Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 28 Dec 2017 14:12:18 +0000
From:      bugzilla-noreply@freebsd.org
To:        freebsd-bugs@FreeBSD.org
Subject:   [Bug 224479] kernel panic in reboot+swapoff sys call
Message-ID:  <bug-224479-8-yThGMweyoN@https.bugs.freebsd.org/bugzilla/>
In-Reply-To: <bug-224479-8@https.bugs.freebsd.org/bugzilla/>
References:  <bug-224479-8@https.bugs.freebsd.org/bugzilla/>

next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D224479

--- Comment #16 from Mark Millard <markmi@dsl-only.net> ---
(In reply to Konstantin Belousov from comment #14)

You have written in the past that for file based
(vnode-based) swap files the system is deadlock
prone in a manor not under significant user control.
>From what I've seen lots of folks set up the file
based swap spaces without having a clue that such
is the status.

I was one of them at one time, following
http://www.wonkity.com/~wblock/docs/html/ssd.html#_filesystems_and_trim
material for SSDs that gave no hint of the issues
involved. If the instructions had told me that I
needed to enable the mode of use because of deadlock
issues that do not happen with partition based swap
spaces, I never would have tied it.

I had swap space based deadlocks vastly faster than
any sdcard wear out would have occurred: the
configuration was simply unreliable over fairly
short time frames. Also, the deadlocks are not
examples of wear-and-tear. (You might want a
better analogy for your point in that respect.)

I view FreeBSD as designed to automatically avoid
deadlocks for swapping only for partition-based
swap spaces.

Lesser points but more tied to this report:

I was expecting that "shutdown -r now" might stop
some processes before initiating the v-node removal,
making such processes no longer sources of swap-in
activity for later stages. (I was not thinking of
any general fix to the deadlocking issues.) reboot,
by contrast, I was expecting leaves more processes
around that might try to swap-in in an untimely
manor. I freely admit my expectations might be
garbage-in/garbage-out.

I was not expecting the sending of SIGSEGV or other
signals to a process that is trying to swap-in
after there is effectively nothing available to
swap-in from.

Does some involved kernel stack need to be in a
swapped out state to have the problem that has
been described? Or can it be a problem when no
kernel stacks are swapped out?

--=20
You are receiving this mail because:
You are the assignee for the bug.=



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-224479-8-yThGMweyoN>