modules/tpm-gpio-reset: TPM GPIO reset audit and assertion tools by tlaurion · Pull Request #2165 · linuxboot/heads

tlaurion · 2026-07-23T20:36:08Z

Adds tpm-gpio-detect (audit) and tpm-gpio-assert (execute) C binaries from the
tlaurion/tpm-gpio-fail fork
(commit 9052dbd, branch minimal-platform-fork), based on
Mate Kukri's original PoC.

modules/tpm-gpio-reset (new, 22 lines)

Builds both C binaries from the fork. Depends on pciutils and tpm2-tools.
Gated behind CONFIG_TPM2_TOOLS=y. Installs into initrd as
bin/tpm-gpio-detect and bin/tpm-gpio-assert.

tpm-gpio-detect (audit -- safe, read-only)

Uses libpci to find the ISA/LPC bridge by device class (no raw MMCFG).
Identifies the PCH platform, reads P2SB SBREG_BAR (hardcoded fallback
when hidden by FSP-S), then checks GPIO pad lock registers via /dev/mem:
PADCFGLOCK and PADCFGLOCKTX per PLTRST# pad (GPP_B13) and ESPI bus pins,
pad mode (NF1/GPIO), per-platform TIER classification. Self-documenting
ASCII table output with register values.

Exit codes: 0=all locked, >0=unlocked count, 2=not vulnerable by design,
126=unknown PCH, 127=hardware error.

tpm-gpio-assert (execute -- TPM may reboot, PCRs cleared, NVRAM preserved)

Calls tpm2 shutdown -c before and tpm2 startup -c after assertion.
Uses libpci to find the PCH, matching kukri's procedure.
Reads PADCFGLOCK/TX status, checks pad mode, performs NF1->GPIO
mode transition to create a PLTRST# reset edge:

All PCRs read via tpm2 pcrread before and after assertion
DW0 writeback verified (0x80000000), DW1 partial locking noted
Mechanism: NF1->GPIO mode transition (confirmed via coreboot GPIO code).
PADRSTCFG (DW0 bits 31:30) selects pad reset trigger, not attack vector.
TPM auto-check via tpm2 pcrread after tpm2 startup -c

Exit codes: 0=DW0 verified, 1=DW0 locked, 2=not vulnerable, 126=unknown
PCH, 127=hardware error. ASCII-only output.

doc/TPM_GPIO_Reset_Approaches.md (new, 442 lines)

Developer reference: 8 approaches attempted on NV4x ADL-P, register
addresses verified against coreboot 26.06, GPIO lock status per
Dasharo/Purism forks, NF1->GPIO mode transition mechanism. TIER ordered:
TIER 1 (confirmed), TIER 2 (unconfirmed), TIER 3 (uncertain), NOT VULNERABLE.

doc/TPM_GPIO_Reset_Vulnerability.md (new, 23 lines)

Points to Heads Wiki Threat Model and Recovery Shell page.

doc/BOARDS_AND_TESTERS.md (+156/-62)

TPM GPIO vulnerability section referencing wiki. Restructured EOL/ESU board table.

initrd/.bash_history (+4 lines)

Recovery shell convenience commands. Removes bash-based PoC script.

doc/faq.md (-120 lines), doc/keys.md, doc/security-model.md

Reduced to wiki cross-references to avoid duplication.

Tested

NV4x ADL-P (0x5182) -- CONFIRMED VULNERABLE (TIER 1)

All GPIO pads unlocked (PADCFGLOCK = 0x00000000)
DW0 in NF1 mode (0x40000400, mode=1)
NF1->GPIO transition verified (DW0 writeback = 0x80000000)
tpm2 shutdown/startup session management correct
PCRs cleared to zero after assertion -- TPM reset confirmed on hardware

All other 13 platform families are untested
Report with tpm-gpio-detect 2>&1 | tee log. See fork issues.

Cross-references

Heads Wiki PR Stable long term kernel is 4.9.38 #224 -- RecoveryShell tools, ADL-P CONFIRMED, flash guides
Heads Wiki Threat Model -- per-board TPM GPIO protection table
tlaurion/tpm-gpio-fail -- fork with 14 PCH families, C source

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

tlaurion · 2026-07-31T05:00:13Z

@Thrilleratplay @gaspar-ilom last review before merge please, conjointly to linuxboot/heads-wiki#224

tlaurion · 2026-07-31T05:05:56Z

Note that Alder Lake work has been made by @Sean-StarLabs on coreboot:

https://review.coreboot.org/q/topic:%22starlabs-lockdown%22

Can someone get a CVE?

Thrilleratplay

LGTM

Adds tpm-gpio-detect (audit) and tpm-gpio-assert (execute) from tlaurion/tpm-gpio-fail, based on Mate Kukri original PoC. Libpci PCI access. PCR output before/after assertion. ASCII output. Module depends on pciutils, tpm2-tools. Built when CONFIG_TPM2_TOOLS=y. TESTED: NV4x ADL-P (0x5182) — PCRs cleared to zero, confirmed. ALL OTHER PLATFORMS UNTESTED. Docs: TPM GPIO vulnerability docs point to Heads Wiki threat model. initrd: .bash_history entries for recovery shell testing. Signed-off-by: Thierry Laurion <insurgo@riseup.net>

tlaurion · 2026-08-02T17:50:49Z

Decided to extend @kukrimate Poc in fork and pack for all TPM2 based boards.

@macpijan nv4x_adl confirmed vulnerable.

From the repo's .bash_history at commit d8f1224:

# TPM GPIO Reset Vulnerability test: mount USB stick rw, audit PLTRST# lock status
mount-usb.sh --mode rw && tpm-gpio-detect 2>&1 | tee /media/tpm-gpio-detect.log
# Assert PLTRST# via P2SB PCR write and check TPM PCRs
tpm-gpio-assert 2>&1 | tee /media/tpm-gpio-assert.log

Fork commit: 0291c12 (Heads module modules/tpm-gpio-reset pins CONFIG_TPM2_TOOLS=y). Two commands: tpm-gpio-detect (audit) then tpm-gpio-assert (execute — handles tpm2 shutdown/startup internally).

echo "=== DETECT ===" && cat /media/user/public/tpm-gpio-detect.log && echo && echo "=== ASSERT ===" && cat /media/user/public/tpm-gpio-assert.log
=== DETECT ===
Found platform AlderLake P with PCH ADL-P PCH!
P2SB SBREG_BAR = 0xfd000000 (platform-specific)

--- PLTRST# Pad Lock Status ---
  GPIO       Signal       PADCFGLOCK             PADCFGLOCKTX          
  GPP_B13    PLTRST#      0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED

--- Pad Configuration ---
DW0 (0x07d0) = 0x40000400, mode: NF1 (bits 13:10=0x1)
NF1 mode -- mode transition CAN create PLTRST# reset edge

--- eSPI Bus Pin Lock Status ---
  GPIO       Signal       PADCFGLOCK             PADCFGLOCKTX          
  GPP_A1     ESPI_IO0     0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED
  GPP_A2     ESPI_IO1     0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED
  GPP_A3     ESPI_IO2     0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED
  GPP_A4     ESPI_IO3     0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED
  GPP_A5     ESPI_CS#     0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED
  GPP_A9     ESPI_CLK     0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED
  GPP_A14    ESPI_RESET#  0x00000000 (UNLOCKED)  0x00000000 (UNLOCKED)   <- UNLOCKED

--- Vulnerability Classification ---
TIER 1 -- CONFIRMED VULNERABLE: ADL-P mobile: NF1 mode confirmed, PLTRST# verified (TIER 1)
Attack confirmed working by community testing.

=== RESULT: unlocked pins found (unlocked: 8) ===

=== ASSERT ===
SBREG_BAR = 0xfd000000 (MEM)

Found Alder Point P

--- PADCFGLOCK Status (port 0x6e, offset 0x80) ---
PADCFGLOCK   = 0x00000000 (UNLOCKED, bit 13 clear)
PADCFGLOCKTX = 0x00000000 (UNLOCKED, bit 13 clear)

--- PCR State Before Assertion ---
  sha256:
    0 : 0x0000000000000000000000000000000000000000000000000000000000000000
    1 : 0x0000000000000000000000000000000000000000000000000000000000000000
    2 : 0xC45991E14F0B09EBFCD96FF16EAD26D7F3596E3F24677F587ED93FE66F3A9D16
    3 : 0x0000000000000000000000000000000000000000000000000000000000000000
    4 : 0x51737C77C481AA22095B38D38FC9FD494B0FFA4EAE7D3AC238082083D0AFD614
    5 : 0x5DCB5FCAF52F32ED2BDF8F7C5CBA01A54343D8356EB5E2B755B635F217F29EBB
    6 : 0x0000000000000000000000000000000000000000000000000000000000000000
    7 : 0x0000000000000000000000000000000000000000000000000000000000000000
    8 : 0x0000000000000000000000000000000000000000000000000000000000000000
    9 : 0x0000000000000000000000000000000000000000000000000000000000000000
    10: 0x0000000000000000000000000000000000000000000000000000000000000000
    11: 0x0000000000000000000000000000000000000000000000000000000000000000
    12: 0x0000000000000000000000000000000000000000000000000000000000000000
    13: 0x0000000000000000000000000000000000000000000000000000000000000000
    14: 0x0000000000000000000000000000000000000000000000000000000000000000
    15: 0x0000000000000000000000000000000000000000000000000000000000000000
    16: 0x0000000000000000000000000000000000000000000000000000000000000000
    17: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    18: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    19: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    20: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    21: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    22: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    23: 0x0000000000000000000000000000000000000000000000000000000000000000

--- TPM Session Preparation ---
Running tpm2 shutdown -c to cleanly shut down TPM session

*** WARNING: About to assert PLTRST# -- TPM may reboot (PCRs clear, NVRAM preserved) ***
*** This test reboots the TPM (PCRs clear, NVRAM preserved).    ***
PCR port=0x6e offset=0x7d0
DW0 (original) = 0x40000400  DW1 (original) = 0x0003c025

--- Pad Mode Check ---
Current mode: NF1 (DW0 bits 13:10 = 0x1)
Mode transition possible: NF1 -> GPIO -> NF1 (creates PLTRST# reset edge)
PLTRST# assertion CAN attempt mode transition reset
DW0 (asserted) = 0x80000000  DW1 (asserted) = 0x00000025
DW1 partially locked (TX control bits) - does not prevent reset
Registers restored
Running tpm2 startup -c to reinitialize TPM

--- PCR State After Assertion ---
  sha256:
    0 : 0x0000000000000000000000000000000000000000000000000000000000000000
    1 : 0x0000000000000000000000000000000000000000000000000000000000000000
    2 : 0x0000000000000000000000000000000000000000000000000000000000000000
    3 : 0x0000000000000000000000000000000000000000000000000000000000000000
    4 : 0x0000000000000000000000000000000000000000000000000000000000000000
    5 : 0x0000000000000000000000000000000000000000000000000000000000000000
    6 : 0x0000000000000000000000000000000000000000000000000000000000000000
    7 : 0x0000000000000000000000000000000000000000000000000000000000000000
    8 : 0x0000000000000000000000000000000000000000000000000000000000000000
    9 : 0x0000000000000000000000000000000000000000000000000000000000000000
    10: 0x0000000000000000000000000000000000000000000000000000000000000000
    11: 0x0000000000000000000000000000000000000000000000000000000000000000
    12: 0x0000000000000000000000000000000000000000000000000000000000000000
    13: 0x0000000000000000000000000000000000000000000000000000000000000000
    14: 0x0000000000000000000000000000000000000000000000000000000000000000
    15: 0x0000000000000000000000000000000000000000000000000000000000000000
    16: 0x0000000000000000000000000000000000000000000000000000000000000000
    17: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    18: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    19: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    20: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    21: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    22: 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
    23: 0x0000000000000000000000000000000000000000000000000000000000000000

=== RESULT: PCRs cleared to zero -- TPM reboot confirmed via sha256 PCR read ===

--- Next Steps ---
NF1->GPIO transition executed.
Report results: https://github.com/tlaurion/tpm-gpio-fail/issues
Include this log output.

updated https://ticket.coreboot.org/issues/576#note-6 with PoC output to show ADL P vulnerable.

… dirty builds PR #2165 changed blobs/m900/README.md and blobs/xx80/README.md, but the cache key only hashed *.sh files under blobs/. The stale cache contained old README content, and workspace attachment overwrote the fresh checkout, making git diff detect modifications and set GIT_STATUS=dirty. This caused all subsequent CircleCI builds to produce -dirty ROM artifacts. Fix: hash all files under blobs/ for the cache key, not just *.sh, so the cache invalidates whenever any tracked blob file changes. Signed-off-by: Thierry Laurion <insurgo@riseup.net>

kukrimate · 2026-08-20T16:00:50Z

Intel seems to have released an article about this: https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/technical-documentation/gpio-configuration-best-practices.html

Copilot AI review requested due to automatic review settings July 23, 2026 20:36

Copilot AI reviewed Jul 23, 2026

tlaurion mentioned this pull request Jul 23, 2026

doc: document TPM GPIO reset vulnerability and per-board lock status #2161

Closed

This comment was marked as duplicate.

Sign in to view

This comment has been minimized.

Sign in to view

tlaurion linked an issue Jul 23, 2026 that may be closed by this pull request

GPIO locking is broken on Kaby Lake and possibly other platforms #2159

Closed

tlaurion requested a review from JonathonHall-Purism July 23, 2026 20:51

tlaurion mentioned this pull request Jul 22, 2026

Heads-threat-model: per-board GPIO TPM reset protection table; uniformise 20 board flashing guide pages linuxboot/heads-wiki#221

Closed

tlaurion force-pushed the doc/tpm-gpio-vuln branch 17 times, most recently from b3e7502 to a2d99a3 Compare July 24, 2026 17:50

tlaurion mentioned this pull request Jul 24, 2026

boards: Add Star Labs Heads targets #2164

Open

14 tasks

tlaurion force-pushed the doc/tpm-gpio-vuln branch 3 times, most recently from f5c6544 to 4aa3965 Compare July 24, 2026 18:39

tlaurion force-pushed the doc/tpm-gpio-vuln branch 5 times, most recently from c527dc3 to d8f1224 Compare July 31, 2026 04:46

tlaurion marked this pull request as draft July 31, 2026 04:57

tlaurion marked this pull request as ready for review July 31, 2026 04:57

tlaurion force-pushed the doc/tpm-gpio-vuln branch 6 times, most recently from dc01c74 to a56a84b Compare July 31, 2026 06:12

Thrilleratplay approved these changes Jul 31, 2026

View reviewed changes

tlaurion force-pushed the doc/tpm-gpio-vuln branch 5 times, most recently from b647ef9 to de96f0c Compare July 31, 2026 17:28

tlaurion force-pushed the doc/tpm-gpio-vuln branch from de96f0c to 3ebc60c Compare July 31, 2026 20:32

tlaurion merged commit a605d04 into linuxboot:master Jul 31, 2026
5 of 10 checks passed

Kondix10 mentioned this pull request Aug 5, 2026

TPM GPIO reset vulnerability Dasharo/dasharo-issues#1908

Open

tlaurion mentioned this pull request Aug 7, 2026

.circleci/config.yml: hash all blobs files in cache key to prevent dirty builds #2183

Merged

tlaurion mentioned this pull request Aug 7, 2026

CircleCI publishes dirty build for clean commit, hash mismatch vs local build #2182

Closed

10 tasks

Navigation Menu

Uh oh!

Uh oh!

modules/tpm-gpio-reset: TPM GPIO reset audit and assertion tools - #2165

modules/tpm-gpio-reset: TPM GPIO reset audit and assertion tools#2165
tlaurion merged 4 commits into
linuxboot:masterfrom
tlaurion:doc/tpm-gpio-vuln

tlaurion commented Jul 23, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

This comment was marked as duplicate.

This comment has been minimized.

This comment has been minimized.

tlaurion commented Jul 31, 2026 •

edited

Loading

Uh oh!

tlaurion commented Jul 31, 2026 •

edited

Loading

Uh oh!

Thrilleratplay left a comment

Uh oh!

Uh oh!

tlaurion commented Aug 2, 2026

Uh oh!

kukrimate commented Aug 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Uh oh!

Conversation

tlaurion commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

modules/tpm-gpio-reset (new, 22 lines)

tpm-gpio-detect (audit -- safe, read-only)

tpm-gpio-assert (execute -- TPM may reboot, PCRs cleared, NVRAM preserved)

doc/TPM_GPIO_Reset_Approaches.md (new, 442 lines)

doc/TPM_GPIO_Reset_Vulnerability.md (new, 23 lines)

doc/BOARDS_AND_TESTERS.md (+156/-62)

initrd/.bash_history (+4 lines)

doc/faq.md (-120 lines), doc/keys.md, doc/security-model.md

Tested

Cross-references

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

This comment was marked as duplicate.

This comment has been minimized.

This comment has been minimized.

tlaurion commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tlaurion commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Thrilleratplay left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

tlaurion commented Aug 2, 2026

Uh oh!

kukrimate commented Aug 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

tlaurion commented Jul 23, 2026 •

edited

Loading

tlaurion commented Jul 31, 2026 •

edited

Loading

tlaurion commented Jul 31, 2026 •

edited

Loading