mkinstall mapped x86_64 and x86 to their Debian names and accepted ppc64le
and ppc64el. Every other architecture, riscv64 included, was logged as
"Unknown arch" on each diskful install, although the install went on with
the name unchanged, which is right for riscv64.
Move the mapping into install_darch, which takes the Debian name from
xCAT::Utils::debian_arch and knows the architectures xCAT installs Ubuntu
on. riscv64 is one of them.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
copycd builds the riscv64 boot loader by running grub-mkimage, which
grub-common ships on every supported Ubuntu release. Nothing declared it, so a
management or service node installed without that package copies riscv64 media
and produces no loader, while DHCP keeps pointing every riscv64 node at the path
where the loader should be.
The declaration belongs to xcat-server, which carries the plugin that runs the
command, so both metapackages inherit it.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
The note shown when boot/grub2/grub2.<arch> is missing told the
administrator it comes from grub2-xcat or the EL installation media. On
Ubuntu neither is true: copycd builds the loader from the grub2 package on
the media, because the image the media carry cannot boot over the network.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
go-xcat stopped on riscv64 before it reached the package manager, so the
installer xCAT documents could not set up the management node the riscv64
packages are built for. The architecture is now accepted alongside the
others.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
archive.ubuntu.com publishes amd64 and i386 only, so a ppc64el or riscv64
node was given an apt mirror carrying no package for it and the installer
could not fetch what the minimal live media lacks. The default is now the
ports archive for those architectures, chosen from the osimage's
architecture rather than the package directory, which is whatever path the
administrator configured. site.ubuntu_apt_mirror still overrides it.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
grub2 reads its configuration as a script, so an unquoted command separator
ends the linux command and everything after it is lost. The Ubuntu
installer seed is written as ds=nocloud-net;s=<url>, so the node booted
without the seed URL and without the arguments that followed it, including
BOOTIF. The installer then found no autoinstall configuration and waited
for someone to answer its questions. A separator that is neither escaped
nor inside a quoted span is now escaped where it stands, which grub2
removes before it hands the line to the kernel. A value the caller escaped
or quoted keeps exactly the form the caller gave it.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
The Ubuntu driver table had no riscv64 entry, so genimage was handed an
empty list and built an image carrying no network module. A node whose NIC
is not built into the kernel then has no interface to fetch its root
filesystem with. The architecture now gets the same drivers the enterprise
Linux table lists for it, plus the overlay module every Ubuntu image needs.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
archive.ubuntu.com publishes amd64 and i386 only, so debootstrap could not
find a single package for a ppc64el or riscv64 netboot image and genimage
failed on every architecture except x86. The default mirror is now the
ports archive for those architectures. site.ubuntu_apt_mirror still
overrides it, for a local mirror that serves every architecture.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
riscv64 nodes have no boot loader unless one reaches /tftpboot/boot/grub2,
and nothing on an Ubuntu management node puts one there. The grub2 image
the media carry cannot serve: it holds a built-in configuration that looks
for the live filesystem, so a node that loads it drops to a grub prompt
instead of reading the configuration nodeset writes. copycd now builds a
netboot image from the grub2 package the media ship, and warns when it
cannot, because the node has no other source for one. An image already in
place is kept only when it is a whole executable image for the
architecture the firmware loads and carries the prefix this boot path
needs; one that is not is removed, so a rebuild that cannot run leaves
nodeset reporting a missing loader rather than serving an unusable one.
The media of every other architecture are untouched.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
24.04 and 26.04 had no riscv64 package list, so a diskless image or an install
for the architecture fell back to the generic list and reached debootstrap
without a kernel or the tools the boot scripts call.
The lists hold the same packages as their x86_64 counterparts. Every one of
them is published for riscv64 in noble and resolute, main or universe.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
The image carries the name service libraries of its architecture, and riscv64
matched neither the x86_64 nor the ppc64el branch. It fell through to the
generic path, which looks for lib/libnss_dns.so.2, so a riscv64 image shipped
without a resolver and the node could not resolve any name.
Ubuntu keeps them in lib/riscv64-linux-gnu, confirmed in the 24.04.4 riscv64
server filesystem.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
The riscv64 live-server image keeps its kernel at casper/vmlinux, where every
other live image keeps casper/vmlinuz, so the probe found no kernel and mkinstall
reported that the install image was missing.
Add the riscv64 candidate pair and let copycd name the architecture the media
reports. Verified against Ubuntu-Server 24.04.4 riscv64, which carries
casper/vmlinux, casper/initrd and casper/install-sources.yaml.
Signed-off-by: Vinícius Ferrão <2031761+viniciusferrao@users.noreply.github.com>
The go-xcat shell behavior tests were written as Perl harnesses, which made the shell assertions harder to read and kept shell-specific setup outside a native shell test framework.
Add BATS to the GitHub Actions dependency set, run BATS tests from the same preserved source tree as the Perl unit suite, and move the go-xcat repository checks into xCAT-test/autotest/bats.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
Both copies refused to run unless /usr/sbin/tcpdump existed. Debian and Ubuntu
install it as /usr/bin/tcpdump, so the rogue DHCP detector never ran there and
the probe reported its tcpdump check as failed.
Resolve tcpdump through PATH and the standard system directories with
CommandUtils::find_executable, run the resolved path, and match that path when
the capture process is killed at the end.
The shared Ubuntu lists serve every release and architecture without a list
of its own, and ospkgs hands the whole list to one apt-get install, so one
unknown name loses every package on it. ntp and ntpdate are gone from 26.04.
libodbc1 is the unixODBC runtime name up to 22.04. libvirt-bin is gone from
20.04 on and qemu-kvm from 22.04 on.
The shared lists keep ntp and ntpdate, so releases up to 24.04 keep the
daemon they had. 26.04 gets release lists that carry chrony. unixodbc
replaces libodbc1 on every release, and the odbcsetup postscript keeps its
runtime. The service lists add libdbd-pg-perl beside libdbd-mysql-perl, as a
service node may run the xCAT database on PostgreSQL. libvirt-daemon-system
with libvirt-clients replaces libvirt-bin from 18.04 on; 12.04, 14.04 and
16.04 keep kvm lists with the old names.
qemu-kvm was a transitional name for the emulator of the host architecture,
and no current release has one name for that. Per-architecture kvm lists
name the native one: qemu-system-x86 on x86_64, qemu-system-ppc on ppc64el,
qemu-system-misc on 24.04 riscv64, with ppc64le linked to ppc64el as the
other lists do. The shared kvm lists fall back to
qemu-system, which carries every emulator, so an architecture without a list
of its own still gets one. On 26.04 riscv64 libvirt-daemon-system depends on
qemu-kvm or qemu-system and nothing provides qemu-kvm, so apt installs that
fallback there whatever the list names. Every kvm list names qemu-utils:
ospkgs installs without recommends, and libvirt needs qemu-img for the qcow2
volumes kvm.pm creates.
The check ran on EL9 only, and only when the version carried a minor number,
so CentOS Stream was never checked. On EL10 a management node without EPEL
or CRB failed inside dnf install with a dependency error instead of the
message that names the missing repository. The CRB message proposed a CentOS
Stream repository file with signature checks disabled, on every
distribution. The probe used dnf list, which an installed copy of the probe
package satisfies with the repository disabled, and which reports a failed
query as a missing repository.
Check EL9 and EL10, with or without a minor version. Probe the enabled
repositories with repoquery for the host architecture and noarch, so a
source repository does not stand in for the binary one, and stop with the
package manager's own error when the query fails. Name the EPEL release
package of the running major version. For CRB, name crb enable from a
current epel-release, which handles Red Hat Enterprise Linux under both
subscription management and RHUI, Rocky Linux, AlmaLinux and CentOS Stream,
and the dnf config-manager command for Oracle Linux.
getcredentials answered xcat_secure_pw only for root, so a postscript
had no way to get the password of another node account from the passwd
table.
xcat_secure_pw:<user> now returns the password field of the passwd row
key=system,username=<user> when <user> is root or a sudoer named in the
postscripts or postbootscripts of the requesting node, its osimage, or
xcatdefaults. A sudoer without a row or without a password gets the
locked field "!", so the node applies the reply as is. Any other user,
an invalid user name, or a failed hash answers with an error instead of
an empty reply. The root request reads the same row as before and keeps
the error reply for a missing row.
genimage translated one architecture for debootstrap, x86_64 to amd64, and
compared against a bareword rather than a string, which only resolves because
the script does not enable strict subs.
Read the name from xCAT::Utils, which genimage already loads. Every
architecture reaches debootstrap with the name it does today.
copycd translated the architecture the Ubuntu media reports with its own
if/elsif chain, and genimage translates the same names back for debootstrap with
another one. Neither can be reused, so a new architecture has to be added to
both.
Put both directions in xCAT::Utils and have copycd read from there. The names
and the fallback do not change: media that xCAT has no name for still leave the
architecture as the media reported it.
The probe spelled out every candidate path twice inside one nested condition,
once to test it and once to assign it, so adding an architecture meant adding
another branch of the same shape. Move the candidates into a table keyed by
architecture family and walk it in order.
Same paths, same precedence, same failure behaviour: a media tree that matches
nothing leaves the caller on the "install image not found" path as before.
The boot flip in compute.subiquity.tmpl addressed port 3002. xcatd's install
monitor listens on site.xcatiport, so a cluster that moves the port loses the
flip and every node PXE-loops back into the installer. The flip now reads
site.xcatiport and keeps 3002 as the default. TABLEBLANKOKAY, because the key is
optional and a plain TABLE lookup of an absent key fails the whole template.
The flip also counted any reply as an accepted request. It now requires the
monitor's "ready" greeting before it sends "next", and "done" afterwards, so a
different service on that port is not read as a flipped node.
subiquity_nfsroot_server in debian.pm called getipaddr without a family. A
dual-stack management node answers with its IPv6 address, and casper takes
everything after the first colon in nfsroot= as the path, so the live filesystem
never mounts. It now asks for IPv4, as dhcp.pm and mknb.pm do.
The DNS setup wrote the xcatmaster name as a nameserver when getent found no
address, which is the case the step exists to prevent. It now keeps the
resolv.conf DHCP gave the live installer.
ubuntu_subiquity_boot_flip.t, debian_subiquity_boot_params.t and
ubuntu_resolvconf_ip.t fail on the parent commit and pass here.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
`makedhcp -q <node>` on Ubuntu's ISC-limited releases answers "no DHCP reservation
found" when it cannot read dhcpd.conf. The operator reads that as a node without a
reservation. An InfiniBand node also gets an answer with no hardware address.
_query_isc_static_host in dhcp.pm read the file with an -r test and dropped a failed
open. It also matched only a "hardware ethernet" line, while _add_isc_static_host
writes "hardware infiniband" for an InfiniBand node and adds a twin declaration
between the same markers.
_read_isc_conf_lines now returns the read error, _query_isc_static_host returns it to
listnode, and listnode answers the caller with an error. The parser accepts any
hardware type and keeps the first declaration of the block. The path of dhcpd.conf and
the distribution name are package variables, so a test can drive the query and
listnode.
dhcp_isc_static_host_query.t covers the InfiniBand address, the twin declaration, the
unreadable file and the listnode answer. It fails without this change.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
xCAT::NTP::Backend->available reported chrony as available on chronyd alone, while makentp
configured chrony only where systemctl was present too. On a host with chronyd and no systemctl
the selector returned chrony with no downgrade, makentp fell through to the ntpd path, and the
admin saw either a silent switch or "Please make sure ntpd is installed".
available now requires chronyd and systemctl for chrony, so the selector answers on the same
terms makentp acts on, and makentp branches on the name alone. choose therefore downgrades to
ntpd, or reports install, in the case it used to pass over. A commands argument injects the
command probe, in the same shape as the existing available argument.
ntp_backend_selection.t covers both commands. Six of its assertions fail without this change.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
The comments around the install monitor respawn retell the failure, defend the
design and repeat the same causal chain in three places. Reduce them to the
facts that are not visible at the site: the ordering rules, why there is no
attempt limit, and what each fork site inherits. The rest is in the commit
messages and the PR.
Comment only. RespawnUtils.pm loses 26 lines and no code changes; xcatd loses
comment lines only. Both unit test files still pass.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
The respawn is forked from the middle of the service loop, so the child
inherits @pendingconnections -- the client sockets the parent has accepted and
not yet handed to a worker. The monitor never serves one, and it outlives the
worker that does, so its copy keeps that client's socket open until the daemon
exits.
Close them in the child, next to the listener and the rescanplugins channel it
already drops.
xcatd_install_monitor.t runs the lifted respawn block against stand-in
descriptors and requires every pending connection to be closed. It fails
without this change.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
The install monitor is forked while generic_reaper is the SIGCHLD handler.
ssl_reaper is only installed once the main service loop starts, and
generic_reaper comes back whenever connections are throttled.
Only ssl_reaper cleared $pid_MON. A death reaped by generic_reaper left
$pid_MON holding a dead pid, and the service loop re-forks only when $pid_MON
is clear, so xcatiport stayed dead for the life of the daemon.
Move that accounting into reap_install_monitor and call it from both reapers.
xcatd_install_monitor.t runs both reapers over a dead child and requires each
to clear $pid_MON and fold the death into the pacing. The generic_reaper case
fails without this change.
Signed-off-by: Daniel Hilst <392820+dhilst@users.noreply.github.com>
Vinícius Ferrão
Vinícius Ferrão
Daniel Hilst