For CRefine, this process is much more complex than for Refine and up,
as the C code both has its own definitions `maxDom` and `numDomains`,
but they are not defined in terms of each other, only numbers.
Similarly, array size types and their corresponding ArrayGuard bounds
checks refer to specific numbers, making a fullproof abstraction impossible.
A reasonably constrained interface to numDomains/maxDomain/maxDom in
Wellformed_C provides a sufficient abstraction to allow the proofs to be
independent of the number of domains (constrained to <= 256). Using the
value_type command allows more abstraction techniques, such as linking
the size of the scheduler queues back to numDomains*numPriorities,
without stating what the numbers are. Finally, for getting past the
ArrayGuard bounds checks, we do leak some information in the form of
`explicit` lemmas. These are the least safe, but short of augmenting the
C parser to re-wrap array sizes into equivalent constants/types, they
constitute a limited risk. Nonetheless, `explicit` lemmas should be used
as sparingly as possible.
Refinement to C proceeds by pretending we don't know the number of
domains, and whenever a control flow decision is made based on
`numDomains > 1`, we follow both branches, as we did for Refine. We also
attempt to avoid clever rewrites such as `(x < 1) = (x = 0)` which mess
up bounds checks into a domain-size array when `numDomains = 1`.
Signed-off-by: Rafal Kolanski <rafal.kolanski@proofcraft.systems>
Previously, the C kernel maintained a global pointer to the IRQ node.
This pointer was only initialised during boot, when the actual IRQ node
was dynamically allocated from untyped memory.
The C kernel now includes a statically allocated IRQ node, which is just
a suitably sized array of CTEs. This commit updates the proofs to verify
this change to the C kernel.
Session-qualified imports will be required for Isabelle2018 and help clarify
the structure of sessions in the build tree.
This commit mainly adds a new set of sessions for lib/, including a Lib
session that includes most theories in lib/ and a few separate sessions for
parts that have dependencies beyond CParser or are separate AFP sessions.
The group "lib" collects all lib/ sessions.
As a consequence, other theories should use lib/ theories by session name,
not by path, which in turns means spec and proof sessions should also refer
to each other by session name, not path, to avoid duplicate theory errors in
theory merges later.
This change was a result of the constant "(tcb_t*)~0" being defined as
0x00000000FFFFFFFF on x86-64 (0 is implicitly a 32-bit integer) rather
than 0xFFFFFFFFFFFFFFFF as expected.
Colloquially known as "invert-fastpath".
Update verification efforts on ARM for the following seL4 changes:
- scheduling decisions done in possibleSwitchTo are moved to the
scheduler
- possibleSwitchTo only checks whether the candidate is valid for a
fast switch, not its priority, accepting possible candidates
immmediately as a switch-to scheduler action
- the scheduler checks the candidate against the current thread and
against the bitmaps before making a decision
- attemptSwitchTo and switchIfRequiredTo are gone
- scheduler is now more complicated, and numerous proofs related to it
are rewritten from scratch
- fast path now checks ready queues via the scheduler bitmaps
- L2 scheduler bitmap order reversed for better cache locality
Many iterations between the kernel and verification teams were needed
to get this right.
In X64 update the following to match the C kernel:
- TCB size-bits (11).
- Endpoint size-bits (4).
- Guard bits (58).
- Message registers.
For all architectures, replace magic numbers with defined constants in
specifications, and as far as possible in proofs:
- tcb_bits in abstract spec.
- tcbBlockSizeBits, cteSizeBits, ntfnSizeBits, epSizeBits in Haskell
spec, Haskell and C refinement proofs.