{"id":"CVE-2026-93168","title":"In the Linux kernel, the following vulnerability has been resolved:\n\ndmaengine: xilinx_dma: Fix CPU stall in xilinx_dma_poll_timeout\n\nCurrently when calling xilinx_dma_poll_timeout with delay_us=0 and a\ncondition that is never fulfilled,…","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\ndmaengine: xilinx_dma: Fix CPU stall in xilinx_dma_poll_timeout\n\nCurrently when calling xilinx_dma_poll_timeout with delay_us=0 and a\ncondition that is never fulfilled,…","severity":"none","vendor":"Linux","product":"Linux","affected":["Linux >= 9495f2648287029fb5545c34a0fa318426ebe84c < 8b5654d317277e6e505c23f8fa7e415237fefc7f","Linux >= 9495f2648287029fb5545c34a0fa318426ebe84c < aa99c4d1d63bbc26a5fc4c667d89b2595743c19d","Linux 4.6"],"published":"2026-09-17","updated":"2026-09-17","sourceUpdated":"2026-09-17T17:18:12.403","source":"NVD","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-93168","references":[{"url":"https://git.kernel.org/stable/c/8b5654d317277e6e505c23f8fa7e415237fefc7f","label":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"},{"url":"https://git.kernel.org/stable/c/aa99c4d1d63bbc26a5fc4c667d89b2595743c19d","label":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"}],"tags":["nvd","cve.org"],"ingestedAt":"2026-09-17T16:21:47.732Z","epss":0.00155,"epssPercentile":0.05119,"slug":"CVE-2026-93168","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\ndmaengine: xilinx_dma: Fix CPU stall in xilinx_dma_poll_timeout\n\nCurrently when calling xilinx_dma_poll_timeout with delay_us=0 and a\ncondition that is never fulfilled, the CPU busy-waits for prolonged time\nand the timeout triggers only with a massive delay causing a CPU stall.\n\nThis happens due to a huge underestimation of wall clock time in\npoll_timeout_us_atomic. Commit 7349a69cf312 (\"iopoll: Do not use\ntimekeeping in read_poll_timeout_atomic()\") changed the behavior to no\nlonger use ktime_get at the expense of underestimation of wall clock\ntime which appears to be very large for delay_us=0. Instead of timing\nout after approximately XILINX_DMA_LOOP_COUNT microseconds, the timeout\ntakes XILINX_DMA_LOOP_COUNT * 1000 * (time that the overhead of the for\nloop in poll_timeout_us_atomic takes) which is in the range of several\nminutes for XILINX_DMA_LOOP_COUNT=1000000. Fix this by using a non-zero\nvalue for delay_us. Use delay_us=10 to keep the delay in the hot path of\nstarting DMA transfers minimal but still avoid CPU stalls in case of\nunexpected hardware failures.\n\nOne-off measurement with delay_us=0 causes the cpu to busy wait around 7\nminutes in the timeout case. After applying this patch with delay_us=10\nthe measured timeout was 1053428 microseconds which is roughly\nequivalent to the expected 1000000 microseconds specified in\nXILINX_DMA_LOOP_COUNT.\n\nAdd a constant XILINX_DMA_POLL_DELAY_US for delay_us value.\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":[]}