{"id":"CVE-2026-98592","published":"2026-10-18T16:17:11.830","lastModified":"2026-10-19T10:19:20.360","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nnet: spacemit: clear TX descriptor on fragment mapping failure\n\nemac_tx_mem_map() writes TX_DESC_0_OWN into the ring descriptor for\nevery slot beyond old_head as soon as that slot's memset()'d local\ncopy is committed with \"*tx_desc_addr = tx_desc\", i.e. before the\nbuffers for that slot have necessarily all been mapped successfully.\nIf emac_tx_map_frag() then fails on a later fragment, the err_free_skb\npath calls emac_free_tx_buf() to unmap and drop the skb, but leaves\nthe already-written descriptor memory untouched, and tx_ring->head is\nnever advanced past old_head (the \"tx_ring->head = head\" store is\nskipped by the goto).\n\nSo a slot between old_head and the rolled-back head can be left with\nTX_DESC_0_OWN set and buffer_addr_{1,2} pointing at DMA mappings that\nemac_free_tx_buf() just tore down, while software considers that slot\nfree again. The next successful emac_tx_mem_map() call only rebuilds\nold_head itself; if the DMA engine auto-advances into the following\ndescriptor once it finishes old_head's packet, it will fetch that\nstale, already-unmapped address.\n\nemac_tx_clean_desc() already treats emac_free_tx_buf() and clearing\nthe descriptor as a pair when reclaiming completed descriptors; do\nthe same in the mapping failure path.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/2d14720beb58870b15a52b236c3ab0be0e06e915","tags":[]},{"url":"https://git.kernel.org/stable/c/3001a103a367c6051c43aff363b6979c10e65347","tags":[]},{"url":"https://git.kernel.org/stable/c/ca9986cc824f2feb79b4a83e5f6f4c24c2140448","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}