Description

In the Linux kernel, the following vulnerability has been resolved: svcrdma: Reject Write/Reply chunks with segcount 0 A peer can send a Write or Reply chunk whose segcount field is zero. xdr_check_write_chunk() only rejects segcount > rc_maxpages, so zero passes the range check, and xdr_inline_decode(stream, 0) returns the current (non-NULL) cursor without advancing. The function returns true and pcl_alloc_write() then links a struct svc_rdma_chunk with ch_segcount == 0 onto rc_write_pcl or rc_reply_pcl. An earlier patch in this series made pcl_for_each_segment() safe for ch_segcount == 0, so this no longer drives the memory walk it used to. Rejecting the malformed frame at the decode boundary is still worthwhile as defense in depth: it keeps degenerate zero-segment chunks off the parsed chunk lists entirely, so any future consumer that walks ch_segments directly cannot observe one, and it makes the zero-floor easy to backport to trees where the macro change is more intrusive. RFC 8166 has no meaning for a Write/Reply chunk that describes no remote buffer, so no legitimate client is affected. xdr_check_reply_chunk() funnels Reply chunks through xdr_check_write_chunk() and inherits the same rejection. pcl_alloc_write() also links each chunk onto the parsed chunk list before filling its segment array. If a future change weakens the segcount-0 rejection, an incomplete chunk is visible to consumers during the fill loop. Reorder so that list_add_tail() follows the segment fill loop, ensuring only fully-populated chunks appear on the list.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "9.1",
  "pubDate": "2026-09-24T16:17:18.253Z",
  "pubdate": "2026-09-24T16:17:18.253Z",
  "executiveSummary": "The vulnerability resides in the Linux kernel's Remote Direct Memory Access (RDMA) service provider (svcrdma).\nIt involves improper validation of the 'segcount' field within Write and Reply RDMA chunks, allowing for the processing of zero-segment chunks.\nThis represents a security flaw where malformed RPC-over-RDMA frames are accepted by the XDR decoding logic, potentially leading to inconsistencies in parsed chunk lists.\nThe primary risk is a potential for memory corruption or logic errors if future consumers of the parsed chunk list assume the presence of valid segments.\nAn unauthenticated remote attacker can trigger this condition by sending a specially crafted RDMA message containing a zero-count chunk.\nExploitation requires network access to the target system hosting an NFS-over-RDMA service; it does not inherently require authentication but depends on the target's acceptance of malformed XDR sequences.",
  "technicalDetails": "The vulnerability originates in the XDR decoding layer of the svcrdma implementation, specifically within 'xdr_check_write_chunk()' and its associated 'xdr_check_reply_chunk()'.\nThe 'segcount' field, which dictates the number of segments in an RDMA Write or Reply chunk, was previously subject only to an upper-bound check against 'rc_maxpages'. Crucially, the check failed to enforce a minimum value of one.\nWhen a peer sends a chunk with 'segcount' set to zero, 'xdr_inline_decode(stream, 0)' succeeds by returning the current cursor position without advancing. This causes the decoder to incorrectly signal that the chunk is valid.\nThe function 'pcl_alloc_write()' then allocates and initializes a 'struct svc_rdma_chunk' with 'ch_segcount == 0'. This structure is subsequently linked into 'rc_write_pcl' or 'rc_reply_pcl'.\nThe attack flow proceeds as follows: An attacker sends a crafted RDMA write request with a chunk header specifying zero segments. The kernel parses this malformed chunk because the logic fails to reject non-positive segment counts. Consequently, an incomplete or degenerate chunk entry is inserted into the parsed chunk list (PCL).\nThis creates an 'inconsistent state' scenario. While recent changes to 'pcl_for_each_segment()' mitigate immediate traversal crashes, the structural integrity of the PCL is compromised. Furthermore, because 'pcl_alloc_write()' historically linked the chunk to the list before populating the segment array, there existed a race condition where consumers could interact with a partially initialized object if the thread was preempted or if further logic bugs allowed access to the list prematurely.\nBy reordering the list operations to occur only after the segment fill loop, the implementation ensures atomicity of the object population. However, the root cause remains the acceptance of malformed XDR-encoded frames at the protocol boundaries. This behavior contradicts RFC 8166 specifications, which do not define semantic meaning for zero-segment chunks, effectively making these degenerate frames a vector for triggering kernel code paths that assume non-zero segment availability."
}