[PATCH v1] erofs-utils: mkfs: fix fingerprint not set in certain modes
Gao Xiang
hsiangkao at linux.alibaba.com
Fri Apr 10 19:03:04 AEST 2026
On 2026/4/10 17:00, Yuezhang.Mo at sony.com wrote:
>> On 2026/4/10 14:05, Yuezhang Mo wrote:
>>> In certain modes, such as "--tar=f --sort=none", data is written to
>>> the image before fingerprint calculation. In this case, ->datasource
>>> will be set to `EROFS_INODE_DATA_SOURCE_NONE`.
>>>
>>> The original `erofs_set_inode_fingerprint()` function only attempts to
>>> read data from a local file or disk buffer; it cannot handle the
>>> `EROFS_INODE_DATA_SOURCE_NONE` case, causing fingerprint setting to be
>>> skipped.
>>>
>>> This patch adds handling for the `EROFS_INODE_DATA_SOURCE_NONE` case,
>>> reading data from the image and calculating the fingerprint.
>>>
>>> Signed-off-by: Yuezhang Mo <Yuezhang.Mo at sony.com>
>>> Reviewed-by: Friendy Su <friendy.su at sony.com>
>>> Reviewed-by: Daniel Palmer <daniel.palmer at sony.com>
>>> ---
>>> lib/inode.c | 20 ++++++++++++++------
>>> 1 file changed, 14 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/lib/inode.c b/lib/inode.c
>>> index 2cfc6c5..51d5266 100644
>>> --- a/lib/inode.c
>>> +++ b/lib/inode.c
>>> @@ -1975,6 +1975,13 @@ static int erofs_set_inode_fingerprint(struct erofs_inode *inode, int fd,
>>>
>>> if (!ishare_xattr_prefix_id)
>>> return 0;
>>> +
>>> + if (inode->datasource == EROFS_INODE_DATA_SOURCE_NONE) {
>>> + ret = erofs_iopen(&vf, inode);
>>> + if (ret)
>>> + return ret;
>>> + }
>>> +
>>> erofs_sha256_init(&md);
>>> do {
>>> u8 buf[32768];
>>> @@ -2018,12 +2025,6 @@ static int erofs_mkfs_begin_nondirectory(const struct erofs_mkfs_btctx *btctx,
>>> goto out;
>>> }
>>>
>>> - if (S_ISREG(inode->i_mode) && inode->i_size) {
>>> - ret = erofs_set_inode_fingerprint(inode, ctx.fd, ctx.fpos);
>>> - if (ret < 0)
>>> - return ret;
>>> - }
>>
>> I vaguely remembered we have to leave it here since
>> otherwise it may impact compressed files.
>>
>> Also EROFS_INODE_DATA_SOURCE_NONE means that mkfs
>> dump will change nothing about that, so I suggest
>>
>> apply erofs_set_inode_fingerprint() to every
>> EROFS_INODE_DATA_SOURCE_NONE user
>> (e.g. in lib/tar.c) instead.
>>
>
> Hi Xiang,
>
> This will bring a bit big code change.
>
> How about changing it like this? It maintains the original
> order and the change is simple.
`EROFS_INODE_DATA_SOURCE_NONE` should assume that metadata
is done as-is, that is the current design principle.
using erofs_iopen() to reuse erofs_pread() is really
unexpected for me TBH.
Thanks,
Gao Xiang
More information about the Linux-erofs
mailing list