kern/55641: Recent changes to random/entropy "pkgsrc devel brick" an Intel Ivy Bridge system, with workaround

[davshao%gmail.com@localhost]

>Number:         55641
>Category:       kern
>Synopsis:       Recent changes to random/entropy "pkgsrc devel brick" an Intel Ivy Bridge system, with workaround
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Sep 03 18:45:00 +0000 2020
>Originator:     David Shao
>Release:        NetBSD current after a couple of weeks ago
>Organization:
>Environment:
NetBSD xxxxxx.xxx 9.99.72 NetBSD 9.99.72 (GENERIC) #1: Thu Sep  3 08:36:53 PDT 2020  xxxxxx.xxx:/usr/obj/sys/arch/amd64/compile/GENERIC amd64
>Description:
After updating NetBSD current a couple of weeks ago, building pkgsrc devel/glib2 would stop on an Intel Ivy Bridge machine (Asus P8H77-V) motherboard, i3 CPU) at a line ending in something like

. output

That is what I mean by "pkgsrc devel bricking".  Rebuilding the system, reinstalling the system, nothing will fix this.

Pressing Ctrl-C, on this system, I discovered building devel/glib2 stopped in a line calling a random number function.  Then I remember seeing on boot a line flashing by talking about entropy.

This problem does not appear at all on an Intel Eagle Lake system and on a Comet Lake system.  Therefore it will not appear across a wide variety of machines.  But from what I am seeing in the mailing lists, it will appear for some number of contributors and will render their machines useless for further pkgsrc development unless worked around.

Why Ivy Bridge?  I believe it introduced the RDRAND function.  It would not surprise me if a first generation implementation had some edge cases.

>How-To-Repeat:

>Fix:
Look up a posting:

HEADS UP: Entropy overhaul

Date: Fri, 1 May 2020 21:10:58 +0000

Apply the following two lines as root as a workaround:

dd if=/dev/urandom of=/dev/random bs=32 count=1
sysctl -w kern.entropy.consolidate=1

In the long run, it seems unlikely a fix will ever be applied since the problem occurs for a small number of older machines.  But there must be some way to communicate to people upgrading to current, or later to a release, to apply the workaround to "pkgsrc devel unbrick" their systems.