INSTALL.md

INSTALLATION INSTRUCTIONS FOR GREENBONE VULNERABILITY MANAGER

Please note: The reference system used by most of the developers is Debian GNU/Linux 'Buster' 10. The build might fail on any other system. Also, it is necessary to install dependent development packages.

Prerequisites for Greenbone Vulnerability Manager

Prerequisites:

• GCC (Debian package: gcc)
• cmake >= 3.0 (Debian package: cmake)
• cJSON >= 1.7.14 (Debian package: libcjson-dev)
• glib-2.0 >= 2.42 (Debian package: libglib2.0-dev)
• gnutls >= 3.2.15 (Debian package: libgnutls28-dev)
• libgvm_base, libgvm_util, libgvm_osp, libgvm_gmp >= 20.08.0 (gvm-libs component)
• PostgreSQL database >= 9.6 (Debian packages: libpq-dev postgresql-server-dev-11)
• pkg-config (Debian package: pkg-config)
• libical >= 1.0.0 (Debian package: libical-dev)
• xsltproc (Debian package: xsltproc)
• gpgme

Install these prerequisites on Debian GNU/Linux 'Buster' 10:

apt-get install gcc cmake libcjson-dev libglib2.0-dev libgnutls28-dev libpq-dev postgresql-server-dev-11 pkg-config libical-dev xsltproc libgpgme-dev

Prerequisites for building documentation:

• Doxygen
• xsltproc (for building the GMP HTML documentation)
• xmltoman (optional, for building man page)

Prerequisites for building tests:

• Cgreen (optional, for building tests)

Please see the section "Prerequisites for Optional Features" below additional optional prerequisites.

Compiling Greenbone Vulnerability Manager

If you have installed required libraries to a non-standard location, remember to set the PKG_CONFIG_PATH environment variable to the location of you pkg-config files before configuring:

export PKG_CONFIG_PATH=/your/location/lib/pkgconfig:$PKG_CONFIG_PATH

Create a build directory and change into it with:

mkdir build
cd build

Then configure the build with:

cmake -DCMAKE_INSTALL_PREFIX=/path/to/your/installation ..

Or (if you want to use the default installation path /usr/local/):

cmake ..

This only needs to be done once.

Thereafter, the following commands are useful:

make                # build the scanner
make doc            # build the documentation
make doc-full       # build more developer-oriented documentation
make tests          # build tests
make install        # install the build
make rebuild_cache  # rebuild the cmake cache

Please note that you may have to execute make install as root, especially if you have specified a prefix for which your user does not have full permissions.

To clean up the build environment, simply remove the contents of the build directory you created above.

Choosing the Connection Type

Greenbone Vulnerability Manager can serve client connections on either a TCP socket or a UNIX domain socket.

The default is a UNIX domain socket at:

<install-prefix>/var/run/gvmd.sock

This location can be overridden with the --unix-socket option, and the permissions of the socket can be specified with the --listen-owner, --listen-group and --listen-mode options.

To use a TCP socket, call gvmd with the --listen option, for example:

gvmd --listen=127.0.0.1

Certificate Generation

All TCP-based communication with Greenbone Vulnerability Manager uses the TLS protocol to establish secure connections and for authentication and authorization. This requires the presence of a certificate infrastructure consisting of a certificate authority (CA) and a server and client certificate signed by the CA.

Greenbone Vulnerability Manager uses a client certificate when connecting to a scanner via the OSP protocol.

The easiest way to generate this certificate is to use the gvm-manage-certs script. A quick way to set up required certificates on the local system is to execute the command gvm-manage-certs -a.

If you intend to use OSP scanners and Manager on separate systems you need to make sure that the mutual trust is properly configured via the TLS certificates. The gvm-manage-certs script can assist you in setting up your infrastructure. Please refer to the documentation provided with the script for usage details.

If certificates have expired or in other ways there is need to update certificates for scanners, please see also section Updating Scanner Certificates.

Configure PostgreSQL Database Backend

Setting up the PostgreSQL database

1. Install Postgres.

   apt install postgresql postgresql-contrib postgresql-server-dev-all

2. Install the pg-gvm extension.

   Install the pg-gvm extension library (https://github.com/greenbone/pg-gvm). For instructions on how to do this, see the README file there.

3. Run cmake and build gvmd as usual.

4. Setup Postgres User and DB (/usr/share/doc/postgresql-common/README.Debian.gz)

   sudo -u postgres bash
   createuser -DRS mattm       # mattm is your OS login name
   createdb -O mattm gvmd

5. Setup permissions.

   sudo -u postgres bash  # if you logged out after step 4
   psql gvmd
   create role dba with superuser noinherit;
   grant dba to mattm;    # mattm is the user created in step 4

6. Make Postgres aware of the gvm libraries if not installed in a ld-aware directory. For example create file /etc/ld.so.conf.d/gvm.conf with appropriate path and then run ldconfig.

7. Run Manager as usual.

8. To run SQL on the database.

   psql gvmd

Switching between releases

There are two factors for developers to consider when switching between releases:

1. gvmd uses C server-side extensions that link to gvm-libs, so Postgres needs to be able to find the version of gvm-libs that goes with gvmd.

   One way to do this is to modify ld.so.conf and run ldconfig after installing the desired gvmd version.

2. The Postgres database "gvmd" must be the version that is supported by gvmd. If it is too high, gvmd will refuse to run. If it is too low gvmd will only run if the database is migrated to the higher version.

   One way to handle this is to switch between different versions of the database using RENAME:

   sudo -u postgres psql -q --command='ALTER DATABASE gvmd RENAME TO gvmd_10;'
   sudo -u postgres psql -q --command='ALTER DATABASE gvmd_master RENAME TO gvmd;'

Analyzing the size of the tables

In case the database grows in size and you want to understand which of the tables is responsible for it, there are two queries to check table sizes:

Biggest relations:

SELECT nspname || '.' || relname AS "relation",
    pg_size_pretty(pg_relation_size(C.oid)) AS "size"
  FROM pg_class C
  LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)
  WHERE nspname NOT IN ('pg_catalog', 'information_schema')
  ORDER BY pg_relation_size(C.oid) DESC
  LIMIT 20;

Biggest tables:

SELECT nspname || '.' || relname AS "relation",
    pg_size_pretty(pg_total_relation_size(C.oid)) AS "total_size"
  FROM pg_class C
  LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)
  WHERE nspname NOT IN ('pg_catalog', 'information_schema')
    AND C.relkind <> 'i'
    AND nspname !~ '^pg_toast'
  ORDER BY pg_total_relation_size(C.oid) DESC
  LIMIT 20;

These queries were taken from https://wiki.postgresql.org/wiki/Disk_Usage

Migrating the Database (e.g. during an upgrade of GVM)

If you have used Manager before (e.g. an older version which got upgraded to a newer major release), you might get the following error in your gvmd.log during startup:

gvmd: database is wrong version

If this is happening you need to migrate the database to the current data model. Use this command to run the migration:

gvmd --migrate

Creating an administrator user for GVM

You can create an administrator user with the --create-user option of gvmd:

gvmd --create-user=myuser

The new user's password is printed on success.

An administrator user can later create further users or administrators via clients like the Greenbone Security Assistant (GSA).

Also, the new user can change their password via GSA.

Set the Feed Import Owner

Certain resources that were previously part of the gvmd source code are now shipped via the feed. An example is the config "Full and Fast".

gvmd will only create these resources if a "Feed Import Owner" is configured:

gvmd --modify-setting 78eceaec-3385-11ea-b237-28d24461215b --value <uuid_of_user>

The UUIDs of all created users can be found using

gvmd --get-users --verbose

Keeping the feeds up-to-date

The gvmd Data, SCAP and CERT Feeds should be kept up-to-date by calling the greenbone-feed-sync script regularly (e.g. via a cron entry).

There are currently two synchronization methods available: The older shell-based one included in the gvmd repository and a newer Python-based one that also handles the VT synchronization.

Python-based sync tool

The currently recommended way of synchronizing the gvmd data feeds is the Python tool "greenbone-feed-sync", which can be found at https://github.com/greenbone/greenbone-feed-sync together with instruction for its installation and usage.

When upgrading to the new synchronization tool, the old script should be removed to avoid conflicts as both are named "greenbone-feed-sync".

Legacy shell script

The legacy feed sync script is deprecated and will be removed in the next major release of gvmd.

Therefore, installation of it is now disabled by default but can be enabled for backward compatibility with the CMake option -DINSTALL_OLD_SYNC_SCRIPTS=ON.

The legacy script has to be run for each type of data:

greenbone-feed-sync --type GVMD_DATA
greenbone-feed-sync --type SCAP
greenbone-feed-sync --type CERT

Please note: The CERT feed sync depends on data provided by the SCAP feed and should be called after syncing the latter. You will need the rsync tool for a successful synchronization.

Configure the default OSPD scanner socket path

By default, Manager tries to connect to the default OSPD scanner via the following path:

/var/run/ospd/ospd.sock

If this path doesn't match your setup you need to change the socket path accordingly.

Get the UUID of the OpenVAS Default scanner:

gvmd --get-scanners

Update the path (example, path needs to be adapted accordingly):

gvmd --modify-scanner=<uuid of OpenVAS Default scanner> --scanner-host=<install-prefix>/var/run/ospd/ospd-openvas.sock

Logging Configuration

By default, Manager writes logs to the file

<install-prefix>/var/log/gvm/gvmd.log

Logging is configured entirely by the file

<install-prefix>/etc/gvm/gvmd_log.conf

The configuration is divided into domains like this one

[md   main]
prepend=%t %p
prepend_time_format=%Y-%m-%d %Hh%M.%S %Z
file=/var/log/gvm/gvmd.log
level=128

The level field controls the amount of logging that is written. The value of level can be

  4  Errors.
  8  Critical situation.
 16  Warnings.
 32  Messages.
 64  Information.
128  Debug.  (Lots of output.)

Enabling any level includes all the levels above it. So enabling Information will include Warnings, Critical situations and Errors.

To get absolutely all logging, set the level to 128 for all domains in the configuration file.

Logging to syslog can be enabled in each domain like:

[md   main]
prepend=%t %p
prepend_time_format=%Y-%m-%d %Hh%M.%S %Z
file=syslog
syslog_facility=daemon
level=128

Optimizing the database

Greenbone Vulnerability Manager offers the command line option --optimize=<name> to run various optimization of the database. The currently supported values for <name> are:

• vacuum

  This option can reduce the file size by freeing some unused storage space in the database. For more information see the documentation for the VACUUM command of the database back-end you are using.

• analyze

  This option updates various internal statistics of the database used to optimize queries. For more information see the documentation for the ANALYZE command of the database back-end you are using.

• add-feed-permissions

  This option adds new read permissions on all feed data objects for the roles defined in the "Feed Import Roles" setting if they do not exist. The new permissions will be owned by the same user as the data objects, usually the feed import owner.

  This does not affect the command permissions, any permissions created for users or groups, or other types of permissions like modify or delete.

• cleanup-config-prefs

  This option removes duplicate preferences from Scan Configs and corrects some broken preference values. For the latter, the NVT preferences in the database must be up to date (if Manager and Scanner are both running, then this should happen automatically).

• cleanup-feed-permissions

  This option removes permissions on all feed data objects for all roles that are not defined in the "Feed Import Roles" setting.

  This does not affect the command permissions, any permissions created for users or groups, or other types of permissions like modify or delete.

• cleanup-port-names

  This cleans up the ports of results as stored in the database by removing parts that do not conform to the format <port>/<protocol>. For example the application name will be removed from a port using the old format telnet (23/tcp), reducing it to the new format 23/tcp. This makes filtering results and delta reports more consistent.

• cleanup-report-formats

  This cleans up references to report formats that have been removed without using the DELETE_REPORT_FORMAT GMP command, for example after a built-in report format has been removed.

• cleanup-result-nvts

  This cleans up results with missing result_nvt entries which can result in filters and overrides not working properly.

• cleanup-result-severities

  This cleans up results with no severity by assigning the default severity set by the user owning the result. All new results should have a severity assigned but this was not ensured in older versions, so this function can be used to correct missing severity scores in older reports.

• cleanup-sequences

  This cleans up id sequences that are likely to run out due to regular feed updates like the ids for config preferences.

• cleanup-tls-certificate-encoding

  This cleans up TLS certificates where the subject or issuer DN is not valid UTF-8.

• migrate-relay-sensors

  If relays are active, this can be used to make sure all sensor type scanners have a matching relay, i.e. OSP sensors have an OSP relay and GMP scanners have a GMP relay. GMP scanners are migrated to OSP sensors if an OSP relay is available.

• rebuild-report-cache

  This clears the cache containing the unfiltered result counts of all reports and fully rebuilds it.

• update-report-cache

  This creates the cache containing the unfiltered result counts of all reports that are not cached yet.

Encrypted Credentials

By default, the Manager stores private key and password parts of target credentials encrypted in the database. This avoids leaking such keys via backups. To be able to do a proper restore of the data, it is important to also backup the encryption key. The easiest way to do this is to create backup of the entire directory tree

<install-prefix>/var/lib/gvm/gvmd/gnupg/

and store it at a safe place independent of the database backups. This needs to be done only once after the key has been created or changed. The Manager creates the key at startup if it does not exist.

To check whether the key has been generated you may use the command:

gpg --homedir <install-prefix>/var/lib/gvm/gvmd/gnupg --list-secret-keys

An example output would be:

sec   2048R/1B55390F 2013-01-18
uid                  GVM Credential Encryption

Your key will have the same user ID (GVM Credential Encryption) but another keyid (1B55390F) and another creation date (2013-01-18).

Older versions of the Manager didn't used encrypted credentials. Thus, for old installations the database may hold a mix of cleartext and encrypted credentials. Note, that after changing a cleartext credential it will be saved encrypted.

To encrypt all existing credentials you may use:

gvmd --encrypt-all-credentials

Key change: If you disable the current key (see also the gpg manual) and create a new key, this command will decrypt using the old but disabled key and then re-encrypt using the new key. The command --decrypt-all-credentials may be used to revert to plaintext credentials:

gpg --homedir /var/lib/gvm/gvmd/gnupg -K

Look for the current key and remember its keyid. Then:

gpg --homedir /var/lib/gvm/gvmd/gnupg --edit-key KEYID

At the prompt enter disable followed by save and quit. Then start gvmd to create a new key:

gvmd

and finally re-encrypt all passwords:

gvmd --encrypt-all-credentials

No encryption: If for backward compatibility reasons encrypted credentials are not desired, the manager must always be started with the option --disable-encrypted-credentials.

Resetting Credentials Encryption Key

If you lost some part of the encryption key, neither a regular migration nor a simple creation might work.

In this case you need to reset the encryption key with the following procedure. There is no way to get the encrypted credentials back. You will need to enter all of them anew afterwards.

Get the key fingerprint:

gpg --homedir <install-prefix>/var/lib/gvm/gvmd/gnupg --list-secret-keys

Remove the secret key:

gpg --homedir=<prefix>/etc/openvas/gnupg --delete-secret-keys KEYID

Remove the key:

gpg --homedir=<prefix>/etc/openvas/gnupg --delete-keys KEYID

Create a new key:

gvmd --create-credentials-encryption-key

Finally, reset all credentials, by hand.

Updating Scanner Certificates

If you have changed the CA certificate used to sign the server and client certificates or the client certificate itself you will need to update the certificates in Manager database as well.

The database can be updated using the following command:

gvmd --modify-scanner <uuid> \
     --scanner-ca-pub <cacert> \
     --scanner-key-pub <clientcert> \
     --scanner-key-priv <clientkey>

Where:

• <uuid>

  refers to the UUID used by OpenVAS Manager to identify the scanner; the UUID can be retrieved with gvmd --get-scanners.

• <cacert>

  refers to the certificate of the CA used to sign the scanner certificate. Leaving this empty will delete the CA certificate of the scanner. This option can be dropped if the scanner uses a certificate that corresponds with the default CA certificate of Manager.

• <clientcert>

  refers to the certificate Manager uses to authenticate when connecting to the scanner. For a default OSP scanner setup with self-signed certificates this would be /var/lib/gvm/CA/clientcert.pem.

• <clientkey> refers to the private key Manager uses to authenticate when connecting to the scanner. For a default OSP scanner setup with self-signed certificates this would be /var/lib/gvm/private/CA/clientkey.pem.

To set just a new default CA certificate:

gvmd --modify-setting 9ac801ea-39f8-11e6-bbaa-28d24461215b \
     --value "`cat /var/lib/gvm/CA/cacert.pem`"

Replace the path to the pem-file with the one of your setup. The UUID is the fixed one of the immutable global setting for the default CA certificate and thus does not need to be changed.

Changing the Maximum Number of Rows per Page

The maximum number of rows returned by the GMP GET commands, like GET_TARGETS, is controlled by the GMP setting "Max Rows Per Page". This setting is an upper limit on the number of resources returned by any GET command, regardless of the value given for rows in the command's filter.

The default value for "Max Rows Per Page" is 1000. 0 indicates no limit.

This setting can not be changed via GMP. However, the gvmd option --modify-setting can be used to change it.

gvmd --modify-setting 76374a7a-0569-11e6-b6da-28d24461215b \
    --value 100

This changes the global value of the setting, and so applies to all users. Adding --user to the command will set a value for maximum rows only for that user.

Prerequisites for Optional Features

Certain features of the Manager also require some programs at run time:

Prerequisites for generating PDF reports:

• pdflatex

  On Debian GNU/Linux 'Stretch' 9 the following packages can be installed to fulfill this prerequisite:

  apt-get install texlive-latex-extra --no-install-recommends
  apt-get install texlive-fonts-recommended
  

Prerequisites for generating HTML reports:

• xsltproc

Prerequisites for generating verinice reports:

• xsltproc, xmlstarlet, zip

Prerequisites for generating credential RPM packages:

• rpm
• fakeroot

Prerequisites for generating credential DEB packages:

• dpkg
• fakeroot

Prerequisites for generating credentials .exe packages:

• makensis (usually distributed as part of nsis)

Prerequisites for generating system reports:

• A program in the PATH, with usage gvmcg seconds type, where seconds is the number of seconds before now that the report covers, and type is the type of report. When called with type titles the script must print a list of possible types, where the name of the type is everything up to the first space and everything else is a title for the report. When called with one of these types, gvmcg must print a PNG in base64 encoding. When called with the special type blank, gvmcg must print a PNG in base64 for the Manager to use when a request for one of the titled types fails. gvmcg may indicate failure by simply refraining from printing.

Prerequisites for signature verification:

• gnupg

Prerequisites for HTTP alerts:

• wget

Prerequisites for Alemba vFire alert:

• A program in the PATH called greenbone_vfire_connector that takes the path to an XML file as described by doc/vfire-data-xml.rnc as an argument.

Prerequisites for Sourcefire Connector alert:

• A program in the PATH called greenbone_sourcefire_connector that takes args IP, port, PKCS12 file and report file in Sourcefire format.

Prerequisites for verinice .PRO Connector alert:

• A program in the PATH called greenbone_verinice_connector that takes args IP, port, username, password and report file in verinice .PRO format.

Prerequisites for SCP alert:

• sshpass
• scp

Prerequisites for Send alert:

• socat

Prerequisites for SNMP alert:

• snmp

Prerequisites for SMB alert:

• python3
• smbclient

Prerequisites for Tipping Point alert:

• python3
• python3-lxml

Prerequisites for key generation on systems with low entropy:

• haveged (or a similar tool)

Prerequisites for S/MIME support (e.g. email encryption):

• GNU privacy guard - S/MIME version (Debian package: gpgsm)

Prerequisites for certificate generation:

• GnuTLS certtool (Debian package: gnutls-bin)

Static code analysis with the Clang Static Analyzer

If you want to use the Clang Static Analyzer (https://clang-analyzer.llvm.org/) to do a static code analysis, you can do so by prefixing the configuration and build commands with scan-build:

scan-build cmake ..
scan-build make

The tool will provide a hint on how to launch a web browser with the results.

It is recommended to do this analysis in a separate, empty build directory and to empty the build directory before scan-build call.