A crash happens when trying to disconnect a reconnecting ctrl:

 1) The network was cut off when the connection was just established,
    scan work hang there waiting for some IOs complete.  Those I/Os were
    retried because we return BLK_STS_RESOURCE to blk in reconnecting.
 2) After a while, I tried to disconnect this connection.  This
    procedure also hangs because it tried to obtain ctrl->scan_lock.
    It should be noted that now we have switched the controller state
    to NVME_CTRL_DELETING.
 3) In nvme_check_ready(), we always return true when ctrl->state is
    NVME_CTRL_DELETING, so those retrying I/Os were issued to the bottom
    device which was already freed.

To fix this, when ctrl->state is NVME_CTRL_DELETING, issue cmd to bottom
device only when queue state is live.  If not, return host path error to
the block layer
Signed-off-by: NRuozhu Li <liruozhu@huawei.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

With unique discovery controller NQNs we cannot distinguish the
subsystem type by the NQN alone, but need to check the subsystem
type, too.
So expose the subsystem type in a new sysfs attribute 'subsystype'.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Reviewed-by: NChaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The current blk_mq_quiesce_queue() and blk_mq_unquiesce_queue() always
stops and starts the queue unconditionally. And there can be concurrent
quiesce/unquiesce coming from different unrelated code paths, so
unquiesce may come unexpectedly and start queue too early.

Prepare for supporting concurrent quiesce/unquiesce from multiple
contexts, so that we can address the above issue.

NVMe has very complicated quiesce/unquiesce use pattern, add one atomic
bit for makeiing sure that blk-mq quiece/unquiesce is always called in
pair.
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211014081710.1871747-5-ming.lei@redhat.comSigned-off-by: NJens Axboe <axboe@kernel.dk>

Add two APIs for stopping and starting admin queue.
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20211014081710.1871747-2-ming.lei@redhat.comSigned-off-by: NJens Axboe <axboe@kernel.dk>

Take advantage of struct io_comp_batch, if passed in to the nvme poll
handler. If it's set, rather than complete each request individually
inline, store them in the io_comp_batch list. We only do so for requests
that will complete successfully, anything else will be completed inline as
before.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Some apple controllers use the command id as an index to implementation
specific data structures and will fail if the value is out of bounds.
The nvme driver's recently introduced command sequence number breaks
this controller.

Provide a quirk so these spec incompliant controllers can function as
before. The driver will not have the ability to detect bad completions
when this quirk is used, but we weren't previously checking this anyway.

The quirk bit was selected so that it can readily apply to stable.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=214509
Cc: Sven Peter <sven@svenpeter.dev>
Reported-by: NOrlando Chamberlain <redecorating@protonmail.com>
Reported-by: NAditya Garg <gargaditya08@live.com>
Signed-off-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Tested-by: NSven Peter <sven@svenpeter.dev>
Link: https://lore.kernel.org/r/20210927154306.387437-1-kbusch@kernel.orgSigned-off-by: NJens Axboe <axboe@kernel.dk>

Preparatory patch in order to reuse nvme_multi_css in the nvme target
code.
Signed-off-by: NAdam Manzanares <a.manzanares@samsung.com>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

When triggering a rescan due to a namespace resize we will be
receiving AENs on every controller, triggering a rescan of all
attached namespaces. If multipath is active only the current path and
the ns_head disk will be updated, the other paths will still refer to
the old size until AENs for the remaining controllers are received.

If I/O comes in before that it might be routed to one of the old
paths, triggering an I/O failure with 'access beyond end of device'.
With this patch the old paths are skipped from multipath path
selection until the controller serving these paths has been rescanned.
Signed-off-by: NHannes Reinecke <hare@suse.de>
[dwagner: - introduce NVME_NS_READY flag instead of NVME_NS_INVALIDATE
          - use 'revalidate' instead of 'invalidate' which
	    follows the zoned device code path.
	  - clear NVME_NS_READY before clearing current_path]
Signed-off-by: NDaniel Wagner <dwagner@suse.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

These values are unused now that the lightnvm support is gone.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>

Now that the lightnvm driver is removed, we don't need a pointer to it's
now non-existent struct.
Signed-off-by: NKeith Busch <kbusch@kernel.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

We cannot detect a (perhaps buggy) controller that is sending us
a completion for a request that was already completed (for example
sending a completion twice), this phenomenon was seen in the wild
a few times.

So to protect against this, we use the upper 4 msbits of the nvme sqe
command_id to use as a 4-bit generation counter and verify it matches
the existing request generation that is incrementing on every execution.

The 16-bit command_id structure now is constructed by:
| xxxx | xxxxxxxxxxxx |
  gen    request tag

This means that we are giving up some possible queue depth as 12 bits
allow for a maximum queue depth of 4095 instead of 65536, however we
never create such long queues anyways so no real harm done.
Suggested-by: NKeith Busch <kbusch@kernel.org>
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Acked-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Reviewed-by: NDaniel Wagner <dwagner@suse.de>
Tested-by: NDaniel Wagner <dwagner@suse.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Lightnvm supports the OCSSD 1.x and 2.0 specs which were early attempts
to produce Open Channel SSDs and never made it into the NVMe spec
proper.  They have since been superceeded by NVMe enhancements such
as ZNS support.  Remove the support per the deprecation schedule.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210812132308.38486-1-hch@lst.deReviewed-by: NMatias Bjørling <mb@lightnvm.io>
Reviewed-by: NJavier González <javier@javigon.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

When the last path to a ns_head drops the current code
removes the ns_head from the subsystem list, but will only
delete the disk itself if the last reference to the ns_head
drops. This is causing an refcounting imbalance eg when
applications have a reference to the disk, as then they'll
never get notified that the disk is in fact dead.
This patch moves the call 'del_gendisk' into nvme_mpath_check_last_path(),
ensuring that the disk can be properly removed and applications get the
appropriate notifications.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

We don't have an nvme status to report if the driver's .queue_rq()
returns an error without dispatching the requested nvme command. Check
the return value from blk_execute_rq() for all passthrough commands so
the caller may know their command was not successful.

If the command is from the target passthrough interface and fails to
dispatch, synthesize the response back to the host as a internal target
error.
Signed-off-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Link: https://lore.kernel.org/r/20210610214437.641245-5-kbusch@kernel.orgSigned-off-by: NJens Axboe <axboe@kernel.dk>

The generic blk_execute_rq() knows how to handle polled completions. Use
that instead of implementing an nvme specific handler.
Signed-off-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Link: https://lore.kernel.org/r/20210610214437.641245-3-kbusch@kernel.orgSigned-off-by: NJens Axboe <axboe@kernel.dk>

For various transports such as fc/tcp/pci it is common to check if
NVMe SGLs are supported or not by the controller.

In this preparation patch we add a helper to avoid the open coding of
such checks in the various transport.
Signed-off-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Now that only one caller is left remove the helpers by restructuring
nvme_pr_command so that it has two helpers for sending a command of to a
given nsid using either the ns_head for multipath, or the namespace
stored in the gendisk.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>

Split multipath support out of nvme_report_zones into a separate helper
and simplify the non-multipath version as a result.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>

nvme_init_identify and thus nvme_mpath_init can be called multiple
times and thus must not overwrite potentially initialized or in-use
fields.  Split out a helper for the basic initialization when the
controller is initialized and make sure the init_identify path does
not blindly change in-use data structures.

Fixes: 0d0b660f ("nvme: add ANA support")
Reported-by: NMartin Wilck <mwilck@suse.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NHannes Reinecke <hare@suse.de>

queue_rq() in pci only checks if the dispatched queue (nvmeq) is ready,
e.g. not being suspended. Since nvme_alloc_admin_tags() in reset flow
restarts the admin queue, users are able to submit admin commands to a
controller before reset_work() completes. Commands submitted under this
condition may interfere with commands that performs identify, IO queue
setup in reset_work(), and may result in a hang described in the
following patch.

As seen in the fabrics, user commands are prevented from being executed
under inproper controller states. We may reuse this logic to maintain a
clear admin queue during reset_work().
Signed-off-by: NTao Chiu <taochiu@synology.com>
Signed-off-by: NCody Wong <codywong@synology.com>
Reviewed-by: NLeon Chien <leonchien@synology.com>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

In multipath case, we should consider namespace attachment with
controllers in a subsystem when we find out the live controller for the
namespace.  This patch manually reverted the commit 3557a440
("nvme: don't bother to look up a namespace for controller ioctls") with
few more updates to nvme_ns_head_chr_ioctl which has been newly updated.

Fixes: 3557a440 ("nvme: don't bother to look up a namespace for
controller ioctls")
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: NMinwoo Im <minwoo.im.dev@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Userspace has not been allowed to I/O to device that's failed to
be initialized.  This patch introduces generic per-namespace character
device to allow userspace to I/O regardless the block device is there or
not.

The chardev naming convention will similar to the existing blkdev naming,
using a ng prefix instead of nvme, i.e.

	- /dev/ngXnY

It also supports multipath which means it will not expose chardev for the
hidden namespace blkdevs (e.g., nvmeXcYnZ).  If /dev/ngXnY is created for
a ns_head, then I/O request will be routed to a specific controller
selected by the iopolicy of the subsystem.
Signed-off-by: NMinwoo Im <minwoo.im.dev@gmail.com>
Signed-off-by: NJavier González <javier.gonz@samsung.com>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Tested-by: NKanchan Joshi <joshi.k@samsung.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

According to the NVMe base spec the KATO commands should be sent
at half of the KATO interval, to properly account for round-trip
times.
As we now will only ever send one KATO command per connection we
can easily use the recommended values.
This also fixes a potential issue where the request timeout for
the KATO command does not match the value in the connect command,
which might be causing spurious connection drops from the target.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Move the multipath block_device_operations to multipath.c, where they
belong.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NJavier González <javier.gonz@samsung.com>
Reviewed-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>

Split out the ioctl code from core.c into a new file.  Also update
copyrights while we're at it.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NJavier González <javier.gonz@samsung.com>
Reviewed-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>

Pass the proper user pointer instead of the not all that useful integer
representation.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NJavier González <javier.gonz@samsung.com>

Return false from nvme_set_disk_name and let the caller set the
non-multipath name instead of duplicating the naming information in two
places.  Also remove the pointless local variables for the disk name
and flags and the not needed ctrl argument.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NJavier González <javier.gonz@samsung.com>

Move the multipath gendisk out of #ifdef CONFIG_NVME_MULTIPATH and add
a new nvme_ns_head_multipath that uses it to check if a ns_head has
a multipath device associated with it.
Signed-off-by: NMinwoo Im <minwoo.im.dev@gmail.com>
[hch: added the IS_ENABLED, converted a few existing users]
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NJavier González <javier.gonz@samsung.com>
Reviewed-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>

Commands that access LBA contents without a data transfer between the
host historically have not had a spec defined upper limit. The driver
set the queue constraints for such commands to the max data transfer
size just to be safe, but this artificial constraint frequently limits
devices below their capabilities.

The NVMe Workgroup ratified TP4040 defines how a controller may
advertise their non-MDTS limits. Use these if provided and default to
the current constraints if not. Since the Dataset Management command
limits are defined in logical blocks, but without a namespace to tell us
the logical block size, the code defaults to the safe 512b size.
Signed-off-by: NKeith Busch <kbusch@kernel.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

All nvme transport drivers preallocate an nvme command for each request.
Assume to use that command for nvme_setup_cmd() instead of requiring
drivers pass a pointer to it. All nvme drivers must initialize the
generic nvme_request 'cmd' to point to the transport's preallocated
nvme_command.

The generic nvme_request cmd pointer had previously been used only as a
temporary copy for passthrough commands. Since it now points to the
command that gets dispatched, passthrough commands must directly set it
up prior to executing the request.
Signed-off-by: NKeith Busch <kbusch@kernel.org>
Reviewed-by: NJens Axboe <axboe@kernel.dk>
Reviewed-by: NHimanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

This is a prep patch so that we can move the identify data structure
related code initialization from nvme_init_identify() into a helper.

Rename the function nvmet_init_identify() to nvmet_init_ctrl_finish().

Next patch will move the nvme_id_ctrl related initialization from newly
renamed function nvme_init_ctrl_finish() into the nvme_init_identify()
helper.
Signed-off-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Use the proper macro instead of hard-coded value.
Signed-off-by: NKanchan Joshi <joshi.k@samsung.com>
Reviewed-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Some Amazon NVMe controllers do not follow the NVMe specification
and are limited to 48-bit DMA addresses.  Add a quirk to force
bounce buffering if needed and limit the IOVA allocation for these
devices.

This affects all current Amazon NVMe controllers that expose EBS
volumes (0x0061, 0x0065, 0x8061) and local instance storage
(0xcd00, 0xcd01, 0xcd02).
Signed-off-by: NFilippo Sironi <sironi@amazon.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The original design to use device-managed resource allocation
doesn't really work as the NVMe controller has a vastly different
lifetime than the hwmon sysfs attributes, causing warning about
duplicate sysfs entries upon reconnection.
This patch reworks the hwmon allocation to avoid device-managed
resource allocation, and uses the NVMe controller as parent for
the sysfs attributes.

Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: NHannes Reinecke <hare@suse.de>
Tested-by: NEnzo Matsumiya <ematsumiya@suse.de>
Tested-by: NDaniel Wagner <dwagner@suse.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

When using nvme native multipathing, if a path related error occurs
during ->queue_rq, the request needs to be completed with
NVME_SC_HOST_PATH_ERROR so that the request can be failed over.

Introduce a helper to complete the command from ->queue_rq in a wait
that invokes nvme_complete_rq.
Signed-off-by: NChao Leng <lengchao@huawei.com>
[hch: renamed, added a return value to clean up the callers a bit]
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Add nvme_cancel_tagset and nvme_cancel_admin_tagset for tear down and
reconnection error handling.
Signed-off-by: NChao Leng <lengchao@huawei.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The only used argument in this function is the "req".
Signed-off-by: NMax Gurtovoy <mgurtovoy@nvidia.com>
Reviewed-by: NMinwoo Im <minwoo.im.dev@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

There are no callers for nvme_reset_ctrl_sync() and
nvme_alloc_request_qid() so that we keep the symbols exported.

Unexport those functions, mark them static and update the header file
respectively.
Signed-off-by: NMinwoo Im <minwoo.im.dev@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Allow ZNS NVMe SSDs to present a read-only namespace when append is not
supported, instead of rejecting the namespace directly.

This allows (i) the namespace to be used in read-only mode, which is not
a problem as the append command only affects the write path, and (ii) to
use standard management tools such as nvme-cli to choose a different
format or firmware slot that is compatible with the Linux zoned block
device.
Signed-off-by: NJavier González <javier.gonz@samsung.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Commands get stuck while Host NVMe-oF controller is in reconnect state.
The controller enters into reconnect state when it loses connection with
the target.  It tries to reconnect every 10 seconds (default) until
a successful reconnect or until the reconnect time-out is reached.
The default reconnect time out is 10 minutes.

Applications are expecting commands to complete with success or error
within a certain timeout (30 seconds by default).  The NVMe host is
enforcing that timeout while it is connected, but during reconnect the
timeout is not enforced and commands may get stuck for a long period or
even forever.

To fix this long delay due to the default timeout, introduce new
"fast_io_fail_tmo" session parameter.  The timeout is measured in seconds
from the controller reconnect and any command beyond that timeout is
rejected.  The new parameter value may be passed during 'connect'.
The default value of -1 means no timeout (similar to current behavior).
Signed-off-by: NVictor Gladkov <victor.gladkov@kioxia.com>
Signed-off-by: NChaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NChao Leng <lengchao@huawei.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>