From: Dan Williams <dan.j.williams@intel.com>
To: Jan Kara <jack@suse.cz>
Cc: linux-nvdimm <linux-nvdimm@lists.01.org>,
Christoph Hellwig <hch@lst.de>,
Matthew Wilcox <mawilcox@microsoft.com>,
Ross Zwisler <ross.zwisler@linux.intel.com>,
Linux MM <linux-mm@kvack.org>,
linux-fsdevel <linux-fsdevel@vger.kernel.org>,
"Luck, Tony" <tony.luck@intel.com>,
linux-xfs <linux-xfs@vger.kernel.org>,
"Darrick J. Wong" <darrick.wong@oracle.com>
Subject: Re: [PATCH 05/11] filesystem-dax: set page->index
Date: Thu, 31 May 2018 14:49:49 -0700 [thread overview]
Message-ID: <CAPcyv4j6zjvzxKb3JUHU+iSrPXybj-uRNFAWXoc46dkMgjyN6g@mail.gmail.com> (raw)
In-Reply-To: <20180531100849.xkwe5xsjpkw6nev7@quack2.suse.cz>
On Thu, May 31, 2018 at 3:08 AM, Jan Kara <jack@suse.cz> wrote:
[..]
>> >> As far as I can see reflink+dax would require teaching kernel code
>> >> paths that ->mapping may not be a singular relationship. Something
>> >> along the line's of what Jerome was presenting at LSF to create a
>> >> special value to indicate, "call back into the filesystem (or the page
>> >> owner)" to perform this operation.
>> >>
>> >> In the meantime the kernel crashes when userspace accesses poisoned
>> >> pmem via DAX. I assume that reworking rmap for the dax+reflink case
>> >> should not block dax poison handling? Yell if you disagree.
>> >
>> > The thing is, up until get_user_pages() vs truncate series ("fs, dax: use
>> > page->mapping to warn if truncate collides with a busy page" in
>> > particular), DAX was perfectly fine with reflinks since we never needed
>> > page->mapping.
>>
>> Sure, but if this rmap series had come first I still would have needed
>> to implement ->mapping. So unless we invent a general ->mapping
>> replacement and switch all mapping users, it was always going to
>> collide with DAX eventually.
>
> Yes, my comment was more in direction that life would be easier if we could
> keep DAX without rmap support but I guess that's just too cumbersome.
I'm open to deeper reworks later. As it stands currently just calling
madvise(..., MADV_HWPOISON) on a DAX mapping causes a page reference
to be leaked because the madvise code has no clue about proper
handling of DAX pages, and consuming real poison leads to a fatal
condition / reset.
I think fixing those bugs with the current rmap dependencies on
->mapping and ->index is step1 and step2 is a longer term solution for
dax rmap that does also allow reflink. I.e. it's an rmap > reflink
argument for now.
>
>> > Now this series adds even page->index dependency which makes
>> > life for rmap with reflinks even harder. So if nothing else we should at
>> > least make sure reflinked filesystems cannot be mounted with dax mount
>> > option for now and seriously start looking into how to implement rmap with
>> > reflinked files for DAX because this noticeably reduces its usefulness.
>>
>> This restriction is already in place. In xfs_reflink_remap_range() we have:
>>
>> /* Don't share DAX file data for now. */
>> if (IS_DAX(inode_in) || IS_DAX(inode_out))
>> goto out_unlock;
>
> OK, good.
>
>> All this said, perhaps we don't need to set ->link, it would just mean
>> a wider search through the rmap tree to find if the given page is
>> mapped. So, I think I can forgo setting ->link if I teach the rmap
>> code to search the entire ->mapping.
>
> I guess you mean ->index in the above. And now when thinking about it I don't
> think it is worth the complications to avoid using ->index.
Ok, and yes I meant ->index... sorry, too much struct page on the
brain presently.
next prev parent reply other threads:[~2018-05-31 21:49 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-05-22 14:39 [PATCH 00/11] mm: Teach memory_failure() about ZONE_DEVICE pages Dan Williams
2018-05-22 14:39 ` [PATCH 01/11] device-dax: convert to vmf_insert_mixed and vm_fault_t Dan Williams
2018-05-22 14:39 ` [PATCH 02/11] device-dax: cleanup vm_fault de-reference chains Dan Williams
2018-05-22 14:39 ` [PATCH 03/11] device-dax: enable page_mapping() Dan Williams
2018-05-23 9:03 ` Jan Kara
2018-05-30 19:54 ` kbuild test robot
2018-05-22 14:39 ` [PATCH 04/11] device-dax: set page->index Dan Williams
2018-05-22 14:39 ` [PATCH 05/11] filesystem-dax: " Dan Williams
2018-05-23 8:40 ` Jan Kara
2018-05-30 1:38 ` Dan Williams
2018-05-30 8:13 ` Jan Kara
2018-05-30 23:21 ` Dan Williams
2018-05-31 10:08 ` Jan Kara
2018-05-31 21:49 ` Dan Williams [this message]
2018-05-22 14:40 ` [PATCH 06/11] filesystem-dax: perform __dax_invalidate_mapping_entry() under the page lock Dan Williams
2018-05-23 9:35 ` Jan Kara
2018-05-23 13:50 ` Dan Williams
2018-05-22 14:40 ` [PATCH 07/11] mm, madvise_inject_error: fix page count leak Dan Williams
2018-05-23 4:19 ` Naoya Horiguchi
2018-05-24 20:55 ` Dan Williams
2018-05-22 14:40 ` [PATCH 08/11] x86, memory_failure: introduce {set, clear}_mce_nospec() Dan Williams
2018-05-22 14:40 ` [PATCH 09/11] mm, memory_failure: pass page size to kill_proc() Dan Williams
2018-05-23 6:41 ` Naoya Horiguchi
2018-05-22 14:40 ` [PATCH 10/11] mm, memory_failure: teach memory_failure() about dev_pagemap pages Dan Williams
2018-05-23 6:48 ` Naoya Horiguchi
2018-05-22 14:40 ` [PATCH 11/11] libnvdimm, pmem: restore page attributes when clearing errors Dan Williams
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAPcyv4j6zjvzxKb3JUHU+iSrPXybj-uRNFAWXoc46dkMgjyN6g@mail.gmail.com \
--to=dan.j.williams@intel.com \
--cc=darrick.wong@oracle.com \
--cc=hch@lst.de \
--cc=jack@suse.cz \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-nvdimm@lists.01.org \
--cc=linux-xfs@vger.kernel.org \
--cc=mawilcox@microsoft.com \
--cc=ross.zwisler@linux.intel.com \
--cc=tony.luck@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).