The XSL fault interrupt takes an mm reference and queues fault_work embedded in the SPA. free_irq() stops new interrupt handlers but does not drain an already queued worker, allowing free_spa() to release its container first. Cancel the work after freeing the interrupt. When cancellation removes a pending instance, also drop the mm reference that the worker would have released. Changes in v2: cancel the work after free_irq() and drop the mm reference when a pending instance was removed, and refresh the comment that explains the ordering. Found by the author's in-house static analysis tool. Fixes: 5ef3166e8a32 ("ocxl: Driver code for 'generic' opencapi devices") Cc: stable@vger.kernel.org Link: https://lore.kernel.org/all/20260928134110.227-1-getshell@seu.edu.cn/ Signed-off-by: Hongyan Xu Signed-off-by: Zehao Gao --- Discovery: found by the author's in-house static analysis tool. drivers/misc/ocxl/link.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/drivers/misc/ocxl/link.c b/drivers/misc/ocxl/link.c index 7749dcd..a0220af 100644 --- a/drivers/misc/ocxl/link.c +++ b/drivers/misc/ocxl/link.c @@ -252,7 +252,11 @@ static irqreturn_t xsl_fault_handler(int irq, void *data) spa->xsl_fault.dsisr = dsisr; spa->xsl_fault.pe_data = *pe_data; schedule = true; - /* mm_users count released by bottom half */ + /* + * The bottom half drops the mm_users reference, or + * release_xsl_irq() does when the work is still pending + * and gets canceled during teardown. + */ } rcu_read_unlock(); if (schedule) @@ -335,6 +339,8 @@ static void release_xsl_irq(struct ocxl_link *link) if (spa->virq) { free_irq(spa->virq, link); + if (cancel_work_sync(&spa->xsl_fault.fault_work)) + mmput(spa->xsl_fault.pe_data.mm); irq_dispose_mapping(spa->virq); } kfree(spa->irq_name); -- 2.50.1.windows.1