CVE-2025-38502

{"id": "CVE-2025-38502", "cveTags": [], "metrics": {}, "published": "2025-08-16T10:15:25.653", "references": [{"url": "https://git.kernel.org/stable/c/19341d5c59e8c7e8528e40f8663e99d67810473c", "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}, {"url": "https://git.kernel.org/stable/c/abad3d0bad72a52137e0c350c59542d75ae4f513", "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}], "vulnStatus": "Awaiting Analysis", "descriptions": [{"lang": "en", "value": "In the Linux kernel, the following vulnerability has been resolved:\n\nbpf: Fix oob access in cgroup local storage\n\nLonial reported that an out-of-bounds access in cgroup local storage\ncan be crafted via tail calls. Given two programs each utilizing a\ncgroup local storage with a different value size, and one program\ndoing a tail call into the other. The verifier will validate each of\nthe indivial programs just fine. However, in the runtime context\nthe bpf_cg_run_ctx holds an bpf_prog_array_item which contains the\nBPF program as well as any cgroup local storage flavor the program\nuses. Helpers such as bpf_get_local_storage() pick this up from the\nruntime context:\n\n ctx = container_of(current->bpf_ctx, struct bpf_cg_run_ctx, run_ctx);\n storage = ctx->prog_item->cgroup_storage[stype];\n\n if (stype == BPF_CGROUP_STORAGE_SHARED)\n ptr = &READ_ONCE(storage->buf)->data[0];\n else\n ptr = this_cpu_ptr(storage->percpu_buf);\n\nFor the second program which was called from the originally attached\none, this means bpf_get_local_storage() will pick up the former\nprogram's map, not its own. With mismatching sizes, this can result\nin an unintended out-of-bounds access.\n\nTo fix this issue, we need to extend bpf_map_owner with an array of\nstorage_cookie[] to match on i) the exact maps from the original\nprogram if the second program was using bpf_get_local_storage(), or\nii) allow the tail call combination if the second program was not\nusing any of the cgroup local storage maps."}, {"lang": "es", "value": "En el kernel de Linux, se ha resuelto la siguiente vulnerabilidad: bpf: Correcci\u00f3n del acceso fuera de los l\u00edmites en el almacenamiento local de cgroup Lonial inform\u00f3 que se puede manipular un acceso fuera de los l\u00edmites en el almacenamiento local de cgroup mediante llamadas de cola. Dados dos programas, cada uno utilizando un almacenamiento local de cgroup con un tama\u00f1o de valor diferente, y un programa realizando una llamada de cola en el otro. El verificador validar\u00e1 cada uno de los programas individuales sin problemas. Sin embargo, en el contexto de tiempo de ejecuci\u00f3n, bpf_cg_run_ctx contiene un bpf_prog_array_item que contiene el programa BPF, as\u00ed como cualquier sabor de almacenamiento local de cgroup que use el programa. Los ayudantes como bpf_get_local_storage() recogen esto del contexto de tiempo de ejecuci\u00f3n: ctx = container_of(current->bpf_ctx, struct bpf_cg_run_ctx, run_ctx); storage = ctx->prog_item->cgroup_storage[stype]; if (stype == BPF_CGROUP_STORAGE_SHARED) ptr = &READ_ONCE(storage->buf)->data[0]; else ptr = this_cpu_ptr(storage->percpu_buf); Para el segundo programa llamado desde el programa adjunto original, esto significa que bpf_get_local_storage() tomar\u00e1 el mapa del programa anterior, no el suyo. Con tama\u00f1os no coincidentes, esto puede resultar en un acceso fuera de los l\u00edmites no deseado. Para solucionar este problema, necesitamos extender bpf_map_owner con una matriz de storage_cookie[] para que coincida con i) los mapas exactos del programa original si el segundo programa usaba bpf_get_local_storage(), o ii) permitir la combinaci\u00f3n de llamadas de cola si el segundo programa no usaba ninguno de los mapas de almacenamiento local de cgroup."}], "lastModified": "2025-08-18T20:16:28.750", "sourceIdentifier": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}