[svlug] Serious Problem With Some LG CD-R Drives -was- DVD burner question

Mark S Bilk mark at cosmicpenguin.com
Tue Nov 4 15:11:41 PST 2003

In-Reply-To: <Pine.LNX.4.44.0311041102100.3404-100000 at slavent.doubleu.com>; from scott at doubleu.com on Tue, Nov 04, 2003 at 11:08:00AM -0800
Organization: http://www.cosmicpenguin.com/911

On Tue, Nov 04, 2003 at 11:08:00AM -0800, Scott Hess wrote:
>I had a 10-pack of DVD-R media (Memorex 4x, I think), and a DVD+RW disc
>(Ridata 2.4x or 4x, I think).  I've had no luck trying to rewrite the
>DVD+RW, or reformat, or anything.  Every time with the media error.
>I'm thinking of going with the LG drive that got the good reviews on Tom's 
>Hardware.  After poking around to see what the net thinks of it WRT Linux, 
>of course.  

Some LG (what does that stand for, anyhow?) CD-R drives have 
firmware that interprets the standard ATAPI FLUSH_CACHE command 
as a "load firmware" command and proceeds to overwrite the 
firmware with garbage, effectively killing the drive.  The 
proprietary LG driver for MS-Windows doesn't give that command,
but the latest Mandrake and probably other distros do.  

I don't know if any LG DVD-R drives have the same problem, nor
if it indicates the company has a negative attitude toward Linux.

Here are some results (edited in some cases for brevity) of a 
groups.google search for:

  lg cache 
with newsgroup = *linux* 


From: root <root at coyote.den>
Subject: Re: Damage CDROM drive by mounting?
Newsgroups: comp.os.linux.hardware
Message-ID: <HHQnb.16394$%e3.12331 at nwrddc03.gnilink.net>
Date: Wed, 29 Oct 2003 15:06:15 GMT

On Tuesday 28 October 2003 22:59 Tim wrote:

> sgarcia at bak.rr.com (Steve) wrote in message
> news:<f547625f.0310160926.dcc96de at posting.google.com>...
> The problem was that the LG CD-ROM manufacturer violated ATAPI
> standards and replaced the FLUSH_CACHE command with the vector for
> re-flashing the firmware.
> So, when Mandrake's Linux was booting, clearing the cache is a
> sensible undertaking, and should have no real effect on a CD-ROM
> drive... but it would clobber the LG CD-ROM drive's firmware.
> So, you could fix it if you re-flash the firmware of the CD-ROM, then
> it would be fixed.  A lot of devices won't let you re-flash them if
> the firmware is messed up (lame design, you must agree).

That also would appear to make it the manufacturers problem if its a
modifcation of the ATAPI interface specs/standards.   If that was my
drive, and its provable, I wouldn't shut up till I had a new fixed
drive in my posession.

Cheers, Gene
A mostly retired old coot


Newsgroups: comp.os.linux.advocacy
From: Ilari Liusvaara <noaddress at porn.org>
Subject: Re: MandrakeSoft identifies reason for CD-ROMs dying,  smh.com.au
Message-ID: <slrnbq5adj.nl6.noaddress at LK_Perkele_IV9.localdomain>
Date: Fri, 31 Oct 2003 18:28:03 +0000

Datagram from Kelsey Bjarnason incoming on netlink socket
<pan.2003. at xxnospamyy.lightspeed.bc.ca>. Dumping datagram.
> Which is a much less interesting question than why a simple cache flush is
> enough to kill a drive.

In some LG CD-ROMs, When OS sends FLUSH_CACHE, it gets interpretted by the

Worse yet, the LG drives do absolutely no checks on the new firmware that
it is OK, but proceed to overwrite the firmware with garbage.


From: Nix <nix-razor-pit at esperi.org.uk>
Newsgroups: uk.comp.os.linux
Subject: Re: Mandrake 9.2 BUG
Date: Thu, 30 Oct 2003 08:23:52 +0000
Message-ID: <87k76nw2uf.fsf at amaterasu.srvr.nix>

On Wed, 29 Oct 2003, Bill Unruh muttered drunkenly:
> LG defined a "flush buffers" code used according to the ATAPI standards
> on CDRW drives as meaning "download flash rom code" on their CDR
> machines. Such a reuse of codes is downright stupid at the best of
> times, and to define a code which means one thing on one set of machines
> to mean something which can destroy the machine on another is more than
> stupid. 

You are much too charitable :(

> (It is like a builder putting a switch on the dash which in cars with
> airconditioning switched on the air conditioner, and cars without
> drained all the oil out of the engine.

What a nice analogy. :)

> It is apparently the packet writing code which uses the flush buffers
> command, and uses the return of the command to test whether the drive is
> read  only or RW. (Not sure of that last statement but is seems to be
> implied by some of the stuff I have read.)

No, it doesn't do that because there'd be no point; no non-RW drives
support packet writing in firmware anyway (what'd be the point?),
and CD-R's are quite within their rights to implement FLUSH CACHE
anyway (and many do).

The problem it's solving is that many CD-RW drives have quite large
buffers and, er, not much of a desire to flush them fast. Without this
you could write something to the disk and still be at risk of data loss
if there was a power cut half an hour later.

(Bear in mind that packet writing turns a CD/RW into something more like
a big floppy, so we have to expect people to use them more like
conventional random- access media, rather than the `burn all things at
once' method which is common now.)

> Thus if you update your kernel to 2.4.22, beware. Mandrake being the
> first distro to use it got caught.

As far as I'm concerned the manufacturer of this crappy firmware got
caught violating the spec. Grotesquely.


From: tim at stonerise.com (Tim)
Newsgroups: comp.os.linux.advocacy
Subject: Re: Life's not so Good with Linux
Date: 29 Oct 2003 13:34:28 -0800
Message-ID: <31554a99.0310291334.38659258 at posting.google.com>

wjbell <wjbell at none.net> wrote in message news:<_F0nb.1261$Q%1.480 at newssvr14.news.prodigy.com>...
> >> Hmmmm... how come windows doesn't send a FLUSH_CACHE and waste the 
> >> drive???

The reason Windows doesn't send the command is because Windows didn't
write the driver, LG did.  LG provides Windows with a full list of
capabilities.  The way to determine whether a CD drive can write is to
issue the FLUSH_CACHE command and see what the result code is.

Windows has no need to determine this because the driver itself,
provided by LG, has their own proprietary methods of determining which
model and capabilities are supported.

It would be nice if LG supplied documentation or released their driver
source-code, but LG clearly has a vested interest in keeping secret
their deviances and defects.

Even to the layperson, FLUSH_CACHE should not conceivably be able to
harm a CD ROM drive in any way...  The cache of a read-only device
should be "flush"able without any serious effects.

The fact that LG turned the harmless FLUSH_CACHE vector into a
potentially device-killing FIRMWARE_UPDATE routine is nothing less
than engineering malpractice.  In fact, any ATAPI compliance software
would similarly have "ruined" the CD-ROMs even in safe-mode!

Basically, they renamed the FORMAT or FDISK command into COPY, so that
anyone doing a DIR will see the system work normally, but anyone doing
the less common COPY command would see that their hard-drives were

The ONLY excuse is malice or incompetence on the part of LG.

More information about the svlug mailing list