On 3/28/23 02:46, Zheng Yejian wrote: > Currently blk_register_tracepoints() WARN for every registration failure > of block tracepoint, however, blk_trace can still be setup-ed. > > It seems better to cleanup registered tracepoints and return error when > a certain registration failed, even if it's a small probability, and > users can know the error and do blk_trace setup again. > > Signed-off-by: Zheng Yejian <zhengyejian1@xxxxxxxxxx> > --- > So far I'm fine with current code as this adds a lot of code. I've thought of doing this when I worked on blktrace extension, but others are okay with it sure .. -ck