samba3.smb2.ioctl

[David Disseldorp]

Hi Andrew,

I've raised https://bugzilla.samba.org/show_bug.cgi?id=11176 to track
this.

On Fri, 20 Mar 2015 18:22:35 +1300, Andrew Bartlett wrote:

> Can someone help explain to me what this test failure means?  I get it
> in the Catalyst Cloud when running make test on git master, building on
> Ubuntu 12.04.  We only see it there, and not on 14.04.  We are backed by
> Ceph RBD volumes, if that helps. 

Which filesystem are you using, and on which architecture?

> Thanks,
> 
> Andrew Bartlett
> 
> [1(0)/3 at 0s] samba3.smb2.ioctl fs_specific(nt4_dc)
> UNEXPECTED(failure): samba3.smb2.ioctl fs_specific.sparse_punch(nt4_dc)
> REASON: Exception: Exception: ../source4/torture/smb2/ioctl.c:3308:
> far_count was 1 (0x1), expected 0 (0x0): unexpected response len

samba3.smb2.ioctl.sparse_punch creates a 4K file, and then hole-punches
the full range via FSCTL_SET_ZERO_DATA / FALLOC_FL_PUNCH_HOLE. This
failure indicates that data is still present (via SEEK_DATA) after the
hole-punch.

> UNEXPECTED(failure): samba3.smb2.ioctl
> fs_specific.sparse_qar_ob1(nt4_dc)
> REASON: Exception: Exception: ../source4/torture/smb2/ioctl.c:4477:
> far_rsp[0].len was 65537 (0x10001), expected 65536 (0x10000): unexpected
> far len

samba3.smb2.ioctl.sparse_qar_ob1 creates a 128K file, and then
hole-punches the latter 64K. This failure indicates that data is still
present after the hole punch.

These tests are currently quite FS specific. I'll address this by having
them check for deallocation-on-hole-punch behaviour and skipping if not
present.

Cheers, David