Re: [f2fs-dev 4/5] f2fs: Key functions to handle inline data

From: Jaegeuk Kim
Date: Mon Oct 28 2013 - 21:07:15 EST


Hi,

2013-10-29 (í), 01:20 +0800, Huajun Li:
> On Mon, Oct 28, 2013 at 8:43 PM, Jaegeuk Kim <jaegeuk.kim@xxxxxxxxxxx> wrote:
> > Hi,
> >
> > 2013-10-26 (í), 00:01 +0800, Huajun Li:
> >> From: Huajun Li <huajun.li@xxxxxxxxx>
> >>
> >> Functions to implement inline data read/write, and move inline data to
> >> normal data block when file size exceeds inline data limitation.
> >>
> >> Signed-off-by: Huajun Li <huajun.li@xxxxxxxxx>
> >> Signed-off-by: Haicheng Li <haicheng.li@xxxxxxxxxxxxxxx>
> >> Signed-off-by: Weihong Xu <weihong.xu@xxxxxxxxx>
> >> ---
> >> fs/f2fs/Makefile | 2 +-
> >> fs/f2fs/f2fs.h | 7 +++
> >> fs/f2fs/inline.c | 144 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
> >> 3 files changed, 152 insertions(+), 1 deletion(-)
> >> create mode 100644 fs/f2fs/inline.c
> >>
> >
> > [snip]
> >
> >> +static int __f2fs_convert_inline_data(struct inode *inode, struct page *page)
> >> +{
> >> + int err;
> >> + struct page *ipage;
> >> + struct dnode_of_data dn;
> >> + void *src_addr, *dst_addr;
> >> + struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
> >> +
> >
> > Here.. f2fs_lock_op(sbi);
> >
> >> + ipage = get_node_page(sbi, inode->i_ino);
> >> + if (IS_ERR(ipage))
> >> + return PTR_ERR(ipage);
> >> +
> >
> > Need to move f2fs_lock_op prior to get_node_page.
> >
> >> + /* i_addr[0] is not used for inline data,
> >
> > Coding style.
> > /*
> > * ...
> > */
> >
> >> + * so reserving new block will not destroy inline data */
> >> + err = f2fs_reserve_block(inode, &dn, 0);
> >> + if (err) {
> >> + f2fs_put_page(ipage, 1);
> >> + f2fs_unlock_op(sbi);
> >> + return err;
> >> + }
> >
> > Need to move this too.
> >
> >> +
> >> + src_addr = inline_data_addr(ipage);
> >> + dst_addr = page_address(page);
> >> + zero_user_segment(page, 0, PAGE_CACHE_SIZE);
> >
> > + space for readability.
> >
> >> + /* Copy the whole inline data block */
> >> + memcpy(dst_addr, src_addr, MAX_INLINE_DATA);
> >> + zero_user_segment(ipage, INLINE_DATA_OFFSET,
> >> + INLINE_DATA_OFFSET + MAX_INLINE_DATA);
> >> + clear_inode_flag(F2FS_I(inode), FI_INLINE_DATA);
> >> + set_raw_inline(F2FS_I(inode),
> >> + (struct f2fs_inode *)page_address(ipage));
>
> Thanks for your comments, I will update them according to above comments.
>
> >
> > --> sync_inode_page()?
> >
> IMO, we just handle file data, so do we still need call sync_inode_page() ?

I think sync_inode_page() is more suitable, since we need to sync
between i_size, i_blocks and its i_flag, inline_data, at this moment.

>
> >> +
> >> + set_page_dirty(ipage);

Need to consider skipping set_page_dirty(ipage).

> >> + f2fs_put_page(ipage, 1);
> >> + set_page_dirty(page);
> >
> > Here... f2fs_unlock_op(sbi);
> >
> > Here, we need to consider data consistency.
> > Let's think about the below scenario.
> > 1. inline_data was written.
> > 2. sync_fs is done.
> > 3. additional data were written.
> > : this triggers f2fs_convert_inline_data(), produces a page, and then
> > unsets the inline flag.
> > 4. checkpoint was done with updated inode page. Note that, checkpoint
> > doesn't guarantee any user data.
> > 5. sudden power-off is occurred.
> > -> Once power-off-recovery is done, we loose the inline_data which was
> > written successfully at step #2.
> >
> > So, I think we need to do f2fs_sync_file() procedure at this moment.
> > Any idea?
> >
>
> Yes, need consider this case carefully, thanks for your reminder.
>
> Considering sudden power-off may happen before f2fs_sync_file() is
> called, so how about following solutions:
> ...
> /* Copy the whole inline data block */
> memcpy(dst_addr, src_addr, MAX_INLINE_DATA);
>
> do f2fs_sync_file() procedure ( Or do write_data_page() procedure ? )

Ya, but it may still have some conditions to do this or not.
One of them is checking whether previous inline data should be recovered
or not, for example.

>
> zero_user_segment(ipage, INLINE_DATA_OFFSET, INLINE_DATA_OFFSET +
> MAX_INLINE_DATA);
> clear_inode_flag(F2FS_I(inode), FI_INLINE_DATA);
> set_raw_inline(F2FS_I(inode), (struct f2fs_inode *)page_address(ipage));
> ...
>
> >> +
> >> + return 0;
> >> +}
> >> +
> >> +int f2fs_convert_inline_data(struct inode *inode,
> >> + struct page *p, unsigned flags)
> >> +{
> >> + int err;
> >> + struct page *page;
> >> +
> >> + if (p && !p->index) {
> >> + page = p;
> >> + } else {
> >> + page = grab_cache_page_write_begin(inode->i_mapping, 0, flags);
> >> + if (IS_ERR(page))
> >> + return PTR_ERR(page);
> >> + }
> >> +
> >> + err = __f2fs_convert_inline_data(inode, page);
> >> +
> >> + if (p && !p->index)
> >> + f2fs_put_page(page, 1);
> >> +
> >> + return err;
> >> +}
> >> +
> >> +int f2fs_write_inline_data(struct inode *inode,
> >> + struct page *page, unsigned size)
> >> +{
> >> + void *src_addr, *dst_addr;
> >> + struct page *ipage;
> >> + struct dnode_of_data dn;
> >> + int err;
> >> +
> >> + set_new_dnode(&dn, inode, NULL, NULL, 0);
> >> + err = get_dnode_of_data(&dn, 0, LOOKUP_NODE);
> >> + if (err)
> >> + return err;
> >> + ipage = dn.inode_page;
> >> +
> >> + src_addr = page_address(page);
> >> + dst_addr = inline_data_addr(ipage);
> >> +
> >> + zero_user_segment(ipage, INLINE_DATA_OFFSET,
> >> + INLINE_DATA_OFFSET + MAX_INLINE_DATA);
> >> + memcpy(dst_addr, src_addr, size);
> >> + if (!f2fs_has_inline_data(inode))
> >> + set_inode_flag(F2FS_I(inode), FI_INLINE_DATA);
> >> + set_raw_inline(F2FS_I(inode),
> >> + (struct f2fs_inode *)page_address(ipage));
> >> + set_page_dirty(ipage);
> >> +
> >> + if ((F2FS_I(inode)->i_xattr_nid && inode->i_blocks == 2) ||
> >> + (!F2FS_I(inode)->i_xattr_nid && inode->i_blocks == 1)) {
> >> + f2fs_put_dnode(&dn);
> >> + return 0;
> >> + }
> >> +
> >> + /* Release the first data block if it is allocated */
> >> + truncate_data_blocks_range(&dn, 1);
> >> + f2fs_put_dnode(&dn);
> >> +
> >> + return 0;
> >> +}
> >
> > --
> > Jaegeuk Kim
> > Samsung
> >
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html

--
Jaegeuk Kim
Samsung

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/