Re: [PATCH v3 RESEND] f2fs: introduce sb.required_features to store incompatible features

From: Chao Yu
Date: Mon Aug 12 2019 - 03:16:08 EST




On 2019/8/9 23:26, Jaegeuk Kim wrote:
> On 08/06, Chao Yu wrote:
>> On 2019/8/6 10:11, Jaegeuk Kim wrote:
>>> On 08/06, Chao Yu wrote:
>>>> On 2019/8/6 9:24, Jaegeuk Kim wrote:
>>>>> On 08/06, Chao Yu wrote:
>>>>>> On 2019/8/6 8:35, Jaegeuk Kim wrote:
>>>>>>> On 08/02, Chao Yu wrote:
>>>>>>>> On 2019/8/2 6:35, Jaegeuk Kim wrote:
>>>>>>>>> On 08/01, Chao Yu wrote:
>>>>>>>>>> On 2019/8/1 12:22, Jaegeuk Kim wrote:
>>>>>>>>>>> On 07/31, Chao Yu wrote:
>>>>>>>>>>>> On 2019/7/31 7:18, Jaegeuk Kim wrote:
>>>>>>>>>>>>> On 07/29, Chao Yu wrote:
>>>>>>>>>>>>>> From: Chao Yu <yuchao0@xxxxxxxxxx>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Later after this patch was merged, all new incompatible feature's
>>>>>>>>>>>>>> bit should be added into sb.required_features field, and define new
>>>>>>>>>>>>>> feature function with F2FS_INCOMPAT_FEATURE_FUNCS() macro.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Then during mount, we will do sanity check with enabled features in
>>>>>>>>>>>>>> image, if there are features in sb.required_features that kernel can
>>>>>>>>>>>>>> not recognize, just fail the mount.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Signed-off-by: Chao Yu <yuchao0@xxxxxxxxxx>
>>>>>>>>>>>>>> ---
>>>>>>>>>>>>>> v3:
>>>>>>>>>>>>>> - change commit title.
>>>>>>>>>>>>>> - fix wrong macro name.
>>>>>>>>>>>>>> fs/f2fs/f2fs.h | 15 +++++++++++++++
>>>>>>>>>>>>>> fs/f2fs/super.c | 10 ++++++++++
>>>>>>>>>>>>>> include/linux/f2fs_fs.h | 3 ++-
>>>>>>>>>>>>>> 3 files changed, 27 insertions(+), 1 deletion(-)
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>>>>>>>>>>>>>> index a6eb828af57f..b8e17d4ddb8d 100644
>>>>>>>>>>>>>> --- a/fs/f2fs/f2fs.h
>>>>>>>>>>>>>> +++ b/fs/f2fs/f2fs.h
>>>>>>>>>>>>>> @@ -163,6 +163,15 @@ struct f2fs_mount_info {
>>>>>>>>>>>>>> #define F2FS_CLEAR_FEATURE(sbi, mask) \
>>>>>>>>>>>>>> (sbi->raw_super->feature &= ~cpu_to_le32(mask))
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> +#define F2FS_INCOMPAT_FEATURES 0
>>>>>>>>>>>>>> +
>>>>>>>>>>>>>> +#define F2FS_HAS_INCOMPAT_FEATURE(sbi, mask) \
>>>>>>>>>>>>>> + ((sbi->raw_super->required_features & cpu_to_le32(mask)) != 0)
>>>>>>>>>>>>>> +#define F2FS_SET_INCOMPAT_FEATURE(sbi, mask) \
>>>>>>>>>>>>>> + (sbi->raw_super->required_features |= cpu_to_le32(mask))
>>>>>>>>>>>>>> +#define F2FS_CLEAR_INCOMPAT_FEATURE(sbi, mask) \
>>>>>>>>>>>>>> + (sbi->raw_super->required_features &= ~cpu_to_le32(mask))
>>>>>>>>>>>>>> +
>>>>>>>>>>>>>> /*
>>>>>>>>>>>>>> * Default values for user and/or group using reserved blocks
>>>>>>>>>>>>>> */
>>>>>>>>>>>>>> @@ -3585,6 +3594,12 @@ F2FS_FEATURE_FUNCS(lost_found, LOST_FOUND);
>>>>>>>>>>>>>> F2FS_FEATURE_FUNCS(sb_chksum, SB_CHKSUM);
>>>>>>>>>>>>>> F2FS_FEATURE_FUNCS(casefold, CASEFOLD);
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> +#define F2FS_INCOMPAT_FEATURE_FUNCS(name, flagname) \
>>>>>>>>>>>>>> +static inline int f2fs_sb_has_##name(struct f2fs_sb_info *sbi) \
>>>>>>>>>>>>>> +{ \
>>>>>>>>>>>>>> + return F2FS_HAS_INCOMPAT_FEATURE(sbi, F2FS_FEATURE_##flagname); \
>>>>>>>>>>>>>> +}
>>>>>>>>>>>>>> +
>>>>>>>>>>>>>> #ifdef CONFIG_BLK_DEV_ZONED
>>>>>>>>>>>>>> static inline bool f2fs_blkz_is_seq(struct f2fs_sb_info *sbi, int devi,
>>>>>>>>>>>>>> block_t blkaddr)
>>>>>>>>>>>>>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>>>>>>>>>>>>>> index 5540fee0fe3f..3701dcce90e6 100644
>>>>>>>>>>>>>> --- a/fs/f2fs/super.c
>>>>>>>>>>>>>> +++ b/fs/f2fs/super.c
>>>>>>>>>>>>>> @@ -2513,6 +2513,16 @@ static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
>>>>>>>>>>>>>> return -EINVAL;
>>>>>>>>>>>>>> }
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> + /* check whether current kernel supports all features on image */
>>>>>>>>>>>>>> + if (le32_to_cpu(raw_super->required_features) &
>>>>>>>>>>>>>
>>>>>>>>>>>>> ...
>>>>>>>>>>>>> #define F2FS_FEATURE_VERITY 0x0400 /* reserved */
>>>>>>>>>>>>> ...
>>>>>>>>>>>>> #define F2FS_FEATURE_CASEFOLD 0x1000
>>>>>>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x1BFF
>>>>>>>>>>>>>
>>>>>>>>>>>>> if (le32_to_cpu(raw_super->required_features) & ~F2FS_FEATURE_SUPPORT) {
>>>>>>>>>>>>> ...
>>>>>>>>>>>>> return -EINVAL;
>>>>>>>>>>>>> }
>>>>>>>>>>>>
>>>>>>>>>>>> Um, I thought .required_features are used to store new feature flags from 0x0.
>>>>>>>>>>>>
>>>>>>>>>>>> All 'F2FS_FEATURE_SUPPORT' bits should be stored in sb.feature instead of
>>>>>>>>>>>> sb.required_features, I'm confused...
>>>>>>>>>>>
>>>>>>>>>>> I'm thinking,
>>>>>>>>>>>
>>>>>>>>>>> f2fs-tools sb->required_features f2fs F2FS_FEATURE_SUPPORT
>>>>>>>>>>> v0 0 v0 no_check -> ok
>>>>>>>>>>> v1 0x1BFF v0 no_check -> ok
>>>>>>>>>>> v0 0 v1 0x1BFF -> ok
>>>>>>>>>>> v1 0x1BFF v1 0x1BFF -> ok
>>>>>>>>>>> v2 0x3BFF v1 0x1BFF -> fail
>>>>>>>>>>> v1 0x1BFF v2 0x3BFF -> ok
>>>>>>>>>>> v2 0x3BFF v2 0x3BFF -> ok
>>>>>>>>>>
>>>>>>>>>> I see, it's a bit waste for 0x1FFF low bits in sb->required_features. Why not
>>>>>>>>>> leaving 0x0FFF in sb->feature w/o sanity check. And make all new incompatible
>>>>>>>>>> features (including casefold) adding into sb->required_features.
>>>>>>>>>
>>>>>>>>> I don't think we can define like this, and we still have 32bits feature filed.
>>>>>>>>> This would give another confusion to understand. VERITY is reserved only now.
>>>>>>>>>
>>>>>>>>> #define F2FS_FEATURE_CASEFOLD 0x0001
>>>>>>>>
>>>>>>>> Oops, so you want to make .required_features being almost a mirror of .feature,
>>>>>>>> and do sanity check on it... I can see now. :P
>>>>>>>>
>>>>>>>> If so, why not just use .feature:
>>>>>>>
>>>>>>> Sometimes, we don't need to set the flag, but not required at some point.
>>>>>>> (e.g., verify)
>>>>>>
>>>>>> Sorry, I'm not sure whether I have understood your point... :(
>>>>>>
>>>>>> IIUC of your point, we have defined F2FS_FEATURE_SUPPORT (0x0BFF) which excludes
>>>>>> F2FS_FEATURE_VERITY (0x0400) feature bit, then once verity feature merged in
>>>>>> kernel, we can add it into F2FS_FEATURE_SUPPORT, any problem we may face here?
>>>>>
>>>>> I was thinking the cases like "don't care features" made by mkfs. For example,
>>>>> mkfs can set F2FS_FEATURE_BLKZONED, which doesn't need f2fs being supported.
>>>>
>>>> Yes, I can understand this.
>>>>
>>>> So F2FS_FEATURE_SUPPORT can exclude them directly?
>>>
>>> No, I'd like to control it via mkfs. Kernel always needs to say what they can
>>> support, IIUC your point.
>>
>> Oh, it's different macros, we will define F2FS_INCOMPAT_FEATURE_SUPPORT 0x19B9,
>> and F2FS_ALL_FEATURE_SUPPORT (0x0001 | 0x0002 ... ).
>>
>> In sanity check function, we only check .feature with
>> F2FS_INCOMPAT_FEATURE_SUPPORT. And of course, meanwhile kernel will say it
>> supports features in F2FS_ALL_FEATURE_SUPPORT.
>
> Still I don't get the point where we need multiple macros. Why not user just can

We don't have to, but I just meant all supported and incompatible features are
different in kernel side.

> give required_feature in mkfs and check it by kernel later?

Yes, we can, I just doubt that we don't need the field .required_features...

Because .feature has included all features user wanted, and
F2FS_INCOMPAT_FEATURE_SUPPORT has included all features (need layout change)
kernel need to check. I didn't get the point why we need additional field, am I
missing something?

Thanks,

>
>>
>> Thanks,
>>
>>>
>>>>
>>>> excluded:
>>>>
>>>> #define F2FS_FEATURE_BLKZONED 0x0002
>>>> #define F2FS_FEATURE_ATOMIC_WRITE 0x0004
>>>> #define F2FS_FEATURE_LOST_FOUND 0x0200
>>>>
>>>> included:
>>>>
>>>> #define F2FS_FEATURE_ENCRYPT 0x0001
>>>> #define F2FS_FEATURE_EXTRA_ATTR 0x0008
>>>> #define F2FS_FEATURE_PRJQUOTA 0x0010
>>>> #define F2FS_FEATURE_INODE_CHKSUM 0x0020
>>>> #define F2FS_FEATURE_FLEXIBLE_INLINE_XATTR 0x0040
>>>> #define F2FS_FEATURE_QUOTA_INO 0x0080
>>>> #define F2FS_FEATURE_INODE_CRTIME 0x0100
>>>> #define F2FS_FEATURE_SB_CHKSUM 0x0800
>>>> //#define F2FS_FEATURE_VERITY 0x0400 /* reserved */
>>>> #define F2FS_FEATURE_CASEFOLD 0x1000
>>>>
>>>> #define F2FS_FEATURE_SUPPORT 0x19B9
>>>>
>>>> Thanks,
>>>>
>>>>>
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> kernel tool
>>>>>>>> v5.2 .. 1.12
>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x0BFF
>>>>>>>>
>>>>>>>> v5.3 .. 1.13
>>>>>>>> #define F2FS_FEATURE_CASEFOLD 0x1000
>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x1BFF
>>>>>>>>
>>>>>>>> v5.4 .. 1.14
>>>>>>>> #define F2FS_FEATURE_CASEFOLD 0x1000
>>>>>>>> #define F2FS_FEATURE_COMPRESS 0x2000
>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x3BFF
>>>>>>>>
>>>>>>>> f2fs-tools sb->feature f2fs F2FS_FEATURE_SUPPORT
>>>>>>>>
>>>>>>>> [enable all features in tools]
>>>>>>>> v1.12 0x0BFF v5.2 no_check -> ok
>>>>>>>> v1.12 0x0BFF v5.3 0x1BFF -> ok
>>>>>>>> v1.12 0x0BFF v5.4 0x3BFF -> ok
>>>>>>>>
>>>>>>>> v1.13 0x1BFF v5.2 that's issue we need to fix
>>>>>>>> v1.13 0x1BFF v5.3 0x1BFF -> ok
>>>>>>>> v1.13 0x1BFF v5.4 0x3BFF -> ok
>>>>>>>>
>>>>>>>> v1.14 0x3BFF v5.2 that's issue we need to fix
>>>>>>>> v1.14 0x3BFF v5.3 0x1BFF -> fail
>>>>>>>> v1.14 0x3BFF v5.4 0x3BFF -> ok
>>>>>>>>
>>>>>>>> Or am I missing something?
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>>
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Then that would be:
>>>>>>>>>>
>>>>>>>>>> kernel tool
>>>>>>>>>> v5.2 .. 1.12
>>>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x0000
>>>>>>>>>>
>>>>>>>>>> v5.3 .. 1.13
>>>>>>>>>> #define F2FS_FEATURE_CASEFOLD 0x0001
>>>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x0001
>>>>>>>>>>
>>>>>>>>>> v5.4 .. 1.14
>>>>>>>>>> #define F2FS_FEATURE_CASEFOLD 0x0001
>>>>>>>>>> #define F2FS_FEATURE_COMPRESS 0x0002
>>>>>>>>>> #define F2FS_FEATURE_SUPPORT 0x0003
>>>>>>>>>>
>>>>>>>>>> f2fs-tools sb->required_features f2fs F2FS_FEATURE_SUPPORT
>>>>>>>>>>
>>>>>>>>>> v1.12 0x0000 v5.2 no_check -> ok
>>>>>>>>>> v1.12 0x0000 v5.3 0x0001 -> ok
>>>>>>>>>> v1.12 0x0000 v5.4 0x0003 -> ok
>>>>>>>>>>
>>>>>>>>>> v1.13 0x0001 v5.2 that's issue we need to fix
>>>>>>>>>> v1.13 0x0001 v5.3 0x0001 -> ok
>>>>>>>>>> v1.13 0x0001 v5.4 0x0003 -> ok
>>>>>>>>>>
>>>>>>>>>> v1.14 0x0003 v5.2 that's issue we need to fix
>>>>>>>>>> v1.14 0x0003 v5.3 0x0001 -> fail
>>>>>>>>>> v1.14 0x0003 v5.4 0x0003 -> ok
>>>>>>>>>>
>>>>>>>>>> And all compatible features can be added into sb->feature[_VERITY, ....].
>>>>>>>>>>
>>>>>>>>>> Would that okay to you?
>>>>>>>>>>
>>>>>>>>>> Thanks,
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Thanks,
>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>> + ~F2FS_INCOMPAT_FEATURES) {
>>>>>>>>>>>>>> + f2fs_info(sbi, "Unsupported feature: %x: supported: %x",
>>>>>>>>>>>>>> + le32_to_cpu(raw_super->required_features) ^
>>>>>>>>>>>>>> + F2FS_INCOMPAT_FEATURES,
>>>>>>>>>>>>>> + F2FS_INCOMPAT_FEATURES);
>>>>>>>>>>>>>> + return -EINVAL;
>>>>>>>>>>>>>> + }
>>>>>>>>>>>>>> +
>>>>>>>>>>>>>> /* Check checksum_offset and crc in superblock */
>>>>>>>>>>>>>> if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
>>>>>>>>>>>>>> crc_offset = le32_to_cpu(raw_super->checksum_offset);
>>>>>>>>>>>>>> diff --git a/include/linux/f2fs_fs.h b/include/linux/f2fs_fs.h
>>>>>>>>>>>>>> index a2b36b2e286f..4141be3f219c 100644
>>>>>>>>>>>>>> --- a/include/linux/f2fs_fs.h
>>>>>>>>>>>>>> +++ b/include/linux/f2fs_fs.h
>>>>>>>>>>>>>> @@ -117,7 +117,8 @@ struct f2fs_super_block {
>>>>>>>>>>>>>> __u8 hot_ext_count; /* # of hot file extension */
>>>>>>>>>>>>>> __le16 s_encoding; /* Filename charset encoding */
>>>>>>>>>>>>>> __le16 s_encoding_flags; /* Filename charset encoding flags */
>>>>>>>>>>>>>> - __u8 reserved[306]; /* valid reserved region */
>>>>>>>>>>>>>> + __le32 required_features; /* incompatible features to old kernel */
>>>>>>>>>>>>>> + __u8 reserved[302]; /* valid reserved region */
>>>>>>>>>>>>>> __le32 crc; /* checksum of superblock */
>>>>>>>>>>>>>> } __packed;
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> --
>>>>>>>>>>>>>> 2.22.0
>>>>>>>>>>>>> .
>>>>>>>>>>>>>
>>>>>>>>>>> .
>>>>>>>>>>>
>>>>>>>>> .
>>>>>>>>>
>>>>>>> .
>>>>>>>
>>>>> .
>>>>>
>>> .
>>>
> .
>