Add EBPFOpts.PacketStartOffset for non-zero-based PacketStart - #45
Closed
Dhiver wants to merge 1 commit into
Closed
Conversation
ToEBPF's indirect packet-load guard assumed PacketStart points at offset 0 of the packet. It bounds the data-dependent index against maxStartOffset() = maxPacketOffset - length + 1, relying on the verifier's rule that a packet pointer only gets a range when off + umax_value <= MAX_PACKET_OFF (0xFFFF). When the caller sets PacketStart to a pointer that already carries a fixed offset (e.g. it points at the L3 header rather than the start of the frame), that fixed offset is added on top of the guarded index. An indirect load with a data-dependent index (such as walking TCP options) can then push off + umax_value past MAX_PACKET_OFF, and the verifier rejects the program with "invalid access to packet" even though the access is safe. Add EBPFOpts.PacketStartOffset so callers can declare PacketStart's fixed offset. maxStartOffsetWithBase() reserves that many bytes of the packet-offset budget in the indirect guard so the bounded range stays within MAX_PACKET_OFF. It fails closed (returns 0, i.e. noMatch) when no room remains, rather than emitting an unbounded, unverifiable load. maxStartOffsetWithBase also rejects a negative base offset (PacketStart cannot precede the packet) and fails closed, so bogus input never widens the guard's budget. PacketStartOffset defaults to 0, preserving existing behaviour.
arthurfabre
added a commit
that referenced
this pull request
Jul 27, 2026
The verifier rejects any programs that accesses packets with offsets greater than 0xFFFF. For absolute loads we can detect this statically, for indirect loads we insert a runtime check. But our checks didn't account for any pre-existing offset to the packet pointer the caller passes in! (eg because the caller has parsed / skipped the ethernet header). This must be accounted for in the maximum offset check. Add an option to both EBPFOpts and COpts to allow users to specify the maximum offset the packet pointer may have, so it can be included in the static analysis for absolute loads, and the runtime checks for indirect loads. Thank you to @Dhiver for figuring this out in #45.
arthurfabre
added a commit
that referenced
this pull request
Jul 27, 2026
The verifier rejects any programs that accesses packets with offsets greater than 0xFFFF. For absolute loads we can detect this statically, for indirect loads we insert a runtime check. But our checks didn't account for any pre-existing offset to the packet pointer the caller passes in! (eg because the caller has parsed / skipped the ethernet header). This must be accounted for in the maximum offset check. Add an option to both EBPFOpts and COpts to allow users to specify the maximum offset the packet pointer may have, so it can be included in the static analysis for absolute loads, and the runtime checks for indirect loads. Thank you to @Dhiver for figuring this out in #45.
arthurfabre
added a commit
that referenced
this pull request
Jul 27, 2026
The verifier rejects any programs that accesses packets with offsets greater than 0xFFFF. For absolute loads we can detect this statically, for indirect loads we insert a runtime check. But our checks didn't account for any pre-existing offset to the packet pointer the caller passes in! (eg because the caller has parsed / skipped the ethernet header). This must be accounted for in the maximum offset check. Add an option to both EBPFOpts and COpts to allow users to specify the maximum offset the packet pointer may have, so it can be included in the static analysis for absolute loads, and the runtime checks for indirect loads. Thank you to @Dhiver for figuring this out in #45.
Collaborator
|
Thanks for digging into this! We need to handle this when generating C too, and for absolute loads. We also need to exercise this in tests that load programs that use indirect loads with packet offsets. I've done that in #47. |
arthurfabre
added a commit
that referenced
this pull request
Jul 28, 2026
The verifier rejects any programs that accesses packets with offsets greater than 0xFFFF. For absolute loads we can detect this statically, for indirect loads we insert a runtime check. But our checks didn't account for any pre-existing offset to the packet pointer the caller passes in! (eg because the caller has parsed / skipped the ethernet header). This must be accounted for in the maximum offset check. Add an option to both EBPFOpts and COpts to allow users to specify the maximum offset the packet pointer may have, so it can be included in the static analysis for absolute loads, and the runtime checks for indirect loads. Thank you to @Dhiver for figuring this out in #45.
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
2 participants
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.
ToEBPF's indirect packet-load guard assumed PacketStart points at offset 0 of the packet. It bounds the data-dependent index against maxStartOffset() = maxPacketOffset - length + 1, relying on the verifier's rule that a packet pointer only gets a range when off + umax_value <= MAX_PACKET_OFF (0xFFFF).
When the caller sets PacketStart to a pointer that already carries a fixed offset (e.g. it points at the L3 header rather than the start of the frame), that fixed offset is added on top of the guarded index. An indirect load with a data-dependent index (such as walking TCP options) can then push off + umax_value past MAX_PACKET_OFF, and the verifier rejects the program with "invalid access to packet" even though the access is safe.
Add EBPFOpts.PacketStartOffset so callers can declare PacketStart's fixed offset. maxStartOffsetWithBase() reserves that many bytes of the packet-offset budget in the indirect guard so the bounded range stays within MAX_PACKET_OFF. It fails closed (returns 0, i.e. noMatch) when no room remains, rather than emitting an unbounded, unverifiable load. maxStartOffsetWithBase also rejects a negative base offset (PacketStart cannot precede the packet) and fails closed, so bogus input never widens the guard's budget.
PacketStartOffset defaults to 0, preserving existing behaviour.