ima pcr question

ima pcr question

[Jerry Snitselaar]

We had a report of messages from ima saying "Error communicating with
TPM".  Looking into it a bit, it looks like with some Dell systems
(possibly others as well) in the bios they can set the hash algorithm
being used. In this case with that set to sha256 the messages
appear. Flipping the system to using sha1 makes them disappear.
Looking at the ima code, ima_calc_boot_aggregate_tfm hard codes using
sha1. Should that be changed to use whatever the default is in the
config, or possibly find out from the tpm what algorithm is being used?

Regards,
Jerry

Re: ima pcr question

[Mimi Zohar]

On Mon, 2020-01-13 at 17:06 -0700, Jerry Snitselaar wrote:
> We had a report of messages from ima saying "Error communicating with
> TPM".  Looking into it a bit, it looks like with some Dell systems
> (possibly others as well) in the bios they can set the hash algorithm
> being used. In this case with that set to sha256 the messages
> appear. Flipping the system to using sha1 makes them disappear.
> Looking at the ima code, ima_calc_boot_aggregate_tfm hard codes using
> sha1. Should that be changed to use whatever the default is in the
> config, or possibly find out from the tpm what algorithm is being used?

The ima-ng template contains two digests.  The first digest is the
value being extended into the TPM, while the second digest is either
the boot aggregate or file data hash.  It sounds like the problem is
with the first digest.  Changing the boot-aggregate to use sha256
might be a good idea, but probably won't fix the problem.

Mimi

Re: ima pcr question

[Jerry Snitselaar]

On Mon Jan 13 20, Mimi Zohar wrote:
>On Mon, 2020-01-13 at 17:06 -0700, Jerry Snitselaar wrote:
>> We had a report of messages from ima saying "Error communicating with
>> TPM".  Looking into it a bit, it looks like with some Dell systems
>> (possibly others as well) in the bios they can set the hash algorithm
>> being used. In this case with that set to sha256 the messages
>> appear. Flipping the system to using sha1 makes them disappear.
>> Looking at the ima code, ima_calc_boot_aggregate_tfm hard codes using
>> sha1. Should that be changed to use whatever the default is in the
>> config, or possibly find out from the tpm what algorithm is being used?
>
>The ima-ng template contains two digests.  The first digest is the
>value being extended into the TPM, while the second digest is either
>the boot aggregate or file data hash.  It sounds like the problem is
>with the first digest.  Changing the boot-aggregate to use sha256
>might be a good idea, but probably won't fix the problem.
>
>Mimi
>

The error message is coming from ima_pcrread, and the tpm_digest that gets passed
by ima_calc_boot_aggregate_tfm to ima_pcrread is declared:

	struct tpm_digest d = { .alg_id = TPM_ALG_SHA1, .digest = {0} };

According to Dell their default BIOS setting is to use sha256. What they see with that
setting is:

[ 5.475036] ima: Error Communicating to TPM chip
[ 5.475083] tsc: Refined TSC clocksource calibration: 3311.999 MHz
[ 5.475092] clocksource: tsc: mask: 0xffffffffffffffff max_cycles: 0x2fbd936b72f, max_idle_ns: 440795283163 ns
[ 5.475118] ima: Error Communicating to TPM chip
[ 5.475165] ima: Error Communicating to TPM chip
[ 5.475235] clocksource: Switched to clocksource tsc
[ 5.475266] ima: Error Communicating to TPM chip
[ 5.475311] ima: Error Communicating to TPM chip
[ 5.475341] ima: Error Communicating to TPM chip
[ 5.475371] ima: Error Communicating to TPM chip
[ 5.475402] ima: Error Communicating to TPM chip
[ 5.489049] ima: No architecture policies found

Re: ima pcr question

[James Bottomley]

On Fri, 2020-01-17 at 15:29 -0700, Jerry Snitselaar wrote:
> On Mon Jan 13 20, Mimi Zohar wrote:
> > On Mon, 2020-01-13 at 17:06 -0700, Jerry Snitselaar wrote:
> > > We had a report of messages from ima saying "Error communicating
> > > with
> > > TPM".  Looking into it a bit, it looks like with some Dell
> > > systems
> > > (possibly others as well) in the bios they can set the hash
> > > algorithm
> > > being used. In this case with that set to sha256 the messages
> > > appear. Flipping the system to using sha1 makes them disappear.
> > > Looking at the ima code, ima_calc_boot_aggregate_tfm hard codes
> > > using
> > > sha1. Should that be changed to use whatever the default is in
> > > the
> > > config, or possibly find out from the tpm what algorithm is being
> > > used?
> > 
> > The ima-ng template contains two digests.  The first digest is the
> > value being extended into the TPM, while the second digest is
> > either
> > the boot aggregate or file data hash.  It sounds like the problem
> > is
> > with the first digest.  Changing the boot-aggregate to use sha256
> > might be a good idea, but probably won't fix the problem.
> > 
> > Mimi
> > 
> 
> The error message is coming from ima_pcrread, and the tpm_digest that
> gets passed
> by ima_calc_boot_aggregate_tfm to ima_pcrread is declared:
> 
> 	struct tpm_digest d = { .alg_id = TPM_ALG_SHA1, .digest = {0}
> };
> 
> According to Dell their default BIOS setting is to use sha256. What
> they see with that
> setting is:
> 
> [ 5.475036] ima: Error Communicating to TPM chip
> [ 5.475083] tsc: Refined TSC clocksource calibration: 3311.999 MHz
> [ 5.475092] clocksource: tsc: mask: 0xffffffffffffffff max_cycles:
> 0x2fbd936b72f, max_idle_ns: 440795283163 ns
> [ 5.475118] ima: Error Communicating to TPM chip
> [ 5.475165] ima: Error Communicating to TPM chip
> [ 5.475235] clocksource: Switched to clocksource tsc
> [ 5.475266] ima: Error Communicating to TPM chip
> [ 5.475311] ima: Error Communicating to TPM chip
> [ 5.475341] ima: Error Communicating to TPM chip
> [ 5.475371] ima: Error Communicating to TPM chip
> [ 5.475402] ima: Error Communicating to TPM chip
> [ 5.489049] ima: No architecture policies found

The implication seems to be that on a DELL setting the bios default to
sha256 turns off the TPM's sha1 pcr banks ... is that the case?

tssgetcapability -cap 5

should confirm or deny this.

James

Re: ima pcr question

[Jerry Snitselaar]

On Fri Jan 17 20, James Bottomley wrote:
>On Fri, 2020-01-17 at 15:29 -0700, Jerry Snitselaar wrote:
>> On Mon Jan 13 20, Mimi Zohar wrote:
>> > On Mon, 2020-01-13 at 17:06 -0700, Jerry Snitselaar wrote:
>> > > We had a report of messages from ima saying "Error communicating
>> > > with
>> > > TPM".  Looking into it a bit, it looks like with some Dell
>> > > systems
>> > > (possibly others as well) in the bios they can set the hash
>> > > algorithm
>> > > being used. In this case with that set to sha256 the messages
>> > > appear. Flipping the system to using sha1 makes them disappear.
>> > > Looking at the ima code, ima_calc_boot_aggregate_tfm hard codes
>> > > using
>> > > sha1. Should that be changed to use whatever the default is in
>> > > the
>> > > config, or possibly find out from the tpm what algorithm is being
>> > > used?
>> >
>> > The ima-ng template contains two digests.  The first digest is the
>> > value being extended into the TPM, while the second digest is
>> > either
>> > the boot aggregate or file data hash.  It sounds like the problem
>> > is
>> > with the first digest.  Changing the boot-aggregate to use sha256
>> > might be a good idea, but probably won't fix the problem.
>> >
>> > Mimi
>> >
>>
>> The error message is coming from ima_pcrread, and the tpm_digest that
>> gets passed
>> by ima_calc_boot_aggregate_tfm to ima_pcrread is declared:
>>
>> 	struct tpm_digest d = { .alg_id = TPM_ALG_SHA1, .digest = {0}
>> };
>>
>> According to Dell their default BIOS setting is to use sha256. What
>> they see with that
>> setting is:
>>
>> [ 5.475036] ima: Error Communicating to TPM chip
>> [ 5.475083] tsc: Refined TSC clocksource calibration: 3311.999 MHz
>> [ 5.475092] clocksource: tsc: mask: 0xffffffffffffffff max_cycles:
>> 0x2fbd936b72f, max_idle_ns: 440795283163 ns
>> [ 5.475118] ima: Error Communicating to TPM chip
>> [ 5.475165] ima: Error Communicating to TPM chip
>> [ 5.475235] clocksource: Switched to clocksource tsc
>> [ 5.475266] ima: Error Communicating to TPM chip
>> [ 5.475311] ima: Error Communicating to TPM chip
>> [ 5.475341] ima: Error Communicating to TPM chip
>> [ 5.475371] ima: Error Communicating to TPM chip
>> [ 5.475402] ima: Error Communicating to TPM chip
>> [ 5.489049] ima: No architecture policies found
>
>The implication seems to be that on a DELL setting the bios default to
>sha256 turns off the TPM's sha1 pcr banks ... is that the case?
>
>tssgetcapability -cap 5
>
>should confirm or deny this.
>
>James
>

I believe so, I'm waiting to get access to the system here to double check.
Before with the intel stack, tpm2_pcrlist -s would return both sha1 and sha256,
but the plain tpm2_pcrlist command would show only banks for one or the other
depending on which setting was in the bios. For the other it would just print
out the algorithm and nothing else.

I should be able to run the tss2 command later today.

Regards,
Jerry

Re: ima pcr question

[Jerry Snitselaar]

On Sat Jan 18 20, Jerry Snitselaar wrote:
>On Fri Jan 17 20, James Bottomley wrote:
>>On Fri, 2020-01-17 at 15:29 -0700, Jerry Snitselaar wrote:
>>>On Mon Jan 13 20, Mimi Zohar wrote:
>>>> On Mon, 2020-01-13 at 17:06 -0700, Jerry Snitselaar wrote:
>>>> > We had a report of messages from ima saying "Error communicating
>>>> > with
>>>> > TPM".  Looking into it a bit, it looks like with some Dell
>>>> > systems
>>>> > (possibly others as well) in the bios they can set the hash
>>>> > algorithm
>>>> > being used. In this case with that set to sha256 the messages
>>>> > appear. Flipping the system to using sha1 makes them disappear.
>>>> > Looking at the ima code, ima_calc_boot_aggregate_tfm hard codes
>>>> > using
>>>> > sha1. Should that be changed to use whatever the default is in
>>>> > the
>>>> > config, or possibly find out from the tpm what algorithm is being
>>>> > used?
>>>>
>>>> The ima-ng template contains two digests.  The first digest is the
>>>> value being extended into the TPM, while the second digest is
>>>> either
>>>> the boot aggregate or file data hash.  It sounds like the problem
>>>> is
>>>> with the first digest.  Changing the boot-aggregate to use sha256
>>>> might be a good idea, but probably won't fix the problem.
>>>>
>>>> Mimi
>>>>
>>>
>>>The error message is coming from ima_pcrread, and the tpm_digest that
>>>gets passed
>>>by ima_calc_boot_aggregate_tfm to ima_pcrread is declared:
>>>
>>>	struct tpm_digest d = { .alg_id = TPM_ALG_SHA1, .digest = {0}
>>>};
>>>
>>>According to Dell their default BIOS setting is to use sha256. What
>>>they see with that
>>>setting is:
>>>
>>>[ 5.475036] ima: Error Communicating to TPM chip
>>>[ 5.475083] tsc: Refined TSC clocksource calibration: 3311.999 MHz
>>>[ 5.475092] clocksource: tsc: mask: 0xffffffffffffffff max_cycles:
>>>0x2fbd936b72f, max_idle_ns: 440795283163 ns
>>>[ 5.475118] ima: Error Communicating to TPM chip
>>>[ 5.475165] ima: Error Communicating to TPM chip
>>>[ 5.475235] clocksource: Switched to clocksource tsc
>>>[ 5.475266] ima: Error Communicating to TPM chip
>>>[ 5.475311] ima: Error Communicating to TPM chip
>>>[ 5.475341] ima: Error Communicating to TPM chip
>>>[ 5.475371] ima: Error Communicating to TPM chip
>>>[ 5.475402] ima: Error Communicating to TPM chip
>>>[ 5.489049] ima: No architecture policies found
>>
>>The implication seems to be that on a DELL setting the bios default to
>>sha256 turns off the TPM's sha1 pcr banks ... is that the case?
>>
>>tssgetcapability -cap 5
>>
>>should confirm or deny this.
>>
>>James
>>
>
>I believe so, I'm waiting to get access to the system here to double check.
>Before with the intel stack, tpm2_pcrlist -s would return both sha1 and sha256,
>but the plain tpm2_pcrlist command would show only banks for one or the other
>depending on which setting was in the bios. For the other it would just print
>out the algorithm and nothing else.
>
>I should be able to run the tss2 command later today.
>
>Regards,
>Jerry

with sha1 selected:

[root@dell-per830-01 ~]# tssgetcapability -cap 5
2 PCR selections
     hash TPM_ALG_SHA1
     TPMS_PCR_SELECTION length 3
     ff ff ff 
     hash TPM_ALG_SHA256
     TPMS_PCR_SELECTION length 3
     00 00 00 

with sha256 selected:

[root@dell-per830-01 ~]# tssgetcapability -cap 5
2 PCR selections
     hash TPM_ALG_SHA1
     TPMS_PCR_SELECTION length 3
     00 00 00 
     hash TPM_ALG_SHA256
     TPMS_PCR_SELECTION length 3
     ff ff ff 

Re: ima pcr question

[James Bottomley]

On Sat, 2020-01-18 at 16:20 -0700, Jerry Snitselaar wrote:
> On Sat Jan 18 20, Jerry Snitselaar wrote:
> > On Fri Jan 17 20, James Bottomley wrote:
[...]
> > > The implication seems to be that on a DELL setting the bios
> > > default to sha256 turns off the TPM's sha1 pcr banks ... is that
> > > the case?
> > > 
> > > tssgetcapability -cap 5
> > > 
> > > should confirm or deny this.
> > > 
> > > James
> > > 
> > 
> > I believe so, I'm waiting to get access to the system here to
> > double check. Before with the intel stack, tpm2_pcrlist -s would
> > return both sha1 and sha256, but the plain tpm2_pcrlist command
> > would show only banks for one or the other depending on which
> > setting was in the bios. For the other it would just print
> > out the algorithm and nothing else.
> > 
> > I should be able to run the tss2 command later today.
> > 
> > Regards,
> > Jerry
> 
> with sha1 selected:
> 
> [root@dell-per830-01 ~]# tssgetcapability -cap 5
> 2 PCR selections
>      hash TPM_ALG_SHA1
>      TPMS_PCR_SELECTION length 3
>      ff ff ff 
>      hash TPM_ALG_SHA256
>      TPMS_PCR_SELECTION length 3
>      00 00 00 
> 
> with sha256 selected:
> 
> [root@dell-per830-01 ~]# tssgetcapability -cap 5
> 2 PCR selections
>      hash TPM_ALG_SHA1
>      TPMS_PCR_SELECTION length 3
>      00 00 00 
>      hash TPM_ALG_SHA256
>      TPMS_PCR_SELECTION length 3
>      ff ff ff 

OK, so that confirms the suspicion.  The only active bank is the one
you've selected in the bios ... I suppose it was done to avoid having
to measure through more than one bank, but it does mean IMA must cope
in the case the sha1 bank isn't active.

James

Re: ima pcr question

[Mimi Zohar]

On Sat, 2020-01-18 at 15:51 -0800, James Bottomley wrote:
> On Sat, 2020-01-18 at 16:20 -0700, Jerry Snitselaar wrote:
> > On Sat Jan 18 20, Jerry Snitselaar wrote:
> > > On Fri Jan 17 20, James Bottomley wrote:
> [...]
> > > > The implication seems to be that on a DELL setting the bios
> > > > default to sha256 turns off the TPM's sha1 pcr banks ... is that
> > > > the case?
> > > > 
> > > > tssgetcapability -cap 5
> > > > 
> > > > should confirm or deny this.
> > > > 
> > > > James
> > > > 
> > > 
> > > I believe so, I'm waiting to get access to the system here to
> > > double check. Before with the intel stack, tpm2_pcrlist -s would
> > > return both sha1 and sha256, but the plain tpm2_pcrlist command
> > > would show only banks for one or the other depending on which
> > > setting was in the bios. For the other it would just print
> > > out the algorithm and nothing else.
> > > 
> > > I should be able to run the tss2 command later today.
> > > 
> > > Regards,
> > > Jerry
> > 
> > with sha1 selected:
> > 
> > [root@dell-per830-01 ~]# tssgetcapability -cap 5
> > 2 PCR selections
> >      hash TPM_ALG_SHA1
> >      TPMS_PCR_SELECTION length 3
> >      ff ff ff 
> >      hash TPM_ALG_SHA256
> >      TPMS_PCR_SELECTION length 3
> >      00 00 00 
> > 
> > with sha256 selected:
> > 
> > [root@dell-per830-01 ~]# tssgetcapability -cap 5
> > 2 PCR selections
> >      hash TPM_ALG_SHA1
> >      TPMS_PCR_SELECTION length 3
> >      00 00 00 
> >      hash TPM_ALG_SHA256
> >      TPMS_PCR_SELECTION length 3
> >      ff ff ff 
> 
> OK, so that confirms the suspicion.  The only active bank is the one
> you've selected in the bios ... I suppose it was done to avoid having
> to measure through more than one bank, but it does mean IMA must cope
> in the case the sha1 bank isn't active.

Agreed, IMA needs to support calculating and extending the different
TPM banks with the correct digest values; and the IMA measurement list
format needs to be updated as well.  Patches have been posted to
address both of these issues, but were not ready to be upstreamed.
 Now that the TPM interface has been addressed, we can at least
address the first issue - calculating and extending the TPM with the
appropriate digest values.

For now, if the TPM SHA1 bank is not enabled, I suggest going into
TPM-bypass mode.  ima_init_digests() already walks the list of enabled
TPM banks.  Updating to check if a SHA1 bank is enabled should be
relatively straight forward.

Mimi