| Age | Commit message (Collapse) | Author |
|---|
|
newlib, because memset is called a LOT in the course of RSA signing, and
we need to understand how much time we're actually spending there.
|
|
|
|
|
|
so the user might be able to copy and paste a generated key into the CLI
for recovery.
Display had been a 32-byte hexdump. Parsing (manual re-entry of the KEK)
had been 8 32-bit little-endian values. This has been a pain point for
literally two years, albeit at a low enough pain level that I've managed
to cringe and ignore it.
|
|
|
|
|
|
This can take long enough (several minutes for h=10) that we do it in a
background task, which is then converted to an RPC dispatch task.
Also add a very limited form of free(), to free the topmost allocation in
the sdram "heap". I don't want to deal with real heap management, but I do
want to be able to recover memory upon deleting a hashsig key, if it's
easy to do so.
|
|
|
|
|
|
|
|
MGMT is the default UART, and no one should have to explicitly refer to
the UART unless they need USER (hsm.c:hal_serial_send_char).
The default UART is now exposed in the header file, so that the
default-using functions can be macros, which saves a few bytes in code
space, and a few microseconds in function call overhead.
|
|
Note: This affects libhal/ks_token.c, which uses the keystore driver directly.
|
|
|
|
|
|
|
|
|
|
|
|
interrupted) rather than LR (the return address from the function we
interrupted).
Also, change u_short and u_int to unsigned short and unsigned int, since
gcc recently decided that those aren't part of the C99 standard.
Finally, add profilable versions of memcpy, memset, and friends, because
they get called a lot in the course of unit testing, and it would be nice
to know who's calling them.
|
|
|
|
|
|
already a user-callback mechanism with HAL_SYSTICK_IRQHandler() and HAL_SYSTICK_Callback().
|
|
|
|
subject to the same forces that made it a good idea in the first place.
commit 2b6b9f8
Change RPC UART to have a high-priority thread monitoring a large(ish) DMA
buffer, because we've observed out-of-order receives under load.
|
|
|
|
|
|
|
|
interrupted) rather than LR (the return address from the function we
interrupted).
Also, change u_short and u_int to unsigned short and unsigned int, since
gcc recently decided that those aren't part of the C99 standard.
Finally, add profilable versions of memcpy, memset, and friends, because
they get called a lot in the course of unit testing, and it would be nice
to know who's calling them.
|
|
|
|
|
|
Clean up Makefiles and initialization code.
|
|
Required minor manual intervention to resolve merge issues git had no
way of understanding: git is clever, but not quite clever enough to
understand that a commit in branch had removed the entire RTOS that a
commit in the other branch was using. No big deal, just a couple of
osDelay() calls needing conversion to HAL_Delay() or task_delay().
|
|
|
|
|
|
There are no priorities and no preemption, so tasks run in a round-robin
fashion, and explicitly yield control.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Apparently it's easier to duplicate source files into multiple project
directories than to write Makefiles that do something sane. Feh.
|
|
|
|
|
|
Fetching a list of keys and all of their metadata isn't an atomic
process, nor, probably, should it be, so we need to cope with things
like a key being deleted via the RPC interface while we're fetching
its metadata for display on the console interface.
|
|
|
|
buffer, because we've observed out-of-order receives under load.
|
|
If hal_rpc_server_dispatch() returns an XDR decode error because the
request packet was too short, don't call Error_Handler() and kill the
dispatch thread, just drop the request.
Add more ibuf_queue entries, but don't panic and kill the dispatch thread
if we can't get one, just drop the incoming character (which will lead to
an XDR decode error if/when we finally get an ibuf).
|
|
|
|
We need to start with a long serial timeout, in order to catch the reboot
messages for a firmware upload (this has to be done through the bootloader).
But once we start sending the file, cut the serial timeout to 1ms. (I've
tested it down to 1us, but that may not work for everyone, and it doesn't
improve performance in a statistically significant way.)
This brings the time to upload a 4.5MB bitstream from 38:23 to 1:25.
|
|
|
