> Also, the fact that the patch series involves multiple file system is > a massive pain. It means I'm going to have to do a separate > regression test --- or preferably, I would ask *you* to run a file > system regression test[1] --- to make sure what is *not* a trivial > patch doesn't break things. Also, it means that this patch series is > going to get more complicated to get into kernel, and we may have to > deal with patch conflicts if this goes in via some third party tree > (such as Andrew's tree). > > [1] https:/thunk.org/gce-xfstests Sure, I will run the regression. > > One way to make life easier is to add the new function with the new > interface first, and then wait a release cycle, and then move file > systems over in independant patches. In last review, you left it to me either to add new function or modify the input parameters of block_page_mkwrite() to return err to caller. As block_page_mkwrite() is getting called from 2 places in ext4 & nilfs, I choose to add new input argument in block_page_mkwrite() rather than introducing new function and put everything in a single commit.
