Summary:
This is a simple CFG transformation that compresses jumps to jumps
into jumps to the final destination. LLVM's simplifycfg pass seemingly
should have eliminated these, but does not.
Reviewed By: jvillard
Differential Revision: D25196728
fbshipit-source-id: 3288dbd8d
Summary:
Currently the symbolic execution options, obtained from the command
line flags and preanalysis, are passed to the Control entry points as
a record, and then passed around within Control as needed. This diff
simplifies the code in Control by replacing the record mechanism with
an additional functor parameter containing the options. The main
benefit is that the bound option no longer needs to be part of the
analysis state.
This change does make the code in Sledge_cli slightly more
complicated, as it now performs a functor application using a
first-class module computed from the command line flags.
Reviewed By: jvillard
Differential Revision: D25196735
fbshipit-source-id: 80e5d5d62
Summary:
Revise the control-flow exploration scheduling algorithm to fix
several issues. The main difference is to change the priority queue to
keep the control edges on the frontier of exploration in sync with the
states that are waiting to be propagated. This fixes several sorts of
issue where the decision of which control and state joins to perform
was unexpected / wrong. Part of keeping the frontier edges and waiting
states in sync is that the waiting states are associated not only with
a destination block, but the stack of that block. This fixes several
issues.
Combined, these changes lead to the algorithm only attempting joins
for which the pointwise max join on depth maps is correct (with the
caveat of no mathematical proof yet).
Reviewed By: jvillard
Differential Revision: D25196733
fbshipit-source-id: db007fe1f
Summary:
The first-order solver sometimes needs to generate fresh variables to
express the solution of equations. It needs to ensure that these
generated variables do not clash. Before this diff, there was a
confusion where new variables were fresh with respect to only the
current set of "universal" variables. This is wrong, and this diff
adds the full set of variables instead.
Reviewed By: jvillard
Differential Revision: D25196732
fbshipit-source-id: afc56834a
Summary:
It can happen that the actual return variable, which is the name of
the variable in the caller's scope that receives the returned value,
clashes with the formals of the callee. The scope of the return
formula was wrong in this case, as the actual return was outscoped via
existential quantification with the rest of the formals after the
return values was passed.
Reviewed By: jvillard
Differential Revision: D25196727
fbshipit-source-id: 94cf25418
Summary:
A context with valid but extraneous equations was being kept. The
extra equations are valid, but might violate vocabulary invariants.
Reviewed By: jvillard
Differential Revision: D25196734
fbshipit-source-id: a8001a075
Summary:
Currently there is a symbolic execution option to ignore exceptional
control flow. This hack does not fit well, and it is unclear how much
backend functionality should take it into consideration. This diff
removes this option and replaces it with an option during model
compilation. This has the advantage of clarifying and simplifying the
backend, with the disadvantage of no longer supporting switching
between exceptions and no-exceptions modes at analysis time. Since the
possibility of ignoring exceptional control flow is due to it not being
ready yet, this is a good trade to make.
Reviewed By: jvillard
Differential Revision: D25146148
fbshipit-source-id: 1f1299ee1
Summary:
Use the equality class information in the symbolic state to resolve
callees of indirect calls.
Reviewed By: jvillard
Differential Revision: D25146160
fbshipit-source-id: a1c39bbe1
Summary:
The callee function of a Call can often be resolved
statically. Currently this is resolution is only done dynamically
during symbolic execution by checking if the callee expression is a
function name and looking up the function in the program. This is
wasted and redundant work. Also, the static resolution code is
duplicated in all the domains.
This diff resolves this by resolving known callees statically at
translation time. This involves:
- add an ICall terminator that is the same as Call is currently
- change Call to use a func callee instead of Exp.t
- make callee field mutable since recursive calls can create cycles
- change the Llair.Term.call constructor to return a thunk to perform
the backpatching once the callee has been translated
- modify the Frontend
+ to determine whether to emit Call or ICall depending on whether
the callee in LLVM is already a Function
+ to record the LLVM function -- backpatch thunk pairs encountered during translation
+ record the mapping of LLVM to LLAIR functions during translation
+ to enumerate the calls to backpatch after all functions have been
translated, and find the LLAIR function corresponding to each LLVM
function and backpatch the call to use it as the callee
+ to handle direct calls to undefined functions, when backpatching
translate such function declarations into undefined functions
Reviewed By: jvillard
Differential Revision: D25146152
fbshipit-source-id: 47d2ca1ff
Summary:
Current code partially tries to handle Invoke, but incorrectly, and
these names are only needed for Call. So this diff revises this to be
slightly simpler and less confusing.
Reviewed By: jvillard
Differential Revision: D25146157
fbshipit-source-id: ead4f6f31
Summary: Rework the intrinsic name detection to detect e.g. llvm.memset.*
Reviewed By: jvillard
Differential Revision: D25146150
fbshipit-source-id: 85ebcfb7a
Summary:
Rename the existing exec_intrinsic that works for calls to intrinsic
functions to exec_intrinsic_func to make room for an exec_intrinsic
that works on intrinsic instructions.
Reviewed By: jvillard
Differential Revision: D25146166
fbshipit-source-id: 80ae3aac9
Summary:
The translation of instruction intrinsics that are `Invoke`d is almost
the same as those that are `Call`ed. The same Llair instruction is
produced, but it is wired into the CFG differently. This diff uses the
translation of instruction intrinsics used for `Call`s for `Invoke`s
as well.
Reviewed By: jvillard
Differential Revision: D25146159
fbshipit-source-id: 85a93d915
Summary:
The code that computes the number of actuals is largely duplicated
between the Call and Invoke cases. But some issues have been fixed in
each that ought to be applied to the other. This factors out and
unifies this computation.
Reviewed By: jvillard
Differential Revision: D25146149
fbshipit-source-id: 78552327a
Summary:
Move the translation of Calls to intrinsics that map to instructions
into a separat function.
No functional change.
Reviewed By: jvillard
Differential Revision: D25146169
fbshipit-source-id: 79c640344
Summary:
Currently intrinsics are treated as functions, with Call instructions
to possibly-undefined functions with known names. This diff adds an
Intrinsic instruction form to express these more directly and:
- avoid the overhead of intrinsics needing to end blocks
- avoid the overhead of the function call machiery
- avoid the complexity of doing the string name lookup to find their
specs, repeatedly
This diff only adds the backend support, the added Intrinsic
instructions are not yet generated by the frontend.
Reviewed By: jvillard
Differential Revision: D25146155
fbshipit-source-id: f24024183
Summary:
Previously, when LLAIR was in SSA form, blocks took parameters just
like functions, and it was sometimes necessary to partially apply a
block to some of the parameters. For example, blocks to which function
calls return would need to accept the return value as an argument, and
sometimes immediately jump to another block passing the rest of the
arguments as well. These "trampoline" blocks were partial applications
of the eventual block to all but the final, return value,
argument.
This partial application mechanism meant that function parameters and
arguments were represented as a stack, with the first argument at the
bottom, that is, in reverse order.
Now that LLAIR is free of SSA, this confusion is no longer needed, and
this diff changes the representation of function formal parameters and
actual arguments to be in the natural order. This also brings Call
instructions in line with Intrinsic instructions, which will make
changing the handling of intrinsics from Calls to Intrinsic less
error-prone.
Reviewed By: jvillard
Differential Revision: D25146163
fbshipit-source-id: d3ed07a45
Summary:
Overwritten variables in move instructions are not impossible. Since
Domain_itv does not handle them, the check should be a `todo` rather
than an `assert false`.
Reviewed By: jvillard
Differential Revision: D25146168
fbshipit-source-id: 13d8587c7
Summary:
It was possible for the scope of a local to be incorrectly restored
when entering it for the first time in a caller after is was shadowed
by a callee. This could happen because scope management in the
analysis relies on renaming variables to adjust the vocabulary of
symbolic states. This was usually done, but optimizations of renaming
with a substitution whose domain is disjoint from the vocabulary of a
formula inadvertantly violated this vocabulary-adjustment
assumption. (Yes, this is too easy to get wrong.)
Reviewed By: jvillard
Differential Revision: D25146162
fbshipit-source-id: 30f2d657f
Summary:
These are not simply handled by `@deriving` since:
- These types are recursive and so some fields need to be ignored
- Some are uniquely identified by one of their fields
- Some fields are mutable, but set only during construction, and need
to be considered by compare, equal, hash, but ppx_hash refuses.
Reviewed By: jvillard
Differential Revision: D24989064
fbshipit-source-id: 7f8d699e5
Summary:
The use of realpath on paths obtained from debug info and the current
working directory is application-usage-specific behavior that does not
belong in the backend library. This diff moves these uses to the
frontend and cli, respectively. Also, the use of realpath in the
frontend is memoized along the same lines as the other frontend
translation functions.
This was also the last use of `core` in the `sledge` library, so the
dependency is moved to `sledge_cli` and `sledge_report`.
Reviewed By: ngorogiannis
Differential Revision: D24989070
fbshipit-source-id: c21b275f5
Summary:
Currently __llair_choice is left undefined, and so executed as skip.
This has the correct behavior, but makes it hard to distinguish from
calls to unintentionally-undefined procedures.
Reviewed By: ngorogiannis
Differential Revision: D24989068
fbshipit-source-id: f62981857
Summary:
Use B/K/M/G/... units for memory quantities, round percentages to 2
decimal places, and robustly detect deltas of close to zero.
Reviewed By: ngorogiannis
Differential Revision: D24989063
fbshipit-source-id: 35f0bccfb
Summary:
Instructions that have multiple specs sometimes have inconsistent
footprints for one or more of their specifications. In such cases, the
handling of the existential and universal vocabularies was sometimes
incorrect before this diff. In particular, the ghosts of a spec need
not appear in the pure approximation of the footprint, which the
previous code incorrectly assumed. Also, the unit of disjunction is
`false` with an empty vocabulary, not the precondition's vocabulary as
the previous code incorrectly used.
Reviewed By: ngorogiannis
Differential Revision: D24989067
fbshipit-source-id: eca3bff55
Summary:
Change `-llair-output` to `-output`, for binary form, and
`-llair-txt-output` to `-llair-output`, for textual form. Also
correspondingly change `.llair` to `.bllair`, for binary, and
`.llair.txt` to `.llair` for text.
This improves command line argument completion, and makes `.llair` the
extension of the files most commonly interacted with.
Reviewed By: ngorogiannis
Differential Revision: D24951506
fbshipit-source-id: ad4c73ca2
Summary:
The `seq` name of this field refers to the expected sort of it's
value, where the others refer to the role they play. So rename
seq (for sequence) to cnt (for contents).
Reviewed By: ngorogiannis
Differential Revision: D24951507
fbshipit-source-id: fd6640517
Summary: It is too easy to mix up multiple arguments of the same type.
Reviewed By: jvillard
Differential Revision: D24934116
fbshipit-source-id: 6e595b26e
Summary:
Global constants have reliable types, and their sizes can be used
instead of storing the size of the initializer separately.
Reviewed By: jvillard
Differential Revision: D24934114
fbshipit-source-id: 2426ab5be
Summary:
In practice this has not been observed to matter so far, since
treating `Splat N` as interpreted or uninterpreted does not matter
when `N` is a literal constant, and code seen so far only uses `Splat`
for zero initializers or memset with literal constant bytes.
Reviewed By: jvillard
Differential Revision: D24934118
fbshipit-source-id: 213e9724e
Josh Berdine