CVE-2026-72466 PUBLISHED

xprtrdma: Fix bcall rep leak and unbounded peek

Assigner: Linux
Reserved: 09.08.2026 Published: 15.08.2026 Updated: 17.08.2026

In the Linux kernel, the following vulnerability has been resolved:

xprtrdma: Fix bcall rep leak and unbounded peek

rpcrdma_is_bcall() decodes a reply's first words to decide whether the frame is a backchannel call. Two issues in that decode path let a short or malformed reply leak the receive buffer and drain the Receive queue.

First, the speculative peek

<pre>p = xdr_inline_decode(xdr, 0); /* five p++ reads follow */ </pre>

asks xdr_inline_decode() for zero bytes, which returns xdr->p without consulting xdr->end. The five subsequent __be32 reads can then walk up to 20 bytes past the wire payload into stale regbuf contents and misclassify the reply as a backchannel call.

Second, after the post-peek

<pre>p = xdr_inline_decode(xdr, 3 * sizeof(*p)); if (unlikely(!p)) return true; </pre>

the short-header arm returns true without calling rpcrdma_bc_receive_call(). The contract with the caller is that a true return transfers ownership of rep to the backchannel path:

<pre>rpcrdma_reply_handler() if (rpcrdma_is_bcall(r_xprt, rep)) return; /* bare return, skips out_post */ ... out_post: rpcrdma_post_recvs(r_xprt, credits + ...); </pre>

Because rpcrdma_bc_receive_call() never ran, no one took rep, but rpcrdma_reply_handler still bare-returns past rpcrdma_rep_put() and rpcrdma_post_recvs(). The rep, with its persistently DMA-mapped receive buffer, is orphaned on rb_all_reps and freed only at transport teardown. This completion reposts nothing, so its slot is reclaimed only when a later forward-channel reply reaches out_post and rpcrdma_post_recvs() allocates a fresh rep to backfill; absent that traffic the Receive queue drains and the peer's Sends draw RNR NAKs.

Fix by consulting xdr->end after the zero-length peek so the five __be32 reads cannot run unless 20 bytes of wire payload remain. A byte-precise comparison against xdr->end is required because a non-4-aligned receive rounds the stream's word count up past the true payload. Also return false from the short-header arm so the reply falls through the normal out_norqst cleanup chain (rpcrdma_rep_put() plus rpcrdma_post_recvs()).

Metrics

CVSS Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 9.8

AV:N - The bug is reached when a remote RDMA peer's inbound message completes in rpcrdma_wc_receive() and rpcrdma_reply_handler() decodes it; a malicious or compromised NFS-over-RDMA server can send crafted short rdma_msg replies over the established RPC/RDMA connection without local access. AC:L - On an active NFSv4.1+ RPC/RDMA session the attacker fully controls inbound replies and can repeatedly send short/malformed rdma_msg frames to hit the zero-length peek and short-header return-true paths, reliably leaking receive buffers and draining the Receive queue. PR:N - Exploitation requires only network reachability as the connected RDMA/RPC peer (e.g., a malicious NFS server); no local account, mount privileges, or namespaced capabilities on the victim NFS client are needed to deliver the malformed RDMA replies. UI:N - Once an NFS-over-RDMA mount and NFSv4.1 backchannel session exist, triggering the flaw needs no additional victim action beyond the attacker sending malicious RDMA receive payloads during normal client-server communication. S:U - The vulnerability and its impact remain within the NFS client's kernel SUNRPC/xprtrdma transport context; it does not cross a VM, container, or IOMMU security boundary to affect other authorities. C:H - xdr_inline_decode(xdr,0) permits up to five subsequent __be32 reads past the wire payload into stale receive-buffer contents, an out-of-bounds read that can misclassify replies and expose remnants of prior RPC/NFS data to unintended processing paths. I:H - Stale-buffer-assisted misclassification can treat a forward-channel reply as a backchannel RPC_CALL, a type confusion that routes attacker-influenced bytes into rpcrdma_bc_receive_call()/NFSv4 callback handling and can corrupt client session state. A:H - Each successful trigger orphans a DMA-mapped rpcrdma_rep and skips rpcrdma_post_recvs(), monotonically draining the RDMA Receive queue until the peer gets RNR NAKs and the NFS-over-RDMA connection becomes unusable, enabling sustained remote denial of service.

Product Status

Vendor Linux
Product Linux
Versions Default: unaffected
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to 0cee8f9c3b14bd6dee9c4310090a7f45b89b834f (excl.)
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to 7afc2f8d2fd9394724df9eaf22ce7a71029a5fba (excl.)
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to 88b5346284a184a6b7d019912232a571d672d3e3 (excl.)
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to 07aa506436be7634e381e1e1f6d0efa9efc81ecc (excl.)
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to d7a2870dde3bb09d51d6b9c877642996ad6b92dd (excl.)
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to 118a16a18c59f7ad8084b2d13988839b669fca10 (excl.)
  • affected from 41c8f70f5a3db7e06179186b6525fd9ee1d7d314 to c7653d5cebc8492c77ec0415b5e9c0fb3e644bc6 (excl.)
Vendor Linux
Product Linux
Versions Default: affected
  • Version 4.14 is affected
  • unaffected from 0 to 4.14 (excl.)
  • unaffected from 5.15.212 to 5.15.* (incl.)
  • unaffected from 6.1.178 to 6.1.* (incl.)
  • unaffected from 6.6.145 to 6.6.* (incl.)
  • unaffected from 6.12.97 to 6.12.* (incl.)
  • unaffected from 6.18.40 to 6.18.* (incl.)
  • unaffected from 7.1.5 to 7.1.* (incl.)
  • unaffected from 7.2 to * (incl.)

References