Re: [PATCH] hwrng: tpm: Do not enable by default

On 09.05.26 17:18, Jarkko Sakkinen wrote:
> On Wed, Apr 29, 2026 at 02:33:20PM +0000, Niedermayr, BENEDIKT wrote:
>> On 10/27/25 20:51, Jarkko Sakkinen wrote:
>>> On Tue, Oct 21, 2025 at 02:46:15PM +0200, Jan Kiszka wrote:
>>>> From: Jan Kiszka <[email protected]>
>>>>
>>>> As seen with optee_ftpm, which uses ms-tpm-20-ref [1], a TPM may write
>>>> the current time epoch to its NV storage every 4 seconds if there are
>>>> commands sent to it. The 60 seconds periodic update of the entropy pool
>>>> that the hwrng kthread does triggers this, causing about 4 writes per
>>>> requests. Makes 2 millions per year for a 24/7 device, and that is a lot
>>>> for its backing NV storage.
>>>>
>>>> It is therefore better to make the user intentionally enable this,
>>>> providing a chance to read the warning.
>>>>
>>>> [1] https://github.com/Microsoft/ms-tpm-20-ref
>>>>
>>>> Signed-off-by: Jan Kiszka <[email protected]>
>>>
>>> Looking at DRBG_* from [1] I don't see anything you describe. If OPTEE
>>> writes NVRAM,  then the implementation is broken.
>>>
>>> Also AFAIK, it is pre-seeded per power cycle. There's nothing that even
>>> distantly relates on using NVRAM.
>>>
>>> [1] https://trustedcomputinggroup.org/wp-content/uploads/TPM-2.0-1.83-Part-4-Supporting-Routines-Code.pdf
>>
>> Hi all,
>>
>> we recently also stumbled over this issue which led me here to this 
>> thread and maybe adding our observations helps to clarify things here a 
>> bit (hopefully) or at least augments the information related to firmware 
>> TPM based implementation based on ms-tpm-20-ref.
>>
>> Based on the optee_ftpm repo, as Jan already described, which currently 
>> references commit 98b60a44aba7 of [1] suffers this exact issue because 
>> of the NV_CLOCK_UPDATE_INTERVAL [2] which is set to "12" and issues a 
>> write for each command after ~4 seconds have passed.
>>
>> This config has been changed to "22" (on current master branch [3]) 
>> which is the allowed maximum when following the TPM spec (chapter 36.3.2 
>> in [4]) which leads to round about 70 minutes, but optee_ftpm didn't 
>> move ahead to this commit, yet.
>> This config exists for being able to adapt the write cycles to the 
>> specific wear conditions of the hardware.
>>
>> Moreover the ms-tpm-20-ref repo seems to not be maintained anymore and 
>> one should rather switch to [6].
>>
>> So there are currently firmware TPM implementations out there that lead 
>> to these frequent writes.
> 
> Really this would need a product and official bug bulletin for it to
> even consider a workaround. Speculation does not count.
> 

The key point Benedikt tries to make here is that the TPM 2 spec forces
any vendor to do something about persisting the last seen time at least
every 70 min. If they didn't do that, then they would violate the space
- arguably a bug. But, correct, it does not tell us anything about how
this happens in a random firmware TPM implementation.

>>
>> AFAIK since the tpm-20-ref implementation basically only supports a file 
>> on disk or RAM backing storage, the optee_ftpm repo [5] provides it's 
>> own _plat_NV* implementations that replace the default ones and finally 
>> call OP-TEE's TEE_* secure storage API, which then routes to whatever 
>> backend OP-TEE is configured with (REE-FS or RPMB) – In our case the RPMB.
>>
>> Because there are currently implementations out there (e.g. start using 
>> optee_ftpm) it may make sense to add this information to the kernel 
>> config's help text at least?
> 
> Your first forum to report such issues is the TPM vendor.

I would still not recommend anyone relying on a firmware TPM to turn on
CONFIG_HW_RANDOM_TPM if there are viable alternatives. In case of the
open source stack with optee_os + optee_ftpm, we know that at least one
exists: CONFIG_HW_RANDOM_OPTEE.

So, if there is no good place to document this in the kernel, maybe it
is worth to document it in optee_ftpm instead.

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center