Commit 9f5c3632 authored by JINMEI Tatuya's avatar JINMEI Tatuya
Browse files

[904] added general description of the script. also ocrrected the file

name (it was given the wrong name in the first commit).
parent f0071203
......@@ -15,6 +15,320 @@
Generator of various types of DNS data in the hex format.
This script reads a human readable specification file (called "spec
file" hereafter) that defines some type of DNS data (an RDATA, an RR,
or a complete message) and dumps the defined data to a separate file
as a "wire format" sequence parsable by the
UnitTestUtil::readWireData() function (currently defined as part of
libdns++ tests). Many DNS related tests involve wire format test
data, so it will be convenient if we can define the data in a more
intuitive way than writing the entire hex sequence by hand.
Here is a simple example. Consider the following spec file:
sections: a
as_rr: True
When the script reads this file, it detects the file specifies a single
component (called "section" here) that consists of a single A RDATA,
which must be dumped as an RR (not only the part of RDATA). It then
dumps the following content:
# A RR ( Class=IN(1) TTL=86400 RDLEN=4)
076578616d706c6503636f6d00 0001 0001 00015180 0004
# Address=
As can be seen, the script automatically completes all variable
parameters of RRs: owner name, class, TTL, RDATA length and data. For
testing purposes many of these will be the same common one (like
"" or, so it would be convenient if we only have
to specify non default parameters. To change the RDATA (i.e., the
IPv4 address), we should add the following line at the end of the spec
Then the last two lines of the output file will be as follows:
# Address=
In some cases we would rather specify malformed data for tests. This
script has the ability to specify broken parameters for many types of
data. For example, we can generate data that would look like an A RR
but the RDLEN is 3 by adding the following line to the spec file:
rdlen: 3
Then the first two lines of the output file will be as follows:
# A RR ( Class=IN(1) TTL=86400 RDLEN=3)
076578616d706c6503636f6d00 0001 0001 00015180 0003
** USAGE ** [-o output_file] spec_file
If the -o option is missing, and if the spec_file has a suffix (such as
in the form of "data.spec"), the output file name will be the prefix
part of it (as in "data"); if -o is missing and the spec_file does not
have a suffix, the script will fail.
A spec file accepted in this script should be in the form of a
configuration file that is parsable by the Python's standard
configparser module. In short, it consists of sections; each section
is identified in the form of [section_name] followed by "name: value"
entries. Lines beginning with # or ; will be treated as comments.
Refer to the configparser module documentation for further details of
the general syntax.
This script has two major modes: the custom mode and the DNS query
mode. The former generates an arbitrary combination of DNS message
header, question section, RDATAs or RRs. It is mainly intended to
generate a test data for a single type of RDATA or RR, or for
complicated complete DNS messages. The DNS query mode is actually a
special case of the custom mode, which is a shortcut to generate a
simple DNS query message (with or without EDNS).
* Custom mode syntax *
By default this script assumes the DNS query mode. To specify the
custom mode, there must be a special "custom" section in the spec
file, which should contain 'sections' entry. This value of this
entryis colon-separated string fields, each of which is either
"header", "question", "edns", "name", or a string specifying an RR
type. For RR types the string is lower-cased string mnemonic that
identifies the type: 'a' for type A, 'ns' for type NS, and so on
(note: in the current implementation it's case sensitive, and must be
lower cased).
Each of these fields is interpreted as a section name of the spec
(configuration), and in that section parameters specific to the
semantics of the field can be configured.
A "header" section specifies the content of a DNS message header.
See the documentation of the DNSHeader class of this module for
configurable parameters.
A "question" section specifies the content of a single question that
is normally to be placed in the Question section of a DNS message.
See the documentation of the DNSQuestion class of this module for
configurable parameters.
An "edns" section specifies the content of an EDNS OPT RR. See the
documentation of the EDNS class of this module for configurable
A "name" section specifies a domain name with or without compression.
This is specifically intended to be used for testing name related
functionalities and would rarely be used with other sections. See the
documentation of the Name class of this module for configurable
In a specific section for an RR or RDATA, possible entries depend on
the type. But there are some common configurable entries. See the
description of the RR class. The most important one would be "as_rr".
It controls whether the entry should be treated as an RR (with name,
type, class and TTL) or only as an RDATA. By default as_rr is
"False", so if an entry is to be intepreted as an RR, an as_rr entry
must be explicitly specified with a value of "True".
Another common entry is "rdlen". It specifies the RDLEN field value
of the RR (note: this is included when the entry is interpreted as
RDATA, too). By default this value is automatically determined by the
RR type and (it has a variable length) from other fields of RDATA, but
as shown in the above example, it can be explicitly set, possibly to a
bogus value for testing against invalid data.
For type specific entries (and their defaults when provided), see the
documentation of the corresponding Python class defined in this
module. In general, there should be a class named the same mnemonic
of the corresponding RR type for each supported type, and they are a
subclass of the RR class. For example, the "NS" class is defined for
RR type NS.
Look again at the A RR example shown at the beginning of this
description. There's a "custom" section, which consists of a
"sections" entry whose value is a single "a", which means the data to
be generated is an A RR or RDATA. There's a corresponding "a"
section, which only specifies that it should be interpreted as an RR
(all field values of the RR are derived from the default).
If you want to generate a data sequence for two ore more RRs or
RDATAs, you can specify them in the form of colon-separated fields for
the "sections" entry. For example, to generate a sequence of A and NS
RRs in that order, the "custom" section would be something like this:
sections: a:ns
and there must be an "ns" section in addtion to "a".
If a sequence of two or more RRs/RDATAs of the same RR type should be
generated, these should be uniquely indexed with the "/" separator.
For example, to generate two A RRs, the "custom" section would be as
sections: a/1:a/2
and there must be "a/1" and "a/2" sections.
Another practical example that would be used for many tests is to
generate data for a complete DNS ressponse message. The spec file of
such an example configuration would look like as follows:
sections: header:question:a
qr: 1
ancount: 1
as_rr: True
With this configuration, this script will generate test data for a DNS
response to a query for containing one corresponding
A RR in the answer section.
* DNS query mode syntax *
If the spec file does not contain a "custom" section (that has a
"sections" entry), this script assumes the DNS query mode. This mode
is actually a special case of custom mode; it implicitly assumes the
"sections" entry whose value is "header:question:edns".
In this mode it is expected that the spec file also contains at least
a "header" and "question" sections, and optionally an "edns" section.
But the script does not warn or fail even if the expected sections are
* Entry value types *
As described above, a section of the spec file accepts entries
specific to the semantics of the section. They generally correspond
to DNS message or RR fields.
Many of them are expected to be integral values, for which either decimal or
hexadecimal representation is accepted, for example:
rr_ttl: 3600
tag: 0x1234
Some others are expected to be string. A string value does not have
to be quated:
but can also be quoated with single quotes:
address: ''
Note 1: a string that can be interpreted as an integer must be quated.
For example, if you want to set a "string" entry to "3600", it should
string: '3600'
instead of
string: 3600
Note 2: a string enclosed with double quotes is not accepted:
# This doesn't work:
address: ""
In general, string values are converted to hexadecimal sequences
according to the semantics of the entry. For instance, a textual IPv4
address in the above example will be converted to a hexadecimal
sequence corresponding to a 4-byte integer. So, in many cases, the
acceptable syntax for a particular string entry value should be
obvious from the context. There are still some exceptional cases
especially for complicated RR field values, for which the
corresponding class documentation should be referenced.
One special string syntax that would be worth noting is domain names,
which would natually be used in many kinds of entries. The simplest
form of acceptable syntax is a textual representation of domain names
such as "" (note: names are always assumed to be
"absolute", so the trailing dot can be omitted). But a domain name in
the wire format can also contain a compression pointer. This script
provides a simple support for name compression with a special notation
of "ptr=nn" where nn is the numeric pointer value (decimal). For example,
if the NSDNAME field of an NS RDATA is specified as follows:
nsname: ns.ptr=12
this script will generate the following output:
# NS name=ns.ptr=12
This script is expected to be extended as we add more support for
various types of RR. It is encouraged to add support for a new type
of RR to this script as we see the need for testing that type. Here
is a simple instruction of how to do that.
Assume you are adding support for "FOO" RR. Also assume that the FOO
RDATA contains a single field named "value".
What you are expected to do is as follows:
- Add a dictionary entry of "'foo': (FOO, {})" to config_param of the
get_config_param() function (this step could be automated; we may do
it in a future version)
- Define a new class named "FOO" inherited from the RR class. Also
define a class variable named "value" for the FOO RDATA field (the
variable name can be different from the field name, but it's
convenient if it can be easily identifiable.) with an appropriate
default value (if possible):
class FOO(RR):
value = 10
The name of the variable will be (automatically) used as the
corresponding entry name in the spec file. So, a spec file that
sets this field to 20 would look like this:
value: 20
- Define the "dump()" method for class FOO. It must call
self.dump_header() (which is derived from class RR) at the
beginning. It then prints the RDATA field values in an appropriate
way. Assuming the value is a 16-bit integer field, a complete
dump() method would look like this:
def dump(self, f):
if self.rdlen is None:
self.rdlen = 2
self.dump_header(f, self.rdlen)
f.write('# Value=%d\\n' % (self.value))
f.write('%04x\\n' % (self.value))
The first f.write() call is not mandatory, but is encouraged to
provide so that the generated files will be more human readable.
Depending on the complexity of the RDATA fields, the dump()
implementation would be more complicated. In particular, if the
RDATA length is variable and the RDLEN field value is not specified
in the spec file, the dump() method is normally expected to
calculate the correct length and pass it to dump_header(). See the
implementation of various derived classes of class RR for actual
import configparser, re, time, socket, sys
from datetime import datetime
from optparse import OptionParser
......@@ -231,12 +545,15 @@ class RR:
derived class for the RR type SOA should be named "SOA".
Configurable parameters are as follows:
- as_rr (bool): Whether or not the data is to be dumped as an RR. False
by default.
- rr_class (string): The RR class of the data. Only meaningful when the
data is dumped as an RR. Default is 'IN'.
- rr_ttl (integer): The TTL value of the RR. Only meaningful when the
data is dumped as an RR. Default is 86400 (1 day).
- as_rr (bool): Whether or not the data is to be dumped as an RR.
False by default.
- rr_name (string): The owner name of the RR. The string must be
interpreted as a valid domain name (compression pointer can be
contained). Default is ''
- rr_class (string): The RR class of the data. Only meaningful
when the data is dumped as an RR. Default is 'IN'.
- rr_ttl (integer): The TTL value of the RR. Only meaningful when
the data is dumped as an RR. Default is 86400 (1 day).
def __init__(self):
......@@ -595,7 +912,7 @@ if __name__ == "__main__":
print_header(output, configfile)
# First try the 'custom' mode; if it fails assume the standard mode.
# First try the 'custom' mode; if it fails assume the query mode.
sections = config.get('custom', 'sections').split(':')
except configparser.NoSectionError:
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment