On Mon, Mar 26, 2018 at 10:16:54 +0200, Michal Privoznik wrote:
On 03/26/2018 09:49 AM, Peter Krempa wrote:
> On Mon, Mar 26, 2018 at 07:16:42 +0200, Michal Privoznik wrote:
>> Unless overridden, libdevmapper logs all messages to stderr.
>> Therefore if something goes wrong in storage_backend_mpath.c or
>> parthelper.c we don't honour user set logging targets.
>>
>> Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
>> ---
>> src/util/virlog.c | 23 +++++++++++++++++++++++
>> 1 file changed, 23 insertions(+)
>
> I think this should be done in the function which registers the mpath
> storage backend. For now it should work as is, but given that we are
> going to split the daemons this code will not really be necessary in
> other places.
>
> And it really does not have to do much with our logging code.
>
Well, in the next patch I'm introducing virFileGetMPathTargets() which
uses dm_* a lot. Now, since the function lives in src/util/ it can be
used by many drivers. Do we expect each one of them to provide dummy
callback on their own? I don't think so. I agree this is not a nice
patch, but I don't really see any better place for it.
I've written another response to my original reply which explains a bit
more.
Basically, the code in the next patch should be put in a new file along
with a virOnce function to silence the logs if you don't want to move it
into the storage driver directly.