Direct extraction from a tivo drive in a pc (not networked)

[jt1134]

Will direct extraction work with a TivoHD? (THIS unit)

yes. you'll need to fetch the source code for the 64bit mfs_* utils and compile it for x86. however, there's not much point in doing so. see my response in the other thread. cross-posting is frowned upon.

[paleGreen1]

Our faithful Sony SAT-T60 now boots up to the Green Screen of Death, after 7+ years of faithful service. I upgraded it to 2 160 GB drives myself years ago (kept the standard kernel), then later added a CacheCard, all of which were working fine until the GSOD starting appearing. I don't want to invest in new hard drives for it, since I can get a new HD DVR for the same price. But I would like to salvage any shows I can from it.

I'm following directions here and here but I'm getting stuck by a "Not a TiVo super block!" message when I run mfs_streams or mfs_info. (tivopart seemed to work: several partitions appeared on the KNOPPIX desktop for hdc and hdd.) I tried the 20050604 image of mfs-utils first, then the 20060503 image.

I'm pretty sure I disabled encryption the last time the software was auto-upgraded (after which I disconnected the phone line so I wouldn't have to keep re-hacking it) but I can't be positive (it was probably a couple years ago). I read about ciphercheck, but that appears to be for S2 units.

Am I doing something wrong? How likely is it that the system just too messed up to recover anything? Anything else I should try?

Thank you!

[Jamie]

...
Am I doing something wrong? How likely is it that the system just too messed up to recover anything? Anything else I should try?

Since this is a series 1, this could be a byte swap issue, although it is also possible the superblock of the MFS is damaged. What is the "magic number" printed in the "Not a TiVo superblock (magic=xxxx)" message?

[paleGreen1]

What is the "magic number" printed in the "Not a TiVo superblock (magic=xxxx)" message?

Thanks for the speedy reply Jamie! It's magic=0x31343735.

Here's my steps:

root@Knoppix:~# export PATH=$PATH:/mnt/hda1/mfs-utils/
root@Knoppix:~# /mnt/hda1/tivopart/tivopart.x86 r /dev/hdc

(hdc1 through hdc16 appear on the desktop)

root@Knoppix:~# /mnt/hda1/tivopart/tivopart.x86 r /dev/hdd

(hdd1 through hdd4 appear on the desktop)

root@Knoppix:~# export MFS_DEVICE="/dev/hdc /dev/hdd"
root@Knoppix:~# mfs_streams
Not a TiVo super block! (magic=0x31343735)

Thanks!

[Jamie]

If you look in mfs.c before that error message, you'll see a set of cases that cover the various expected "magic" values. Since you pulled the drive during a hung GSOD, I'm pretty sure you have an inconsistent file system (fs_inconsistent = 1). Your PC host, I assume, is an Intel PC, so you should have little_endian = 1. Finally, it's an S1, so unless you have the kernel patch to byte swap the IDE bus, you need io_need_bswap(1). So included your magic number along with the case for 0x30333735. If you are lucky, that's all you need to get it to work.

If the magic number change doesn't help, dump the first 512 bytes of partition 10 on the "a" disk with hexdump and post it.

I don't have an S1, and I'm not totally sure the byte swap code is going to do the right thing in your case (e.g. partitions validated with tivopart). Let's see where we are after you make the change suggested above.

[paleGreen1]

Wow. That is very cool -- and very scary since I haven't compiled a C program since college in the early 90's. I'm guessing things have gotten a little easier since then, but I remember pulling my hair out trying to get things to just 'make' successfully. So this will probably be an adventure...

I don't want to ask too many noob questions, but what do I need to do to compile this? I read here to just:

make proto
make i386

Will that do it?

[Jamie]

It's like riding a bike....

I think just "make" should do it. The Makefile will automatically build the proto.h file, and it should default to building for the host (PC) architecture.

[paleGreen1]

I guess things have gotten easier! ;-)

I added the case statement for my magic number with the settings you recommended and did the make. I ended up with a bin.Linux-i586 folder. I re-ran the steps to run mfs_streams and got the following message:

Warning: filesystem is inconsistent. Run fsfix and mfscheck ASAP
Failed to byte swap correctly

Does this confirm your hunch that the byte swap code isn't going to help in my case?

I think I figured out how to do the hexdump all by myself!

root@Knoppix:~# hexdump /dev/hdc10 -n 512
0000000 0000 0000 3735 3134 7d12 fed5 0000 0010
0000010 0000 0001 0000 0040 0000 0400 0008 0000
0000020 7fee 08f0 2f64 6576 2f68 6461 3130 202f
0000030 6465 762f 6864 6131 3120 2f64 6576 2f68
0000040 6461 3132 202f 6465 762f 6864 6131 3320
0000050 2f64 6576 2f68 6461 3134 202f 6465 762f
0000060 6864 6131 3520 2f64 6576 2f68 6462 3220
0000070 2f64 6576 2f68 6462 3300 0000 0000 0000
0000080 0000 0000 0000 0000 0000 0000 0000 0000
*
00000a0 0000 0000 1be3 dc00 0007 a120 0000 0001
00000b0 0000 03e8 0422 0e5d 0000 03e9 0000 0078
00000c0 0000 0000 0000 0461 000f fffe 0000 0001
00000d0 0008 0000 0008 0000 0083 7305 0000 00b0
00000e0 0085 1b82 0000 0080 0000 0000 0000 0000
00000f0 0000 0000 0000 0000 0000 0000 0000 0000
*
0000200

Let me know if I messed something up.

Thank you very much!

[Jamie]

I guess things have gotten easier! ;-)

I added the case statement for my magic number with the settings you recommended and did the make. I ended up with a bin.Linux-i586 folder. I re-ran the steps to run mfs_streams and got the following message:

Warning: filesystem is inconsistent. Run fsfix and mfscheck ASAP
Failed to byte swap correctly

Does this confirm your hunch that the byte swap code isn't going to help in my case?
...

No. There's just one more place the code needed to be changed to recognize this "magic" number. Try the attached mfs.c (also committed to cvs).

[paleGreen1]

I replaced my mfs.c with the new one, but I got some errors during the make:

...
gcc -Wall -I. -I/sw/include -I/sw/include/gnugetopt -I../include -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -DNEED_ALLOCA_H -O3 -DBUILD_DATE=\"`date +%Y/%m/%d`\" -c mfs_info.c -o obj.Linux-i586/mfs_info.o
mfs_info.c: In function 'stream_scan2':
mfs_info.c:251: warning: passing argument 1 of 'mfs_all_inodes' from incompatible pointer type
mfs_info.c:251: error: too few arguments to function 'mfs_all_inodes'
make: *** [obj.Linux-i586/mfs_info.o] Error 1

Comparing the new version of mfs.c with mine revealed several differences beyond the magic number code. Looking at the repository on SourceForge, it looks like a few files have changed since mfs-utils_src-20060503.tar.bz2 was made on May 3 2006:

partition.c           1.6       12 hours      jamiepainter       Support additional "fs_inconsistent" magic numbers.
mfs.c                 1.9       12 hours      jamiepainter       Support additional "fs_inconsistent" magic numbers.
schema-merged-9.3.txt 1.2       16 months     jamiepainter       check in to base
schema-7.3.1.txt      1.1       2 years       jamiepainter       Update scheme to match Makefile.
CHANGES               1.6       2 years       jamiepainter       Darwin-i386 support and misc bug fixes.
mfs_info.c            1.4       2 years       jamiepainter       Darwin-i386 support and misc bug fixes.
io.c                  1.7       2 years       jamiepainter       Darwin-i386 support and misc bug fixes.
schema.c              1.3       2 years       jamiepainter       Darwin-i386 support and misc bug fixes.
tserver.c             1.5       2 years       jamiepainter       Darwin-i386 support and misc bug fixes.

I'm guessing I'll need to download these and try the make again. Is that right?

Thanks!

[Jamie]

Sorry about that. I'm working from the latest version checked in to cvs. You should probably do a CVS checkout of the whole package. There are instructions for that here.

[paleGreen1]

I finally got around to trying this again. I don't have a CVS client up and running so I downloaded everything individually from here.

This time when I did a make I got some warnings but it did complete (something about parameters expecting 'int' but receiving 'u64' or something like that). But when I ran mfs_streams this time (following the same steps as before) I got:

Warning: filesystem is inconsistent. Run fsfix and mfscheck ASAP
crc mismatch len=280 0x7d12fed5 0xe8f7cde2
WARNING: total sectors doesn't match (total=467917824 sb=0)
Segmentation fault

Anything else left to try? I really appreciate your help on this.

[Jamie]

Let's take this to PM. I sent you a PM several days ago.

[marcolisi]

Let's take this to PM. I sent you a PM several days ago.

I need help pls.
I am new at tivo. I downloaded son shows , including italian soccer off the italian channel I have (i am italian) but I can not reverse them on my pc using tivodesktop. :-(
I would like to open the case of tivo, take the hd out and connect it to my pc to get these show out pls.
Can u pls pm me ?

thank u
marco

[stevel]

You won't be able to do that if you didn't disable encryption. You will need to use TiVo Desktop.