linux-toradex.git/drivers/nvme, branch v5.12

nvmet-tcp: fix kmap leak when data digest in use

2021-03-18T04:39:18+00:00

When data digest is enabled we should unmap pdu iovec before handling
the data digest pdu.

Signed-off-by: Elad Grupi 
Reviewed-by: Sagi Grimberg 
Signed-off-by: Christoph Hellwig

nvmet: don't check iosqes,iocqes for discovery controllers

2021-03-18T04:39:02+00:00

From the base spec, Figure 78:

  "Controller Configuration, these fields are defined as parameters to
   configure an "I/O Controller (IOC)" and not to configure a "Discovery
   Controller (DC).

   ...
   If the controller does not support I/O queues, then this field shall
   be read-only with a value of 0h

Just perform this check for I/O controllers.

Fixes: a07b4970f464 ("nvmet: add a generic NVMe target")
Reported-by: Belanger, Martin 
Signed-off-by: Sagi Grimberg 
Reviewed-by: Chaitanya Kulkarni 
Signed-off-by: Christoph Hellwig

nvme-rdma: fix possible hang when failing to set io queues

2021-03-18T04:39:01+00:00

We only setup io queues for nvme controllers, and it makes absolutely no
sense to allow a controller (re)connect without any I/O queues.  If we
happen to fail setting the queue count for any reason, we should not allow
this to be a successful reconnect as I/O has no chance in going through.
Instead just fail and schedule another reconnect.

Reported-by: Chao Leng 
Fixes: 711023071960 ("nvme-rdma: add a NVMe over Fabrics RDMA host driver")
Signed-off-by: Sagi Grimberg 
Reviewed-by: Chao Leng 
Signed-off-by: Christoph Hellwig

nvme-tcp: fix possible hang when failing to set io queues

2021-03-18T04:39:01+00:00

We only setup io queues for nvme controllers, and it makes absolutely no
sense to allow a controller (re)connect without any I/O queues.  If we
happen to fail setting the queue count for any reason, we should not
allow this to be a successful reconnect as I/O has no chance in going
through. Instead just fail and schedule another reconnect.

Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver")
Signed-off-by: Sagi Grimberg 
Signed-off-by: Christoph Hellwig

nvme-tcp: fix misuse of __smp_processor_id with preemption enabled

2021-03-18T04:39:01+00:00

For our pure advisory use-case, we only rely on this call as a hint, so
fix the warning complaints of using the smp_processor_id variants with
preemption enabled.

Fixes: db5ad6b7f8cd ("nvme-tcp: try to send request in queue_rq context")
Fixes: ada831772188 ("nvme-tcp: Fix warning with CONFIG_DEBUG_PREEMPT")
Signed-off-by: Sagi Grimberg 
Reviewed-by: Chaitanya Kulkarni 
Tested-by: Yi Zhang 
Signed-off-by: Christoph Hellwig

nvme-tcp: fix a NULL deref when receiving a 0-length r2t PDU

2021-03-18T04:39:01+00:00

When the controller sends us a 0-length r2t PDU we should not attempt to
try to set up a h2cdata PDU but rather conclude that this is a buggy
controller (forward progress is not possible) and simply fail it
immediately.

Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver")
Reported-by: Belanger, Martin 
Signed-off-by: Sagi Grimberg 
Signed-off-by: Christoph Hellwig

nvme: fix Write Zeroes limitations

2021-03-18T04:38:49+00:00

We voluntarily limit the Write Zeroes sizes to the MDTS value provided by
the hardware, but currently get the units wrong, so fix that.

Fixes: 6e02318eaea5 ("nvme: add support for the Write Zeroes command")
Signed-off-by: Christoph Hellwig 
Reviewed-by: Keith Busch 
Tested-by: Klaus Jensen 
Reviewed-by: Klaus Jensen 
Reviewed-by: Chaitanya Kulkarni 
Reviewed-by: Himanshu Madhani

nvme: allocate the keep alive request using BLK_MQ_REQ_NOWAIT

2021-03-18T04:38:48+00:00

To avoid an error recovery deadlock where the keep alive work is waiting
for a request and thus can't be flushed to make progress for tearing down
the controller.  Also print the error code returned from
blk_mq_alloc_request to help debugging any future issues in this code.

Based on an earlier patch from Hannes Reinecke .

Signed-off-by: Christoph Hellwig 
Reviewed-by: Sagi Grimberg 
Reviewed-by: Chaitanya Kulkarni 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Daniel Wagner

nvme: merge nvme_keep_alive into nvme_keep_alive_work

2021-03-18T04:38:48+00:00

Merge nvme_keep_alive into its only caller to prepare for additional
changes to this code.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Sagi Grimberg 
Reviewed-by: Chaitanya Kulkarni 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Daniel Wagner

nvme-fabrics: only reserve a single tag

2021-03-18T04:38:48+00:00

Fabrics drivers currently reserve two tags on the admin queue.  But
given that the connect command is only run on a freshly created queue
or after all commands have been force aborted we only need to reserve
a single tag.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Sagi Grimberg 
Reviewed-by: Chaitanya Kulkarni 
Reviewed-by: Hannes Reinecke 
Reviewed-by: Daniel Wagner