On 8/15/16 11:38 AM, Zorro Lang wrote:
> From the result of strace, we can see:
> quotactl(0x5808 /* Q_??? */|USRQUOTA, "/dev/mapper/testvg-scratchdev", 0,
> 0x7ffdbbda94c0) = -1 EINVAL (Invalid argument)
> quotactl(Q_XGETQSTAT|USRQUOTA, "/dev/mapper/testvg-scratchdev", 0,
> {version=1, ...}) = 0
>
> So the first call for XFS_GETQSTATV return EINVAL.
>
> In linux kernel quota_getxstatev() function, the logic is:
> /* If this kernel doesn't support user specified version, fail */
> switch (fqs.qs_version) {
> case FS_QSTATV_VERSION1:
> break;
> default:
> return -EINVAL;
> }
>
> So we need to set qa_version to FS_QSTATV_VERSION1, before
> we call XFS_GETQSTATV. And as we talked on IRC, it test passed if I set
>
> sv.qs_version = 1;
>
> at here.
>
> So maybe a V2 patch is needed to fix this problem:)
Yep, kernel commit said:
| Callers of the new quotactl command have to set the version of the data
| structure being passed, and kernel will fill as much data as requested.
| If the kernel does not support the user-space provided version, EINVAL
| will be returned. User-space can reduce the version number and call the same
| quotactl again.
but this isn't documented anywhere else, AFAICT; this call still isn't
documented in the quotactl manpage. o_O
I'll fix it up, thanks for catching it!
-Eric
|