[PATCH v2] erofs-utils: fsck: add --xattrs and --no-xattrs options

Hongzhen Luo hongzhen at linux.alibaba.com
Tue Sep 3 21:10:32 AEST 2024


On 2024/9/3 17:32, Gao Xiang wrote:
>
>
> On 2024/9/3 16:56, Hongzhen Luo wrote:
>> The current `fsck --extract` does not support exporting the extended
>> attributes of files. This patch adds `--xattrs` option to dump the
>> extended attributes, as well as the `--no-xattrs` option to not dump
>> the extended attributes.
>>
>> Signed-off-by: Hongzhen Luo <hongzhen at linux.alibaba.com>
>> ---
>> v2: In addition to "--xattrs", the option "--no-xattrs" has been added.
>> v1: 
>> https://lore.kernel.org/all/20240903073517.3311407-1-hongzhen@linux.alibaba.com/
>> ---
>>   fsck/main.c | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 74 insertions(+)
>>
>> diff --git a/fsck/main.c b/fsck/main.c
>> index 28f1e7e..273bf73 100644
>> --- a/fsck/main.c
>> +++ b/fsck/main.c
>> @@ -9,6 +9,7 @@
>>   #include <utime.h>
>>   #include <unistd.h>
>>   #include <sys/stat.h>
>> +#include <sys/xattr.h>
>>   #include "erofs/print.h"
>>   #include "erofs/compress.h"
>>   #include "erofs/decompress.h"
>> @@ -31,6 +32,7 @@ struct erofsfsck_cfg {
>>       bool overwrite;
>>       bool preserve_owner;
>>       bool preserve_perms;
>> +    bool dump_xattrs;
>>   };
>>   static struct erofsfsck_cfg fsckcfg;
>>   @@ -48,6 +50,8 @@ static struct option long_options[] = {
>>       {"no-preserve-owner", no_argument, 0, 10},
>>       {"no-preserve-perms", no_argument, 0, 11},
>>       {"offset", required_argument, 0, 12},
>> +    {"xattrs", no_argument, 0, 13},
>> +    {"no-xattrs", no_argument, 0, 14},
>>       {0, 0, 0, 0},
>>   };
>>   @@ -98,6 +102,8 @@ static void usage(int argc, char **argv)
>>           " --extract[=X]          check if all files are well 
>> encoded, optionally\n"
>>           "                        extract to X\n"
>>           " --offset=#             skip # bytes at the beginning of 
>> IMAGE\n"
>> +        " --xattrs               dump extended attributes (default)\n"
>> +        " --no-xattrs            do not dump extended attributes\n"
>>           "\n"
>>           " -a, -A, -y             no-op, for compatibility with fsck 
>> of other filesystems\n"
>>           "\n"
>> @@ -225,6 +231,16 @@ static int erofsfsck_parse_options_cfg(int argc, 
>> char **argv)
>>                   return -EINVAL;
>>               }
>>               break;
>> +        case 13:
>> +            if (!fsckcfg.dump_xattrs) {
>> +                erofs_err("--xattrs conflicts with --no-xattrs");
>> +                return -EINVAL;
>> +            }
>
> No need this, the latter option will override the previous ones.
>
>> +            fsckcfg.dump_xattrs = true;
>> +            break;
>> +        case 14:
>> +            fsckcfg.dump_xattrs = false;
>> +            break;
>>           default:
>>               return -EINVAL;
>>           }
>> @@ -411,6 +427,57 @@ out:
>>       return ret;
>>   }
>>   +static int erofsfsck_dump_xattrs(struct erofs_inode *inode)
>> +{
>> +    char *keylst, *key;
>> +    ssize_t kllen;
>> +    int ret;
>> +
>> +    kllen = erofs_listxattr(inode, NULL, 0);
>> +    if (kllen <= 0)
>> +        return kllen;
>> +    keylst = malloc(kllen);
>> +    if (!keylst)
>> +        return -ENOMEM;
>> +    ret = erofs_listxattr(inode, keylst, kllen);
>> +    if (ret < 0)
>> +        goto out;
>> +    for (key = keylst; key < keylst + kllen; key += strlen(key) + 1) {
>> +        void *value = NULL;
>> +        size_t size = 0;
>> +
>> +        ret = erofs_getxattr(inode, key, NULL, 0);
>> +        if (ret < 0)
>> +            break;
>> +        if (ret) {
>> +            size = ret;
>> +            value = malloc(size);
>> +            if (!value) {
>> +                ret = -ENOMEM;
>
> I think error message is needed here.
>
>> +                break;
>> +            }
>> +            ret = erofs_getxattr(inode, key, value, size);
>> +            if (ret < 0) {
>
> Same here.
>
>> +                free(value);
>> +                break;
>> +            }
>> +            if (fsckcfg.extract_path) {
>> +                ret = setxattr(fsckcfg.extract_path, key, value,
>> +                           size, 0);
>> +                if (ret) {
>> +                    free(value);
>
> It seems that is unneeded.
>
>> +                    break;
>> +                }
>> +            } else
>> +                ret = 0;
>
> why not just
>             if (ret) {
>                 erofs_err(...);
>                 break;
>             }
>
> here?
>
Thanks for pointing this out. I will make the changes in the next version.

Thanks,

Hongzhen

>
>> +            free(value);
>> +        }
>> +    }
>> +out:
>> +    free(keylst);
>> +    return ret;
>> +}
>> +
>>   static int erofs_verify_inode_data(struct erofs_inode *inode, int 
>> outfd)
>>   {
>>       struct erofs_map_blocks map = {
>> @@ -909,6 +976,12 @@ static int erofsfsck_check_inode(erofs_nid_t 
>> pnid, erofs_nid_t nid)
>>       if (ret && ret != -ECANCELED)
>>           goto out;
>>   +    if (fsckcfg.dump_xattrs) {
>> +        ret = erofsfsck_dump_xattrs(&inode);
>> +        if (ret)
>> +            return ret;
>> +    }
>> +
>>       /* XXXX: the dir depth should be restricted in order to avoid 
>> loops */
>>       if (S_ISDIR(inode.i_mode)) {
>>           struct erofs_dir_context ctx = {
>> @@ -955,6 +1028,7 @@ int main(int argc, char *argv[])
>>       fsckcfg.overwrite = false;
>>       fsckcfg.preserve_owner = fsckcfg.superuser;
>>       fsckcfg.preserve_perms = fsckcfg.superuser;
>> +    fsckcfg.dump_xattrs= true;
>
> Missing space.
>
> Thanks,
> Gao Xiang
>
>>         err = erofsfsck_parse_options_cfg(argc, argv);
>>       if (err) {


More information about the Linux-erofs mailing list