{"id":"CVE-2026-99164","title":"spi: imx: fix slab corruption from undersized DMA bounce buffers","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\nspi: imx: fix slab corruption from undersized DMA bounce buffers\n\nSince commit faa8e404ad8e (\"spi: imx: support dynamic burst length for\nECSPI DMA mode\"), the driver al…","severity":"none","vendor":"Linux","product":"Linux","affected":["Linux >= faa8e404ad8e686cb98c51dc507fdcacfb8020ce < 2777f6637da02da84c9c1ff92f413ce0f0995084","Linux >= faa8e404ad8e686cb98c51dc507fdcacfb8020ce < 59eb22641f0c526dd064390b11cff99ae87c2858","Linux 7.0"],"published":"2026-10-28","updated":"2026-10-28","sourceUpdated":"2026-10-28T10:35:16.694Z","source":"CVEORG","sourceUrl":"https://www.cve.org/CVERecord?id=CVE-2026-99164","references":[{"url":"https://git.kernel.org/stable/c/2777f6637da02da84c9c1ff92f413ce0f0995084"},{"url":"https://git.kernel.org/stable/c/59eb22641f0c526dd064390b11cff99ae87c2858"}],"tags":["cve.org"],"ingestedAt":"2026-10-28T11:09:40.393Z","slug":"CVE-2026-99164","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\nspi: imx: fix slab corruption from undersized DMA bounce buffers\n\nSince commit faa8e404ad8e (\"spi: imx: support dynamic burst length for\nECSPI DMA mode\"), the driver allocates its private DMA TX/RX buffers\nwith kzalloc(dma_len), where dma_len is only rounded up to 4 bytes, but\nmaps, syncs and unmaps them with DMA_CACHE_ALIGNED_LEN(dma_len), i.e.\nrounded up to the cache line size. As a result, the DMA API operates on\nmemory beyond the end of the allocation.\n\nThis was observed on a custom i.MX8MP platform with an mcp251xfd CAN\ncontroller under heavy RPMSG load, as a series of refcount_t warnings\ntriggered from rpmsg_recv_done() in virtio_rpmsg_bus:\n\n  refcount_t: addition on 0; use-after-free.\n  refcount_t: underflow; use-after-free.\n  refcount_t: saturated; leaking memory.\n\nLeading finally to an \"Unable to handle kernel paging request\" oops.\nReproducing this problem took between 30 minutes and roughly 30 hours\nunder the same workload.\n\nHardware watchpoints on the refcount of a struct rpmsg_endpoint showed\nit was overwritten from 2 to 1 by __pi_memcpy_generic() in the\nirq/155-spi2.0 thread. Additionally it was found that a \"normal\"\nrpmsg_recv_done() took 11µs +/-2µs from its kref_get to kref_put. In\nthe error case the same sequence took about 215µs +/-10µs. Based on\nthis, an LLM was used to identify the size mismatch fixed here.\n\nAllocate the buffers with DMA_CACHE_ALIGNED_LEN(dma_len) so they cover\nthe full mapped length, and use the same length for\ndma_sync_single_for_cpu() to stay consistent with map and unmap.\n\nWith this patch applied, the same workload ran for more than 200 hours\nwithout the problem reproducing.\n\n## Affected\n\n- `Linux >= faa8e404ad8e686cb98c51dc507fdcacfb8020ce < 2777f6637da02da84c9c1ff92f413ce0f0995084`\n- `Linux >= faa8e404ad8e686cb98c51dc507fdcacfb8020ce < 59eb22641f0c526dd064390b11cff99ae87c2858`\n- `Linux 7.0`\n\n## Remediation\n\nRefer to the linked advisories for vendor-supplied fixes and affected version ranges.","depth":"sunlit","depthScore":3,"depthScoreParts":{"impact":2.8,"likelihood":0,"exploitation":0,"ransomware":0},"changes":[]}