Full Virtual TPM for VMs and Containers #4071

shjala · 2024-07-09T10:44:38Z

Problem with existing VTPM service

First of all, VTPM has a misleading name since it is not actually a Virtual TPM, but a HTTP server which through its APIs, exposes a limited set of the host's TPM functionalities to the callers. As a result, there is no TPM device present in the VM (no /dev/tpm* ). This is problematic since some services (for example Azure Identity Service) expects a TPM device is available and is programmed to communicate directly with the TPM for credential transfer and signing purposes.

To fix this issue, we patch the Azure Identity Service to translate its TPM calls to VTPM calls (the patch set is maintained in EVE-TOOLS repository). This approach while working, can lead to maintenance issues. Currently we support AZIOT version 1.2. We claim to support the 1.4 version too, but in reality we revert the Azure Identity Service to a commit between version 1.2 and 1.3 to apply the patches and are lucky that compiled binaries are still compatible with version 1.4.

Full Virtual TPM in VMs and Containers

As mentioned above, using VTPM is not optimal. VTPM comes with its own set of problems¹, in addition it creates a limiting environment and any external program that relies on a TPM requires creating and maintaining its own set of patches. To solve this problem, This PR uses Software TPM (SWTPM) to emulate a Virtual TPM per VMs.

Fortunately SWTPM integrates well with QEMU, in order to emulate a Virtual TPM in the VM, we create a SWTPM instance per VM. The SWTPM instance is configured to use a Unix Domain Socket as a communication line, after passing the socket path to the QEMU virtual TPM configurations, it automatically creates a virtual TPM device for the VM which is accessible like a normal TPM under /dev/tpm* . We can create the same setup for bare metal containers using Linux Virtual TPM Proxy driver to create arbitrary virtual TPM devices on the host, backed by a SWTPM process and mapped to the container as /dev/tpmX device.

SWTPM saves and loads the TPM state on/from the disk, so at the next VM boot all the TPM keys, TPM NVRAM data, etc are present. In addition SWTPM is configured to encrypt the TPM state files using a 256 bit AES key (we use the same key as vault key, this key stored in TPM and access is authorized using a PCR policy, this makes it easier to manage since if we use a separate key for SWTPM states, there is one more key to backup/restore from the controller). We can pass the constant VM id and the encryption key through a KDF to create unique keys for each VM/Container if needed, but it brings no added security. Here is a diagram, showing the basic architecture of the new VTPM:

1: For example it relies on a set of TPM management utilities (TPM-TOOLS) to communicate with the host TPM, this dependency not only increase the attack surface, but also can cause maintenance problem, because TPM-TOOLS is consist of a set of command-line utilities (Not APIs) and a next version of TPM-TOOLS can simply change the command line arguments behavior and break the VTPM functionality without notice.

TODO

Eden Test

lf-edge/eden#1011

Manual Tests

So far I have tested the following with a Ubuntu 22.04.2 LTS :
1- Sign and verify with the TPM using the endorsement hierarchy [successful]

ubuntu@vm01:~$ tpm2_createprimary -C e -c primary.ctx
tpm2_create -G rsa -u rsa.pub -r rsa.priv -C primary.ctx
tpm2_load -C primary.ctx -u rsa.pub -r rsa.priv -c rsa.ctx
echo "my message" > message.dat
tpm2_sign -c rsa.ctx -g sha256 -o sig.rssa message.dat
tpm2_verifysignature -c rsa.ctx -g sha256 -s sig.rssa -m message.dat
name-alg:
  value: sha256
  raw: 0xb
attributes:
  value: fixedtpm|fixedparent|sensitivedataorigin|userwithauth|restricted|decrypt
  raw: 0x30072
type:
  value: rsa
  raw: 0x1
exponent: 65537
bits: 2048
scheme:
  value: null
  raw: 0x10
scheme-halg:
  value: (null)
  raw: 0x0
sym-alg:
  value: aes
  raw: 0x6
sym-mode:
  value: cfb
  raw: 0x43
sym-keybits: 128
rsa: 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
name-alg:
  value: sha256
  raw: 0xb
attributes:
  value: fixedtpm|fixedparent|sensitivedataorigin|userwithauth|decrypt|sign
  raw: 0x60072
type:
  value: rsa
  raw: 0x1
exponent: 65537
bits: 2048
scheme:
  value: null
  raw: 0x10
scheme-halg:
  value: (null)
  raw: 0x0
sym-alg:
  value: null
  raw: 0x10
sym-mode:
  value: (null)
  raw: 0x0
sym-keybits: 0
rsa: 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
name: 000bfe394bb93fc82e2e6fae6f6c4dab6199215674727f5c287aa3641eff4d8f3779
ubuntu@vm01:~$ echo $?
0

2- Sign with the TPM and verify with openssl [successful]

ubuntu@vm01:~$ openssl ecparam -name prime256v1 -genkey -noout -out private.ecc.pem
openssl ec -in private.ecc.pem -out public.ecc.pem -pubout
# Generate a hash to sign
echo "data to sign" > data.in.raw
sha256sum data.in.raw | awk '{ print "000000 " $1 }' | \
xxd -r -c 32 > data.in.digest
# Load the private key for signing
tpm2_loadexternal -Q -G ecc -r private.ecc.pem -c key.ctx
# Sign in the TPM and verify with OSSL
tpm2_sign -Q -c key.ctx -g sha256 -d -f plain -o data.out.signed data.in.digest
openssl dgst -verify public.ecc.pem -keyform pem -sha256 \
-signature data.out.signed data.in.raw
read EC key
writing EC key
Verified OK

Manually Test AzureIoTEdge (Ubuntu 20.04, Aziot 1.4.0 and EVE-TOOLS)

Manually deploying a ubuntu-20.04-server-amd64 and running the test-script I wrote for eden :

ubuntu@vm01:~$ sudo iotedge system status
System services:
    aziot-edged             Running
    aziot-identityd         Running
    aziot-keyd              Running
    aziot-certd             Running
    aziot-tpmd              Running

Manually Test AzureIoTEdge (Ubuntu 20.04, Aziot latest (1.5.x) without EVE-TOOLS, using Virtual TPM)

Manually deploying a ubuntu-22.04-server-amd64 and running the test_make_tpm_keys and then test_ubuntu22.04_aziot_latest :

[00:31:28.097] ubuntu@vm01:~$ sudo iotedge system status
System services:
[00:31:35.783]     aziot-edged             Running
[00:31:35.784]     aziot-identityd         Running
[00:31:35.785]     aziot-keyd              Running
[00:31:35.785]     aziot-certd             Running
[00:31:35.786]     aziot-tpmd              Running
[00:31:35.787]

shjala · 2024-07-16T13:59:22Z

Please don't mind the many commits, I will squash them later. This is more or less ready for review even though there are tasks on the list (I'll do while the review is going on).

eriknordmark · 2024-07-23T18:58:10Z

Isn't /dev/tpm0 the TPM 1.X device, and /dev/tpmrm0 the TPM 2.0 device?

I assume we need to keep the current vtpm linuxkit container around since there are some AZIOT VMs which use the existing eve-tools to access that.

eriknordmark · 2024-07-23T19:05:55Z

sign swtpm EK with actual TPM ek to create a chain of trust

Would it make sense to also add some way for an app instance to find out about the host's TPM? Or will that be implicit in the signature it will see for its EK?
(If need be we can export information to app instances using the meta-data server.)

how to prevent attacker from cloning a swtpm instance (encrypted state)? what is our attack model?

It might make sense to have the encryption key be a function of the App instance UUID. That would prevent someone from reusing it in another VM they control.

Note that I haven't thought through the attack scenario in detail - other than a user with physical access can temporarly remove the disk and mount it on a different computer and copy /persist/foo/A to /persist/foo/B. But if we put this under /persist/vault/ then that possibility goes away.

eriknordmark

It is quite hard to review this PR due to the choice to rename the existing vtpm to ptpm and having the new be called vtmp.

Can we defer that rename until the end of the process so that the diffs make some more sense?

pkg/apparmor/profiles/usr.bin.swtpm

docs/VTPM.md

pkg/pillar/types/locationconsts.go

shjala · 2024-07-24T09:03:14Z

Isn't /dev/tpm0 the TPM 1.X device, and /dev/tpmrm0 the TPM 2.0 device?

/dev/tpmrm0 is same as /dev/tpm0 but with the addition of automatic resource management, basically a tpm2-abrmd in the kernel. It helps managing TPM resources otherwise if raw /dev/tpm0 is used as TPM interface, it is common to exhaust TPM resource fairly quickly and hit errors.

AFAIK TPM version is determined by GetCap packet and examining the returned version, here we only support TPM 2.0 with the --tpm2 arg passed to swtpm.

I assume we need to keep the current vtpm linuxkit container around since there are some AZIOT VMs which use the existing eve-tools to access that.

yes, I will keep it running in the VTPM container alongside the SWTPM till all have upgraded to newer versions of AZIOT with no eve-tools patches.

shjala · 2024-07-25T09:49:35Z

It is quite hard to review this PR due to the choice to rename the existing vtpm to ptpm and having the new be called vtmp.

Can we defer that rename until the end of the process so that the diffs make some more sense?

should be less cluttered now.

pkg/pillar/hypervisor/hypervisor.go

pkg/vtpm/swtpm-vtpm/src/main.go

shjala · 2024-09-02T08:17:56Z

Rebased against master and resolved the merge conflict.
Running Yetus locally it complains about auto generated patch file and invalid flag: --keep-git-dir, there is issue open for the later in Hadolint repo unexpected '-' when using --keep-git-dir hadolint/hadolint#951
commit message linter complains about one line commit messages (which is enough) and body not starting with a capital letter (which I think it is too restrictive, and waste of runners time and energy to push again just to make a letter capital).

OhmSpectator · 2024-09-02T08:32:42Z

Is there a way to temporarily suppress the Yetus warning?

Regarding the capital letter. Yep, in some cases, it can be unnecessary. Still, in others, when you write a sentence, I find it useful as it can signal that the commit message was not carefully written in a "get off me" style.

shjala · 2024-09-02T08:44:37Z

Is there a way to temporarily suppress the Yetus warning?

No I think it is not a rule but parser that has no idea about --keep-git-dir, my haskell is not good enough to send a patch.

OhmSpectator · 2024-09-02T08:52:52Z

I've downloaded the Yetus results from here to check if there are other issues as well, but the reports are empty... Strange.

shjala · 2024-09-02T09:10:19Z

I've downloaded the Yetus results from here to check if there are other issues as well, but the reports are empty... Strange.

Yes I tried that too, here is my local run :

$ make mini-yetus MYETUS_VERBOSE=y
Running mini-yetus
./tools/mini-yetus.sh -f  
[+] Running mini-yetus
[+] No source branch specified. Using the main branch: master
[+] No destination branch specified. Using the current branch: swtpm_in_vm
[+] Running yetus on the changes...
[+] Full results:
.codespellignorelines:10:0:blanks:tabs in line
.codespellignorelines:1:0:blanks:tabs in line
.codespellignorelines:11:0:blanks:tabs in line
.codespellignorelines:12:0:blanks:tabs in line
.codespellignorelines:2:0:blanks:tabs in line
.codespellignorelines:3:0:blanks:tabs in line
.codespellignorelines:4:0:blanks:tabs in line
.codespellignorelines:5:0:blanks:tabs in line
.codespellignorelines:6:0:blanks:tabs in line
.codespellignorelines:7:0:blanks:tabs in line
.codespellignorelines:8:0:blanks:tabs in line
.codespellignorelines:9:0:blanks:tabs in line
.hadolint.yaml:1:1:yamllint: warning: missing document start "---" (document-start)
/home/shah/shah-dev/eve/pkg/pillar/hypervisor/kvm.go:198:0:detsecrets:db3d405b10675998c030223177d42e71b4e7a312:Secret Keyword
/home/shah/shah-dev/eve/pkg/pillar/hypervisor/kvm.go:4:1:revive: should have a package comment https://revive.run/r#package-comments
/home/shah/shah-dev/eve/pkg/pillar/hypervisor/kvm_test.go:2371:0:detsecrets:db3d405b10675998c030223177d42e71b4e7a312:Secret Keyword
/home/shah/shah-dev/eve/pkg/pillar/hypervisor/kvm_test.go:56:0:detsecrets:e7ea4f94cb4af75c6643566ca6d95d9433b8a6f2:Secret Keyword
/home/shah/shah-dev/eve/pkg/pillar/scripts/onboot.sh:121:0:shelldocs:ERROR: function percent_used has no @audience
/home/shah/shah-dev/eve/pkg/pillar/scripts/onboot.sh:121:0:shelldocs:ERROR: function percent_used has no @stability
/home/shah/shah-dev/eve/pkg/pillar/scripts/onboot.sh:139:0:shelldocs:ERROR: function free_space has no @audience
/home/shah/shah-dev/eve/pkg/pillar/scripts/onboot.sh:139:0:shelldocs:ERROR: function free_space has no @stability
/home/shah/shah-dev/eve/pkg/vtpm/Dockerfile:33:25:hadolint: invalid flag: --keep-git-dir
/home/shah/shah-dev/eve/pkg/vtpm/patch/patch-tpm2-tools.diff:11:0:blanks:end of line
/home/shah/shah-dev/eve/pkg/vtpm/patch/patch-tpm2-tools.diff:15:0:blanks:end of line
/home/shah/shah-dev/eve/pkg/vtpm/patch/patch-tpm2-tools.diff:17:0:blanks:end of line
/home/shah/shah-dev/eve/pkg/vtpm/patch/patch-tpm2-tools.diff:24:0:blanks:end of line
/home/shah/shah-dev/eve/pkg/vtpm/patch/patch-tpm2-tools.diff:7:0:blanks:end of line
.syft.yaml:7:1:yamllint: warning: missing document start "---" (document-start)
[+] Results stored in /tmp/tmp.GXzKfYfdpI/yetus-output

OhmSpectator · 2024-09-02T10:11:03Z

Ok, in general, it looks like it is ready to be merged. Does anyone who reviewed the logic have any more comments, or are all satisfied? @milan-zededa, @christoph-zededa, @eriknordmark ?

eriknordmark · 2024-09-02T13:39:13Z

I see there is still two TODOs which are not checked. The alpine tag can be done separately, but I wonder about " make sure each swtpm instance gets its's own unique and persistent seed (and by extension SRK, EK, etc)". Is this done?

shjala · 2024-09-02T14:19:46Z

I see there is still two TODOs which are not checked. The alpine tag can be done separately, but I wonder about " make sure each swtpm instance gets its's own unique and persistent seed (and by extension SRK, EK, etc)". Is this done?

yes it is :

shah@shah:~/shah-dev/eve/swtpm-seed-test$ cat test-swtpm-seed.sh

#!/bin/bash

CWD=$(pwd)
TPM_SRV_PORT1=2000
TPM_CTR_PORT1=2001
TPM_SRV_PORT2=3000
TPM_CTR_PORT2=3001

EK_HANDLE=0x81000001

TPM_STATE1=/tmp/swtpm-seed-test1
TPM_STATE2=/tmp/swtpm-seed-test2

rm -rf $TPM_STATE1
rm -rf $TPM_STATE2
mkdir -p $TPM_STATE1
mkdir -p $TPM_STATE2

flushtpm() {
  tpm2 flushcontext -t
  tpm2 flushcontext -l
  tpm2 flushcontext -s
}

swtpm socket --tpm2 \
    --server port="$TPM_SRV_PORT1" \
    --ctrl type=tcp,port="$TPM_CTR_PORT1" \
    --tpmstate dir="$TPM_STATE1" \
    --flags not-need-init,startup-clear &
PID1=$!

swtpm socket --tpm2 \
    --server port="$TPM_SRV_PORT2" \
    --ctrl type=tcp,port="$TPM_CTR_PORT2" \
    --tpmstate dir="$TPM_STATE2" \
    --flags not-need-init,startup-clear &
PID2=$!

# create first EK and export it
export TPM2TOOLS_TCTI="swtpm:host=localhost,port=$TPM_SRV_PORT1"
tpm2 clear
tpm2 createek -c ek1.ctx
flushtpm
tpm2 readpublic -Q -c ek1.ctx -f pem -o ek1.pem

# create second EK and export it
export TPM2TOOLS_TCTI="swtpm:host=localhost,port=$TPM_SRV_PORT2"
tpm2 clear
tpm2 createek -c ek2.ctx
flushtpm
tpm2 readpublic -Q -c ek2.ctx -f pem -o ek2.pem

kill $PID1
kill $PID2
rm ek1.ctx ek2.ctx
rm -rf $TPM_STATE1 $TPM_STATE2
shah@shah:~/shah-dev/eve/swtpm-seed-test$
shah@shah:~/shah-dev/eve/swtpm-seed-test$
shah@shah:~/shah-dev/eve/swtpm-seed-test$
shah@shah:~/shah-dev/eve/swtpm-seed-test$ ./test-swtpm-seed.sh
shah@shah:~/shah-dev/eve/swtpm-seed-test$ ls
ek1.pem  ek2.pem  test-swtpm-seed.sh
shah@shah:~/shah-dev/eve/swtpm-seed-test$ diff <(openssl rsa -in ek1.pem -pubin -outform DER -pubout | openssl dgst -sha256) \
     <(openssl rsa -in ek2.pem -pubin -outform DER -pubout | openssl dgst -sha256)
writing RSA key
writing RSA key
1c1
< SHA2-256(stdin)= 90d39a7ff6f1b377898d1dc7867d63e4f6c09c24a39b8eabc0d61a3cbc4edfab
---
> SHA2-256(stdin)= 01d4226284fd74c939ebae367dea6d51c22e3f496337ead3b8b52c83a017262f
shah@shah:~/shah-dev/eve/swtpm-seed-test$

pkg/vtpm/build.yml

pkg/apparmor/profiles/usr.bin.vtpm

pkg/pillar/hypervisor/kvm.go

eriknordmark

I added three questions as comments, and it seems like commitlint wants to to have sentences which start with a captital letter.

Running the tests

Just a clean up and renaming old vtpm binary to its new name ptpm. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Move the patch file to a more appropriate location. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Create nessessary directories in /persist and /run for swtpm and vtpm to store their data and set the correct permissions. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Remove unnecessary access to /config and add access to /persist in the vtpm container. /persist is used to store the TPM state. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Add apparmor profiles for swtpm and ptpm and update the existing profiles for vtpm to reflect the changes. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Move old vtpm docs to PTPM.md and update VTPM.md with new content reflecting the new vtpm service. Signed-off-by: Shahriyar Jalayeri <[email protected]>

…e VMs This commit brings SWTPM in EVE, in order to emulate Virtual TPM in the VMS, making TPM accessible in Vms like a normal TPM device under /dev/tpm* . SWTPM saves and loads the TPM state on/from the disk, so at the next VM boot all the TPM keys, TPM NVRAM data, etc are present. In addition SWTPM is configured to encrypt the TPM state files using a 256-bit AES key. We use the same key as vault key, this key stored in TPM and access to it is authorized using a PCR policy, this makes it easier to manage since if we use a separate key for SWTPM states, there is one more key to backup/restore from the controller. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Restrict vtpm access to needed directories only, it doesn't need access to all of the /persist and /run, just /persist/swtpm and /run/swtpm. Signed-off-by: Shahriyar Jalayeri <[email protected]>

Just in case the swtpm takes a bit longer to start up, due to the system being busy or TPM being slow to release the encryption key, increase the timeout to 10 seconds. Signed-off-by: Shahriyar Jalayeri <[email protected]>

shjala requested a review from eriknordmark as a code owner July 9, 2024 10:44

shjala marked this pull request as draft July 9, 2024 10:44

github-actions bot requested a review from christoph-zededa July 9, 2024 10:45

shjala changed the title ~~[WIP] virtual TPM for vm and containers~~ [WIP] Virtual TPM for VMs and Containers Jul 9, 2024

github-actions bot requested review from OhmSpectator, rene, rouming and rucoder July 10, 2024 08:43

shjala changed the title ~~[WIP] Virtual TPM for VMs and Containers~~ [PoC] Virtual TPM for VMs and Containers Jul 11, 2024

shjala force-pushed the swtpm_in_vm branch from 89958d5 to 7bed560 Compare July 15, 2024 13:29

shjala changed the title ~~[PoC] Virtual TPM for VMs and Containers~~ Full Virtual TPM for VMs and Containers Jul 15, 2024

shjala marked this pull request as ready for review July 16, 2024 13:58

shjala force-pushed the swtpm_in_vm branch 3 times, most recently from d17afe0 to e34bc2d Compare July 17, 2024 08:53

eriknordmark reviewed Jul 23, 2024

View reviewed changes

pkg/apparmor/profiles/usr.bin.swtpm Outdated Show resolved Hide resolved

eriknordmark reviewed Jul 23, 2024

View reviewed changes

docs/VTPM.md Outdated Show resolved Hide resolved

eriknordmark reviewed Jul 23, 2024

View reviewed changes

docs/VTPM.md Outdated Show resolved Hide resolved

eriknordmark reviewed Jul 23, 2024

View reviewed changes

pkg/pillar/types/locationconsts.go Outdated Show resolved Hide resolved

shjala force-pushed the swtpm_in_vm branch from a53723c to 07a27a8 Compare July 25, 2024 09:29

github-actions bot requested a review from eriknordmark July 25, 2024 09:37

christoph-zededa reviewed Jul 25, 2024

View reviewed changes

pkg/pillar/hypervisor/hypervisor.go Outdated Show resolved Hide resolved

christoph-zededa reviewed Jul 25, 2024

View reviewed changes

pkg/pillar/hypervisor/hypervisor.go Outdated Show resolved Hide resolved

christoph-zededa reviewed Jul 25, 2024

View reviewed changes

pkg/vtpm/swtpm-vtpm/src/main.go Outdated Show resolved Hide resolved

rouming approved these changes Sep 2, 2024

View reviewed changes

eriknordmark reviewed Sep 2, 2024

View reviewed changes

pkg/vtpm/build.yml Outdated Show resolved Hide resolved

eriknordmark reviewed Sep 2, 2024

View reviewed changes

pkg/apparmor/profiles/usr.bin.vtpm Show resolved Hide resolved

eriknordmark reviewed Sep 2, 2024

View reviewed changes

pkg/pillar/hypervisor/kvm.go Outdated Show resolved Hide resolved

eriknordmark approved these changes Sep 2, 2024

View reviewed changes

shjala force-pushed the swtpm_in_vm branch 2 times, most recently from b13bc8e to 7839ea7 Compare September 3, 2024 08:16

shjala requested a review from eriknordmark September 3, 2024 08:17

shjala added 9 commits September 3, 2024 11:18

VTPM : Rename vtpm_api.proto to api.proto and rename vtpm binary to ptpm

c4e2e28

Just a clean up and renaming old vtpm binary to its new name ptpm. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Move patch-tpm2-tools.diff to patch/patch-tpm2-tools.diff

1337847

Move the patch file to a more appropriate location. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Create swtpm and vtpm directories on boot

378f68f

Create nessessary directories in /persist and /run for swtpm and vtpm to store their data and set the correct permissions. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Make /persist available in vtpm container

c722c90

Remove unnecessary access to /config and add access to /persist in the vtpm container. /persist is used to store the TPM state. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Update apparmor profiles for swtpm and ptpm

352022c

Add apparmor profiles for swtpm and ptpm and update the existing profiles for vtpm to reflect the changes. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Update docs for PTPM and VTPM

be3eb9e

Move old vtpm docs to PTPM.md and update VTPM.md with new content reflecting the new vtpm service. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Restrict vtpm access to neede directories only

b08cc66

Restrict vtpm access to needed directories only, it doesn't need access to all of the /persist and /run, just /persist/swtpm and /run/swtpm. Signed-off-by: Shahriyar Jalayeri <[email protected]>

VTPM : Increase swtpm launch timeout to 10 seconds

c5298f1

Just in case the swtpm takes a bit longer to start up, due to the system being busy or TPM being slow to release the encryption key, increase the timeout to 10 seconds. Signed-off-by: Shahriyar Jalayeri <[email protected]>

shjala force-pushed the swtpm_in_vm branch from 7839ea7 to c5298f1 Compare September 3, 2024 08:20

github-actions bot requested a review from rouming September 3, 2024 08:22

eriknordmark merged commit 0c5c458 into lf-edge:master Sep 3, 2024
25 of 26 checks passed

This was referenced Sep 4, 2024

Add vTPM test lf-edge/eden#1011

Draft

Qemu : add patch to fix bios linker loader assert #4204

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Full Virtual TPM for VMs and Containers #4071

Full Virtual TPM for VMs and Containers #4071

shjala commented Jul 9, 2024 •

edited

Loading

shjala commented Jul 16, 2024 •

edited

Loading

eriknordmark commented Jul 23, 2024

eriknordmark commented Jul 23, 2024

eriknordmark left a comment

shjala commented Jul 24, 2024

shjala commented Jul 25, 2024

shjala commented Sep 2, 2024

OhmSpectator commented Sep 2, 2024

shjala commented Sep 2, 2024

OhmSpectator commented Sep 2, 2024

shjala commented Sep 2, 2024

OhmSpectator commented Sep 2, 2024

eriknordmark commented Sep 2, 2024

shjala commented Sep 2, 2024

eriknordmark left a comment

Full Virtual TPM for VMs and Containers #4071

Full Virtual TPM for VMs and Containers #4071

Conversation

shjala commented Jul 9, 2024 • edited Loading

Problem with existing VTPM service

Full Virtual TPM in VMs and Containers

TODO

Eden Test

Manual Tests

Manually Test AzureIoTEdge (Ubuntu 20.04, Aziot 1.4.0 and EVE-TOOLS)

Manually Test AzureIoTEdge (Ubuntu 20.04, Aziot latest (1.5.x) without EVE-TOOLS, using Virtual TPM)

shjala commented Jul 16, 2024 • edited Loading

Please don't mind the many commits, I will squash them later. This is more or less ready for review even though there are tasks on the list (I'll do while the review is going on).

eriknordmark commented Jul 23, 2024

eriknordmark commented Jul 23, 2024

eriknordmark left a comment

Choose a reason for hiding this comment

shjala commented Jul 24, 2024

shjala commented Jul 25, 2024

shjala commented Sep 2, 2024

OhmSpectator commented Sep 2, 2024

shjala commented Sep 2, 2024

OhmSpectator commented Sep 2, 2024

shjala commented Sep 2, 2024

OhmSpectator commented Sep 2, 2024

eriknordmark commented Sep 2, 2024

shjala commented Sep 2, 2024

eriknordmark left a comment

Choose a reason for hiding this comment

shjala commented Jul 9, 2024 •

edited

Loading

shjala commented Jul 16, 2024 •

edited

Loading