NetBSD Problem Report #59560

From gson@gson.org  Sun Jul 27 14:02:50 2025
Return-Path: <gson@gson.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
	 key-exchange X25519 server-signature RSA-PSS (2048 bits)
	 client-signature RSA-PSS (2048 bits))
	(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 5644F1A923C
	for <gnats-bugs@gnats.NetBSD.org>; Sun, 27 Jul 2025 14:02:50 +0000 (UTC)
Message-Id: <20250727140241.AA780253F03@guava.gson.org>
Date: Sun, 27 Jul 2025 17:02:41 +0300 (EEST)
From: gson@gson.org (Andreas Gustafsson)
Reply-To: gson@gson.org (Andreas Gustafsson)
To: gnats-bugs@NetBSD.org
Subject: evbarm tmpfs /tmp is too small for ATF tests
X-Send-Pr-Version: 3.95

>Number:         59560
>Category:       port-evbarm
>Synopsis:       evbarm tmpfs /tmp is too small for ATF tests
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-evbarm-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Jul 27 14:05:00 +0000 2025
>Last-Modified:  Sat Apr 25 09:00:01 +0000 2026
>Originator:     Andreas Gustafsson
>Release:        NetBSD-current, source date >= 2024.12.27.19.02.19
>Organization:
>Environment:
System: NetBSD
Architecture: arm
Machine: evbarm
>Description:

The number of failing tests on the evbarm-earmv7hf testbed increased sharply
with this commit:

  2024.12.27.19.02.19 christos src/distrib/utils/embedded/conf/evbarm.conf 1.43
  2024.12.27.19.02.19 christos src/distrib/utils/embedded/conf/evbmips.conf 1.5
  2024.12.27.19.02.19 christos src/distrib/utils/embedded/conf/evbppc.conf 1.2
  2024.12.27.19.02.19 christos src/distrib/utils/embedded/conf/riscv.conf 1.3
  2024.12.27.19.02.19 christos src/distrib/utils/embedded/conf/x86.conf 1.12
  2024.12.27.19.02.19 christos src/distrib/utils/embedded/mkimage 1.85

as seen here:

  https://releng.netbsd.org/b5reports/evbarm-earmv7hf/commits-2024.12.html#2024.12.27.19.02.19

This commit had the effect of adding a tmpfs mount for /tmp which was
previously on the root file system.  The tmpfs is limited to 25% of
RAM and the testbed VM has 128 MB of RAM, so /tmp is now only 32 MB,
which is too little to reliably run the ATF tests.

See PR 54886 for some possibly relevant discussion about a similar issue
on i386.

>How-To-Repeat:

Run the ATF tests on an evbarm system with 128 MB of RAM.

>Fix:

>Audit-Trail:
From: Martin Husemann <martin@duskware.de>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: port-evbarm/59560: evbarm tmpfs /tmp is too small for ATF tests
Date: Sun, 27 Jul 2025 16:27:22 +0200

 On Sun, Jul 27, 2025 at 02:05:00PM +0000, Andreas Gustafsson wrote:
 > This commit had the effect of adding a tmpfs mount for /tmp which was
 > previously on the root file system.  The tmpfs is limited to 25% of
 > RAM and the testbed VM has 128 MB of RAM, so /tmp is now only 32 MB,
 > which is too little to reliably run the ATF tests.

 I run the sh3 atf tests on a machine with 

 Filesystem     Size   Used  Avail %Cap Mounted on
 [..]
 tmpfs           32M   588K    31M   1% /tmp


 and I would prefer if we keep tests running in such environments (skipping
 tests that need more when we identify them).

 The hardware only has 64MB ram and it can't take any additional.

 Martin

From: "Andreas Gustafsson" <gson@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/59560 CVS commit: pkgsrc/misc/py-anita
Date: Sat, 25 Apr 2026 08:07:43 +0000

 Module Name:	pkgsrc
 Committed By:	gson
 Date:		Sat Apr 25 08:07:43 UTC 2026

 Modified Files:
 	pkgsrc/misc/py-anita: Makefile distinfo

 Log Message:
 misc/py-anita: update to 2.18

 Changes from 2.17:

 Support distribution URLs having a query string and/or fragment
 component

 On platforms other than NetBSD, invoke simh-vax as just vax.

 Support upcoming sysinst changes to set the the date and time as part
 of the NetBSD/vax install.

 Store the ATF output files under /var/tmp on the guest, because
 /tmp may be a tmpfs which is too small in low-memory configurations
 (NetBSD PR 59560).

 Increase default disk size to 2G, as 1536M is no longer enough to run
 the ATF tests on amd64.


 To generate a diff of this commit:
 cvs rdiff -u -r1.81 -r1.82 pkgsrc/misc/py-anita/Makefile
 cvs rdiff -u -r1.57 -r1.58 pkgsrc/misc/py-anita/distinfo

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: Andreas Gustafsson <gson@gson.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: port-evbarm/59560: evbarm tmpfs /tmp is too small for ATF tests
Date: Sat, 25 Apr 2026 11:55:39 +0300

 The problem was exacerbated on the testbeds by anita keeping the ATF
 output files in /tmp, in addition to ATF itself using /tmp as the work
 directory for the tests.

 This part is addressed in anita 2.18 by moving the output files to
 /var/tmp.  The testbed has been running a pre-release of anita 2.18
 for a few weeks, and is now updated to 2.18.

 The ATF work directories remain on /tmp, which is still rather small,
 and this is still causing some tests to be skipped, but at least they
 are no longer failing by the hundreds.

 If you run the ATF tests under qemu using the "virt" machine type
 instead of "vexpress-a15", with the same 128 MB of memory, the tmpfs
 somehow gets created with a size even smaller than the nominal 25% =
 32 MB, and a large number of tests again fail.

 I still think the best and most complete fix for this would be to not
 configure a tmpfs /tmp by default, like we did before christos' commit
 of 2024.12.27.19.02.19, at least not on small-memory systems.
 -- 
 Andreas Gustafsson, gson@gson.org

Home	PR Database Search