On Tue, Apr 25, 2023 at 01:11:53PM +0300, Kirill A . Shutemov wrote: > On Tue, Apr 25, 2023 at 08:14:14AM +0100, Lorenzo Stoakes wrote: > > GUP does not correctly implement write-notify semantics, nor does it > > guarantee that the underlying pages are correctly dirtied, which could lead > > to a kernel oops or data corruption when writing to file-backed mappings. > > > > This is only relevant when the mappings are file-backed and the underlying > > file system requires folio dirty tracking. File systems which do not, such > > as shmem or hugetlb, are not at risk and therefore can be written to > > without issue. > > > > Unfortunately this limitation of GUP has been present for some time and > > requires future rework of the GUP API in order to provide correct write > > access to such mappings. > > > > In the meantime, we add a check for the most broken GUP case - > > FOLL_LONGTERM - which really under no circumstances can safely access > > dirty-tracked file mappings. > > > > Suggested-by: Jason Gunthorpe <jgg@xxxxxxxxxx> > > Signed-off-by: Lorenzo Stoakes <lstoakes@xxxxxxxxx> > > --- > > v3: > > - Rebased on latest mm-unstable as of 24th April 2023. > > - Explicitly check whether file system requires folio dirtying. Note that > > vma_wants_writenotify() could not be used directly as it is very much focused > > on determining if the PTE r/w should be set (e.g. assuming private mapping > > does not require it as already set, soft dirty considerations). > > Hm. Okay. Have you considered having a common base for your case and > vma_wants_writenotify()? Code duplication doesn't look good. > I did and I actually started implementing something for the same reason, however I wondered whether it was worth it for essentially 3 clauses that are shared between the two. On second thoughts, it is painful to have this duplicated, so let me take another look. > > -- > Kiryl Shutsemau / Kirill A. Shutemov
