Floating Point Exception at startup with iSCSI target => Fixed in 1.5.6

[benjamin9999]

i686 Linux 2.6.22.1 SMP on AMD X2. nearly up-to-date gentoo linux i686

starting using VBoxVRDP:

root@clean1 ~ # /opt/vbox/VBoxVRDP -startvm testvbox VirtualBox VRDP Headless Interface (C) 2005-2007 innotek GmbH All rights reserved

[1] 12702 floating point exception /opt/vbox/VBoxVRDP -startvm testvbox

if i remove the iscsi target, it works.

the target runs IET 0.4.15 and works fine with other initiators.

how can i debug?

[benjamin9999]

for variety sake, i also setup archlinux, x86-64, and attempted to create the same machine and iSCSI for hda also causes floating point exception.

end of vbox.log looks like: 00:01:14.275 End of CPUID dump 00:01:14.277 Debug: HCPhys32BitPD=0000000000666000 aHCPhysPaePDs={0000000035c33000,0000000035d34000,0000000035d35000,0000000049946000} H 00:01:14.277 Debug: HCPhysInterPD=0000000000660000 HCPhysInterPaePDPTR=0000000000663000 HCPhysInterPaePML4=0000000000665000 00:01:14.277 Debug: apInterPTs={0000000000661000,0000000000662000} apInterPaePTs={0000000035cd1000,0000000035cc8000} apInterPaePDs={0000 00:01:14.277 REM: Loading /opt/VirtualBox-1.4.0/VBoxREM2.rel at 0x00000000411e9010 (30096968 bytes) 00:01:14.277 REM: (gdb) add-symbol-file /opt/VirtualBox-1.4.0/VBoxREM2.rel 0x00000000411e9010 00:01:14.360 TM: cTSCTicksPerSecond=0x781a79a8 (2015001000) fTSCVirtualized=true fTSCUseRealTSC=false fMaybeUseOffsettedHostTSC=false 00:01:14.360 CoreCode: R3=00002aaaaaaab000 R0=00002aaaaaaab000 GC=a0223000 Phys=00000000005f8000 cb=0x1000 00:01:14.372 PIT: mode=3 count=0x10000 (65536) - 18.20 Hz (ch=0) 00:01:14.373 Shared Folders service loaded. 00:01:14.394 iSCSI: login to target iqn.2001-11.com.rqdq:core.sd.testvbox successful 00:01:14.394 iSCSI: target address 10.2.2.10, target name iqn.2001-11.com.rqdq:core.sd.testvbox, SCSI LUN 0 <eof>

and kernel says... VBoxVRDP[13411] trap divide error rip:2aaaabf07586 rsp:411e6870 error:0

[benjamin9999]

forget this one, my target was actually broken.

[benjamin9999]

in this case, the target could connect, but no LU's are returned. perhaps a better error handler for this than FPU error would be nice.

[Sander van Leeuwen]

Could you retry with 1.5.0 and include a stack trace if it still faults?

[Frank Mehnert]

This bug was fixed in SVN. This divide-by-zero occurs for misconfigured iSCSI targets. Either the media sector count or the media sector size were set to zero. We display an appropriate error message now.