This should only be done in the top-level function.
Also, we need to know if we indeed are the top-level function or not: for
example, when mbedtls_ecp_muladd() calls mbedtls_ecp_mul(), the later should
not reset ops_done. This is handled by the "depth" parameter in the restart
context.
When a restartable function calls another restartable function, the current
ops_count needs to be shared to avoid either doing too many operations or
returning IN_PROGRESS uselessly. So it needs to be in the top-level context
rather than a specific sub-context.
This was intended to detect aborted operations, but now that case is handled
by the caller freeing the restart context.
Also, as the internal sub-context is managed by the callee, no need for the
caller to free/reset the restart context between successful calls.
Following discussion in the team, it was deemed preferable for the restart
context to be explicitly managed by the caller.
This commits in the first in a series moving in that directly: it starts by
only changing the public API, while still internally using the old design.
Future commits in that series will change to the new design internally.
The test function was simplified as it no longer makes sense to test for some
memory management errors since that responsibility shifted to the caller.
It's going to be convenient for each function that can generate a
MBEDTLS_ERR_ECP_IN_PROGRESS on its own (as opposed to just passing it around)
to have its own restart context that they can allocate and free as needed
independently of the restart context of other functions.
For example ecp_muladd() is going to have its own restart_muladd context that
in can managed, then when it calls ecp_mul() this will manage a restart_mul
context without interfering with the caller's context.
So, things need to be renames to avoid future name clashes.
From a user's perspective, you want a "basic operation" to take approximately
the same amount of time regardless of the curve size, especially since max_ops
is a global setting: otherwise if you pick a limit suitable for P-384 then
when you do an operation on P-256 it will return way more often than needed.
Said otherwise, a user is actually interested in actual running time, and we
do the API in terms of "basic ops" for practical reasons (no timers) but then
we should make sure it's a good proxy for running time.
Ok, so the original plan was to make mpi_inv_mod() the smallest block that
could not be divided. Updated plan is that the smallest block will be either:
- ecp_normalize_jac_many() (one mpi_inv_mod() + a number or mpi_mul_mpi()s)
- or the second loop in ecp_precompute_comb()
With default settings, the minimum non-restartable sequence is:
- for P-256: 222M
- for P-384: 341M
This is within a 2-3x factor of originally planned value of 120M. However,
that value can be approached, at the cost of some performance, by setting
ECP_WINDOW_SIZE (w below) lower than the default of 6. For example:
- w=4 -> 166M for any curve (perf. impact < 10%)
- w=2 -> 130M for any curve (perf. impact ~ 30%)
My opinion is that the current state with w=4 is a good compromise, and the
code complexity need to attain 120M is not warranted by the 1.4 factor between
that and the current minimum with w=4 (which is close to optimal perf).
This is the easy part: with the current steps, all information between steps
is passed via T which is already saved. Next we'll need to split at least the
first loop, and maybe calls to normalize_jac_many() and/or the second loop.
Separating main computation from filling of the auxiliary array makes things
clearer and easier to restart as we don't have to remember the in-progress
auxiliary array.
Previously there were only two states:
- T unallocated
- T allocated and valid
Now there are three:
- T unallocated
- T allocated and in progress
- T allocated and valid
Introduce new bool T_ok to distinguish the last two states.
Free it as soon as it's no longer needed, but as a backup free it in
ecp_group_free(), in case ecp_mul() is not called again after returning
ECP_IN_PROGRESS.
So far we only remember it when it's fully computed, next step is to be able
to compute it in multiple steps.
In case of argument change, freeing everything is not the most efficient
(wastes one free()+calloc()) but makes the code simpler, which is probably
more important here
We'll need to store MPIs and other things that allocate memory in this
context, so we need a place to free it. We can't rely on doing it before
returning from ecp_mul() as we might return MBEDTLS_ERR_ECP_IN_PROGRESS (thus
preserving the context) and never be called again (for example, TLS handshake
aborted for another reason). So, ecp_group_free() looks like a good place to
do this, if the restart context is part of struct ecp_group.
This means it's not possible to use the same ecp_group structure in different
threads concurrently, but:
- that's already the case (and documented) for other reasons
- this feature is precisely intended for environments that lack threading
An alternative option would be for the caller to have to allocate/free the
restart context and pass it explicitly, but this means creating new functions
that take a context argument, and putting a burden on the user.
The plan is to count basic operations as follows:
- call to ecp_add_mixed() -> 11
- call to ecp_double_jac() -> 8
- call to mpi_mul_mpi() -> 1
- call to mpi_inv_mod() -> 120
- everything else -> not counted
The counts for ecp_add_mixed() and ecp_double_jac() are based on the actual
number of calls to mpi_mul_mpi() they they make.
The count for mpi_inv_mod() is based on timing measurements on K64F and
LPC1768 boards, and are consistent with the usual very rough estimate of one
inversion = 100 multiplications. It could be useful to repeat that measurement
on a Cortex-M0 board as those have smaller divider and multipliers, so the
result could be a bit different but should be the same order of magnitude.
The documented limitation of 120 basic ops is due to the calls to mpi_inv_mod()
which are currently not interruptible nor planned to be so far.
This is the first step towards making verify_chain() iterative. While from a
readability point of view the current recursive version is fine, one of the
goals of this refactoring is to prepare for restartable ECC integration, which
will need the explicit stack anyway.
Besides avoiding near-duplication, this avoids having three generations of
certificate (child, parent, grandparent) in one function, with all the
off-by-one opportunities that come with it.
This also allows to simplify the signature of verify_child(), which will be
done in next commit.
This is from the morally 5th (and soon obsolete) invocation of this function
in verify_top().
Doing this badtime-skipping when we search for a parent in the provided chain
is a change of behaviour, but it's backwards-compatible: it can only cause us
to accept valid chains that we used to reject before. Eg if the peer has a
chain with two version of an intermediate certificate with different validity
periods, the first non valid and the second valid - such cases are probably
rare or users would have complained already, but it doesn't hurt to handle it
properly as it allows for more uniform code.
This may look like a behaviour change because one check has been added to the
function that was previously done in only one of the 3 call sites. However it
is not, because:
- for the 2 call sites in verify(), the test always succeeds as path_cnt is 0.
- for the call site in verify_child(), the same test was done later anyway in
verify_top()
There are 3 instance that were replaced, but 2 instances of variants of this
function exist and will be handled next (the extra parameter that isn't used
so far is in preparation for that):
- one in verify_child() where path_cnt constraint is handled too
- one in verify_top() where there is extra logic to skip parents that are
expired or future, but only if there are better parents to be found
This is a slight change of behaviour in that the previous condition was:
- same subject
- signature matches
while the new condition is:
- exact same certificate
However the documentation for mbedtls_x509_crt_verify() (note on trust_ca)
mentions the new condition, so code that respected the documentation will keep
working.
In addition, this is a bit faster as it doesn't check the self-signature
(which never needs to be checked for certs in the trusted list).
When we're looking for a parent, in trusted CAs, 'top' should be 1.
This only impacted which call site for verify_top() was chosen, and the error
was then fixed inside verify_top() by iterating over CAs again, this time
correctly setting 'top' to 1.
This is the beginning of a series of commits refactoring the chain
building/verification functions in order to:
- make it simpler to understand and work with
- prepare integration of restartable ECC
md() already checks for md_info == NULL. Also, in the future it might also
return other errors (eg hardware errors if acceleration is used), so it make
more sense to check its return value than to check for NULL ourselves and then
assume no other error can occur.
Also, currently, md_info == NULL can never happen except if the MD and OID modules
get out of sync, or if the user messes with members of the x509_crt structure
directly.
This commit does not change the current behaviour, which is to treat MD errors
the same way as a bad signature or no trusted root.
Modify the function mbedtls_x509_csr_parse_der() so that it checks the
parsed CSR version integer before it increments the value. This prevents
a potential signed integer overflow, as these have undefined behaviour
in the C standard.
Rename the macro MBEDTLS_PLATFORM_SETUP_ALT to
MBEDTLS_PLATFORM_SETUP_TEARDOWN_ALT to make the name more descriptive
as this macro enables/disables both functions.
Add the following two functions to allow platform setup and teardown
operations for the full library to be hooked in:
* mbedtls_platform_setup()
* mbedtls_platform_teardown()
An mbedtls_platform_context C structure is also added and two internal
functions that are called by the corresponding setup and teardown
functions above:
* mbedtls_internal_platform_setup()
* mbedtls_internal_plartform_teardown()
Finally, the macro MBEDTLS_PLATFORM_SETUP_ALT is also added to allow
mbedtls_platform_context and internal function to be overriden by the
user as needed for a platform.
The previous commit bd5ceee484f201b90a384636ba12de86bd330cba removed
the definition of the global constants
- mbedtls_test_ca_crt_rsa_len,
- mbedtls_test_cli_crt_rsa_len,
- mbedtls_test_ca_crt_rsa, and
- mbedtls_test_cli_crt_rsa.
This commit restores these to maintain ABI compatibility.
Further, it was noticed that without SHA256_C being enabled the
previous code failed to compile because because the SHA1 resp. SHA256
certificates were only defined when the respective SHAXXX_C options
were set, but the emission of the global variable mbedtls_test_ca_crt
was unconditionally defined through the SHA256
certificate. Previously, the RSA SHA1 certificate was unconditionally
defined and used for that.
As a remedy, this commit makes sure some RSA certificate is defined
and exported through the following rule:
1. If SHA256_C is active, define an RSA SHA256 certificate and export
it as mbedtls_test_ca_crt. Also, define SHA1 certificates only if
SHA1_C is set.
2. If SHA256_C is not set, always define SHA1 certificate and export
it as mbedtls_test_ca_crt.
Fix a resource leak on windows platform, in mbedtls_x509_crt_parse_path,
in case a failure. when an error occurs, goto cleanup, and free the
resource, instead of returning error code immediately.
Protecting the ECP hardware acceleratior with mutexes is inconsistent with the
philosophy of the library. Pre-existing hardware accelerator interfaces
leave concurrency support to the underlying platform.
Fixes#863
If we didn't walk the whole chain, then there may be any kind of errors in the
part of the chain we didn't check, so setting all flags looks like the safe
thing to do.
* restricted/iotssl-1398:
Add ChangeLog entry
Ensure application data records are not kept when fully processed
Add hard assertion to mbedtls_ssl_read_record_layer
Fix mbedtls_ssl_read
Simplify retaining of messages for future processing
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.
