| Message ID | 20240329212444.395559-5-michael.roth@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | gmem fix-ups and interfaces for populating gmem pages | expand |
On 3/29/24 10:24 PM, Michael Roth wrote: > filemap users like guest_memfd may use page cache pages to > allocate/manage memory that is only intended to be accessed by guests > via hardware protections like encryption. Writes to memory of this sort > in common paths like truncation may cause unexpected behavior such > writing garbage instead of zeros when attempting to zero pages, or > worse, triggering hardware protections that are considered fatal as far > as the kernel is concerned. > > Introduce a new address_space flag, AS_INACCESSIBLE, and use this > initially to prevent zero'ing of pages during truncation, with the > understanding that it is up to the owner of the mapping to handle this > specially if needed. > > Link: https://lore.kernel.org/lkml/ZR9LYhpxTaTk6PJX@google.com/ > Cc: Matthew Wilcox <willy@infradead.org> > Suggested-by: Sean Christopherson <seanjc@google.com> > Signed-off-by: Michael Roth <michael.roth@amd.com> Hm somehow it seems like a rather blunt solution to a fairly specific issue on one hand, and on the other hand I'm not sure whether there are other places (not yet triggered) that should now take into account the flag to keep its promise. But as long as it gets the job done, and can be replaced later with something better... Acked-by: Vlastimil Babka <vbabka@suse.cz> > --- > include/linux/pagemap.h | 1 + > mm/truncate.c | 3 ++- > 2 files changed, 3 insertions(+), 1 deletion(-) > > diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h > index e8ac0b32f84d..a7c3f43d1d22 100644 > --- a/include/linux/pagemap.h > +++ b/include/linux/pagemap.h > @@ -207,6 +207,7 @@ enum mapping_flags { > AS_STABLE_WRITES, /* must wait for writeback before modifying > folio contents */ > AS_UNMOVABLE, /* The mapping cannot be moved, ever */ > + AS_INACCESSIBLE, /* Do not attempt direct R/W access to the mapping */ > }; > > /** > diff --git a/mm/truncate.c b/mm/truncate.c > index 725b150e47ac..c501338c7ebd 100644 > --- a/mm/truncate.c > +++ b/mm/truncate.c > @@ -233,7 +233,8 @@ bool truncate_inode_partial_folio(struct folio *folio, loff_t start, loff_t end) > * doing a complex calculation here, and then doing the zeroing > * anyway if the page split fails. > */ > - folio_zero_range(folio, offset, length); > + if (!(folio->mapping->flags & AS_INACCESSIBLE)) > + folio_zero_range(folio, offset, length); > > if (folio_has_private(folio)) > folio_invalidate(folio, offset, length);
diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h index e8ac0b32f84d..a7c3f43d1d22 100644 --- a/include/linux/pagemap.h +++ b/include/linux/pagemap.h @@ -207,6 +207,7 @@ enum mapping_flags { AS_STABLE_WRITES, /* must wait for writeback before modifying folio contents */ AS_UNMOVABLE, /* The mapping cannot be moved, ever */ + AS_INACCESSIBLE, /* Do not attempt direct R/W access to the mapping */ }; /** diff --git a/mm/truncate.c b/mm/truncate.c index 725b150e47ac..c501338c7ebd 100644 --- a/mm/truncate.c +++ b/mm/truncate.c @@ -233,7 +233,8 @@ bool truncate_inode_partial_folio(struct folio *folio, loff_t start, loff_t end) * doing a complex calculation here, and then doing the zeroing * anyway if the page split fails. */ - folio_zero_range(folio, offset, length); + if (!(folio->mapping->flags & AS_INACCESSIBLE)) + folio_zero_range(folio, offset, length); if (folio_has_private(folio)) folio_invalidate(folio, offset, length);
filemap users like guest_memfd may use page cache pages to allocate/manage memory that is only intended to be accessed by guests via hardware protections like encryption. Writes to memory of this sort in common paths like truncation may cause unexpected behavior such writing garbage instead of zeros when attempting to zero pages, or worse, triggering hardware protections that are considered fatal as far as the kernel is concerned. Introduce a new address_space flag, AS_INACCESSIBLE, and use this initially to prevent zero'ing of pages during truncation, with the understanding that it is up to the owner of the mapping to handle this specially if needed. Link: https://lore.kernel.org/lkml/ZR9LYhpxTaTk6PJX@google.com/ Cc: Matthew Wilcox <willy@infradead.org> Suggested-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Michael Roth <michael.roth@amd.com> --- include/linux/pagemap.h | 1 + mm/truncate.c | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-)