[issue2066] tmpsf corruption

[Joachim de Groot (via DragonFly issue tracker)]

New submission from Joachim de Groot <jdegroot at web.de>:

I've noticed that files stored on tmpfs gets corrupted when running the
hammer.sh installation script on a swapless machine. 

Steps to reproduce:
1. Set up a virtual machine (tested on KVM and VirtualBox), 512 MB RAM
2. Boot from the installation CD (tested with 2.8.2 and 2.10.1)
3. Log in as root
4. # cp /usr/share/example/rconfig/hammer.sh .
5. Set the correct disk name in hammer.sh on line 18 
6. # csh hammer.sh

=> the script stops at line 154. Viewing the file reveals that it contains garbage. 

During the execution of the script the following messages are logged on the
console when / is being copied to /mnt (see attached screenshot): 

swap_pager_getswapspace: failed
tmpfs_write bwrite error 12
tmpfs_write bwrite error 12

Funny enough, the same issue occurs when enabling the newly-created swap device
after line 66 in hammer.sh. The swap space gets used and the messages are not
emitted, but nevertheless the script terminated too early and contains only garbage.

----------
files: tmpfs_error.png
messages: 9886
nosy: jaydg
status: unread
title: tmpsf corruption

_____________________________________________________
DragonFly issue tracker <bugs at lists.dragonflybsd.org>
<http://bugs.dragonflybsd.org/issue2066>
_____________________________________________________Attachment:
tmpfs_error.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: png00000.png
Type: application/octet-stream
Size: 9801 bytes
Desc: "Description: PNG image"
URL: <http://lists.dragonflybsd.org/pipermail/bugs/attachments/20110507/f2c5641c/attachment-0017.obj>