Hacker News Reader

Comment by swinglock | Hacker News Reader

https://github.com/openzfs/zfs/discussions/14793

So... WD/SanDisk only tested the SN770’s firmware with then-contemporary Windows, different behaviour from Linux/ZoL led to breakage, and now Windows too has changed its behaviour enough to expose the breakage?

(I’m still wondering if the successor, SN7100, is affected, as it’s not mentioned in that thread either way. The upmarket DRAMful SN850X seems not to be affected.)

Or not? The bug here looks different from that one, which the same outlet reported on[1] earlier.

[1] https://www.neowin.net/news/wd-ssds-still-block-windows-11-2...

Who can say until the dust has settled? My bet is on defective hardware. Maybe they can patch the firmware, or Microsoft (and Linux!) add workarounds in the kernel. But just because a kernel might do such a thing does not make it at fault when it didn't. At least not before the problem makes itself known.

This could just as well as a bug, be the Windows kernel now being better optimized to take advantage of fast storage media. Perhaps the software cache, scheduling, and what not became a bottleneck, and they fixed that. Unless it's part of the NVMe standard or what have you that the OS IO scheduler should not attempt writing "too fast", it's hard to blame it in good faith.

I'm not entirely sure it's the same issue; all the users there seem to have reflashed the SSDs to use 4k sectors and there are many reports of problems going away or never happening if 512b sector size is used. I'm doubtful Windows users are flashing their drives to 4k sectors at the same rate as ZFS-heads.

I think you misunderstood that. Configuring your filesystem to use a specific sector size (or whatever term is used for various file systems) does not modify the disk firmware. I'm not seeing it anyway.

No, that is one thing you can do (in zfs this is just setting the recordsize property) but that is not what I am talking about. I am talking about reflashing to have the firmware present logical 4k sectors instead of logical 512b sectors. Theoretically this boosts throughput and while it would technically increase latency (you'd have to rewrite a bigger sector for any change) in practice these drives all have physical sectors actually larger than 4k so this isn't an issue. It isn't a full firmware flash but it's a low-level reconfiguration of the firmware (independent of filesystem sector size), and it seems there are bugs in this mode.

https://support-en.wd.com/app/answers/detailweb/a_id/20968/~...

Thanks, now I see.

My point is various drives are bad and causes issues not only with Windows, be it with some or all firmwares from the manufacturer. Among them the only specific drive mentioned in the article. If WD advertises this mode then it ought to work.

Absolutely but my point (or rather, guess) was that it's a separate issue from the one well-documented in that GitHub link as I don't think random Windows users are doing this reconfiguration and the GH link seems to only affect reconfigured drives.

Reply on news.ycombinator.com