On Wed 24-04-19 11:13:48, Dan Williams wrote: > On Wed, Apr 24, 2019 at 10:38 AM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote: > > > > On Wed, Apr 24, 2019 at 10:13:15AM -0700, Dan Williams wrote: > > > I think unaligned addresses have always been passed to > > > vmf_insert_pfn_pmd(), but nothing cared until this patch. I *think* > > > the only change needed is the following, thoughts? > > > > > > diff --git a/fs/dax.c b/fs/dax.c > > > index ca0671d55aa6..82aee9a87efa 100644 > > > --- a/fs/dax.c > > > +++ b/fs/dax.c > > > @@ -1560,7 +1560,7 @@ static vm_fault_t dax_iomap_pmd_fault(struct > > > vm_fault *vmf, pfn_t *pfnp, > > > } > > > > > > trace_dax_pmd_insert_mapping(inode, vmf, PMD_SIZE, pfn, entry); > > > - result = vmf_insert_pfn_pmd(vma, vmf->address, vmf->pmd, pfn, > > > + result = vmf_insert_pfn_pmd(vma, pmd_addr, vmf->pmd, pfn, > > > write); > > > > We also call vmf_insert_pfn_pmd() in dax_insert_pfn_mkwrite() -- does > > that need to change too? > > It wasn't clear to me that it was a problem. I think that one already > happens to be pmd-aligned. Why would it need to be? The address is taken from vmf->address and that's set up in __handle_mm_fault() like .address = address & PAGE_MASK. So I don't see anything forcing PMD alignment of the virtual address... Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
