Vulnerability-Lookup

GHSA-HW95-CHC9-4W36

Vulnerability from github – Published: 2025-08-22 18:31 – Updated: 2025-11-26 18:31

Details

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

nvmet: pci-epf: Do not complete commands twice if nvmet_req_init() fails

Have nvmet_req_init() and req->execute() complete failed commands.

Description of the problem: nvmet_req_init() calls __nvmet_req_complete() internally upon failure, e.g., unsupported opcode, which calls the "queue_response" callback, this results in nvmet_pci_epf_queue_response() being called, which will call nvmet_pci_epf_complete_iod() if data_len is 0 or if dma_dir is different from DMA_TO_DEVICE. This results in a double completion as nvmet_pci_epf_exec_iod_work() also calls nvmet_pci_epf_complete_iod() when nvmet_req_init() fails.

Steps to reproduce: On the host send a command with an unsupported opcode with nvme-cli, For example the admin command "security receive" $ sudo nvme security-recv /dev/nvme0n1 -n1 -x4096

This triggers a double completion as nvmet_req_init() fails and nvmet_pci_epf_queue_response() is called, here iod->dma_dir is still in the default state of "DMA_NONE" as set by default in nvmet_pci_epf_alloc_iod(), so nvmet_pci_epf_complete_iod() is called. Because nvmet_req_init() failed nvmet_pci_epf_complete_iod() is also called in nvmet_pci_epf_exec_iod_work() leading to a double completion. This not only sends two completions to the host but also corrupts the state of the PCI NVMe target leading to kernel oops.

This patch lets nvmet_req_init() and req->execute() complete all failed commands, and removes the double completion case in nvmet_pci_epf_exec_iod_work() therefore fixing the edge cases where double completions occurred.

Severity

5.5 (Medium)


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

Show details on source website

JSON

To clipboard

{
  "affected": [],
  "aliases": [
    "CVE-2025-38658"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-908"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2025-08-22T16:15:40Z",
    "severity": "MODERATE"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nnvmet: pci-epf: Do not complete commands twice if nvmet_req_init() fails\n\nHave nvmet_req_init() and req-\u003eexecute() complete failed commands.\n\nDescription of the problem:\nnvmet_req_init() calls __nvmet_req_complete() internally upon failure,\ne.g., unsupported opcode, which calls the \"queue_response\" callback,\nthis results in nvmet_pci_epf_queue_response() being called, which will\ncall nvmet_pci_epf_complete_iod() if data_len is 0 or if dma_dir is\ndifferent from DMA_TO_DEVICE. This results in a double completion as\nnvmet_pci_epf_exec_iod_work() also calls nvmet_pci_epf_complete_iod()\nwhen nvmet_req_init() fails.\n\nSteps to reproduce:\nOn the host send a command with an unsupported opcode with nvme-cli,\nFor example the admin command \"security receive\"\n$ sudo nvme security-recv /dev/nvme0n1 -n1 -x4096\n\nThis triggers a double completion as nvmet_req_init() fails and\nnvmet_pci_epf_queue_response() is called, here iod-\u003edma_dir is still\nin the default state of \"DMA_NONE\" as set by default in\nnvmet_pci_epf_alloc_iod(), so nvmet_pci_epf_complete_iod() is called.\nBecause nvmet_req_init() failed nvmet_pci_epf_complete_iod() is also\ncalled in nvmet_pci_epf_exec_iod_work() leading to a double completion.\nThis not only sends two completions to the host but also corrupts the\nstate of the PCI NVMe target leading to kernel oops.\n\nThis patch lets nvmet_req_init() and req-\u003eexecute() complete all failed\ncommands, and removes the double completion case in\nnvmet_pci_epf_exec_iod_work() therefore fixing the edge cases where\ndouble completions occurred.",
  "id": "GHSA-hw95-chc9-4w36",
  "modified": "2025-11-26T18:31:01Z",
  "published": "2025-08-22T18:31:22Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-38658"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/746d0ac5a07d5da952ef258dd4d75f0b26c96476"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/a535c0b10060bc8c174a7964b0f98064ee0c4774"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H",
      "type": "CVSS_V3"
    }
  ]
}

CVE-2025-38658 (GCVE-0-2025-38658)

Vulnerability from cvelistv5 – Published: 2025-08-22 16:01 – Updated: 2026-08-05 12:03

Title

nvmet: pci-epf: Do not complete commands twice if nvmet_req_init() fails

Summary

In the Linux kernel, the following vulnerability has been resolved: nvmet: pci-epf: Do not complete commands twice if nvmet_req_init() fails Have nvmet_req_init() and req->execute() complete failed commands. Description of the problem: nvmet_req_init() calls __nvmet_req_complete() internally upon failure, e.g., unsupported opcode, which calls the "queue_response" callback, this results in nvmet_pci_epf_queue_response() being called, which will call nvmet_pci_epf_complete_iod() if data_len is 0 or if dma_dir is different from DMA_TO_DEVICE. This results in a double completion as nvmet_pci_epf_exec_iod_work() also calls nvmet_pci_epf_complete_iod() when nvmet_req_init() fails. Steps to reproduce: On the host send a command with an unsupported opcode with nvme-cli, For example the admin command "security receive" $ sudo nvme security-recv /dev/nvme0n1 -n1 -x4096 This triggers a double completion as nvmet_req_init() fails and nvmet_pci_epf_queue_response() is called, here iod->dma_dir is still in the default state of "DMA_NONE" as set by default in nvmet_pci_epf_alloc_iod(), so nvmet_pci_epf_complete_iod() is called. Because nvmet_req_init() failed nvmet_pci_epf_complete_iod() is also called in nvmet_pci_epf_exec_iod_work() leading to a double completion. This not only sends two completions to the host but also corrupts the state of the PCI NVMe target leading to kernel oops. This patch lets nvmet_req_init() and req->execute() complete all failed commands, and removes the double completion case in nvmet_pci_epf_exec_iod_work() therefore fixing the edge cases where double completions occurred.

Severity

8.8 (High)


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

Assigner

Linux mitre root

References

2 references

URL	Tags
https://git.kernel.org/stable/c/a535c0b10060bc8c1…
https://git.kernel.org/stable/c/746d0ac5a07d5da95…

Impacted products

2 products

Vendor	Product	Version
Linux	Linux	Affected: 0faa0fe6f90ea59b10d1b0f15ce0eb0c18eff186 , < a535c0b10060bc8c174a7964b0f98064ee0c4774 (git) Affected: 0faa0fe6f90ea59b10d1b0f15ce0eb0c18eff186 , < 746d0ac5a07d5da952ef258dd4d75f0b26c96476 (git)
Linux	Linux	Affected: 6.14 Unaffected: 0 , < 6.14 (semver) Unaffected: 6.16.1 , ≤ 6.16.* (semver) Unaffected: 6.17 , ≤ * (original_commit_for_fix)

Show details on NVD website

JSON

To clipboard

{
  "containers": {
    "cna": {
      "affected": [
        {
          "defaultStatus": "unaffected",
          "product": "Linux",
          "programFiles": [
            "drivers/nvme/target/pci-epf.c"
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "vendor": "Linux",
          "versions": [
            {
              "lessThan": "a535c0b10060bc8c174a7964b0f98064ee0c4774",
              "status": "affected",
              "version": "0faa0fe6f90ea59b10d1b0f15ce0eb0c18eff186",
              "versionType": "git"
            },
            {
              "lessThan": "746d0ac5a07d5da952ef258dd4d75f0b26c96476",
              "status": "affected",
              "version": "0faa0fe6f90ea59b10d1b0f15ce0eb0c18eff186",
              "versionType": "git"
            }
          ]
        },
        {
          "defaultStatus": "affected",
          "product": "Linux",
          "programFiles": [
            "drivers/nvme/target/pci-epf.c"
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "vendor": "Linux",
          "versions": [
            {
              "status": "affected",
              "version": "6.14"
            },
            {
              "lessThan": "6.14",
              "status": "unaffected",
              "version": "0",
              "versionType": "semver"
            },
            {
              "lessThanOrEqual": "6.16.*",
              "status": "unaffected",
              "version": "6.16.1",
              "versionType": "semver"
            },
            {
              "lessThanOrEqual": "*",
              "status": "unaffected",
              "version": "6.17",
              "versionType": "original_commit_for_fix"
            }
          ]
        }
      ],
      "cpeApplicability": [
        {
          "nodes": [
            {
              "cpeMatch": [
                {
                  "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
                  "versionEndExcluding": "6.16.1",
                  "versionStartIncluding": "6.14",
                  "vulnerable": true
                },
                {
                  "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
                  "versionEndExcluding": "6.17",
                  "versionStartIncluding": "6.14",
                  "vulnerable": true
                }
              ],
              "negate": false,
              "operator": "OR"
            }
          ]
        }
      ],
      "descriptions": [
        {
          "lang": "en",
          "value": "In the Linux kernel, the following vulnerability has been resolved:\n\nnvmet: pci-epf: Do not complete commands twice if nvmet_req_init() fails\n\nHave nvmet_req_init() and req-\u003eexecute() complete failed commands.\n\nDescription of the problem:\nnvmet_req_init() calls __nvmet_req_complete() internally upon failure,\ne.g., unsupported opcode, which calls the \"queue_response\" callback,\nthis results in nvmet_pci_epf_queue_response() being called, which will\ncall nvmet_pci_epf_complete_iod() if data_len is 0 or if dma_dir is\ndifferent from DMA_TO_DEVICE. This results in a double completion as\nnvmet_pci_epf_exec_iod_work() also calls nvmet_pci_epf_complete_iod()\nwhen nvmet_req_init() fails.\n\nSteps to reproduce:\nOn the host send a command with an unsupported opcode with nvme-cli,\nFor example the admin command \"security receive\"\n$ sudo nvme security-recv /dev/nvme0n1 -n1 -x4096\n\nThis triggers a double completion as nvmet_req_init() fails and\nnvmet_pci_epf_queue_response() is called, here iod-\u003edma_dir is still\nin the default state of \"DMA_NONE\" as set by default in\nnvmet_pci_epf_alloc_iod(), so nvmet_pci_epf_complete_iod() is called.\nBecause nvmet_req_init() failed nvmet_pci_epf_complete_iod() is also\ncalled in nvmet_pci_epf_exec_iod_work() leading to a double completion.\nThis not only sends two completions to the host but also corrupts the\nstate of the PCI NVMe target leading to kernel oops.\n\nThis patch lets nvmet_req_init() and req-\u003eexecute() complete all failed\ncommands, and removes the double completion case in\nnvmet_pci_epf_exec_iod_work() therefore fixing the edge cases where\ndouble completions occurred."
        }
      ],
      "metrics": [
        {
          "cvssV3_1": {
            "baseScore": 8.8,
            "baseSeverity": "HIGH",
            "vectorString": "CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H",
            "version": "3.1"
          },
          "scenarios": [
            {
              "lang": "en",
              "value": "AV:A - The vulnerable code is the NVMe PCI *endpoint* target, driven entirely by NVMe commands submitted by a separate root-complex host across the PCIe link, so the attacker is external software on a directly-attached peer machine rather than a user of the endpoint system. In the deployments this driver targets (disaggregated/composable PCIe storage, and endpoint devices VFIO-passed-through to untrusted guests) no physical manipulation is needed at attack time \u2014 only presence on the same logically adjacent PCIe fabric.\nAC:L - A single NVMe command with an opcode the target does not implement deterministically fails nvmet_req_init() and produces the double completion; the commit itself gives a one-line nvme-cli reproducer. No timing, memory-layout, or victim-state condition outside the attacker\u0027s control is involved.\nPR:N - NVMe over PCIe has no authentication or authorization step \u2014 the host merely enables CC.EN, creates queues, and writes a doorbell \u2014 so the attacker holds no credentials at all on the vulnerable endpoint system. Any command submitter on the attached host or a passed-through guest reaches the failing path.\nUI:N - The endpoint\u0027s SQ-polling work item picks up the malformed command and dispatches it automatically. No action by any user or administrator on the victim endpoint system is required.\nS:U - The corruption is confined to the endpoint kernel\u0027s own heap and driver state, the same security authority that contains the vulnerable code. The cross-machine nature of the attack is already captured by the Attack Vector metric.\nC:H - The dangling CQ list entry causes nvmet_pci_epf_cq_work() to re-read a freed iod and build a completion entry from it (including a dereference of a possibly reallocated iod-\u003esq), which is memcpy_toio()\u0027d straight into the attacker host\u0027s completion queue, leaking reused kernel heap contents. More broadly, the double free and iod aliasing yield general heap-corruption primitives leverageable for arbitrary kernel memory disclosure.\nI:H - The same iod is returned to the mempool twice, so two concurrent commands are handed the identical object and its data_segs buffers are kfree()\u0027d twice in shared kmalloc caches, with the attacker controlling those buffers\u0027 sizes and contents via PRP/SGL lists and DMA. A double free plus use-after-free write on an aliased object is a well-established path to controlled kernel writes and control-flow hijack.\nA:H - The commit explicitly states the double completion \"corrupts the state of the PCI NVMe target leading to kernel oops,\" and CONFIG_DEBUG_LIST additionally trips on the duplicate list_add. The attacker can repeat the triggering command at will to panic or wedge the endpoint system."
            }
          ]
        }
      ],
      "providerMetadata": {
        "dateUpdated": "2026-08-05T12:03:45.053Z",
        "orgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
        "shortName": "Linux"
      },
      "references": [
        {
          "url": "https://git.kernel.org/stable/c/a535c0b10060bc8c174a7964b0f98064ee0c4774"
        },
        {
          "url": "https://git.kernel.org/stable/c/746d0ac5a07d5da952ef258dd4d75f0b26c96476"
        }
      ],
      "title": "nvmet: pci-epf: Do not complete commands twice if nvmet_req_init() fails",
      "x_generator": {
        "engine": "bippy-1.2.0"
      }
    }
  },
  "cveMetadata": {
    "assignerOrgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
    "assignerShortName": "Linux",
    "cveId": "CVE-2025-38658",
    "datePublished": "2025-08-22T16:01:01.651Z",
    "dateReserved": "2025-04-16T04:51:24.031Z",
    "dateUpdated": "2026-08-05T12:03:45.053Z",
    "state": "PUBLISHED"
  },
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2"
}

Sightings

Author	Source	Type	Date	Other

Nomenclature

Seen: The vulnerability was mentioned, discussed, or observed by the user.
Confirmed: The vulnerability has been validated from an analyst's perspective.
Published Proof of Concept: A public proof of concept is available for this vulnerability.
Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
Not confirmed: The user expressed doubt about the validity of the vulnerability.
Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Detection rules are retrieved from Rulezet.

The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base model, served locally by ML-Gateway. They have not been verified by an analyst and are provided for guidance only.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.

Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.