aboutsummaryrefslogtreecommitdiff
path: root/safety.md
blob: 8a03652eb68d6638597c8b610d8888fc20cd8475 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
Each item in Wikipedia's [list of types of memory errors](https://en.wikipedia.org/wiki/Memory_safety#Types_of_memory_errors) and what Crowbar does to prevent them.

# Access errors

## Buffer overflow

Crowbar addresses buffer overflow with bounds checking.
In C, the type `char *` can point to a single character, a null-terminated string of unknown length, a buffer of fixed size, or nothing at all.
In Crowbar, the type `char *` can only point to either a single character or nothing at all.
If a buffer is declared as `char[50] name;` then it has type `char[50]`, and can be implicitly converted to `(char[50])*`, a pointer-to-50-chars.
If memory is dynamically allocated, it works as follows:

```crowbar
void process(size_t bufferSize, char[bufferSize] buffer) {
    // do some work with buffer, given that we know its size
}

int main(int argc, (char[1024?])[argc] argv) {
    size_t bufferSize = getBufferSize();
    (char[bufferSize])* buffer = malloc(bufferSize);
    process(bufferSize, buffer);
    free(buffer);
}
```

Note that `malloc` as part of the Crowbar standard library has signature `(void[size])* malloc(size_t size);` and so no cast is needed above.
In C, `buffer` in `main` would have type pointer-to-VLA-of-char, but `buffer` in `process` would have type VLA-of-char, and this conversion would emit a compiler warning.
However, in Crowbar, a `(T[N])*` is always implicitly convertible to `T[N]`, so no warning exists.
(This is translated into C by dereferencing `buffer` in `main`.)

Note as well that the type of `argv` is complicated.
This is because the elements of `argv` have unconstrained size.
TODO figure out if that's the right way to handle that

## Buffer over-read

bounds checking again

## Race condition

uhhhhh 🤷‍♀️

## Page fault

bounds checking, dubious-pointer checking

## Use after free

`free(x);` not followed by `x = NULL;` is a compiler error.
`owned` and `borrowed` keywords 

# Uninitialized variables

forbid them in syntax

## Null pointer dereference

dubious-pointer checking

## Wild pointers

dubious-pointer checking

# Memory leak

## Stack exhaustion

uhhhhhh 🤷‍♀️

## Heap exhaustion

that counts as error handling, just the `malloc`-shaped kind

## Double free

this is just use-after-free but the use is calling free on it

## Invalid free

don't do that

## Mismatched free

how does that even happen

## Unwanted aliasing

uhhh don't do that?