Mounting cd/dvd drive in targetcli

[devedse]

I'm currently working on setting up an iSCSI disk system to install and boot windows from. Ideally I'd simply want to mount the Windows iso disk, boot into WinPE using iPXE, and run the setup.

The only step I currently can't get working is mounting the Windows image as a CD/DVD disk.

In another iSCSI tool (TGT) I did the following:

<target iqn.2024-11.local.vm-iscsibeast:windows-installer>
        backing-store /root/iso/en-us_windows_11_business_editions_version_24h2_updated_june_2025_x64_dvd_4cf132a4.iso
</target>

This allowed the windows pe environment to detect the iSCSI disk as a CD/DVD drive.

However with targetcli I've not been able to get this to work besides manually extracting the .iso, creating an iSCSI drive, copying over the files, mounting that into WinPE. However that still leaves the disadvantage that Windows see's this as a hard disk rather then a DVD drive. (Thus allowing to select it as a potential installation target and all that stuff you don't want).

Is there some kind of way to either solve or work around this?

[gonzoleeman]

I just tested this, and it worked fine for me.

I set up a fileio backend with the ISO image as backing store. And it worked just fine.

[devedse]

What commands did you do to create that? And was it also visible in WinPE as a DVD Drive?

[gonzoleeman]

What commands? I used targetcli. I don't have the window handy (it's a virtual machine I've already shutdown). Are you familiar with targetcli?

I used the same exact commands I use when creating any iscsi target using a file backend (fileio), except that I pointed the backend at the ISO file instead of any empty (possibly sparse) file, as usual. I named my backing mysio, so it's at /backstores/fileio/myiso.

Once the /backstores/fileio/myiso back-end is created, I created a new iscsi target, set the ACL to my initiator name, and set the lun0 to point at /backstores/fileio/myiso. The portal defaults to "all interfaces" by default.

Then I just discovered and connected to it. All from Linux.

No, I didn't try this from Windows. I can't imagine what that would take, but I don't even have a PC, so there's that.

[devedse]

I used these commands to set it up. As you can see I've been playing around with all kinds of variables because I couldn't get WinPE to detect the thing as an actual CD drive. It wouldn't show it.

#!/bin/bash
# setup-windows11-iscsi.sh

ISO_PATH="/root/iso/en-us_windows_11_business_editions_version_24h2_updated_june_2025_x64_dvd_4cf132a4.iso"
TARGET_NAME="iqn.2024-11.local.vm-iscsibeast:windows-installer"
SERVER_IP=$(hostname -I | awk '{print $1}')

echo "=== Setting Up Windows 11 iSCSI Target ==="

# Check if the ISO exists
if [ ! -f "$ISO_PATH" ]; then
    echo "ERROR: Windows ISO not found at $ISO_PATH"
    echo "Please place your ISO in /root/iso/"
    exit 1
fi

# Check if the target already exists and overwrite if it does
echo "Checking for existing targets..."
if targetcli /iscsi ls 2>/dev/null | grep -q "${TARGET_NAME}"; then
    echo "Removing existing Windows installer target..."
    
    # Step-by-step removal for consistency
    echo "Step 1: Disabling target portal group..."
    targetcli /iscsi/${TARGET_NAME}/tpg1 disable 2>/dev/null || echo "  (already disabled or doesn't exist)"
    
    echo "Step 2: Removing LUNs..."
    targetcli /iscsi/${TARGET_NAME}/tpg1/luns delete 0 2>/dev/null || echo "  (LUN doesn't exist or already removed)"
    
    echo "Step 3: Removing iSCSI target..."
    targetcli /iscsi delete ${TARGET_NAME} 2>/dev/null || echo "  (target doesn't exist or already removed)"
    
    echo "✓ Previous iSCSI target removed."
fi

# Check if the backstore already exists
if targetcli /backstores/fileio ls 2>/dev/null | grep -q "win11iso"; then
    echo "Step 4: Removing existing backstore..."
    targetcli /backstores/fileio delete win11iso 2>/dev/null || echo "  (backstore doesn't exist or already removed)"
    echo "✓ Previous backstore removed."
fi

echo "Step 5: Saving configuration..."
targetcli saveconfig 2>/dev/null || echo "  (failed to save config)"

echo "Creating new iSCSI target for Windows ISO..."
targetcli <<EOF
/backstores/fileio create name=win11iso file_or_dev=${ISO_PATH} write_back=false
/backstores/fileio/win11iso set attribute emulate_model_alias=1
/backstores/fileio/win11iso set attribute emulate_write_cache=0
/backstores/fileio/win11iso set attribute emulate_fua_write=0
/backstores/fileio/win11iso set attribute emulate_fua_read=0
/backstores/fileio/win11iso set attribute emulate_tpu=0
/backstores/fileio/win11iso set attribute emulate_tpws=0
/backstores/fileio/win11iso set attribute emulate_caw=0
/backstores/fileio/win11iso set attribute emulate_3pc=0
/backstores/fileio/win11iso set attribute emulate_rest_reord=0
/backstores/fileio/win11iso set attribute is_nonrot=1
/backstores/fileio/win11iso set attribute block_size=2048
/iscsi create ${TARGET_NAME}
/iscsi/${TARGET_NAME}/tpg1/luns create /backstores/fileio/win11iso
/iscsi/${TARGET_NAME}/tpg1 set attribute authentication=0 generate_node_acls=1 demo_mode_write_protect=1
/iscsi/${TARGET_NAME}/tpg1 set parameter DefaultTime2Wait=120
/iscsi/${TARGET_NAME}/tpg1 set parameter DefaultTime2Retain=300
/iscsi/${TARGET_NAME}/tpg1 set parameter MaxOutstandingR2T=8
/iscsi/${TARGET_NAME}/tpg1 set parameter MaxRecvDataSegmentLength=262144
/iscsi/${TARGET_NAME}/tpg1 enable
saveconfig
exit
EOF

echo "Setting device type to CD-ROM..."
targetcli /backstores/fileio/win11iso set attribute emulate_pr=0
targetcli /backstores/fileio/win11iso set attribute emulate_ua_intlck_ctrl=0

echo ""
echo "✓ Windows ISO has been configured as an iSCSI target!"
echo "  Target Name: ${TARGET_NAME}"
echo "  ISO Path: ${ISO_PATH}"
echo "  Target is now available for PXE boot."
echo ""
echo "Add this to your PXE file:"
echo "iscsi:${SERVER_IP}::::${TARGET_NAME}"

[gonzoleeman]

I fear part of the problem you are having is that you are trying to mount a ISO images as a drive. I don't believe that's ever going to work.

When I tested from Linux (not Windows), I did not get a local CD/DVD drive device. I got a disk device. And when I mounted it, it was like mounting a real disk drive. Linux is able to boot form an ISO Image (via iPXE, for example), or an actual CD/DVD drive with an ISO inserted into it.

Again, I have no idea how windoze works, but in Linux I can actually insert the ISO CD/DVD media in a CD/DVD drive, and export the drive (i.e. not the ISO image). In this case, when I access it remotely using iSCSI, I get what appears to be a local CD/DVD drive, which an ISO inserted into it.

So if you wish to import a CD/DVD drive into a system, I don't believe exporting an ISO will do what you want.

For the "tgt" target, you said:

This allowed the windows pe environment to detect the iSCSI disk as a CD/DVD drive.

I'm surprised by this, since you didn't tell tgt what type of thing to export.

[devedse]

Tgt seems to have some defaults based on the file path. I believe you can also configure it manually.

[devedse]

@gonzoleeman , I did a bit more research into this with some AI thrown into the mix, and this is what it came up with why this works on TGT and not on TargetCLI (by digging through the codebases):

Explicit Device Type Specification: TGT uses the --device-type or -Y parameter to explicitly specify the device type when creating a LUN:


tgtadm --op new --mode logicalunit --tid 1 --lun 1 --device-type=cd
Device Type Constants: In TGT's scsi.h, device types are defined as:
    tgtadm --op new --mode logicalunit --tid 1 --lun 1 --device-type=cd
TYPE_DISK = 0x00 (Regular disk)
TYPE_MMC = 0x05 (CD/DVD - MultiMedia Commands)
MMC Implementation: TGT has a complete MMC (MultiMedia Commands) implementation in mmc.c that:

Sets the device type to TYPE_MMC
Implements DVD/CD-specific SCSI commands
Uses MMC backing store type
Sets device attributes like removable = 1 and readonly = 1
Handles DVD profiles (PROFILE_DVD_ROM, PROFILE_DVD_PLUS_R)
Backing Store Integration: The MMC device template forces the use of "mmc" backing store:
    /* MMC devices always use mmc backingstore */
    bst = get_backingstore_template("mmc");

[maurizio-lombardi]

Hello @devedse , I noticed your email on the target-devel mailing list, did you see the patch that I proposed here ?

To make that patch usable, targetcli needs to be patched as well... but if you just want to make a test, you could try to modify fd_get_device_type() to always return TYPE_ROM and tell me how Windows reacts?

[maurizio-lombardi]

@devedse Nevermind, I tested it myself. It works on Linux but Windows is still unhappy. Maybe adding support to the capabilities & mechanical status mode page may improve the situation, I will try that.
I just hope Windows doesn't require an extensive support to MMC because it's not a trivial amount of work...

[devedse]

@maurizio-lombardi , sorry, only saw your last comment just now. Thanks for helping me figure this thing out 😄 . If there's anything else I can test, let me know.

Just to understand your initial suggestion, I assume the easiest way of going about that would be is actually modifying the python code in the installed targetcli-fb folder?