Re: [PATCH] ceph: make the lost+found dir accessible by kernel client

[Xiubo Li <xiubli@redhat.com>]

On 2021/4/20 20:51, Jeff Layton wrote:
> On Tue, 2021-04-20 at 10:02 +0800, Xiubo Li wrote:
>> On 2021/4/20 0:09, Jeff Layton wrote:
>>> On Mon, 2021-04-19 at 10:32 +0800, xiubli@redhat.com wrote:
>>>> From: Xiubo Li <xiubli@redhat.com>
>>>>
>>>> Inode number 0x4 is reserved for the lost+found dir, and the app
>>>> or test app need to access it.
>>>>
>>>> URL: https://tracker.ceph.com/issues/50216
>>>> Signed-off-by: Xiubo Li <xiubli@redhat.com>
>>>> ---
>>>>    fs/ceph/super.h              | 3 ++-
>>>>    include/linux/ceph/ceph_fs.h | 7 ++++---
>>>>    2 files changed, 6 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/fs/ceph/super.h b/fs/ceph/super.h
>>>> index 4808a1458c9b..0f38e6183ff0 100644
>>>> --- a/fs/ceph/super.h
>>>> +++ b/fs/ceph/super.h
>>>> @@ -542,7 +542,8 @@ static inline int ceph_ino_compare(struct inode *inode, void *data)
>>>>    
>>>>
>>>>
>>>>
>>>>    static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
>>>>    {
>>>> -	if (vino.ino < CEPH_INO_SYSTEM_BASE && vino.ino != CEPH_INO_ROOT) {
>>>> +	if (vino.ino < CEPH_INO_SYSTEM_BASE && vino.ino != CEPH_INO_ROOT &&
>>>> +	    vino.ino != CEPH_INO_LOST_AND_FOUND ) {
>>>>    		WARN_RATELIMIT(1, "Attempt to access reserved inode number 0x%llx", vino.ino);
>>>>    		return true;
>>>>    	}
>>>> diff --git a/include/linux/ceph/ceph_fs.h b/include/linux/ceph/ceph_fs.h
>>>> index e41a811026f6..57e5bd63fb7a 100644
>>>> --- a/include/linux/ceph/ceph_fs.h
>>>> +++ b/include/linux/ceph/ceph_fs.h
>>>> @@ -27,9 +27,10 @@
>>>>    #define CEPH_MONC_PROTOCOL   15 /* server/client */
>>>>    
>>>>
>>>>
>>>>
>>>>    
>>>>
>>>>
>>>>
>>>> -#define CEPH_INO_ROOT   1
>>>> -#define CEPH_INO_CEPH   2       /* hidden .ceph dir */
>>>> -#define CEPH_INO_DOTDOT 3	/* used by ceph fuse for parent (..) */
>> Hi Jeff,
>>
>> Please fix the "CEPH_INO_DOTDOT" when you folding this patch. The inode
>> number 3 is not _DOTDOT any more. This was introduced by an very old
>> commit(dd6f5e105d85e) but I couldn't find the related change about this
>> in ceph code.
>>
>> It should be:
>>
>> #define CEPH_INO_GLOBAL_SNAPREALM 3
>>
>>
> Sounds good. I can fold that change into the patch.
>
> What should we use for a comment there? I took a look at the MDS code
> and it wasn't clear to me what the global_snaprealm is actually for...

Maybe the commit comment title is a good choice from Zheng's initial commit:

commit 0f31d49de20bd40836dceebb06e12763a1def5e0
Author: Yan, Zheng <ukernel@gmail.com>
Date:   Thu Oct 26 15:32:29 2017 +0800

     mds: attach inode with multiple hardlinks to dummy global snaprealm

     The dummy global snaprealm includes all snapshots in the filesystem.
     For any later snapshot, mds will COW the inode and preserve snap data.
     These snap data will cover any possible snapshot on remote linkages
     of the inode.




>>>> +#define CEPH_INO_ROOT           1
>>>> +#define CEPH_INO_CEPH           2 /* hidden .ceph dir */
>>>> +#define CEPH_INO_DOTDOT         3 /* used by ceph fuse for parent (..) */
>>>> +#define CEPH_INO_LOST_AND_FOUND 4 /* lost+found dir */
>>>>    
>>>>
>>>>
>>>>
>>>>    /* arbitrary limit on max # of monitors (cluster of 3 is typical) */
>>>>    #define CEPH_MAX_MON   31
>>> Thanks Xiubo,
>>>
>>> For some background, apparently cephfs-data-scan can create this
>>> directory, and the clients do need access to it. I'll fold this into the
>>> original patch that makes these inodes inaccessible (ceph: don't allow
>>> access to MDS-private inodes).
>>>
>>> Cheers!
>>