kern/59792: 5.2-BETA: softupdates enabled on / kills installation

[dsf at globalcrossing.net]

>Number:         59792
>Category:       kern
>Synopsis:       5.2-BETA: softupdates enabled on / kills installation
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Sat Nov 29 07:50:12 PST 2003
>Closed-Date:
>Last-Modified:
>Originator:     Dan Foster
>Release:        FreeBSD 5.2-BETA-20031127-JPSNAP i386
>Organization:
New York Web Works
>Environment:
System: FreeBSD woot.catbert.org 5.2-BETA-20031127-JPSNAP FreeBSD 5.2-BETA-20031127-JPSNAP #0: Thu Nov 27 01:28:59 GMT 2003 root at ushi.jp.freebsd.org:/usr/obj/usr/src/sys/GENERIC i386

	P4/3.0 GHz/800FSB/HT, FreeBSD 5.2-BETA-20031127-JPSNAP
>Description:
	I have given this PR the category critical/high because it
	completely breaks installation in an unrecoverable way, and
	5.2-RELEASE is slated to be the first production-quality
	FreeBSD v5 release, so...

	5.2-BETA installer crashes into DDB when it's about to start
	extracting files from base/base.ag if softupdates is enabled
	for the / filesystem.

	Due to the nature of failure (installation failure), I have
	been unable to obtain a saved crash dump file. I have also
	manually transcribed the DDB backtrace.

	It does not appear to be an hardware issue given the particular
	backtrace output, and is also reproducible at will on a variety
	of drives (internal, external) as well as different controllers
	and busses (ATA, Firewire/ATA, USB/ATA), and has also been seen
	by at least two other people according to a search of the
	freebsd-current mailing list in the past year.

	I have only been testing 5.1-CURRENT since about September 15
	of this year (2 1/2 months now) and it was present even back
	then, and apparently for some time -- evident from other bug
	reports on freebsd-current from earlier. So this does not appear
	to be a new or recently introduced problem.

	I would be more than happy to assist with any verification
	(install tests) as I have a dedicated external hard drive for
	things like this.

9073664 bytes read from base dist, chunk  7 of 35 @ 1814.7 KBytes/sec.
kernel: type 12 trap, code=0
Stopped at	_mtx_assert+0x4e:	movl	0x1c(%ebx),%eax
db> trace
_mtx_assert(0,1,c08afff0,63e,1000>) at mtx_assert+0x4e
vm_page_set_invalid(c22540d0,0,1000,554,0) at vm_page_set_invalid+0x35
brelse(d3505a60,0,e2242b90,c70a1a00,cbcf0000) at brelse+0x52b
softdep_disk_io_initiation(d35056f0,c6cf0000,e2242bc4,246,c089e678) at
	softdesk_disk_io_initialization+0xc4
spec_xstrategy(c721e30c,d35056f0,d5b,c089e678,d35056f0) at spec_xstrategy+0x117
spec_specstrategy(e2242bf8,f4,c2278310,4,d35056f0) at spec_specstrategy+0x72
ufs_strategy(e2242c3c,e2242c6c,c06b718d,e2242c3c,1) at ufs_strategy+0x13c
ufs_vnoperate(e2242c3c,1,c089e678,360,c0980340) at ufs_vnoperate+0x18
bwrite(d35056f0,0,c089e678,69b,880aa40) at bwrite+0x44d
vfs_bio_awrite(d35056f0,0,c089e678,88e,c6cf0000) at vfs_bio_awrite+0x252
flushbufqueues(0,0,c089e678,819,64) at flushbufqueues+0x247
buf_daemon(0,e2242d48,c0896058,311,fcd2e824) at buf_daemon+0xdc
fork_exit(c06b97c0,0,e2242d48) at fork_exit+0xb4
fork_trampoline() at fork_trampoline+0x8
--- trap 0x1, eip = 0, esp = 0xe2242d7c, ebp = 0 ---
db>

>How-To-Repeat:
	Enable softupdates for the / filesystem then start installation.
>Fix:
	Current work-around is to NOT enable softupdates for the / filesystem
	only during installation time. It may be possible to safely enable
	softupdates for / post-installation but I have not tested this yet.

	If this cannot be fixed in time for 5.2-RELEASE, then I would
	recommend an update to the installer that disallows users from
	enabling software updates for / and also a mention of this in
	the install notes. This only applies to an installation from CD,
	floppy, or network and NOT from build/installworld; that particular
	case works fine.
>Release-Note:
>Audit-Trail:
>Unformatted: