On Mon, 22 Jun 2026 at 12:19, Bruce Richardson <[email protected]> wrote: > Running an AI correctness review reports a potential issue with ifpga bus > after this patch. Maybe worth a double check. > > /Bruce > > In ifpga_alloc_afu_dev (called at scan time, before any probe), > afu_dev->intr_handle is allocated unconditionally. > > In the old ifpga_cleanup, the "goto free:" label ran for all devices, so > intr_handle was freed unconditionally. In the new ifpga_cleanup, > intr_handle is only freed via ifpga_unplug_device, which is only called > when rte_dev_is_probed() returns true. For any ifpga device discovered > during scan but never successfully probed (no matching driver, failed > probe, blocked devargs), intr_handle is leaked. The fix should free it > unconditionally in ifpga_cleanup, outside the rte_dev_is_probed block.
My AI friend did not see it and it gets resolved when using the generic helper. But yes good catch. -- David Marchand
