This change upgrades to GCC 12.3 and GNU binutils 2.42. The GNU linker
appears to have changed things so that only a single de-duplicated str
table is present in the binary, and it gets placed wherever the linker
wants, regardless of what the linker script says. To cope with that we
need to stop using .ident to embed licenses. As such, this change does
significant work to revamp how third party licenses are defined in the
codebase, using `.section .notice,"aR",@progbits`.
This new GCC 12.3 toolchain has support for GNU indirect functions. It
lets us support __target_clones__ for the first time. This is used for
optimizing the performance of libc string functions such as strlen and
friends so far on x86, by ensuring AVX systems favor a second codepath
that uses VEX encoding. It shaves some latency off certain operations.
It's a useful feature to have for scientific computing for the reasons
explained by the test/libcxx/openmp_test.cc example which compiles for
fifteen different microarchitectures. Thanks to the upgrades, it's now
also possible to use newer instruction sets, such as AVX512FP16, VNNI.
Cosmo now uses the %gs register on x86 by default for TLS. Doing it is
helpful for any program that links `cosmo_dlopen()`. Such programs had
to recompile their binaries at startup to change the TLS instructions.
That's not great, since it means every page in the executable needs to
be faulted. The work of rewriting TLS-related x86 opcodes, is moved to
fixupobj.com instead. This is great news for MacOS x86 users, since we
previously needed to morph the binary every time for that platform but
now that's no longer necessary. The only platforms where we need fixup
of TLS x86 opcodes at runtime are now Windows, OpenBSD, and NetBSD. On
Windows we morph TLS to point deeper into the TIB, based on a TlsAlloc
assignment, and on OpenBSD/NetBSD we morph %gs back into %fs since the
kernels do not allow us to specify a value for the %gs register.
OpenBSD users are now required to use APE Loader to run Cosmo binaries
and assimilation is no longer possible. OpenBSD kernel needs to change
to allow programs to specify a value for the %gs register, or it needs
to stop marking executable pages loaded by the kernel as mimmutable().
This release fixes __constructor__, .ctor, .init_array, and lastly the
.preinit_array so they behave the exact same way as glibc.
We no longer use hex constants to define math.h symbols like M_PI.
At least in neovim, `│vi:` is not recognized as a modeline because it
has no preceding whitespace. After fixing this, opening a file yields
an error because `net` is not an option. (`noet`, however, is.)
This change fixes stderr to be unbuffered. Added hardware AES on ARM64
to help safeguard against timing attacks. The curl.com command will be
somewhat more pleasant to use.
This change fixes a nasty regression caused by
80b211e314 which deadlocked.
This change also causes MbedTLS to prefer the ChaCha ciphersuite on
older CPUs that don't have AES hardware instructions.
This change enables SSL compression. It significantly reduces the
network load of the testing infrastructure, for free, since this
revision didn't need to change any runit protocol code. However we
turn it off by default in redbean since no browsers support it.
It turns out that some TLSv1.0 clients (e.g. curl command on RHEL5) will
send an SSLv2-style ClientHello. These types of clients are usually ten+
years old and were designed to interop with servers ten years older than
them. Your redbean is now able to interop with these clients even though
redbean doesn't actually support SSLv2 or SSLv3. Please note that the -B
flag may be passed to disable this along with TLSv1.0, TLSv1.1, 3DES, &c
The following Lua APIs have been added to redbean:
- ProgramSslCompression(bool)
- ProgramSslCiphersuite(name:str)
- ProgramSslPresharedKey(key:str,identity:str)
Lastly the DHE ciphersuites have been enabled. IANA recommends DHE and
with old clients like RHEL5 it's the only perfect forward secrecy they
implement.
redbean will now set Referer-Policy to no-referrer-when-downgrade on
text/html responses by default. There's better explanations on the bits
of security redbean is offering. In short, it's 128+ for modern clients
and 112+ for legacy. If the -B flag is used then it's 192+ for modern
and 150+ for non-EC.
One of the disadvantages of x25519 and ℘256 is it only provides 126 bits
of security, so that seems like a weak link in the chain, if we're using
ECDHE-ECDSA-AES256-GCM-SHA384. The U.S. government wants classified data
to be encrypted using a curve at least as strong as ℘384, which provides
192 bits of security, but if you read the consensus of stack exchange it
would give you the impression that ℘384 is three times slower.
This change (as well as the previous one) makes ℘384 three times as fast
by tuning its modulus and multiplication subroutines with new tests that
should convincingly show: the optimized code behaves the same way as the
old code. Some of the diff noise from the previous change is now removed
too, so that our vendored fork can be more easily compared with upstream
sources. So you can now have stronger cryptography without compromises.
℘384 modulus Justine l: 28𝑐 9𝑛𝑠
℘384 modulus MbedTLS NIST l: 127𝑐 41𝑛𝑠
℘384 modulus MbedTLS MPI l: 1,850𝑐 597𝑛𝑠
The benchmarks above show the improvements made by secp384r1() which is
an important function since it needs to be called 13,000 times whenever
someone establishes a connection to your web server. The same's true of
Mul6x6Adx() which is able to multiply 384-bit numbers in 73 cycles, but
only if your CPU was purchased after 2014 when Broadwell was introduced
This change makes SSL virtual hosting possible. You can now load
multiple certificates for multiple domains and redbean will just
figure out which one to use, even if you only have 1 ip address.
You can also use a jumbo certificate that lists all your domains
in the the subject alternative names.
This change also makes performance improvements to MbedTLS. Here
are some benchmarks vs. cc1920749e
BEFORE AFTER (microsecs)
suite_ssl.com 2512881 191738 13.11x faster
suite_pkparse.com 36291 3295 11.01x faster
suite_x509parse.com 854669 120293 7.10x faster
suite_pkwrite.com 6549 1265 5.18x faster
suite_ecdsa.com 53347 18778 2.84x faster
suite_pk.com 49051 18717 2.62x faster
suite_ecdh.com 19535 9502 2.06x faster
suite_shax.com 15848 7965 1.99x faster
suite_rsa.com 353257 184828 1.91x faster
suite_x509write.com 162646 85733 1.90x faster
suite_ecp.com 20503 11050 1.86x faster
suite_hmac_drbg.no_reseed.com 19528 11417 1.71x faster
suite_hmac_drbg.nopr.com 12460 8010 1.56x faster
suite_mpi.com 687124 442661 1.55x faster
suite_hmac_drbg.pr.com 11890 7752 1.53x faster
There aren't any special tricks to the performance imporvements.
It's mostly due to code cleanup, assembly and intel instructions
like mulx, adox, and adcx.
This change boosts SSL handshake performance from 2,627 to ~10,000 per
second which is the same level of performance as NGINX at establishing
secure connections. That's impressive if we consider that redbean is a
forking frontend application server. This was accomplished by:
1. Enabling either SSL session caching or SSL tickets. We choose to
use tickets since they reduce network round trips too and that's
a more important metric than wrk'ing localhost.
2. Fixing mbedtls_mpi_sub_abs() which is the most frequently called
function. It's called about 12,000 times during an SSL handshake
since it's the basis of most arithmetic operations like addition
and for some strange reason it was designed to make two needless
copies in addition to calling malloc and free. That's now fixed.
3. Improving TLS output buffering during the SSL handshake only, so
that only a single is write and read system call is needed until
blocking on the ping pong.
redbean will now do a better job wiping sensitive memory from a child
process as soon as it's not needed. The nice thing about fork is it's
much faster than reverse proxying so the goal is to use the different
address spaces along with setuid() to minimize the risk that a server
key will be compromised in the event that application code is hacked.
Your redbean can now interoperate with clients that require TLS crypto.
This is accomplished using a protocol polyglot that lets us distinguish
between HTTP and HTTPS regardless of the port number. Certificates will
be generated automatically, if none are supplied by the user. Footprint
increases by only a few hundred kb so redbean in MODY=tiny is now 1.0mb
- Add lseek() polyfills for ZIP executable
- Automatically polyfill /tmp/FOO paths on NT
- Fix readdir() / ftw() / nftw() bugs on Windows
- Introduce -B flag for slower SSL that's stronger
- Remove mbedtls features Cosmopolitan doesn't need
- Have base64 decoder support the uri-safe alternative
- Remove Truncated HMAC because it's forbidden by the IETF
- Add all the mbedtls test suites and make them go 3x faster
- Support opendir() / readdir() / closedir() on ZIP executable
- Use Everest for ECDHE-ECDSA because it's so good it's so good
- Add tinier implementation of sha1 since it's not worth the rom
- Add chi-square monte-carlo mean correlation tests for getrandom()
- Source entropy on Windows from the proper interface everyone uses
We're continuing to outperform NGINX and other servers on raw message
throughput. Using SSL means that instead of 1,000,000 qps you can get
around 300,000 qps. However redbean isn't as fast as NGINX yet at SSL
handshakes, since redbean can do 2,627 per second and NGINX does 4.3k
Right now, the SSL UX story works best if you give your redbean a key
signing key since that can be easily generated by openssl using a one
liner then redbean will do all the things that are impossibly hard to
do like signing ecdsa and rsa certificates that'll work in chrome. We
should integrate the let's encrypt acme protocol in the future.
Live Demo: https://redbean.justine.lol/
Root Cert: https://redbean.justine.lol/redbean1.crt
