[PATCH v2] virtmanager: Fix legacy firmware installation error
Cole Robinson
crobinso at redhat.com
Tue Apr 6 22:03:56 UTC 2021
On 3/10/21 4:02 PM, Charles Arnold wrote:
> When we choose /usr/share/qemu/bios.bin or /usr/share/qemu/bios-256k.bin
> in Customize configuration before install, show this error message:
>
> Unable to complete install: 'operation failed: unable to find any master var store for loader: /usr/share/qemu/bios.bin'
>
> Traceback (most recent call last):
> File "/usr/share/virt-manager/virtManager/asyncjob.py", line 65, in cb_wrapper
> callback(asyncjob, *args, **kwargs)
> File "/usr/share/virt-manager/virtManager/createvm.py", line 2081, in _do_async_install
> installer.start_install(guest, meter=meter)
> File "/usr/share/virt-manager/virtinst/install/installer.py", line 721, in start_install
> domain = self._create_guest(
> File "/usr/share/virt-manager/virtinst/install/installer.py", line 669, in _create_guest
> domain = self.conn.createXML(install_xml or final_xml, 0)
> File "/usr/lib64/python3.8/site-packages/libvirt.py", line 4347, in createXML
> raise libvirtError('virDomainCreateXML() failed')
> libvirt.libvirtError: operation failed: unable to find any master var store for loader: /usr/share/qemu/bios.bin
>
> This patch checks the loader for bios.bin or bios-256k.bin and prevents them
> from begin presented as UEFI files (type pflash) but instead as type 'rom'.
>
Sorry for the late response. From your domcapabilities mail I guess suse
is shipping firmware.json files for bios*.bin? I guess that's acceptable
but it's a little strange for historical usage, no one should really be
manually choosing between bios.bin and bios-256k.bin manually for
example. Anyways...
> Signed-off-by: Charles Arnold <carnold at suse.com>
> Signed-off-by: Yu-Chen Cho <acho at suse.com>
>
> ---
> virtManager/object/domain.py | 18 +++++++++---------
> virtinst/guest.py | 13 +++++++++++++
> 2 files changed, 22 insertions(+), 9 deletions(-)
>
> diff --git a/virtManager/object/domain.py b/virtManager/object/domain.py
> index cc2f506d..21841ada 100644
> --- a/virtManager/object/domain.py
> +++ b/virtManager/object/domain.py
> @@ -671,16 +671,16 @@ class vmmDomain(vmmLibvirtObject):
>
> if loader != _SENTINEL:
> if loader is None:
> - # Implies seabios, aka the default, so clear everything
> - guest.os.loader = None
> - guest.os.loader_ro = None
> - guest.os.loader_type = None
> - guest.os.nvram = None
> - guest.os.nvram_template = None
> + # Implies the default, so clear everything
> + guest.set_legacy_path(None)
> else:
> - # Implies UEFI
> - guest.set_uefi_path(loader)
> - guest.disable_hyperv_for_uefi()
> + if "bios.bin" in loader or "bios-256k.bin" in loader:
> + # Implies Custom: seabios, not UEFI
> + guest.set_legacy_path(loader)
> + else:
> + # Implies UEFI
> + guest.set_uefi_path(loader)
> + guest.disable_hyperv_for_uefi()
>
> if nvram != _SENTINEL:
> guest.os.nvram = nvram
> diff --git a/virtinst/guest.py b/virtinst/guest.py
> index 34da3e8c..52f49252 100644
> --- a/virtinst/guest.py
> +++ b/virtinst/guest.py
> @@ -566,6 +566,19 @@ class Guest(XMLBuilder):
> self.os.machine)
> self.os.machine = "q35"
>
> + def set_legacy_path(self, path):
> + """
> + Configure legacy BIOS for QEMU.
> + """
> + self.os.loader = path
> + if (path):
> + self.os.loader_ro = True
> + else:
> + self.os.loader_ro = None
> + self.os.loader_type = None
> + self.os.nvram = None
> + self.os.nvram_template = None
> +
This still has the same problem I pointed out in previous mails, that
`pytest --cov` now regresses. Anything added to virtinst/ needs unit
test coverage.
I think the simpler thing to do here is extend the 'if loader is None'
in details.py to handle bios*.bin too. Then the logic stays self
contained in virt-manager which is the only one that needs this code
path so far, and the change will be smaller.
Thanks,
Cole
More information about the virt-tools-list
mailing list