VirtualBox

Opened 15 years ago

Closed 12 years ago

#2850 closed defect (fixed)

OS/2 guest USB stops working with VB > 2.0.4

Reported by: Trevor Hemsley Owned by:
Component: USB Version: VirtualBox 3.0.6
Keywords: Cc:
Guest type: other Host type: Linux

Description (last modified by Frank Mehnert)

I have an OS/2 guest system using all the latest OS/2 drivers running on a Centos 5.2 32 bit Linux system. When running VirtualBox 2.0.4 everything works correctly and the guest can see the 2 USB external disks and the data on them. If I upgrade to 2.0.6 or 2.1.0 then I lose the 2 external USB disks from the guest machine.

On Centos 5.2, the usbfs file system is mounted from /etc/rc.sysinit at boot time and I have amended this script to contain the following lines

if [ ! -d /proc/bus/usb ]; then

modprobe usbcore >/dev/null 2>&1 && mount -n -t usbfs /proc/bus/usb /proc/bus/usb -o devgi

d=501,devmode=664 else

mount -n -t usbfs /proc/bus/usb /proc/bus/usb -o devgid=501,devmode=664

fi

This works correctly as far as 2.0.4 but fails as soon as I upgrade to 2.0.6+. If I rpm -e the newer release and re-install 2.0.4 then everything works correctly again.

Attachments (8)

VBox.log (39.9 KB ) - added by Trevor Hemsley 15 years ago.
Vbox.log from OS/2 boot under 2.1.0
VBox.2.log (40.5 KB ) - added by Trevor Hemsley 14 years ago.
3.0.6 vbox log
lvm-on-3.0.6.png (9.9 KB ) - added by Trevor Hemsley 14 years ago.
lvm screen contents under VBox 3.0.6
lvm-on-2.0.4.png (9.8 KB ) - added by Trevor Hemsley 14 years ago.
lvm screen on VBox 2.0.4 reuploaded correct
config.os2 (8.2 KB ) - added by Trevor Hemsley 14 years ago.
config.sys from this system
config.2.os2 (8.2 KB ) - added by Trevor Hemsley 14 years ago.
config.sys from this system
VBox.3.log (38.1 KB ) - added by Trevor Hemsley 14 years ago.
Everything working on 2.0.4
VBox.4.log (44.1 KB ) - added by Trevor Hemsley 14 years ago.
log file from3.0.8 without USB device attached then attached via menu

Download all attachments as: .zip

Change History (53)

by Trevor Hemsley, 15 years ago

Attachment: VBox.log added

Vbox.log from OS/2 boot under 2.1.0

comment:1 by Trevor Hemsley, 15 years ago

Have tried every release of VB since 2.0.4 and all fail to allow use of USB hard disks under OS/2. 2.1.4 is better than most but still does not work. With 2.1.4, the disk appears when I run lvm but it shows as a "Compatibility" volume and it's not possible to get OS/2 to recognise the partitioning of the disk. In addition, attempting to use the drive from OS/2 running under VB 2.1.4 will corrupt the lvm meta data on the disk - according to a product called DFSee, the lvm CRC info is incorrect and is set to 0xffffffff.

comment:2 by Trevor Hemsley, 14 years ago

Still not working with VB 3.0.6

USB disks attach to the guest correctly and can be assigned via the menu but will not be seen by the guest O/S at all. An lvm /rediscoverprm just hangs forever.

This seems to work on a Windows XP host but fails on a Centos 5 host. The guest disk was identical as I copied the VDI from one machine to the other and booted it up.

Uninstalling 3.0.6 and reinstalling 2.0.4 (the last release that worked correctly) immediately fixes the problem with no other changes required (apart from restoring the xml files from the backup copies).

comment:3 by Frank Mehnert, 14 years ago

Version: VirtualBox 2.1.0VirtualBox 3.0.6

Thank you for this detailed report. Looking at your log file, there are two USB devices (04b4:6830 and 067b:2507) which are attached to your guest. Are these the two disks? You can check the USB IDs with lsusb or cat /proc/bus/usb/devices. AFAICS, the only USB-related change was to introduce the iPhone fix. Maybe this somehow conflicts with your hard disks. As you are on CentOS 5 I assume that you are still on Linux 2.6.18, correct?

comment:4 by Trevor Hemsley, 14 years ago

04B4:6830 is one of them. The other one seems to be 07AB:FCCB which reports itself as "Freecom Technologies". It's possible that I have changed the USB caddy as I've had a couple crap out in the last couple of years.

This is a self compiled kernel with the .config taken from a standard Centos 5 kernel and then had make oldconfig run against it followed by a make rpm to generate RPMs ready to be installed. It reports itself as

[root@trevor4 ~]# uname -a Linux trevor4 2.6.29.4 #4 SMP Fri May 29 23:00:08 BST 2009 i686 i686 i386 GNU/Linux

I can try newer kernels generated using the same method if that would be useful or I can install the standard Centos 2..6.18-164 kernel if that would aid in debugging this. I look after Linux systems for a living so am fairly comfortable with hacking them about :)

comment:5 by michaln, 14 years ago

Yes, trying a stock kernel would be definitely worthwhile. Also, what about other guests (Windows, Linux)?

I am pretty skeptical that the iPhone fix would be relevant here, but then the OS/2 USB stack could be extremely weird. Trying a different guest should answer that question. A Windows host OTOH is so different that whether it works or not for the same guest is not very significant.

comment:6 by Trevor Hemsley, 14 years ago

Hi Michal, by 'stock' I presume you mean a distribution kernel as opposed to a stock vanilla kernel from kernel.org (which is what I am currently running). I will try tonight if I get time. I'll try with everything that I can find - I have a 2.6.31.1 kernel compiled here and ready to go and have also just installed the kernel-PAE-2.6.18-164 kernel from Centos. Will try the Centos one first and see if that makes any difference.

comment:7 by Trevor Hemsley, 14 years ago

Right, first boot on kernel-PAE-2.6.18-164 with a fresh install of VirtualBox 3.0.6 i386. OS/2 boots OK but lvm /rediscoverprm hangs immediately when run. I will attach the latest log file above. If the file changes if/when lvm /rediscoverprm ends then I will upload any extra lines that it appends in the meantime as a separate file.

by Trevor Hemsley, 14 years ago

Attachment: VBox.2.log added

3.0.6 vbox log

comment:8 by Trevor Hemsley, 14 years ago

lvm /rediscoverprm finished in about 10 minutes but no further lines we added to the log file. Neither of the two USB disks attached appear to be accessible (i.e. df -m shows nothing for J: or K: which is where they normally appear when it works).

Invoking lvm itself took about 20/25 minutes to show anything. I'll take a screenshot of this and how it looks under 2.0.4 so that you can see the difference. The disks _are_ seen but just don't work :(

by Trevor Hemsley, 14 years ago

Attachment: lvm-on-3.0.6.png added

lvm screen contents under VBox 3.0.6

by Trevor Hemsley, 14 years ago

Attachment: lvm-on-2.0.4.png added

lvm screen on VBox 2.0.4 reuploaded correct

comment:9 by Trevor Hemsley, 14 years ago

The most interesting differences I can see between the two are that on 3.0.6 both disks are seen as 2MB larger than they are on 2.0.4. On 3.0.6 they are also seen as 'Removable' drives which I think indicates that they are being treated as enormous floppy drives.On 2.0.4 they are both recognised as LVM volumes but on 3.0.6 they both show as Compatibility volumes.

comment:10 by michaln, 14 years ago

Yes, I meant the default kernel that ships with CentOS - basically whatever comes with your distro. I understand that did not make any discernible difference (correct?); that's good to know.

The description of the problem sounds rather odd, and I'm beginning to wonder if it's strictly USB related at all. And FWIW, the size differences you saw could be the size of a disk vs. the size of a filesystem on that disk.

It would help if you could try a different guest OS, then we would know if it's OS/2 guest specific or not (I suspect it is, but can't tell for sure).

Also, what is your OS/2 setup exactly? I have a MCP2 VM here, but I haven't tried using USB with OS/2 ever since I bought a Mac a few years ago :) Could you perhaps attach a copy of your OS/2 VM's CONFIG.SYS?

comment:11 by Trevor Hemsley, 14 years ago

I am fairly sure that I tried a Centos guest under a Centos host as an experiment a few months ago and that worked ok but I would have to repeat the installation to verify this. I will attach a copy of my config.sys later on when I have access to the machine. It is WSeB ACP1 but patched up to date with all the latest fixpacks and drivers - can supply bldlevel output for \os2\boot\usb* if that would be helpful. They will be the last ones that IBM released though but I guess it'll be helpful just to make sure.

comment:12 by michaln, 14 years ago

Given how many versions of OS/2 USB drivers are floating around, it would be good to know what versions exactly you're using. File sizes and dates of the USB drivers would be helpful.

As an alternative, if you were willing to upload your OS/2 image to us (preferably using "export appliance" functionality in VBox 3.x), that might be a lot simpler.

comment:13 by Trevor Hemsley, 14 years ago

I suspect the upload of this one might be a bit large - it's on a 160GB virtual disk :)

I think I have the same issue on an ECS install though and that's only on a 2GB disk which might zip down quite a lot.

Let's start with info about driver versions and see if that helps first.

comment:14 by Trevor Hemsley, 14 years ago

[D:\os2\boot]bldlevel usbmsd.add Build Level Display Facility Version 6.10.480 Oct 6 2000 (C) Copyright IBM Corporation 1993-2000 Signature: @#IBM:10.162#@ OS/2 USB MSD Class Adapter Driver Vendor: IBM Revision: 10.162 File Version: 10.162 Description: OS/2 USB MSD Class Adapter Driver

[D:\os2\boot]bldlevel usbd.sys Build Level Display Facility Version 6.10.480 Oct 6 2000 (C) Copyright IBM Corporation 1993-2000 Signature: @#IBM:10.162#@ OS/2 USB Device Driver Vendor: IBM Revision: 10.162 File Version: 10.162 Description: OS/2 USB Device Driver

[D:\os2\boot]hcimonit You have 1 PCI USB OHCI host controller(s) You have 1 PCI USB EHCI host controller(s)

[D:\os2\boot]bldlevel usbehcd.sys Build Level Display Facility Version 6.10.480 Oct 6 2000 (C) Copyright IBM Corporation 1993-2000 Signature: @#IBM:10.162#@ OS/2 USB EHCI compliant Device Driver Vendor: IBM Revision: 10.162 File Version: 10.162 Description: OS/2 USB EHCI compliant Device Driver

[D:\os2\boot]bldlevel usbohcd.sys Build Level Display Facility Version 6.10.480 Oct 6 2000 (C) Copyright IBM Corporation 1993-2000 Signature: @#IBM:10.162#@ OS/2 USB OHCI compliant Device Driver Vendor: IBM Revision: 10.162 File Version: 10.162 Description: OS/2 USB OHCI compliant Device Driver

comment:15 by Trevor Hemsley, 14 years ago

Yuck

[D:\os2\boot]bldlevel usbmsd.add
Build Level Display Facility Version 6.10.480 Oct  6 2000
(C) Copyright IBM Corporation 1993-2000
Signature:       @#IBM:10.162#@  OS/2 USB MSD Class Adapter Driver
Vendor:          IBM
Revision:        10.162
File Version:    10.162
Description:     OS/2 USB MSD Class Adapter Driver

[D:\os2\boot]bldlevel usbd.sys
Build Level Display Facility Version 6.10.480 Oct  6 2000
(C) Copyright IBM Corporation 1993-2000
Signature:       @#IBM:10.162#@  OS/2 USB Device Driver
Vendor:          IBM
Revision:        10.162
File Version:    10.162
Description:     OS/2 USB Device Driver

[D:\os2\boot]hcimonit
You have 1 PCI USB OHCI host controller(s)
You have 1 PCI USB EHCI host controller(s)

[D:\os2\boot]bldlevel usbehcd.sys
Build Level Display Facility Version 6.10.480 Oct  6 2000
(C) Copyright IBM Corporation 1993-2000
Signature:       @#IBM:10.162#@  OS/2 USB EHCI compliant Device Driver
Vendor:          IBM
Revision:        10.162
File Version:    10.162
Description:     OS/2 USB EHCI compliant Device Driver

[D:\os2\boot]bldlevel usbohcd.sys
Build Level Display Facility Version 6.10.480 Oct  6 2000
(C) Copyright IBM Corporation 1993-2000
Signature:       @#IBM:10.162#@  OS/2 USB OHCI compliant Device Driver
Vendor:          IBM
Revision:        10.162
File Version:    10.162
Description:     OS/2 USB OHCI compliant Device Driver

by Trevor Hemsley, 14 years ago

Attachment: config.os2 added

config.sys from this system

by Trevor Hemsley, 14 years ago

Attachment: config.2.os2 added

config.sys from this system

comment:16 by Trevor Hemsley, 14 years ago

One more for luck

[D:\]bldlevel os2krnl
Build Level Display Facility Version 6.10.480 Oct  6 2000
(C) Copyright IBM Corporation 1993-2000
Signature:       @#IBM:14.104a#@_SMP  IBM OS/2 Kernel
Vendor:          IBM
Revision:        14.104
File Version:    14.104
Description:     _SMP  IBM OS/2 Kernel

comment:17 by michaln, 14 years ago

Thanks for the info, though the last bit leaves me a little confused. Are you actually running the SMP kernel even though the VM has just one CPU? I'd recommend running the W4/UNI kernel (for MCP/ACP) to eliminate that as a possible source of problems. Besides, the SMP kernel will only slow things down on a single-CPU machine.

Here's something I don't understand: Your log file from 2.1.0 (why not a log from working 2.0.4?) contains two "OCHI: root hub reset completed with VINF_SUCCESS" at the point when the guest initializes USB. However, the 3.0.6 log contains no such messages. I don't know if that's significant on its own, but I see that your config.sys has USBOHCD.SYS commented out. Are you *sure* that is a valid OS/2 USB configuration? EHCI does not provide a complete USB host controller implementation, it always requires a companion OHCI or UHCI controller as a fallback.

In the meantime, I ran some trials of my own, using MCP2 + FP4 as a basis. With the 10.162 USBBASIC and USBMSD packages, I was unable to get the VM to boot. After some trial and error I determined that the VM hangs in USBOHCD.SYS for some reason. However, using USBOHCD.SYS build 10.70 (dated 5/21/2002) avoids the hang and the VM successfully boots.

I didn't have much luck with USB devices on either Linux or Windows hosts though. The VM captures the USB device, but there is no USB activity. Running "lvm /rediscoverprm" returns immediately and the guest doesn't appear to notice the USB device at all. This was with unreleased development VBox code though...

On my Mac laptop, using VBox 3.0.6, the same OS/2 VM was able to see and use a 1GB USB memory stick, I could read and write it from OS/2 and use it to exchange data between the OS/2 VM and the Mac host (plain FAT16 filesystem).

There is certainly something odd going on with OS/2 USB inside VirtualBox, but then the OS/2 USB stack in general doesn't exactly inspire confidence. If you know of some solid OS/2 diagnostic tools which can examine the internal state of the USB stack (which devices are attached etc.), please let us know.

BTW, is there supposed to be any difference between the config.os2 and config.2.os2 files you uploaded? (I'm guessing there isn't)

comment:18 by Trevor Hemsley, 14 years ago

I'm running the SMP kernel because it is the only one that works with the ACPI PSD. I've tried all kernels and this is the only one that works with VirtualBox and allows USB and networking to work - and they both require acpi.add to function.

The lack of usbohcd.sys is a similar problem but I don't remember if this one was to do with VBox or if it was required by the underlying hardware when this machine ran native on the same hardware. With usbohcd.sys then the boot never completes - just hangs before SNAP initialises. Ah, yes, I see you had the same problem :)

The two config.os2s are due to me pressing the back button and then getting distracted and forgetting that I'd already uploaded it. The 2nd could be deleted but I don't have the ability to do that.

The log file from 2.1.0 probably should have been accompanied by one from 2.0.4 to show how it looks when it works. I'll generate one and upload it

by Trevor Hemsley, 14 years ago

Attachment: VBox.3.log added

Everything working on 2.0.4

comment:19 by Trevor Hemsley, 14 years ago

Uploaded and this one only has one root hub reset in it vs the two in both the other files. The 3.0.6 log does contain the msgs but they are EHCI not OHCI. The EHCI ones do not appear in either of the other two logs.

comment:20 by Trevor Hemsley, 14 years ago

I tried to put in basedev=usbohcd.sys but with 10.162 of this file then it stalls at boot time on VB 2.0.4. I did a quick bit of archaeology and found a copy of 10.084 which worked on VB 2.0.4. Reinstalled 3.0.6 and rebooted the VM and it's still running lvm /rediscoverprm as I type. Not looking hopeful :(

The last few lines of the log say

00:00:28.877 Audio: set_record_source ars=0 als=0 (not implemented) 00:00:28.883 Audio: set_record_source ars=0 als=0 (not implemented) 00:00:30.847 Audio: set_record_source ars=4 als=4 (not implemented) 00:00:30.857 EHCI: Hardware reset 00:00:31.077 EHCI: root hub reset completed with VINF_SUCCESS 00:00:31.971 EHCI: USB Operational 00:00:32.627 OHCI: USB Operational 00:00:32.673 PIT: mode=2 count=0x10000 (65536) - 18.20 Hz (ch=0) 00:00:37.879 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=b1c00000 w=640 h=480 bpp=8 cbLine=0x280 00:00:41.060 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=00000000 w=720 h=400 bpp=0 cbLine=0x0 00:00:41.691 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=b1c00000 w=1600 h=1200 bpp=32 cbLine=0x1900

comment:21 by Trevor Hemsley, 14 years ago

Grrr, &*@$ing wiki formatting

00:00:28.876 Audio: set_record_source ars=0 als=0 (not implemented)
00:00:28.877 Audio: set_record_source ars=0 als=0 (not implemented)
00:00:28.883 Audio: set_record_source ars=0 als=0 (not implemented)
00:00:30.847 Audio: set_record_source ars=4 als=4 (not implemented)
00:00:30.857 EHCI: Hardware reset
00:00:31.077 EHCI: root hub reset completed with VINF_SUCCESS
00:00:31.971 EHCI: USB Operational
00:00:32.627 OHCI: USB Operational
00:00:32.673 PIT: mode=2 count=0x10000 (65536) - 18.20 Hz (ch=0)
00:00:37.879 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=b1c00000 w=640 h=480 bpp=8 cbLine=0x280
00:00:41.060 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=00000000 w=720 h=400 bpp=0 cbLine=0x0
00:00:41.691 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=b1c00000 w=1600 h=1200 bpp=32 cbLine=0x1900

comment:22 by michaln, 14 years ago

Is acpi.add some kind of eCS thing? I can safely say that it's not required for any of the USB packages from IBM, and it's not needed for IBM networking either.

It's good to know that the 10.162 USBOHCD.SYS is behaving the same in VBox 2.0.4. I wonder if that's actually something IBM broke, since the older builds seem to work in VBox, or at least don't hang. I don't have anything between 10.084 and 10.162, so hard to tell when USBOHCD.SYS stopped working.

Anyway I do see some problems, but not lvm /rediscoverprm hanging or taking forever... it's more like OS/2 doesn't see the devices at all (when it doesn't work). Have you tried booting the VM without any USB devices and attaching them later?

comment:23 by Trevor Hemsley, 14 years ago

Acpi.PSD is a replacement for OS2APIC.PSD and enables SMP support on more recent motherboards that use ACPI tables for enumerating multiple cpus instead of the older type of BIOS data that os2apic.psd understands. It's necessary for any multi-core machine that's been released in about the last 4 or 5 years if you want OS/2 to see more than one core of your processor.

Long term, if I can get 3.0.6 installed I would love to use more than one cpu within OS/2.

I'll try the trick with attaching the USB device after boot but I think I already did so in the past. Worth double checking though.

comment:24 by Trevor Hemsley, 14 years ago

Attempt #4517 :)

Installed VB 3.0.8 tonight and this changes the symptoms of the problem a bit. lvm /rediscoverprm still takes an age to complete and no disk shows up afterwards. Running lvm afterwards does show the drive and this time it is marked as [REMOVABLE 1] which is different I think.

Prior to booting with 3.0.8 I had made some changes to the OS/2 boot by replacing PSD=ACPI.PSD with PSD=OS2APIC.PSD (the standard IBM supplied SMP support driver). The bootup under VB 2.0.4 still worked after this and the USB drives were still OK. However I did discover why I use PSD=ACPI.PSD at this point as my networking died. Everything looked OK but all attempts to ping another host on my own subnet via IP address failed completely. I then booted under VB 3.0.8 and the network now works again without making any other changes. Will get a screenshot of lvm later so it can be compared to the ones I previously uploaded.

comment:25 by Trevor Hemsley, 14 years ago

Oh well, it appears old age is playing tricks with my memory - the lvm screen is identical between 3.0.8 and 3.0.6.

comment:26 by Trevor Hemsley, 14 years ago

But SMP seems to work from an initial quick test!

comment:27 by Trevor Hemsley, 14 years ago

A test using a boot without the USB drive attached followed by attaching it via the menu results in lvm /rediscoverprm returning instantly but no trace of the device being found. Log from this attempt with 3.0.8 is attached.

by Trevor Hemsley, 14 years ago

Attachment: VBox.4.log added

log file from3.0.8 without USB device attached then attached via menu

comment:28 by Trevor Hemsley, 14 years ago

Ok, a second lvm /rediscoverprm now hangs on the same boot up as before. Perhaps I just hadn't given it enough time to settle or something or maybe it just needed a second kick. Still finds nothing.

comment:29 by Trevor Hemsley, 14 years ago

I've just been off reading the bug reports about the iPhone fix you talked about and have a kernel compile running now having changed the #define they talk about in those tickets to 131072. Will see if that makes any difference.

comment:30 by michaln, 14 years ago

So, did that #define make any difference? I doubt it, but would be good to know.

FWIW, the problem with the last USBOHCD.SYS is an interesting conflux of bugs. In its initialization code, the IBM OHCI driver does not respect the OHCI specification which states that all register accesses must be aligned and DWORD sized. And VirtualBox rejects such accesses completely, while actual hardware probably still delivers the expected results, at least for reads.

comment:31 by Trevor Hemsley, 14 years ago

Sorry, I didn't get a chance to try this out until today. So the only difference I made was to uninstall the 2.6.31.1 kernel that I had on there and replace it with the identical one but with the #define set to 131072. Initial results seem to be very promising, lvm /rediscoverprm returns immediately and the drive mounts and the files seem to be accessible. So far so good. I haven't had much of a chance to test much else yet.

For what it's worth, in among all that lot of experiments I did discover that I have the only working configuration of ACPI.PSD, os2krnl and associated drivers that actually allow USB hard drives to work on an OS/2 guest at all. Changing acpi.psd for os2apic.psd (and the eCS resource.sys for the older IBM resource.sys since they relate to the PSD drivers) stops USB from functioning. With IBM os2apic.psd and resource.sys then USB appears to sort of work but responses from the drives take ages and it gives the appearance that the hard disk is in perpetual error recovery. OS2APIC.PSD with the eCS resource.sys does not work at all. I have to use the SMP os2krnl to get networking to work.

comment:32 by michaln, 14 years ago

If you are absolutely certain that the maximum URB size limit is what makes the difference, we'll have to reproduce the problem and take a closer look. So far I had no luck at all on a Linux host (the guest did not appear to see any USB devices at all), and on my Windows and OS X boxes, USB in an OS/2 guest worked.

FWIW, I am using a MCP2 guest with an updated W4 kernel. I can safely state that neither networking nor USB in OS/2 requires acpi.psd, *but* I am using a more or less unmodified (just updated) IBM OS/2 install. I don't know what kind of funny stuff eCS did.

comment:33 by Trevor Hemsley, 14 years ago

I have a disk image of an ecs vm here that I could send to you to experiment with. It's not the machine that I generally use which is IBM OS/2 ACP1 plus xr_c005. This one is ecs 2.0 RC6 and the bzipped version of the ecs.vdi is about 256MB in size. Don't think you really want that attached to a ticket though! I have just tested it under VBox 2.0.4 and it definitely allows USB hard disks to be attached. Last time I tried this on VBox 3.x it did not work on that but I have not tried it under this URB modified kernel yet.

comment:34 by michaln, 14 years ago

No, you definitely shouldn't attach to this ticket :) Please try to verify that the USB support in your VM is a) broken with VBox 2.0.6, and b) works again when the kernel URB size limit is raised. If that is indeed the case, we'll provide space for you to upload.

comment:35 by Trevor Hemsley, 14 years ago

Tonight's testing.

Booted with kernel 2.6.31.1 with USB URB size 131072 and running the eCS guest I talked about. VBox 2.0.4 works Vbox 2.0.6 works VBox 3.0.8 works

Unfortunately I forgot to save the 2.6.31.1 kernel RPM that I built with the standard USB URB size so I tested with the 2.6.29.4 kernel that I was running a couple of weeks ago. This had no modifications at all. VBox 2.0.4 boots but no USB VBox 3.0.8 boots but no USB

I checked my previous log files and this was not working in VBox.4.log uploaded on the 5th Oct which shows kernel 2.6.31.1 and VBox 3.0.8 not working.

comment:36 by Trevor Hemsley, 14 years ago

[trevor@trevor4 VDI]$ cat ecs.vdi | bzip2 -c -9 > ecs.vdi.bz2 [trevor@trevor4 VDI]$ ll total 159929340 -rw-r--r-- 1 trevor trevor 576725504 Oct 22 21:20 ecs.vdi -rw-rw-r-- 1 trevor trevor 257530903 Oct 22 21:54 ecs.vdi.bz2

Will need about 256MB to upload this and on my connection that will take a while :( I take it that you'd also like the ecs.xml file that matches this?

comment:37 by michaln, 14 years ago

Wait. You said about the unmodified 2.6.29.4 kernel: "VBox 2.0.4 boots but no USB" - is that really correct?!

The XML file would help, but it's much better to use the "File/Export Appliance" menu entry in VBox 3.0.8. That will compress the disk image and store the relevant information about the VM. Please e-mail me (michaln at sun dot com) and I'll let you know where to upload the resulting file(s).

comment:38 by Trevor Hemsley, 14 years ago

It was late :( 2.0.4 works on all kernel releases.

I reinstalled 3.0.8 just to do the export and have an ecs.ovf and an ecs.vmdk file of about 290MB. Email on its way.

comment:39 by michaln, 14 years ago

Image received, thanks. I tested it on my Linux box (32-bit Fedora 9) with a USB stick but did *not* observe any problem. The VM could access the media, regardless of whether it was present at boot or inserted later.

I don't know what is different, if it's my host OS or the USB media. If you haven't already, I recommend you try different USB media; they could well behave differently.

comment:40 by michaln, 14 years ago

This problem could be related to a PCI interrupt routing mismatch that I've just been investigating in VBox 3.1.4. You could go to the OS/2 Hardware Manager and check the interrupt assigned to the EHCI device. If it's 11, this is likely the same interrupt routing bug. If it's 10, then it's something else. If the interrupt is neither 10 nor 11, I have no idea what's going on.

comment:41 by Trevor Hemsley, 14 years ago

It is indeed 11.

comment:42 by Trevor Hemsley, 14 years ago

Hi Michal, not sure if your fix made it into 3.1.6 but I tried the x86_64 version of that last night and it was still unable to see any USB hard disks. It did look like the fix was there from a reading of the changelog but sadly it appears to have made no difference to this problem. But thanks for remembering the issue and hoping that you'd fixed it :)

comment:43 by Trevor Hemsley, 14 years ago

Michal, when you said that you had "tested it on my Linux box (32-bit Fedora 9) with a USB stick but did *not* observe any problem", was this a USB stick of less than 8GB? Apparently these are treated as large floppy devices by OS/2 and devices larger than 8GB are treated as hard disks. All my tests have been performed with 2.5" USB attached hard drive cases with laptop drives installed or 3.5" cases with 3.5" drives in. All tests have been with devices between 40GB and 160GB in size.

comment:44 by Trevor Hemsley, 12 years ago

Fixed by 4.2.0 at last! Thanks.

comment:45 by Frank Mehnert, 12 years ago

Description: modified (diff)
Resolution: fixed
Status: newclosed

Thanks for the feedback! I'm not sure VirtualBox 4.2 is really the first version which fixed this problem, at least we didn't do any specific fix. But anyway, the positive result is more important.

Note: See TracTickets for help on using tickets.

© 2023 Oracle
ContactPrivacy policyTerms of Use