Fix Blob::offset upper-bound checks to be strict (< instead of <=)#7102
Open
Chessing234 wants to merge 1 commit intoBVLC:masterfrom
Open
Fix Blob::offset upper-bound checks to be strict (< instead of <=)#7102Chessing234 wants to merge 1 commit intoBVLC:masterfrom
Chessing234 wants to merge 1 commit intoBVLC:masterfrom
Conversation
Blob::offset(n, c, h, w) guards the inputs with CHECK_LE(n, num()), CHECK_LE(c, channels()), CHECK_LE(h, height()), CHECK_LE(w, width()). A valid index is in [0, dim), so these allow n == num(), c == channels() etc. to pass — one past the end of the buffer, producing a stale/OOB linear offset from `((n * channels() + c) * height() + h) * width() + w`. The vector overload of offset() already uses the correct pattern (CHECK_LT(indices[i], shape(i))). Switch the four legacy-accessor checks to CHECK_LT to match. Fixes BVLC#6391.
Chessing234
added a commit
to Chessing234/caffe
that referenced
this pull request
Apr 20, 2026
…ndex Issue BVLC#6391. The four-argument Blob::offset(n, c, h, w) has lower-bound guards of the form: CHECK_GE(n, 0); CHECK_GE(channels(), 0); CHECK_GE(height(), 0); CHECK_GE(width(), 0); Only the first line checks the index. The other three check the *dimensions* of the blob (channels(), height(), width()), which for any constructed blob are structurally non-negative -- these are vacuous asserts that fire only if the blob itself is malformed, and never guard the index the caller passed in. A negative c/h/w is silently accepted and then used to compute ((n * channels() + c) * height() + h) * width() + w pointing outside the buffer. The vector overload a few lines below already gets this right: CHECK_GE(indices[i], 0); CHECK_LT(indices[i], shape(i)); This change only touches the three misnamed CHECK_GE calls so the lower bounds match the vector overload and the bounds the call actually needs. Upper-bound hardening (CHECK_LE -> CHECK_LT) is out of scope; it is handled separately by BVLC#7102.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #6391.
Bug
In `include/caffe/blob.hpp`, the four-argument `Blob::offset(n, c, h, w)` guards its inputs with:
```cpp
CHECK_GE(n, 0);
CHECK_LE(n, num());
CHECK_GE(channels(), 0);
CHECK_LE(c, channels());
CHECK_GE(height(), 0);
CHECK_LE(h, height());
CHECK_GE(width(), 0);
CHECK_LE(w, width());
```
Root cause
A valid index is in `[0, dim)`, so `CHECK_LE(n, num())` allows `n == num()` (and likewise for `c/h/w`) to pass. The method then computes
```cpp
((n * channels() + c) * height() + h) * width() + w
```
which at `n == num()` is exactly one past the buffer's last element. The vector overload directly below already uses the correct pattern:
```cpp
CHECK_GE(indices[i], 0);
CHECK_LT(indices[i], shape(i));
```
Fix
Switch the four legacy-accessor upper bounds to `CHECK_LT` so the valid range is `[0, dim)`, matching the vector overload. The lower-bound `CHECK_GE` lines (which #6391 does not flag) are left alone to keep the change minimal.