clamav/docs/UserManual/development.md

ClamAV Development

This page aims to provide information useful when developing, debugging, or profiling ClamAV.

Building ClamAV for Development

Below are some recommendations for building ClamAV so that it's easy to debug:

Running ./configure

Suggestions:

• Modify the CFLAGS variable as follows (assuming you're build with gcc):

  • Include gdb debugging information (-ggdb). This will make it easier to debug with gdb

  • Disable optimizations (-O0). This will ensure the line numbers you see in gdb match up with what is actually being executed.

• Run configure with the following options:

  • --prefix=`pwd`/build: This will cause make install to install into the specified directory to avoid potentially tainting a release install of ClamAV that you may have.

  • --enable-debug: This will define CL_DEBUG, which mostly just enables additional print statements that are useful for debugging

  • --enable-check: Enables the unit tests, which can be run with 'make check'

  • --enable-coverage: If using gcc, sets -fprofile-arcs -ftest-coverage so that code coverage metrics will get generated when the program is run. Note that the code inserted to store program flow data may show up in any generated flame graphs or profiling output, so if you don't care about code coverage, omit this

  • --enable-libjson: Enables libjson, which enables the --gen-json option. The json output contains additional metadata that might be helpful when debugging.

  • --enable-static --disable-shared: This will only build libclamav and the supporting libraries as static libraries, and will result in the clamscan that is built having this code embedded. This is useful for running programs like gprof which don't handle profiling code in shared objects.

  • --with-systemdsystemunitdir=no: Don't try to register clamd as a systemd service

  • You might want to include the following flags also so that the optional functionality is enabled: --enable-experimental --enable-clamdtop --enable-libjson --enable-milter --enable-xml --enable-pcre. Note that this may require you to install additional development libraries.

  • I ran into problems building with llvm on Ubuntu 18.04, so add --disable-llvm

Altogether, the following configure command can be used:

CFLAGS="-ggdb -O0" ./configure --prefix=`pwd`/built --enable-debug --enable-check --enable-coverage --enable-libjson --enable-static --disable-shared --with-systemdsystemunitdir=no --enable-experimental --enable-clamdtop --enable-libjson --enable-xml --enable-pcre --disable-llvm

To satisify all library dependencies, something like this should work (from Ubuntu 18.04):

sudo apt-get install git gcc libxml2-dev libssl-dev make libmilter-dev libcurl4-openssl-dev libjson-c-dev check pkgconf libncurses5-dev libpcre3-dev g++ libtool libbz2-dev

Running make

Run the following to finishing building. -j2 in the code below is used to indicate that the build process should use 2 cores. Increase this if your machine is more powerful.

make -j2
make install

Also, you can run 'make check' to run the unit tests

Downloading the Official Ruleset

If you plan to use custom rules for testing, you can invoke clamscan via ./built/bin/clamscan, specifying your custom rule files via -d parameters.

If you want to download the official ruleset to use with clamscan, do the following:

1. Run mkdir -p built/share/clamav
2. Comment out line 8 of etc/freshclam.conf.sample
3. Run ./built/bin/freshclam --config-file etc/freshclam.conf.sample

General Debugging

Useful clamscan Flags

The following are useful flags to include when debugging clamscan:

• --debug --verbose: Print lots of helpful debug information

• --gen-json: Print some additional debug information in a JSON format

• --statistics=pcre --statistics=bytecode: Print execution statistics on any PCRE and bytecode rules that were evaluated

• --dev-performance: Print per-file statistics regarding how long scanning took and the times spent in various scanning stages

• --detect-broken: This will attempt to detect broken executable files. If an executable is determined to be broken, some functionality might not get invoked for the sample, and this could be an indication of an issue parsing the PE header or file. This causes those binary to generate an alert instead of just continuing on.

• --max-filesize=2000M --max-scansize=2000M --max-files=2000000 --max-recursion=2000000 --max-embeddedpe=2000M --max-htmlnormalize=2000000 --max-htmlnotags=2000000 --max-scriptnormalize=2000000 --max-ziptypercg=2000000 --max-partitions=2000000 --max-iconspe=2000000 --max-rechwp3=2000000 --pcre-match-limit=2000000 --pcre-recmatch-limit=2000000 --pcre-max-filesize=2000M: Effectively disables all file limits and maximums for scanning. This is useful if you'd like to ensure that all files in a set get scanned, and would prefer clam to just run slowly or crash rather than skip a file because it encounters one of these thresholds

The following are useful flags to include when debugging rules that you're writing:

• -d: Allows you to specify a custom ClamAV rule file from the command line

• --bytecode-unsigned: If you are testing custom bytecode rules, you'll need this flag so that clamscan actually runs the bytecode signature

• --all-match: Allows multiple signatures to match on a file being scanned

• --leave-temps --tmpdir=/tmp: By default, ClamAV will attempt to extract embedded files that it finds, normalize certain text files before looking for matches, and unpack packed executables that it has unpacking support for. These flags tell ClamAV to write these intermediate files out to the directory specified. Usually when a file is written, it will mention the file name in the --debug output, so you can have some idea at what stage in the scanning process a tmp file was created.

• --dump-certs: For signed PE files that match a rule, display information about the certificates stored within the binary. Note - sigtool has this functionality as well and doesn't require a rule match to view the cert data

Useful sigtool Flags

sigtool pulls in libclamav and provides shortcuts to doing tasks that clamscan does behind the scenes. These can be really useful when writing a signature or trying to get information about a signature that might be causing FPs or performance problems.

The following sigtool flags can be useful when debugging:

• --unpack: Unpack the specified CVD/CLD file

• --decode: Given a ClamAV signature from STDIN, show a more user-friendly representation of it

• --hex-dump: Given a sequence of bytes from STDIN, print the hex equivalent

• --mdb: Generate section hashes of the specified file

• --imp: Generate import hashes of the specified file

• --html-normalise: Normalize the specified HTML file in the way that clamscan will before looking for rule matches. This makes it either to write rules that will actually match.

• --ascii-normalise: Normalized the specified ASCII text file in the way that clamscan will before looking for rule matches

• --print-certs: Print the Authenticode signatures of any PE files specified. This is useful when writing signature-based .crb rule files.

• --vba: Extract VBA/Word6 macro code

Using gdb

Given that you might want to pass a lot of arguments to gdb, consider taking advantage of the --args parameter. For example:

gdb --args ./built/bin/clamscan -d /tmp/test.ldb -d /tmp/blacklist.crb -d --dumpcerts --debug --verbose --max-filesize=2000M --max-scansize=2000M --max-files=2000000 --max-recursion=2000000 --max-embeddedpe=2000M --max-iconspe=2000000 f8f101166fec5785b4e240e4b9e748fb6c14fdc3cd7815d74205fc59ce121515

When using ClamAV without libclamav statically linked, if you set breakpoints on libclamav functions by name, you'll need to make sure to indicate that the breakpoints should be resolved after libraries have been loaded.

For other documentation about how to use gdb, check out the following resources:

• A Guide to gdb
• gdb Quick Reference

Hunting for Memory Leaks

You can easily hunt for memory leaks with valgrind. Check out this guide to get started:

• Valgrind Quick Start If checking for leaks, be sure to run clamscan with samples that will hit as many of the unique code paths in the code you are testing. An example invocation is as follows:

valgrind --leak-check=full ./built/bin/clamscan -d /tmp/test.ldb --leave-temps --tempdir /tmp/test --debug --verbose /tmp/upx-samples/ > /tmp/upx-results-2.txt 2>&1

Alternatively, on Linux, you can use glibc's built-in leak checking functionality:

MALLOC_CHECK_=7 ./built/bin/clamscan

See the mallopt man page for more details

Computing Code Coverage

gcov/lcov can be used to produce a code coverage report indicating which lines of code were executed on a single run or by multiple runs of clamscan. NOTE: for these metrics to be collected, ClamAV needs to have been configured with the --enable-coverage option.

First, run the following to zero out all of the performance metrics:

lcov -z --directory . --output-file coverage.lcov.data

Next, run ClamAV through whatever test cases you have. Then, run lcov again to collect the coverage data as follows:

lcov -c --directory . --output-file coverage.lcov.data

Finally, run the genhtml tool that ships with lcov to produce the code coverage report:

genhtml coverage.lcov.data --output-directory report

The report directory will have an index.html page which can be loaded into any web browser.

For more information, visit the lcov webpage

Profiling - Flame Graphs

FlameGraph is a great tool for generating interactive flame graphs based collected profiling data. The github page has thorough documentation on how to use the tool, but an overview is presented below:

First, install perf, which on Linux can be done via:

apt-get install linux-tools-common linux-tools-generic linux-tools-`uname -r`

Modify the system settings to allow perf record to be run by a standard user:

$ sudo su
# cat /proc/sys/kernel/perf_event_paranoid 
# echo "1" > /proc/sys/kernel/perf_event_paranoid 
# exit

Invoke clamscan via perf record as follows, and run perf script to collect the profiling data:

perf record -F 100 -g -- ./built/bin/clamscan -d /tmp/test.ldb /tmp/2aa6b18d509090c60c3e4ecdd8aeb16e5f149807e3404c86892112710eab576d
perf script > out.perf

The '-F' parameter indicates how many samples should be collected during program execution. If your scan will take a long time to run, a lower value should be sufficient. Otherwise, consider choosing a higher value (on Ubuntu 18.04, 7250 is the max frequency, but it can be increased via /proc/sys/kernel/perf_event_max_sample_rate.

Check out the FlameGraph project and run the following commands to generate the flame graph:

perl stackcollapse-perf.pl ../clamav-devel/out.perf > /tmp/out.folded
perl flamegraph.pl /tmp/out.folded > /tmp/test.svg

The SVG that is generated is interactive, but some viewers don't support this. Be sure to open it in a web browser like Chrome to be able to take full advantage of it.

Profiling - Callgrind

Callgrind is a profiling tool included with valgrind. This can be done by prepending valgrind --tool=callgrind to the clamscan command. kcachegrind is a follow-on tool that will graphically present the profiling data and allow you to explore it visually, although if you don't already use KDE you'll have to install lots of extra packages to use it.

System Call Tracing / Fault Injection

strace can be used to track the system calls that are performed and provide the number of calls / time spent in each system call. This can be done by prepending strace -c to a clamscan command. Results will look something like this:

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
 95.04    0.831430          13     62518           read
  3.22    0.028172          14      2053           munmap
  0.69    0.006005           3      2102           mmap
  0.28    0.002420           7       344           pread64
  0.16    0.001415           5       305         1 openat
  0.13    0.001108           3       405           write
  0.11    0.000932          23        40           mprotect
  0.07    0.000632           2       310           close
  0.07    0.000583           9        67        30 access
  0.05    0.000395           1       444           lseek
  0.04    0.000344           2       162           fstat
  0.04    0.000338           1       253           brk
  0.03    0.000262           1       422           fcntl
  0.02    0.000218          16        14           futex
  0.01    0.000119           1       212           getpid
  0.01    0.000086          14         6           getdents
  0.00    0.000043           7         6           dup
  0.00    0.000040           1        31           unlink
  0.00    0.000038          19         2           rt_sigaction
  0.00    0.000037          19         2           rt_sigprocmask
  0.00    0.000029           1        37           stat
  0.00    0.000022          11         2           prlimit64
  0.00    0.000021          21         1           sysinfo
  0.00    0.000020           1        33           clock_gettime
  0.00    0.000019          19         1           arch_prctl
  0.00    0.000018          18         1           set_tid_address
  0.00    0.000018          18         1           set_robust_list
  0.00    0.000013           0        60           lstat
  0.00    0.000011           0        65           madvise
  0.00    0.000002           0        68           geteuid
  0.00    0.000000           0         1           execve
  0.00    0.000000           0         1           uname
  0.00    0.000000           0         1           getcwd
------ ----------- ----------- --------- --------- ----------------
100.00    0.874790                 69970        31 total

strace can also be used for cool things like system call fault injection. For instance, I was curious whether the 'read' bytecode API call was implemented in such a way that the underlying read system call could handle EINTR being returned (which can happen periodically). To test this, I wrote the following bytecode rule:

VIRUSNAME_PREFIX("BC.Heuristic.Test.Read.Passed")
VIRUSNAMES("")
TARGET(0)

SIGNATURES_DECL_BEGIN
DECLARE_SIGNATURE(zeroes)
SIGNATURES_DECL_END

SIGNATURES_DEF_BEGIN
DEFINE_SIGNATURE(zeroes, "0:0000")
SIGNATURES_DEF_END

bool logical_trigger()
{
    return matches(Signatures.zeroes);
}

#define READ_S(value, size) if (read(value, size) != size) return 0;

int entrypoint(void)
{
    char buffer[65536];
    int i;

    for (i = 0; i < 256; i++)
    {
        debug(i);
        debug("\n");
        READ_S(buffer, sizeof(buffer));
    }

    foundVirus("");
    return 0;
}

I compiled the rule, made a test file to match against, and ran it under strace to determine what underlying read system call was used for the bytecode read function:

clambc-compiler read_test.bc
dd if=/dev/zero of=/tmp/zeroes bs=65535 count=256
strace clamscan -d read_test.cbc --bytecode-unsigned /tmp/zeroes

It uses pread64 under the hood, so the following command could be used for fault injection:

strace -e fault=pread64:error=EINTR:when=20+10 clamscan -d read_test.cbc --bytecode-unsigned /tmp/zeroes 

This command tells strace to skip the first 20 pread64 calls (these appear to be used by the loader, which didn't seem to handle EINTR correctly) but to inject EINTR for every 10th call afterward. We can see the injection in action and that the system call is retried successfully:

pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15007744) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15073280) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15138816) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15204352) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15269888) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15335424) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15400960) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15466496) = 65536
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15532032) = 65536
pread64(3, 0x7f6a7ff43000, 65536, 15597568) = -1 EINTR (Interrupted system call) (INJECTED)
pread64(3, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 65536, 15597568) = 65536

More documentation on this feature can be found in this presentation from FOSDEM 2017.