ATA messages via SCSI layer

[ link: ata-via-scsi | tags: ata scsi linux kernel | updated: Sat, 05 May 2007 12:45:01 ]

I've been working on a contract for Symbio Technologies for the last month. They are makers of a few thin client terminals.

My work for Symbio involves talking to a SATA hard disk using ATA command set. What makes this a bit more interesting is that /dev/hda is the way of the past. New devices are covered by libata drives which fit into the SCSI subsystem.

So, the challenge for me was how to send raw ATA messages using the SCSI layer to the SATA drive. Besides the fact that the interface is sparsely documented, it was pretty easy.

Specs

Let's start off with ATA commands set. You will need to get a copy of the ATA/ATAPI-7 spec from the T13 committee (for some reason these links don't work anymore). Any command you send to the device will look like what's outlined in this document.

As mentioned above, the new way of interfacing with the ATA devices is via SATA. There is even a documented translation method called SCSI ATA Translation (or SAT). This is maintained by the T10 committee, and you want to download the SAT2 document.

WARNING

Before you proceed... I have no idea what I am talking about. The stuff below may make your disk explode.

IDE subsystem

First let's review the history.

Once upon a time there was IDE, and we had a whole slew of IDE devices under CONFIG_IDE in the kernel .config file.

To communicate with these devices you would use one of the IOCTL's:

  • HDIO_DRIVE_TASK - can issue commands that have LBA commands, cannot read or write data,
  • HDIO_DRIVE_CMD - can issue commands that read blocks of data, cannot send commands with LBA parameters, and
  • HDIO_DRIVE_TASKFILE - can issue pretty much anything, but is more complicated to use.

Each of these IOCTLs passes down a single buffer that is used for control, input and output data.

As you can see the first two are useless if you want to do any real communication. They do seem to be helpful for some SMART commands and for some other STATUS commands. For anything else you need to use HDIO_DRIVE_TASKFILE.

  • Here is another example which gets the SMART STATUS from the drive.

    I am using this because SMART STATUS is a no data command, so we can use the HDIO_DRIVE_TASK ioctl.

    This may make your eyes bleed, but truly this is how you use this interface.

    int fd, rc;
    uint8_t buf[7];
    
    
    fd = open ("/dev/hda", O_RDWR);
    
    
    buf[0] = WIN_SMART;     // command id
    buf[1] = SMART_STATUS;  // feature id
    buf[2] = 1;             // number of sectors
    buf[3] = 0x00;          // LBA low
    buf[4] = 0x4F;          // LBA mid
    buf[5] = 0xC2;          // LBA high
    buf[6] = 0;             // device select
    
    
    rc = ioctl (fd, HDIO_DRIVE_TASK, buf);
    
    
    printf ("status %02x\n", buf[0]);
    printf ("error  %02x\n", buf[1]);
    printf ("n sect %02x\n", buf[2]);
    printf ("LBA L  %02x\n", buf[3]);
    printf ("LBA M  %02x\n", buf[4]);
    printf ("LBA H  %02x\n", buf[5]);
    printf ("select %02x\n", buf[6]);
    

    Thing to note is that the descriptor size is 7 bytes, and there is no data transfered in this command. If you give the wrong command or feature you may cause a SATA reset and not be able to talk to your drive anymore.

    It's pretty awful of an interface, and it gets a bit worse...

  • Here is an example of getting the IDENTIFY message from the drive:

    Here I am using an IDENTIFY request as it returns 1 block of data from the device.

    int fd, rc;
    uint8_t buf[4 + 512];
    
    
    fd = open ("/dev/hda", O_RDWR);
    
    
    buf[0] = WIN_IDENTIFY;  // command id
    buf[1] = 1;             // number of sectors
    buf[2] = 0;             // feature id
    buf[3] = 1;             // number of sectors
    
    
    rc = ioctl (fd, HDIO_DRIVE_CMD, buf);
    
    
    printf ("status %02x\n", buf[0]);
    printf ("error  %02x\n", buf[1]);
    printf ("n sect %02x\n", buf[2]);
    

    In this interface the descriptor is 4 bytes, and it's followed by a multiple of 512 blocks for the data transfer. You can only read with this interface, and you can cause SATA resets if you're not careful.

    But, wait... it gets worse.

  • Here is an example of getting SMART log records...

    Here I am using SMART READ VALUES command which returns one block of data...

    int fd, rc;
    uint8_t buf[4 + 512];
    
    
    fd = open ("/dev/hda", O_RDWR);
    
    
    buf[0] = WIN_SMART;         // command id
    buf[1] = 0;                 // LBA low
    buf[2] = SMART_READ_VALUES; // feature id
    buf[3] = 1;                 // number of sectors
    
    
    rc = ioctl (fd, HDIO_DRIVE_CMD, buf);
    
    
    printf ("status %02x\n", buf[0]);
    printf ("error  %02x\n", buf[1]);
    printf ("n sect %02x\n", buf[2]);
    

    I said it gets worse. It's worse because the encoding of the descriptor depends on the command id. And a lot of this is not documented, so you have to read the code.

Now onto the new interface...

SCSI subsystem

libata is the name of a set of drivers and some glue code that allows the SCSI layer to talk to ATA devices. From user land, for quite some time SATA drives have been showing up as /dev/sda (a SCSI disk device). More recently this was extended to PATA devices, and some old drivers were ported to libata.

libata does support HDIO_DRIVE_CMD and HDIO_DRIVE_TASK for backwards compatibility -- mainly for hdparm, hdtemp, smartmontools, etc. However it does not support HDIO_DRIVE_TASKFILE, and it is suggested that people use SG_IO instead.

The SCSI ioctl SG_IO allows the same features as the IDE TASKFILE, but sends the commands through the SCSI subsystem. Since the SCSI and ATA commands are nothing alike libata maintains translation code to do this for common commands, as mandated by the SAT2 T10 committee document referenced above.

If a command does not fit into the SAT2 translation you can wrap the ATA command with an SG_IO header and pass it using ATA16 pass-through commands.

I am not going to talk about the easy case of using SCSI ioctls, for which there are SAT mappings, to talk to the ATA device... that's easy and pretty well documented. For examples see the source code for the sginfo tool from sg3-utils package.

I will however show how to send the three different types of commands:

  • First, an example of how to issue a no data command:

    int fd, rc;
    struct sg_io_hdr sg_io;
    uint8_t cdb[16];
    uint8_t sense[32];
    
    
    // under most circumstances /dev/sda maps to /dev/sg0
    // only SCSI generic interfaces can use SG_IO ioctl
    // see output of: sginfo -l
    fd = open ("/dev/sg0", O_RDWR);
    
    
    memset (&sg_io, 0, sizeof(sg_io));
    memset (&cdb, 0, sizeof(cdb));
    memset (&sense, 0, sizeof(sense));
    
    
    sg_io.interface_id    = 'S';
    sg_io.cmdp            = cdb;
    sg_io.cmd_len         = sizeof(cdb);
    sg_io.dxferp          = NULL;
    sg_io.dxfer_len       = 0;
    sg_io.dxfer_direction = SG_DXFER_NONE;
    sg_io.sbp             = sense;
    sg_io.mx_sb_len       = sizeof(sense);
    sg_io.timeout         = 5000;   // 5 seconds
    
    
    cdb[0] = 0x85;                  // pass-through ATA16 command (no translation)
    cdb[1] = (3 << 1);              // no data
    cdb[2] = 0x20;                  // no data
    cdb[4] = feature_id;            // ATA feature ID
    cdb[6] = 0;                     // number of sectors
    cdb[7] = lba_low >> 8;
    cdb[8] = lba_low;
    cdb[9] = lba_mid >> 8;
    cdb[10] = lba_mid;
    cdb[11] = lba_high >> 8;
    cdb[12] = lba_high;
    cdb[14] = command_id;           // ATA command ID
    
    
    rc = ioctl (fd, SG_IO, &sg_io);
    
    
    // check expected magic
    if (sense[0] != 0x72 || sense[7] != 0x0e || sense[8] != 0x09
                    || sense[9] != 0x0c || sense[10] != 0x00)
            error;
    
    
    printf ("error  = %02x", sense[11]);    // 0x00 means success
    printf ("status = %02x", sense[21]);    // 0x50 means success
    
  • Next, an example of how to issue a data-in (or from device) command:

    int fd, rc;
    struct sg_io_hdr sg_io;
    uint8_t cdb[16];
    uint8_t sense[32];
    
    
    fd = open ("/dev/sg0", O_RDWR);
    
    
    memset (&sg_io, 0, sizeof(sg_io));
    memset (&cdb, 0, sizeof(cdb));
    memset (&sense, 0, sizeof(sense));
    
    
    sg_io.interface_id    = 'S';
    sg_io.cmdp            = cdb;
    sg_io.cmd_len         = sizeof(cdb);
    sg_io.dxferp          = data_in_buffer;
    sg_io.dxfer_len       = data_in_length;         // multiple of 512
    sg_io.dxfer_direction = SG_DXFER_FROM_DEV;
    sg_io.sbp             = sense;
    sg_io.mx_sb_len       = sizeof(sense);
    sg_io.timeout         = 5000;                   // 5 seconds
    
    
    cdb[0] = 0x85;                  // pass-through ATA16 command (no translation)
    cdb[1] = (4 << 1);              // data-in
    cdb[2] = 0x2e;                  // data-in
    cdb[4] = feature_id;            // ATA feature ID
    cdb[6] = 1;                     // number of sectors
    cdb[7] = lba_low >> 8;
    cdb[8] = lba_low;
    cdb[9] = lba_mid >> 8;
    cdb[10] = lba_mid;
    cdb[11] = lba_high >> 8;
    cdb[12] = lba_high;
    cdb[14] = command_id;           // ATA command ID
    
    
    rc = ioctl (fd, SG_IO, &sg_io);
    
    
    // check expected magic
    if (sense[0] != 0x72 || sense[7] != 0x0e || sense[8] != 0x09
                    || sense[9] != 0x0c || sense[10] != 0x00)
            error;
    
    
    printf ("error  = %02x", sense[11]);    // 0x00 means success
    printf ("status = %02x", sense[21]);    // 0x50 means success
    
  • And last, an example of how to issue a data-out command:

    int fd, rc;
    struct sg_io_hdr sg_io;
    uint8_t cdb[16];
    uint8_t sense[32];
    
    
    fd = open ("/dev/sg0", O_RDWR);
    
    
    memset (&sg_io, 0, sizeof(sg_io));
    memset (&cdb, 0, sizeof(cdb));
    memset (&sense, 0, sizeof(sense));
    
    
    sg_io.interface_id    = 'S';
    sg_io.cmdp            = cdb;
    sg_io.cmd_len         = sizeof(cdb);
    sg_io.dxferp          = data_out_buffer;
    sg_io.dxfer_len       = data_out_length;        // multiple of 512
    sg_io.dxfer_direction = SG_DXFER_TO_DEV;
    sg_io.sbp             = sense;
    sg_io.mx_sb_len       = sizeof(sense);
    sg_io.timeout         = 5000;                   // 5 seconds
    
    
    cdb[0] = 0x85;                  // pass-through ATA16 command (no translation)
    cdb[1] = (5 << 1);              // data-out
    cdb[2] = 0x26;                  // data-out
    cdb[4] = feature_id;            // ATA feature ID
    cdb[6] = 1;                     // number of sectors
    cdb[7] = lba_low >> 8;
    cdb[8] = lba_low;
    cdb[9] = lba_mid >> 8;
    cdb[10] = lba_mid;
    cdb[11] = lba_high >> 8;
    cdb[12] = lba_high;
    cdb[14] = command_id;           // ATA command ID
    
    
    rc = ioctl (fd, SG_IO, &sg_io);
    
    
    // check expected magic
    if (sense[0] != 0x72 || sense[7] != 0x0e || sense[8] != 0x09
                    || sense[9] != 0x0c || sense[10] != 0x00)
            error;
    
    
    printf ("error  = %02x", sense[11]);    // 0x00 means success
    printf ("status = %02x", sense[21]);    // 0x50 means success
    

Hope that some of this was useful. Please let me know if I missed something.

Bart's Blog

/ata-via-scsi

About

I am an embedded Linux software developer and consultant operating under Jukie Networks Inc in Ottawa, Canada.

RSS Feed - Full Content

RSS Feed - Full Content

ipv6 ready

Valid XHTML 1.0 Transitional

Created with Vim

Created with Perl




Tags

abiword adam android apt asm ata audio bacula bash bios board-game boot bug busybox bzr c code conference debian debug desktop devel disk dns dpkg dream email embedded find firefox font fs fun g1 gdb geode git git scm git-find git-graft go google gpg grub hardware html hw inkscape ion3 ipsec ipv6 ipv6summit.ca irc irssi joke kdb kernel klips kqemu kvm laptop lastfm ldap lighttpd linux lua lvm mail make mercurial meta mindblown mmc mouse-free mp3 mpd music mutt neovim net nexusone nfs nvim oclug ogre ols oom OOo openswan opteron osx ottawa perl photography power pxe qemu raid redhat rogers ruby sata sbc scm screen script scsi security sheeva shell skype spam squid ssd ssh suse svn tags talk thinkpad ubuntu uml unix urxvt usb uzbl vi video vim vimgit virt vpn vserver web weechat wifi wind wmii wmiirc-lua work x x86 x86emu xen xen-box-setup xfs xterm zsh

Posts

[ 20240214132102 ]
new git learnings

[ 20240203164450 ]
new vim config

[ 20181212145129 ]
improving find -exec efficiency

[ 20130409092758 ]
automount mmcblk devices

[ 20130316190507 ]
tunnelbroker vs IRC

[ 20110616180255 ]
HDD -> SDD

[ 20110517204617 ]
how to manually create a 6in4 tunnel

[ 20110501202915 ]
Presentation slides published!

[ 20110320190458 ]
What is IPv6, and why should I care?

[ 20110309230106 ]
Ottawa IPv6 Summit 2011

[ 20110309225149 ]
I am now an IPv6 Sage

[ 20101214145558 ]
It's a Holiday Miracle

[ 20101103095130 ]
ipv6 on your desktop in 2 steps

[ 20101020113026 ]
growing a live LVM volume

[ 20100915175649 ]
distributing DNS list through radvd

[ 20100908173402 ]
Canadian ipv6 drought

[ 20100903190623 ]
ipv6 certification

[ 20100723192259 ]
m4a to mp3

[ 20100723134522 ]
git 1.7.2 is out

[ 20100523092001 ]
console=ttyS0 with grub2

[ 20100426162644 ]
How many times is my function used within an executable?

[ 20100412100135 ]
vmlinux on Ubuntu

[ 20100328152951 ]
where your WIND coverage ends

[ 20100326113354 ]
Nexus One live in Ottawa on WIND Mobile

[ 20100311133411 ]
the WIND excitement

[ 20100302160743 ]
sata hotswap pico-HOWTO

[ 20100227093435 ]
serving http content out of a git repo

[ 20100225104049 ]
pimped out zsh prompt

[ 20100221184126 ]
live termcasting of your terminal over telnet

[ 20100215083725 ]
Debian on UBIFS upgrade on SheevaPlug

[ 20100214205532 ]
nexus one

[ 20100130150527 ]
skype on Debian Linux (64bit)

[ 20100125191021 ]
I am so peeved at Rogers

[ 20091205130455 ]
running really nice

[ 20091120225902 ]
notmuch for vim

[ 20091113102221 ]
squid and apt

[ 20091104195344 ]
using WIP branches to save every edit

[ 20091104194146 ]
using git workflows to avoid loosing intermediate changes

[ 20091019183503 ]
reflections on ACM Reflections

[ 20091013174248 ]
bacula rejected Hello command

[ 20091007235046 ]
pimping out git log

[ 20091002094400 ]
virtualization primer

[ 20090922110756 ]
adding an external encrypted volume under Debian

[ 20090905112529 ]
the dreaded process of rooting Rogers Dream

[ 20090830174551 ]
Cyanogen's recipe for Cupcake/Donut-like pastry

[ 20090827100232 ]
simpler android rooting

[ 20090821095427 ]
androids don't like water

[ 20090820215358 ]
prettier function tracing

[ 20090816151854 ]
Baking a cupcake

[ 20090814131559 ]
Hello Android!

[ 20090813100025 ]
sim unlocking a G1

[ 20090731185522 ]
rooting the droid

[ 20090727210307 ]
getting into android

[ 20090714094340 ]
importing an old project into git

[ 20090702113222 ]
Why pick Git?

[ 20090628131255 ]
select loop for X events

[ 20090625181315 ]
portable printf

[ 20090623225800 ]
switching to uzbl

[ 20090622223147 ]
Linux Symposium

[ 20090622214023 ]
bringing git-format-patch to bzr

[ 20090612215638 ]
nfs local caching with fscache and cachefilesd on Lenny

[ 20090610202041 ]
Scott Chacon smacks git around

[ 20090610150039 ]
how would you read a file into an array of lines

[ 20090609215141 ]
libguestfs

[ 20090609000208 ]
tiding up the PATH

[ 20090608232531 ]
only showing relevant messages in mutt by default

[ 20090608010405 ]
git-vim hacking

[ 20090530223801 ]
mark-yank-urls: fix bug allowing shell to interpret the url

[ 20090509212648 ]
wmiirc-lua v0.2.8 release

[ 20090509113125 ]
wmiirc-lua moving to github

[ 20090504101605 ]
two terminals one PWD

[ 20090501172645 ]
splitting files out of a commit

[ 20090409155905 ]
git workflow: git amend

[ 20090401112030 ]
how old are these files in git?

[ 20090330181138 ]
sles 11 on kvm

[ 20090323194942 ]
android true type font

[ 20090322203939 ]
popen with stdin, stdout, and stderr

[ 20090320214228 ]
shrinking URLs

[ 20090304004744 ]
readlater

[ 20081112150409 ]
splitting patches with git

[ 20081011081638 ]
creating busybox symlinks

[ 20081002215121 ]
wmiirc-lua v0.2.5 release

[ 20080916155113 ]
git-svn strangeness

[ 20080915112959 ]
installing git man pages quickly

[ 20080913112345 ]
wmiirc-lua v0.2.4 release

[ 20080825100454 ]
Kernel Walkthroughs - booting

[ 20080813210205 ]
Linux Kernel Booting

[ 20080719211329 ]
printable OLS/2008 schedule

[ 20080715214447 ]
wmiirc-lua updates

[ 20080713194704 ]
Git Screencast

[ 20080713143429 ]
four steps to reproducible Debian installs

[ 20080705150651 ]
USB2.0 enclosure benchmark

[ 20080703230924 ]
Linux Kernel Walkthroughs posted

[ 20080702113602 ]
Introducing the Ottawa Ruby folks to Git

[ 20080628160732 ]
Authenticating Linux against OSX LDAP directory

[ 20080627142123 ]
Canada Day Events 2008

[ 20080613162541 ]
Linux Kernel Walkthroughs

[ 20080510083828 ]
is my usb device connected to a fast port?

[ 20080509111534 ]
show more git info on zsh prompt

[ 20080430104202 ]
git-vim

[ 20080412100337 ]
color your word

[ 20080404105620 ]
show current git branch on zsh prompt (2)

[ 20080303200359 ]
how to track multiple svn branches in git

[ 20080301134220 ]
fixing X for GeodeLX

[ 20080108002540 ]
kvm nfs hang

[ 20080107160836 ]
screen -c relative path bug

[ 20080105132854 ]
WeeChat spell suggestions

[ 20071219221358 ]
show current git branch in zsh

[ 20071217141037 ]
wmiirc-lua kitchen sink repository

[ 20071212100316 ]
protecting sshd from OOM killer

[ 20071204234232 ]
wmiirc-lua v0.2.1 remembers a bit more

[ 20071013205336 ]
wmiirc-lua v0.2 has suspend and raw modes

[ 20070929112345 ]
wmiirc-lua debianization

[ 20070924104140 ]
zsh tip of the day - global aliases

[ 20070915094213 ]
wmiirc-lua v0.1.1

[ 20070913130838 ]
comparing two directories

[ 20070909204125 ]
git-rebase --interactive

[ 20070908115905 ]
zsh tab completion awesomeness

[ 20070902000736 ]
wmiirc in lua

[ 20070831150306 ]
debugging with -dbg libraries

[ 20070831142646 ]
svn status like output in git

[ 20070829141847 ]
Git Cheat Sheet

[ 20070821142038 ]
switching to abiword

[ 20070811105746 ]
forwarding ssh and X through screen

[ 20070807112531 ]
git-svnup

[ 20070724082355 ]
reducing power consumption

[ 20070722123734 ]
Makefile template

[ 20070722002649 ]
less, colourful

[ 20070719162359 ]
irssi docs

[ 20070716114553 ]
qemu eats up /dev/shm

[ 20070710214512 ]
git-clean in svn land

[ 20070705113139 ]
ipw2200 not working

[ 20070627191916 ]
git slides updated

[ 20070618220649 ]
unpopular debian packages on my system

[ 20070611125852 ]
git-svn with multiple branches

[ 20070517085321 ]
Linus on Git at Google

[ 20070510134551 ]
vim modelines insecure

[ 20070504205042 ]
bios disassembler

[ 20070504124124 ]
dd hex arguments

[ 20070503013555 ]
urxvt mouseless url yanking

[ 20070502211941 ]
gitdiff.vba v2

[ 20070419234350 ]
india

[ 20070418155857 ]
zsh fun

[ 20070418143632 ]
pipe to pastey.net

[ 20070418094151 ]
vimgrep alias

[ 20070416202545 ]
mouse-free

[ 20070406141850 ]
ATA messages via SCSI layer

[ 20070330221019 ]
GITDiff vim plugin

[ 20070329011735 ]
git presentation for OCLUG

[ 20070328123631 ]
fixing vim's [[ and ]] for bad code

[ 20070316092236 ]
pxeboot and nfsroot with debian

[ 20070312134706 ]
etc snapshots with git

[ 20070307004041 ]
remote power switch

[ 20070222215355 ]
klips-less openswan git tree

[ 20070221041316 ]
git caching for v1.5.x

[ 20070218002214 ]
klips loses zlib

[ 20070209172606 ]
vim and linux CodingStyle

[ 20070207205427 ]
my kqemu install

[ 20070204100100 ]
leaner meaner openswan

[ 20070115111917 ]
wmii+ruby xlock action

[ 20070112131252 ]
cloning xterms in wmii+ruby

[ 20070102010551 ]
wmii+ruby talk for OCLUG

[ 20061228220641 ]
dump and restore

[ 20061218100219 ]
C style

[ 20061101002027 ]
fetching all git branches from remote

[ 20061028111607 ]
local caching for git repos

[ 20061020145437 ]
automatic version creation with git

[ 20061018213306 ]
wmii w/ ruby wmiirc

[ 20061018201907 ]
small fonts

[ 20061007151802 ]
google-codesearch from vim

[ 20060928020813 ]
shell commands

[ 20060920093957 ]
letting screen apps use the ssh-agent

[ 20060908223613 ]
mpdscribble stream support

[ 20060907125149 ]
glGo on ubuntu/dapper amd64

[ 20060906163240 ]
lbdb and mutt

[ 20060902145643 ]
vim tutorial

[ 20060902135722 ]
fixing your terminal

[ 20060828124713 ]
apt-get pdiffs

[ 20060824224842 ]
256 colour xterms

[ 20060824152658 ]
dynamic IPcomp

[ 20060824145428 ]
inkscape++

[ 20060818150516 ]
tags/cscope for system headers

[ 20060805131557 ]
opteron 170, part 4

[ 20060805101941 ]
opteron 170, part 3

[ 20060803233234 ]
opteron 170, part 2

[ 20060802210126 ]
opteron 170

[ 20060729144129 ]
OLS keysigning / 2006

[ 20060728105500 ]
git-find findings

[ 20060727162941 ]
starting on git-find

[ 20060727113632 ]
git-graft and git-find brainstorm

[ 20060726224531 ]
pretty function tracing

[ 20060713174723 ]
uml and multiple network segments

[ 20060707182236 ]
lastfm artist and title to clipboard

[ 20060706162256 ]
reverting a git changeset

[ 20060629212003 ]
user #3

[ 20060628083456 ]
firefox crashes with form input

[ 20060621151402 ]
git vs hg

[ 20060612222204 ]
ldap account management

[ 20060612194523 ]
stupid ldap

[ 20060608092157 ]
rpm hell is right

[ 20060605095726 ]
OpenSSH VPNs

[ 20060604114317 ]
Lenovo lost a customer

[ 20060601234010 ]
generating html colourized sourcecode

[ 20060601211716 ]
ion3 greatness and acting on X selections

[ 20060526085644 ]
software RAID10 performance

[ 20060525234148 ]
learning to love git

[ 20060524121638 ]
recent vim7 articles

[ 20060516095748 ]
bootstrapping debian on my sbc

[ 20060428145140 ]
entropy injection

[ 20060423140628 ]
Adam

[ 20060414202507 ]
converting mp3s to CD

[ 20060413232836 ]
secure apt-get

[ 20060412194423 ]
xen domain0 on debian

[ 20060410220525 ]
LVM2 on RAID1 mirror

[ 20060410102824 ]
building a RHEL4 kernel w/ kdb support

[ 20060407230939 ]
xen on debian

[ 20060407230818 ]
Upgraded look

[ 20060330131334 ]
Flattered by a copy

[ 20060328165153 ]
vim7 from source

[ 20060308123539 ]
Perl, Catalyst, CPAN, and Debian

[ 20060308123302 ]
last.fm

[ 20060128124841 ]
carcassonne and zombies

[ 20060120135931 ]
Election Humour

[ 20050925130002 ]
ldap on debian

[ 20050528190034 ]
Error while mapping shared library sections

[ 20050124130158 ]
IRC over email gateway

[ 20050110225522 ]
brute force attacks sshd?

[ 20050108095026 ]
LDAP authentication (part 1)

[ 20041124130146 ]
sawfish workspace themes

[ 20041113082651 ]
Mini-DV to divx using mencoder

[ 20041004084525 ]
notes on vserver

[ 20040922104334 ]
fast kernel logging

[ 20040803104122 ]
Digital Rebel for sale... GONE

[ 20040714202912 ]
OpenOffice resources

[ 20040603175746 ]
photo editing

[ 20040601082817 ]
WRT54G

[ 20040503205227 ]
Digital Rebel

[ 20040420200136 ]
Open Office templates

[ 20040326082602 ]
bash vi editing mode

[ 20040315204142 ]
debian install CDs

[ 20040312155542 ]
change of jobs

[ 20040308091554 ]
spamassasin extras

[ 20040305163216 ]
cool debian tools

[ 20040305155708 ]
first post!





Bart Trojanowski
http://www.jukie.net/~bart
bart@jukie.net