{
  "id": "CVE-2026-89596",
  "url": "https://poc.griffinguard.io/cve/CVE-2026-89596",
  "published_at": "2026-09-11T19:44:58.797Z",
  "title": "forcedeth: fix off-by-one when saving/restoring non-PCI config space",
  "vendor": "Linux",
  "product": "Linux",
  "cvss": {
    "score": 7.1,
    "source": "cna",
    "vector": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H"
  },
  "kev": {
    "in_kev": false,
    "added_at": null,
    "days_after_publication": null
  },
  "first_poc": null,
  "poc_repo_count": 0,
  "exploit_count": 0,
  "suspicious_repo_count": 0,
  "edge_device": null,
  "description": "In the Linux kernel, the following vulnerability has been resolved:\n\nforcedeth: fix off-by-one when saving/restoring non-PCI config space\n\nnv_suspend() and nv_resume() walk the non-PCI configuration space with\n\n\tfor (i = 0; i \u003c= np-\u003eregister_size/sizeof(u32); i++)\n\nwhich runs one iteration too many. saved_config_space is declared as\n\n\tu32 saved_config_space[NV_PCI_REGSZ_MAX/4];\n\nand NV_PCI_REGSZ_VER3 is equal to NV_PCI_REGSZ_MAX (0x604), so on a VER3\ndevice register_size/sizeof(u32) is exactly the array length and the last\niteration addresses one element past the end.\n\nThe element it lands on is np-\u003ename_rx[0..3]: saved_config_space[] is\nfollowed immediately by char name_rx[IFNAMSIZ + 3], and char needs no\npadding. Nothing observable is corrupted by that, because nv_request_irq()\nrewrites name_rx with sprintf() before it is ever passed to request_irq().\nThe bug is the out-of-bounds access itself, which UBSAN reports and which\nCONFIG_UBSAN_TRAP=y turns into a trap that aborts the running kernel code,\nplus an MMIO read and, on resume, an MMIO writel() to base + 0x604, one\ndword past the range the driver mapped:\n\n\tnp-\u003ebase = ioremap(addr, np-\u003eregister_size);\n\nVER1 and VER2 devices stay inside the array, but they too get the stray\nread and the stray write one dword past their own window.\n\nCaught by UBSAN on an Apple Macmini3,1 (MCP79) during a deep S3 cycle.\nThe splat below is trimmed: the build path in the file name, the CPU\nand taint lines, the Workqueue line, the \"?\" hint frames, and the\nframes below device_suspend are all cut. The kernel was tainted, with\nan out-of-tree nouveau and CPU_OUT_OF_SPEC; forcedeth itself was the\nstock module.\n\n  UBSAN: array-index-out-of-bounds in drivers/net/ethernet/nvidia/forcedeth.c:6225:25\n  index 385 is out of range for type 'u32 [385]'\n  Call Trace:\n   dump_stack_lvl+0x5d/0x80\n   ubsan_epilogue+0x5/0x2b\n   __ubsan_handle_out_of_bounds.cold+0x54/0x59\n   __this_module+0xe398c/0xe9010 [forcedeth]\n   pci_pm_suspend+0x80/0x170\n   dpm_run_callback+0x51/0x160\n   device_suspend+0x1a2/0x4a0\n   ...\n\nBoth loops are hit. UBSAN reports each source location only once per module\nload (__ubsan_handle_out_of_bounds() calls suppress_report(), which does\ntest_and_set_bit(REPORTED_BIT, ...) on the struct source_location), so the\ntwo splats land in the first S3 cycle after the module is loaded and later\ncycles are silent even though the access still runs off the end every time.\nIn that first cycle line 6225 is reported from pci_pm_suspend and line 6240\nfrom pci_pm_resume.\n\nThe same off-by-one was fixed in nv_get_regs() by commit ba9aa134287f\n(\"forcedeth: fix buffer overflow\") in 2012; these two loops were missed.\nThe suspend and resume side was reported on LKML in September 2013 by Marc\nWeber, with the same analysis and the same one-character fix, but the patch\nwas attached rather than sent inline and the thread ended there.\n\nUse \u003c instead of \u003c=, which saves and restores exactly register_size bytes.",
  "cvss_by_source": {
    "cna": 7.1
  },
  "repos": [],
  "remediation": {
    "affected": [
      {
        "vendor": "Linux",
        "product": "Linux",
        "ranges": [
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c effd589568b2c4399894a23000286210c60af6d7",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c 702aa4a44c87026b97b1da5c5ca80e1fa8a9875a",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c fc71c6c3d3e3bb17474dcd1162aea6783a452a64",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c c4d5953582463f45e48b14ef9815e5fb636f7e1f",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c 9379f8527ca60d92715c90b32f1b477de9ec32cb",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c 0c3f4544ff3873594c8af1b907b4ac4e6d5ce005",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c c4f196bfeedd71e56aba4b63bd8f919edb2f7056",
          "\u003e= 1a1ca86158eee303af5270338695f90bc7ae02b3, \u003c 9393f1d656a79693e0c123ff7bc7c5c0f708046d"
        ],
        "fixed_in": [
          "effd589568b2c4399894a23000286210c60af6d7",
          "702aa4a44c87026b97b1da5c5ca80e1fa8a9875a",
          "fc71c6c3d3e3bb17474dcd1162aea6783a452a64",
          "c4d5953582463f45e48b14ef9815e5fb636f7e1f",
          "9379f8527ca60d92715c90b32f1b477de9ec32cb",
          "0c3f4544ff3873594c8af1b907b4ac4e6d5ce005",
          "c4f196bfeedd71e56aba4b63bd8f919edb2f7056",
          "9393f1d656a79693e0c123ff7bc7c5c0f708046d"
        ]
      },
      {
        "vendor": "Linux",
        "product": "Linux",
        "ranges": [
          "2.6.27"
        ],
        "fixed_in": [
          "0",
          "5.10.270",
          "5.15.221",
          "6.1.188",
          "6.6.157",
          "6.12.109",
          "6.18.50",
          "7.2.4",
          "7.3"
        ]
      }
    ],
    "upgrade": [
      "Upgrade Linux to effd589568b2c4399894a23000286210c60af6d7 or 702aa4a44c87026b97b1da5c5ca80e1fa8a9875a or fc71c6c3d3e3bb17474dcd1162aea6783a452a64 or c4d5953582463f45e48b14ef9815e5fb636f7e1f or 9379f8527ca60d92715c90b32f1b477de9ec32cb or 0c3f4544ff3873594c8af1b907b4ac4e6d5ce005 or c4f196bfeedd71e56aba4b63bd8f919edb2f7056 or 9393f1d656a79693e0c123ff7bc7c5c0f708046d or later.",
      "Upgrade Linux to 0 or 5.10.270 or 5.15.221 or 6.1.188 or 6.6.157 or 6.12.109 or 6.18.50 or 7.2.4 or 7.3 or later."
    ]
  },
  "exploit_refs": [],
  "hunt_profile": [],
  "rules": []
}
