'Bug in kernel' message?

[louielouie]

Using the latest build 1.10.5 on Mageia 5.1
Opened the movie "Passengers", got this message but still a successful backup in what seemed like a reasonable amount of time:
...
Error 'Posix error - Unknown error 1073741823' occurred while reading '/dev/sr0' at offset '3274307584'
Unable to open file '/BDMV/STREAM/00001.m2ts' in OS mode due to a bug in OS Kernel. This can be worked around, but read speed may be very slow.
...

Wasn't sure how much of a concern this was. I have had trouble with "Gravity" but the read errors have been read errors with previous versions, also

[NullNix]

Using the latest build 1.10.5 on Mageia 5.1
Opened the movie "Passengers", got this message but still a successful backup in what seemed like a reasonable amount of time:
...
Error 'Posix error - Unknown error 1073741823' occurred while reading '/dev/sr0' at offset '3274307584'
Unable to open file '/BDMV/STREAM/00001.m2ts' in OS mode due to a bug in OS Kernel. This can be worked around, but read speed may be very slow.
...

Wasn't sure how much of a concern this was. I have had trouble with "Gravity" but the read errors have been read errors with previous versions, also

A few months ago, I asked what the bug in the kernel was so I could fix it. No response. I guess, whatever the bug is, Mike is satisfied with the workaround, but in that case it's a shame MakeMKV is still complaining about it all the time. (It is possible that upstream has already considered a fix but thinks whatever-it-is is not a bug, but there is no sign of a bug report anywhere I've looked. I've asked around and none of the block layer maintainers I've chatted to can recall ever seeing a bug report associated with this -- though it's possible that one turned up and they didn't realise what it related to.)

FWIW, I don't think the 'unknown error' is related to the 'bug in OS Kernel': I routinely see the latter message, but have never seen the former, which frankly looks more like an uninitialized variable than anything else. That's not an errno, and the value is *way* too high for it to some sort of bug where makemkv wrongly considers a short read to be an error (nobody would try to read 1GiB in a single read() call, would they? 'cos that's just not going to work).

[mike admin]

The Linux kernel fails to update the mounted device size. Insert DVD, eject, insert blu-ray, and linux kernel would still report 4GB size for an underlying block device. There is no stable repro, it just happens from time to time. MakeMKV reads inaccessible sectors using direct scsi commands, bypassing block layer. Usually, this turns out to be significantly slower.

[NullNix]

The Linux kernel fails to update the mounted device size. Insert DVD, eject, insert blu-ray, and linux kernel would still report 4GB size for an underlying block device. There is no stable repro, it just happens from time to time. MakeMKV reads inaccessible sectors using direct scsi commands, bypassing block layer. Usually, this turns out to be significantly slower.

Hmm. Interesting! You say 'mounted': you're not ejecting a disk while the fs is still mounted on it, are you? 'cos makemkv seems to try to unmount first, and frankly if you swap disks under a mounted fs you can expect much bigger problems, up to and including a kernel panic.

I'll see if I can observe this happening next time makemkv complains (which is, as you note, not too often).