Re: [PATCH 1/2] io_uring: split alloc and add of overflow

[Jens Axboe <axboe@kernel.dk>]

On 5/16/25 10:31 AM, Caleb Sander Mateos wrote:
>> @@ -1442,10 +1462,11 @@ void __io_submit_flush_completions(struct io_ring_ctx *ctx)
>>                     unlikely(!io_fill_cqe_req(ctx, req))) {
>>                         if (ctx->lockless_cq) {
>>                                 spin_lock(&ctx->completion_lock);
>> -                               io_req_cqe_overflow(req);
>> +                               io_req_cqe_overflow(req, GFP_ATOMIC);
>>                                 spin_unlock(&ctx->completion_lock);
>>                         } else {
>> -                               io_req_cqe_overflow(req);
>> +                               gfp_t gfp = ctx->lockless_cq ? GFP_KERNEL : GFP_ATOMIC;
> 
> This is in the else case of an if (ctx->lockless_cq). Isn't
> ctx->lockless_cq known to be false?

Indeed! Actually this part needs to be split too. I'll redo this one.

-- 
Jens Axboe