NetBSD-Bugs archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: kern/47030 (Installation stalls during set extraction)
The following reply was made to PR kern/47030; it has been noted by GNATS.
From: Andreas Gustafsson <gson%gson.org@localhost>
To: jdolecek%NetBSD.org@localhost
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: kern/47030 (Installation stalls during set extraction)
Date: Mon, 14 Nov 2016 20:13:33 +0200
jdolecek%NetBSD.org@localhost wrote:
> I can repeat the problem using 6.1 isos using qemu 2.7.0 from pkgsrc, but not
> any more with 7.0.1. Also on my test machine I can't repeat it any more
> after NVMe driver fixes. So I think this problem is actually fixed by changes
> in 7.x.
> Please confirm if you still see this problem with recent kernels.
I don't have an easy way of testing just a recent kernel with a system
otherwise identical to the the one the bug was originally reported for.
Instead, I'm testing full release builds from date-based checkouts of
mainline sources, as that is what I have an automated setup for.
I did some initial tests with NetBSD/sparc under qemu-current but they
failed to produce a conclusive result because I kept getting "esp0:
!TC on DATA XFER" errors, aka PR 46530.
I then switched to NetBSD/i386, running under qemu-current with 32M
virtual RAM and a 2G virtual disk. After some bisections, it looks
like the problems of installations stalling disappeared with the
commit
2012.11.20.20.17.43 jakllsch src/distrib/utils/sysinst/Attic/bsddisklabel.c 1.57
This commit disabled WAPBL by default. In other words, the problem
appears to be triggered by installing on a file system with WAPBL
enabled.
To confirm this, I tested installing -current/i386 (source date
2016.11.12.20.56.49) with following patch to re-enable WAPBL:
Index: src/usr.sbin/sysinst/bsddisklabel.c
===================================================================
RCS file: /bracket/repo/src/usr.sbin/sysinst/bsddisklabel.c,v
retrieving revision 1.2
diff -u -r1.2 bsddisklabel.c
--- src/usr.sbin/sysinst/bsddisklabel.c 3 Aug 2014 16:09:38 -0000 1.2
+++ src/usr.sbin/sysinst/bsddisklabel.c 14 Nov 2016 15:32:54 -0000
@@ -133,8 +133,9 @@
asprintf(&buf, "%s", mountpt);
strlcpy(p->pi_mount, buf, sizeof p->pi_mount);
p->pi_flags |= PIF_MOUNT;
- /* Default to UFS2. */
+ /* Default to logging, UFS2. */
if (p->pi_fstype == FS_BSDFFS) {
+ p->pi_flags |= PIF_LOG;
#ifdef DEFAULT_UFS2
#ifndef HAVE_UFS2_BOOT
if (strcmp(mountpt, "/") != 0)
and the installation once again stalled.
So, my conclusion is that the bug is still there in the -current
kernel, but is hidden by WAPBL no longer being enabled by default in
sysinst.
--
Andreas Gustafsson, gson%gson.org@localhost
Home |
Main Index |
Thread Index |
Old Index