Never a panacea: ModSecurity drawbacks

I’ve written before of using ModSecurity for reducing bots traffic, especially for those bots that are not important to the success of a site, like almost all of the so-called “marketing bots”. Unfortunately, installing and setting up ModSecurity with the defaults parameter can cause quite a bit of headaches, especially for technically-oriented applications.

There are indeed quite a few different drawbacks to the use of that module, in particular related to the Core Rule Set that ships with it; some of the rules are quite taxing to the web server (since it has to parse eventually a lot of data), and others are simply hitting false positives quite easily.

For instance, the rule with id 960017 (Host header is a numeric IP address) while very valid usually breaks for the Nagios HTTP check, while the very draconian 950005 will stop any application to receive posts that talks about most Unix paths, including /etc. Luckily enough, mod_security does have means to handle whitelisting from rules with multiple methods: you can use rules that hit on user-agent (bad idea for whitelisting) or source IP (better), or you can use Apache environments.

For instance, my Typo instance has the following entry in its vhost definition to apply both my antispam rules and exclude the draconian rule:

Include /etc/apache2/vhosts.d/modsec_antispam.include
SecRuleRemoveById 950005

(Yes there are a few things that still needs to be cleared up, especially regarding the trackbacks that should probably have different antispam rules from the comments; in particular, trackbacks probably shouldn’t arrive from browsers at all).

So unfortunately, before modsec can be set up as standard piece of software for Apache servers, time has to pass…

(Val)grinding your code

One of of the most exoteric tool when working on software projects that developers should be using more often than they do, is certainly Valgrind. It’s a multi-faced tool: timing profiler, memory profiler, race conditions detector, … it can do all this because it’s mainly a simulated CPU, which executes the code, keeping track of what happens, and replacing some basic functions like the allocation routines.

So this works pretty well, to a point; unfortunately it does not always work out that well because there are a few minor issues: it needs to know about all the instructions used in the programs (otherwise it raises SIGILL crashes), and right now it does not know about SSE4 on x86 and x86-64 processors. And it has to know about the places in the C library where it cannot correctly keep scores via a series of suppression files. Turns out it currently fails at properly knowing about the changes in glibc 2.10, which makes it pretty difficult to find the real issues right in this moment with the most bleeding edge tools available.

But beside these problems with the modern tools and hardware, there are other interesting edges; one of these, the one I’m going to write about today, relate to the way the memory checker tool identifies leaks in the software: all the memory areas that cannot be reached in some way at the end of the execution, are considered leaked. Unfortunately this often enough counts in the memory areas that are allocated at one point and would just be freed at the end of the process, which can be said not to be a leak at all.

All the memory areas that are only to be freed at the end of a process don’t really need to be freed at all: the operating system will take care of that; leaks are those where the memory is used up and never freed nor used during execution; especially bad are those leaks that happen inside either iterative cycles or upon similar inputs, because they build up.

Actually freeing up the resources that does not need to be freed up can increase the size of the executable section of a binary (executable or library) for no good reason during production usage of the software, it is thus often considered a bad idea. At the same time, freeing them up makes it easier to find the actual bad leaks in the software. My choice for this is to write the actual source code to do the freeing up, but only compile it when debug code is enabled, just like assertions and debug output.

This way, the production builds won’t have unneeded code, while the debug builds will give proper results out on tools like valgrind. But it doesn’t stop at freeing the resources at the end of the main() function, because sometimes you don’t preserve reference to the actual memory areas allocated during initialization (since they are not strictly needed), or you just keep them as static local variables, which are both reported as unreachable by the end of the process, and are not easy to free up. To work this out, in feng, I’ve been using a slightly more sophisticated solution, making use of a feature of recent GCC and ICC versions: destructor functions.

Destructor functions are called within the fini execution path, after the main() and before the process is unloaded and the resource freed automatically. Freeing the memory at that point is the ideal situation to make sure that all the resources are freed up properly. Of course, this reduces the cases when the code is enabled to having both debug system enabled, and a compiler that supports the destructor functions. But this is a broad enough definition to allow proper development. A simple check with autoconf and the thing is done.

By switching the local static variables with unit-static variables (which, once compiled, are basically the same thing), and adding a cleanup function, almost all memory can be freed without having an explicit deallocation function. There is still the matter of initializing these memory areas; in this case there are two options: either you just initialize them explicitly, or you initialize them the first time they are needed. Both these options allow for the size of the structure to be chosen at runtime (from a configuration file), for example for the number of configuration contexts to initialize. The third way, that is to initialize them with constructor functions, also works out well for fixed-size memory areas, but since I don’t really have a good reason for which they could be better than the alternative the to just call a function during initialization, I don’t have any reason to take them into consideration.

For dynamic initialization, the interface usually used is the “once” interface, which also allows for multithreaded applications to have run-once blocks of code without having to fiddle with mutexes and locks; this is implemented by the pthread_once_t handling functions from the POSIX thread interface, or GOnce as provided by glib. This allows delayed allocation and initialization of structures which is especially useful for parts of the code that are not used constantly but just in some specific cases.

All the code can be found in your local feng repository or the nearest git server!

CrawlBot Wars

Everybody who ever wanted to write a “successful website” (or more recently, thanks to the Web 2.0 hype, a “successful blog”) knows the bless and curse of crawlers, or bots, that are unleashed by all kind of entities to scan the web, and report the content back to their owners.

Most of these crawlers are handled by search engines, such as Google, Microsoft Live Search, Yahoo! and so on. With the widespread use of feeds, at least Google and Yahoo! added to their standard crawler bots also feed-specific crawlers that are used to aggregate blogs and other feeds into nice interfaces for their users (think Google Reader). Together with this kind of crawlers, though, there are less useful, sometimes nastier crawlers that either don’t respond to search engines, or respond to search engines whose ethical involvement makes somewhat wonder.

Good or bad, at the end of the day you might not want some bots to crawl your site; some Free Software -bigots- activists some time ago wanted, for instance, to exclude the Microsoft bot from their sites (while I have some other ideas), but there are certain bots that are even more useful to block, like the so-called “marketing bots”.

You might like Web 2.0 or you might not, but certainly lots of people found the new paradigm of Web as a gold mind to make more money out of content others have written – incidentally these are not, like RIAA, MPAA and SIAE insist, the “pirates” that copy music and movies, but rather companies whose objective is to provide other companies with marketing research and data based on content of blogs and similar services. While some people might be interested in getting their blog scanned by these crawlers either way, I’d guess that for most users who host their own blog this is just a waste of bandwidth: the crawlers tend to be quite pernicious since they don’t use If-Modified-Since or Etag headers in their request, and even when they do, they tend to make quite a few requests on the feeds per hour (compare this with Google’s Feedfetcher bot that requires at most one copy of the same feed per hour – well, if it isn’t confused by multiple compatibility redirects like it unfortunately is with my main blog).

While there is a voluntary exclusion protocol (represented by the omni-present robots.txt file), only actually “good” robots do consider that, while evil or rogue robots can simply ignore it. Also, it might be counter-productive to block rogue robots even when they do look at it. Say that a rogue robot wants your data, and to pass as a good one is advertising itself in the User-Agent string, complete with a link to a page explaining what it’s supposedly be doing, and accepting the exclusion. If you exclude it in robots.txt you can give it enough information to choose a _different_ User-Agent string that is not listed in the exclusion protocol.

One way to deal with the problem is by blocking the requests at the source, answering straight away with an HTTP 403 (Access Denied) on the web server when making a request. When using the Apache web server, the easiest way to do this is by using modsecurity and a blacklist rule for rogue robots, similar to the antispam system I’ve been using for a few months already. The one problem I see with this is that Apache’s mod_rewrite seem to be executed _before_ mod_security, which means that for any request that is rewritten by compatibility rules (moved, renamed, …) there is first a 301 response and just after that an actual 403.

I’m currently working on compiling such a blacklist by analysing the logs of my server, the main problem is deciding which crawlers to block and which to keep. When the description page explicitly states they are marketing research, blocking them is quite straightforward; when they _seem_ to provide an actual search service, that’s more shady, and it turns down to checking the behaviour of the bot itself on the site. And then there are the vulnerability scanners.

Still, it doesn’t stop here: given that in the Google description of GoogleBot they provide a (quite longish to be honest) method to verify that a bot is actually GoogleBot as it advertises itself to be, one has to assume that there are rogue bots out there trying to pass for GoogleBot or other good and lecit bot. This is very likely the case because some website that are usually visible only by registered users make an exception for search engine crawlers to access and index their content.

Especially malware, looking for backdoors into a web application, is likely to forge the User-Agent of a known good search engine bot (that is likely _not_ blocked by the robots.txt exclusion list), so that it doesn’t fire up any alarm in the logs. So finding “fake” search engine bots is likely to be an important step in securing a webserver running webapplications, may them be trusted or not.

As far as I know there is currently no way in Apache to check that a request actually does come from the bot it’s declared to come from. The nslookup method that Google suggests works fine for a forensic analysis but it’s almost impossible to perform properly with Apache itself, and not even modsecurity, by itself, can do much about that. On the other hand, there is one thing in the recent 2.5 versions of modsecurity that can be probably used to implement an actually working check: the LUA scripts loading. Which is what I’m going to work on as soon as I find some extra free time.

Autotools Come Home

With our experience as Gentoo developers, me and Luca have had experience with a wide range of build systems; while there are obvious degrees of goodness/badness in build system worlds, we express our preference for autotools over most of the custom build systems, and especially over cmake-based build systems, that seem to be high on the tide thanks to KDE in the last two years.

I have recently written my views on build systems: in which I explain why I dislike CMake and why I don’t mind it when it is replacing a very-custom bad build system. The one reason I gave for using CMake is the lack of support for Microsoft Visual C++ Compiler, which is needed by some type of projects under Windows (GCC still lacks way too many features); this starts to become a moot point.

Indeed if you look at the NEWS file for the latest release (unleashed yesterday) 1.11, there is this note:

– The `depcomp’ and `compile’ scripts now work with MSVC under MSYS.

This means that when running configure scripts under MSYS (which means having most of the POSIX/GNU tools available under the Windows terminal prompt), it’s possible to use the Microsoft compiler, thanks to the compile wrapper script. Of course this does not mean the features are on par with CMake yet, mostly because all the configure scripts I’ve seen up to now seem to expect GCC or compatible compilers, which means that it will require for more complex tests, and especially macro archives, to replace the Visual Studio project files. Also, CMake having a fairly standard way to handle options and extra dependencies, can have a GUI to select those, where autotools are still tremendously fragmented in that regard.

Additionally, one of the most-recreated and probably useless features, the Linux-style quiet-but-not-entirely build, is now implemented directly in automake through the silent-make option. While I don’t see much point in calling that a killer feature I’m sure there are people who are interested in seeing that.

While many people seem to think that autotools are dead and that they should disappear, there is actually fairly active development behind them, and the whole thing is likely going to progress and improve over the next months. Maybe I should find the time to try making the compile wrapper script work with Borland’s compiler too, of which I have a license; it would be one feature that CMake is missing.

At any rate, I’ll probably extend my autotools guide for automake 1.11, together with a few extras, in the next few days. And maybe I can continue my Autotools Mythbuster series that I’ve been writing on my blog for a while.

Profiling Proficiency

Trying to assess the impact of the Ragel state machine generator on software, I’ve been trying to come down with a quick way to seriously benchmark some simple testcases, making sure that the results are as objective as possible. Unfortunately, I’m not a profiling expert; to keep with the alliteration in the post’s title, I’m not proficient with profilers.

Profiling can be done either internally to the software (manually or with GCC’s help), externally with a profiling software, or in the kernel with a sampling profiler. Respectively, you’d be found using the rdtsc instruction (on x86/amd64), the gprof command, valgrind and one of either sysprof or oprofile. Turns out that almost all these options don’t give you anything useful in truth, and you have to decide whether to get theoretical data, or practical timing, and in both cases, you don’t have a very useful result.

Of that list of profiling software, the one that looked more promising was actually oProfile; especially more interesting since the AMD Family 10h (Barcelona) CPUs supposedly have a way to have accurate reporting of instruction executed, which should combine the precise timing reported by oProfile with the invariant execution profile that valgrind provides. Unfortunately oprofile’s documentation is quite lacking, and I could find nothing to get that “IBS” feature working.

Since oprofile is a sampling profiler, it has a few important points to be noted: it requires kernel support (which in my case required a kernel rebuild because I had profiling support disabled), it requires root to set up and start profiling, through a daemon process, by default it profiles everything running in the system, which on a busy system running a tinderbox might actually be _too_ much. Support for oprofile in Gentoo also doesn’t seem to be perfect, for instance there is no standardised way to start/stop/restart the daemon, which might not sound that bad for most people, but is actually useful because sometimes you forget to have the daemon running and you try to start it time and time again and it doesn’t seem to work as you expect. Also, for some reason it stores the data in /var/lib instead of /var/cache; this wouldn’t be a problem if it wasn’t that if you don’t pay enough attention you can easily see your /var/lib filesystem filling up (remember: it runs at root so it bypasses the root-reserved space).

More worrisome, you’ll never get proper profiling on Gentoo yet, at least for AMD64 systems. The problem is that all sampling profilers (so the same holds true for sysprof too) require frame pointer information; the well-known -fomit-frame-pointer flag, that allows to save precious registers on x86, and used to break debug support, can become a problem, as Mart pointed out to me. The tricky issue is that since a few GCC and GDB versions, the frame pointer is no longer needed to be able to get complete backtraces of processes being debugged; this meant that, for instance on AMD64, the flag is now automatically enabled by -O2 and higher. On some architecture, this is still not a problem to sample-based profiling, but on AMD64 it is. Now, the testcases I had to profile are quite simple and minimal and only call into the C library (and that barely calls the kernel to read the input files), so I only needed to have the C library built with frame pointers to break down the functions; unfortunately this wasn’t as easy as I hoped.

The first problem is that the Gentoo ebuild for glibc does have a glibc-omitfp USE flag; but does not seem to explicitly disable frame pointer omitting, just explicitly enable it. Changing that, thus, didn’t help; adding -fno-omit-frame-pointer also does not work properly because flags are (somewhat obviously) filtered down; adding that flag to the whitelist in flag-o-matic does not help either. I have yet to see it working as it’s supposed to, so for now oprofile is just accumulating dust.

My second choice was to use gprof together with GCC’s profiling support, but that only seems to provide a breakdown in percentage of execution time, which is also not what I was aiming for. The final option was to fall back to valgrind, in particular the callgrind tool. Unfortunately the output of that tool is not human-readable, and you usually end up using software like KCacheGrind to be able to parse it; but since this system is KDE-free I couldn’t rely on that; and also KCacheGrind does not provide an easy way to extract data out of it to compile benchmark tables and tests.

On the other hand, the callgrind_annotate script does provide the needed information, although still in a not exactly software-parsable fashion; indeed to find the actual information I had to look at the main() calltrace and then select the function I’m interested in profiling, that provided me with the instruction count of execution. Unfortunately this does not really help me as much as I was hoping, since it tells me nothing on how much time will it take for the CPU to execute the instruction (which is related to the amount of cache and other stuff like that), but it’s at least something to start with.

I’ll be providing the complete script, benchmark and hopefully results once I can actually have a way to correlate them to real-life situations.