Re: [PATCH 5/8] btrfs: check if the fsid in the primary sb and copy sb are same

[Anand Jain <anand.jain@oracle.com>]

On 03/27/2018 04:49 PM, Nikolay Borisov wrote:
> 
> 
> On 26.03.2018 11:27, Anand Jain wrote:
>> During the btrfs dev scan make sure that other copies of superblock
>> contain the same fsid as the primary SB. So that we bring to the
>> user notice if the superblock has been overwritten.
>>
>>   mkfs.btrfs -fq /dev/sdc
>>   mkfs.btrfs -fq /dev/sdb
>>   dd if=/dev/sdb of=/dev/sdc count=4K skip=64K seek=64K obs=1 ibs=1
>>   mount /dev/sdc /btrfs
>>
>> Caveat: Pls note that older btrfs-progs do not wipe the non-overwriting
>> stale superblock like copy2 if a smaller mkfs.btrfs -b  <size>  is created.
>> So this patch in the kernel will report error. The workaround is to wipe
>> the superblock manually, like
>>   dd if=/dev/zero of=<dev> seek=274877906944 ibs=1 obs=1 count4K
>> OR apply btrfs-progs patch
>>    btrfs-progs: wipe copies of the stale superblock beyond -b size
>> which shall find and wipe the non overwriting superblock.
>>
>> Signed-off-by: Anand Jain <anand.jain@oracle.com>
>> ---
>>   fs/btrfs/volumes.c | 60 ++++++++++++++++++++++++++++++++++++++++++------------
>>   1 file changed, 47 insertions(+), 13 deletions(-)
>>
>> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
>> index ed22f0a3d239..45dd0674571b 100644
>> --- a/fs/btrfs/volumes.c
>> +++ b/fs/btrfs/volumes.c
>> @@ -1198,40 +1198,74 @@ static int btrfs_read_disk_super(struct block_device *bdev, u64 bytenr,
>>   int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
>>   			  struct btrfs_fs_devices **fs_devices_ret)
>>   {
>> +	struct btrfs_super_block *disk_super_primary;
>>   	struct btrfs_super_block *disk_super;
>>   	struct btrfs_device *device;
>>   	struct block_device *bdev;
>>   	struct page *page;
>>   	int ret = 0;
>> -	u64 bytenr;
>> +	int i;
>>   
>> -	/*
>> -	 * we would like to check all the supers, but that would make
>> -	 * a btrfs mount succeed after a mkfs from a different FS.
>> -	 * So, we need to add a special mount option to scan for
>> -	 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
>> -	 */
>> -	bytenr = btrfs_sb_offset(0);
>>   	flags |= FMODE_EXCL;
>>   
>>   	bdev = blkdev_get_by_path(path, flags, holder);
>>   	if (IS_ERR(bdev))
>>   		return PTR_ERR(bdev);
>>   
>> -	ret = btrfs_read_disk_super(bdev, bytenr, &page, &disk_super);
>> -	if (ret)
>> +	disk_super_primary = kzalloc(sizeof(*disk_super_primary), GFP_KERNEL);
>> +	if (!disk_super_primary) {
>> +		ret = -ENOMEM;
>>   		goto error_bdev_put;
>> +	}
>> +
>> +	/*
>> +	 * We would like to check all the supers and use one good copy,
>> +	 * but that would make a btrfs mount succeed after a mkfs from
>> +	 * a different FS.
>> +	 * So, we need to add a special mount option to scan for
>> +	 * later supers, using BTRFS_SUPER_MIRROR_MAX instead.
>> +	 * So, just validate if all copies of the superblocks are ok
>> +	 * and have the same fsid.
>> +	 */
>> +	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
>> +		u64 bytenr = btrfs_sb_offset(i);
>> +
>> +		ret = btrfs_read_disk_super(bdev, bytenr, &page, &disk_super);
>> +		if (ret) {
>> +			if (i == 0)
>> +				goto error_kfree;
>> +			/* copy2 is optional */
>> +			ret = 0;
>> +			continue;
>> +		}
>> +
>> +		if (i == 0) {
>> +			memcpy(disk_super_primary, disk_super,
>> +			       sizeof(*disk_super_primary));
>> +			btrfs_release_disk_super(page);
>> +			continue;
> 
> Doing the memcpy is enough here, the bottom of the loop already releases
> the disk page and continues on the next iteration.

  The page map happens inside btrfs_read_disk_super(), we
  need unmap before going for the next superblock.


>> +		} else if (memcmp(disk_super_primary->fsid, disk_super->fsid,
>> +			   BTRFS_FSID_SIZE)) {
>> +			pr_err("BTRFS (device %pg): superblock fsid missmatch "\
>> +				"primary %pU copy%d %pU", bdev,
>> +				disk_super_primary->fsid, i, disk_super->fsid);
>> +			ret = -EINVAL;
>> +			btrfs_release_disk_super(page);
>> +			goto error_kfree;
>> +		}
>> +		btrfs_release_disk_super(page);
> 
> I'd say split the "read first sb" from the loop, because alway want to
> read it and return an error if it fails. And then have the loop begin at
> i = 1 and handle only the possible mirrors of the sb. That would clean
> up the nested 'if' in handling the ret.
> 
> Also you could introduce another struct *page primary_page where you
> read the first super block. That way you save a memcpy + kzalloc but
> you'd have to always free it on function exit so I am not sure how much
> value it brings in terms of readability.

  Right. Also with this approach I don't have to kzallo() anymore. Will fix.

Thanks, Anand

>> +	}
>>   
>>   	mutex_lock(&uuid_mutex);
>> -	device = device_list_add(path, disk_super);
>> +	device = device_list_add(path, disk_super_primary);
>>   	if (IS_ERR(device))
>>   		ret = PTR_ERR(device);
>>   	else
>>   		*fs_devices_ret = device->fs_devices;
>>   	mutex_unlock(&uuid_mutex);
>>   
>> -	btrfs_release_disk_super(page);
>> -
>> +error_kfree:
>> +	kfree(disk_super_primary);
>>   error_bdev_put:
>>   	blkdev_put(bdev, flags);
>>   
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>