lkml.org 
[lkml]   [2010]   [Nov]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH RFC] vmalloc: eagerly clear ptes on vunmap
On Mon, Nov 29, 2010 at 12:32:11PM -0800, Jeremy Fitzhardinge wrote:
> When unmapping a region in the vmalloc space, clear the ptes immediately.
> There's no point in deferring this because there's no amortization
> benefit.
>
> The TLBs are left dirty, and they are flushed lazily to amortize the
> cost of the IPIs.
>
> This specific motivation for this patch is a regression since 2.6.36 when
> using NFS under Xen, triggered by the NFS client's use of vm_map_ram()
> introduced in 56e4ebf877b6043c289bda32a5a7385b80c17dee. XFS also uses
> vm_map_ram() and could cause similar problems.

I do wonder whether there are cache benefits from batching page table
updates, especially the batched per cpu maps (and in your version they
get double-cleared as well). I think this patch is good, but I think
perhaps making it configurable would be nice.

So... main question, does it allow Xen to use lazy flushing and avoid
vm_unmap_aliases() calls?


>
> Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
> Cc: Nick Piggin <npiggin@kernel.dk>
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index a3d66b3..9960644 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -566,7 +566,6 @@ static void __purge_vmap_area_lazy(unsigned long *start, unsigned long *end,
> if (va->va_end > *end)
> *end = va->va_end;
> nr += (va->va_end - va->va_start) >> PAGE_SHIFT;
> - unmap_vmap_area(va);
> list_add_tail(&va->purge_list, &valist);
> va->flags |= VM_LAZY_FREEING;
> va->flags &= ~VM_LAZY_FREE;
> @@ -616,6 +615,8 @@ static void purge_vmap_area_lazy(void)
> */
> static void free_unmap_vmap_area_noflush(struct vmap_area *va)
> {
> + unmap_vmap_area(va);
> +
> va->flags |= VM_LAZY_FREE;
> atomic_add((va->va_end - va->va_start) >> PAGE_SHIFT, &vmap_lazy_nr);
> if (unlikely(atomic_read(&vmap_lazy_nr) > lazy_max_pages()))
> @@ -944,8 +945,10 @@ static void vb_free(const void *addr, unsigned long size)
> BUG_ON(vb->free);
> spin_unlock(&vb->lock);
> free_vmap_block(vb);
> - } else
> + } else {
> spin_unlock(&vb->lock);
> + vunmap_page_range((unsigned long)addr, (unsigned long)addr + size);
> + }
> }
>
> /**
> @@ -988,7 +991,6 @@ void vm_unmap_aliases(void)
>
> s = vb->va->va_start + (i << PAGE_SHIFT);
> e = vb->va->va_start + (j << PAGE_SHIFT);
> - vunmap_page_range(s, e);
> flush = 1;
>
> if (s < start)
>


\
 
 \ /
  Last update: 2010-11-30 13:45    [W:0.071 / U:0.104 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site