lkml.org 
[lkml]   [2004]   [Mar]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [RFC][PATCH 1/3] radix priority search tree - objrmap complexity fix
    On Mon, Mar 29, 2004 at 07:12:34PM +0100, Hugh Dickins wrote:
    > On Mon, 29 Mar 2004, Andrea Arcangeli wrote:
    > >
    > > Here a further update for xfs:
    > >
    > > --- sles/fs/xfs/linux/xfs_vnode.h.~1~ 2004-03-29 18:33:20.047028592 +0200
    > > +++ sles/fs/xfs/linux/xfs_vnode.h 2004-03-29 19:02:37.101915648 +0200
    > > @@ -601,8 +601,8 @@ static __inline__ void vn_flagclr(struct
    > > * Some useful predicates.
    > > */
    > > #define VN_MAPPED(vp) \
    > > - (!list_empty(&(LINVFS_GET_IP(vp)->i_mapping->i_mmap)) || \
    > > - (!list_empty(&(LINVFS_GET_IP(vp)->i_mapping->i_mmap_shared))))
    > > + (!prio_tree_empty(&(LINVFS_GET_IP(vp)->i_mapping->i_mmap)) || \
    > > + (!prio_tree_empty(&(LINVFS_GET_IP(vp)->i_mapping->i_mmap_shared))))
    > > #define VN_CACHED(vp) (LINVFS_GET_IP(vp)->i_mapping->nrpages)
    > > #define VN_DIRTY(vp) mapping_tagged(LINVFS_GET_IP(vp)->i_mapping, \
    > > PAGECACHE_TAG_DIRTY)
    >
    > Needs also to check
    > !list_empty(&(LINVFS_GET_IP(vp)->i_mapping->i_mmap_nonlinear))
    >
    > Various arches need a similar conversion too (and use page_mapping(page)
    > rather than page->mapping: see arch and include/asm in my anobjrmap 3/6).
    >
    > Those arches which do more than test list_empty (now prio_tree_empty),
    > arm and parisc (I think that's all): look as if they can take full

    I've noticed arm and parisc, luckily no arm/parisc user tried my tree
    yet ;).

    > advantage of the prio tree; and I hope we can ignore the nonlinears
    > in those cases - if a page is mapped in a nonlinear vma it may suffer
    > from D-cache aliasing inconsistencies if also mapped elsewhere in
    > that user address space, never mind. Is that reasonable?

    some arch was setting a max file offset multiple in the mmap API to
    avoid aliasing issues too, nonlinear broke it off, not sure if that is
    being taken into account, but certainly having a i_mmap_nonlinear will
    facilitate the life of those archs.

    > > and really some other bigger tree needs this part too (not a mainline
    > > issue).
    > >
    > > --- sles/fs/xfs/dmapi/dmapi_xfs.c.~1~ 2004-03-29 18:33:03.781501328 +0200
    > > +++ sles/fs/xfs/dmapi/dmapi_xfs.c 2004-03-29 18:58:57.754261560 +0200
    > > @@ -228,17 +228,21 @@ prohibited_mr_events(
    > > struct address_space *mapping = LINVFS_GET_IP(vp)->i_mapping;
    > > int prohibited = (1 << DM_EVENT_READ);
    > > struct vm_area_struct *vma;
    > > + struct prio_tree_iter iter;
    > >
    > > if (!VN_MAPPED(vp))
    > > return 0;
    > >
    > > #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
    > > down(&mapping->i_shared_sem);
    > > - list_for_each_entry(vma, &mapping->i_mmap_shared, shared) {
    > > + vma = __vma_prio_tree_first(&mapping->i_mmap_shared, &iter, 0, ULONG_MAX);
    > > + while (vma) {
    > > if (!(vma->vm_flags & VM_DENYWRITE)) {
    > > prohibited |= (1 << DM_EVENT_WRITE);
    > > break;
    > > }
    > > +
    > > + vma = __vma_prio_tree_next(vma, &mapping->i_mmap_shared, &iter, 0, ULONG_MAX);
    > > }
    > > up(&mapping->i_shared_sem);
    > > #else
    >
    > This looks horrid (not your change, the original), and would need to look
    > at nonlinears too; but I thought this was what i_writecount < 0 is for?

    no idea what's the point of this stuff, Christoph maybe wants to
    elaborate.
    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

    \
     
     \ /
      Last update: 2005-03-22 14:02    [W:4.344 / U:0.016 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site