This applies any mac that might be necessary, ensures that we have
enough space in the WPACKET to perform the encryption and sets up the
SSL3_RECORD ready for that encryption.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
Only tls13_meth.c needs to handle adding record padding. All other
*_meth.c files can ignore it.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
The KTLS cipher function is a no-op so it doesn't matter if we call it.
We shouldn't special case KTLS in tls_common.c
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
We introduce a new function to prepare the record header. KTLS has its own
version since this is done by the kernel.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
Remove TLSv1.3 specific processing of the record type out of tls_common.c
and into tls13_meth.c
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
We move some protocol specific code for write buffer and WPACKET allocation
and initialisation out of tls_common.c and into the protocol specific files.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
This field was used to track whether a cipher ctx was valid for writing
or not, and also whether we should write out plaintext alerts. With the new
record layer design we no longer need to track whether a cipher ctx is valid
since the whole record layer will be aborted if it is not. Also we have a
different mechanism for tracking whether we should write out plaintext
alerts. Therefore this field is removed from the SSL object.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
Most of this was unnecessary anyway since DTLS isn't using these codepaths.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
We also clean up some of the KTLS code while we are doing it now that all
users of KTLS have been moved to the new write record layer.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19343)
We also convert to passing COMP_METHOD rather than SSL_COMP to the record
layer. The former is a public type while the latter is internal only - and
the only thing we need from SSL_COMP is the method.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19217)
Since OPENSSL_malloc() and friends report ERR_R_MALLOC_FAILURE, and
at least handle the file name and line number they are called from,
there's no need to report ERR_R_MALLOC_FAILURE where they are called
directly, or when SSLfatal() and RLAYERfatal() is used, the reason
`ERR_R_MALLOC_FAILURE` is changed to `ERR_R_CRYPTO_LIB`.
There were a number of places where `ERR_R_MALLOC_FAILURE` was reported
even though it was a function from a different sub-system that was
called. Those places are changed to report ERR_R_{lib}_LIB, where
{lib} is the name of that sub-system.
Some of them are tricky to get right, as we have a lot of functions
that belong in the ASN1 sub-system, and all the `sk_` calls or from
the CRYPTO sub-system.
Some extra adaptation was necessary where there were custom OPENSSL_malloc()
wrappers, and some bugs are fixed alongside these changes.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19301)
Move the multiblock code into a separate file and introduce the usage of
record_functions_st for some write functions.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
We use the returned data to decide how to split the data we want to write
into records.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
This also means we can convert SSLfatal calls to RLAYERfatal
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
There were a small number of references to the SSL_CONNECTION that can
be removed easily and replaced with record layer equivalents.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
Previously we were referencing the block_padding value through the
SSL_CONNECTION. Now it is held within OSSL_RECORD_LAYER.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
We wrap the callback and pass it to the record layer via the dispatch
array, in order to avoid accessing it directly via SSL_CONNECTION.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
In all cases we should be able to replace this with a simple check
against rl->version.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
This flag can now be managed entirely by the new record layer code so we
move it into ossl_record_layer_st.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
Now that we are no longer recursively addinng the prefix record this
doesn't seem necessary any more. We always add it every time we do
tls_write_records.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
This calculation is based on lots of information from state machine and
elsewhere that the record layer cannot access. In reality it is sufficient
to simply tell the record layer what version to use.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
We retain a numwpipes for now in the old record layer structure for use
by DTLS. This will eventually be removed when DTLS moves over to the new
way of doing things.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
We use the record layer reference to the BIO rather than the SSL object
reference. This removes an unneeded SSL object usage.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
This removes unnecessary usage of the SSL object from the record layer.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
This isn't a record layer responsibility so should be removed from
write_records.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
Make sure we set the write record layer method and create the object
where appropriate. Move the newly restructured writing code into the
record layer object.
For now we are cheating and still accessing the underlying SSL_CONNECTION
object. This will be removed in subsequent commits.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
The new write record layer architecture splits record writing into
a "write_records" call and a "retry_write_records" call - where multiple
records can be sent to "write_records" in one go. We restructure the code
into that format in order that future commits can move these functions into
the new record layer more easily.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19198)
Leak sanitizer reports following leak for ssl-test-new subtest
4-tlsv1_2-both-compress:
==335733==ERROR: LeakSanitizer: detected memory leaks
Direct leak of 17728 byte(s) in 1 object(s) allocated from:
#0 0x3ff9fbba251 in malloc (/usr/lib64/libasan.so.8+0xba251)
#1 0x3ff9f71744f in tls_do_uncompress ssl/record/methods/tls_common.c:868
#2 0x3ff9f7175bd in tls_default_post_process_record ssl/record/methods/tls_common.c:896
#3 0x3ff9f715ee7 in tls_get_more_records ssl/record/methods/tls_common.c:773
#4 0x3ff9f712209 in tls_read_record ssl/record/methods/tls_common.c:958
#5 0x3ff9f6ef73f in ssl3_read_bytes ssl/record/rec_layer_s3.c:1235
#6 0x3ff9f776165 in tls_get_message_header ssl/statem/statem_lib.c:1198
#7 0x3ff9f74709b in read_state_machine ssl/statem/statem.c:624
#8 0x3ff9f74709b in state_machine ssl/statem/statem.c:478
#9 0x3ff9f662e61 in SSL_do_handshake ssl/ssl_lib.c:4430
#10 0x100c55d in do_handshake_step test/helpers/handshake.c:775
#11 0x100c55d in do_connect_step test/helpers/handshake.c:1134
#12 0x100e85b in do_handshake_internal test/helpers/handshake.c:1544
#13 0x1011715 in do_handshake test/helpers/handshake.c:1738
#14 0x101d1a7 in test_handshake test/ssl_test.c:543
#15 0x1027875 in run_tests test/testutil/driver.c:370
#16 0x1008393 in main test/testutil/main.c:30
#17 0x3ff9cc2b871 in __libc_start_call_main (/usr/lib64/libc.so.6+0x2b871)
#18 0x3ff9cc2b94f in __libc_start_main_alias_2 (/usr/lib64/libc.so.6+0x2b94f)
#19 0x100864f (/code/openssl/test/ssl_test+0x100864f)
Direct leak of 17728 byte(s) in 1 object(s) allocated from:
#0 0x3ff9fbba251 in malloc (/usr/lib64/libasan.so.8+0xba251)
#1 0x3ff9f71744f in tls_do_uncompress ssl/record/methods/tls_common.c:868
#2 0x3ff9f7175bd in tls_default_post_process_record ssl/record/methods/tls_common.c:896
#3 0x3ff9f715ee7 in tls_get_more_records ssl/record/methods/tls_common.c:773
#4 0x3ff9f712209 in tls_read_record ssl/record/methods/tls_common.c:958
#5 0x3ff9f6ef73f in ssl3_read_bytes ssl/record/rec_layer_s3.c:1235
#6 0x3ff9f776165 in tls_get_message_header ssl/statem/statem_lib.c:1198
#7 0x3ff9f74709b in read_state_machine ssl/statem/statem.c:624
#8 0x3ff9f74709b in state_machine ssl/statem/statem.c:478
#9 0x3ff9f662e61 in SSL_do_handshake ssl/ssl_lib.c:4430
#10 0x100c55d in do_handshake_step test/helpers/handshake.c:775
#11 0x100c55d in do_connect_step test/helpers/handshake.c:1134
#12 0x1010b09 in do_handshake_internal test/helpers/handshake.c:1550
#13 0x1011715 in do_handshake test/helpers/handshake.c:1738
#14 0x101d1a7 in test_handshake test/ssl_test.c:543
#15 0x1027875 in run_tests test/testutil/driver.c:370
#16 0x1008393 in main test/testutil/main.c:30
#17 0x3ff9cc2b871 in __libc_start_call_main (/usr/lib64/libc.so.6+0x2b871)
#18 0x3ff9cc2b94f in __libc_start_main_alias_2 (/usr/lib64/libc.so.6+0x2b94f)
#19 0x100864f (/code/openssl/test/ssl_test+0x100864f)
SUMMARY: AddressSanitizer: 35456 byte(s) leaked in 2 allocation(s).
Fix this by freeing the SSL3_RECORD structure inside the OSSL_RECORD_LAYER.
Signed-off-by: Juergen Christ <jchrist@linux.ibm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19030)
When a server responds to a second TLSv1.3 ClientHello it is required to
set the legacy_record_version to 0x0303 (TLSv1.2). The client is required
to ignore that field even if it is wrong. The recent changes to the read
record layer in PR #18132 made the record layer stricter and it was
checking that the legacy_record_version was the correct value. This
caused connection failures when talking to buggy servers that set the
wrong legacy_record_version value.
We make us more tolerant again.
Fixes#19051
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19058)
Otherwise ssl3_cipher() cannot work properly.
Fixes Coverity CID 1509401
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19027)
or on record layer that is to be freed anyway.
Fixes Coverity CID 1509402, 1509403
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19027)
The current libssl code always ensures that the callbacks are non-null.
However, the record layer itself wasn't checkthing this. We ensure it does.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Some minor formatting cleanups and other minor tweaks.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Some macros were redefined in ssl3_cbc.c. We remove the redefinitions
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
The SSL_AD_NO_ALERT value was defined in two places. We centralise its
definition.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Some functions in the record layer were called rlayer_*, but most were
called tls_*. We standardise on the latter.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
The 1 in DTLS1 is confusing and is removed. We also tweak the structure
to always be able to track 64 packets regardless of whether we are on a
32 bit or 64 bit system.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
The value for epoch was being represented internally via various types:
uint16_t, unsigned short, unsigned int, unsigned long
We standardise on uint16_t
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
The read field is no longer used and can be safely removed.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
We would like the capability for the options/mode/read_ahead settings
to be updateable after the record layer object has been instantiated.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
The tls13encryption is an internal test that reaches inside libssl
to test encryption/decryption of records. It needs to be amended for the
new code structure so that it is testing the equivalent things as before.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
This allows querying of the record layer to get a human readable state
string out. This resolves two outstanding TODO comments and enables us
to remove the rstate variable from s->rlayer.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
This resolves an outstanding "TODO" item.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
If read_ahead data is left over when a record layer closes down, there is
already code present to push into the "next" BIO. So the TODO removed here
is no longer relevant.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Now everything has been moved to the record layer the additional check
for setting up the read buffer is not needed and can be removed.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Now that the read record layer has moved to the new architecture we can
clear some of the old stuff away.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
This also fixes ssl3_pending while we are at it
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Some TODO(RECLAYER) comments are no longer necessary and can be removed.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
We had some temporary function pointers in OSSL_RECORD_METHOD which were
only necessary during the process of refactoring the read record layer.
These are no longer required so can be removed.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Remove references to dtls1_* and just use dtls_*
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
We add unprocessed DTLS records to the unprocessed record queue. When
the record layer closes down we write the unprocessed records to the
next record layer object.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Similarly with read_hash and write_hash. In the new model we have a
separate record layer object for reading and writing. Therefore we don't
need to distinguish between reading and writing inside the record layer
object in the encryption and md ctxs.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Don't use the message callback from the SSL object. Instead we use a
wrapper callback so that the record layer does not need to be aware of the
SSL object.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
This removes some references to the SSL object from the record layer.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Pass the max fragment length to the record layer when it is applicable
to avoid the need to go through the SSL object.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
We remove the dependency on the SSL object. Instead we check if the
record layer has already set an alert code.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
We no longer have to go through the SSL object to discover whether EtM has
been negotiated.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Instead we introduce RLAYER_USE_EXPLICIT_IV(). This removes a dependency
on the SSL object.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Remove SSL_IS_TLS13() and hello_retry_request
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
When validate_record_header() gets called it should be able to rely on
the fact that an SSLv2 record version means that the record was received in
SSLv2 format.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Sometimes data read by a record layer in one epoch is actually intended for
the next epoch. For example in a TLS with read_ahead, the read_ahead data
could contain a KeyUpdate message followed by application data encrypted
with new keys. Therefore we implement a mechanism for passing this data
across the epochs.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
This also moves other protocol specific code to the protocol specific
files.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18132)
Matt Caswell
Richard Levitte
Tomas Mraz
Juergen Christ