CD Puppy2.12 no boot in "SCSI" drives (i875/ICH5) > resolved

[JohnRoberts]

Hi!!!

I am a newb with Puppy. I burned the 2.12 ISO (MD5 verified...) on a CD-R. It refuses to boot on my PC. I have tested it on 3 other PCs and it works fine!! On my PC which is homebrew with following data:

Albatron PX875P Pro V2.0 (i875P/ICH5) + P4/Celeron (not Celeron D, hence no Prescott core...) 2.6GHz/FSB400 + 4x512MB DDR333 Transcend DIMMs (dual channel)+ Albatron NVidia GeForce FX5200 AGPx8 VGA + Seagate ST3200822AS 200GB SATA HD + SAMSUNG SP0411N 40GB PATA HD mounted via a PCI UATA controller (SiL0680 Medley based...) + Teac CD-W552G CD-R/W + NEC ND-3550A DVD-R/W + std floppy + std PS/2 keyboard + Logitech compatible PS/2 trackball on a Philips 170T4 monitor.

CD-R begins to boot normally (default boot...), then reaches:
"Now executing 'init' script in initial-ramdisk"
"Loading kernel modules... done"
"Looking for Puppy in done"
"Looking for pup_save.3fs in hda1... sda1... scd0... scd1... sda1 done"

and then crashes with error:

"ERROR cannot find Puppy on 'idecd' boot media.
PUPMODE=1 PDEV=
Exited to initial-ramdisk (/dev/ram0)"
and then I just get the prompt ("#") (with message "...guru can debug...", etc.).

The same CD-R I tried (and worked perfectly) on 3 other PCs ranging from a 3 year old P4/Northwood on a Chaintech i845PE based mobo with 2GB of RAM to a VIA KM266/ CeleronD with 512MB RAM and even a fresh Athlon64 on an NForce4 mobo with 2GB of RAM...
Any clues about this???

P.S. The same machine that Puppy will not boot, I tried with SUSE 10.0 LiveDVD and boots without any problem...Oh, also on both HDs filesystem is NTFS. Native OS is Win2k/SP4


Many Thanks in advance

[caesirian]

maybe puppy is not compatible with ur ram?¿

i dont know, but when u boot puppy from the cd it runs totally in ram, u couuld run puppy without hard disc... so.. we should start analizing the problem from there...


good luck

[rcrsn51]

Try copying the pup_212.sfs file from the CD to the root of your hda1 partition and rebooting.

[Gn2]

Puppy is having Probs locating your drives:
BTW - are Sata & Pata hard drives on an "add-in" controller ?

What does

Code: Select all

 fdisk -l

> show Re locations & drive letters (hd- ???)
Which optical was used to boot - for ease of detection:
cable first IDE contoller for hard drives, second for opticals -
In that manner, flawed probe scripts are aided in probes.

(Dev/hdc would be 2nd controller, primary)
Normally (to Linux) that would not matter - but all depends on how any boot or install script probes, ID's locations.

When it exits to CLI prompt - use output of fdisk to tell Puppy
full path to CD disc containing the sfs files

HTH

[JohnRoberts]

RAM incompatibility?? That's a strange one, I say... The Chaintech, i845PE-based mobo PC I tried (and worked...) has also Transcend DDR333 DIMMs. It sounds a bit strange to be something to do with RAM. Analyzing the data I have (and NOT being an expert...), the interesting part is the "...cannot find Puppy on 'idecd' boot media..." message. Since - apparently - Puppy IS INDEED on the CD-R, there must be something wrong with the device handling/enumeration. Since Puppy exists, the initialization procedure searches in the wrong place...Well, I don't know...I need definitely more feedback on the way CD Puppy boots, i.e. when it reaches the point of the crash event what exactly has taken place concerning boot-up procedure? Are there any logfiles to look for?

[JohnRoberts]

I was thinking along similar lines Gn2 (and we were writing simultaneously it seems ). I need some time to check...I am writing from the target PC...will revert ASAP. SATA HD is connected directly to mobo. PATA is connected via the UATA-133 controller. Primary HD is SATA. If I remember correctly (let me peek through the window...) CD-R/W and DVD-R/W share same IDE channel, CD-R/W is master, DVD-R/W slave. I don't remember if it is IDE1 or IDE2 but anyway the other IDE is unusable because i875 (this is an Intel trick...) emulates SATA as being "the other" IDE channel. That's why to use another PATA drive, I could not connect directly, and had to resort to the PCI connection.

[caesirian]

do u have puppy installed or this is the first time ur trying to boot it in that machine??

[caesirian]

now i remember i had some troubles in the past booting different live-cds
try puting ur cd drive as a slave of ur hard drive (if u can) if not leave the drive cd alone without the dvd

or try different combination til it work

[JohnRoberts]

This is no HD installation. It is Live-CD only...
Tried the fdisk -l and found some interesting results
It lists my 2 HDs as following:

My 40GB PATA HD which is normally my secondary disk is listed as dev/hda1
My primary 200GB SATA disk is dev/sda1

Should I see in fdisk -l the optical drives as well or not?

To guestimate, scd0 and scd1 would be the optical drives (or at least were the script supposes that they are...). How can I check this?

It wouldn't make much sense in recabling. SATA has only one place to go. This leaves effectively only one IDE channel and leaves 2 opticals + PATA HD hence UATA PCI controller needed. Since PATA HD is only secondary, I keep the native available IDE channel for the opticals. I think it would be rather error prone to mount any of the opticals on the PCI controller...I am also not favoring mounting HDs and opticals sharing common IDE channels. Don't know if anyone had stable systems using this set-up, I usually prefer NOT sharing channels between opticals and HDs.

At the point that the script crashes it is looking for pup_212.sfs on the Live-CD, am I right?

[GuestToo]

you can type:

fdisk -l /dev/scd0 /dev/scd1

probedisk is the program that would probably be used to detect your drives ... just type probedisk

i don't think Puppy has support for scsi drives yet ... i don't think Puppy can boot from a scsi cd drive

[JohnRoberts]

/dev/scd0 and /dev/scd1 means that the opticals are regarded as SCSI??? (very weird!!!...will check)

[Gn2]

Correct - Puppy initrd image cannot locate the optical holding your booted CD disc - & only the hard drives are listed via fdisk

Code: Select all

Dmesg|less

will give boot probe results
(That odd lookng vertical seperator in above - "pipes" output so Msg may be paused & read.

Or > (lscpci) may show (puppy includes some other tool never noted before) uhid or something ?

"Re-cabling" > nor altering BIOS boot sequence > should never be needed !

Please try dmesg output to ID your opticals,
"point" Puppy to device used to boot the liveCD -

Mount manually if needed

Code: Select all

 mkdir /hd_xx
  mount -t iso9660 /dev/hd_xx  /your_created_mount_point

BTW - HOW did my post get placed a few down - when first viewed - no other had yet replied ?

More than the B/B is acting up !

[GuestToo]

well, Puppy might think they are scsi drives, or at least that they need scsi emulation, which might be about the same thing ... on a typical pc with ide drives, the hard drive would probably be /dev/hda and the optical drive might be /dev/hdc

i don't know if you can type "disktype /dev/scd0" if you get a command prompt when Puppy crashes, it might give you some information

[JohnRoberts]

Yes!!! Mystery is just beginning to unravel...Checked with probedisk, test-eide and test-scsi and for some reason Puppy thinks CD-R/W and DVD-R/W are SCSI devices!!! At this point I need some serious help to fix this...

[JohnRoberts]

The thread sequence changes faster than I can type Tks Gn2 will try ASAP. They are indeed /dev/scd0 and /dev/scd1...

[Gn2]

Scsi emulation - should only be needed for interface burning -(2.4_xx kernels ) or USB cold/hot-plugging and file system "character" devices
(vs "true" block devices)

Linux addresses all in one or alternate modes - all are "seen" as "devices" !
One is associated under true block device sub-systems - others may be "virtual" read/write F/S types.

=(tmpfs - /dev/shm etal ) > all may be mounted so Kernel may access.
You are correct- to not "share" an IDE controller w/hard drives and opticals - UDMA speeds will automatically revert to use slowest device capabilities !

IOW if a H/drive has I.E UDMA133 speed (no IDE optical may match > hence the advantage of true Scsi devices) -
faster I/O reads/writes of hard drive are nullified.

Hdparm [man hdparm) "tweaks" may help - but never match
(Above is DMA drive capabilities as supplied by OEM & reported to kernel via hdparm options. )
.
HTH

[JohnRoberts]

Results are:

> Dmesg doesn't work
(I cannot seem to find it...)

>Opticals are definitely /dev/scd0 and /dev/scd1

>"fdisk -l /dev/scd0" returns a 87MB partition (obviously corresponding to the CD-R...)

>"mount" is looking for /etc/fstab which isn't there...(no luck there...)

>I think I found the script...It must be /sbin/init. If this holds it is rather long and I will have a careful look at to locate the section in question...(which makes sense if it is looking for pup_212.sfs to be in 'idecd' and not '/dev/scd0')

>I must be heading for work soon (local time is 05:50), I will revert ASAP

Meanwhile Gn2 could you clarify usage of mount??
Does "mount -t iso9660 /dev/scd0" make sense??

[Gn2]

J-R

Yes that should work - but remember to first create a (optional) "mount point"
Your sample - tells Linux to mount - but not where yet and fstab has not been created ( bootup incomplete)
> as did the "missing" (my bad) "dmesg" !

Sorry for mis-direction.

[JohnRoberts]

I failed to mention there is one more message appearing after the error on exit to #:

• /bin/sh: can't access tty; job control turned off

Is this just a side effect from script exiting abnormally?

On mounting the CD-R, I tried this on #:

Code: Select all

 mkdir /mnt/idecd 

followed by:

Code: Select all

 mount -t iso9660 /dev/scd0 /mnt/idecd 

and mount returns

• mount: /dev/scd0 is write protected mounting read-only

At this point I can normally access the CD-R filesystem through /mnt/idecd.

I tried rerunning the '/sbin/init' script manually but without success. It aborts with the same error msg.

Where to next? How do I complete the booting procedure?

[Gn2]

http://www.murga-linux.com/puppy/viewto ... hlight=sbm

http://btmgr.sourceforge.net/download.html

Should work - best wishes.