-
Notifications
You must be signed in to change notification settings - Fork 352
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Check that offset is not too big, check projection offset to be inbounds #447
Comments
Or maybe projections can never leave the bounds... I have not been able to exploit this. We currently miss the UB in use std::mem;
fn main() {
let slice: &[u8; 5] = &[0; 5];
let slice: &[u8; 10] = unsafe { mem::transmute(slice) };
let _x = &slice[7];
} but IMHO this is a case for validating values during computation, not for adding bounds checks to projections. |
Also see #437 |
See the discussion at rust-lang/nomicon#149 (comment) and rust-lang/nomicon#149 (comment): for now we probably don't actually need to do any checks on the offsets here if we instead check that when we dereference ( It will be interesting to see if there is code in our test suite / libstd violating this. Also, here's another example: fn main() {
let local = 5u8;
let ptr = (&local as *const u8).wrapping_sub(1) as *const (u8, u8);
let _ref = unsafe { &(*ptr).1 };
} And with rust-lang/rfcs#2582, here's another one: fn main() {
let ptr: *const (u8, u8) = 16usize as *const _;
let ptr2 = &raw const (*ptr).1;
} |
Miri: Check that a ptr is aligned and inbounds already when evaluating `*` This syncs Miri with what the Nomicon and the Reference say, and resolves rust-lang/miri#447. Also this would not have worked without rust-lang#62982 due to new cycles. ;) r? @oli-obk
Miri: Check that a ptr is aligned and inbounds already when evaluating `*` This syncs Miri with what the Nomicon and the Reference say, and resolves rust-lang/miri#447. Also this would not have worked without rust-lang#62982 due to new cycles. ;) r? @oli-obk
Miri: Check that a ptr is aligned and inbounds already when evaluating `*` This syncs Miri with what the Nomicon and the Reference say, and resolves rust-lang/miri#447. Also this would not have worked without rust-lang#62982 due to new cycles. ;) r? @oli-obk
Miri: Check that a ptr is aligned and inbounds already when evaluating `*` This syncs Miri with what the Nomicon and the Reference say, and resolves rust-lang/miri#447. Also this would not have worked without rust-lang#62982 due to new cycles. ;) r? @oli-obk
Miri: Check that a ptr is aligned and inbounds already when evaluating `*` This syncs Miri with what the Nomicon and the Reference say, and resolves rust-lang/miri#447. Also this would not have worked without rust-lang#62982 due to new cycles. ;) r? @oli-obk
adjust tests for eager pointer checks on deref The Miri side of rust-lang/rust#63075. Fixes #447.
According to rust-lang/rust#53676, there is an upper bound to what you can do with
offset
. miri should check that.Related question: Shouldn't the
place_field
method in the miri engine checkpointer_offset_inbounds
? That will be a heavy perf hit, but I think we might currently be missing out on some UB.The text was updated successfully, but these errors were encountered: