On Thu, Oct 15, 2020 at 10:06:51AM +0100, Christoph Hellwig wrote: > Don't we also need to handle the new return value in a few other places > like cachefiles_read_reissue swap_readpage? Maybe those don't get > called on the currently converted instances, but just leaving them > without handling AOP_UPDATED_PAGE seems like a time bomb. Er, right. And nobh_truncate_page(), and read_page(). And then I noticed the bug in cachefiles_read_reissue(). Sigh. Updated patch series coming soon.
