On Sat, Jun 06, 2020 at 02:30:37AM +0800, Gao Xiang wrote: ... > > > > Alternately-- if you want to sanity check that b_pag and the buffer > > belong to the same ag, why not do that in xfs_buf_find for all the > > buffers? > > Since that modification doesn't relate to this patch though (since > the purpose of this patch is not add ASSERT to xfs_buf_find). > > If in that way, I think we can just kill all these ASSERTs. Add some to the previous words: What I really concern is to avoid introduce some regression from this trivial patch. That is the original reason why I added these assertions and tested with stress. If some of these seems useful to the codebase, I could leave them in this patch (since these are related to the modification of this patch). Otherwise, I'd suggest kill them all though as I mentioned earilier in the reply of v1. Thanks, Gao Xiang
