The SSL/TLS module maintains a number of internally used pointers
`out_hdr`, `out_len`, `out_iv`, ..., indicating where to write the
various parts of the record header.
These pointers have to be kept in sync and sometimes need update:
Most notably, the `out_msg` pointer should always point to the
beginning of the record payload, and its offset from the pointer
`out_iv` pointing to the end of the record header is determined
by the length of the explicit IV used in the current record
protection mechanism.
This commit introduces functions deducing these pointers from
the pointers `out_hdr` / `in_hdr` to the beginning of the header
of the current outgoing / incoming record.
The flexibility gained by these functions will subsequently
be used to allow shifting of `out_hdr` for the purpose of
packing multiple records into a single datagram.
For now, just check that it causes us to fragment. More tests are coming in
follow-up commits to ensure we respect the exact value set, including when
renegotiating.
Note: no interop tests in ssl-opt.sh for now, as some of them make us run into
bugs in (the CI's default versions of) OpenSSL and GnuTLS, so interop tests
will be added later once the situation is clarified. <- TODO
This will allow fragmentation to always happen in the same place, always from
a buffer distinct from ssl->out_msg, and with the same way of resuming after
returning WANT_WRITE
- take advantage of the fact that we're only called for first send
- put all sanity checks at the top
- rename and constify shortcut variables
- improve comments
`mbedtls_ssl_get_record_expansion()` is supposed to return the maximum
difference between the size of a protected record and the size of the
encapsulated plaintext.
It had the following two bugs:
(1) It did not consider the new ChaChaPoly ciphersuites, returning
the error code #MBEDTLS_ERR_SSL_INTERNAL_ERROR in this case.
(2) It did not correctly estimate the maximum record expansion in case
of CBC ciphersuites in (D)TLS versions 1.1 and higher, in which
case the ciphertext is prefixed by an explicit IV.
This commit fixes both bugs.
The length to the debug message could conceivably leak through the time it
takes to print it, and that length would in turn reveal whether padding was
correct or not.
The basis for the Lucky 13 family of attacks is for an attacker to be able to
distinguish between (long) valid TLS-CBC padding and invalid TLS-CBC padding.
Since our code sets padlen = 0 for invalid padding, the length of the input to
the HMAC function, and the location where we read the MAC, give information
about that.
A local attacker could gain information about that by observing via a
cache attack whether the bytes at the end of the record (at the location of
would-be padding) have been read during MAC verification (computation +
comparison).
Let's make sure they're always read.
For the situation where the mbedTLS device has limited RAM, but the
other end of the connection doesn't support the max_fragment_length
extension. To be spec-compliant, mbedTLS has to keep a 16384 byte
incoming buffer. However the outgoing buffer can be made smaller without
breaking spec compliance, and we save some RAM.
See comments in include/mbedtls/config.h for some more details.
(The lower limit of outgoing buffer size is the buffer size used during
handshake/cert negotiation. As the handshake is half-duplex it might
even be possible to store this data in the "incoming" buffer during the
handshake, which would save even more RAM - but it would also be a lot
hackier and error-prone. I didn't really explore this possibility, but
thought I'd mention it here in case someone sees this later on a mission
to jam mbedTLS into an even tinier RAM footprint.)
As a protection against the Lucky Thirteen attack, the TLS code for
CBC decryption in encrypt-then-MAC mode performs extra MAC
calculations to compensate for variations in message size due to
padding. The amount of extra MAC calculation to perform was based on
the assumption that the bulk of the time is spent in processing
64-byte blocks, which is correct for most supported hashes but not for
SHA-384. Correct the amount of extra work for SHA-384 (and SHA-512
which is currently not used in TLS, and MD2 although no one should
care about that).
* development: (504 commits)
Fix minor code style issues
Add the uodate to the soversion to the ChangeLog
Fix the ChangeLog for clarity, english and credit
Update version to 2.9.0
ecp: Fix binary compatibility with group ID
Changelog entry
Change accepted ciphersuite versions when parsing server hello
Remove preprocessor directives around platform_util.h include
Fix style for mbedtls_mpi_zeroize()
Improve mbedtls_platform_zeroize() docs
mbedtls_zeroize -> mbedtls_platform_zeroize in docs
Reword config.h docs for MBEDTLS_PLATFORM_ZEROIZE_ALT
Organize CMakeLists targets in alphabetical order
Organize output objs in alfabetical order in Makefile
Regenerate errors after ecp.h updates
Update ecp.h
Change variable bytes_written to header_bytes in record decompression
Update ecp.h
Update ecp.h
Update ecp.h
...
Rename to mbedtls_ssl_get_async_operation_data and
mbedtls_ssl_set_async_operation_data so that they're about
"async operation data" and not about some not-obvious "data".
When a handshake step starts an asynchronous operation, the
application needs to know which SSL connection the operation is for,
so that when the operation completes, the application can wake that
connection up. Therefore the async start callbacks need to take the
SSL context as an argument. It isn't enough to let them set a cookie
in the SSL connection, the application needs to be able to find the
right SSL connection later.
Also pass the SSL context to the other callbacks for consistency. Add
a new field to the handshake that the application can use to store a
per-connection context. This new field replaces the former
context (operation_ctx) that was created by the start function and
passed to the resume function.
Add a boolean flag to the handshake structure to track whether an
asynchronous operation is in progress. This is more robust than
relying on the application to set a non-null application context.
Change the signature of mbedtls_ssl_handshake_free again. Now take the
whole SSL context as argument and not just the configuration and the
handshake substructure.
This is in preparation for changing the asynchronous cancel callback
to take the SSL context as an argument.
Conflict resolution:
* ChangeLog: put the new entry from my branch in the proper place.
* include/mbedtls/error.h: counted high-level module error codes again.
* include/mbedtls/ssl.h: picked different numeric codes for the
concurrently added errors; made the new error a full sentence per
current standards.
* library/error.c: ran scripts/generate_errors.pl.
* library/ssl_srv.c:
* ssl_prepare_server_key_exchange "DHE key exchanges": the conflict
was due to style corrections in development
(4cb1f4d49c) which I merged with
my refactoring.
* ssl_prepare_server_key_exchange "For key exchanges involving the
server signing", first case, variable declarations: merged line
by line:
* dig_signed_len: added in async
* signature_len: removed in async
* hashlen: type changed to size_t in development
* hash: size changed to MBEDTLS_MD_MAX_SIZE in async
* ret: added in async
* ssl_prepare_server_key_exchange "For key exchanges involving the
server signing", first cae comment: the conflict was due to style
corrections in development (4cb1f4d49c)
which I merged with my comment changes made as part of refactoring
the function.
* ssl_prepare_server_key_exchange "Compute the hash to be signed" if
`md_alg != MBEDTLS_MD_NONE`: conflict between
ebd652fe2d
"ssl_write_server_key_exchange: calculate hashlen explicitly" and
46f5a3e9b4 "Check return codes from
MD in ssl code". I took the code from commit
ca1d742904 made on top of development
which makes mbedtls_ssl_get_key_exchange_md_ssl_tls return the
hash length.
* programs/ssl/ssl_server2.c: multiple conflicts between the introduction
of MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS and new auxiliary functions and
definitions for async support, and the introduction of idle().
* definitions before main: concurrent additions, kept both.
* main, just after `handshake:`: in the loop around
mbedtls_ssl_handshake(), merge the addition of support for
MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS and SSL_ASYNC_INJECT_ERROR_CANCEL
with the addition of the idle() call.
* main, if `opt.transport == MBEDTLS_SSL_TRANSPORT_STREAM`: take the
code from development and add a check for
MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS.
* main, loop around mbedtls_ssl_read() in the datagram case:
take the code from development and add a check for
MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS; revert to a do...while loop.
* main, loop around mbedtls_ssl_write() in the datagram case:
take the code from development and add a check for
MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS; revert to a do...while loop.
In mbedtls_ssl_get_key_exchange_md_tls1_2, add an output parameter for
the hash length. The code that calls this function can currently do
without it, but it will need the hash length in the future, when
adding support for a third-party callback to calculate the signature
of the hash.
New compile-time option MBEDTLS_SSL_ASYNC_PRIVATE_C, enabling
callbacks to replace private key operations. These callbacks allow the
SSL stack to make an asynchronous call to an external cryptographic
module instead of calling the cryptography layer inside the library.
The call is asynchronous in that it may return the new status code
MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS, in which case the SSL stack returns
and can be later called where it left off.
This commit introduces the configuration option. Later commits will
implement the feature proper.
This function is declared in ssl_internal.h, so this is not a public
API change.
This is in preparation for mbedtls_ssl_handshake_free needing to call
methods from the config structure.
This commit removes all the static occurrencies of the function
mbedtls_zeroize() in each of the individual .c modules. Instead the
function has been moved to utils.h that is included in each of the
modules.
In mbedtls_ssl_derive_keys, don't call mbedtls_md_hmac_starts in
ciphersuites that don't use HMAC. This doesn't change the behavior of
the code, but avoids relying on an uncaught error when attempting to
start an HMAC operation that hadn't been initialized.
The _ext suffix suggests "new arguments", but the new functions have
the same arguments. Use _ret instead, to convey that the difference is
that the new functions return a value.
Conflict resolution:
* ChangeLog: put the new entries in their rightful place.
* library/x509write_crt.c: the change in development was whitespace
only, so use the one from the iotssl-1251 feature branch.
A previous commit changed the record encryption function
`ssl_encrypt_buf` to compute the MAC in a temporary buffer
and copying the relevant part of it (which is strictly smaller
if the truncated HMAC extension is used) to the outgoing message
buffer. However, the change was only made in case Encrypt-Then-MAC
was enabled, but not in case of MAC-Then-Encrypt. While this
doesn't constitute a problem, for the sake of uniformity this
commit changes `ssl_encrypt_buf` to compute the MAC in a temporary
buffer in this case, too.
* restricted/pr/403:
Correct record header size in case of TLS
Don't allocate space for DTLS header if DTLS is disabled
Improve debugging output
Adapt ChangeLog
Add run-time check for handshake message size in ssl_write_record
Add run-time check for record content size in ssl_encrypt_buf
Add compile-time checks for size of record content and payload
In a previous PR (Fix heap corruption in implementation of truncated HMAC
extension #425) the place where MAC is computed was changed from the end of
the SSL I/O buffer to a local buffer (then (part of) the content of the local
buffer is either copied to the output buffer of compare to the input buffer).
Unfortunately, this change was made only for TLS 1.0 and later, leaving SSL
3.0 in an inconsistent state due to ssl_mac() still writing to the old,
hard-coded location, which, for MAC verification, resulted in later comparing
the end of the input buffer (containing the computed MAC) to the local buffer
(uninitialised), most likely resulting in MAC verification failure, hence no
interop (even with ourselves).
This commit completes the move to using a local buffer by using this strategy
for SSL 3.0 too. Fortunately ssl_mac() was static so it's not a problem to
change its signature.
In case truncated HMAC must be used but the Mbed TLS peer hasn't been updated
yet, one can use the compile-time option MBEDTLS_SSL_TRUNCATED_HMAC_COMPAT to
temporarily fall back to the old, non-compliant implementation of the truncated
HMAC extension.
The truncated HMAC extension as described in
https://tools.ietf.org/html/rfc6066.html#section-7 specifies that when truncated
HMAC is used, only the HMAC output should be truncated, while the HMAC key
generation stays unmodified. This commit fixes Mbed TLS's behavior of also
truncating the key, potentially leading to compatibility issues with peers
running other stacks than Mbed TLS.
Details:
The keys for the MAC are pieces of the keyblock that's generated from the
master secret in `mbedtls_ssl_derive_keys` through the PRF, their size being
specified as the size of the digest used for the MAC, regardless of whether
truncated HMAC is enabled or not.
/----- MD size ------\ /------- MD size ----\
Keyblock +----------------------+----------------------+------------------+---
now | MAC enc key | MAC dec key | Enc key | ...
(correct) +----------------------+----------------------+------------------+---
In the previous code, when truncated HMAC was enabled, the HMAC keys
were truncated to 10 bytes:
/-10 bytes-\ /-10 bytes-\
Keyblock +-------------+-------------+------------------+---
previously | MAC enc key | MAC dec key | Enc key | ...
(wrong) +-------------+-------------+------------------+---
The reason for this was that a single variable `transform->maclen` was used for
both the keysize and the size of the final MAC, and its value was reduced from
the MD size to 10 bytes in case truncated HMAC was negotiated.
This commit fixes this by introducing a temporary variable `mac_key_len` which
permanently holds the MD size irrespective of the presence of truncated HMAC,
and using this temporary to obtain the MAC key chunks from the keyblock.
Previously, MAC validation for an incoming record proceeded as follows:
1) Make a copy of the MAC contained in the record;
2) Compute the expected MAC in place, overwriting the presented one;
3) Compare both.
This resulted in a record buffer overflow if truncated MAC was used, as in this
case the record buffer only reserved 10 bytes for the MAC, but the MAC
computation routine in 2) always wrote a full digest.
For specially crafted records, this could be used to perform a controlled write of
up to 6 bytes past the boundary of the heap buffer holding the record, thereby
corrupting the heap structures and potentially leading to a crash or remote code
execution.
This commit fixes this by making the following change:
1) Compute the expected MAC in a temporary buffer that has the size of the
underlying message digest.
2) Compare to this to the MAC contained in the record, potentially
restricting to the first 10 bytes if truncated HMAC is used.
A similar fix is applied to the encryption routine `ssl_encrypt_buf`.
* development: (30 commits)
update README file (#1144)
Fix typo in asn1.h
Improve leap year test names in x509parse.data
Correctly handle leap year in x509_date_is_valid()
Renegotiation: Add tests for SigAlg ext parsing
Parse Signature Algorithm ext when renegotiating
Minor style fix
config.pl get: be better behaved
config.pl get: don't rewrite config.h; detect write errors
Fixed "config.pl get" for options with no value
Fix typo and bracketing in macro args
Ensure failed test_suite output is sent to stdout
Remove use of GNU sed features from ssl-opt.sh
Fix typos in ssl-opt.sh comments
Add ssl-opt.sh test to check gmt_unix_time is good
Extend ssl-opt.h so that run_test takes function
Always print gmt_unix_time in TLS client
Restored note about using minimum functionality in makefiles
Note in README that GNU make is required
Fix changelog for ssl_server2.c usage fix
...
Previously, if `MBEDTLS_SSL_RENEGOTIATION` was disabled, incoming handshake
messages in `mbedtls_ssl_read` (expecting application data) lead to the
connection being closed. This commit fixes this, restricting the
`MBEDTLS_SSL_RENEGOTIATION`-guard to the code-paths responsible for accepting
renegotiation requests and aborting renegotiation attempts after too many
unexpected records have been received.
This commit reconciles the code path responsible for resending the
final DTLS handshake flight with the path for handling resending of
the other flights.
DTLS records from previous epochs were incorrectly checked against the
current epoch transform's minimal content length, leading to the
rejection of entire datagrams. This commit fixed that and adapts two
test cases accordingly.
Internal reference: IOTSSL-1417
- Enhances the documentation of mbedtls_ssl_get_bytes_avail (return
the number of bytes left in the current application data record, if
there is any).
- Introduces a new public function mbedtls_ssl_check_pending for
checking whether any data in the internal buffers still needs to be
processed. This is necessary for users implementing event-driven IO
to decide when they can safely idle until they receive further
events from the underlying transport.
Give a note on the debugging output on the following occasions:
(1) The timer expires in mbedtls_ssl_fetch_input
(2) There's more than one records within a single datagram
This commit fixes the following case: If a client is both expecting a
SERVER_HELLO and has an application data record that's partially
processed in flight (that's the situation the client gets into after
receiving a ServerHelloRequest followed by ApplicationData), a
subsequent call to mbedtls_ssl_read will set keep_current_message = 1
when seeing the unexpected application data, but not reset it to 0
after the application data has been processed. This commit fixes this.
It also documents and suggests how the problem might be solved in a
more structural way on the long run.
This commit adds a hard assertion to mbedtls_ssl_read_record_layer
triggering if both ssl->in_hslen and ssl->in_offt are not 0. This
should never happen, and if it does, there's no sensible way of
telling whether the previous message was a handshake or an application
data message.
There are situations in which it is not clear what message to expect
next. For example, the message following the ServerHello might be
either a Certificate, a ServerKeyExchange or a CertificateRequest. We
deal with this situation in the following way: Initially, the message
processing function for one of the allowed message types is called,
which fetches and decodes a new message. If that message is not the
expected one, the function returns successfully (instead of throwing
an error as usual for unexpected messages), and the handshake
continues to the processing function for the next possible message. To
not have this function fetch a new message, a flag in the SSL context
structure is used to indicate that the last message was retained for
further processing, and if that's set, the following processing
function will not fetch a new record.
This commit simplifies the usage of this message-retaining parameter
by doing the check within the record-fetching routine instead of the
specific message-processing routines. The code gets cleaner this way
and allows retaining messages to be used in other situations as well
without much effort. This will be used in the next commits.
This commit adds four tests to tests/ssl-opt.sh:
(1) & (2): Check behaviour of optional/required verification when the
trusted CA chain is empty.
(3) & (4): Check behaviour of optional/required verification when the
client receives a server certificate with an unsupported curve.
This commit changes the behaviour of mbedtls_ssl_parse_certificate
to make the two authentication modes MBEDTLS_SSL_VERIFY_REQUIRED and
MBEDTLS_SSL_VERIFY_OPTIONAL be in the following relationship:
Mode == MBEDTLS_SSL_VERIFY_REQUIRED
<=> Mode == MBEDTLS_SSL_VERIFY_OPTIONAL + check verify result
Also, it changes the behaviour to perform the certificate chain
verification even if the trusted CA chain is empty. Previously, the
function failed in this case, even when using optional verification,
which was brought up in #864.
* gilles/IOTSSL-1330/development:
Changelog entry for the bug fixes
SSLv3: when refusing renegotiation, stop processing
Ignore failures when sending fatal alerts
Cleaned up double variable declaration
Code portability fix
Added changelog entry
Send TLS alerts in many more cases
Skip all non-executables in run-test-suites.pl
SSL tests: server requires auth, client has no certificate
Balanced braces across preprocessor conditionals
Support setting the ports on the command line
By default, keep allowing SHA-1 in key exchange signatures. Disabling
it causes compatibility issues, especially with clients that use
TLS1.2 but don't send the signature_algorithms extension.
SHA-1 is forbidden in certificates by default, since it's vulnerable
to offline collision-based attacks.
Default to forbidding the use of SHA-1 in TLS where it is unsafe: for
certificate signing, and as the signature hash algorithm for the TLS
1.2 handshake signature. SHA-1 remains allowed in HMAC-SHA-1 in the
XXX_SHA ciphersuites and in the PRF for TLS <= 1.1.
For easy backward compatibility for use in controlled environments,
turn on the MBEDTLS_TLS_DEFAULT_ALLOW_SHA1 compiled-time option.
* hanno/sig_hash_compatibility:
Improve documentation
Split long lines
Remember suitable hash function for any signature algorithm.
Introduce macros and functions to characterize certain ciphersuites.
According to RFC5246 the server can indicate the known Certificate
Authorities or can constrain the aurhorisation space by sending a
certificate list. This part of the message is optional and if omitted,
the client may send any certificate in the response.
The previous behaviour of mbed TLS was to always send the name of all the
CAs that are configured as root CAs. In certain cases this might cause
usability and privacy issues for example:
- If the list of the CA names is longer than the peers input buffer then
the handshake will fail
- If the configured CAs belong to third parties, this message gives away
information on the relations to these third parties
Therefore we introduce an option to suppress the CA list in the
Certificate Request message.
Providing this feature as a runtime option comes with a little cost in
code size and advantages in maintenance and flexibility.
This commit changes `ssl_parse_signature_algorithms_ext` to remember
one suitable ( := supported by client and by our config ) hash
algorithm per signature algorithm.
It also modifies the ciphersuite checking function
`ssl_ciphersuite_match` to refuse a suite if there
is no suitable hash algorithm.
Finally, it adds the corresponding entry to the ChangeLog.
In many places in TLS handling, some code detects a fatal error, sends
a fatal alert message, and returns to the caller. If sending the alert
fails, then return the error that triggered the alert, rather than
overriding the return status. This effectively causes alert sending
failures to be ignored. Formerly the code was inconsistently sometimes
doing one, sometimes the other.
In general ignoring the alert is the right thing: what matters to the
caller is the original error. A typical alert failure is that the
connection is already closed.
One case which remains not handled correctly is if the alert remains
in the output buffer (WANT_WRITE). Then it won't be sent, or will be
truncated. We'd need to either delay the application error or record
the write buffering notice; to be done later.
The TLS client and server code was usually closing the connection in
case of a fatal error without sending an alert. This commit adds
alerts in many cases.
Added one test case to detect that we send the alert, where a server
complains that the client's certificate is from an unknown CA (case
tracked internally as IOTSSL-1330).
Fix an incorrect condition in ssl_check_ctr_renegotiate() that compared
64 bits of record counter instead of 48 bits as described in RFC 6347
Section 4.3.1. This would cause the function's return value to be
occasionally incorrect and the renegotiation routines to be triggered
at unexpected times.
Fixes many typos, and errors in comments.
* Clarifies many comments
* Grammar correction in config.pl help text
* Removed comment about MBEDTLS_X509_EXT_NS_CERT_TYPE.
* Comment typo fix (Dont => Don't)
* Comment typo fix (assure => ensure)
* Comment typo fix (byes => bytes)
* Added citation for quoted standard
* Comment typo fix (one complement => 1's complement)
The is some debate about whether to prefer "one's complement", "ones'
complement", or "1's complement". The more recent RFCs related to TLS
(RFC 6347, RFC 4347, etc) use " 1's complement", so I followed that
convention.
* Added missing ")" in comment
* Comment alignment
* Incorrect comment after #endif
Certificates with unsupported algorithms in the certificate chain
prevented verification even if a certificate before the unsupported
ones was already trusted.
We change the behaviour to ignoring every certificate with unknown
(unsupported) signature algorithm oid when parsing the certificate
chain received from the peer.
