From: Filipe Manana <fdmanana@gmail.com>
To: Qu Wenruo <wqu@suse.com>
Cc: linux-btrfs <linux-btrfs@vger.kernel.org>,
Ritesh Harjani <riteshh@linux.ibm.com>
Subject: Re: [Patch v2 40/42] btrfs: fix a crash caused by race between prepare_pages() and btrfs_releasepage()
Date: Wed, 28 Apr 2021 11:56:18 +0100 [thread overview]
Message-ID: <CAL3q7H5P79kEqWUnN2QKG92N3u7+G0uWbmeC0yT1LypV63MAYA@mail.gmail.com> (raw)
In-Reply-To: <20210427230349.369603-41-wqu@suse.com>
On Wed, Apr 28, 2021 at 12:06 AM Qu Wenruo <wqu@suse.com> wrote:
>
> [BUG]
> When running generic/095, there is a high chance to crash with subpage
> data RW support:
> assertion failed: PagePrivate(page) && page->private, in fs/btrfs/subpage.c:171
> ------------[ cut here ]------------
> kernel BUG at fs/btrfs/ctree.h:3403!
> Internal error: Oops - BUG: 0 [#1] SMP
> CPU: 1 PID: 3567 Comm: fio Tainted: G C O 5.12.0-rc7-custom+ #17
> Hardware name: Khadas VIM3 (DT)
> Call trace:
> assertfail.constprop.0+0x28/0x2c [btrfs]
> btrfs_subpage_assert+0x80/0xa0 [btrfs]
> btrfs_subpage_set_uptodate+0x34/0xec [btrfs]
> btrfs_page_clamp_set_uptodate+0x74/0xa4 [btrfs]
> btrfs_dirty_pages+0x160/0x270 [btrfs]
> btrfs_buffered_write+0x444/0x630 [btrfs]
> btrfs_direct_write+0x1cc/0x2d0 [btrfs]
> btrfs_file_write_iter+0xc0/0x160 [btrfs]
> new_sync_write+0xe8/0x180
> vfs_write+0x1b4/0x210
> ksys_pwrite64+0x7c/0xc0
> __arm64_sys_pwrite64+0x24/0x30
> el0_svc_common.constprop.0+0x70/0x140
> do_el0_svc+0x28/0x90
> el0_svc+0x2c/0x54
> el0_sync_handler+0x1a8/0x1ac
> el0_sync+0x170/0x180
> Code: f0000160 913be042 913c4000 955444bc (d4210000)
> ---[ end trace 3fdd39f4cccedd68 ]---
>
> [CAUSE]
> Although prepare_pages() calls find_or_create_page(), which returns the
> page locked, but in later prepare_uptodate_page() calls, we may call
> btrfs_readpage() which unlocked the page.
>
> This leaves a window where btrfs_releasepage() can sneak in and release
> the page.
>
> This can be proven by the dying ftrace dump:
> fio-3567 : prepare_pages: r/i=5/257 page_offset=262144 private=1 after set extent map
> fio-3536 : __btrfs_releasepage.part.0: r/i=5/257 page_offset=262144 private=1 clear extent map
> fio-3567 : prepare_uptodate_page.part.0: r/i=5/257 page_offset=262144 private=0 after readpage
> fio-3567 : btrfs_dirty_pages: r/i=5/257 page_offset=262144 private=0 NOT PRIVATE
>
> [FIX]
> In prepare_uptodate_page(), we should not only check page->mapping, but
> also PagePrivate() to ensure we are still hold a correct page which has
> proper fs context setup.
>
> Reported-by: Ritesh Harjani <riteshh@linux.ibm.com>
> Tested-by: Ritesh Harjani <riteshh@linux.ibm.com>
> Signed-off-by: Qu Wenruo <wqu@suse.com>
> ---
> fs/btrfs/file.c | 12 +++++++++++-
> 1 file changed, 11 insertions(+), 1 deletion(-)
>
> diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
> index 45ec3f5ef839..70a36852b680 100644
> --- a/fs/btrfs/file.c
> +++ b/fs/btrfs/file.c
> @@ -1341,7 +1341,17 @@ static int prepare_uptodate_page(struct inode *inode,
> unlock_page(page);
> return -EIO;
> }
> - if (page->mapping != inode->i_mapping) {
> +
> + /*
> + * Since btrfs_readpage() will get the page unlocked, we have
I find the phrasing slightly confusing - saying btrfs_readpage() will
get the page unlocked, gives the idea we pass it an unlocked page.
Saying that btrfs_readpage() unlocks the page before it returns is more clear.
> + * a window where fadvice() can try to release the page.
The race is far more generic and is related to another task calling
btrfs_releasepage() before we are able to lock the page again.
Can happen due to memory pressure, page migration, etc - certainly not
specific to a concurrent fadvise() (and not fadvice) call.
So I would mention a concurrent btrfs_releasepage() call and not fadvise().
> + * Here we check both inode mapping and PagePrivate() to
> + * make sure the page is not released.
> + *
> + * The priavte flag check is essential for subpage as we need
priavte -> private
Other than that it looks good.
Thanks.
> + * to store extra bitmap using page->private.
> + */
> + if (page->mapping != inode->i_mapping || !PagePrivate(page)) {
> unlock_page(page);
> return -EAGAIN;
> }
> --
> 2.31.1
>
--
Filipe David Manana,
“Whether you think you can, or you think you can't — you're right.”
next prev parent reply other threads:[~2021-04-28 10:56 UTC|newest]
Thread overview: 117+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-04-27 23:03 [Patch v2 00/42] btrfs: add data write support for subpage Qu Wenruo
2021-04-27 23:03 ` [Patch v2 01/42] btrfs: scrub: fix subpage scrub repair error caused by hardcoded PAGE_SIZE Qu Wenruo
2021-05-13 22:57 ` David Sterba
2021-05-13 23:32 ` Qu Wenruo
2021-04-27 23:03 ` [Patch v2 02/42] btrfs: make free space cache size consistent across different PAGE_SIZE Qu Wenruo
2021-04-27 23:03 ` [Patch v2 03/42] btrfs: remove the unused parameter @len for btrfs_bio_fits_in_stripe() Qu Wenruo
2021-05-13 22:58 ` David Sterba
2021-05-13 23:07 ` David Sterba
2021-04-27 23:03 ` [Patch v2 04/42] btrfs: allow btrfs_bio_fits_in_stripe() to accept bio without any page Qu Wenruo
2021-04-27 23:03 ` [Patch v2 05/42] btrfs: refactor submit_extent_page() to make bio and its flag tracing easier Qu Wenruo
2021-05-13 23:03 ` David Sterba
2021-05-21 11:06 ` Johannes Thumshirn
2021-05-21 11:26 ` Qu Wenruo
2021-05-21 13:30 ` David Sterba
2021-04-27 23:03 ` [Patch v2 06/42] btrfs: make subpage metadata write path to call its own endio functions Qu Wenruo
2021-04-27 23:03 ` [Patch v2 07/42] btrfs: pass btrfs_inode into btrfs_writepage_endio_finish_ordered() Qu Wenruo
2021-05-13 23:06 ` David Sterba
2021-05-13 23:35 ` Qu Wenruo
2021-05-21 14:27 ` Josef Bacik
2021-05-21 20:22 ` David Sterba
2021-05-22 0:24 ` Qu Wenruo
2021-05-23 7:40 ` Qu Wenruo
2021-05-23 13:43 ` Josef Bacik
2021-05-23 13:50 ` Qu Wenruo
2021-05-23 14:08 ` Josef Bacik
2021-04-27 23:03 ` [Patch v2 08/42] btrfs: make Private2 lifespan more consistent Qu Wenruo
2021-04-27 23:03 ` [Patch v2 09/42] btrfs: refactor how we finish ordered extent io for endio functions Qu Wenruo
2021-05-13 23:11 ` David Sterba
2021-04-27 23:03 ` [Patch v2 10/42] btrfs: update the comments in btrfs_invalidatepage() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 11/42] btrfs: introduce btrfs_lookup_first_ordered_range() Qu Wenruo
2021-05-13 23:13 ` David Sterba
2021-04-27 23:03 ` [Patch v2 12/42] btrfs: refactor btrfs_invalidatepage() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 13/42] btrfs: rename PagePrivate2 to PageOrdered inside btrfs Qu Wenruo
2021-04-27 23:03 ` [Patch v2 14/42] btrfs: pass bytenr directly to __process_pages_contig() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 15/42] btrfs: refactor the page status update into process_one_page() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 16/42] btrfs: provide btrfs_page_clamp_*() helpers Qu Wenruo
2021-04-27 23:03 ` [Patch v2 17/42] btrfs: only require sector size alignment for end_bio_extent_writepage() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 18/42] btrfs: make btrfs_dirty_pages() to be subpage compatible Qu Wenruo
2021-04-27 23:03 ` [Patch v2 19/42] btrfs: make __process_pages_contig() to handle subpage dirty/error/writeback status Qu Wenruo
2021-04-27 23:03 ` [Patch v2 20/42] btrfs: make end_bio_extent_writepage() to be subpage compatible Qu Wenruo
2021-04-27 23:03 ` [Patch v2 21/42] btrfs: make process_one_page() to handle subpage locking Qu Wenruo
2021-04-27 23:03 ` [Patch v2 22/42] btrfs: introduce helpers for subpage ordered status Qu Wenruo
2021-04-27 23:03 ` [Patch v2 23/42] btrfs: make page Ordered bit to be subpage compatible Qu Wenruo
2021-04-27 23:03 ` [Patch v2 24/42] btrfs: update locked page dirty/writeback/error bits in __process_pages_contig Qu Wenruo
2021-04-27 23:03 ` [Patch v2 25/42] btrfs: prevent extent_clear_unlock_delalloc() to unlock page not locked by __process_pages_contig() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 26/42] btrfs: make btrfs_set_range_writeback() subpage compatible Qu Wenruo
2021-04-27 23:03 ` [Patch v2 27/42] btrfs: make __extent_writepage_io() only submit dirty range for subpage Qu Wenruo
2021-04-27 23:03 ` [Patch v2 28/42] btrfs: make btrfs_truncate_block() to be subpage compatible Qu Wenruo
2021-04-27 23:03 ` [Patch v2 29/42] btrfs: make btrfs_page_mkwrite() " Qu Wenruo
2021-04-27 23:03 ` [Patch v2 30/42] btrfs: reflink: make copy_inline_to_page() " Qu Wenruo
2021-04-27 23:03 ` [Patch v2 31/42] btrfs: fix the filemap_range_has_page() call in btrfs_punch_hole_lock_range() Qu Wenruo
2021-04-27 23:03 ` [Patch v2 32/42] btrfs: don't clear page extent mapped if we're not invalidating the full page Qu Wenruo
2021-04-27 23:03 ` [Patch v2 33/42] btrfs: extract relocation page read and dirty part into its own function Qu Wenruo
2021-04-27 23:03 ` [Patch v2 34/42] btrfs: make relocate_one_page() to handle subpage case Qu Wenruo
2021-04-27 23:03 ` [Patch v2 35/42] btrfs: fix wild subpage writeback which does not have ordered extent Qu Wenruo
2021-04-27 23:03 ` [Patch v2 36/42] btrfs: disable inline extent creation for subpage Qu Wenruo
2021-05-04 4:28 ` Qu Wenruo
2021-04-27 23:03 ` [Patch v2 37/42] btrfs: skip validation for subpage read repair Qu Wenruo
2021-04-27 23:03 ` [Patch v2 38/42] btrfs: allow submit_extent_page() to do bio split for subpage Qu Wenruo
2021-04-27 23:03 ` [Patch v2 39/42] btrfs: reject raid5/6 fs " Qu Wenruo
2021-04-28 14:22 ` Neal Gompa
2021-04-28 23:11 ` Qu Wenruo
2021-05-12 22:04 ` David Sterba
2021-04-27 23:03 ` [Patch v2 40/42] btrfs: fix a crash caused by race between prepare_pages() and btrfs_releasepage() Qu Wenruo
2021-04-28 10:56 ` Filipe Manana [this message]
2021-04-27 23:03 ` [Patch v2 41/42] btrfs: fix the use-after-free bug in writeback subpage helper Qu Wenruo
2021-05-06 23:46 ` Qu Wenruo
2021-05-07 4:57 ` Ritesh Harjani
2021-05-07 5:14 ` Qu Wenruo
2021-05-10 8:38 ` Qu Wenruo
2021-05-10 12:29 ` Ritesh Harjani
2021-05-10 13:10 ` Qu Wenruo
2021-05-11 10:48 ` Ritesh Harjani
2021-05-11 11:15 ` Qu Wenruo
2021-05-12 1:49 ` Qu Wenruo
2021-05-12 7:09 ` Ritesh Harjani
2021-05-13 16:33 ` Ritesh Harjani
2021-05-13 21:36 ` Ritesh Harjani
2021-05-13 23:41 ` Qu Wenruo
2021-05-14 15:08 ` Ritesh Harjani
2021-05-14 17:53 ` Ritesh Harjani
2021-05-14 22:22 ` Qu Wenruo
2021-05-15 9:59 ` Ritesh Harjani
2021-05-15 10:15 ` Qu Wenruo
2021-05-25 4:43 ` Ritesh Harjani
2021-05-25 5:52 ` Qu Wenruo
2021-05-25 6:14 ` Qu Wenruo
2021-05-25 9:23 ` Ritesh Harjani
2021-05-25 9:45 ` Qu Wenruo
2021-05-25 9:49 ` Qu Wenruo
2021-05-25 10:20 ` Ritesh Harjani
2021-05-25 11:41 ` Qu Wenruo
2021-05-25 13:02 ` Ritesh Harjani
2021-05-26 5:29 ` Ritesh Harjani
2021-05-26 5:58 ` Qu Wenruo
2021-05-26 13:45 ` Ritesh Harjani
2021-05-28 8:26 ` Qu Wenruo
2021-05-28 8:59 ` Ritesh Harjani
2021-05-28 10:25 ` Qu Wenruo
2021-05-30 1:50 ` Qu Wenruo
2021-04-27 23:03 ` [Patch v2 42/42] btrfs: allow read-write for 4K sectorsize on 64K page size systems Qu Wenruo
2021-05-12 22:18 ` [Patch v2 00/42] btrfs: add data write support for subpage David Sterba
2021-05-12 23:48 ` Qu Wenruo
2021-05-13 2:21 ` Qu Wenruo
2021-05-13 22:54 ` David Sterba
2021-05-14 1:41 ` Qu Wenruo
2021-05-14 2:26 ` riteshh
2021-05-14 10:28 ` riteshh
2021-05-14 11:28 ` David Sterba
2021-05-14 14:38 ` riteshh
2021-05-14 11:30 ` David Sterba
2021-05-14 22:25 ` David Sterba
2021-05-14 22:45 ` Qu Wenruo
2021-05-14 23:05 ` David Sterba
2021-05-14 23:17 ` Qu Wenruo
2021-05-17 13:22 ` David Sterba
2021-05-17 23:20 ` Qu Wenruo
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=CAL3q7H5P79kEqWUnN2QKG92N3u7+G0uWbmeC0yT1LypV63MAYA@mail.gmail.com \
--to=fdmanana@gmail.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=riteshh@linux.ibm.com \
--cc=wqu@suse.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.