Currently, if nvme_scan_ns_list fails, nvme_scan_work will fall back to
a sequential scan. nvme_scan_ns_list can fail for a variety of reasons,
e.g. a transient transport issue, and the resulting sequential scan can
be extremely expensive on controllers reporting an NN value close to the
maximum allowed (> 4 billion). Avoid sequential scans wherever possible
by only falling back to them in two cases:

- When the NVMe version supported (VS) value reported by the device is
  older than NVME_VS(1, 1, 0), before which support of Identify NS List
  not required.
- When the Identify NS List command fails with the DNR bit set in the
  status. This is to accommodate (non-compliant) devices which report a
  VS value which implies support for Identify NS List, but nevertheless
  do not support the command. Such devices will most likely fail the
  command with the DNR bit set.

The third case is when the device claims support for Identify NS List
but the command fails with DNR not set. In such cases, fallback to
sequential scan is potentially expensive and likely unnecessary, as a
retry of the list scan should succeed. So this change skips the fallback
in this third case.
Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

when starting error recovery there might be a authentication work
running, and it involves I/O commands. Given the controller is tearing
down there is no chance for the I/O to complete other than timing out
which may unnecessarily take a full io timeout.

So first tear down the queues, fail/cancel all inflight I/O (including
potentially authentication) and only then stop authentication. This
ensures that failover is not stalled due to blocked authentication I/O.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

when starting error recovery there might be a authentication work
running, and it involves I/O commands. Given the controller is tearing
down there is no chance for the I/O to complete other than timing out
which may unnecessarily take a full io timeout.

So first tear down the queues, fail/cancel all inflight I/O (including
potentially authentication) and only then stop authentication. This
ensures that failover is not stalled due to blocked authentication I/O.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

It triggered the queue authentication work elements in parallel, but
the ctrl authentication work itself completes when all of them
completes. Hence wait for queues auth completions.

This also makes nvme_auth_stop simply a sync cancel of ctrl
dhchap_auth_work.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>

only ctrl deletion calls nvme_auth_free, which was stopped prior in the
teardown stage, so there is no possibility that it should ever run when
nvme_auth_free is called. As a result, we can remove a local chap pointer
variable.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

We know exactly how many dhchap contexts we will need, there is no need
to hold a list that we need to protect with a mutex. Convert to
a dynamically allocated array. And dhchap_context access state is
maintained by the chap itself.

Make dhchap_auth_mutex protect only the ctrl host_key and ctrl_key
in a fine-grained lock such that there is no long lasting acquisition
of the lock and no need to take/release this lock when flushing
authentication works.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>

ctrl ctrl_key member may be overwritten from a sysfs context driven
by the user. Once a queue local copy was created, use that instead
to minimize checks on a shared resource.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Now that the chap context is reset upon completion, this is no longer
needed. Also remove nvme_auth_reset as no callers are left.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

These are now redundant as the dhchap context is
removed after authentication completes.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>

We don't want to keep authentication sensitive info in memory for unlimited
amount of time.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

We want to guarantee that we have chap buffers when a controller
reconnects under memory pressure. Add a mempool specifically
for that.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>

dhchap structure is per-queue, it is wasteful to keep it for the entire
lifetime of the queue. Allocate it dynamically and get rid of it after
authentication. We don't need kzalloc because all accessors are clearing
it before writing to it.

Also, remove redundant chap buf_size which is always 4096, use a define
instead.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

No one passes NVME_QID_ANY to nvme_auth_negotiate.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Replace ctrl ctrl_key/host_key only after nvme_auth_generate_key is successful.
Also, this fixes a bug where the keys are leaked.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

nvme_auth_generate_key can fail, don't ignore it upon initialization.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

host_response, host_key, ctrl_key and sess_key are
freed in nvme_auth_reset_dhchap which is called from
nvme_auth_free_dhchap.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

The connect sequence will re-authenticate.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Only the nvme module calls it.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Use nvme_ctrl_auth_work and nvme_queue_auth_work for better
readability.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>

nvme_auth_[reset|free] operate on the controller while
__nvme_auth_[reset|free] operate on a chap struct (which maps to a queue
context). Rename it for clarity.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Unbind a device driver when a reset fails is very unusual behavior.
Just shut the controller down and leave it in dead state if we fail
to reset it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

nvme_reset_work is a little fragile as it needs to handle both resetting
a live controller and initializing one during probe.  Split out the initial
probe and open code it in nvme_probe and leave nvme_reset_work to just do
the live controller reset.

This fixes a recently introduced bug where nvme_dev_disable causes a NULL
pointer dereferences in blk_mq_quiesce_tagset because the tagset pointer
is not set when the reset state is entered directly from the new state.
The separate probe code can skip the reset state and probe directly and
fixes this.

To make sure the system isn't single threaded on enabling nvme
controllers, set the PROBE_PREFER_ASYNCHRONOUS flag in the device_driver
structure so that the driver core probes in parallel.

Fixes: 98d81f0d ("nvme: use blk_mq_[un]quiesce_tagset")
Reported-by: Gerd Bayer <gbayer@linux.ibm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Check that a HMB is wanted into the allocation helper instead of the
caller.  This makes life simpler for an upcoming second caller.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

Move the OACS check and the error checking into nvme_dbbuf_dma_alloc so
that an upcoming second caller doesn't have to duplicate this boilerplate
code.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

nvme_pci_configure_admin_queue is called right after nvme_pci_enable, and
it's work is undone by nvme_dev_disable.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Move setting of low-level constant parameters from nvme_reset_work to
nvme_pci_alloc_ctrl.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Add a helper that allocates the nvme_dev structure up to the point where
we can call nvme_init_ctrl.  This pairs with the free_ctrl method and can
thus be used to cleanup the teardown path and make it more symmetric.

Note that this now calls nvme_init_ctrl a lot earlier during probing,
which also means the per-controller character device shows up earlier.
Due to the controller state no commnds can be send on it, but it might
make sense to delay the cdev registration until nvme_init_ctrl_finish.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Add a helper to create the iod mempool.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

nvme_dbbuf_dma_free frees dma coherent memory, so it must not be called
after ->remove has returned.  Fortunately there is no way to use it
after shutdown as no more I/O is possible so it can be moved.  Similarly
the iod_mempool can't be used for a device kept alive after shutdown, so
move it next to freeing the PRP pools.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Once the controller is shutdown no one can access the admin queue.  Tear
it down in nvme_dev_remove_admin, which matches the flow in the other
drivers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Allow the transport driver to override the attribute groups for the
control device, so that the PCIe driver doesn't manually have to add a
group after device creation and keep track of it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

Nothing about the TCG Opal support is PCIe transport specific, so move it
to the core code.  For this nvme_init_ctrl_finish grows a new
was_suspended argument that allows the transport driver to tell the OPAL
code if the controller came out of a suspend cycle.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: James Smart <jsmart2021@gmail.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

nvme_passthrough_end can race with a reset, which can lead to
racing stores to the cels xarray as well as further shengians
with upcoming more complicated initialization.

So drop the call and just log that the controller capabilities
might have changed and a reset could be required to use the new
controller capabilities.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>

While the specification allows devices to either deallocate data
or to actually write zeroes on any Write Zeroes command, many SSDs
only do the sensible thing and deallocate data when the DEAC bit
is specific.  Set it when it is supported and the caller doesn't
explicitly opt out of deallocation.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>

Allow all identify-namespace variants (CNS 00h, 05h and 08h) without
requiring CAP_SYS_ADMIN. The information (retrieved using id-ns) is
needed to form IO commands for passthrough interface.
Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Currently both io and admin commands are kept under a
coarse-granular CAP_SYS_ADMIN check, disregarding file mode completely.

$ ls -l /dev/ng*
crw-rw-rw- 1 root root 242, 0 Sep  9 19:20 /dev/ng0n1
crw------- 1 root root 242, 1 Sep  9 19:20 /dev/ng0n2

In the example above, ng0n1 appears as if it may allow unprivileged
read/write operation but it does not and behaves same as ng0n2.

This patch implements a shift from CAP_SYS_ADMIN to more fine-granular
control for io-commands.
If CAP_SYS_ADMIN is present, nothing else is checked as before.
Otherwise, following rules are in place
- any admin-cmd is not allowed
- vendor-specific and fabric commmand are not allowed
- io-commands that can write are allowed if matching FMODE_WRITE
permission is present
- io-commands that read are allowed

Add a helper nvme_cmd_allowed that implements above policy.
Change all the callers of CAP_SYS_ADMIN to go through nvme_cmd_allowed
for any decision making.
Since file open mode is counted for any approval/denial, change at
various places to keep file-mode information handy.
Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

sizeof( struct nvmefc_ls_rcv_op ) = 64
sizeof( union nvmefc_ls_requests ) = 1024
sizeof( union nvmefc_ls_responses ) = 128

So, in nvme_fc_rcv_ls_req(), 1216 bytes of memory are requested when
kzalloc() is called.

Because of the way memory allocations are performed, 2048 bytes are
allocated. So about 800 bytes are wasted for each request.

Switch to 3 distinct memory allocations, in order to:
   - save these 800 bytes
   - avoid zeroing this extra memory
   - make sure that memory is properly aligned in case of DMA access
    ("fc_dma_map_single(lsop->rspbuf)" just a few lines below)
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

There is no need to have a separate slab cache for each namespace,
and having separate ones creates duplicate debugs file names as well.

Fixes: d5eff33e ("nvmet: add simple file backed ns support")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>

In order to test queue number changes we need to make sure that the
host reconnects. Because only when the host disconnects from the
target the number of queues are allowed to change according the spec.

The initial idea was to disable and re-enable the ports and have the
host wait until the KATO timer expires, triggering error
recovery. Though the host would see a DNR reply when trying to
reconnect. Because of the DNR bit the connection is dropped
completely. There is no point in trying to reconnect with the same
parameters according the spec.

We can force to reconnect the host is by deleting all controllers. The
host will observe any newly posted request to fail and thus starts the
error recovery but this time without the DNR bit set.
Signed-off-by: Daniel Wagner <dwagner@suse.de>
Reviewed-by: Chaitanya Kulkarni  <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Use try_cmpxchg instead of cmpxchg (*ptr, old, new) == old in
nvmet_update_sq_head.  x86 CMPXCHG instruction returns success in ZF flag, so
this change saves a compare after cmpxchg (and related move instruction in
front of cmpxchg).

Also, try_cmpxchg implicitly assigns old *ptr value to "old" when cmpxchg
fails. There is no need to re-read the value in the loop.

Note that the value from *ptr should be read using READ_ONCE to prevent
the compiler from merging, refetching or reordering the read.

No functional change intended.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>