ACPI Debugging (3) - Debug ACPI Tables with Firmware Test Suite (FWTS)

Previous two articles, ACPI AML Debugger in Ubuntu 18.04 & Debug AML (DSD & SSDT) with ACPICA Utilities, discuss techniques for AML debugging; however ACPI specification includes many other important tables. Firmware Test Suite (FWTS) can play a big role in debugging ACPI implementation.

While other ACPI tables include fixed fields and lengths, it doesn't mean their contents cannot go wrong. In fact, many systems have many incorrect ACPI tables in their firmware. This can happen when systems are developed by integrating code from various parties without being modified specifically for customised hardware.

A test suite is a good solution to catch errors without involving too many engineering resources as it can identify errors automatically or semi-automatically. In particular, Firmware Test Suite is one of the best test suite (did I also mention it is freely available too?) and it is the recommended ACPI SCT by UEFI forum, who is the owner of the specification.

What is Firmware Test Suite (FWTS) ?

Firmware Test Suite (FWTS) is a test suite that performs sanity checks on firmware. It is intended to identify BIOS, UEFI, ACPI and many other errors and if appropriate it will try to explain the errors and give advice to help workaround or fix firmware bugs.

More information can be found at its official website.

Installing FWTS

Installation for Ubuntu Linux is as below:

1. Add stable PPA for latest FWTS (optional) - sudo add-apt-repository ppa:firmware-testing-team/ppa-fwts-stable
2. Install FWTS - sudo apt install fwts

FWTS packages for Other Linux distros can be found here.

Running Firmware Test Suite

FWTS is a command-line tool and its usage is "fwts cmds" (root access is usually required). The below is an example of checking FACP table
~$ sudo fwts fadt Running 1 tests, results appended to results.log Test: FADT Fixed ACPI Description Table tests. ACPI FADT Description Table flag info. 1 info only FADT checksum test. 1 passed FADT revision test. 1 warning ACPI FADT Description Table tests. 24 passed, 2 warnings, 2 skipped Test FADT SCI_EN bit is enabled. 1 passed Test FADT reset register. 2 passed ~$ sudo fwts mcfg Running 1 tests, results appended to results.log Test: MCFG PCI Express* memory mapped config space test. Validate MCFG table. 1 passed Validate MCFG PCI config space. 1 passed ~$ sudo fwts checksum Running 1 tests, results appended to results.log Test: ACPI table checksum test. ACPI table checksum test. 20 passed ~$ sudo fwts acpitables Running 1 tests, results appended to results.log Test: ACPI table headers sanity tests. Test ACPI headers. 17 passed, 1 failed

More test details can be found on its references.

Running Multiple Tests

FWTS can also run multiple tests by appending tests one after another as below:

~$ sudo fwts fadt mcfg checksum acpitables Running 4 tests, results appended to results.log Test: FADT Fixed ACPI Description Table tests. ACPI FADT Description Table flag info. 1 info only FADT checksum test. 1 passed FADT revision test. 1 warning ACPI FADT Description Table tests. 24 passed, 2 warnings, 2 skipped Test FADT SCI_EN bit is enabled. 1 passed Test FADT reset register. 2 passed Test: MCFG PCI Express* memory mapped config space test. Validate MCFG table. 1 passed Validate MCFG PCI config space. 1 passed Test: ACPI table checksum test. ACPI table checksum test. 20 passed Test: ACPI table headers sanity tests. Test ACPI headers. 17 passed, 1 failed

Running All ACPI Tests

FWTS provides a way to execute all tests in a category. For example, sudo fwts --acpitests run all ACPI tests all together. A similar command is sudo fwts --uefitests which will not be discussed in details here.

Let's See Some Examples...

FWTS is used for real projects. Here are some examples of tests we have run.

Checking FADT

Running fadt includes 6 tests that check whether system firmware/BIOS implements FACP table correctly.

Especially, the sizes of FACP table grow with ACPI specs, but some firmware/BIOS updates the FACP major and minor versions without adding the new fields in later ACPI spec. This can be caught by FWTS with ease. An example is Test 3 in results.log shown below.

~$ sudo fwts fadt ; cat results.log Running 1 tests, results appended to results.log Test: FADT Fixed ACPI Description Table tests. ACPI FADT Description Table flag info. 1 info only FADT checksum test. 1 passed FADT revision test. 1 warning ACPI FADT Description Table tests. 24 passed, 2 warnings, 2 skipped Test FADT SCI_EN bit is enabled. 1 passed Test FADT reset register. 2 passed Results generated by fwts: Version V19.01.00 (2019-01-16 03:58:48). Some of this work - Copyright (c) 1999 - 2019, Intel Corp. All rights reserved. Some of this work - Copyright (c) 2010 - 2019, Canonical. Some of this work - Copyright (c) 2016 - 2019, IBM. Some of this work - Copyright (c) 2017 - 2019, ARM Ltd. This test run on 24/01/19 at 17:16:21 on host Linux earth 4.15.0-43-generic #46-Ubuntu SMP Thu Dec 6 14:45:28 UTC 2018 x86_64. Command: "fwts fadt". Running tests: fadt. fadt: FADT Fixed ACPI Description Table tests. -------------------------------------------------------------------------------- FADT X_FIRMWARE_CTRL 64 bit pointer was zero, falling back to using FIRMWARE_CTRL 32 bit pointer. Test 1 of 6: ACPI FADT Description Table flag info. FADT: flag states WBINVD is set WBINVD_FLUSH is not set PROC_C1 is set P_LVL2_UP is not set PWR_BUTTON is not set SLP_BUTTON is set FIX_RTC is not set RTC_S4 is set TMR_VAL_EXT is not set DCK_CAP is not set RESET_REG_SUP is set SEALED_CASE is not set HEADLESS is not set CPU_SW_SLP is not set PCI_EXP_WAK is not set USE_PLATFORM_CLOCK is set S4_RTC_STS_VALID is set REMOTE_POWER_ON_CAPABLE is set FORCE_APIC_CLUSTER_MODEL is not set FORCE_APIC_PHYSICAL_DESTINATION_MODE is not set HW_REDUCED_ACPI is not set LOW_POWER_S0_IDLE_CAPABLE is not set FADT: IA-PC Boot Architecture flag states LEGACY_DEVICES is not set 8042 is not set VGA_NOT_PRESENT is not set MSI_NOT_SUPPORTED is not set PCIE_ASPM_CONTROLS is not set CMOS_RTC_NOT_PRESENT is not set FADT: ARM Boot Architecture flag states PSCI_COMPLIANT is not set PSCI_USE_HVC is not set Test 2 of 6: FADT checksum test. PASSED: Test 2, FADT checksum is correct Test 3 of 6: FADT revision test. FADT revision: 5.0 FADT table length: 268 WARNING: Test 3, FADT revision is outdated: 5.0 ADVICE: The most recent revision of the FADT defined in the ACPI specification is 6.2. While older revisions of the FADT can be used, newer ones may enable additional functionality that cannot be used until the FADT is updated. Test 4 of 6: ACPI FADT Description Table tests. PASSED: Test 4, Only one of FIRMWARE_CTRL and X_FIRMWARE_CTRL is non-zero. PASSED: Test 4, FADT 32 bit DSDT and 64 bit X_DSDT both point to the same physical address (0xd8fef1b0). ADVICE: While it is not correct to use both of the 32- and 64-bit DSDT address fields in recent versions of ACPI, they are at least the same address, which keeps the kernel from getting confused. At some point, the 32-bit DSDT address may get ignored so it is recommended that the FADT be upgraded to only use the 64-bit X_DSDT field. In the meantime, however, ACPI will still behave correctly. PASSED: Test 4, FADT first reserved field is one (compatible with ACPI 1.0). PASSED: Test 4, FADT second reserved field is zero. FADT Preferred PM Profile: 1 (Desktop) PASSED: Test 4, FADT has a valid preferred PM profile. FADT indicates ACPI IS NOT in reduced hardware mode. FADT SCI_INT is 9 PASSED: Test 4, FADT SMI_CMD indicates System Management Mode is supported, and the SCI Interrupt is non-zero. PASSED: Test 4, FADT SMI ACPI enable command is non-zero, and SMM is supported. PASSED: Test 4, FADT SMI ACPI disable command is non-zero, and SMM is supported. PASSED: Test 4, FADT indicates we are not in reduced hardware mode, and required FACS is present. PASSED: Test 4, FADT S4BIOS_REQ command is not set and FACS indicates it is not supported. PASSED: Test 4, FADT required PM1A_EVT_BLK field is non-zero PASSED: Test 4, FADT 32- and 64-bit PM1A_EVT_BLK fields are at least equal. ADVICE: Both FADT 32- and 64-bit PM1A_EVT_BLK fields are being used, but only one should be non-zero. However, they are at least equal so the kernel will at least have a usable value. SKIPPED: Test 4, FADT PM1B_EVT_BLK not being used. PASSED: Test 4, FADT required PM1A_CNT_BLK field is non-zero PASSED: Test 4, FADT 32- and 64-bit PM1A_CNT_BLK fields are at least equal. ADVICE: Both FADT 32- and 64-bit PM1A_CNT_BLK fields are being used, but only one should be non-zero. However, they are at least equal so the kernel will at least have a usable value. SKIPPED: Test 4, FADT PM1B_CNT_BLK not being used. PASSED: Test 4, FADT 32- and 64-bit PM2_CNT_BLK fields are at least equal. ADVICE: Both FADT 32- and 64-bit PM2_CNT_BLK fields are being used, but only one should be non-zero. However, they are at least equal so the kernel will at least have a usable value. PASSED: Test 4, FADT 32- and 64-bit PM_TMR_BLK fields are at least equal. ADVICE: Both FADT 32- and 64-bit PM_TMR_BLK fields are being used, but only one should be non-zero. However, they are at least equal so the kernel will at least have a usable value. PASSED: Test 4, FADT PM1_EVT_LEN >= 4. PASSED: Test 4, FADT PM1_CNT_LEN >= 2. PASSED: Test 4, FADT PM2_CNT_LEN >= 1. PASSED: Test 4, FADT PM_TMR_LEN is 4. PASSED: Test 4, FADT GPE0_BLK_LEN non-zero and a non-negative multiple of 2: 16. PASSED: Test 4, FADT GPE1_BLK_LEN is zero and GPE1_BLK is not supported. FADT GPE1_BASE is 0 Using P_BLK address of 0x1810 WARNING: Test 4, FADT P_LVL2_LAT is > 100 (101) but a P_BLK is defined. This implies a C2 state is not supported, but there is a P_BLK register block defined which implies there might be a C2 state that works. There is not enough information to determine if this is expected or not. WARNING: Test 4, FADT P_LVL3_LAT is > 1000 (1001) but a P_BLK is defined. This implies a C3 state is not supported, but there is a P_BLK register block defined which implies there might be a C3 state that works. There is not enough information to determine if this is expected or not. FADT FLUSH_SIZE is 1024 FADT FLUSH_STRIDE is 16 FADT DUTY_OFFSET is 0 FADT DUTY_WIDTH is 0 FADT DAY_ALRM is 13 FADT MON_ALRM is 0 FADT CENTURY is 50 PASSED: Test 4, FADT SLEEP_CONTROL_REG is null and not available when not in reduced hardware mode. PASSED: Test 4, FADT SLEEP_STATUS_REG is null and not available when not in reduced hardware mode. FADT Hypervisor Vendor Identity is 0 Test 5 of 6: Test FADT SCI_EN bit is enabled. FADT is greater than ACPI version 1.0 PASSED: Test 5, SCI_EN bit in PM1a Control Register Block is enabled. Test 6 of 6: Test FADT reset register. PASSED: Test 6, FADT reset register width is 8 bits wide as expected. PASSED: Test 6, FADT register bit offset is 0 as expected. ================================================================================ 28 passed, 0 failed, 3 warnings, 0 aborted, 2 skipped, 1 info only. ================================================================================ 28 passed, 0 failed, 3 warnings, 0 aborted, 2 skipped, 1 info only. Test Failure Summary ================================================================================ Critical failures: NONE High failures: NONE Medium failures: NONE Low failures: NONE Other failures: NONE Test |Pass |Fail |Abort|Warn |Skip |Info | ---------------+-----+-----+-----+-----+-----+-----+ fadt | 28| | | 3| 2| 1| ---------------+-----+-----+-----+-----+-----+-----+ Total: | 28| 0| 0| 3| 2| 1| ---------------+-----+-----+-----+-----+-----+-----+

 

Checking Table Checksum

Another common error in system firmware/BIOS is ACPI table checksum. When built by AML compiler, ex. iasl, table checksum is updated; however, it is a common practice that system firmware/BIOS modifies ACPI tables for hardware customisation during boot time. This can corrupt table checksum if firmware/BIOS does not update it accordingly.

Below is an example to verify the checksum fields of all ACPI tables.

~$ sudo fwts checksum ; cat results.log Running 1 tests, results appended to results.log Test: ACPI table checksum test. ACPI table checksum test. 20 passed Results generated by fwts: Version V19.01.00 (2019-01-16 03:58:48). Some of this work - Copyright (c) 1999 - 2019, Intel Corp. All rights reserved. Some of this work - Copyright (c) 2010 - 2019, Canonical. Some of this work - Copyright (c) 2016 - 2019, IBM. Some of this work - Copyright (c) 2017 - 2019, ARM Ltd. This test run on 24/01/19 at 17:23:54 on host Linux earth 4.15.0-43-generic #46-Ubuntu SMP Thu Dec 6 14:45:28 UTC 2018 x86_64. Command: "fwts checksum". Running tests: checksum. checksum: ACPI table checksum test. -------------------------------------------------------------------------------- Test 1 of 1: ACPI table checksum test. FADT X_FIRMWARE_CTRL 64 bit pointer was zero, falling back to using FIRMWARE_CTRL 32 bit pointer. PASSED: Test 1, Table RSDP has correct checksum 0x90. PASSED: Test 1, Table RSDP has correct extended checksum 0x8e. PASSED: Test 1, Table XSDT has correct checksum 0x21 PASSED: Test 1, Table DSDT has correct checksum 0xfc PASSED: Test 1, Table FACP has correct checksum 0xc9 PASSED: Test 1, Table APIC has correct checksum 0xbf PASSED: Test 1, Table FPDT has correct checksum 0x73 PASSED: Test 1, Table SLIC has correct checksum 0x6a PASSED: Test 1, Table SSDT has correct checksum 0x36 PASSED: Test 1, Table SSDT has correct checksum 0xb7 PASSED: Test 1, Table SSDT has correct checksum 0x3e PASSED: Test 1, Table HPET has correct checksum 0xa9 PASSED: Test 1, Table SSDT has correct checksum 0x3e PASSED: Test 1, Table MCFG has correct checksum 0x9c PASSED: Test 1, Table SSDT has correct checksum 0x31 PASSED: Test 1, Table ASF! has correct checksum 0xb8 PASSED: Test 1, Table BGRT has correct checksum 0x46 PASSED: Test 1, Table DMAR has correct checksum 0x37 PASSED: Test 1, Table RSDT has correct checksum 0xbd PASSED: Test 1, Table FACP has correct checksum 0xa1 ================================================================================ 20 passed, 0 failed, 0 warning, 0 aborted, 0 skipped, 0 info only. ================================================================================ 20 passed, 0 failed, 0 warning, 0 aborted, 0 skipped, 0 info only. Test Failure Summary ================================================================================ Critical failures: NONE High failures: NONE Medium failures: NONE Low failures: NONE Other failures: NONE Test |Pass |Fail |Abort|Warn |Skip |Info | ---------------+-----+-----+-----+-----+-----+-----+ checksum | 20| | | | | | ---------------+-----+-----+-----+-----+-----+-----+ Total: | 20| 0| 0| 0| 0| 0| ---------------+-----+-----+-----+-----+-----+-----+

Next Topic...

FWTS can analyze ACPI tables from other systems. This can be done with fwts --dumpfile=acpidump.log. More details will be discussed in the next blog.

ACPI Debugging (2) - Debug AML (DSDT & SSDT) with ACPICA Utilities

Using acpidbg on Ubuntu 18.04 x64 can be quite handy; however, the Linux kernel with ACPI_DEBUGGER is not always available, such as on Ubuntu for ARM. In such cases, acpica also provides a set of utilities, named acpica-tools, for ACPI debugging.

Installation

The acpica-tools can be installed by

•  sudo apt install acpica-tool

The latest source code can be downloaded from either of below, and it can be compiled and installed by “make” followed by “sudo make install“.

• Windows: https://acpica.org/downloads/binary-tools
• Ubuntu: sudo apt install acpica-tool

ACPICA Tools

The acpica-tools consist of the following utilities: acpibin, acpiexamples, acpihelp, acpisrc, iasl, acpidump, acpiexec, acpinames and acpixtract. This article focuses on how to use iasl, acpidump, acpiexec and acpixtract because they are commonly used in debugging.

• acpidump - collect ACPI tables from a running system
• acpixtract - extract ACPI tables from an acpidump file
• acpiexec - emulate ACPI tables from extracted acpi tables
• iasl - compile & disassemble ACPI tables

A Case Study - Airplane Mode

The airplane-mode button on laptops is usually implemented by both system BIOS and an OS driver. More specifically, it requires an ACPI device to be “present” in DSDT (or SSDT) table and a corresponding device driver. The figure below demonstrates how to debug if the airplane mode button fails to work.

Each computer brand has its specific implementation(s) and corresponding device driver(s). Some examples are listed below. Similarly, other driver source code can be found in Linux kernel.

• Acer - acer-wireless
• ASUS - asus-wireless
• Dell - dell-rbtn, intel-hid, intel-vbtn
• HP - hp-wireless
• Lenovo - idealpad-laptop, thinkpad_acpi
• Xiaomi - hp-wireless

Let’s use a Dell system as an example but the same technique also applies for others. To understand whether the airplane mode driver is loaded, one can run

$ lsmod | grep -E 'dell_rbtn|intel_hid|intel_vbtn' intel_hid 16384 0 sparse_keymap 16384 2 intel_hid,dell_wmi

As expected, the intel-hid is used on this particular Dell system.

As seen above, the “intel-hid” is used on this particular Dell system.
Let’s assume we know intel-hid is supposed to be used on this Dell system but it is not loaded for now. We can debug BIOS ASL/AML code using ACPICA utilities following the below steps.

Find Out ACPI Device's (intel-hid's) Hardware ID

A quick online search shows intel-hid.c is for an ACPI device with _HID “INT33D5”.

Extract and Disassemble ACPI Tables

• Get all ACPI tables: sudo acpidump > acpi.log
• Extract DSDT and SSDT: acpixtract acpi.log
• Disassemble tables: iasl -d *.dat

Check Whether INT33D5 is "Present"

• Find INT33D5 in DSDT or SSDT: grep INT33DT *.dsl

$ grep INT33D5 *.dsl dsdt.dsl: Name (_HID, "INT33D5") // _HID: Hardware ID
Once found, let's use vi to see more details in DSDT such as the device name (HIDD) as below:

Scope (_SB) { Device (HIDD) { Name (_HID, "INT33D5") // _HID: Hardware ID // ... skipped Method (_STA, 0, Serialized) // _STA: Status { If ((OIDE () >= One)) { Return (0x0F) } Else { Return (Zero) } } ...

• Check device status, i.e. Method (_STA)

In addition to the device declaration, Linux ACPI device driver is loaded when _STA returns “present”. As in the above ASL code, _STA can be present (return 0x0F) or absent (return Zero). One can continue to trace OIDE() -> OSID() and so on to find out what _STA returns.

Alternatively, using acpiexec is easier for this task.

Evaluate _SB.HIDD._STA with acpiexec

Using acpiexec is the same as acpidbg, but it loads ACPI tables from files and runs AML in emulation mode, i.e. it does not touch hardware registers or actual memory.

• Load ACPI tables: acpiexec *.dat
• Find HIDD path: find HIDD
• Execute HIDD's _STA: execute _SB.HIDD._STA

If _SB.HIDD._STA returns 0xF but intel-hid is not loaded, something is wrong with Linux kernel, ex. intel-hid is not compiled or is blacklisted. In this case, this should be forwarded to a Linux kernel engineer.

If _SB.HIDD._STA returns 0, we can continue to use acpiexec to find out OIDE() returns. In fact, we can use acpiexec to trace _SB._HIDD._STA

• Trace _SB.HIDD._STA: debug _SB.HIDD._STA

Tutorial of using acpiexec is the same as acpidbg and can be found in the previous blog.

Final Words

ACPICA tools are very useful for debugging ACPI bugs, and using it well can save much time. This tutorial touches surface of what ACPICA tools can do and more documents can be found on acpica.org website.

ACPI Debugging (1) - ACPI AML Runtime Debugger in Ubuntu 18.04 (x64)

ACPICA is an open-source project that provides an operating system (OS)-independent reference implementation. It also contains a list of utilities such as ASL compiler (iasl), acpiexec (an AML emulator) and so on. However, debugging AML on Linux in runtime wasn't provided in ACPICA ... not until Linux Kernel 4.13.

Enabling AML Debugging

The aml-debugger.txt, the instruction for enabling AML runtime debugger, is available at Documentation/acpi/ in Linux kernel source code. In short, two things are required for AML runtime debugging

• Enable AML debugging (CONFIG_ACPI_DEBUGGER=y & CONFIG_ACPI_DEBUGGER_USER=m) when compiling Linux kernel

• Compile an utility, acpidbg from Linux kernel (I uploaded a copy here)

While compiling a custom-build kernel is nothing new to kernel developers, it is often inconvenient for system firmware / BIOS developers who need to verify ACPI AML implementation in their firmware. Fortunately, Ubuntu 18.04 (x64) and later enable these config by default, and one can simply execute acpidbg on Ubuntu 18.04 - even on Ubuntu Live from USB too!

Running AML Debugging

Executing acpidbg on Ubuntu 18.04 (x64) is straight-forward
$ sudo ./acpidbg -

and "help" shows a list supported commands
- help General-Purpose Commands: ... ... Namespace Access Commands: ... ... Control Method Execution Commands: ... ...

Examples

acpidbg is particularly handy when one needs to evaluate any ACPI AML objects during runtime - especially ones that change under different conditions. This may be explained by some examples below:

Example 1 - To determine AC power status in runtime.

1. Find _PSR objects
2. Evaluate _PSR when AC is disconnected
3. Evaluate _PSR when AC is connected.

- find _PSR \_SB.PCI0.LPC.EC.AC._PSR Method 00000000c53d0a47 01 Args 0 Len 0023 Aml 00000000eab88cb2 - execute \_SB.PCI0.LPC.EC.AC._PSR Evaluating \_SB.PCI0.LPC.EC.AC._PSR Evaluation of \_SB.PCI0.LPC.EC.AC._PSR returned object 00000000c46fa555, external buffer length 18 [Integer] = 0000000000000000 - execute \_SB.PCI0.LPC.EC.AC._PSR Evaluating \_SB.PCI0.LPC.EC.AC._PSR Evaluation of \_SB.PCI0.LPC.EC.AC._PSR returned object 00000000c46fa555, external buffer length 18 [Integer] = 0000000000000001

acpidbg also works with complex objects such as packages.
Example 2 - To determine battery information and status.

1. Find _BIF and _BST objects
2. Evaluate _BIF and _BST objects (note _BST changes dynamically).

- find _BIF \_SB.PCI0.LPC.EC.BAT0._BIF Method 00000000817541c3 01 Args 0 Len 0040 Aml 000000009abb252e - execute \_SB.PCI0.LPC.EC.BAT0._BIF Evaluating \_SB.PCI0.LPC.EC.BAT0._BIF Evaluation of \_SB.PCI0.LPC.EC.BAT0._BIF returned object 00000000c46fa555, external buffer length 170 [Package] Contains 13 Elements: [Integer] = 0000000000000000 [Integer] = 000000000000ABE0 [Integer] = 0000000000009E34 [Integer] = 0000000000000001 [Integer] = 0000000000003B60 [Integer] = 00000000000007E9 [Integer] = 00000000000000C8 [Integer] = 0000000000000001 [Integer] = 0000000000000001 [String] Length 07 = "00HW027" [String] Length 05 = " 409" [String] Length 03 = "LiP" [String] Length 03 = "SMP" - find _BST \_SB.PCI0.LPC.EC.BAT0._BST Method 00000000ed98c5ba 01 Args 0 Len 001D Aml 00000000a48eaeb6 - execute \_SB.PCI0.LPC.EC.BAT0._BST Evaluating \_SB.PCI0.LPC.EC.BAT0._BST Evaluation of \_SB.PCI0.LPC.EC.BAT0._BST returned object 00000000c46fa555, external buffer length 78 [Package] Contains 4 Elements: [Integer] = 0000000000000001 [Integer] = 00000000000010C6 [Integer] = 00000000000088B8 [Integer] = 0000000000003FC9

acpidbg can decode bit fields and save time on counting bits. Try to run acpidbg to evaluate any _PLD objects and you will see what I mean :).

[Presentation] Firmware Test Suite - Uses, Development, Contribution and GPL

I did a presentation of "Firmware Test Suite - Uses, Development, Contribution and GPL" in UEFI Plugfest in Taipei on Nov 1 2017. The presentation is shared @ slideshare and can be viewed below:

Firmware Test Suite - Uses, Development, Contribution and GPL from Alex Hung

Why cannot ACPI _REV object be used to detect Linux anymore?

Linux community is never a fan of distinguishing Windows & Linux in BIOS (see my 2012's blog Why does Linux Say It Is not Linux in _OSI("Linux")?). However, many BIOS engineers seem to see the needs of to workaround BIOS bugs for Linux regardlessly. After realising _OSI("Linux") cannot be used by default, some creative BIOS engineers figured ACPI's _REV object could be used for this purpose some years ago, based on the following facts:

1. _REV definition in ACPI spec 5.1
5.7.4 \_REV (Revision Data Object) This predefined object evaluates to the revision of the ACPI Specification that the specified \_OS implements as a DWORD. Larger values are newer revisions of the ACPI specification. Arguments: None Return Value: An Integer containing the revision of the currently executing ACPI implementation

2. Windows returns 2 & Linux returns ACPI revision
If (_REV == 2) { // This is Windows } Elseif (_REV == 4 || _REV == 5) { // This is Linux }

While it is obvious it was Windows which did not follow ACPI spec, nothing can stop BIOS engineers from mis-using _REV as a mechanism to detect Linux in BIOS ASL codes.

It did not take long when ACPI Spec Work Group (ASWG), the owner and the publisher of ACPI spec, realised that many systems are mis-using _REV objects in their firmware. Eventually a spec contributer submitted a change to _REV object in ACPI 6.0 and later, and the _REV object is now defined as below:

5.7.4 \_REV (Revision Data Object) This predefined object evaluates to an Integer (DWORD) representing the revision of the ACPI Specification implemented by the specified \_OS. Arguments: None Return Value: An Integer representing the revision of the currently executing ACPI implementation. 1. Only ACPI 1 is supported, only 32-bit integers. 2. ACPI 2 or greater is supported. Both 32-bit and 64-bit integers are supported. Actual integer width depends on the revision of the DSDT (revision < 2 means 32-bit. >= 2 means 64-bit). Other values - Reserved

After ACPI 6.0 was published, the corresponding changes are also made to Linux kernel - the evaluation of _REV now returns 2 in kernel 4.1 and later (for example, Ubuntu 15.10 with kernel 4.2).

In order to provide backward compatibility, a new kernel parameter "acpi_rev_override" was introduced for those who still want to use _REV = 5 in their BIOS ASL code, but most of major Linux distributions do not include this kernel parameter by default, and it should be used with cautions.

If you are using one of the systems with _REV in the firmware, you can develop a kernel patch to add your system to the _REV quirk list acpi_rev_dmi_table in drivers/acpi/blacklist.c as below, like the patch I submitted.

static struct dmi_system_id acpi_rev_dmi_table[] __initdata = {
        {
         .callback = dmi_enable_rev_override,
         .ident = "DELL XPS 13 (2015)",
         .matches = {
                      DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
                      DMI_MATCH(DMI_PRODUCT_NAME, "XPS 13 9343"),
                },
        },
        ...

• DMI_SYS_VENDOR can be identified by sudo dmidecode -s system-manufacturer 
• DMI_PRODUCT_NAME can be identified by sudo dmidecode -s system-product-name.

Uitlity: Reading and Writing PCI, IO and memory with Linux Firmware Debug Kit (lfdk) in Linux

Many BIOS engineers are familiar with ru.exe in DOS and R/W everything in Windows.  One frequently asked question is whether there is an equivalent utility in Linux.  The answer is "sort of".

Linux Firmware Debug Kit (lfdk) provides read/write functions to PCI[1], IO and memory spaces.

lfdk's deb package is available but it is outdated. I usually download [3] and compile source code, and run ldfk manually in a terminal (ctrl + alt + T if you are running Ubuntu +Unity) as below:

1. sudo apt-get install -f git libncurses-dev
2. git clone https://github.com/alexhungce/lfdk1.git
3. cd lfdk
4. make
5. cd bin
6. sudo insmod ./lfdd_drv.ko
7. sudo ./lfdk

A familiar screen will pop up

The navigation  is to use arrow keys and hotkeys L, M and I to different spaces as instructed on the bottom of the lfdk.



[1] Built-in lspci command can be used to read PCI configuration space, too
[2] This is tested in Ubuntu Linux 12.04 to 15.04.
[3] One can also download the source code from sourceforge.net above.

[Presentation] BIOS, Linux and Firmware Test Suite in-between

I was invited to present "BIOS, Linux and Firmware Test Suite in-between" in Debian Taiwan's MiniDeb Conf in Taipei, Taiwan on Nov 9, 2013. The presentation is shared @ slideshare and can be viewed below:



Firmware Test Suite is an ongoing project that targets to automate firmware and hardware related tests. It receives some attentions from various parties including hardware, software and system vendors. (I also presented the similar topic in UEFI Plugfest in Sept, 2013, and the presentation slides are available @ http://uefi.org/learning_center/presentationsandvideos).

Info Session: Improving BIOS Compatibility on Linux

I had a chance to present a BIOS-related topic "Improving BIOS Compatibility on Linux" in workplace, and the presentation is publicly available. Hope it may interest someone. :-)

Why does Linux Say It Is not Linux in _OSI("Linux")?

First, let me explain what _OSI is.

Section 5.7.2 of ACPI specification defines it as below:
===================================================
_OSI (Operating System Interfaces)

This object provides the platform with the ability to query OSPM to determine the set of ACPI related interfaces, behaviors, or features that the operating system supports.

...

_OSI provides the platform with the ability to support new operating system versions and their associated features when they become available. OSPM can choose to expose new functionality based on the _OSI argument string. That is, OSPM can use the strings passed into _OSI to ensure compatibility between older platforms and newer operating systems by maintaining known compatible behavior for a platform.

===================================================

In theory, BIOS can use _OSI to provide different functionality for different operating systems to a specific platform.

Now, the answer to the question "Why does Linux Say It Is not Linux in _OSI("Linux")?" is - because Linux kernel wants to provide better quality and user-experiences.

Why is the differentiation is not desirable for Linux? Linux source code (drivers/acpi/osl.c) gives a history how the design decision was made as below:

/*
 * The story of _OSI(Linux)
 *
 * From pre-history through Linux-2.6.22,
 * Linux responded TRUE upon a BIOS OSI(Linux) query.
 *
 * Unfortunately, reference BIOS writers got wind of this
 * and put OSI(Linux) in their example code, quickly exposing
 * this string as ill-conceived and opening the door to
 * an un-bounded number of BIOS incompatibilities.
 *
 * For example, OSI(Linux) was used on resume to re-POST a
 * video card on one system, because Linux at that time
 * could not do a speedy restore in its native driver.
 * But then upon gaining quick native restore capability,
 * Linux has no way to tell the BIOS to skip the time-consuming
 * POST -- putting Linux at a permanent performance disadvantage.
 * On another system, the BIOS writer used OSI(Linux)
 * to infer native OS support for IPMI!  On other systems,
 * OSI(Linux) simply got in the way of Linux claiming to
 * be compatible with other operating systems, exposing
 * BIOS issues such as skipped device initialization.
 *
 * So "Linux" turned out to be a really poor chose of
 * OSI string, and from Linux-2.6.23 onward we respond FALSE.
 *
 * BIOS writers should NOT query _OSI(Linux) on future systems.
 * Linux will complain on the console when it sees it, and return FALSE.
 * To get Linux to return TRUE for your system  will require
 * a kernel source update to add a DMI entry,
 * or boot with "acpi_osi=Linux"
 */ 

ACPICA User Guide and Programmer Reference also has a similar discussion in Section 8.1.7.2 "Why ACPICA responds TRUE to _OSI (Windows)"

===================================================
ACPICA responds TRUE to all known Windows strings because ACPICA attempts to be fully compatible with the Windows implementation of ACPI. On the other hand, ACPICA responds FALSE to other operating system strings (such as “Linux”, “FreeBSD”, or “HP-UX”) because doing so has been seen to often cause serious problems. For example, on many platforms, the only path through the ASL code that has been fully tested by the manufacturer is in fact the path for “Windows”. By responding TRUE to other operating system strings, the ASL may execute paths that have had only limited or even no evaluation.
===================================================

To summarize, returning FALSE in _OSI("Linux") forces better software quality from two perspectives.

1. Not to "workaround" kernel errors in BIOS code - if Linux bugs are found, it should be updated to Linux kernel developers to have them fixed. 

2. To ensure Linux kernel executes code path that is well-tested - the same path Windows would execute. 

In the case of that _OSI("Linux") is really needed, it can be forced with kernel parameter acpi_osi=Linux as in the above comments.

Some OEM/ODM ship their systems with acpi_osi=Linux in their customized images; however, the solution has a downside. If a particular bug is not fixed in future kernel, or it is not a kernel bug but a hardware compatibility bug and cannot be fixed, a clean Linux installation by an end-user without adding this parameters will expose this bug. 

Instead, a better solution would be to have a BIOS setup option for workaround. If the bug is fixed, an end-user can  disable the workaround. A clean installation without acpi_osi=Linux will not expose the bug, either.

_OSI Strings in Ubuntu 

_OSI does not return true for a specific string.  Instead, it returns true for a list of strings.  As in Windows, _OSI in Linux returns true up to a Windows version.  For example, kernel 3.13 (used in Ubuntu 14.04) would report _OSI = true for Windows 7 (ex. "Windows 2012" and all versions before.  The below table lists Windows versions and their supports in Ubuntu Linux

_OSI argument	Windows version	Supported in Ubuntu
Windows 2000	Windows 2000	Pre-Dapper 6.06 LTS
Windows 2001	Windows XP	Pre-Dapper 6.06 LTS
Windows 2001	Windows XP SP1	Pre-Dapper 6.06 LTS
Windows 2001.1	Windows Server 2003	Pre-Dapper 6.06 LTS
Windows 2001 SP2	Windows XP SP2	Pre-Dapper 6.06 LTS
Windows 2001.1 SP1	Windows 2001.1 SP1	Hardy 8.04 LTS
Windows 2006	Windows Vista	Hardy 8.04 LTS
Windows 2006.1	Windows Server 2008	Lucid 10.04 LTS
Windows 2006 SP1	Windows Vista SP1	Lucid 10.04 LTS
Windows 2006 SP2	Windows Vista SP2	Natty 11.04
Windows 2009	Windows 7 and Server 2008 R2	Lucid 10.04 LTS
Windows 2012	Windows 8 and Server 2012	Raring 13.04
Windows 2013	Windows 8.1 and Server 2012 R2	Utopic 14.10
Windows 2015	Windows 10	Wily 15.10