Re: [RFC][PATCH] mm: Fix RLIMIT_MEMLOCK

[KOSAKI Motohiro]

> diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
> index a841123..f8f47dc 100644
> --- a/drivers/infiniband/core/umem.c
> +++ b/drivers/infiniband/core/umem.c
> @@ -137,17 +137,22 @@ struct ib_umem *ib_umem_get (struct ib_ucontext *context, unsigned long addr,
>
>         down_write(&current->mm->mmap_sem);
>
> -       locked     = npages + current->mm->pinned_vm;
> +       locked     = npages + mm_locked_pages(current->mm);
>         lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
>
>         if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) {
>                 ret = -ENOMEM;
> -               goto out;
> +               goto err;
>         }
>
>         cur_base = addr & PAGE_MASK;
> +       umem->start_addr = cur_base;
> +       umem->end_addr   = cur_base + npages;
> +
> +       ret = mm_mpin(umem->start_addr, umem->end_addr);
> +       if (ret)
> +               goto err;

I believe RLIMIT_MEMLOCK should be checked within mm_mpin().


> +static inline unsigned long mm_locked_pages(struct mm_struct *mm)
> +{
> +       return mm->pinned_vm + mm->locked_vm;
> +}

This is acceptable. but if we create mm_locked_pages(), /proc should
also use this.
Otherwise pinning operation magically decrease VmLck field in
/proc/pid/status and people
get a confusion.



> @@ -310,9 +309,49 @@ static int mlock_fixup(struct vm_area_struct *vma, struct vm_area_struct **prev,
>          * Keep track of amount of locked VM.
>          */
>         nr_pages = (end - start) >> PAGE_SHIFT;
> -       if (!lock)
> -               nr_pages = -nr_pages;
> -       mm->locked_vm += nr_pages;
> +
> +       /*
> +        * We should only account pages once, if VM_PINNED is set pages are
> +        * accounted in mm_struct::pinned_vm, otherwise if VM_LOCKED is set,
> +        * we'll account them in mm_struct::locked_vm.
> +        *
> +        * PL  := vma->vm_flags
> +        * PL' := newflags
> +        * PLd := {pinned,locked}_vm delta
> +        *
> +        * PL->PL' PLd
> +        * -----------
> +        * 00  01  0+
> +        * 00  10  +0
> +        * 01  11  +-
> +        * 01  00  0-
> +        * 10  00  -0
> +        * 10  11  00
> +        * 11  01  -+
> +        * 11  10  00
> +        */

This comment is too cryptic. :-)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/