Another vulnerability

[Dave Emery <die@pig.die.com>]

With Intel, Hackers Check In When Bugs Check Out 

By Alexander Wolfe

SANTA CLARA, Calif. -- Intel's BIOS Update technology to quickly
fix bugs that crop up in its microprocessors without having to recall
the
chips may contain a Trojan horse -- a hole that could potentially enable
hackers to wreak havoc on the company's CPUs -- said a BIOS expert
familiar with the technology.However, other industry experts said they
believe Intel is staking out a pace-setting position with its
bug-busting
technology. They give the semiconductor giant kudos for using BIOS
Update to reduce the impact of bugs in the face of a verification crisis
that
makes it increasingly difficult to ensure that microprocessors with tens
of
millions of transistors are validated and free of flaws.

BIOS Update is a hidden feature that can fix bugs in Pentium Pro and
Pentium II CPUs by patching the microcode inside the microprocessor.
When the processor boots up, the BIOS loads the patches, which are
contained in a 2,048-byte-long BIOS Update data block that is supplied
by Intel.
"The problem is, the BIOS cannot verify whether the BIOS Update data
block contains real microcode or not," claimed one BIOS expert, who
requested anonymity. "As long as the header and the checksum are okay,
the BIOS will load that microcode into the microprocessor. Some hacker
could actually wipe out microcode in the CPU. There is nothing that can
prevent this."

Intel doesn't see such a scenario as a realistic threat, pointing to the
fact
that the BIOS Update data block is encrypted. "We've spent quite a lot
of time thinking about such scenarios to make sure we had sufficient
mechanisms in place so you couldn't introduce your own flavor of BIOS
Update into the processor," said Ajay Malhortra, a technical marketing
manager based here at Intel's microprocessor group. "Not only is the
data block containing the microcode patch encrypted, but once the
processor examines the header of the BIOS update, there are two levels
of encryption in the processor that must occur before it will
successfully
load the update."

But Intel's biggest security feature may lie in keeping the technical
details
behind its BIOS Update technology a closely guarded secret. "There is
no documentation," said Frank Binns, an architect in Intel's
microprocessor group. "It's not as if you can get an Intel 'Red Book'
with
this stuff written down. It's actually in the heads of less than 10
people in
the whole of Intel."

However, some experts remain unconvinced. "This is just like any other
technology -- if you want to reverse-engineer it, you can," said Ed
Curry,
president of Lone Star Evaluation Laboratories, a Georgetown, Texas
microprocessor benchmarking and testing company. "You can do it by
brute force, or a hacker could obtain information from someone inside
the
company or someone who had access to the documentation."

Indeed, Curry, who said he's made presentations on computer-security
issues to the U.S. Defense Department, said he believes microprocessor
hardware in general is much more vulnerable to hacking than is commonly
believed.

"This is the big hole in our government security programs," he said.
"They
don't look at hardware as well as they should; they only look at
software.
This goes beyond desktop computers. You have to remember that
microprocessors are now embedded in our weapons systems."

Nevertheless, it's widely believed that it would be tough for a hacker
to
fake a complete microcode patch, in no small measure because it's also
very difficult to obtain documentation that details the internal
representation -- word widths and usage of all the bits -- of Pentium
Pro
microcode. In the era of the 8086 and 8088, microcode documentation
was readily available. But such information is provided to selected
developers only under tight nondisclosure restrictions. "It's a tightly
held
secret," Intel's Binns said.

New-Tech Jitters

However, it is seen as more feasible for a hacker to successfully fake
the
header and checksum portion of the BIOS Update data block --
something that could still cause the microprocessor to crash or lock up.

According to another BIOS expert, talk of potential Trojan horses might
be nothing more than jitters about new technology. "This is a new thing
in
the market," said the expert, who likened it to the early days of flash
BIOS.

"There was a great fear factor when the industry started using flash
BIOSes," he said, "where concerns were raised that somebody could go
in and destroy a system by flashing in a new BIOS containing an errant
piece of code. I think today there's a fear that someone will play
around
with this BIOS Update feature and try to cause havoc with Intel's
CPUs."

As an added security precaution, some BIOS manufacturers limit access
to their software. "As a matter of policy, we don't make our BIOS code
available to anyone other than a system vendor or motherboard
manufacturer," said Thomas Benoit, corporate marketing manager at
BIOS vendor Phoenix Technologies, Natick, Mass. "We don't believe
anyone should be twiddling the bits in our BIOS code."

Irrespective of Trojan horse scenarios, many experts see Intel's
bug-busting technology as a boon. "This feature benefits everyone -- it
shouldn't be viewed as a liability, but as an asset," said Mark Huffman,
marketing manager at American Megatrends, in Norcross, Ga. "It allows
you to be able to update your processor without pulling it out of the
system. Obviously, you can flash in a new BIOS a lot quicker than you
can pop the case, pop the CPU and wait for a replacement."

Indeed, BIOS Update has already been successfully used in the field to
fix glitches in Pentium Pro-class CPUs, according to an Intel spokesman
and to sources at several major BIOS vendors.

"Yes, it is used," said an engineer at one vendor. "I personally know of
five different things in the Pentium Pro related to multiprocessing,
system
management interrupt and other areas."

"I think it'll be very useful," Phoenix Technology's Benoit said. "It's
really to Intel's benefit that BIOS vendors are implementing this
feature."

"It's a very good feature," said Laurent Gharda, vice president of
marketing at BIOS vendor Award Software International, in Mountain
View, Calif. "The downside is going to be lower performance, perhaps.
But the upside is avoiding a chip recall, as took place a few years
ago."
Intel's Pentium was recalled in January 1995 following the revelation of
a
bug in the processor's floating-point divide operations.

Moreover, some say BIOS Update may signal the start of an
industrywide trend. "These new Pentium-class clone CPUs that have
recently been announced -- like the Centaur microprocessor -- they're
going to do the same type of process," said Huffman at American
Megatrends. Centaur -- officially the IDT-C6 -- is made by Centaur
Technology, an Austin, Texas-based subsidiary of Integrated Device
Technology. It was introduced in May and delivered to beta customers in
Taiwan last month.

At Centaur, a spokesman said, "The current silicon we are sampling has
that capability, but in the production version of the chip we are
dropping
the feature, because it necessitates an increased die size."

As a result, any bugs that crop up will have to be fixed via a mask
revision
-- a path the spokesman described as preferable. "Ideally, you want to
do
fixes by mask changes," he said. "That way, you'll have clean silicon
moving forward. Otherwise, you have lots of different versions of BIOS
floating around." But Centaur can easily add the feature back into
future
versions, if it wishes.

For its part, Advanced Micro Devices of Sunnyvale, Calif., does not have
the feature in its K5 and K6 microprocessors, according to a company
spokesman. "There are some errata that can't be fixed by a BIOS update
-- specifically, a hardwired instruction can't be changed." He added
that
AMD has the ability to add the feature into future designs, if it deems
it
necessary.

Still, Huffman of American Megatrends thinks the BIOS Update feature
has legs. "I think you'll see a trend toward CPU manufacturers
incorporating this capability so they can perform microcode updates in
the
field," he said. "It gives them more flexibility in their manufacturing
process -- they can keep their fab lines running and don't have to stop
them to make a mask change and switch to a new stepping every time
there's an
erratum. More important, they don't have to recall the stepping that has
the bug. They can just issue a BIOS update."

Intel doesn't tell the BIOS vendors what bugs are being fixed in any
given BIOS Update. However, there appears to be a way to figure that
out.

 "It's true you can't see what's happening from a binary standpoint,"
the
BIOS expert who requested anonymity said. "But Intel does release
errata along with the update, which gives an explanation of what the
update is for. To that extent, you know what they're fixing, though you
don't know the exact binary details of what's occurring."

Although the BIOS Update feature is firmly in place in the Pentium Pro
and Pentium II families, Intel declined to comment on whether it is
being
used in Pentiums with the MMX multimedia extensions. Looking ahead,
deciding whether to implement the technology in future CPU families will
involve architectural considerations that extend far beyond a desire to
bust
bugs.

"We're just learning the power this technology really has," Intel's
Malhortra said. "In concert with that, we're also becoming more aware of
some of its limitations. For example, the trade-off between die size
that's
used for microcode-patchable space [i.e., for the BIOS Update feature]
vs. die size that can be devoted to performance enhancements or to
additional micro-architectural features is a tough one."

Validation Boost

"One could make the argument that, with improved validation processes,
you won't need to expand silicon real estate devoted to the
microcode-patch feature, because early validation would catch the bulk
of
problems," Malhortra added.

Nevertheless, there's a growing concern that microprocessor bugs could
become a bigger problem as 64-bit CPU architectures -- which will be
orders of magnitude more difficult to validate than current designs --
are
introduced toward the end of the decade.

"It's becoming abundantly clear that the ability to manufacture in high
volume and to provide a reliable product through validation are somewhat
mutually exclusive," Intel's Binns said. "It takes a fairly large amount
of
time to wring all the errata out of a processor. Fixing errata by making
changes to silicon is OK, if you can make those changes quickly.
Unfortunately, with the complexity of the processors we've got today,
that's not acceptable. The smarter we can get with features like this,
the
less errata we bring to market. And if we do see errata after we ship,
we
can correct them in situ."

----- End of forwarded message from Richard Crisp -----