Re: [patch 2.6.13-rc4] fix get_user_pages bug

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


Hugh Dickins <[email protected]> wrote:
> On Mon, 1 Aug 2005, Linus Torvalds wrote:
> > 
> > that "continue" will continue without the spinlock held, and now do 
> Yes, I was at last about to reply on that point and others.
> I'll make those comments in a separate mail to Nick and all.
> > Instead, I'd suggest changing the logic for "lookup_write". Make it 
> > require that the page table entry is _dirty_ (not writable), and then 
> Attractive, I very much wanted to do that rather than change all the
> arches, but I think s390 rules it out: its pte_mkdirty does nothing,
> its pte_dirty just says no.
> Whether your patch suits all other uses of (__)follow_page I've not
> investigated (and I don't see how you can go without the set_page_dirty
> if it was necessary before);

That was introduced 19 months ago by the s390 guys (see patch below).  I
don't really see why Martin decided to mark the page software-dirty at that

It's a nice thing to do from the VM dirty-memory accounting POV, but I
don't see that it's essential.

> but at present see no alternative to
> something like Nick's patch, though I'd much prefer smaller.
> Or should we change s390 to set a flag in the pte just for this purpose?

That would be a good approach IMO, if possible.

From: Martin Schwidefsky <[email protected]>

Fix endless loop in get_user_pages() on s390.  It happens only on s/390
because pte_dirty always returns 0.  For all other architectures this is an

In the case of "write && !pte_dirty(pte)" follow_page() returns NULL.  On all
architectures except s390 handle_pte_fault() will then create a pte with
pte_dirty(pte)==1 because write_access==1.  In the following, second call to
follow_page() all is fine.  With the physical dirty bit patch pte_dirty() is
always 0 for s/390 because the dirty bit doesn't live in the pte.


 mm/memory.c |   21 +++++++++++++--------
 1 files changed, 13 insertions(+), 8 deletions(-)

diff -puN mm/memory.c~s390-16-follow_page-lockup-fix mm/memory.c
--- 25/mm/memory.c~s390-16-follow_page-lockup-fix	2004-01-18 22:36:00.000000000 -0800
+++ 25-akpm/mm/memory.c	2004-01-18 22:36:00.000000000 -0800
@@ -651,14 +651,19 @@ follow_page(struct mm_struct *mm, unsign
 	pte = *ptep;
 	if (pte_present(pte)) {
-		if (!write || (pte_write(pte) && pte_dirty(pte))) {
-			pfn = pte_pfn(pte);
-			if (pfn_valid(pfn)) {
-				struct page *page = pfn_to_page(pfn);
-				mark_page_accessed(page);
-				return page;
-			}
+		if (write && !pte_write(pte))
+			goto out;
+		if (write && !pte_dirty(pte)) {
+			struct page *page = pte_page(pte);
+			if (!PageDirty(page))
+				set_page_dirty(page);
+		}
+		pfn = pte_pfn(pte);
+		if (pfn_valid(pfn)) {
+			struct page *page = pfn_to_page(pfn);
+			mark_page_accessed(page);
+			return page;


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at
Please read the FAQ at

[Index of Archives]     [Kernel Newbies]     [Netfilter]     [Bugtraq]     [Photo]     [Gimp]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Video 4 Linux]     [Linux for the blind]
  Powered by Linux