The Linux-ready Firmware Developer Kit team is pleased to announce the
release R1 of the kit. In this release many bugs have been fixed and several
key enhancements have been done to help the ease of use of the kit, and
several new tests have been added.
The Linux-ready Firmware Developer Kit is a tool to test how well Linux
works together with the firmware (BIOS or EFI) of your machine, and is
designed for use by both firmware development teams and Linux kernel
hackers to prevent and diagnose firmware bugs.
Summary
=======
Enhancements
* Inclusion of the Linux 2.6.19 kernel for the latest hardware support
* Include several Linux distribution kernels for testing
* Many bugfixes
* The Serial-console mode now detects speed automatically
* Prototype IA64 support
* Automatic mode
New Tests
=========
* Suspend-Resume
* basic HPET test
* P-state coordination
* Thermal Trip Points
* 32/64 FADT test
* SSDT AML test
* Microcode version test
* DMI table test
* VMX enabled test
* OS/2 gap test
* APIC edge/level test
* PCI Express maxreadreq test
* _SUN test
You can download this latest release of the kit from
http://www.linuxfirmwarekit.org
The Linux-ready Firmware Developer Kit team
Jacob Pan
Rolla Selbak
Arjan van de Ven
Changes in R1 of the Linux-ready Firmware Developer Kit
-------------------------------------------------------
Features
--------
Linux distribution kernels
Several people have asked us to make it easy to also test vendor
kernels with the BIOS, and not just the vanilla kernel.org kernel.
The standard ISO image contains 3 such kernels now, and the build
system allows adding your own very easily.
General improvements
--------------------
Infrastructure
The mini-OS on the CD has been upgraded to the Fedora Core 6 rpms.
This allows us to leverage the suspend/resume infrastructure in FC6.
Kernel
The kernel on the CD image has been updated to version 2.6.19
Serial console
Thomas Renninger from Novell/SuSE contributed code that detects the
selected serial console speed automatically, rather than using a
fixed speed as R0 did.
Prototype IA64 support
The firmware kit runs on IA64 now; however no ISO image is available
yet.
Automatic mode
The firmware kit can now run in "auto pilot" mode, where the entire
kit is non-interactive and just saves the results before exiting.
Build system
The build process has been changed to link against the libraries
that will be on the final CD image, rather than the ones on the
system. This should make the build process a lot more portable
than it was in R0.
New tests
---------
Suspend-Resume
This manual test allows you to test if Suspend (to ram), and
more importantly, Resume works on the machine
basic HPET test
This test verifies if the kernel detects the HPET component of
the chipset properly. Firmware writers are strongly encouraged
to default-enable HPET to allow Linux to do fast and accurate
timekeeping.
P-state coordination
Current dual core processors have a behavior that the effective
frequency of both cores in a package is the maximum of the set
frequencies of the cores. This is called "Hardware coordination".
However some firmware does not get this right and uses "software
coordination" where both cores run at the speed that was programmed
into either of the cores.
Thermal Trip Points
Frank Seidel from Novell/SuSE has contributed a test for the thermal
trip points in a system.
32/64 FADT test
The FADT ACPI table is sometimes available in a 32 bit and a 64 bit
version. This test verifies that the common fields in these tables
are identical; they need to be since they describe the same properties.
SSDT AML test
In addition to the DSDT, some firmware has SSDT tables which are
supplemental AML tables to the DSDT. The firmware kit now performs
the same checks on the SSDT tables as it does on the DSDT.
Microcode version test
Intel CPUs have a feature called "microcode update", which makes it
possible to field update some non-performance critical behaviors of
the processor. The firmware is responsible to load a recent enough
version of this microcode into the processor; this test checks if
a more recent version exists.
DMI table test
The DMI table is used by Linux in several places, for example for
model specific workarounds. This test checks the DMI table against
a set of common mistakes, such as out-of-range values and reference
values that are copied but not adjusted from a reference
implementation.
VMX enabled test
Some firmware disables the VMX Virtualization extensions entirely;
this is unfortunate since Xen and KVM cannot use these extensions on
the machine. This test checks for this.
OS/2 gap test
For old versions of the OS/2* operating system, the firmware needed
to leave a gap in memory between 15Mb and 16Mb. However this gap
breaks various bootloaders that are used by Linux distributions,
and should never be enabled by default.
APIC edge/level test
When using apics, the legacy interrupts on a system should be
edge-triggered, while non-legacy interrupts should be level-triggered.
If this is not done correctly, interrupts get either lost or cause
an interrupt storm, both can hang the Linux kernel.
PCI Express maxreadreq test
PCI Express cards have a tunable buffer size with a default size of
128. The firmware is responsible for programming this buffer size to
a higher value during POST for optimal performance. This test
verifies that all PCI Express cards in a system have a higher tuned
setting. This test was suggested by Roland Dreier from Cisco after
a major Infiniband performance issue was diagnosed to be a non-tuned
maxreadreq.
_SUN test
PCI slots may have numbers (this is needed to tell the user "the
card in slot 4 can now be hot-unplugged safely). Some firmware
has assigned duplicate numbers to some slots, and this prevents
PCI/PCI-E hotplug from working reliably.