| Message ID | 20220701142310.2188015-20-glider@google.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Add KernelMemorySanitizer infrastructure | expand |
On Fri, 1 Jul 2022 at 16:24, Alexander Potapenko <glider@google.com> wrote: > > This is a hack to reduce stackdepot pressure. Will it cause false negatives or other issues? If not, I'd just call it an optimization and not a hack. > struct mmu_gather contains 7 1-bit fields packed into a 32-bit unsigned > int value. The remaining 25 bits remain uninitialized and are never used, > but KMSAN updates the origin for them in zap_pXX_range() in mm/memory.c, > thus creating very long origin chains. This is technically correct, but > consumes too much memory. > > Unpoisoning the whole structure will prevent creating such chains. > > Signed-off-by: Alexander Potapenko <glider@google.com> Acked-by: Marco Elver <elver@google.com> > --- > Link: https://linux-review.googlesource.com/id/I76abee411b8323acfdbc29bc3a60dca8cff2de77 > --- > mm/mmu_gather.c | 10 ++++++++++ > 1 file changed, 10 insertions(+) > > diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c > index a71924bd38c0d..add4244e5790d 100644 > --- a/mm/mmu_gather.c > +++ b/mm/mmu_gather.c > @@ -1,6 +1,7 @@ > #include <linux/gfp.h> > #include <linux/highmem.h> > #include <linux/kernel.h> > +#include <linux/kmsan-checks.h> > #include <linux/mmdebug.h> > #include <linux/mm_types.h> > #include <linux/mm_inline.h> > @@ -265,6 +266,15 @@ void tlb_flush_mmu(struct mmu_gather *tlb) > static void __tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, > bool fullmm) > { > + /* > + * struct mmu_gather contains 7 1-bit fields packed into a 32-bit > + * unsigned int value. The remaining 25 bits remain uninitialized > + * and are never used, but KMSAN updates the origin for them in > + * zap_pXX_range() in mm/memory.c, thus creating very long origin > + * chains. This is technically correct, but consumes too much memory. > + * Unpoisoning the whole structure will prevent creating such chains. > + */ > + kmsan_unpoison_memory(tlb, sizeof(*tlb)); > tlb->mm = mm; > tlb->fullmm = fullmm; > > -- > 2.37.0.rc0.161.g10f37bed90-goog >
diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c index a71924bd38c0d..add4244e5790d 100644 --- a/mm/mmu_gather.c +++ b/mm/mmu_gather.c @@ -1,6 +1,7 @@ #include <linux/gfp.h> #include <linux/highmem.h> #include <linux/kernel.h> +#include <linux/kmsan-checks.h> #include <linux/mmdebug.h> #include <linux/mm_types.h> #include <linux/mm_inline.h> @@ -265,6 +266,15 @@ void tlb_flush_mmu(struct mmu_gather *tlb) static void __tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm) { + /* + * struct mmu_gather contains 7 1-bit fields packed into a 32-bit + * unsigned int value. The remaining 25 bits remain uninitialized + * and are never used, but KMSAN updates the origin for them in + * zap_pXX_range() in mm/memory.c, thus creating very long origin + * chains. This is technically correct, but consumes too much memory. + * Unpoisoning the whole structure will prevent creating such chains. + */ + kmsan_unpoison_memory(tlb, sizeof(*tlb)); tlb->mm = mm; tlb->fullmm = fullmm;
This is a hack to reduce stackdepot pressure. struct mmu_gather contains 7 1-bit fields packed into a 32-bit unsigned int value. The remaining 25 bits remain uninitialized and are never used, but KMSAN updates the origin for them in zap_pXX_range() in mm/memory.c, thus creating very long origin chains. This is technically correct, but consumes too much memory. Unpoisoning the whole structure will prevent creating such chains. Signed-off-by: Alexander Potapenko <glider@google.com> --- Link: https://linux-review.googlesource.com/id/I76abee411b8323acfdbc29bc3a60dca8cff2de77 --- mm/mmu_gather.c | 10 ++++++++++ 1 file changed, 10 insertions(+)