Samsung SSD issues.

kronatus · 2025-10-05 02:15:42

Hi all,

This problem started a while back, but due to my own negligence I've only started debugging it now. I have three Samsung SSD's, one boots linux, one boots windows, and the last is new (they are two 860 EVOs and an 870 EVO, respectively). A while ago, my linux boot started acting strangely, where periodically the boot would fail, stating `[Failed] to start file system check on /dev/disk/by-uuid/....` fallback to the emergency initramfs where I would run fsck manually and then reboot, being able to use the machine again. Basically the problem and solution posted here: https://bbs.archlinux.org/viewtopic.php?id=223633

A week ago I ran into an issue where, during large downloads, the file system would just crash entirely, and I would have to reboot. After swapping out SATA ports and re flashing linux, it's no longer crashing completely, but instead I get the following errors in `dmesg`

[ 1733.296473] I/O error, dev sda, sector 14210048 op 0x0:(READ) flags 0x80700 phys_seg 81 
prio class 2

and

 
[ 1747.562667] ata2.00: exception Emask 0x10 SAct 0x0 SErr 0x4880000 action 0xe frozen 
[ 1747.562684] ata2.00: irq_stat 0x0c000040, interface fatal error, connection status changed
[ 1747.562690] ata2: SError: { 10B8B LinkSeq DevExch } 
[ 1747.562700] ata2.00: failed command: READ DMA EXT 
[ 1747.562704] ata2.00: cmd 25/00:c0:40:d9:bd/00:04:00:00:00/e0 tag 13 dma 622592 in res 50/00:00:3f:d9:bd/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) 
[ 1747.562723] ata2.00: status: { DRDY }

I've sense swapped out the SATA cables and the SATA ports that the SSD is attached to on the motherboard. I've also also recreated the above errors on the EVO 870. Further, I have turned on the noncq option, in addition to the typical quircks that get applied:

[    0.754836] ata3.00: Model 'Samsung SSD 860 QVO 1TB', rev 'RVQ02B6Q', applying quirks: noncqtrim zeroaftertrim noncqonati nolpmonati
[    0.754847] ata3.00: FORCE: modified (noncq)
[    0.754982] ata3.00: supports DRM functions and may not be fully accessible
[    0.754990] ata3.00: ATA-11: Samsung SSD 860 QVO 1TB, RVQ02B6Q, max UDMA/133
[    0.754997] ata3.00: 1953525168 sectors, multi 1: LBA48 NCQ (not used)
[    0.756465] ata2.00: Model 'Samsung SSD 860 EVO 1TB', rev 'RVT02B6Q', applying quirks: noncqtrim zeroaftertrim noncqonati nolpmonati
[    0.756475] ata2.00: FORCE: modified (noncq)
[    0.756609] ata2.00: supports DRM functions and may not be fully accessible
[    0.756618] ata2.00: ATA-11: Samsung SSD 860 EVO 1TB, RVT02B6Q, max UDMA/133
[    0.756625] ata2.00: 1953525168 sectors, multi 1: LBA48 NCQ (not used)
[    0.758362] ata6.00: Model 'Samsung SSD 870 EVO 1TB', rev 'SVT03B6Q', applying quirks: noncqtrim zeroaftertrim noncqonati nolpmonati
[    0.758373] ata6.00: FORCE: modified (noncq)

Also, as far as I can tell, this does not happen on windows at all, or at least, if it does, it gets covered up entirely by windows. I've also ran the extended SMART self test on the linux SSD, but am running another one right now and will attach when its finished.

Last edited by kronatus (2025-10-05 02:16:58)

seth · 2025-10-05 13:43:29

Please post your complete system journal for a boot covering those errors

sudo journalctl -b | curl -F 'file=@-' 0x0.st

They're on the bus, it's not likely that this is an internal problem w/ the drive, ie. SMART related but still show the "smartctl -a /dedv/sda"
Also make sure that "sector 14210048" is not constant but that it fails on random sector reads.

Wild guess, power related?
https://wiki.archlinux.org/title/Power_ … Management - try to enforce medium_power
Also "pcie_aspm=off", https://wiki.archlinux.org/title/Kernel_parameters

tcn · 2025-10-05 19:37:09

Will not be of help but I once had a boot from a 970 PRO NVMe just stop at:

:: running hook [btrfs]
Scanning for Btrfs filesystem
registered /dev/nvme...
:: performing fsck on 'PARTUUID=...'
:: mounting 'PARTUUID=...' on real root

And then it just stopped. After some 15 mins or so I eventually rebooted and it just worked again since.

cryptearth · 2025-10-05 20:08:21

last time I've seen such errors I ended up replacing 3 drives

seth · 2025-10-05 20:11:10

On "LinkSeq DevExch"?
Broken plug? Fried firmware?

Arch Linux

#1 2025-10-05 02:15:42

Samsung SSD issues.

#2 2025-10-05 13:43:29

Re: Samsung SSD issues.

#3 2025-10-05 19:37:09

Re: Samsung SSD issues.

#4 2025-10-05 20:08:21

Re: Samsung SSD issues.

#5 2025-10-05 20:11:10

Re: Samsung SSD issues.

Board footer