On Wed, May 06, 2020 at 11:03:06AM -0700, Yang Shi wrote: > > diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h > > index 55199cb5bd66..1169e2428dd7 100644 > > --- a/include/linux/pagemap.h > > +++ b/include/linux/pagemap.h > > @@ -205,15 +205,33 @@ static inline int page_cache_add_speculative(struct page *page, int count) > > return __page_cache_add_speculative(page, count); > > } > > > > +static inline gfp_t thp_gfpmask(gfp_t gfp) > > +{ > > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE > > + /* We'd rather allocate smaller pages than stall a page fault */ > > + gfp |= GFP_TRANSHUGE_LIGHT; > > This looks not correct. GFP_TRANSHUGE_LIGHT may set GFP_FS, but some > filesystem may expect GFP_NOFS, i.e. in readahead path. Apologies, I overlooked this mail. In one of the prerequisite patches for this patch set (which is now merged as f2c817bed58d9be2051fad1d18e167e173c0c227), we call memalloc_nofs_save() in the readahead path. That ensures all allocations will have GFP_NOFS set by the time the page allocator sees them. Thanks for checking on this.