[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[dm-devel] Re: [PATCH] Handle multipath paths in a path group properly during pg_init



Chandra Seetharaman wrote:
Resending the patch to get in patchwork...
-------
The problem reported by Moger Babu was caused due to the architectural
change made when we moved from dm hardware handler to SCSI hardware
handler.

Thanks Babu for finding and reporting the bug.

All of the hardware handlers, do have a state now, and they are set to
active and (some form of) inactive. All of them have prep_fn, which use
this "state" to fail the I/O without it ever being sent to the device.

As Babu has noted in his email, the pg_init/activate is sent on only one
path and the "state" of that path is changed appropriately to "active"
while other paths in the same path group are never changed as they never
got an "activate".

Attached is a patch (compiled, tested, but not clean yet), which makes
changes in the dm-multipath layer to send an "activate" on each paths in
the path groups.

Mike (Anderson) and I had a discussion about whether to implement this
in the dm-mulitpath layer or in the SCSI hardware handler layer and we
came to a conclusion that it is best suited to be in the dm-mulitpath
layer as it is the one that knows the relationship between different
paths.

If it were to be done at the Hardware handler layer, then the hardware
handler may end up having a different notion of the path relationship
and hence may not work as expected by the dm-multipath layer.

This patch has been tested by Hannes in EMC storage. Babu and I tested it
in LSI storage.
----------

Signed-off-by: Chandra Seetharaman <sekharan us ibm com>

Acked-by: Hannes Reinecke <hare suse de>

Cheers,

Hannes
--
Dr. Hannes Reinecke		      zSeries & Storage
hare suse de			      +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Markus Rex, HRB 16746 (AG Nürnberg)


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]