A question about alloc_pages()

A question about alloc_pages()

[David Howells]

Hi,

When alloc_pages() is asked to allocate a block of pages (order > 0), should I
be able to expect that page_count(pages[0]) will be 1, and page_count() for
all the other pages will be 0?

As far as I can see, nothing in the allocator alters what's in the page count
for pages beyond the first when pages are freed, and checks are made that
these are 0 upon freeing, so it looks to me like this ought to be the case.

However, I have a report that sometimes this isn't true, and I'm wondering if
the allocator can't be relied on in this way, or whether there's a bug
somewhere keeping a reference to a released page.

David

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: A question about alloc_pages()

[Christoph Lameter]

David Howells wrote:

> When alloc_pages() is asked to allocate a block of pages (order > 0), should I
> be able to expect that page_count(pages[0]) will be 1, and page_count() for
> all the other pages will be 0?

Correct.

> However, I have a report that sometimes this isn't true, and I'm wondering if
> the allocator can't be relied on in this way, or whether there's a bug
> somewhere keeping a reference to a released page.

Must be a bug.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: A question about alloc_pages()

[David Howells]

Christoph Lameter <cl@linux-foundation.org> wrote:

> Must be a bug.

Seems I wasn't returning the blocks of pages correctly: the first page needed
its page_count() setting to 1 and the an order-N sized page block had to be
order-N aligned before calling __free_pages().  However, if you don't turn on
CONFIG_DEBUG_VM, it appears to work, but produces some odd effects.

David

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>