Recover failed SD card

[JD]

There is a free utility called testdisk.
I used it to recover files from a dead drive that would not mount
and would not re-format.
I recovered about 200+GB of important files.


On 02/23/2019 09:53 PM, Polytropon wrote:
> On Sat, 23 Feb 2019 17:50:27 +0100, Andrea Venturoli wrote:
>> A customer of mine gave me an SD card which is quite surely failing.
> >From what you're describing, it _has_ failed already. ;-)
>
>
>
>> I'm trying to recover what I can.
> Did you ask your customer to recover from his backups?
> Yes, of course you did. :-)
>
> Okay, let's see:
>
>
>
>> I first tried using an USB based reader: altough the SD card should be
>> 4GB in size, dd just copies 121MB. So does recoverdisk.
> This will probably apply to _any_ copying tool. For things
> where dd fails, I often try dd_rescue and ddrescue (two
> different programs). They can lower the block size and
> perform several retries, but that won't help as the device
> reports "end of medium" at 121 MB. And you cannot read
> beyond the end of medium. That particular information is
> provided by the medium itself.
>
>
>
>> "camcontrol readcap /dev/da3" gives:
>>> Last Block: 248319, Block Length: 512 bytes
>> which agains means about 121MB.
> This is "correct" according to what the card identifies as.
>
>
>
>> I put the card in another box and i get:
>>> mmcsd0: 127MB <SD  0.0 SN 00000000 MFG 00/0000 by 0x0000>
>>> at mmc0 0.4MHz/4bit/65535-block
> Above assumption is confirmed.
>
>
>
>> Is there any way I can get beyond this 121-127MB limit and read what I
>> can of the rest?
>> I looked into camcontrol's man page, but came up with no idea.
> That's probably not directly possible. The media controller
> (the "computer" on the SD card) has decided to turn itself
> into a 127 MB card, for whatever reason. A _possible_ reason
> is that the card is worn out, causing malfunctions. The same
> happens to SSDs, but those usually turn into "read-only media"
> to preserve the existing information. But for the price of
> an SD card, you cannot expect this.
>
>
>
>> The card should hold pictures, so I could go ahead with photorec once I
>> got an even partial image.
> That is the recommended approach. Maybe you can already recover
> a fraction of the images stored on the card.
>
>
>
> There is another possibility, but it's actually _very_ hard
> to do, and it's not guaranteed to work:
>
> Obtain an identical SD card. It has to be "as identical as
> possible": same control unit, same memory unit, same firmware
> revision (yes, there's "a whole computer with hard- and software"
> on that thing!). Transplant the old memory chip to the new
> card, removing its new memory chip (empty) beforehand. Then
> try another identification.
>
> If it's a micro-SD card, don't inhale the chip. ;-)
>
> As I mentioned, this is quite problematic, because the new
> controlller might see a worn-out memory chip and act the same
> as the old one. The firmware is closed source, so you don't
> have a chance to _know_ how it will act.
>
>
>
>
>