SpinRite

[Joep]

The amount of Dunning Kruger in this forum is amazing! Look at this, https://forums.grc.com/threads/pre-rele ... post-10855.

"Occasionally, I give digital forensic training to LEA in my country. Once, I mentioned your software to help solve the slowness in old HDDs that needed to be forensically copied."

Is it just me or is this just plain stupid to teach a class to run Spinrite on a drive that has not yet been forensically cloned?!

[fzabkar]

I noticed that exact same post. The stupidity of some people never ceases to amaze me.

[Joep]

This seems very similar to the 4K sector issue you observed earlier Franc:
https://www.grc.com/x/news.exe/2024.01. ... bad.log.gz

In this case it's a SSD with as far as I can tell a 16K page size:

Code: Select all

  | Sector 85,557,312 (8.7591%) was found but, despite SpinRite's every      |
  | effort, none of its data could ever be recovered.                        |
  |--------------------------------------------------------------------------|
  | Sector 85,557,344 (8.7591%) Despite many attempts, this sector of the    |
  | media refused to be written and re-read.                                 |
  |--------------------------------------------------------------------------|
  | Sector 85,557,344 (8.7591%) was found but, despite SpinRite's every      |
  | effort, none of its data could ever be recovered.                        |
  |--------------------------------------------------------------------------|
  | Sector 85,557,376 (8.7591%) Despite many attempts, this sector of the    |
  | media refused to be written and re-read. 

[fzabkar]

What about sectors 85,557,313 - 85,557,343? If there were a 4K issue, why weren't there 8 consecutive LBAs in the log?

FWIW, my 860 Evo has a 512B physical sector size. My Crucial MX500 is 4KB.

I notice that SpinRite now reports whether the drive supports long reads (data + ECC).

Code: Select all

  |                                              features: SMART SECURE POWR |
  |  transfers:         32,768 sector      long ops: no  : WCACH LOOKAHD HPA |        <--  no long ops
  |  ATA speed:     96,645,755 byte/s      feature detail: 746B7D01 7469BC01 |

I also notice that SpinRite appears to be media agnostic, which is surprising, given its history as a HDD repair tool. The rotation rate reported by an SSD should be 1. In fact my 860 Evo reports a value of 1 in Identify Device word #217.

Code: Select all

| adapter vendor: Silicon Image Inc       rotation rate: not specified     |

[fzabkar]

Every Model of drive in this repository where long ops = YES

https://gitlab.com/api/v4/projects/4272 ... =build-job

Most of these are older models. I believe that some newer models that purport to support long reads actually return gibberish (according to Scott Dwyer's feedback at HDD Guru).

Edit:

I notice that ST16000NM001G-2KK103 is in the list. How can that be? It's an AF 16TB drive.

Code: Select all

4GB SATA Flash Drive
GBRLB031XVECCM       43W7685 44E915
HGST HTS541010A9E680
HGST HTS541075A9E680
HGST HTS721010A9E630
Hitachi HDS721616PLA380
HTS541010G9SA00
IBM-DSCM-11000
IC35L060AVV207-0
IC35L060AVVA07-0
Maxtor 33073U4
Maxtor 53073U6
Maxtor 5T040H4
MAXTOR 6L020J1
Maxtor 6Y080M0
Maxtor 6Y120P0
Maxtor 6Y160M0
Maxtor 94610U6
MKNSSDCR120GB
MTRON MSD-SATA3025
PQI IDE DiskOnModule
SanDisk SD8TN8U256G1001
SSD 1TB
SSD SATAIII 16GB
ST16000NM001G-2KK103     <-- Advanced Format, 16TB ???
ST3160023A
ST3160815AS
ST3250310AS
ST3320620AS
ST340810A
ST3750630AS
ST3750640AS
ST380013AS
ST380815AS
ST380819AS
ST9120821A
ST9120822AS
TEAM L7 EVO SSD 120GB
TOSHIBA MQ01ABD050
TOSHIBA MQ01ABD075
TOSHIBA MQ01ABD100
Virtium - StorFly VSFA25CC016G-200
VRFS21025GBCQSMAP1
WDC WD2500JD-98HBC0
WDC WD400AB-00BPA1
WDC WD600AB-22BVA0
WDC WD600BB-22JHA0
WDC WD800BB-53BSA0
WDC WDS500G1B0A-00H9H0

[fzabkar]

AISI, the errors in the log for the 870 Evo would be consistent with something like this:

Code: Select all

Bad        Good       Good             Good

Region 0   Region 1   Region 2   ...   Region 31

LBA x      LBA x+1    LBA x+2    ...   LBA x+31
LBA x+32   LBA x+33   LBA x+34   ...   LBA x+63
LBA x+64
...

[Joep]

What about sectors 85,557,313 - 85,557,343? If there were a 4K issue, why weren't there 8 consecutive LBAs in the log?

I was more thinking along the lines that SSD smallest unit that can be read is a page, on lower level, you can request a sector but to get that an entire page is read. I think these have 16KB or 32 * 512 sector pages. All entries in the log happen to be page aligned (assuming I am correct on 16KB pages).

I'm not certain yet this is significant but again looks like more than coincidence.

[fzabkar]

What about sectors 85,557,313 - 85,557,343? If there were a 4K issue, why weren't there 8 consecutive LBAs in the log?

I was more thinking along the lines that SSD smallest unit that can be read is a page, on lower level, you can request a sector but to get that an entire page is read. I think these have 16KB or 32 * 512 sector pages. All entries in the log happen to be page aligned (assuming I am correct on 16KB pages).

I'm not certain yet this is significant but again looks like more than coincidence.

I'm just wondering how it is that only the first sector of each page is bad. To me, the errors only make sense of the translator assigns each block of 32 consecutive LBAs to the first sectors of 32 separate pages. This would enable the SSD firmware to read 32 LBAs in parallel. Moreover, the errors would then be confined to a single page rather than 32 pages.

[fzabkar]

What about sectors 85,557,313 - 85,557,343? If there were a 4K issue, why weren't there 8 consecutive LBAs in the log?

I was more thinking along the lines that SSD smallest unit that can be read is a page, on lower level, you can request a sector but to get that an entire page is read. I think these have 16KB or 32 * 512 sector pages. All entries in the log happen to be page aligned (assuming I am correct on 16KB pages).

I'm not certain yet this is significant but again looks like more than coincidence.

I'm just wondering how it is that only the first sector of each page is bad. To me, the errors only make sense if the translator assigns each block of 32 consecutive LBAs to the first sectors of 32 separate pages. This would enable the SSD firmware to read 32 LBAs in parallel. Moreover, the errors would then be confined to a single page rather than 32 pages.

[Joep]

OK, gloves come off.

I was replying to a this thread where some guy had a problem with the drives in a Drobo, found out that none of his PCs will run SpinRite which he just purchased and then was made to jump through hoops to get it running.

A typical XY problem.

Actual problem: lost access to data on a Drobo.

The fake, secondary issue: The wild goose chase in which they now try solve by getting him to buy a Zimba board and that very likely will not help restore access to the data.

I post a comment that this looks like a typical XY problem and that he'd better had invested his money in buying a file recovery tool that words with Drobo. I pointed him towards UFS RAID.

What happened next:

1. My comment is deleted
2. I am on post-approval. Everything I comment or post has to be vetted by a moderator.

Threads like these are prove enough for me that he was not interested in solving a customer's problem.

I have now deleted my account with that fraudulent forum and will see what I will do to expose the charlatan.

Thread is here: https://forums.grc.com/threads/running- ... -hdd.1478/