linux.git/drivers/nvme/host/fc.c, branch v4.12

nvme-fc: fix missing put reference on controller create failure

2017-06-07T09:08:53+00:00

The failure case, of a create controller request, called
nvme_uninit_ctrl() but didn't do a put to allow the nvme
controller to be deleted.

Signed-off-by: James Smart 
Signed-off-by: Christoph Hellwig

nvme-fc: on lldd/transport io error, terminate association

2017-06-07T09:08:52+00:00

Per FC-NVME, when lldd or transport detects an i/o error, the
connection must be terminated, which in turn requires the association
to be termianted.  Currently the transport simply creates a nvme
completion status of transport error and returns the io. The FC-NVME
spec makes the mandate as initiator and host, depending on the error,
can get out of sync on outstanding io counts (sqhd/sqtail).

Implement the association teardown on lldd or transport detected
errors.

Signed-off-by: James Smart 
Signed-off-by: Christoph Hellwig 
Reviewed-by: Sagi Grimberg

nvme: replace is_flags field in nvme_ctrl_ops with a flags field

2017-05-26T06:54:16+00:00

So that we can have more flags for transport-specific behavior.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Keith Busch

nvme_fc: remove extra controller reference taken on reconnect

2017-05-22T18:55:29+00:00

fix extra controller reference taken on reconnect by moving
reference to initial controller create

Signed-off-by: James Smart 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Johannes Thumshirn 
Signed-off-by: Christoph Hellwig

nvme_fc: correct nvme status set on abort

2017-05-22T18:55:29+00:00

correct nvme status set on abort. Patch that changed status to being actual
nvme status crossed in the night with the patch that added abort values.

Signed-off-by: James Smart 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Johannes Thumshirn 
Signed-off-by: Christoph Hellwig

nvme_fc: set logging level on resets/deletes

2017-05-22T18:55:29+00:00

Per the review by Sagi on:
http://lists.infradead.org/pipermail/linux-nvme/2017-April/009261.html

Looked at existing warn vs info vs err dev_xxx levels for the messages
printed on reconnects and deletes:
- Resets due to error and resets transitioned to deletes are dev_warn
- Other reset/disconnect messages are dev_info
- Removed chatty io queue related messages

Signed-off-by: James Smart 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Johannes Thumshirn 
Signed-off-by: Christoph Hellwig

nvme_fc: revise comment on teardown

2017-05-22T18:55:29+00:00

Per the recommendation by Sagi on:
http://lists.infradead.org/pipermail/linux-nvme/2017-April/009261.html

An extra reference was pointed out.  There's no issue with the
references, but rather a literal interpretation of what the comment
is saying.

Reword the comment to avoid confusion.

Signed-off-by: James Smart 
Reviewed-by: Johannes Thumshirn 
Reviewed-by: Hannes Reinecke 
Signed-off-by: Christoph Hellwig

nvme_fc: Support ctrl_loss_tmo

2017-05-22T18:55:29+00:00

Sync with Sagi's recent addition of ctrl_loss_tmo in the core fabrics
layer.

Remove local connect limits and connect_attempts variable.
Use fabrics new nr_connects variable and use of nvmf_should_reconnect()
Refactor duplicate reconnect failure code.

Addresses review comment by Sagi on controller reset support:
http://lists.infradead.org/pipermail/linux-nvme/2017-April/009261.html

Signed-off-by: James Smart 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Johannes Thumshirn 
Signed-off-by: Christoph Hellwig

nvme_fc: get rid of local reconnect_delay

2017-05-22T18:55:29+00:00

Remove the local copy of reconnect_delay.
Use the value in the controller options directly.

Signed-off-by: James Smart 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Johannes Thumshirn 
Signed-off-by: Christoph Hellwig

nvme-fc: stop queues on error detection

2017-05-20T16:11:34+00:00

Per the recommendation by Sagi on:
http://lists.infradead.org/pipermail/linux-nvme/2017-April/009261.html

Rather than waiting for reset work thread to stop queues and abort the ios,
immediately stop the queues on error detection. Reset thread will restop
the queues (as it's called on other paths), but it does not appear to have
a side effect.

Signed-off-by: James Smart 
Reviewed-by: Johannes Thumshirn 
Signed-off-by: Christoph Hellwig 
Signed-off-by: Jens Axboe