pATA disk responds to Identify but panics on any block_read attempts

[DiTBho]

Code: [Select]

ata1.00: CFA: ST650211CF, 3.04, max PIO4
ata1.00: 9767520 sectors, multi 0: LBA
ata1.00: revalidation failed rno=-2)
ata1.00: revalidation failed rno=-2)
ata1.00: revalidation failed rno=-2)
ata1.00: disable device
(linux kernel v6.02)

Code: [Select]

my-mon/mips, v3.1, Feb 2025

# cf
pATA drive probing ... true
00000000: 84 8a 25 da 00 00 00 10 7e 00 02 00 00 3f 00 95 | ..%..........?..
00000010: 0a 60 00 00 20 20 20 20 20 20 20 20 20 20 33 4d | .`..          3M
00000020: 45 34 4e 58 59 43 20 20 00 03 01 00 00 04 33 2e | E4NXYC  ......3.
00000030: 30 34 20 20 20 20 53 54 36 35 30 32 31 31 43 46 | 04    ST650211CF
00000040: 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 |
00000050: 20 20 20 20 20 20 20 20 20 20 20 20 20 20 80 10 |               ..
00000060: 00 00 0a 00 00 00 02 00 00 00 00 03 25 da 00 10 | ............%...
00000070: 00 3f 0a 60 00 95 01 00 0a 60 00 95 00 00 01 00 | .?.`.....`......
00000080: 00 03 00 00 00 00 00 78 00 78 00 00 00 00 00 00 | .......x.x......
00000090: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000a0: 00 00 00 00 70 69 50 0c 40 00 70 69 10 0c 40 00 | [email protected]..@.
000000b0: 00 00 00 00 00 00 40 40 00 00 40 0d 80 80 00 00 | ......@@..@.....
000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000100: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000110: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000120: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000130: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000140: 81 4a 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | .J..............
00000150: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000160: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000170: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000180: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000190: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
LBA28={0..9767519} blocks
disk_blocks=9767520
disk_size~=4G 673M
model number : ST650211CF

# de 0
disk_read(LBA, block_n, mem_win_size
disk_ata_read(),LBA=0
disk_ata_piomode_block_get(),LBA=0
disk_ans=false

# cf
no sign of life from CF drive
pATA drive probing ... false
#
(my mon)

I don't understand this pATA device. Bought qty=4 disks, all the same behavior.
They all respond correctly to the ATA identify command (you can see what it responds to above), but they all fail every time it tries to read a block!
I tried with the Linux kernel, but it returns a strange error, "revalidation failed rno=-2)" (what?!?)
Also, after a read command, the disk goes crazy and doesn't respond at all. I have to power cycle it to get it back.

Have anyone ever seen anything like this? 

[DiTBho]

being a ~5GB device, I am using LBA28bit.

[DiTBho]

my mon uses ATA PIO with polling

[I wanted a rude username]

I once saw such a failure with an SD card (the flash had failed but the OS could still query the controller), however I suspect your issue is not that.

Have you tried changing some of the variables? E.g.:

• Different IDE cable
• Different IDE controller
• Different (ideally period-correct) motherboard
• Older kernel

[Nominal Animal]

The error comes from the Linux kernel drivers/ata/libata-core.c:ata_dev_revalidate(), and errno -2 = -ENOENT.  The -ENOENT must come from libata-core.c:ata_dev_read_id() or libata-core:ata_dev_configure(), because all the other error cases return -ENODEV = -19.

[DiTBho]

The error comes from the Linux kernel drivers/ata/libata-core.c:ata_dev_revalidate(), and errno -2 = -ENOENT. 
The -ENOENT must come from libata-core.c:ata_dev_read_id() or libata-core:ata_dev_configure(), because all the other error cases return -ENODEV = -19.

kernel 6.02, libata-core -> ata -> CF
but umm, I don't understand this error.
revalidate()? what? 

The thing that leaves me most surprised is that I wrote a low-level firmware, and I see that both the two seagate disks, two microdrives actually, respond to the Identify ATA command, but they panic as soon as I try to read a sector.

My firmware works perfectly with Hitachi drives, for which Linux also works.

With those two Seagate disks, Linux works exactly like my firmware: that is, it can identify the disk, it returns some information { serial number, model number of sectors, ... }, but then as soon as it tries to read sector 0 ... the disk freezes.

[DiTBho]

I removed the Seagate disk from the router, and put it into a USB_CF adapter.

Code: [Select]

macmini2-intel ~ # hexedit /dev/sdd
00000000   45 52 02 00  00 95 0A 60  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ER.....`................................
00000028   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
00000050   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
00000078   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
000000A0   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
000000C8   73 75 70 70  61 20 74 65  73 74 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  suppa test..............................
000000F0   64 6F 65 73  20 69 74 20  77 6F 72 6B  3F 3F 3F 3F  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  does it work????........................
00000118   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
00000140   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
00000168   00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  00 00 00 00  ........................................
it's successfully recognized, and registered as block device, /dev/sdd (SCSI emulation, pATA disk).
I am able to read and write sectors on GNU/Linux

This leads me to ask: what does the USB_CF adapter do? How do I initialize it?!?
On the router the disk is connected directly to the IO port, which is memory mapped, so ...
... I know exactly what information I'm sending and receiving through my monitor,
and I know it panics as soon as I try to make it do a PIO_read command...

Perplexed

[DiTBho]

Code: [Select]

macmini2-intel ~ # badblocks -svw /dev/sdd
Checking for bad blocks in read-write mode
From block 0 to 4883759
...

let's stress and verify it ... 

[Nominal Animal]

CompactFlash – SDHC/SDXC maps ATA commands to Secure Digital commands.  They're not the same, but they map well enough.

The error:

When an ATA bus is found, drivers/ata/libata-core.c:ata_bus_probe() is called.  The devices are probed and identified in a funky order (because of master/slave signaling and possible bridges).  This initial identification uses ata_dev_read_id(dev, &dev->class, ATA_READID_POSTRESET, dev->id).

Next, the drivers/ata/libata-eh.c:ata_set_mode() is called on each device, to set timings and issue SET FEATURES (ATA command 0xEF).  If the ata port does not have its own ops->set_mode function, then drivers/ata/libata-core.c:ata_do_set_mode() gets called.  It first sets transfer masks (see libata-core.c:ata_dev_xfermask(), libata-core.c:ata_force_xfermask()), then PIO timings, then DMA timings, and finally updates their transfer mode by calling drivers/ata/libata-core.c:ata_dev_set_mode(), which calls drivers/ata/libata-core.c:ata_dev_revalidate(dev, ATA_DEV_UNKNOWN, 0).  (Note that ata_class_enabled(ATA_DEV_UNKNOWN) is false, so the "fail early if" case will not trigger.)

That function ends up calling drivers/ata/libata-core.c:ata_dev_reread_id(dev, 0), which does the second ata_dev_read_id(dev, &dev->class, 0, dev->id).  (You might wish to change the ata_dev_dbg()'s in it to ata_dev_err(), so you get the exact reason.)  The other possibility is that the -ENOENT is returned by the call to drivers/ata/libata-core.c:ata_dev_configure().

Now, in drivers/ata/libata-core.c:ata_dev_read_id() we find an interesting comment about why the ATA_READID_POSTRESET is needed: (SRST mode,) RESET, IDENTIFY, ..., and so on.  Perhaps your Seagates are just crappy pre-ATA4, and require some woo-woo-special sequence to work?

I would definitely change drivers/ata/libata-core.c:ata_dev_read_id() to ata_dev_printk(dev, KERN_ERR,  "failed to IDENTIFY (%s, err_mask=0x%x)\n", reason, err_mask); unconditionally in the err_out: error path, to pinpoint where the failure occurs.

[I wanted a rude username]

I removed the Seagate disk from the router

Adding to what Nominal Animal has said, this is highly relevant information ... should have been mentioned up front. Odds on that router was never designed to spin magnetic disks ... it's quite possible that it can't deliver the startup and/or seek current ... that would line up with the failure mode.

Have you verified that when installed in the router, the drive's actually spinning?

[DiTBho]

Odds on that router was never designed to spin magnetic disks ...

Mikrotik advertised their router as
                       "On RB532 build you can also install a CompactFlash module or Microdrive drive
                        which can be used as an alternative boot device or an additional storage medium.
                        Warning! It does not support hot insert of CompactFlash/Microdrive devices."

it's quite possible that it can't deliver the startup and/or seek current ... that would line up with the failure mode.

but in this case, shouldn't the disk refuse to respond to the ATA "Identify command"?
Say, the electronic is powered, but the motor doesn't have enough current to spin,
in this case the electronic should notify the problem by reporting "diagnostic failure" or something?

Isn't there a specific ATA "Diagnostic command" to know if the motor is spinning or not?

Have you verified that when installed in the router, the drive's actually spinning?

The Hitachi 4G, which works perfectly, makes no noise when its disk spins, so umm dunno with the Seagate

[DiTBho]

CompactFlash – SDHC/SDXC maps ATA commands to Secure Digital commands.  They're not the same, but they map well enough.

I am very confused. I cannot find a good complete document on the subject.
I bought a book called "PC hardware bible".
It explains something about pATA disks but in a very very superficial way

No examples, no in-depth analysis.

Worse still Googling you find tons of confused articles.
Better information from blogs of people who implemented something for their vintage MPUs.
But, once again, it's very basic.

I would definitely change drivers/ata/libata-core.c:ata_dev_read_id() to ata_dev_printk(dev, KERN_ERR,  "failed to IDENTIFY (%s, err_mask=0x%x)\n", reason, err_mask); unconditionally in the err_out: error path, to pinpoint where the failure occurs.

Good idea 

[DiTBho]

Code: [Select]

macmini2-intel ~ # badblocks -svw /dev/sdd
Checking for bad blocks in read-write mode
From block 0 to 4883759
Testing with pattern 0xaa: done
Reading and comparing: done
Testing with pattern 0x55: done
Reading and comparing: done
Testing with pattern 0xff: done
Reading and comparing: done
Testing with pattern 0x00: done
Reading and comparing: done
Pass completed, 0 bad blocks found. (0/0/0 errors)
The Seagate 5GB on the CF_usb adapter passed all the badblocks tests!


Is there any CF-to-CF adapter you know of that provides an alternative way to supply power to the microdrive?

[DiTBho]

cf spec, v1.4
 cf-spec-v1.4.pdf (637.22 kB - downloaded 114 times.)

[Nominal Animal]

ATA = "AT Attachment" is developed by Technical Committee T13 at www.t13.org; the latest standard is ACS-5, ISO/IEC 17760-105:2024.

SD card standards are managed by SD Association at www.sdcard.org.

Odds on that router was never designed to spin magnetic disks

ST650211CF = 5 GB Seagate ST1, a miniature 1" HDD in CompactFlash II form factor (CF itself using parallel ATA, aka IDE); datasheet (PDF).  It draws less than 350mA at 3.3V or 5.0V (averaged over 10ms), which the RB532A should be able to handle fine, given good enough DC supply brick.

Is there any CF-to-CF adapter you know of that provides an alternative way to supply power to the microdrive?

You could kludge one together, using 0.050" = 1.27mm pitch 2×25-pin male-female connectors.  Pins 13 and 38 are the VCC, but it could be either 3.3V or 5.0V.  (You'll want to check your CF card socket, as otherwise you can burn sensitive I/O pins.)
For example, using a short length of 50-wire 0.025"/0.635mm pitch flat cable (like 3M 3749/50) with Harwin M50-3302542 and M50-3802542 IDC connectors.

For the RB532A, consider using the higher supply voltage setting, and a higher voltage DC supply.  It's supposed to have pretty beefy DC-DC converters; the two watts shouldn't be a problem.

[DiTBho]

Does anyone happen to have the datasheet of the Hitachi HMS360404D5CF00?

I'd like to see the supported ATA commands, in details

[DiTBho]

The Hitachi HMS360404D5CF00 works very well

Code: [Select]

# cf_probe
ata_disk_init
pATA drive probing ... true
00000000: 84 8a 1f 00 00 00 00 10 7e 00 02 00 00 3f 00 7a | .............?.z
00000010: 10 00 00 00 20 20 20 20 20 20 20 20 20 20 20 20 | ....
00000020: 4e 32 4c 48 55 45 59 41 00 03 01 00 00 04 44 4e | N2LHUEYA......DN
00000030: 34 4f 43 41 32 41 48 4d 53 33 36 30 34 30 34 44 | 4OCA2AHMS360404D
00000040: 35 43 46 30 30 20 20 20 20 20 20 20 20 20 20 20 | 5CF00
00000050: 20 20 20 20 20 20 20 20 20 20 20 20 20 20 80 20 |               .
00000060: 00 00 0e 00 00 00 02 00 00 00 00 03 1f 00 00 10 | ................
00000070: 00 3f 10 00 00 7a 01 00 10 00 00 7a 00 00 00 00 | .?...z.....z....
00000080: 00 03 00 00 00 00 00 78 00 78 00 00 00 00 00 00 | .......x.x......
00000090: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000a0: 00 1e 00 12 70 68 50 0c 40 00 70 48 10 0c 40 00 | [email protected]..@.
000000b0: 00 00 00 00 00 00 40 60 00 00 00 00 00 00 00 00 | ......@`........
000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000000f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000100: 00 00 00 02 00 00 01 11 00 00 00 00 00 00 00 00 | ................
00000110: 00 00 00 00 00 64 00 00 00 00 00 00 00 00 00 00 | .....d..........
00000120: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000130: 00 00 00 00 00 00 00 00 00 00 00 00 04 b0 00 01 | ................
00000140: 81 00 80 01 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000150: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000160: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000170: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000180: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00000190: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
000001f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
LBA28={0..7999487} blocks
disk_blocks=7999488
disk_size~=3G 834M
block_n=7M 644K
model number : HMS360404D5CF00
adv PIO mode 00000300
Features/Command sets supported(1) 00006870
Features/Command sets supported(2) 00000c50
Features/Command sets supported(3) 00000040
Features/Command sets supported(4) 00004870
Features/Command sets supported(5) 00000c10
Features/Command sets supported(6) 00000040
Ultra DMA support and current mode 00000000
Advanced power management value 00006040
CFA power mode(1) 00000081
Capabilities bits 00000e00 -> is_LBA_capable=true
block_size=0 ??? assuming 512byte
wtf? sec size(1) 00000000
wtf? sec size(2) 00000000
added more checks

Code: [Select]

# parts
found mac-part, blocksize=512, scanning ... done

Partition map..................... 32bit  1  63  31K 512
bootstrap......................... 32bit  64  1600  800K
kernel............................ 32bit  1664  65536  32M
swap.............................. 32bit  67200  524288  256M
rootfs............................ 32bit  591488  7408000  3G 545M 192K
#
added mac-part, this part works well 

I will go back with the Seagate ST1.

[DiTBho]

ST650211CF [..] the VCC could be either 3.3V or 5.0V [..]

So the datasheet says.

The drive receives DC power (+3.3V or +5V) through the CompactFlash interface connector for ST650211CF and ST625211CF models, and (+3.3V only) through the Flex interface connector for ST650211FX and ST625211FX models.

However, on the wikipedia, this describes exactly what I see

Code: [Select]

Certain bus-powered CF card readers lack the power needed to run a Microdrive although they do take CF II cards.
When using such a device, it will usually be detected by the host, but errors will occur once the user attempts to access the drive

detected -> ATA command "identify": success
access the drive -> ATA command "read_sector": failure, drive panic!

Today I noticed on the back of the MD a little label that says "5V" ... 

[Nominal Animal]

Interesting.  If I'm not horribly mistaken (see RB500 series User Manual), the RB532A miniPCIe slots are powered by a (common) 3.3V 4.25A supply.

[DiTBho]

Yup, the rb532a miniPCI is 3.3V only
I checked, the 5V power is not implemented.

The manual doesn't say if the CF module is 3.3V or 5V.
I suspect it's 3.3V ... Can't measure at the moment

But umm ... that also means some MicroDrives are OEM
OEMs are 5V only, and are true-IDE-only.

[DiTBho]

Some "OEM Only" drives use the CompactFlash form factor but only provide a 5V IDE/ATA interface.
These will not work in readers or devices which expect a 3.3V interface and full CompactFlash functionality

look at the Seagate ST1

• it perfectly works with the USB_CF adapter, which operates at 5V. Badblocks passed, zero error reported
• it panics on the rb532a at every ATA command that involves the spinning motor and heads (sector_read, sector_verify, etc ...), but it responds to the ATA identify command, which only involves the control board of the microdrive but not the spinning motor

My guess...the rb532a CF module must be 3.3V.

Ordered a 2GB microdrive made by GS Magicstor. Paid 10 euro on eBay.
Magicstor is a Chinese manufacturer subsidiary of GS Magic, Inc..
The datasheet of their microdrives says
- 2GB 3.3V/5V
- 4GB 3.3V-only

According to Google, there is no OEM version.

With Hitachi... it's hell:
not only there are also OEM versions,
but also custom IBM-made versions for their MIPS WinCE-v1.0 laptop (only sold in Japan),
marked "IBM Microdrive" but made with Hitachi parts,
who also made the Apple iPod drive,
and whatever Apple, IBM or Hitachi microdrive you look at ...
... they are not easy to identify as "OEM" or "non-OEM", as there is no indication on the label!

Some Hitachi microdrives made for Apple iPod have a log on the front,
usually the Apple logo, which means it is an OEM drive, 5V only, True-IDE only,
but other Hitachi microdrives have no indication at all.

The datasheet is very vague, and reports approximate things compared to what I observe!

[DiTBho]

The CF specs say CF cards "should" have to comply with ATA specs, and work with both 3.3V and 5V.
Not respected by OEM.
Not respected by Hitachi and Seagate

Seagate doesn't say it in the datasheets, but when it says 5V on some microdrives, well ..
... well coincidentally those are the ones that don't work on 3.3V only machines, like the RB532A.