From: Chao Yu <chao@kernel.org>
To: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: linux-kernel@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH 2/6] f2fs: do not expose unwritten blocks to user by DIO
Date: Fri, 3 Dec 2021 14:43:07 +0800 [thread overview]
Message-ID: <9b637c29-f9c7-5928-cf2e-346253bb00c7@kernel.org> (raw)
In-Reply-To: <YakMtRif1f0cXT6R@google.com>
On 2021/12/3 2:13, Jaegeuk Kim wrote:
> On 12/02, Chao Yu wrote:
>> On 2021/11/17 5:45, Jaegeuk Kim wrote:
>>> DIO preallocates physical blocks before writing data, but if an error occurrs
>>> or power-cut happens, we can see block contents from the disk. This patch tries
>>> to fix it by 1) turning to buffered writes for DIO into holes, 2) truncating
>>> unwritten blocks from error or power-cut.
>>>
>>> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
>>> ---
>>> fs/f2fs/data.c | 5 ++++-
>>> fs/f2fs/f2fs.h | 5 +++++
>>> fs/f2fs/file.c | 24 +++++++++++++++++++++++-
>>> 3 files changed, 32 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
>>> index 3b27fb7daa8b..7ac1a39fcad2 100644
>>> --- a/fs/f2fs/data.c
>>> +++ b/fs/f2fs/data.c
>>> @@ -1543,8 +1543,11 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
>>> flag != F2FS_GET_BLOCK_DIO);
>>> err = __allocate_data_block(&dn,
>>> map->m_seg_type);
>>> - if (!err)
>>> + if (!err) {
>>> + if (flag == F2FS_GET_BLOCK_PRE_DIO)
>>> + file_need_truncate(inode);
>>> set_inode_flag(inode, FI_APPEND_WRITE);
>>> + }
>>> }
>>> if (err)
>>> goto sync_out;
>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>>> index be871a79c634..14bea669f87e 100644
>>> --- a/fs/f2fs/f2fs.h
>>> +++ b/fs/f2fs/f2fs.h
>>> @@ -654,6 +654,7 @@ enum {
>>> #define FADVISE_KEEP_SIZE_BIT 0x10
>>> #define FADVISE_HOT_BIT 0x20
>>> #define FADVISE_VERITY_BIT 0x40
>>> +#define FADVISE_TRUNC_BIT 0x80
>>> #define FADVISE_MODIFIABLE_BITS (FADVISE_COLD_BIT | FADVISE_HOT_BIT)
>>> @@ -681,6 +682,10 @@ enum {
>>> #define file_is_verity(inode) is_file(inode, FADVISE_VERITY_BIT)
>>> #define file_set_verity(inode) set_file(inode, FADVISE_VERITY_BIT)
>>> +#define file_should_truncate(inode) is_file(inode, FADVISE_TRUNC_BIT)
>>> +#define file_need_truncate(inode) set_file(inode, FADVISE_TRUNC_BIT)
>>> +#define file_dont_truncate(inode) clear_file(inode, FADVISE_TRUNC_BIT)
>>> +
>>> #define DEF_DIR_LEVEL 0
>>> enum {
>>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>>> index 4bf77a5bf998..ec8de0662437 100644
>>> --- a/fs/f2fs/file.c
>>> +++ b/fs/f2fs/file.c
>>> @@ -960,10 +960,21 @@ int f2fs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
>>> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>> filemap_invalidate_lock(inode->i_mapping);
>>> + /*
>>> + * Truncate stale preallocated blocks used by the previous DIO.
>>> + */
>>> + if (file_should_truncate(inode)) {
>>> + err = f2fs_truncate(inode);
>>> + if (err)
>>> + goto out_unlock;
>>> + file_dont_truncate(inode);
>>> + }
>>> +
>>> truncate_setsize(inode, attr->ia_size);
>>> if (attr->ia_size <= old_size)
>>> err = f2fs_truncate(inode);
>>> +out_unlock:
>>> /*
>>> * do not trim all blocks after i_size if target size is
>>> * larger than i_size.
>>> @@ -4257,6 +4268,13 @@ static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter)
>>> /* If it will be an out-of-place direct write, don't bother. */
>>> if (dio && f2fs_lfs_mode(sbi))
>>> return 0;
>>> + /*
>>> + * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into
>>> + * buffered IO, if DIO meets any holes.
>>> + */
>>> + if (dio && i_size_read(inode) &&
>>> + (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode))))
>>> + return 0;
>>> /* No-wait I/O can't allocate blocks. */
>>> if (iocb->ki_flags & IOCB_NOWAIT)
>>> @@ -4366,10 +4384,14 @@ static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
>>> if (preallocated > 0 && i_size_read(inode) < target_size) {
>>> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>> filemap_invalidate_lock(inode->i_mapping);
>>> - f2fs_truncate(inode);
>>> + if (!f2fs_truncate(inode))
>>> + file_dont_truncate(inode);
>>> filemap_invalidate_unlock(inode->i_mapping);
>>> up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>>> + } else {
>>> + file_dont_truncate(inode);
>>
>> How about this case:
>>
>> - touch file
>> - DIO write [0, 8kb] to file
>> - preallocate 2 physical blocks
>> - set FADVISE_TRUNC_BIT
>> - SPO
>> - BUFIO write [0, 4kb] to file
>> - file_dont_truncate -- it leaks unwritten [4kb, 8kb] to user after
>> truncating file to 8kb
>
> i_size should be 4kb, no?
- BUFIO write [0, 4kb] to file
- file_dont_truncate()
i_size should be 4kb, w/ FADVISE_TRUNC_BIT
- setattr 8kb
i_size should be 8kb, last [4kb, 8kb] is unwritten
Thanks,
>
>>
>> Thanks,
>>
>>> }
>>> +
>>> clear_inode_flag(inode, FI_PREALLOCATED_ALL);
>>> if (ret > 0)
>>>
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
next prev parent reply other threads:[~2021-12-03 6:43 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-11-16 21:45 [f2fs-dev] [PATCH 1/6] f2fs: rework write preallocations Jaegeuk Kim
2021-11-16 21:45 ` [f2fs-dev] [PATCH 2/6] f2fs: do not expose unwritten blocks to user by DIO Jaegeuk Kim
2021-11-17 15:16 ` Chao Yu
2021-11-17 19:32 ` Jaegeuk Kim
2021-12-02 2:39 ` Chao Yu
2021-12-02 18:13 ` Jaegeuk Kim
2021-12-03 6:43 ` Chao Yu [this message]
2021-12-03 22:53 ` Jaegeuk Kim
2021-12-04 19:01 ` [f2fs-dev] [PATCH 2/6 v2] " Jaegeuk Kim
2021-11-16 21:45 ` [f2fs-dev] [PATCH 3/6] f2fs: reduce indentation in f2fs_file_write_iter() Jaegeuk Kim
2021-12-02 2:58 ` Chao Yu
2021-11-16 21:45 ` [f2fs-dev] [PATCH 4/6] f2fs: fix the f2fs_file_write_iter tracepoint Jaegeuk Kim
2021-12-02 2:59 ` Chao Yu
2021-11-16 21:45 ` [f2fs-dev] [PATCH 5/6] f2fs: implement iomap operations Jaegeuk Kim
2021-12-02 3:10 ` Chao Yu
2021-12-02 4:15 ` Eric Biggers
2021-12-02 14:04 ` Chao Yu
2021-12-02 18:25 ` Eric Biggers
2021-12-02 18:57 ` Jaegeuk Kim
2021-12-02 19:00 ` Jaegeuk Kim
2021-12-02 20:57 ` Eric Biggers
2021-12-02 21:19 ` Jaegeuk Kim
2021-12-02 21:28 ` Jaegeuk Kim
2021-12-03 6:51 ` Chao Yu
2021-12-30 6:43 ` Chao Yu
2022-01-04 21:15 ` Jaegeuk Kim
2021-11-16 21:45 ` [f2fs-dev] [PATCH 6/6] f2fs: use iomap for direct I/O Jaegeuk Kim
2021-12-10 23:58 ` Jaegeuk Kim
2021-11-17 13:31 ` [f2fs-dev] [PATCH 1/6] f2fs: rework write preallocations Chao Yu
2021-11-17 19:27 ` Jaegeuk Kim
2021-11-18 6:47 ` Chao Yu
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=9b637c29-f9c7-5928-cf2e-346253bb00c7@kernel.org \
--to=chao@kernel.org \
--cc=jaegeuk@kernel.org \
--cc=linux-f2fs-devel@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
/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).