{
  "cveId": "CVE-2026-99127",
  "url": "https://patchblox.com/security/cve/CVE-2026-99127",
  "summary": "CVE-2026-99127 is fixed in linux on 1 platform, first by DEBIAN-CVE-2026-99127, released November 3, 2026. It is not reported as exploited. Patch Radar priority: Routine.",
  "title": "page_pool: keep frag_offset aligned for odd-sized requests",
  "description": "In the Linux kernel, the following vulnerability has been resolved:  page_pool: keep frag_offset aligned for odd-sized requests  page_pool_alloc_frag_netmem() rounds the requested fragment size with  \tsize = ALIGN(size, dma_get_cache_alignment());  dma_get_cache_alignment() returns 1 unless the architecture defines ARCH_DMA_MINALIGN, which DMA-coherent architectures such as x86 do not. There the ALIGN() is a no-op and pool-\u003Efrag_offset advances by the raw, unrounded size.  A single caller asking for an odd size then leaves frag_offset misaligned for every fragment carved out of that page afterwards.  The pool is shared, so the damage is not confined to the caller that caused it.  The per-cpu system_page_pool used by generic XDP hits this. skb_pp_cow_data() allocates its fragments with the raw packet length:  \tsize = min_t(u32, len, PAGE_SIZE); \ttruesize = size; \tpage = page_pool_dev_alloc(pool, \u0026page_off, \u0026truesize);  leaving frag_offset odd for whatever is carved out of that page next.  Its own head allocation is already aligned -- SKB_HEAD_ALIGN(size) plus the XDP_PACKET_HEADROOM its callers pass -- so it is a later user of the shared pool that pays: page_pool_dev_alloc_va() returns a misaligned buffer, napi_build_skb() installs it as skb-\u003Ehead, and skb_shinfo(skb) == skb-\u003Ehead \u002B skb-\u003Eend is misaligned with it.  skb_shinfo()-\u003Edataref is a 4-byte atomic_t at offset 0x20, so the atomic_inc() in __skb_clone() straddles a cache line.  On x86 with split lock detection -- fatal for kernel split locks by default -- this panics the machine:    Oops: Split lock detected   RIP: 0010:skb_clone\u002B0x154/0x1e0   Call Trace:    \u003CIRQ\u003E    raw_local_deliver\u002B0x1ed/0x2c0    ip_protocol_deliver_rcu\u002B0x54/0x1c0    ip_local_deliver_finish\u002B0x85/0x100    ip_local_deliver\u002B0x67/0x100    __netif_receive_skb_one_core\u002B0x85/0xa0    process_backlog\u002B0x87/0x130  Reproduced by attaching any generic-mode XDP program to loopback and opening a RAW IPPROTO_UDP socket, which makes raw_local_deliver() clone every locally delivered UDP packet; ordinary DNS traffic then triggers it, roughly once per 2500 clones.  Observed on 6.12.101 and 7.1.8.  Tracing page_pool_alloc_frag_netmem() over one such run shows the amplification -- two odd-sized requests, nine misaligned offsets:    requested size \u0026 7:   0: 17035    5: 1    7: 1   frag_offset \u0026 7:      0: 17028    3: 1    4: 1    5: 1    6: 1    7: 5  and skb_pp_cow_data() returning heads that were aligned on entry:    head 0xffff8f4c86aeac00 -\u003E 0xffff8f4c53a9a9c4 (\u00267=4)   head 0xffff8f4d6a8a42c0 -\u003E 0xffff8f4c4f7b7a45 (\u00267=5)  Round the fragment size up to at least the alignment struct skb_shared_info requires, so fragments are always suitably aligned for the objects callers build on them.  Architectures needing a larger DMA alignment keep it.  This also makes the remainder computed in page_pool_alloc_netmem(),  \t*size = max_size - *offset;  aligned, since max_size is a power of two -- which fixes the matching misalignment of skb-\u003Eend.  Verified with a controlled A/B under QEMU/KVM: same tree, same config, same compiler, same rootfs and identical traffic, differing only by this patch.  A SEC(\u0022xdp.frags\u0022) XDP_PASS program on lo plus UDP datagrams larger than max_head_size drives skb_pp_cow_data()\u0027s fragment loop, which passes raw packet lengths to the pool.  Measured at the return of skb_pp_cow_data():                            unpatched   patched   skb_pp_cow_data calls       40800     40800   misaligned skb-\u003Ehead         1120         0   dataref at line offset \u003E60     80         0  The last row counts the accesses that actually fault: skb_shinfo()-\u003Edataref sits at head\u002Bend\u002B0x20 and is a 4-byte atomic, so \u0060lock incl\u0060 splits a 64-byte cache line only when that address lands at offset 61..63.  All 80 occurrences were at offset 61; the panic reported above was at offset 62.  Eliminating the misalignment removes every one of them.  Same class of bug as commit 3bed3cc4156e (\u0022net: Do not allocate page fragments t ---truncated---",
  "published": "2026-10-28T11:17:32Z",
  "severity": "Moderate",
  "cvss": 5.5,
  "epss": 0.00173,
  "epssPercentile": 0.06786,
  "priority": "Routine",
  "priorityTier": 4,
  "exploited": false,
  "vendorExploited": false,
  "disclosed": false,
  "vendorAdvisory": "https://msrc.microsoft.com/update-guide/vulnerability/CVE-2026-99127",
  "fixes": [
    {
      "platform": "linux",
      "product": "linux",
      "release": "Debian 13.5 (trixie)",
      "fixedVersion": "6.12.115-2",
      "fix": "DEBIAN-CVE-2026-99127",
      "url": "https://patchblox.com/security/fix/linux/debian-cve-2026-99127",
      "released": "2026-11-03T18:58:21Z"
    },
    {
      "platform": "linux",
      "product": "linux",
      "release": "Debian 14",
      "fixedVersion": "7.2.13-1",
      "fix": "DEBIAN-CVE-2026-99127",
      "url": "https://patchblox.com/security/fix/linux/debian-cve-2026-99127",
      "released": "2026-11-03T18:58:21Z"
    }
  ],
  "noFixYet": [],
  "dataAsOf": "2026-11-08T19:12:23Z",
  "attribution": {
    "text": "Data: Patchblox Patch Radar",
    "url": "https://patchblox.com/security/patch-radar",
    "terms": "https://patchblox.com/security/patch-radar/use-the-data"
  }
}