It might seem redundant to perform error control at both the Data Link layer and the Transport layer, but each operates at a fundamentally different scope, and neither alone can guarantee true end-to-end reliability.

Data Link layer error control is only per-hop (link-local)

• The Data Link layer's error checking (e.g. CRC) only verifies that a frame arrived correctly across a single physical link — from one node to its immediate neighbor (e.g. host to router, or router to router).
• Once a router receives a frame, verifies it, and strips the Data Link header, it re-encapsulates the data into a brand-new frame (with a new CRC) for the next link. This new frame's correctness only re-verifies that this specific hop was error-free.
• Crucially, the router's own internal processing (buffer memory corruption, software bugs, hardware faults during forwarding) is not covered by any Data Link layer check — an error introduced inside a router between de-framing and re-framing would go completely undetected by Data Link layer checks alone.

Transport layer error control provides genuine end-to-end assurance

• The Transport layer's error checking (e.g. TCP checksum) is computed once at the original source and verified only at the final destination — it isn't recomputed and reset at every intermediate hop.
• This means it can catch errors introduced anywhere along the entire path — not just on the physical wire, but also from bugs, memory corruption, or misconfigured devices inside any of the intermediate routers themselves.
• It is the only layer that gives the actual communicating applications (not just adjacent link partners) a guarantee that the data they exchange is correct.

Why not rely on just one layer

• Relying only on the Data Link layer leaves the "inside the router" gap uncovered, as explained above.
• Relying only on the Transport layer, without any Data Link layer error control, would mean that a badly corrupted frame due to noise on a single hop wouldn't be caught (and potentially retransmitted) until the entire end-to-end path's data arrives and fails the final Transport-layer check — far less efficient than catching and locally retransmitting the error immediately at the hop where it occurred.

Conclusion: Data Link layer error control provides fast, local, per-hop reliability (catching and fixing noise-induced errors quickly, hop by hop), while Transport layer error control provides the only genuine end-to-end guarantee, covering failures that occur anywhere in the path — including inside intermediate devices. The two are complementary, not redundant.