在 2024/2/27 18:27, David Hildenbrand 写道:
On 26.02.24 18:32, Alex Williamson wrote:
On Tue, 27 Feb 2024 01:14:54 +0800
Yisheng Xie <ethan.xys@xxxxxxxxxxxxxxxxx> wrote:
在 2024/2/27 00:14, Alex Williamson 写道:
On Tue, 27 Feb 2024 00:01:06 +0800
Yisheng Xie<ethan.xys@xxxxxxxxxxxxxxxxx> wrote:
We meet a warning as following:
WARNING: CPU: 99 PID: 1766859 at mm/gup.c:209
try_grab_page.part.0+0xe8/0x1b0
CPU: 99 PID: 1766859 Comm: qemu-kvm Kdump: loaded Tainted: GOE
5.10.134-008.2.x86_64 #1
^^^^^^^^
Does this issue reproduce on mainline? Thanks,
I have check the code of mainline, the logical seems the same as my
version.
so I think it can reproduce if i understand correctly.
I obviously can't speak to what's in your 5.10.134-008.2 kernel, but I
do know there's a very similar issue resolved in v6.0 mainline and
included in v5.10.146 of the stable tree. Please test. Thanks,
This commit, to be precise:
commit 873aefb376bbc0ed1dd2381ea1d6ec88106fdbd4
Author: Alex Williamson <alex.williamson@xxxxxxxxxx>
Date: Mon Aug 29 21:05:40 2022 -0600
vfio/type1: Unpin zero pages
There's currently a reference count leak on the zero page. We
increment
the reference via pin_user_pages_remote(), but the page is later
handled
as an invalid/reserved page, therefore it's not accounted against the
user and not unpinned by our put_pfn().
Introducing special zero page handling in put_pfn() would
resolve the
leak, but without accounting of the zero page, a single user could
still create enough mappings to generate a reference count overflow.
The zero page is always resident, so for our purposes there's
no reason
to keep it pinned. Therefore, add a loop to walk pages returned from
pin_user_pages_remote() and unpin any zero pages.
BUT
in the meantime, we also have
commit c8070b78751955e59b42457b974bea4a4fe00187
Author: David Howells <dhowells@xxxxxxxxxx>
Date: Fri May 26 22:41:40 2023 +0100
mm: Don't pin ZERO_PAGE in pin_user_pages()
Make pin_user_pages*() leave a ZERO_PAGE unpinned if it
extracts a pointer
to it from the page tables and make unpin_user_page*()
correspondingly
ignore a ZERO_PAGE when unpinning. We don't want to risk
overrunning a
zero page's refcount as we're only allowed ~2 million pins on it -
something that userspace can conceivably trigger.
Add a pair of functions to test whether a page or a folio is a
ZERO_PAGE.
So the unpin_user_page_* won't do anything with the shared zeropage.
(likely, we could revert 873aefb376bbc0ed1dd2381ea1d6ec88106fdbd4)
Thanks for your detail info. BTW, do we need handle all of the
pagereserved page?
