Re: [RFC 12/18] limits: track RLIMIT_MEMLOCK actual max

On 06/13/16 20:43, Alex Williamson wrote:

On Mon, 13 Jun 2016 22:44:19 +0300
Topi Miettinen [off-list ref] wrote:

quoted

Track maximum size of locked memory, presented in /proc/self/limits.

Signed-off-by: Topi Miettinen <redacted>
---
 arch/ia64/kernel/perfmon.c                 |  1 +
 arch/powerpc/kvm/book3s_64_vio.c           |  1 +
 arch/powerpc/mm/mmu_context_iommu.c        |  1 +
 drivers/infiniband/core/umem.c             |  1 +
 drivers/infiniband/hw/hfi1/user_pages.c    |  1 +
 drivers/infiniband/hw/qib/qib_user_pages.c |  1 +
 drivers/infiniband/hw/usnic/usnic_uiom.c   |  2 ++
 drivers/misc/mic/scif/scif_rma.c           |  1 +
 drivers/vfio/vfio_iommu_spapr_tce.c        |  2 ++
 drivers/vfio/vfio_iommu_type1.c            |  2 ++
 include/linux/sched.h                      | 10 ++++++++--
 kernel/bpf/syscall.c                       |  6 ++++++
 kernel/events/core.c                       |  1 +
 mm/mlock.c                                 |  7 +++++++
 mm/mmap.c                                  |  3 +++
 mm/mremap.c                                |  3 +++
 16 files changed, 41 insertions(+), 2 deletions(-)

...  

quoted

diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index 2ba1942..4c6e7a3 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c

@@ -312,6 +312,8 @@ static long vfio_pin_pages(unsigned long vaddr, long npage,
 		}
 	}
 
+	bump_rlimit(RLIMIT_MEMLOCK, (current->mm->locked_vm + i) << PAGE_SHIFT);
+
 	if (!rsvd)
 		vfio_lock_acct(i);
 

Not all cases passing through here bump rlimit (see: rsvd), there's an
entire case above the other end of this closing bracket that does bump
rlimit but returns before here, and I wonder why we wouldn't just do
this in our vfio_lock_acct() accounting function anyway.  Thanks,

Yes, just make disable_hugepages case go to end of function.

-Topi

Alex