{ "schema_version": "1.6.1", "id": "RPI-CVE-2025-38439", "modified": "2026-10-09T11:39:44Z", "upstream": [ "CVE-2025-38439" ], "published": "2025-07-25T15:35:02Z", "summary": "bnxt_en: Set DMA unmap len correctly for XDP_REDIRECT", "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nbnxt_en: Set DMA unmap len correctly for XDP_REDIRECT\n\nWhen transmitting an XDP_REDIRECT packet, call dma_unmap_len_set()\nwith the proper length instead of 0. This bug triggers this warning\non a system with IOMMU enabled:\n\nWARNING: CPU: 36 PID: 0 at drivers/iommu/dma-iommu.c:842 __iommu_dma_unmap+0x159/0x170\nRIP: 0010:__iommu_dma_unmap+0x159/0x170\nCode: a8 00 00 00 00 48 c7 45 b0 00 00 00 00 48 c7 45 c8 00 00 00 00 48 c7 45 a0 ff ff ff ff 4c 89 45\nb8 4c 89 45 c0 e9 77 ff ff ff <0f> 0b e9 60 ff ff ff e8 8b bf 6a 00 66 66 2e 0f 1f 84 00 00 00 00\nRSP: 0018:ff22d31181150c88 EFLAGS: 00010206\nRAX: 0000000000002000 RBX: 00000000e13a0000 RCX: 0000000000000000\nRDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000\nRBP: ff22d31181150cf0 R08: ff22d31181150ca8 R09: 0000000000000000\nR10: 0000000000000000 R11: ff22d311d36c9d80 R12: 0000000000001000\nR13: ff13544d10645010 R14: ff22d31181150c90 R15: ff13544d0b2bac00\nFS: 0000000000000000(0000) GS:ff13550908a00000(0000) knlGS:0000000000000000\nCS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033\nCR2: 00005be909dacff8 CR3: 0008000173408003 CR4: 0000000000f71ef0\nPKRU: 55555554\nCall Trace:\n\n? show_regs+0x6d/0x80\n? __warn+0x89/0x160\n? __iommu_dma_unmap+0x159/0x170\n? report_bug+0x17e/0x1b0\n? handle_bug+0x46/0x90\n? exc_invalid_op+0x18/0x80\n? asm_exc_invalid_op+0x1b/0x20\n? __iommu_dma_unmap+0x159/0x170\n? __iommu_dma_unmap+0xb3/0x170\niommu_dma_unmap_page+0x4f/0x100\ndma_unmap_page_attrs+0x52/0x220\n? srso_alias_return_thunk+0x5/0xfbef5\n? xdp_return_frame+0x2e/0xd0\nbnxt_tx_int_xdp+0xdf/0x440 [bnxt_en]\n__bnxt_poll_work_done+0x81/0x1e0 [bnxt_en]\nbnxt_poll+0xd3/0x1e0 [bnxt_en]", "severity": [ { "type": "CVSS_V3", "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H" } ], "affected": [ { "package": { "ecosystem": "Raspberry Pi OS:12", "name": "linux" }, "versions": [ "1:6.1.47-1+rpt4" ], "ecosystem_specific": { "binaries": [ { "binary_name": "linux-image-6.1.0-rpi3-rpi-2712", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v6", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v7", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v7l", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v8", "binary_version": "1:6.1.47-1+rpt4" } ] }, "database_specific": { "per_version": { "1:6.1.47-1+rpt4": { "linux_commit": "655fc658a15ae7a6f37103754adb39ba52a9a14e", "upstream_version": "6.1.47", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.1.146", "compiled_filter": false } } } } ], "references": [ { "type": "ADVISORY", "url": "https://www.cve.org/CVERecord?id=CVE-2025-38439" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/e260f4d49370c85a4701d43c6d16b8c39f8b605f" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/16ae306602163fcb7ae83f2701b542e43c100cee" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/8d672a1a6bfc81fef9151925c9c0481f4acf4bec" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/f9eaf6d036075dc820520e1194692c0619b7297b" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/5909679a82cd74cf0343d9e3ddf4b6931aa7e613" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/f154e41e1d9d15ab21300ba7bbf0ebb5cb3b9c2a" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/50dad9909715094e7d9ca25e9e0412b875987519" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/3cdf199d4755d477972ee87110b2aebc88b3cfad" } ], "database_specific": { "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git", "upstream_modified": "2026-08-05T12:19:11Z", "severity": { "source": "kernel CNA", "cvss_v3_score": 9.8, "cvss_v3_severity": "Critical", "scenario": "AV:N - The faulty unmap is reached purely by inbound network traffic \u2014 a received packet drives the XDP program to XDP_REDIRECT, `xdp_do_redirect()` \u2192 `bq_xmit_all()` \u2192 `bnxt_xdp_xmit()` \u2192 `__bnxt_xmit_xdp_redirect()`, with the corrupt `dma_unmap_len` consumed in `bnxt_tx_int_xdp()` on the NAPI completion. A remote attacker on the internet sending packets through an XDP-redirect load balancer/router built on a Broadcom NetXtreme NIC triggers it with no local access.\nAC:L - Once the (common datacenter) configuration of a bnxt_en NIC with an attached XDP program and IOMMU enabled is present, *every* redirected frame deterministically records `len = 0` and unmaps incorrectly \u2014 no race, no memory-layout grooming, no timing window is involved. The attacker simply sends traffic, and can send it at line rate to reproduce and to drive IOVA/page-table desynchronization.\nPR:N - The entire path from packet receive through `xdp_do_redirect()` to `bnxt_tx_int_xdp()` runs in softirq/NAPI context on behalf of an anonymous remote peer; there is no credential check, socket, or capability gate anywhere along it. The attacker needs no account or authentication on the target host.\nUI:N - No victim action is required \u2014 the redirect and the subsequent TX completion happen automatically in the driver's NAPI poll as soon as the attacker's packets arrive.\nS:U - The mis-sized DMA unmap, the leaked IOMMU translation, and the resulting corruption/disclosure all land in the host kernel's own memory and its own IOVA domain, within a single security authority. No VM, sandbox, or guest-to-host boundary is crossed, so scope is unchanged.\nC:H - Because `dma_unmap_single(len=0)` leaves the IOMMU translation in place (or removes only part of it) while `xdp_return_frame()` releases the page, and the mis-sized `free_iova_fast()` recycles that IOVA into the wrong size bin, a later `dma_map_single()` on the same IOVA collides with the surviving PTE \u2014 which Intel's `__domain_mapping()` retains while still returning success \u2014 so the NIC reads the previously freed, now-reallocated page and transmits arbitrary kernel/user memory onto the wire.\nI:H - The same IOVA-to-physical desynchronization means device DMA lands on physical pages the kernel has already freed and handed to other subsystems, with the surviving translation permitting access the kernel believes was revoked; combined with `switch_to_super_page()` promotion tearing down page tables covering neighbouring live mappings, this is a memory-corruption primitive whose content is attacker-supplied packet data.\nA:H - The bug reproducibly fires `WARN_ON(unmapped != size)` in `__iommu_dma_unmap()` from IRQ/NAPI context on every redirected frame \u2014 a per-packet splat that panics outright under `panic_on_warn` \u2014 and the under-sized IOVA frees progressively exhaust the device's IOVA space until `dma_map_single()` fails and the NIC stops forwarding traffic." } } }