On 06/13/22 19:33, Richard W.M. Jones wrote:
On Mon, Jun 13, 2022 at 07:01:35PM +0200, Laszlo Ersek wrote:
Register a firstboot script, for installing the guest agent with the guest's own package manager -- that is, "Guest_packages.install_command".
For installing the package, network connectivity is required. Check it first with "nmcli" (also checking whether NetworkManager is running), then with "systemd-networkd-wait-online" (dependent on systemd-networkd). Note that NetworkManager and systemd-networkd are never supposed to be enabled at the same time.
The source domain's SELinux policy may not allow our firstboot service to execute the package's installation scripts (if any). For that reason, temporarily disable SELinux around package installation.
After installation, register another script for launching the agent.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2028764 Signed-off-by: Laszlo Ersek <lersek@redhat.com> ---
Notes: v2:
- drop the "pkg_mgmt" helper variable; replace the call to "g#inspect_get_package_management" with the already populated "inspect.i_package_management" [Rich]
- replace "sleep 60" with an open-coded "nmcli + sleep" loop, plus systemd-networkd-wait-online [Rich]:
> #!/bin/sh > if conn=$(nmcli networking connectivity); then > tries=0 > while > test $tries -lt 30 && > test full != "$conn" > do > sleep 1 > tries=$((tries + 1)) > conn=$(nmcli networking connectivity) > done > elif systemctl -q is-active systemd-networkd; then > /usr/lib/systemd/systemd-networkd-wait-online \ > -q --timeout=30 > fi
- refresh submodule checkout against now-upstream libguestfs-common commit 9e990f3e4530 ("mlcustomize: factor out pkg install/update/uninstall from guestfs-tools", 2022-06-09)
convert/convert_linux.ml | 78 +++++++++++++++++++- common | 2 +- 2 files changed, 77 insertions(+), 3 deletions(-)
diff --git a/convert/convert_linux.ml b/convert/convert_linux.ml index 2ddbc07aa86a..59d143bdda4b 100644 --- a/convert/convert_linux.ml +++ b/convert/convert_linux.ml @@ -562,8 +562,82 @@ let convert (g : G.guestfs) source inspect keep_serial_console _ = name = qga_pkg ) inspect.i_apps in if not has_qemu_guest_agent then - (* FIXME -- install qemu-guest-agent here *) - () + try + let inst_cmd = Guest_packages.install_command [qga_pkg] + inspect.i_package_management in + + (* Use only the portable filename character set in this. *) + let selinux_enforcing = "/root/virt-v2v-fb-selinux-enforcing" + and timeout = 30 in + let fbs = + Firstboot.add_firstboot_script g inspect.i_root + in + info (f_"The QEMU Guest Agent will be installed for this guest at \ + first boot."); + + (* Wait for the network to come online in the guest (best effort). + *) + fbs "wait online" + (sprintf "#!/bin/sh\n\ + if conn=$(nmcli networking connectivity); then\n\ + \ \ tries=0\n\ + \ \ while\n\ + \ \ \ \ test $tries -lt %d &&\n\ + \ \ \ \ test full != \"$conn\"\n\ + \ \ do\n\ + \ \ \ \ sleep 1\n\ + \ \ \ \ tries=$((tries + 1))\n\ + \ \ \ \ conn=$(nmcli networking connectivity)\n\ + \ \ done\n\ + elif systemctl -q is-active systemd-networkd; then\n\ + \ \ /usr/lib/systemd/systemd-networkd-wait-online \\\n\ + \ \ \ \ -q --timeout=%d\n\
I was trying to work out what the \-spaces do here. Tabs?
Nope :) The first (original) problem is that I very much dislike when we put scripts and other verbatim texts (multiline strings) to the very left in the OCaml source; it breaks up the otherwise great indentation of the OCaml code. Therefore I finish the lines in the string with \ -- subsequently, OCaml throws away (from the string) the newline character, plus *all* leading whitespace on the next line. "All whitespace" is the second problem then; I want the compiler to throw away those space chars that are meant for the OCaml identation, but not those that indent the embedded shell script! Therefore, I need to escape at least the very first space character of the shell script's indentation. In theory, the further escapes would not be necessary; so for example rather than \ \ while\n\ I should be able to write \ while\n\ But I kind of disliked this discrepancy. It seemed more consistent to say: "OCaml indentation (to be stripped): no escapes; shell script indentation (not to be stripped): escaped".
+ fi\n" timeout timeout); + + (* Disable SELinux temporarily around package installation. Refer to + * <https://bugzilla.redhat.com/show_bug.cgi?id=2028764#c7> and + * <https://bugzilla.redhat.com/show_bug.cgi?id=2028764#c8>. + *) + fbs "setenforce 0" + (sprintf "#!/bin/sh\n\ + rm -f %s\n\ + if command -v getenforce >/dev/null &&\n\ + \ \ test Enforcing = \"$(getenforce)\"\n\ + then\n\ + \ \ touch %s\n\ + \ \ setenforce 0\n\ + fi\n" selinux_enforcing selinux_enforcing); + fbs "install qga" inst_cmd; + fbs "setenforce restore" + (sprintf "#!/bin/sh\n\ + if test -f %s; then\n\ + \ \ setenforce 1\n\ + \ \ rm -f %s\n\ + fi\n" selinux_enforcing selinux_enforcing); + + (* Start the agent now and at subsequent boots. The following + * commands should work on both sysvinit distros / distro versions + * (regardless of "/etc/rc.d/" vs. "/etc/init.d/" being the scheme + * in use) and systemd distros (via redirection to systemctl). + * + * On distros where the chkconfig command is redirected to + * systemctl, the chkconfig command is likely superfluous. That's + * because on systemd distros, the QGA package comes with such + * runtime dependencies / triggers that the presence of the + * virtio-serial port named "org.qemu.guest_agent.0" automatically + * starts the agent during (second and later) boots. However, even + * on such distros, the chkconfig command should do no harm. + *) + fbs "start qga" + (sprintf "#!/bin/sh\n\ + service %s start\n\ + chkconfig %s on\n" qga_pkg qga_pkg) + with + | Guest_packages.Unknown_package_manager msg + | Guest_packages.Unimplemented_package_manager msg -> + warning (f_"The QEMU Guest Agent will not be installed. The \ + install command for package ‘%s’ could not be created: \ + %s.") qga_pkg msg
and configure_kernel () = (* Previously this function would try to install kernels, but we diff --git a/common b/common index 48527b8768d7..9e990f3e4530 160000 --- a/common +++ b/common @@ -1 +1 @@ -Subproject commit 48527b8768d7e010552beb62500a46ad940bca9a +Subproject commit 9e990f3e4530df3708d176bc50e0bc68cf07d3ff -- 2.19.1.3.g30247aa5d201
Reviewed-by: Richard W.M. Jones <rjones@redhat.com>
Thanks! Laszlo
Rich.