Chapter 8. Bug fixes
This part describes bugs fixed in Red Hat Enterprise Linux 9.2 that have a significant impact on users.
8.1. Installer and image creation
The installer now displays correct total disk space in Custom partitioning with multipath or DDF RAID devices
Previously, when Custom partitioning was selected in Installer on a system with a multipath or DDF RAID device, the total disk space was not reported correctly and member disk devices were listed as available for partitioning.
With this update, the Custom partitioning in Installer reports correct value for total disk space and only allows to use the DDF RAID or multipath device as a whole.
The installer now adds configuration options correctly into the yum repo files
Previously, the installer did not add configuration options correctly into yum repo files while including and excluding packages from additional installation repositories. With this update, yum repo files are created correctly. As a result, using the
--includepkgs= options in the
repo kickstart command now excludes or includes the specified packages during installation as expected.
filename DHCP option no longer blocks downloading the
kickstart file for installation
Previously, when building a path for getting the kickstart file from an NFS server, the installer did not consider the
filename DHCP option. As a consequence, the installer did not download the kickstart file and was blocking the installation process. With this update, the
filename DHCP option correctly constructs a path to the kickstart file. As a result, the kickstart file is downloaded properly, and the installation process starts correctly.
The installer now creates a new GPT disk layout while custom partitioning
Previously, the installer did not change the disk layout to GPT when
inst.gpt was specified on the kernel command line, and the user removed all partitions from a disk with the MBR disk layout on the custom partitioning spoke. As a consequence, the MBR disk layout remained on the disk.
With this update, the installer creates a new GPT disk layout on the disk if
inst.gpt is specified on the kernel command line, and all partitions are removed from a disk on the custom partitioning spoke.
Installer now lists all
PPC PreP Boot or
BIOS Boot partitions during custom partitioning
Previously, when adding multiple
PPC PreP Boot or
BIOS Boot partitions during custom partitioning, the Custom Partitioning screen displayed only one partition of a related type. As a consequence, the Custom Partitioning screen did not reflect the real state of the intended partitioning layout, making the partitioning process difficult and non-transparent.
With this update, the Custom Partitioning screen correctly displays all
PPC PreP Boot or
BIOS Boot partitions in the partitions list. As a result, users can now better understand and manage the intended partitioning layout.
8.2. Subscription management
Subscription manager no longer denies registration and fetching of Red Hat content
subscription-manager operated in container mode when run under OpenShift Container Platform (OCP) because of improved container detection logic in RHEL 9. As a consequence, the system was unable to use the provided subscription credentials and therefore not fetching Red Hat content.
This update fixed the container detection logic so that
subscription-manager running under OCP does not detect the system (that is the running pod) as a container. As a result, you can now use the provided subscription credentials or even register using your own credentials to fetch Red Hat content from an OpenShift container.
subscription-manager no longer retains nonessential text in the terminal
Starting with RHEL 9.1,
subscription-manager displays progress information while processing any operation. Previously, for some languages, typically non-Latin, progress messages did not clean up after the operation finished. With this update, all the messages are cleaned up properly when the operation finishes.
If you have disabled the progress messages before, you can re-enable them by entering the following command:
# subscription-manager config --rhsm.progress_messages=1
8.3. Software management
RPM no longer hangs during a transaction involving the
fapolicyd service restart
Previously, if you tried to update a package that caused the
fapolicyd service to be restarted, for example,
systemd, the RPM transaction stopped responding because the
fapolicyd plug-in failed to communicate with the
With this update, the
fapolicyd plug-in now correctly communicates with the
fapolicyd daemon. As a result, RPM no longer hangs during a transaction which involves the
fapolicyd service restart.
Reverting a DNF upgrade transaction is now possible for a package group or environment
dnf history rollback command failed when attempting to revert an upgrade transaction for a package group or an environment.
With this update, the issue has been fixed, and you can now revert the DNF upgrade transaction for a package group or environment.
Security DNF upgrade is now possible for packages that change their architecture through the upgrade
Patch for BZ#2108969 introduced with RHBA-2022:8295 caused a regression where DNF upgrade using security filters skipped packages that changed their architecture from or to
noarch through the upgrade. Consequently, the missing security upgrades for these packages could leave the system in a vulnerable state.
With this update, the issue has been fixed, and security DNF upgrade no longer skips packages that change architecture from or to
Qt message QM files with 3-letter names are now packaged when an RPM package is being built or rebuilt
find-lang.sh script could not find Qt message QM files (
.qm) with names consisting of 3 characters. Consequently, these files were not added to an RPM package.
With this update, the issue has been fixed, and the 3-letter Qt message QM files can now be packaged when building or rebuilding an RPM.
8.4. Shells and command-line tools
ReaR handles excluded DASDs on the IBM Z architecture correctly
Previously on the IBM Z architecture, ReaR reformatted all connected Direct Access Storage Devices (DASD) during the recovery process, including those DASDs that users excluded from the saved layout and did not intend to restore their content. As a consequence, if you excluded some DASDs from the saved layout, their data were lost during system recovery. With this update, ReaR no longer formats excluded DASDs during system recovery, including the device from which the ReaR rescue system was booted (using the zIPL bootloader). You are also prompted to confirm the DASD formatting script before ReaR reformats DASDs. This ensures that the data on excluded DASDs survive a system recovery.
ReaR no longer fails to restore non-LVM XFS filesystems
Previously, when you used ReaR to restore a non-LVM XFS filesystems with certain settings and disk mapping, ReaR created the file system with the default settings instead of the specified settings. For example, if you had a file system with the
swidth parameters set to non-zero values and you restored the file system using ReaR with disk mapping, the file system would be created with default
swidth parameters ignoring the specified values. As a consequence, ReaR failed during mounting the filesystem with specific XFS options. With this update, ReaR correctly restores the file system with the specified settings.
wsmancli handles HTTP 401 Unauthorized statuses correctly
wsmancli utility for managing systems using Web Services Management protocol now handles authentication to better conform to RFC 2616.
Previously, when connecting to a service that requires authentication, the
wsmancli command returned the error message
Authentication failed, please retry immediately after receiving an HTTP 401 Unauthorized response, for example, because of incomplete credentials. To proceed,
wsmancli prompted you to provide both the username and the password, even in situations where you had already provided a part of your credentials.
With this update,
wsmancli requires only credentials that were not previously provided. As a result, the first authentication attempt does not display any error message. An error message is displayed only after you provide the complete credentials and authentication fails.
USBGuard saves rules even if RuleFile is not defined
Previously, if the
RuleFile configuration directive in USBGuard was set but
RuleFolder was not, the rule set could not be changed. With this update, you can now change the rule set even if RuleFolder is set but RuleFile is not. As a result, you can modify the permanent policy in USBGuard to permanently save newly added rules.
python-sqlalchemy rebased to 1.4.45
python-sqlalchemy package has been rebased to version 1.4.45, which provides many bug fixes over version 1.4.37. Most notably, this version contains a fix for a critical memory bug in the cache key generation.
crypto-policies now disable NSEC3DSA for BIND
Previously, the system-wide cryptographic policies did not control the NSEC3DSA algorithm in the BIND configuration. Consequently, NSEC3DSA, which does not meet current security requirements, was not disabled on DNS servers. With this update, all cryptographic policies disable NSEC3DSA in the BIND configuration by default.
SECLEVEL=3 now works with PSK cipher suites
Previously, pre-shared key (PSK) cipher suites were not recognized as performing perfect forward secrecy (PFS) key exchange methods. As a consequence, the
DHE-PSK cipher suites did not work with OpenSSL configured to
SECLEVEL=3, for example, when the system-wide cryptographic policy was set to
FUTURE. The new version of the
openssl package fixes this problem.
Clevis now correctly skips commented-out devices in
Previously, Clevis tried to unlock commented-out devices in the
crypttab file, causing the
clevis-luks-askpass service to run even if the device was not valid. This caused unnecessary service runs and made it difficult to troubleshoot.
With this fix, Clevis ignores commented-out devices. Now, if an invalid device is commented out, Clevis does not attempt to unlock it, and
clevis-luks-askpass finishes appropriately. This makes it easier to troubleshoot and reduces unnecessary service runs.
Clevis no longer requests too much entropy from
pwmake password generation utility displayed unwanted warnings when Clevis used
pwmake to create passwords for storing data in
LUKS metadata, which caused Clevis to use lower entropy. With this update, Clevis is limited to 256 entropy bits provided to
pwmake, which eliminates an unwanted warning and uses the correct amount of entropy.
USBGuard no longer causes a confusing warning
Previously, a race condition could happen in USBGuard when a parent process finished sooner than the first child process. As a consequence,
systemd reported that a process was present with a wrongly identified parent PID (PPID). With this update, a parent process waits for the first child process to finish in working mode. As a result,
systemd no longer reports such warnings.
OOM killer no longer terminates
usbguard.service file did not contain a definition of the
OOMScoreAdjust option for the
systemd service. Consequently, when the system was low on resources, the
usbguard-daemon process could be terminated before other unprivileged processes. With this update,
usbguard.service file now includes
OOMScoreAdjust setting, which prevents the Out-of-Memory (OOM) killer terminate the
usbguard-daemon process prematurely.
logrotate no longer incorrectly signals Rsyslog in log rotation
Previously, the argument order was incorrectly set in the
logrotate script, which caused a syntax error. This resulted in
logrotate not correctly signaling Rsyslog during log rotation.
With this update, the order of the arguments in
logrotate is fixed and
logrotate signals Rsyslog correctly after log rotation even when the
POSIXLY_CORRECT environment variable is set.
imklog no longer calls
free() on missing objects
imklog module called a
free() function on an already freed object. Consequently,
imklog could cause a segmentation fault. With this update, the object is no longer freed twice.
fagenrules --load now works correctly
fapolicyd service did not correctly handle the signal hang up (SIGHUP). Consequently,
fapolicyd terminated after receiving SIGHUP, and the
fagenrules --load command did not work correctly. This update contains a fix for the problem. As a result,
fagenrules --load now works correctly, and rule updates no longer require manual restarts of
Scans and remediations correctly ignore SCAP Audit rules Audit key
Previously, Audit watch rules that were defined without an Audit key (
-F key) encountered the following problems:
- The rules were marked as non-compliant even if other parts of the rule were correct.
- Bash remediation fixed the path and permissions of the watch rule, but it did not add the Audit key correctly.
Remediation sometimes did not fix the missing key, returning an
errorinstead of a
This affected the following rules:
With this update, the Audit key has been removed from checks and from Bash and Ansible remediations. As a result, inconsistencies caused by the key field during checking and remediating no longer occur, and auditors can choose these keys arbitrarily to make searching Audit logs easier.
Keylime no longer fails attestation of systems that access multiple IMA-measured files
Previously, if a system that runs the Keylime agent accessed multiple files measured by the Integrity Measurement Architecture (IMA) in quick succession, the Keylime verifier incorrectly processed the IMA log additions. As a consequence, the running hash did not match the correct Platform Configuration Register (PCR) state, and the system failed attestation. This update fixes the problem and systems that quickly access multiple measured files no longer fail attestation.
Keylime policy generation script no longer causes a segmentation fault and core dump
create_mb_refstate script generates policies for measured boot attestation in Keylime. Previously,
create_mb_refstate incorrectly calculated the data length in the
DevicePath field. As a consequence, the script tried to access invalid memory using the incorrectly calculated length, which resulted in a segmentation fault and core dump.
This update, which has been published in advisory RHBA-2022:105318-02, prevents the segmentation fault when processing the measured boot event log. As a consequence, you can generate a measured boot policy.
TPM certificates no longer cause Keylime registrar to crash
Previously, some certificates in the Keylime TPM certificate store were malformed x509 certificates and caused the Keylime registrar to crash. This update fixes the problem, and Keylime registrar no longer crashes due to malformed ceritficates.
NetworkManager now preserves IP addresses during reapply before acquiring a new DHCP lease
Previously, after changing the connection settings and then using
nmcli device reapply command, NetworkManager did not preserve the DHCP lease. Consequently, the IP address got removed temporarily. With this fix, NetworkManager preserves the DHCP lease and uses it until the lease expires or the client requests a new one. As a result, when the
nmcli device reapply command restarts DHCP client, it does not temporarily remove the IP address.
firewalld service now triggers the
ipset deprecation warning only when using direct rules
firewalld service used the deprecated
ipset kernel module when it was not necessary. Consequently, RHEL logged the module’s deprecation warning which could be misleading because the
ipset feature of
firewalld is not deprecated. With this update,
firewalld only uses the deprecated
ipset module and logs the warning if the user explicitly uses
ipsets with the
HNV interface now displays the options after reboot
nmcli utility created a Hybrid Network Virtualization (HNV) bond by using NetworkManager API. Consequently, after a reboot, the HNV bond lost the primary port setting. With this fix,
nmcli now uses
hcnmgr to set bonding options for the primary port. The
hcnmgr utility supports migration of live partitions with Single Root Input/Output Virtualization (SR-IOV) for hybrid networks. As a result, the HNV bond interface displays the
active slave/primary_reselect option after reboot.
FADump enabled with Secure Boot works correctly
Previously, when Firmware Assisted Dump (FADump) was enabled in the Secure Boot environment and any of the booting components exceeded the allocated memory region, system reboots caused a GRUB Out of Memory (OOM) state. This update provides a fix in
kexec-tools so that Secure Boot and FADump work together correctly.
8.8. Boot loader
grubby now passes arguments to a new kernel correctly
When you add a new kernel using the
grubby tool and do not specify any arguments, or leave the arguments blank,
grubby will not pass any arguments to the new kernel and
root will not be set. Using the
--copy-default options ensures new arguments are appended to the default arguments.
8.9. File systems and storage
Installer creating LUKSv2 devices with sector size of 512 bytes
Previously, the RHEL installer created LUKSv2 devices with 4096 bytes sectors if the disk had 4096 bytes physical sectors. With this update, installer now creates LUKSv2 devices with sector size of 512 bytes to offer better disk compatibility with different physical sector sizes used together in one LVM volume group even when the LVM physical volumes are encrypted.
supported_speeds sysfs attribute reports correct speed values
Previously, because of an incorrect definition in the
qla2xxx driver, the
supported_speeds sysfs attribute for the HBA reported 20 Gb/s speed instead of the expected 64 Gb/s speed. Consequently, if the HBA supported 64 Gb/s link speed, the
supported_speeds sysfs value was incorrect, which affected the reported speed value.
With this update, the
supported_speeds sysfs attribute for HBA reports the correct speed values, which are 16 Gb/s, 32 Gb/s, and 64 Gb/s. You can view the speed values by executing the
cat /sys/class/fc_host/host*/supported_speeds command.
8.10. High availability and clusters
pcs no longer allows you to modify cluster properties that should not be changed
pcs command line interface allowed you to modify cluster properties that should not be changed or for which change does not take effect. With this fix,
pcs no longer allows you to modify these cluster properties:
pcs now displays cluster properties that are not explicitly configured
pcs command to display the value of a specific cluster property did not list values that are not explicitly configured in the CIB. With this fix, if a cluster property is not set
pcs displays the default value for the property.
Cluster resources that call
crm_mon now stop cleanly at shutdown
crm_mon utility returned a nonzero exit status while Pacemaker was in the process of shutting down. Resource agents that called
crm_mon in their monitor action, such as
ocf:heartbeat:pqsql, could incorrectly return a failure at cluster shutdown. With this fix,
crm_mon returns success even if the cluster is in the process of shutting down. Resources that call
crm_mon now stop cleanly at cluster shutdown.
OCF resource agent metadata actions can now call
crm_node without causing unexpected fencing
As of RHEL 8.5, OCF resource agent metadata actions blocked the controller and
crm_node queries performed controller requests. As a result, if an agent’s metadata action called
crm_node, it blocked the controller for 30 seconds until the action timed out. This could cause other actions to fail and the node to be fenced.
With this fix, the controller now performs metadata actions asynchronously. An OCF resource agent metadata action can now call
crm_node without issue.
Pacemaker now rechecks resource assignments immediately when resource order changes
As of RHEL 8.7, Pacemaker did not recheck resource assignments when the order of resources in the CIB changed with no changes to the resource definition. If configuration reordering would cause resources to move, that would not take place until the next natural transition, up to the value of
cluster-recheck-interval-property. This could cause issues if resource stickiness is not configured for a resource.
With this change, Pacemaker rechecks resource assignments when the order of the resources in the CIB changes, as it did for earlier Pacemaker releases. The cluster now responds immediately to these changes, if needed.
Enabling a single resource and monitoring operation no longer enables monitoring operations for all resources in a resource group
Previously, after unmanaging all resources and monitoring operations in a resource group, managing one of the resources in that group along with its monitoring operation re-enabled the monitoring operations for all resources in the resource group. This could trigger unexpected cluster behavior.
With this fix, managing a resource and re-enabling its monitoring operation re-enables the monitoring operation for that resource only and not for the other resources in a resource group.
8.11. Compilers and development tools
DNS lookup can now succeed even when some CNAME records are invalid
glibc DNS stub resolver treated CNAME records with owner names that are not host names as DNS packet errors. Consequently, the DNS query failed because of the DNS packet errors. With this update, the
glibc stub resolver now skips invalid CNAME records and the corresponding alias information is not extracted. Therefore, DNS lookups can now succeed even if the server response includes a CNAME chain that contains a domain name that is not a host name.
golang now supports 4096 bit keys in x509 FIPS mode
golang did not support the 4096 bit keys in x509 FIPS mode. Consequently, when the user used 4096 bit keys the program crashed. With this update,
golang now supports 4096 bit keys in x509 FIPS mode.
You can install SciPy using
pip on all architectures
openblas-devel package did not contain a pkg-config file for the OpenBLAS library. As a consequence, in certain scenarios, it was impossible to determine the compiler and linker flags using the
pkgconf utility while compiling with OpenBLAS. For example, this caused a failure of the
pip install scipy command on the 64-bit IBM Z and IBM Power Systems, Little Endian architectures.
This update adds the
openblas.pc file to the
openblas-devel package on all supported architectures. As a result, you can install the SciPy library using the
pip package installer.
Note that in RHEL 9, it is recommended to build your applications against the
flexiblas-devel package and link your projects to the FlexiBLAS wrapper library.
tzset function in
glibc now sets the daylight variable to a non-zero value if there is any DST rule in the TZ data
tzset function in
glibc would set the daylight variable to 0 if the last DST transition in the time zone data file did not result in a clock change due to a simultaneous change in the standard time offset. Consequently, when applications use the daylight variable to check if DST was ever active, they do not get the right result and perform incorrect actions based on this information. To fix this, the
tzset function now sets the daylight variable to a non-zero value if there is any DST rule in the time zone data, regardless of offset. As a result, applications now observe the presence of DST rules regardless of offset changes.
OpenJDK RSAPSSSignature implementation now validates RSA keys before using them
Previously, the RSAPSSSignature implementation in OpenJDK did not fully check if given RSA keys could be used by the SunRSASign provider before attempting to use them, which would result in errors when using custom security providers. The bug is now fixed and, as a result, the RSAPSSSignature implementation now validates RSA keys and allows other providers to handle these keys when it cannot.
The OpenJDK XML signature provider is now functional in FIPS mode
Previously, the OpenJDK XML signature provider was unable to operate in FIPS mode. As a result of enhancements to FIPS mode support the OpenJDK XML signature provider is now enabled in FIPS mode.
OpenJDK in FIPS mode no longer experiences unexpected errors with certain PKCS#11 tokens
Previously, some PKCS#11 tokens were not fully initialized before use by OpenJDK in FIPS mode resulting in unexpected errors. With this upgrade, these errors are now expected and handled by the FIPS support code.
8.12. Identity Management
Authentication to external IdPs that require a client secret is now possible
Previously, SSSD did not properly pass client secrets to external identity providers (IdPs). Consequently, authentication failed against external IdPs that you previously configured with the
ipa idp-add --secret command to require a client secret. With this update, SSSD passes the client secret to the IdP and users can authenticate.
IdM now supports setting hostmasks for
sudo rules using Ansible
ipa sudorule-add-host command allowed setting a hostmask to be used by the
sudo rule, but this option was not present in the
ansible-freeipa package. With this update, you can now use the
hostmask variable to define a list of hostmasks to which a particular
sudo rule, defined in Identity Management (IdM), applies.
As a result, you can now automate setting host masks for IdM
sudo rules with Ansible.
dscreate utility now works correctly when it uses a custom path with the
Previously, an instance that used custom directory paths failed to start because the custom directories had a wrong SELinux label. As a consequence, SELinux denied access to these directories and the instance was not created. With this release,
dscreate utility sets correct SELinux labels for the custom instance directories.
A password change for the Directory Server replication manager account now works correctly
Previously, after a password change, Directory Server did not properly update the password cache for the replication agreement. As a consequence, when you changed the password for the replication manager account, the replication failed. With this update, Directory Server updates the cache properly and, as a result, the replication works as expected.
The IdM client installer no longer specifies the TLS CA configuration in the
Previously, the IdM client installer specified the TLS CA configuration in the
ldap.conf file. With this update, OpenLDAP uses the default trust store and the IdM client installer does not set up the TLS CA configuration in the
8.13. Red Hat Enterprise Linux System Roles
nbde_client System Role now correctly handles different names of
nbde_client System Role has been updated to handle the systems on which the
systemd unit has a different name. The role now correctly works with different names of
clevis-luks-askpass on managed nodes, which requires unlocking also LUKS-encrypted volumes that mount late in the boot process.
ha_cluster System Role logs no longer display unencrypted passwords and secrets
ha_cluster System Role accepts parameters that can be passwords or other secrets. Previously, some of the tasks would log their inputs and outputs. As a result, the role logs could contain unencrypted passwords and other secrets.
With this update, the tasks have been changed to use the Ansible
no_log: true directive and the task output is no longer displayed in the role logs. The
ha_cluster System Role logs no longer contain passwords and other secrets. While this update protects secure information, the role logs now provide less information that you can use when debugging your configuration.
Clusters configured with
ha_cluster System Role to use SBD and not start on boot now work correctly
Previously, if a user configured a cluster using the
ha_cluster System Role to use SBD and not start on boot, then the SBD service was disabled and SBD did not start. With this fix, the SBD service is always enabled if a cluster is set to use SBD whether or not the cluster is configured to start on boot.
Enabling implicit files provider to fix
cockpit-session-recording SSSD configuration
A disabled SSSD implicit files provider caused the
cockpit-session-recording modules to create an invalid System Security Services Daemon (SSSD) configuration. This update unconditionally enables the files provider and as a result, the SSSD configuration created by
cockpit-session-recording now works as expected.
nbde_client_clevis role no longer reports traceback to users
nbde_client_clevis role sometimes failed in exception, causing a traceback and reporting sensitive data, such as the
encryption_password field, back to the user. With this update, the role no longer reports sensitive data, only the appropriate error messages.
stonith-watchdog-timeout property with the
ha_cluster System Role now works in a stopped cluster
Previously, when you set the
stonith-watchdog-timeout property with the
ha_cluster System Role in a stopped cluster, the property reverted to its previous value and the role failed. With this fix, configuring the
stonith-watchdog-timeout property by using the
ha_cluster System Role works properly.
Network traffic is now directed through the intended network interface when using
initscripts with the
networking RHEL System Role
Previously, when using the
initscripts provider, the routing configuration for network connections did not specify the output device that the traffic should go through. Consequently, the kernel could use a different output device than the user intended. Now, if the network interface name is specified in the playbook for the connection, it is used as the output device in the route configuration file. This aligns the behavior with NetworkManager, which configures the output device in routes when activating profiles on devices. As a result, the users can ensure that the traffic is directed through the intended network interface.
selinux role now now manages policy modules idempotently
selinux role copied an existing module to the managed node every time, reporting a change even when the module was already present. With this update, the
selinux role checks if the module has been installed on the managed node, and does not attempt to copy and install the module if it is already installed.
System time on nested VMs now works reliably
Previously, system time on nested virtual machines (VMs) in some cases desynchronised from the Level 0 and level 1 hosts. This also sometimes caused the nested VM to become unresponsive or terminate unexpectedly.
With this update, the time handling code in the KVM host kernel code has been fixed, which prevents the described errors from occurring.
VMs on IBM Z no longer fail to start when using
memfd memory backing
Previously, on IBM Z hosts, virtual machines (VMs) failed to boot if they were configured to use the
memfd type of hugepage memory backing, for example as follows:
<memoryBacking> <hugepages/> <source type='memfd'/> </memoryBacking>
With this update, the underlying cause has been fixed, and the affected VMs now start correctly.
VNC can now reliably connect to UEFI VMs after migration
Previously, if you enabled or disabled a message queue while migrating a virtual machine (VM), the Virtual Network Computing (VNC) client failed to connect to the VM after the migration was complete.
This problem affected only UEFI-based VMs that used the Open Virtual Machine Firmware (OVMF).
The problem has been fixed, and the VNC client now reliably connects to UEFI VMs after the migration is complete.
The installer shows the expected system disk to install RHEL on VM
Previously, when installing RHEL on a VM using
virtio-scsi devices, it was possible that these devices did not appear in the installer because of a
device-mapper-multipath bug. Consequently, during installation, if some devices had a serial set and some did not, the
multipath command was claiming all the devices that had a serial. Due to this, the installer was unable to find the expected system disk to install RHEL in the VM.
With this update,
multipath correctly sets the devices with no serial as having no World Wide Identifier (WWID) and ignores them. On installation,
multipath only claims devices that
multipathd uses to bind a multipath device, and the installer shows the expected system disk to install RHEL in the VM.