{"id":"CVE-2026-98803","published":"2026-10-18T16:17:34.350","lastModified":"2026-10-19T10:19:44.350","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nnet: stmmac: fix rx Scatter-Gather support\n\nWhen a received frame is larger than dma_buf_sz, the DMA scatters it\nacross multiple RX descriptors (rx Scatter-Gather). The secondary RX\nbuffer (sec_page) was only allocated and programmed when split-header\n(SPH) was active, so for regular frames buffer2 was neither allocated\nnor backed by a valid mapping. As soon as an incoming frame overflowed\nbuffer1, the DMA wrote the overflow into the unmapped secondary-buffer\naddress, triggering an SMMU translation fault on IOMMU-based platforms:\n\narm-smmu 15000000.iommu: Unhandled context fault: fsr=0x402, iova=0x00000000, fsynr=0x7f0011, cbfrsynra=0x1c90, cb=11\narm-smmu 15000000.iommu: FSR    = 00000402 [Format=2 TF], SID=0x1c90\narm-smmu 15000000.iommu: FSYNR0 = 007f0011 [S1CBNDX=127 WNR PLVL=1]\n\nEnable scatter-gather for non-SPH frames on cores that can program an\nindependent secondary RX buffer (GMAC4/XGMAC): allocate and mark buffer2\nas valid in stmmac_init_rx_buffers() and stmmac_rx_refill(), and account\nfor it in the buffer length computation. Legacy cores have no set_sec_addr\nop, so they keep buffer2 disabled.\n\nSince buffer2 is handed to the DMA at page offset 0, the page pool sync\nwindow is widened to cover both buffers in every mode: offset is set to 0\nand max_len to dma_buf_sz + stmmac_rx_offset().\n\nThe FCS can straddle the buffer1/buffer2 boundary and a descriptor\nboundary, so it is now stripped from the tail of the assembled frame with\npskb_trim() instead of from a single buffer, avoiding an unsigned underflow\nfor frames that overflow a buffer by 1..3 bytes. For single-buffer frames\nthe XDP program must not see the FCS, so it is removed from the XDP buffer\nbefore the program runs.\n\nNative XDP currently only supports single-buffer (linear) frames. An\noversized frame accepted by the MAC (jumbo enabled) is received via\nbuffer2; the XDP program only sees buffer1, so on a TX/REDIRECT verdict\nthe frame is forwarded truncated. XDP multi-buffer support to handle\nthis case is planned as a follow-up.\n\nAF_XDP zero-copy RX is not covered by this change: a ZC queue still\nprograms buffer2 at DMA address 0 (and XGMAC has no buffer2-valid bit),\nso an oversized frame overflowing buffer1 can still trigger the same\nSMMU translation fault. Handling is planned as a follow-up.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/358d5ac5a5d57ce21c3814c4e3e8f3fbcb7304b4","tags":[]},{"url":"https://git.kernel.org/stable/c/88095728511e0df9c89528944f5a1762ad298983","tags":[]},{"url":"https://git.kernel.org/stable/c/a7cff44543f66f3a19151060cb549691d191d44c","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}