != NULL);
> +  // This should take around 80ms
> +
> +  // Read first 64 bits
> +  if (!GetRandomNumber64 (Rand)) {
> +    return FALSE;
> +  }
> +
> +  // Read second 64 bits
> +  return GetRandomNumber64 (++Rand);
> +}
> diff --git a/MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf b/MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf
> new file mode 100644
> index 000000000000..34dea0152497
> --- /dev/null
> +++ b/MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf
> @@ -0,0 +1,40 @@
> +## @file
> +#  Instance of RNG (Random Number Generator) Library.
> +#
> +#  BaseRng Library that uses the TimerLib to provide reasonably random numbers.
> +#  Do NOT use this on a production system as this uses the system performance
> +#  counter rather than a true source of random in addition to having a weak
> +#  random algorithm. This is provided primarily as a source of entropy for
> +#  OpenSSL for platforms that do not have a good built in RngLib as this
> +#  emulates what was done before (though it isn't perfect).
> +#
> +#  Copyright (c) Microsoft Corporation. All rights reserved.<BR>
> +#
> +#  SPDX-License-Identifier: BSD-2-Clause-Patent
> +#
> +#
> +##
> +
> +[Defines]
> +  INF_VERSION                    = 0x00010005

Please use the latest version (1.27 iirc, which you don't actually need 
to convert to hex format like above)

> +  BASE_NAME                      = BaseRngLibTimerLib
> +  MODULE_UNI_FILE                = BaseRngLibTimerLib.uni
> +  FILE_GUID                      = 74950C45-10FC-4AB5-B114-49C87C17409B
> +  MODULE_TYPE                    = BASE
> +  VERSION_STRING                 = 1.0
> +  LIBRARY_CLASS                  = RngLib
> +  CONSTRUCTOR                    = BaseRngLibConstructor
> +
> +#
> +#  VALID_ARCHITECTURES           = IA32 X64

Drop this unless it is accurate.

> +#
> +
> +[Sources]
> +  RngLibTimer.c
> +
> +[Packages]
> +  MdePkg/MdePkg.dec
> +
> +[LibraryClasses]
> +  BaseLib
> +  TimerLib
> diff --git a/MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.uni b/MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.uni
> new file mode 100644
> index 000000000000..766a8e0ddf97
> --- /dev/null
> +++ b/MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.uni
> @@ -0,0 +1,17 @@
> +// /** @file
> +// Instance of RNG (Random Number Generator) Library.
> +//
> +// BaseRng Library that uses TimerLib's performance counter

Please refer to this library as RngLib, to avoid confusion. Adding 
'base' to qualify it is fine.

> +// to provide random numbers.
> +//
> +// Copyright (c) Microsoft Corporation

year?

> +//
> +// SPDX-License-Identifier: BSD-2-Clause-Patent
> +//
> +// **/
> +
> +
> +#string STR_MODULE_ABSTRACT             #language en-US "Instance of RNG Library"
> +
> +#string STR_MODULE_DESCRIPTION          #language en-US "BaseRng Library that uses the TimerLib to provide low-entropy random numbers"
> +
> diff --git a/MdePkg/MdePkg.dsc b/MdePkg/MdePkg.dsc
> index 472fa3777412..d7ba3a730909 100644
> --- a/MdePkg/MdePkg.dsc
> +++ b/MdePkg/MdePkg.dsc
> @@ -62,6 +62,8 @@
>     MdePkg/Library/BasePostCodeLibPort80/BasePostCodeLibPort80.inf
>     MdePkg/Library/BasePrintLib/BasePrintLib.inf
>     MdePkg/Library/BaseReportStatusCodeLibNull/BaseReportStatusCodeLibNull.inf
> +  MdePkg/Library/BaseRngLibTimerLib/BaseRngLibTimerLib.inf
> +  MdePkg/Library/BaseRngLibNull/BaseRngLibNull.inf

Please match the indentation of the surrounding lines.

>     MdePkg/Library/BaseSerialPortLibNull/BaseSerialPortLibNull.inf
>     MdePkg/Library/BaseSynchronizationLib/BaseSynchronizationLib.inf
>     MdePkg/Library/BaseTimerLibNullTemplate/BaseTimerLibNullTemplate.inf
> @@ -69,7 +71,6 @@
>     MdePkg/Library/BaseUefiDecompressLib/BaseUefiTianoCustomDecompressLib.inf
>     MdePkg/Library/BaseSmbusLibNull/BaseSmbusLibNull.inf
>     MdePkg/Library/BaseSafeIntLib/BaseSafeIntLib.inf
> -  MdePkg/Library/BaseRngLibNull/BaseRngLibNull.inf
>   
>     MdePkg/Library/DxeCoreEntryPoint/DxeCoreEntryPoint.inf
>     MdePkg/Library/DxeCoreHobLib/DxeCoreHobLib.inf
> 


  reply	other threads:[~2020-08-13 12:09 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-08-12 22:43 [PATCH v6 0/5] Use RngLib instead of TimerLib for OpensslLib Matthew Carlson
2020-08-12 22:43 ` [PATCH v6 1/5] MdePkg: TimerRngLib: Added RngLib that uses TimerLib Matthew Carlson
2020-08-13 12:09   ` Ard Biesheuvel [this message]
2020-08-13 18:59     ` Matthew Carlson
2020-08-12 22:43 ` [PATCH v6 2/5] MdePkg: BaseRngLibDxe: Add RngLib that uses RngDxe Matthew Carlson
2020-08-13  9:10   ` Liming Gao
2020-08-13 12:19   ` Ard Biesheuvel
2020-08-13 19:18     ` Matthew Carlson
2020-08-12 22:43 ` [PATCH v6 3/5] OvmfPkg: Add RngLib based on TimerLib for Crypto Matthew Carlson
2020-08-13  8:34   ` Laszlo Ersek
2020-08-12 22:43 ` [PATCH v6 4/5] ArmVirtPkg: Add RngLib based on TimerLib for CryptoPkg Matthew Carlson
2020-08-13  8:31   ` Laszlo Ersek
2020-08-12 22:43 ` [PATCH v6 5/5] CryptoPkg: OpensslLib: Use RngLib to generate entropy in rand_pool Matthew Carlson
2020-08-13 12:21   ` Ard Biesheuvel
2020-08-13 15:14 ` [PATCH v6 0/5] Use RngLib instead of TimerLib for OpensslLib Yao, Jiewen
2020-08-13 19:42   ` Matthew Carlson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-list from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d2b05ea0-fe26-a6e5-60db-ef52ca1bc69d@arm.com \
    --to=devel@edk2.groups.io \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox