Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 20 Dec 2024 11:54:10 -0800
From:      bob prohaska <fbsd@www.zefox.net>
To:        Mark Millard <marklmi@yahoo.com>
Cc:        freebsd-arm@freebsd.org
Subject:   Re: armv7, silent hang, low swap, high priorities
Message-ID:  <Z2XLYmOFoQ6wYw_X@www.zefox.net>
In-Reply-To: <633E900F-9C67-4956-A628-3A1549A40AF8@yahoo.com>
References:  <Z2Wk8DRQjEdygNlW@www.zefox.net> <633E900F-9C67-4956-A628-3A1549A40AF8@yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Dec 20, 2024 at 09:24:10AM -0800, Mark Millard wrote:
> On Dec 20, 2024, at 09:10, bob prohaska <fbsd@www.zefox.net> wrote:
> 
> > An old (v 1.1) Pi2 running armv7 -current often hangs during buildworld.
> > 
> > It doesn't respond to enter-tilda-control-B and has to be powercycled.
> > Once rebooted, buildworld can be restarted and makes further progress,
> > sometimes to completion. Swap was configured in two partitions, one
> > on microSD and one on mechanical hard disk, thus both over-provisioned 
> > and wildly unequal in speed. Root is on the mechanical disk. 
> > 
> > An orphan top window surprised me by reporting very high priorities
> > in the 130 range but swap usage was relatively low, less than 100MB.
> > That seemed an odd combination. Usually priorities that high are
> > associated with severe memory pressure. Could the mismatched swap
> > speeds have confused the scheduler? There were no console warnings. 
> > 
> > The machine was fully loaded with a -j4 buildworld, so the fact of 
> > the hang isn't hugely surprising. The top display, at
> > http://www.zefox.net/~fbsd/rpi2/crashes/20241219/top_display
> > seemed internally inconsitent.  Does it surprise anybody else? 
> > 
> > I've since removed the microSD swap, leaving only the mechanical disk
> > swap, to see if that changes the hang behavior.
> 
> 
> QUOTE
> . . .
> The nice utility runs utility at an altered scheduling priority, by in-
> crementing its "nice" value by the specified increment, or a default
> value of 10. The lower the nice value of a process, the higher its
> scheduling priority.
> END QUOTE
> 
> Renice, setpriority(), etc. agree.
> 
> Those 120+ figures are LOWER priority than the under 100 figures,
> not higher.

It never occurred to me that "PRI" might be in units of niceness.

Thank you!

bob prohaska




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Z2XLYmOFoQ6wYw_X>