Hi,

I am just trying out mhdd and so far it's really nice ( invaluable ).

However I would like to have some similar functionality under linux, and thus have decided to give it a simple go at writing such a program.
I looked at the forum and it seems the source is out of the question .... but basicly what is needed is direct access to the disc controller ?

Anyhow, I'm of trying to get some direct access and to keep linux out of it ....

I would really like to ask if its ok to ask some questions ?

I have done devicedrivers, and system programming before, but not IDE programming ... and there are some issues like power on/off that I would like to implement ...

But I would like some feedback
Also if I get 'raw' access to the controller ... what would theoretically be needed to get mhdd to run on it ?

It is very interessant, I think it is possible to send the command ATA directly to HDD with ioctl, or better solution make a kernel's module for send command ATA.
I would like use the functions of these tools: HDD Regenerator, SpinRite, MHDD and Victoria on linux.
http://smartmontools.sourceforge.net/ is a good start for study.

_Now_ there is MHDD for Linux. WHDD.
https://github.com/krieger-od/whdd

Gentoo Linux users have sys-apps/whdd ebuilds in "maksbotan" overlay.
http://git.overlays.gentoo.org/gitweb/? ... dd;hb=HEAD

There're no packages for other linux distros yet. Help in package maintenance (and testing) is highly appreciated.

Thanks - WHDD seems like a nice lookalike for a little of the MHDD functionality under Linux, but to be clear, it is not a workalike for MHDD.

Can you explain how the "zerofill" functionality of WHDD is different to something like:


except for the extra timing info from WHDD, and the thread having raised priority (which could also be done with "nice" when running dd)?

Thank you for trying it out!
Your point is valid. The app is mostly User-Interface thing, visualizing timing of disk access.

WHDD has now got to central portage tree as sys-apps/whdd

But that's not fair to say it is not "workalike". Sure, functionality is limited at the moment - "software is never finished" (C). If you want to see some function, please request it. If you meant you found app design flaw, that prevents it from doing as much as MHDD, then please warn me about it straightly.

Thanks for explaining the "aim" of the software.


I politely disagree. I meant exactly what I said - that is not a criticism of what you have done.


Very true! I know this from my own development work too.


For example, WHDD is not using direct ATA register access - it is using OS read() and write() calls. Therefore it does not report the values in those ATA registers (UNC, ABRT etc.) for every command. This is very different from MHDD.

It also means that WHDD is affected by whatever retries and other error handling / error recovery behaviour is coded in the Linux HBA drivers. I have already found (and forced the manufacturer to correct) faults in Linux HBA drivers, but of course there will be others. Those types of issues affect WHDD. They do not affect MHDD.

Again, this is not a criticism - it is just that any application running on an OS as complex as Linux has some "wins" (more OS services / APIs to use) and some "loses" (lack of easy direct hardware access).

For MHDD, several of its features rely on direct hardware access - e.g. how will you do disk verify to check real disk performance? Instead of verify commands, WHDD sends disk read commands (via the OS read() call), which is therefore affected by any interface, device driver, scheduling, IRQ handling (etc.) bottlenecks (and, as I mentioned before, also the HBA driver and midlayer error handling). Unless WHDD does direct ATA register access, then it cannot be a "workalike" for MHDD, because it does not work like MHDD. There are other examples, but I hope you see what I mean.

As I said before, this is not a criticism of WHDD - it simply works differently to MHDD and so, for some people, it could not be a replacement.

Thanks for sharing the info about the program

WHDD has got massive update.
Now it is capable of scanning device, and copying data from it, using ATA commands!
Copying function has beed added, with error-resilient strategies to copy all the reachable data with maximum speed and least harm to defective source disk.

It has got homepage: http://whdd.org
There is info how to get the tool, demo video, discussion maillists and so on.

Nice.

Is it part of any distros, yet?

_________________
Hard Disk Drive (HDD), Solid State Drive (SSD, SATA, NVMe, etc), USB Flash Drive and RAID Data Recovery Specialist in Massachusetts

Looks pretty cool. Does the clone process save the log somewhere so that it can pickup where it left off if the system needs to be rebooted?

_________________
Luke
Recovery Force Data Recovery

There are packages for ArchLinux and OpenSuse, and Gentoo ebuilds. But currently they are outdated. I've contacted all package maintainers, but packages updating will take its time. For now, please consider building from latest Git revision.


No log yet, but i have been informed about such concept and it is in TODO.
Regarding reboots, i want to say that in ATA operation mode, reading defective areas does not lead to system freeze (which happens in POSIX operation mode), at last i haven't experienced any such hangs.

Thanks

_________________
Hard Disk Drive (HDD), Solid State Drive (SSD, SATA, NVMe, etc), USB Flash Drive and RAID Data Recovery Specialist in Massachusetts

From what I can in the video, it is reading in a fairly large block size. Will it read down to a single sector so that every sector possible is read?

_________________
Luke
Recovery Force Data Recovery

Yes, the operation block size is 256 sectors, or 131072 bytes. I was strongly recommended to not consider trying to access in smaller units in spite of not harming the device. If there is a defect which prevents reading, there is a reasonable probablity that it is spread to more than one hardware sector. Accessing that area again and again to read it thoroughly sector-by-sector may lead to further damaging the device. In such case you should proceed with other tools.

Currently there's no good way to figure out which storage areas left unread after whdd's copying procedure. But you may workaround it: before copying, fill the destination disk/file with certain byte value pattern, like "de ad de ad ...". It will remain there on areas which were not read.

I am happy to answer any questions anywhere, but i am a bit confused to hijack the thread, after all it is not dedicated to whdd.
You are welcome to discuss anything on dedicated maillists whdd-devel and whdd-users: http://whdd.org/?page_id=8

Main idea of WHDD copier - fast copy of alive area without _digging_ into badblocks, that causes self-destroy of HDD with concentric scratches and dead heads.
Hard cases we leave for a pro .
Maybe some another smart and detailed algorithms will be added in the future. Or maybe not
Thank you.

WHDD 2.0 released.
Packages for Debian, Gentoo, Slackware and Ubuntu are available.

Hi

Seems promising.
Can you make a tutorial for instalation on ubuntu?
The one on your website doesn't work.
Thanks

_________________
http://www.pclab.com.pt facebook.com/PCLAB.A.T
ACELab partner

There are debian packages available. There's also a tutorial on installation from git. Which of them you failed to use? Please describe the actual problem in details.
I kindly ask you to proceed discussion of your problem in whdd-users maillist, see http://whdd.org/contact/