while debugging the disconnect issue, discovered that the timeout
routine in `std` was extremely wastefully blasting the name server
to connect to the ticktimer every time it was called. yowch!
this patch does the following:
- cache the ticktimer connection in an AtomicU32
- check if the transaction is done on the "Fast path" before
doing any waiting
- check again after fetching time - at this point, most transacitons
are actually likely done
- only then, do the wait loop
this greatly speeds up screen drawing, it seems.
the library is a bit too specific to a given application
without the generic API being exposed. i.e. the way it was
originally generated would imply that every scenario for
the driver would need to be implemented in the library-side
HAL, when in reality, I think it makes much more sense for
the customization of registers (especially when debugging)
to live near the device side.
This is a ball of Claude-generated code which was guided by
the datasheets, some trait APIs and other prompt info.
The prompt used to generate the code is appended to this file.
The result is structurally pretty similar to the accelerometer
driver templates I've seen provided by the vendor, except in Rust
and with Xous idioms, so I think this is a pretty reasonable
whack at reducing boilerplate. There's way more features in these
accelerometers than I have time to write getters/setters for.
The few functions I have used look good insofar as getter/setters,
but there are some oddly specific defaults in the file that should
probably be corrected, probably due to over-prompting of the code
generator. I'll work through these over time.
this is a prelude to clock tuning the internal bus networks
for bandwidth. The "trivial" thing of just cranking the speed
up to the edge and letting the dividers didn't seem to help
performance much, though; this could be due to the clock synchronizers
beating against each other.
- remove bounds checking for accelerating blits that are already checked
- fix the command ordering for display power-on to remove "noise"
during power-on of the display.
- oem-baosec-lite board support
- more flash memory types
- transparent polarity switching on reset line
The reset line in particular is having its polarity switched
because the I/Os pull-up by default, and we'd like the reset
line for peripherals to be asserted on reset (this prevents
glitching on e.g. displays while the CPU boots).
I decided i feel too uncomfortable doing any state changes based
on unvalidated data. I can't see a downside to moving the check
to only after the signature has been verified.
one minor issue with this implemnetation is it requires dev
images to have an anti-rollback counter that matches or exceed
the last signed image seen.
This is more of a developr ergonomics issue - this will certainly
lead to some surprise by developers who are trying to take
production devices and "downgrade" them to an old dev tree
that has an old anti-rollback firmware number in it.
However, I think this is navigable and probably the simplicity
of not having an "escape" path for dev-signed imgaes is worth it.
All a developer has to do is to set the anti-rollback version
number to be equal to or greater than what was reported in
the device.
this is a sanity check number that can be used by firmware
to vet a counter value before incrementing to it. it's particularly
useful to thwart wear-out attacks agaisnt anti-rollback firmware
versions, i.e. by specifying an extremely high anti-rollback number
one might be able to cause a reset-to-0 failure and this prevents
that from happening by limiting the maximum value of the the
counter that can be requested.
The IFR copy of pubkeys is truly actually indelible because
of the no-exceptions write inhibit byte. Add code to check
against that copy.
Note that the IFR region's downside is the presence of
the write-inhibit byte means the check is complicated
as the full key is striped across two locations, but
this code handles that.
The prior implementation tried to use the UDMA streaming mode
to get Rx data out of the UART.
The problem is that with streaming mode, it's hard to tell how much
data is actually available. This is because the thing that sets up
the stream sets "some number of bytes" as the fetch depth of the
DMA Rx, but all you can read is "number of bytes remaining" out
of the DMA interface.
So when 0 bytes are remaining, you can fetch some data, but you don't
know *how much* to fetch. And if someone else fetched partial data
before you, you don't know how much partial data was fetched. This
could be resolved with some gnarly, highly concurrent state that
gets incremented on every fetch.
Or, we could use poll mode. There's a reason i think it's in the
block, because the Rx DMA mode is more or less not usable. Poll
mode you read a config register, see if there's valid data, and if
so, when you read the data register, it's pulled out of a FIFO.
The FIFO is in hardware and it's 4-deep. As long as you don't
need more than a depth of 4, you're rock solid.
This is a lot easier than the UDMA Rx mode, so I've just gone ahead
and deprecated the Rx mode. I could only ever get the Rx mode
to work to receive a single character *anyways*, so despite its
potential to receive thousands of bytes without CPU intervention,
it was only ever used as a 1-deep buffer because of the inability
to know exactly how much was in the buffer. Thus, the poll mode
is actually an improvement over the DMA mode, in practice.
More importantly, it means the Rx link is much more stable in
GDB kernel mode, if the baud rate is cranked down to 115200 (at
1M baud, the kernel still can't respond quite fast enough to
reliably drain it).
this is essentially the getter/setter that translates
BIO api calls into messages that would be sent to the BIO
server implementation. The BIO server implementation would
have the actual BIO hardware calls in it. These calls, if
written correctly, also double as the no-std implemnetation
of the BIO API.
the problem with it being default in most configurations is that
it pulls in a sha-2 and ed25519 dependency into routines that
never rely on them. now libs that rely on the sigcheck need to
specify the feature explicitly to pull in the depedency.
- backup RAM has an erratum where two locations are corruptable
and should not be used between soft resets to store data
- the reserved 32 bits in the bureg are now accessible as flags.
these flag bits are used between boots to coordinate system state
information.
this allows us to refer to it in boot1 if we eventually decide
it's useful to do it there.
however, as a matter of policy, boot1 checks loader signatures
against *boot1*'s public key block - so if the loader is to be
signed with a different key, boot1 needs to be updated.
actually there's no reason i think to have it print all
four checks - just have it print on two of them, just in
case for some reason the prints could be used to time a glitch.
The "attack" counter notches up when a board goes through a
"slow" powerdown, as would be the case on a system that has
a lot of capacitance relative to the chip. This is a false
signal.
It's hard to tell this apart from a glitch attack, so a
heuristic was applied where at least two glitches have to
be seen before declaring we're under attack.
Note that in paranoid mode, the board instantly resets
when the power gets too low, so the counter doesn't notch
up.
A print statement indicating when a mesh violation was
added - this is to help debug chips that have a weak/marginal
mesh. Because the mesh is so long, small manufacturing defects
in the mesh are also picked up by the mesh sensor...
this reduces the amount of prints for a key range being
erased so if the debug flags are on boot isn't delayed
while the debug prints. instead, a total tally of the
erased keys for a key range is printed.
this is a big, ugly, multi-file commit but it turns out when
you start shaking one part of the clock tree everything rattles
along with it.
Basically this updates the clock tree to use a search for an
optimal setting instead of a hueristic/table lookup, and then
a bunch of small bugs in the clock settings were found that had
to be proopagated to dependent crates.
i2c clock set now seems ... more correct, the frequency is
not overclocking the bus (was at 600khz before, now just below 400khz)
the uart equations finally sees sane
the spi clocks check out
the only inconsistency left is that i can't seem to get a fractional
division on peri clk. it would be really nice to be able to do a
fractional divide and get a little more peformance out of the spi
busses but i think it just doesn't work that way - the "fractional"
division is actually more of a "time-average frequency is correct"
but when you go into fractional mode it's spending some time pulsing
at 2x the rate and some time at 1x the rate to get to the time average
frequency, and when you're at the 2x rate things break.
a slice of key material is written to backup RAM from boot0
all the way to the loader, to prove to the loader that secret
keys were erased in a higher privilege state (loader then erases
the key, and this key is considered leakable without harm to
security).
The problem is that writing the key to backup RAM in boot0
was fixing the hash on the backup block, which is before
the check happens to see if it was a cold power-on. This would
defeat the RTC setting & backup setup logic.
THe solution to this is to exclude that one key from the hashing
area and special case it. A set of routines and API names were
added to help enforce this case.
this makes i2c accessible without having to rely explicitly
on the server itself, making the library callable in more contexts.
necessitated avoiding a name space collision and moving the
I2C hardware drivers to a stucture named I2cDriver; the I2C
object now unambiguously refers to the userspace interface
to I2C.
This routine pares down the clock setting routine to hopefully
a core minimum set of side-effects while allowing the system
to boot.
It also takes the register bank offsets as arguments, as well
as various functions to accomplish different tasks as argument,
so that it can be used in different contexts (such as in `std`).
Finally, the BIO clock rate can be set to slow or fast. The power
cost of the fast setting is about 12mW. If you're not needing
the high speed features of the BIO, slow setting is probably what
you want.
Currently, the system is configured such that after boot1, the
BIO is running in slow mode (to save power for battery powered
configurations); and then the clock set calls in baremetal and
dabao bump this up to the fast speed since these two configurations
are assumed to be running off of a USB wire with ample power.
glitches can trigger interrupts from several sources:
- ECC errors on multiple blocks
- consistency check errors on hardened logic
- light sensor
- voltage sensor
- glue cells
These primitives respond to these glitches in an interrupt
context. At the moment, by default a fairly lax approach is
taken to responding to the glitches: partially, we want to make
sure that the sensors aren't too sensitive and trigger by accident.
The final thresholds can be adjusted fairly readily.
In addition, the mesh can trigger an interrupt, and typically,
that interrupt is a false reading (because the mesh takes a
long time to settle, and we boot the system while the mesh
settles). Thus, we take an approach of polling the mesh
effectively, and when interrupts happen due to the mesh we
just clear it and ignore it and allow the polling routine
to manage the mesh. The mesh is theorized to be a static
attack, anyways.
this is used by the key erase proof routine, but will be
useful in general for storing data in the buram
also, the ranges of slices are currently defined at
the top of the file but might be moved somewhere more
maintainable later.
RRAM writing got bollards added to it, as RRAM write could
be a glitch target: if a glitch could manipulate the
address somehow to stick stuff at an arbitrary location,
during write, that would be bad. However, it's probably
also reasonably dangerous to try and glitch RRAM writes
because there's a chance you spray over off-target items
and thus trigger a signature check failure on your next
boot.
Signature checking parameters are now consolidated into
a single "SecurityConfiguration" record. This allows the
sum total of security state transitions to be summarized in
the bao1x-api/src/pubkeys/mod.rs file, easing maintenance
and audit.
Reaction to the signature check is split into a `hardened_erase_policy`
API call which can take the results of the signature check and
react to it if necessary.
This allows the signature check routine to be used for both
audit (where reaction is not necessary) and for live testing
of the code (where reaction is required). It also simplifies
the implementation of double-checking signatures in the case
of paranoid mode.
Another important improvement is that the signature checking
API outputs the target address for the jump, but masked with
an XOR of the function code of the signed block. The purpose
of the XOR mask isn't to obfuscate the value; rather, it's
to prevent there being a single, simple, fixed-constant
jump instruction laying around that a glitcher can use as
a target: instead, a pair of registers must be loaded
with the target and its mask. This is an example of what
the final jump looks like:
```
60003a8e: ffffffff .word 0xffffffff
60003a92: 0440006f j 60003ad6 <bao1x_hal::hardening::die>
60003a96: ffffffff .word 0xffffffff
60003a9a: 167d addi a2,a2,-1
60003a9c: fe71 bnez a2,60003a78 <rust_entry+0x2f74>
60003a9e: a00d j 60003ac0 <rust_entry+0x2fbc>
60003aa0: 0360006f j 60003ad6 <bao1x_hal::hardening::die>
60003aa4: ffffffff .word 0xffffffff
60003aa8: 02e0006f j 60003ad6 <bao1x_hal::hardening::die>
60003aac: ffffffff .word 0xffffffff
60003ab0: 0260006f j 60003ad6 <bao1x_hal::hardening::die>
60003ab4: ffffffff .word 0xffffffff
60003ab8: 01e0006f j 60003ad6 <bao1x_hal::hardening::die>
60003abc: ffffffff .word 0xffffffff
60003ac0: 0149c533 xor a0,s3,s4
60003ac4: 600805b7 lui a1,0x60080
60003ac8: d0058593 addi a1,a1,-768 # 6007fd00 <_etext+0x6e9dc>
60003acc: 82aa mv t0,a0
60003ace: 852e mv a0,a1
60003ad0: 4581 li a1,0
60003ad2: 8282 jr t0
```
So here you can see the alternating bollards in the region that
will cause any glitches that don't land exactly right to
either halt the machine or cause a reaction.
Line 0x6003ac0 is where the xor happens to compute the target
address. Should a glitcher land on that address, `s3` and `s4`
will need to be loaded with the target address and a value when
XOR'd with it that yields a meaningful result. While there is
a non-zero chance that this can be arranged, it's definitely
not as easy as targeting a jump-to-fixed address exit sequence.
- key revocation check is fully duplicated and tested twice
- bollards & random delays added throughout the code. Random
delays require passing in a csprng object from the calling
environment.
- the primary go/no-go decision on whether a signature passes
or not is unfortunately not hardenable because the API is in
a third party crate. However, the regions around that call are
random-delayed and surrounded with bollards
- the state bits that determine whether the system is secure
and/or in developer mode are implemented using HardenedBool
- erase_secrets has some delays added to it as well as bollards.
However, it does remain a weakness that an attacker
could pull RRAM power during the erase cycle by just isolating
the RRAM power supply and putting a fast current limit on it,
thus preventing effective erasure by starving the circuit
of power.
Nothing we can really do about an active attack on the power
supply to prevent erasure, I think. It is unfortunately required
by design to have an isolated power supply for the RRAM to perform
elevated-voltage forming at CP time, and so, it should always
be possible to limit the current going into it thus preventing,
at least somewhat, any programming operations on the array.
One way counters are used ot record some security-critical properties,
such as key revocations and paranoid mode.
Two methods of retrieving the counters so that they are hardened
are provide here. Both of them rely on the fact that any value
other than 0 for the counter means it is set.
`hardened_get()` retrieves the counter value and sums it several
times. It does this *twice*, and returns *both* values as a tuple.
By returning both values as a tuple, it reduces the likelihood
that the compiler optimizes out the redundant read. The receiver
should implement serial checks that operate on the hardened_get()
result (i.e., test one value of the tuple, then test the other
value of the tuple redundantly, instead of OR'ing them together
into a single test)
`hardened_get2()` is for counters that are fully duplicated in
the array. For extremely critical counters, they are stored in
two physically separate elements. This is done to make bypass
of the counter harder than just FIB'ing out a single bit in
the memory array. In the case of `hardened_get2()`, a tuple
is returned, but instead of reading the same value twice, the
tuple corresponds to the summed value over each of the redundant
counters.
This library contains the following primitives:
check_pll() - checks that the PLL is the primary clock source.
This is important because allowing direct clocking of the system
from the external clock pin makes glitching much, much easier.
die() - a routine that implements a reaction to attacks. At the
moment, we just note attacks and then halt the machine. Eventually,
this could be upgraded to be more reactive, e.g. erasing secrets
or bricking the chip.
Csprng - this is a chacha8 cipher seeded with ring oscillator
TRNG data, plus whatever tiny bit of entropy we can scavenge
from uninitialized RAM (to make it so that it's not as trivial
as cutting the RO wire and tying it to 0 to bypass the library).
There's also a tiny health check on the RO before using it, again,
to make it harder than just cutting the RO wire. The sole purpose
of the Csprng is to generate random delays that frustrate glitching
attacks. The delay is currently tuned to a range of 0-255 iterations
of a loop.
paranoid_mode() - turns on fully automatic reactive sensor
countermeasures. Whenever a voltage glitch is detected, the system
will just reboot.
mesh countermeasures - these set up the secure mesh and the
reaction policy to tamper detection on the secure meshes
apply_attack_policy() - an API call that applies the currently
determined countermeasure in case that the detected attacks
counter has gone above a threshold. The reason we don't simply
react right away is that some countermeasures are fairly
sensitive, and *also*, very importantly, the chip *is* packaged
in a CSP and thus especially for dabao users you will expect
to see random trips of the attack sensors due to environmental
noise (photoflashes, nearby lightning strikes, etc.).
The -A1 stepping is going to tie the IFR access control
bits for data and key slots together. This is due to an RTL
bug where the key slot maps were typo'd and they ended
up mapping to the same IFR ACL.
We "could" continue to use key slots with a rule that
key and data slots would have to coordinate with each
other and not use the same slot number.
However, because of this linkage, and because there are
other bugs suspected in the keyslot implementation, an
executive decision was made to disregard the key slot
feature in the chip.
The main downsides of this are:
- fewer slots (2048 vs 4096)
- no key access linkage to HMAC auth on SCE
I *think* that's it.
Fewer slots - 2048 slots is still a lot of slots, about half are
used now, I think this is plenty.
Key access linkage to HMAC auth on SCE - I have reasons to
suspect this feature is broken anyways for subtle reasons
related to malleability of the HMAC round constants. Even
without that brokenness, it's hard to use HMAC auth to manage
the system because it requires per-device unique keys: so,
we're using a PKI scheme and no symmetric secrets to handle
access controls.
Removing the key slots also greatly simplifies the code
base analytically. This culling removes the key slot
abstraction the "simple" way, but without the weirdness
of the keyslots (in particular the key ID chaining which
makes them different from data slots) I could remove even
more abstractions and simplify the code.
One upside is now all elements can enjoy the write mode
behavior which can force the slot to be only 0->1 writeable,
which could be useful for other things down the road.
This was tested to not interfere with Dabao-Alpha 0
bootloaders.
make it long enough that it's unlikely to have two identical
ones randomly connected to the same machine
but short enough that it's likely/conceivable to have a collision
there is a lock that happens in serial console interactions with
local echo where the instant feedback of a character entered causes
the USB object to be locked when the interrupt handler enters to
receive a character that was just sent.
The fix to this is to add the ability to disable interrupts before
entering a critical section (such as enqueuing data for a write)
and then re-enabling after tha section is done.
identify_flash_reset_qpi's behavior has been revised to
include setting all relevant parameters so that the device
that is identified behaves identically to the model device
(e.g. Macronix), to the extent that it is possible. This
includes setting the QE bit.
The routine to check QE has been renamed to `flash_ensure_eq()`
to reflect that it is safe to call multiple times: it checks
if the bit is already set, and if so, doesn't do anything.
Thus it can be routinely called without worry of wearing
out the device.
xtx devices still not fully working, but at least able to switch
in and out of spim modes in the face of uncertainty over which
device type is present.
rram handler for `std` requires the following additional features
over baremetal:
- mapped pages list, because a process has to "own" a page of RRAM
before writing to it. It's anticipated that the rram writer hardware is
"owned" by the keystore, so all write-access will be mediated by a
service in that PID
- irq-context handling of writes. this prevents writes from being
detached from their staging data, or other concurrent writes
from conflicting.