linux-toradex.git/drivers/nvme, branch v4.15

nvme-pci: take sglist coalescing in dma_map_sg into account

2018-01-17T21:05:35+00:00

Some iommu implementations can merge physically and/or virtually
contiguous segments inside sg_map_dma.  The NVMe SGL support does not take
this into account and will warn because of falling off a loop.  Pass the
number of mapped segments to nvme_pci_setup_sgls so that the SGL setup
can take the number of mapped segments into account.

Reported-by: Fangjian (Turing) 
Fixes: a7a7cbe3 ("nvme-pci: add SGL support")
Signed-off-by: Christoph Hellwig 
Reviewed-by: Keith Busch 
Reviewed-by: Sagi Grimberg 
Signed-off-by: Jens Axboe

nvme-pci: check segement valid for SGL use

2018-01-17T21:05:33+00:00

The driver needs to verify there is a payload with a command before
seeing if it should use SGLs to map it.

Fixes: 955b1b5a00ba ("nvme-pci: move use_sgl initialization to nvme_init_iod()")
Reported-by: Paul Menzel 
Reviewed-by: Paul Menzel 
Signed-off-by: Keith Busch 
Signed-off-by: Christoph Hellwig 
Signed-off-by: Jens Axboe

nvme-fabrics: initialize default host->id in nvmf_host_default()

2018-01-08T09:52:03+00:00

The field was uninitialized before use.

Signed-off-by: Ewan D. Milne 
Signed-off-by: Christoph Hellwig

nvme-fcloop: avoid possible uninitialized variable warning

2017-12-29T09:37:21+00:00

The kbuild test robot send mail of a potential use of an uninitialized
variable - "tport" in fcloop_delete_targetport() which then calls
__targetport_unreg() which uses the variable. It will never be the
case it is uninitialized as the call to __targetport_unreg() only
occurs if there is a valid nport pointer. And at the time the nport
pointer is assigned, the tport variable is set.

Remove the warning by assigning a NULL value initially.

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

nvme-mpath: fix last path removal during traffic

2017-12-29T09:32:58+00:00

In case our last path is removed during traffic, we can end up requeueing
the bio(s) but never schedule the actual requeue work as upper layers
still have open handles on the mpath device node.

Fix this by scheduling requeue work if the namespace being removed is
the last path in the ns_head path list.

Fixes: 32acab3181c7 ("nvme: implement multipath access to nvme subsystems")
Signed-off-by: Sagi Grimberg 
Signed-off-by: Christoph Hellwig

nvme-rdma: fix concurrent reset and reconnect

2017-12-29T09:32:58+00:00

Now ctrl state machine allows to transition from RESETTING to
RECONNECTING.  In nvme-rdma when we receive a rdma cm DISONNECTED event,
we trigger nvme_rdma_error_recovery. This happens also when we execute a
controller reset, issue a cm diconnect request and receive a cm
disconnect reply, as a result, the reset work and the error recovery work
can run concurrently.

Until now the state machine prevented from the error recovery work from
running as a result of a controller reset (RESETTING -> RECONNECTING was
not allowed).

To fix this, we adopt the FC state machine approach, we always transition
from LIVE to RESETTING and only then to RECONNECTING.  We do this both
for the error recovery work and the controller reset work:

 1. transition to RESETTING
 2. teardown the controller association
 3. transition to RECONNECTING

This will restore the protection against reset work and error recovery work
from concurrently running together.

Fixes: 3cec7f9de448 ("nvme: allow controller RESETTING to RECONNECTING transition")
Signed-off-by: Sagi Grimberg 
Signed-off-by: Christoph Hellwig

nvme: fix sector units when going between formats

2017-12-29T09:31:05+00:00

If you format a device with a 4k sector size back to 512 bytes, the queue
limit values for physical block size and minimum IO size were not getting
updated; only the logical block size was being updated.  This patch adds
code to update the physical block and IO minimum sizes.

Signed-off-by: Jeff Lien 
Reviewed-by: Martin K. Petersen 
Signed-off-by: Christoph Hellwig

nvme-pci: move use_sgl initialization to nvme_init_iod()

2017-12-29T09:31:04+00:00

A flag "use_sgl" of "struct nvme_iod" has been used in nvme_init_iod()
without being set to any value. It seems like "use_sgl" has been set
in either nvme_pci_setup_prps() or nvme_pci_setup_sgls() which occur
later than nvme_init_iod().

Make "iod->use_sgl" being set in a proper place, nvme_init_iod().
Also move nvme_pci_use_sgls() up above nvme_init_iod() to make it
possible to be called by nvme_init_iod().

Signed-off-by: Minwoo Im 
Reviewed-by: Sagi Grimberg 
Signed-off-by: Christoph Hellwig

nvme: setup streams after initializing namespace head

2017-12-15T14:18:07+00:00

Fixes a NULL pointer dereference.

Reported-by: Arnav Dawn 
Signed-off-by: Keith Busch 
Reviewed-by: Martin K. Petersen 
Signed-off-by: Christoph Hellwig

nvme: check hw sectors before setting chunk sectors

2017-12-15T14:18:07+00:00

Some devices with IDs matching the "stripe" quirk don't actually have
this quirk, and don't have an MDTS value. When MDTS is not set, the
driver sets the max sectors to UINT_MAX, which is not a power of 2,
hitting a BUG_ON from blk_queue_chunk_sectors. This patch skips setting
chunk sectors for such devices.

Signed-off-by: Keith Busch 
Reviewed-by: Martin K. Petersen 
Signed-off-by: Christoph Hellwig