[Bug 281020] Cannot play sound tracks on mixed-mode CDs (read() fails with EINVAL)

[b***@freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=281020

Christos Margiolis <***@freebsd.org> changed:

What |Removed |Added
----------------------------------------------------------------------------
Assignee|***@FreeBSD.org |***@FreeBSD.org
CC| |***@freebsd.org

[b***@freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=281020

Warner Losh <***@FreeBSD.org> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@FreeBSD.org

--- Comment #1 from Warner Losh <***@FreeBSD.org> ---
commit dd78f43259efd551f76a6cb7744bce49c215863e
Author: Andriy Gapon <***@FreeBSD.org>
Date: Thu Aug 29 07:51:11 2019 +0000

scsi_cd: make the media check asynchronous

This makes the media check process asynchronous, so we no longer block
in cdstrategy() to check for media.

PR: 219857
Obtained from: ken
MFC after: 3 weeks

Is the highlighted commit. A quick glance at the driver suggests that Damian's
analysis is correct and we need to adjust our block size handling code.

[b***@freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=281020

--- Comment #2 from Damjan Jovanovic <***@gmail.com> ---
The bug still exists in the latest CURRENT.

Commit dd78f43259efd551f76a6cb7744bce49c215863e is difficult to revert on top
of CURRENT, "git revert" has many conflicts, I only succeeded by reverting it
manually, and when reverted, the bug still exists, so (unless I broke it
somehow) that commit doesn't seem to be a regression. Maybe FreeBSD's cd driver
was always buggy with mixed-mode CDs.

[b***@freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=281020

Mark Johnston <***@FreeBSD.org> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@FreeBSD.org
Status|New |Open

--- Comment #3 from Mark Johnston <***@FreeBSD.org> ---
It sounds like we're not sure if this is a regression or not?

It might be useful to see the output printed when scsi_cd probes the CD.