On 9/13/19 12:25 PM, Kashyap Chamarthy wrote:
On Mon, Aug 05, 2019 at 06:14:20PM +0200, Michal Privoznik wrote:
[...]
Michal Prívozník (5):
virfirmware: Expose and define autoptr for virFirmwareFree
qemu_firmware: Document qemuFirmwareGetSupported
qemu_firmware: Extend qemuFirmwareGetSupported to return FW paths
qemufirmwaretest: Test FW path getting through
qemuFirmwareGetSupported()
qemu: Use FW descriptors to report FW image paths
[...]
Tested-by: Kashyap Chamarthy <kchamart@xxxxxxxxxx>
Thanks, but I've pushed it yesterday and thus I can no longer add you to
the commit messages. Sorry.
I've just tested this patchset on Fedora 30. (I too can reproduce the
behaviour Cole saw - duplicate 'secboot' binaries.)
Build libvirt with this:
$> git describe
v5.7.0-107-gb6e6d35f3f
Stop the system libvirt daemons:
$> systemctl stop libvirtd virtlockd virtlogd
Start the daemons built from Git:
$> sudo ./run src/virtlockd &
$> sudo ./run src/virtlogd &
$> sudo ./run src/libvirtd &
Make sure your EDK2/OVMF RPM has the 'secboot' binaries/VARS files:
$> rpm -q edk2-ovmf
edk2-ovmf-20190501stable-3.fc30.noarch
$> rpm -ql edk2-ovmf | grep secboot
/usr/share/OVMF/OVMF_CODE.secboot.fd
/usr/share/OVMF/OVMF_VARS.secboot.fd
/usr/share/edk2/ovmf/OVMF_CODE.secboot.fd
/usr/share/edk2/ovmf/OVMF_VARS.secboot.fd
(The top two files are a symlink to the bottom two.)
Before invoking domCapabilities API, ensure the relevant firmware
descriptor files for x86_64 have the secboot binary listed:
$> grep CODE.secboot /usr/share/qemu/firmware/40-edk2-ovmf-x64-sb-enrolled.json /usr/share/qemu/firmware/50-edk2-ovmf-x64-sb.json
/usr/share/qemu/firmware/40-edk2-ovmf-x64-sb-enrolled.json: "filename": "/usr/share/edk2/ovmf/OVMF_CODE.secboot.fd",
/usr/share/qemu/firmware/50-edk2-ovmf-x64-sb.json: "filename": "/usr/share/edk2/ovmf/OVMF_CODE.secboot.fd",
Yeah, I've posted a patch for that:
https://www.redhat.com/archives/libvir-list/2019-September/msg00485.html
The reason is that (I suspect) 40-edk2-ovmf-x64-sb-enrolled.json and
50-edk2-ovmf-x64-sb.json have the same _CODE but different _VARS.
Therefore, in qemuFirmwareGetSupported() (which is called when
constructing domcaps) we effectivelly see two different firmwares:
https://libvirt.org/git/?p=libvirt.git;a=blob;f=src/qemu/qemu_firmware.c;h=f316a26a5ba5c14d993fa8b96df32eba3c1f0997;hb=HEAD#l1536
Therefore, they are appended onto @fws list which is then passed to
virQEMUCapsFillDomainLoaderCaps() in virQEMUCapsFillDomainOSCaps() which
doesn't deduplicate them.
But the same thing can happen even without FW descriptors: ./configure
--with-loader-nvram="/dev/null:X:/dev/null:Y" (a very dumb config, but
serves the point) which declares [{CODE = "/dev/null", NVRAM = "X"},
{CODE = "/dev/null", NVRAM "Y"}] pair. In both cases the FW image is
the same and therefore it would be printed twice. To test this with
latest git just revert e9d51a221c1871 because now libvirt picks FW
descriptors and thus configure arg (or correspondinf qemu.conf knob) is
ignrored.
Michal
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list
