| 1 | Managing device boot order with bootindex properties |
| 2 | ==================================================== |
| 3 | |
| 4 | QEMU can tell QEMU-aware guest firmware (like the x86 PC BIOS) |
| 5 | which order it should look for a bootable OS on which devices. |
| 6 | A simple way to set this order is to use the ``-boot order=`` option, |
| 7 | but you can also do this more flexibly, by setting a ``bootindex`` |
| 8 | property on the individual block or net devices you specify |
| 9 | on the QEMU command line. |
| 10 | |
| 11 | The ``bootindex`` properties are used to determine the order in which |
| 12 | firmware will consider devices for booting the guest OS. If the |
| 13 | ``bootindex`` property is not set for a device, it gets the lowest |
| 14 | boot priority. There is no particular order in which devices with no |
| 15 | ``bootindex`` property set will be considered for booting, but they |
| 16 | will still be bootable. |
| 17 | |
| 18 | Some guest machine types (for instance the s390x machines) do |
| 19 | not support ``-boot order=``; on those machines you must always |
| 20 | use ``bootindex`` properties. |
| 21 | |
| 22 | There is no way to set a ``bootindex`` property if you are using |
| 23 | a short-form option like ``-hda`` or ``-cdrom``, so to use |
| 24 | ``bootindex`` properties you will need to expand out those options |
| 25 | into long-form ``-drive`` and ``-device`` option pairs. |
| 26 | |
| 27 | Example |
| 28 | ------- |
| 29 | |
| 30 | Let's assume we have a QEMU machine with two NICs (virtio, e1000) and two |
| 31 | disks (IDE, virtio): |
| 32 | |
| 33 | .. parsed-literal:: |
| 34 | |
| 35 | |qemu_system| -drive file=disk1.img,if=none,id=disk1 \\ |
| 36 | -device ide-hd,drive=disk1,bootindex=4 \\ |
| 37 | -drive file=disk2.img,if=none,id=disk2 \\ |
| 38 | -device virtio-blk-pci,drive=disk2,bootindex=3 \\ |
| 39 | -netdev type=user,id=net0 \\ |
| 40 | -device virtio-net-pci,netdev=net0,bootindex=2 \\ |
| 41 | -netdev type=user,id=net1 \\ |
| 42 | -device e1000,netdev=net1,bootindex=1 |
| 43 | |
| 44 | Given the command above, firmware should try to boot from the e1000 NIC |
| 45 | first. If this fails, it should try the virtio NIC next; if this fails |
| 46 | too, it should try the virtio disk, and then the IDE disk. |
| 47 | |
| 48 | Limitations |
| 49 | ----------- |
| 50 | |
| 51 | Some firmware has limitations on which devices can be considered for |
| 52 | booting. For instance, the x86 PC BIOS boot specification allows only one |
| 53 | disk to be bootable. If boot from disk fails for some reason, the x86 BIOS |
| 54 | won't retry booting from other disk. It can still try to boot from |
| 55 | floppy or net, though. In the case of s390x BIOS, the BIOS will try up to |
| 56 | 8 total devices, any number of which may be disks or virtio-net devices. |
| 57 | |
| 58 | Sometimes, firmware cannot map the device path QEMU wants firmware to |
| 59 | boot from to a boot method. It doesn't happen for devices the firmware |
| 60 | can natively boot from, but if firmware relies on an option ROM for |
| 61 | booting, and the same option ROM is used for booting from more then one |
| 62 | device, the firmware may not be able to ask the option ROM to boot from |
| 63 | a particular device reliably. For instance with the PC BIOS, if a SCSI HBA |
| 64 | has three bootable devices target1, target3, target5 connected to it, |
| 65 | the option ROM will have a boot method for each of them, but it is not |
| 66 | possible to map from boot method back to a specific target. This is a |
| 67 | shortcoming of the PC BIOS boot specification. |
| 68 | |
| 69 | Mixing bootindex and boot order parameters |
| 70 | ------------------------------------------ |
| 71 | |
| 72 | Note that it does not make sense to use the bootindex property together |
| 73 | with the ``-boot order=...`` (or ``-boot once=...``) parameter. The guest |
| 74 | firmware implementations normally either support the one or the other, |
| 75 | but not both parameters at the same time. Mixing them will result in |
| 76 | undefined behavior, and thus the guest firmware will likely not boot |
| 77 | from the expected devices. |