Conversation
…terminator, remove file pointer cases 1. Add headers, Adding missing headers: For obvious reasons. 2. Remove cases without null terminator: Both clang and g++ do not permit strings to be allocated that are declared to be shorter than the actual initializing expression. Since this is a C++ rule, we rule them out. 3. File pointer manipulation functions (e.g. fgets): Not required by the rule.
…w-nodes-MISRA-C++-2023-Memory-Experimental' into jeongsoolee09/MISRA-C++-2023-Memory-Experimental
These contain false positives due to the limitation of the status quo of the query.
MichaelRFairhurst
left a comment
MichaelRFairhurst
left a comment
There was a problem hiding this comment.
So close to ready!
| CallocFunctionCall() { this.isCallocCall() } | ||
|
|
||
| override int getMinNumBytes() { | ||
| result = lowerBound(this.getArgument(0)) * lowerBound(this.getArgument(1)) |
There was a problem hiding this comment.
Perhaps we should file a bug to come back to this.
In theory, it would be great to have two versions of the query: one where we know with certainty that the resulting pointer is out of bounds if flow analysis is correct -- we assume the maximum allocation size and the smallest pointer offsets. Then another where we suspect a possible invalid pointer, where we assume the minimum allocation size and the largest pointer offsets. These could share most behavior and would have different precisions.
In the meantime, lets ship!
|
|
||
| newtype TArrayAllocation = | ||
| TStackAllocation(ArrayDeclaration arrayDecl) or | ||
| TDynamicAllocation(NarrowedHeapAllocationFunctionCall narrowedAlloc) |
There was a problem hiding this comment.
Let's file a bug to come back to the third kind of "allocation," which is just taking the address of a non-array variable or lvalue.
int x = 0;
int *p = &x; // p is essentially a buffer of size 1
Partly I say let's come back because we would need to be careful to distinguish:
int x = 0;
int arr[5] = {0};
int *p1 = &x; // generally, taking an address to anything should be a buffer of size 1
int *p2 = &arr[0]; // except this
// Note that any lvalue expression can create a "buffer" of size 1, not just variables:
int &f() { return x; }
int *p3 = &f(); // also a "buffer" of size 1
int *p4 = &*p3; // also a "buffer" of size 1
| */ | ||
| int getOffset() { | ||
| if this.asPointerArithmetic() instanceof PointerSubExpr | ||
| then result = -this.getOffsetExpr().getValue().toInt() |
There was a problem hiding this comment.
Another thing to file is that this currently only works on constant values, but in the future we could extend this to use range analysis.
There was a problem hiding this comment.
Good point. Introducing range analysis should be careful, otherwise it might generate a lot of noise. This is out of scope of this PR and should be reserved for later.
| sink.getNode() = end.getBasePointerNode() | ||
| | | ||
| srcOffset = start.getOffset() and | ||
| sinkOffset = end.getOffset() and |
There was a problem hiding this comment.
This overwrites the previous offset, but they should add up.
For example:
int arr[5];
int *p = arr;
int p1 = p + 3; // offset: 3, length: 5
int p2 = p1 + 2; // offset: 5, length: 5Currently, this will produce sinkOffset = 2 for the last line
There was a problem hiding this comment.
Nice!
Only thought now, do still need srcOffset, sinkOffset to be in the table / to be predicate parameters?
In the base case, the srcOffset and sinkOffset come from start and end, not from srcSinkLengthMap.
In the recursive case, the srcOffset from the previous iteration is unused (srcSinkLengthMap(_, start, /*here -> */ _, ...). The sinkOffset from the previous iteration is only bound to be the new srcOffset, which we just determined wasn't used in the next iteration.
srcOffset and sinkOffset are then also not used by the select
|
Two things to note about the multidimensional arrays:
|
MichaelRFairhurst
left a comment
MichaelRFairhurst
left a comment
There was a problem hiding this comment.
literally just minor tweaks!! This looks 🔥 great!
| /** | ||
| * This module provides classes and predicates for analyzing the size of buffers | ||
| * or objects from their base or a byte-offset, and identifying the potential for | ||
| * expressions accessing those buffers to overflow. |
There was a problem hiding this comment.
In this case, can we have c/common/src/codingatandards/c/OutOfBounds.qll import cpp.OutOfBounds ? A simple wrapper import file would be a reasonable easy refactor.
(or find each c query that imports OutOfBounds and update those)?
I think it's probably reasonable
| * @precision medium | ||
| * @problem.severity error | ||
| * @tags external/misra/id/rule-8-7-1 | ||
| * scope/system |
There was a problem hiding this comment.
looks like we're missing correctness
| /** | ||
| * Gets the declared length of this array. | ||
| */ | ||
| int getLength() { result = length } |
There was a problem hiding this comment.
This and the int length; on line 26 can be deleted now, right?
| sink.getNode() = end.getBasePointerNode() | ||
| | | ||
| srcOffset = start.getOffset() and | ||
| sinkOffset = end.getOffset() and |
There was a problem hiding this comment.
Nice!
Only thought now, do still need srcOffset, sinkOffset to be in the table / to be predicate parameters?
In the base case, the srcOffset and sinkOffset come from start and end, not from srcSinkLengthMap.
In the recursive case, the srcOffset from the previous iteration is unused (srcSinkLengthMap(_, start, /*here -> */ _, ...). The sinkOffset from the previous iteration is only bound to be the new srcOffset, which we just determined wasn't used in the next iteration.
srcOffset and sinkOffset are then also not used by the select
| array + | ||
| 4; // NON_COMPLIANT: pointer points more than one beyond the last element | ||
| int *invalid2 = | ||
| array - 1; // NON_COMPLIANT: pointer is outside boundary [FALSE_NEGATIVE] |
There was a problem hiding this comment.
No longer a false negative! 🎉
| strcat(buf1, " "); // NON_COMPLIANT - not null terminated | ||
| strcat(buf2, " "); // COMPLIANT | ||
| strcat(buf3, " "); // COMPLIANT | ||
| strcat(buf4, "12345"); // NON_COMPLIANT |
There was a problem hiding this comment.
Hmm, this is actually a FN
| "description": "Pointers obtained as result of performing arithmetic should point to an initialized object, or an element right next to the last element of an array.", | ||
| "kind": "path-problem", | ||
| "name": "Pointer arithmetic shall not form an invalid pointer", | ||
| "precision": "medium", |
There was a problem hiding this comment.
I'd be very tempted to say "high" precision.
Nicely done :)
| "severity": "error", | ||
| "short_name": "PointerArithmeticFormsAnInvalidPointer", | ||
| "tags": [ | ||
| "scope/system" |
There was a problem hiding this comment.
tags should probably include correctness and security, for this and below
| "tags": [ | ||
| "scope/system" | ||
| ] | ||
| }, |
There was a problem hiding this comment.
Consider adding an implementation scope that we only handle constant offsets for increased precision
| ) | ||
| select end, src, sink, | ||
| "This pointer accesses element at index " + totalOffset + | ||
| " while the underlying object has length " + length + "." |
There was a problem hiding this comment.
One small thought on verbiage.
While "Object" is the right term here, generally speaking, I'm not sure devs think of objects as something that has a "length"... I'd say objects have a "size," and that size is in bytes, making "length" perhaps doubly confusing.
If you want to go with "length," I'd probably suggest "array" or "buffer." Though it could be confusing to say that &x is either.
Maybe something like, "Pointer formed that points to element X of an object contains Y elements" ?
Description
Implement Memory1 (
RULE-8-7-1).Change request type
.ql,.qll,.qlsor unit tests)Rules with added or modified queries
RULE-8-7-1Release change checklist
A change note (development_handbook.md#change-notes) is required for any pull request which modifies:
If you are only adding new rule queries, a change note is not required.
Author: Is a change note required?
🚨🚨🚨
Reviewer: Confirm that format of shared queries (not the .qll file, the
.ql file that imports it) is valid by running them within VS Code.
Reviewer: Confirm that either a change note is not required or the change note is required and has been added.
Query development review checklist
For PRs that add new queries or modify existing queries, the following checklist should be completed by both the author and reviewer:
Author
As a rule of thumb, predicates specific to the query should take no more than 1 minute, and for simple queries be under 10 seconds. If this is not the case, this should be highlighted and agreed in the code review process.
Reviewer
As a rule of thumb, predicates specific to the query should take no more than 1 minute, and for simple queries be under 10 seconds. If this is not the case, this should be highlighted and agreed in the code review process.