On Mon, Oct 14, 2019 at 09:22:24AM +0200, Christoph Hellwig wrote: > On Fri, Oct 11, 2019 at 05:41:45PM -0700, Darrick J. Wong wrote: > > > -#else > > > -#define xlog_state_callback_check_state(l) ((void)0) > > > -#endif > > > > So, uh... does this debugging functionality just disappear? Is it > > unnecessary? It's not clear (to me anyway) why it's ok for the extra > > checking to go away. > > Lets ask the counter question: What kind of bug do you think this > check would catch? Dunno, that's why I'm asking /you/ :) I think the answer is that DO_CALLBACK is pretty pointless since we already have a CALLBACK state, and the removed debugging function checks among all the iclogs for a state that shouldn't be possible (getting ready to be doing callbacks when there are other iclogs further along in the list that are still syncing or ioerror'd)? Particularly since we probably end up moving the iclog to the actual CALLBACK state pretty quickly anyway? --D
