Hi, I am new to the forum, but I have used MHDD several times.

I have this hard drive, Seagate ST900325AS, that has a lot of bad sectors, and I want to recover the data. The problem is that the drive locks up (does not respond to commands) if I ask it to read a certain sector. If I scan the drive with MHDD then all sectors after that are "ABRT". If I then power cycle the drive, it starts responding and I can scan the other sectors.

I found one such sector and erased it, now the scan continues until it hits another such sector. The problem is that then the log is full of ABRT errors (but there is not way to identify them as opposed to the UNC errors), I then have to scan multiple sectors one by one until I hit the very bad one (that causes the drive to lock up).

So, is there a way to make MHDD stop when it encounters the ABRT error? It would help me to identify the very bad sectors (one prior to the first ABRT error)? or at least write the log file with a distinction between UNC and ABRT errors?

The regular bad sectors form some weird pattern, it would be interesting to see if the very bad ones form a pattern too

MHDD is useful for diagnostic and not so much for "fixing" bad sectors and recovering data.

The drive has multiple issues and recommend that you stop tampering with it, because it will lead to killing it completely.

As your good observation indicates, the key is the power cycle status.

Best to seek professional service as only a professional advanced DR tool can handle the job.

_________________
Hard Disk Drive, SSD, USB Drive and RAID Data Recovery Specialist in Massachusetts

No money for the professional, so I might as well try it.

So, I managed to find the description of the binary log format and wrote a program that outputs the first address or an ABRT and the previous address. Now I only have to run MAKELOG, wait until the drive locks up, exit, run my program and get two addresses. The very bad sector is between those addresses, so just a few short scans narrow it down to one sector. Erase, makelog, repeat.

After I find all of them, I'll use ddrescue or similar to recover the data.

Not exactly the best plan to resolve this issue, but worth to try if can't afford pro service. Just remember ddrescue or dd_rescue have their own limitation.

Just a warning, this symptomp is more likely failing head(s) and (also) defect list problem in its firmware, not all sectors with "ABRT" error can be "erased" or "remapped" in many cases.

So far, there have not been any sectors with ABRT. That is, after trying to read some sector (and failing with UNC), the drive locks up and then every next sector fails with ABRT. However, if I power cycle the drive, I can read the other sectors as long as I do not step on that landmine.

Example:

Drive fails sectors 1'025, 3'060 ... with UNC but continues to work. After failing sector 6'358'935 with UNC the drive stops responding and fails all next sectors with ABRT. If I power cycle the drive and begin the scan at sector 6'358'936, the drive reads it and the other sectors normally (with some UNCs). If I power cycle the drive and try to read sector 6'358'935, the drive fails ith UNC and stops responding (which means that the problem is with that particular sector, not time after power cycle or number of sectors read). Now, if I erase the sector 6'358'935 then I can scan sectors 0 - 6'635'900 and so on without the drive locking up. This means that the defect relocation probably works fine (or writing the sector solved the problem in place).

Yes, I know. You erased that "non responsive" sectors so you can scan further and do cloning afterwards. Just remember an HDD defect reallocation table(s) has its own limit. If this limit is reached and there are no more spare sectors left in SA, the drive will stop functioning. That's why as labtech explain above, MHDD is not a good idea when dealing with this problem.

I have to mention also head(s) could easily fail during your procedure

I've run into this as well... You should be worried about trying to read near known clusters of bad sectors because it's not always just magnetic corruption. People have killed heads and platters by doing this! If you must read the whole drive out including the nearby sectors, then it's best to try the error-free areas first. That way if it doesn't come back after a power cycle, you at least have those. It also means that you'll feel a bit safer trying to narrow down the good/bad 'gaps' since you've already got most of the drive. But if your data is valuable this is a kind of "if it's broke, don't break it even more!" situation. Also, many programs have a 'reverse' option to read from the end first.

If you had the money, I'd just say get a professional to do it. Obviously though, you're in the situation where 'value of data' < 'price of recovery'.