Cloning/Fixing Failed HDD

I would have given up if this much work was needed to repair one of my discs, after all it's only telly and I can stream on catchup most things I recorded recently anyway. Fortunately fixdisk has always worked for me.
An interesting admission! Having previously been gainfully employed by the BBC and Channel 4 the expression "It's only television!" is one I'm familiar with. Now everything's working again I'm looking forward to the second half of last year's Crucible snooker and various TPTV series I'd gotten behind with. I'll try not to record quite so much this year, I think.
 
An interesting admission! Having previously been gainfully employed by the BBC and Channel 4 the expression "It's only television!" is one I'm familiar with. Now everything's working again I'm looking forward to the second half of last year's Crucible snooker and various TPTV series I'd gotten behind with. I'll try not to record quite so much this year, I think.
I should add I run two HDR Fox T2s at my house and set everything to record on both boxes. One uses AR and the other padding, this gives me redundancy of which recording method works for that programme. I mostly watch HD channels so being major channels AR usually works. But this setup also gives me redundancy against a disc failure.
 
Would this HD be OK for a HDR FOX T2, if using dd to clone the original HD (on another machine)?
Perhaps not. I was guided by MS Copilot's advice and bought two Seagate Skyhawk ST1000VX013 from Amazon. Our cloned, patched and otherwise fixed HDD has been in service for two weeks now without issue.

Quick Answer: The best modern replacement drives for a Humax HDR‑Fox T2 are surveillance‑grade HDDs such as the Western Digital Purple or Seagate Skyhawk. Both are designed for continuous AV workloads, run cool and quiet, and handle the constant read/write cycles of a PVR far better than standard desktop drives.

🔍 Why Surveillance Drives?​

· Optimized for 24/7 recording — Humax units write constantly when buffering live TV.
· Low vibration & noise — important in a living room environment.
· Error recovery tuned for streaming — avoids long retries that cause playback stutter.
· Better longevity — higher workload ratings than consumer desktop drives.

📊 Recommended Models (UK Availability)​

Drive​
Capacity​
RPM​
Cache​
Workload Rating​
Price (UK)​
Notes​
WD Purple WD10PURZ1TB540064MB180 TB/yr~£68.63 (Amazon UK)Excellent entry‑level replacement, quiet and reliable.
WD Purple WD20PURZ2TB540064MB180 TB/yr~£90–£100Popular upgrade size; Humax supports up to 2TB.
Seagate Skyhawk ST1000VX0051TB590064MB180 TB/yr~£65–£75Designed for DVR/NVR systems, similar to WD Purple.
Seagate Skyhawk ST2000VX0152TB590064MB180 TB/yr~£95–£110Slightly faster spindle speed, good for multi‑stream recording.
 
OK many thanks for that. I am seeing fixdisk report a lot of stuff but is apparently successful fixing it

Please select option: fixdisk
Any additional options (-h for list or press return for none): -y
Are you sure you wish to run the hard disk checker (-y)? [Y/N] y
Running /bin/fix-disk

Checking disk sda (512 byte sectors)

Unmounted /dev/sda1
Unmounted /dev/sda2
Unmounted /dev/sda3


Running short disk self test


Pending sector error(s) found

LBA has not yet been found
A long test is required - this could take 3 hour(s) 33 minutes
Do you wish to continue? [Y/N]: y
Running long disk self test
Error at LBA 1430982485

/dev/sda:
re-writing sector 1430982485: succeeded

Running select disk self test
Error at LBA 1430995055

/dev/sda:
re-writing sector 1430995055: succeeded

Running select disk self test
Error at LBA 1430996810

/dev/sda:
re-writing sector 1430996810: succeeded

etc

then after a few hours

/dev/sda:
re-writing sector 1915909217: succeeded

Running select disk self test


Checking partition tables...

MBR Status: MBR only
GPT Status: not present

Using superblock 0 on sda1
Using superblock 0 on sda2
Using superblock 0 on sda3
Dev: /dev/sda LBA: 1430982485
LBA: 1430982485 is on partition /dev/sda2, start: 2104515, bad sector offset: 1428877970
dumpe2fs 1.42.13 (17-May-2015)
Using superblock 0
Block size: 4096
LBA 1430982485 maps to file system block 178609746 on /dev/sda2

Checking to see if this block is in use...
debugfs 1.42.13 (17-May-2015)
 
Last edited:
I am seeing fixdisk report a lot of stuff but is apparently successful fixing it
Others are really in a better position to comment on this but - in case it's helpful - here's my experience of running fixdisk on my failing original HDD whose LBA 0 was irrecoverable:

After "re-writing sector 0: succeeded" fixdisk repeatedly ran the short disk self test, identifying errors at various blocks and asking if I wished to attempt repair of the bad block, to which I said yes and the re-writing succeeded. In order, the sectors were 558, 2174623, 2174610, 2174623 (again), 2174621, 2174621 (again), 2174617. When it again flagged an error at 2174621 I declined the repair attempt. It then checked the partition tables (MBR Status/GPT Status not present), then searched for partitions finding partitions 1, 2 and 3. It then asked if I'd like to attempt repair of the partition table, to which I said yes. It then re-read the partition table and said "New partition table has been created". It then reported further details before saying partition sda1 was an unrecognised partition type and fixdisk aborted terminating the session with exit status 1.

On the clone I had to dig deeper and use fsck on the individual partitions via the CLI option in CF Maintenance Mode. After this work I opted to create a new sda1* by formatting a spare HDD then copied that, sector-for-sector to the clone, then I had to edit the LBA 0 partition table to match reality. Result: no further demand to format the HDD from the Humax and job done.

*I'm not sure whether sda1 should have been recognised.
 
One wonders why LBA0 in particular. I know Humax doesn't run winXP :) but there is a well known issue with winXP and SSDs, which tend to die after about a year of 24/7 operation, with "cannot find NTLDR" and where does NTLDR reside? At the start of the disk. AFAIK winXP does write to the start of the disk at 1Hz or so. And I wonder if the Humax does the same, due to some other running process? That is why a normal HD (not SSD) is better for these systems.

That very last bit of fixdisk
Checking to see if this block is in use...
debugfs 1.42.13 (17-May-2015)

is taking a very long time...
 
Last edited:
I think it's got stuck... I will leave it until tomorrow morning and then turn it off/on.

Hmmm... spoke too soon. It is still checking more blocks if they are in use.
 
Last edited:
I think it's got stuck...
It hasn't. Just leave it.
I will leave it until tomorrow morning and then turn it off/on.
Turning it off/on just means you have to start again from the beginning. I guess you can work out how that's not going to make it finish any faster.
 
Last edited:
It did finish by the morning.

The new firmware is amazing. Much nicer than what I was running years ago. Decryption is neat too. I do hardware/software development too (arm32, asm+C) and I can see somebody has done a huge amount of work on this project! The way Humax derived the key from the S/N and the MAC is hilarious. Actually I have seen products which auto-generate their MAC with a hash of the (unique) CPU on-chip S/N ;)
 
The new firmware is amazing. Much nicer than what I was running years ago.
The firmware is the same as original (with just a few tweaks. It's the additional software which makes the difference (the firmware tweaks enable the software to launch, plus exposing a few hooks into the hardware operation).
 
Has anyone managed to mod the code in the FLASH? AIUI the custom code is disk-resident and co-exists with the FLASH based stuff.
 
Has anyone managed to mod the code in the FLASH? AIUI the custom code is disk-resident and co-exists with the FLASH based stuff.
That's the distinction (or hair that is being split) here between Custom Firmware (in flash) and the packages added to it. Even some of the packages live in flash, see the fix-flash-packages diagnostic. Also the linux kernel (in flash) has been built from updated source code, which is how it supports GPT hard discs and much larger sizes than Humax supported. The one thing that isn't rebuilt is the humaxtv application which runs the UI, records and plays back things etc.

A LOT of work has been done. Less in recent years as the main author has had to step back for health reasons.
 
Amazing work! I have been involved in something vaguely similar: a system written in Ruby on Rails, which one cannot get anyone to maintain today (under £2k/day) which had to be functionally expanded with code written in PHP and which accesses the database. Then a lot of stuff was added with cron jobs etc. With that we have the Ruby source of course, whereas you did not have the source. I gathered it was linux based originally (most "consumer IT" boxes are). I guess you must have discovered a way to log in (putty etc) and see what is running where.
 
Presumably the box is a linux board with a filesystem in FLASH, and the code resides in there, otherwise you would not be able to replace the HD (unless it had a 2nd SATA controller and a "dd" feature).

I had this recently with a Lenovo mini PC where I had to clone an SSD but it had just one SATA. I had to use an old PC to clone it. I would do the same to replace the HD in the Humax, because I don't want to lose the recordings. Clonezilla is pretty good but it would not clone this SSD (win11 machine, weird GPT partitions etc). But dd worked.
 
whereas you did not have the source.
Actually for the linux kernel we do have the source, Humax were required to publish it due to open source licencing. It wasn't quite right they missed out a couple of security fixes, but af123 the author of much of this stuff figured it out and added that back in plus some other kernel security fixes.

What we don't have the source for is the humaxtv app which runs the UI and recordings and playback etc.
 
What is the hardware? There must be two tuners so two RF / downsample / ADC channels. Freeview is not encrypted and they probably encrypt in software (because CSS is simple) on the way to the HD. What I wonder is, in addition to a standard 80x86 (or ARM?) motherboard, what else is there?
 
What is the hardware? There must be two tuners so two RF / downsample / ADC channels. Freeview is not encrypted and they probably encrypt in software (because CSS is simple) on the way to the HD. What I wonder is, in addition to a standard 80x86 (or ARM?) motherboard, what else is there?
It's a MIPS processor. I don't know other details except yes there are two tuners.
 
What we don't have the source for is the humaxtv app which runs the UI and recordings and playback etc.
I would have thought that with tools like Ghidra around these days that wouldn't be a big obstacle. It is reputed to be able to generate pretty good high level language source code by examining the program structure, although I haven't tried.
 
Back
Top