Re: + mm-revert-thp-make-madv_hugepage-check-for-mm-def_flags.patch added to -mm tree

From: Oleg Nesterov
Date: Wed Feb 26 2014 - 09:51:32 EST


On 02/26, Christian Borntraeger wrote:
>
> On 26/02/14 00:53, akpm@xxxxxxxxxxxxxxxxxxxx wrote:
> > Subject: + mm-revert-thp-make-madv_hugepage-check-for-mm-def_flags.patch added to -mm tree
> > To: athorlton@xxxxxxx,aarcange@xxxxxxxxxx,borntraeger@xxxxxxxxxx,ebiederm@xxxxxxxxxxxx,gerald.schaefer@xxxxxxxxxx,hannes@xxxxxxxxxxx,heiko.carstens@xxxxxxxxxx,kirill.shutemov@xxxxxxxxxxxxxxx,mgorman@xxxxxxx,mingo@xxxxxxxxxx,oleg@xxxxxxxxxx,pbonzini@xxxxxxxxxx,peterz@xxxxxxxxxxxxx,riel@xxxxxxxxxx,rientjes@xxxxxxxxxx,schwidefsky@xxxxxxxxxx,viro@xxxxxxxxxxxxxxxxxx
> > From: akpm@xxxxxxxxxxxxxxxxxxxx
> > Date: Tue, 25 Feb 2014 15:53:13 -0800
> >
> >
> > The patch titled
> > Subject: mm: revert "thp: make MADV_HUGEPAGE check for mm->def_flags"
> > has been added to the -mm tree. Its filename is
> > mm-revert-thp-make-madv_hugepage-check-for-mm-def_flags.patch
> >
> > This patch should soon appear at
> > http://ozlabs.org/~akpm/mmots/broken-out/mm-revert-thp-make-madv_hugepage-check-for-mm-def_flags.patch
> > and later at
> > http://ozlabs.org/~akpm/mmotm/broken-out/mm-revert-thp-make-madv_hugepage-check-for-mm-def_flags.patch
>
>
> NAK.
>
> Since 2012 qemu does call "qemu_madvise(new_block->host, size, QEMU_MADV_HUGEPAGE);" for all kvm pages.
> (commit ad0b5321f1f797274603ebbe20108b0750baee94 Call MADV_HUGEPAGE for guest RAM allocations) so this
> breaks any recent kvm guest on s390.

Well, I can't really discuss the changes in arch/s390.

But perhaps qemu can be changed to avoid MADV_HUGEPAGE on s390 ?
Otherwise I'd suggest the change below.

Oleg.


--- x/mm/huge_memory.c
+++ x/mm/huge_memory.c
@@ -1968,8 +1968,6 @@ out:
int hugepage_madvise(struct vm_area_struct *vma,
unsigned long *vm_flags, int advice)
{
- struct mm_struct *mm = vma->vm_mm;
-
switch (advice) {
case MADV_HUGEPAGE:
/*
@@ -1977,8 +1975,16 @@ int hugepage_madvise(struct vm_area_stru
*/
if (*vm_flags & (VM_HUGEPAGE | VM_NO_THP))
return -EINVAL;
- if (mm->def_flags & VM_NOHUGEPAGE)
+
+/*
+ * MADV_HUGEPAGE after PRCTL_THP_DISABLE is broken on s390 because
+ * qemu blindly does madvise(MADV_HUGEPAGE) after s390_enable_sie().
+ */
+#ifdef CONFIG_S390
+ if (vma->vm_mm->def_flags & VM_NOHUGEPAGE)
return -EINVAL;
+#endif
+
*vm_flags &= ~VM_NOHUGEPAGE;
*vm_flags |= VM_HUGEPAGE;
/*

--
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/