Both related to Val lists constructed as arguments to events that were
not freed because the event function was never called (e.g. no handlers).
Addresses #574
- The check for the root_certs variable was recreating the certificate
store over and over again which would eventually consume lots of memory
in the x509_stores internal (c++) variable. The index check uses the
Val pointer value for comparison now.
- The value of the content-length headers has now been removed
but it could be added back locally at an installation by a user.
- Added fields to indicate if some parsing interruption happened
during the body transfer.
- Closes#581
- If a protocol violation happens, Bro now logs what it has seen
up until the protocol violation and deletes the c$ssl record
so that a long lived connection with a protocol violation does
continue to hold the memory.
- Removed an notice definition from the base SSL scripts.
- Moved a logging stream ID into the export section for known-services
and bumped priority for creating the stream.
- Adding configuration knobs for the SQL injection attack detection
script and renaming the HTTP::SQL_Injection_Attack notice to
HTTP::SQL_Injection_Attack_Against
- Bumped priority when creating Known::CERTS_LOG.
- The test needed to append the second notice the event queue so that
the notice would be suppressed correctly.
- Setting $suppress_for manually in apply_policy since it doesn't pick
up the &default when directly creating the record inside the NOTICE call.
- Duplicate notices are discovered with the new Notice::Info
field $identifier. It's a string that is left up to the
notice implementor to define which would indicate a
fundamentally duplicate notice. The field is optional and
if it's not included it's not possible for notice
suppression to take place.
- Duplicate notices are suppressed by default for the interval
defined by the Notice::default_suppression_interval variable
(1 hour by default).
- A new notice action was defined ACTION_NO_SUPPRESS to prevent
suppression for a specific notice instance. A convenience set
named not_suppressed_types was also created to not suppress
entire notice types.
- A new field was added to the PolicyItem type to modify the length
of time a notice should be suppressed if the predicate matches.
The field is named $suppress_for. This name makes the code more
readable like this: $suppress_for = 1day
- New events were created to give visibility into the notice
framework's suppression activity.
- event Notice::begin_suppression(n: Notice::Info)
- event Notice::suppressed(n: Notice::Info)
- event Notice::end_suppression(n: Notice::Info)
- The suppression.bro script doesn't have a baseline because
it is causing a segfault in Bro. This one test is the
reason that this is being integrated into a branch instead
of master.
This required adding the ability for RecordVal::CoerceTo functions to
optionally allow orphaning fields. The default is to not allow it, but
now before writing to a log, the value of the stream's columns is coerced
down, if necessary, before passing it on to the filter's path_func.
Addresses #600.
This is because libmagic can return different results depending on
installation.
Note, this still doesn't fix all problems: smtp-entities sets md5
depending on the mime-type returned by libmagic ...
Log rotation is now controlled directly through Filter records.
Also addressed a TODO in the default_path_func regarding the
LogMgr::AddFilter function generating internal filter path
suggestions/fallbacks. Now, if the user doesn't explicitly set a filter
path, the filter's path will be the result of the first call to
default_path_func (happens during the first write to the log). And in
that case the path suggestion argument to the path_func is an empty
string.
The traces for external test-suites are no longer kept inside the
repositories themselves but downloaded separately via curl. This is
because git is pretty bad at dealing with large files.
See the README for more information.
- This is an attempt at fixing the memory issues brought about by
the introduction of the new SSL analyzer. My initial testing
shows a hefty memory saving.
* origin/topic/gilbert/rand-pool:
Updating tests.
Updated uid pools to use integer values instead of strings.
Updating tests.
Test no longer relevant. Need a way to generate and test collisions.
A few minor tweaks to make code less braindead. Fixed-length piece of pool name now only used to hash when determinism is not required; otherwise, whole pool name is used. Note that collisions between pool name hashes will lead to sensitivity to initialization order within the UID generator.
Testing long (>32 character) pool names.
Simple test to verify various pools are not affecting each other.
Some working code. Adds UID pools classified by string. Just compiles and runs; need to go back through and make sure this code is actually doing what I want it to do.
Note, I've removed the collision detection. Seems unlikely to occur
and even if, it's not really that bad.
- The "private interface" section is still tracked, but no longer rendered.
We can judge the usefulness of it later and add back if there's demand.
- Documentation of event handlers was being treated the same as event
declarations. Now handlers are tracked separately, but not currently
rendered in the generated doc output since usefulness is questionable.
- Reorganize top-level 'doc' Makefile target so submodules can easily
add their own doc-generating routines to it. e.g. the Bro project
makes a placeholder 'doc' target, then adds 'restdoc', 'sphinxdoc';
later Broccoli can add it's own target as a dependency for generating
API docs.
- Fixed generated docs for BIFs not being organized under a base/
subdirectory like the original source files.
- Fixed documentation style for function parameters not applying to
functions declared as record fields.
- Misc. script documentation tweaks to address warnings given by Sphinx.
