Messages in this thread | | | Date | Thu, 12 Dec 2019 08:55:12 -0800 | From | Jaegeuk Kim <> | Subject | Re: [f2fs-dev] [PATCH 6/6] f2fs: set I_LINKABLE early to avoid wrong access by vfs |
| |
On 12/11, Chao Yu wrote: > On 2019/12/11 9:31, Jaegeuk Kim wrote: > > On 12/11, Chao Yu wrote: > >> On 2019/12/11 9:21, Jaegeuk Kim wrote: > >>> On 12/10, Chao Yu wrote: > >>>> On 2019/12/10 6:23, Jaegeuk Kim wrote: > >>>>> This patch moves setting I_LINKABLE early in rename2(whiteout) to avoid the > >>>>> below warning. > >>>>> > >>>>> [ 3189.163385] WARNING: CPU: 3 PID: 59523 at fs/inode.c:358 inc_nlink+0x32/0x40 > >>>>> [ 3189.246979] Call Trace: > >>>>> [ 3189.248707] f2fs_init_inode_metadata+0x2d6/0x440 [f2fs] > >>>>> [ 3189.251399] f2fs_add_inline_entry+0x162/0x8c0 [f2fs] > >>>>> [ 3189.254010] f2fs_add_dentry+0x69/0xe0 [f2fs] > >>>>> [ 3189.256353] f2fs_do_add_link+0xc5/0x100 [f2fs] > >>>>> [ 3189.258774] f2fs_rename2+0xabf/0x1010 [f2fs] > >>>>> [ 3189.261079] vfs_rename+0x3f8/0xaa0 > >>>>> [ 3189.263056] ? tomoyo_path_rename+0x44/0x60 > >>>>> [ 3189.265283] ? do_renameat2+0x49b/0x550 > >>>>> [ 3189.267324] do_renameat2+0x49b/0x550 > >>>>> [ 3189.269316] __x64_sys_renameat2+0x20/0x30 > >>>>> [ 3189.271441] do_syscall_64+0x5a/0x230 > >>>>> [ 3189.273410] entry_SYSCALL_64_after_hwframe+0x49/0xbe > >>>>> [ 3189.275848] RIP: 0033:0x7f270b4d9a49 > >>>>> > >>>>> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> > >>>>> --- > >>>>> fs/f2fs/namei.c | 27 +++++++++++++-------------- > >>>>> 1 file changed, 13 insertions(+), 14 deletions(-) > >>>>> > >>>>> diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c > >>>>> index a1c507b0b4ac..5d9584281935 100644 > >>>>> --- a/fs/f2fs/namei.c > >>>>> +++ b/fs/f2fs/namei.c > >>>>> @@ -797,6 +797,7 @@ static int __f2fs_tmpfile(struct inode *dir, struct dentry *dentry, > >>>>> > >>>>> if (whiteout) { > >>>>> f2fs_i_links_write(inode, false); > >>>>> + inode->i_state |= I_LINKABLE; > >>>>> *whiteout = inode; > >>>>> } else { > >>>>> d_tmpfile(dentry, inode); > >>>>> @@ -867,6 +868,12 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> F2FS_I(old_dentry->d_inode)->i_projid))) > >>>>> return -EXDEV; > >>>>> > >>>>> + if (flags & RENAME_WHITEOUT) { > >>>>> + err = f2fs_create_whiteout(old_dir, &whiteout); > >>>>> + if (err) > >>>>> + return err; > >>>>> + } > >>>> > >>>> To record quota info correctly, we need to create whiteout inode after > >>>> dquot_initialize(old_dir)? > >>> > >>> __f2fs_tmpfile() will do it. > >> > >> Okay. > >> > >> Any comments on below question? > >> > >>> > >>>> > >>>>> + > >>>>> err = dquot_initialize(old_dir); > >>>>> if (err) > >>>>> goto out; > >>>>> @@ -898,17 +905,11 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> } > >>>>> } > >>>>> > >>>>> - if (flags & RENAME_WHITEOUT) { > >>>>> - err = f2fs_create_whiteout(old_dir, &whiteout); > >>>>> - if (err) > >>>>> - goto out_dir; > >>>>> - } > >>>>> - > >>>>> if (new_inode) { > >>>>> > >>>>> err = -ENOTEMPTY; > >>>>> if (old_dir_entry && !f2fs_empty_dir(new_inode)) > >>>>> - goto out_whiteout; > >>>>> + goto out_dir; > >>>>> > >>>>> err = -ENOENT; > >>>>> new_entry = f2fs_find_entry(new_dir, &new_dentry->d_name, > >>>>> @@ -916,7 +917,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> if (!new_entry) { > >>>>> if (IS_ERR(new_page)) > >>>>> err = PTR_ERR(new_page); > >>>>> - goto out_whiteout; > >>>>> + goto out_dir; > >>>>> } > >>>>> > >>>>> f2fs_balance_fs(sbi, true); > >>>>> @@ -948,7 +949,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> err = f2fs_add_link(new_dentry, old_inode); > >>>>> if (err) { > >>>>> f2fs_unlock_op(sbi); > >>>>> - goto out_whiteout; > >>>>> + goto out_dir; > >>>>> } > >>>>> > >>>>> if (old_dir_entry) > >>>>> @@ -972,7 +973,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> if (IS_ERR(old_page)) > >>>>> err = PTR_ERR(old_page); > >>>>> f2fs_unlock_op(sbi); > >>>>> - goto out_whiteout; > >>>>> + goto out_dir; > >>>>> } > >>>>> } > >>>>> } > >>>>> @@ -991,7 +992,6 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> f2fs_delete_entry(old_entry, old_page, old_dir, NULL); > >>>>> > >>>>> if (whiteout) { > >>>>> - whiteout->i_state |= I_LINKABLE; > >>>>> set_inode_flag(whiteout, FI_INC_LINK); > >>>>> err = f2fs_add_link(old_dentry, whiteout); > >>>> > >>>> [ 3189.256353] f2fs_do_add_link+0xc5/0x100 [f2fs] > >>>> [ 3189.258774] f2fs_rename2+0xabf/0x1010 [f2fs] > >>>> > >>>> Does the call stack point here? if so, we have set I_LINKABLE before > >>>> f2fs_add_link(), why the warning still be triggered? > >> > >> Am I missing something? > > > > Not sure exactly tho, I suspect some races before/after unlock_new_inode(). > > Alright, I doubt some races on whiteout->i_state updating, as we set I_LINKABLE > w/o holding inode.i_lock. > > Could you have a try with holding i_lock?
I don't see the warning with this patch, and jumped to another issue. I'd like to take a look at that later.
Thanks,
> > Thanks, > > > > >> > >> Thanks, > >> > >>>> > >>>> Thanks, > >>>> > >>>>> if (err) > >>>>> @@ -1027,15 +1027,14 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, > >>>>> f2fs_unlock_op(sbi); > >>>>> if (new_page) > >>>>> f2fs_put_page(new_page, 0); > >>>>> -out_whiteout: > >>>>> - if (whiteout) > >>>>> - iput(whiteout); > >>>>> out_dir: > >>>>> if (old_dir_entry) > >>>>> f2fs_put_page(old_dir_page, 0); > >>>>> out_old: > >>>>> f2fs_put_page(old_page, 0); > >>>>> out: > >>>>> + if (whiteout) > >>>>> + iput(whiteout); > >>>>> return err; > >>>>> } > >>>>> > >>>>> > >>> . > >>> > > . > >
| |