Re: [PATCH v2] libxl: Reset toolstack_save file position in libxl

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 21/05/14 17:04, Ian Jackson wrote:
> Andrew Cooper writes ("Re: [PATCH v2] libxl: Reset toolstack_save file position in libxl"):
>> On 19/05/2014 19:36, Jason Andryuk wrote:
>>> toolstack_save data is written to a temporary file in libxl and read
>>> back in libxl-save-helper.  The file position must be reset prior to
>>> reading the file, which is done in libxl-save-helper with lseek.
>>>
>>> lseek is unsupported for pipes and sockets, so a wrapper passing such an
>>> fd to libxl-save-helper fails the lseek.  Moving the lseek to libxl
>>> avoids the error, allowing the save to continue.
>>>
>>> Signed-off-by: Jason Andryuk <andryuk@aero.org>
>>> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
>> Ian: As noted previously, this should be a backport candidate.
> We normally do backports only for bugs (and not even all bugs).  I'm
> afraid that I don't see how the current arrangements are a bug.
>
> See my other mail.  Perhaps the answer to that will explain to me how
> the current are a bug, rather than simply less convenient and flexible
> than they could be.
>
> Thanks,
> Ian.

It is a layering violation.  libxl-save-helper cannot assume that the fd
is a file, so shouldn't seek on it.

libxl however should prepare a "socket-like" fd to the save-helper,
which means rewinding the file itself.


As indicated, "toolstack_data" is heading very much in the direction of
/dev/null with the migration v2 patches, as it itself is a gross
layering violation. If you don't deem this worthy for backport, its not
the end of the world.

~Andrew