We have an nvme device that has a concept of a stripe size. IO requests
that do not transfer data crossing a stripe boundary has greater
performance compared to IO that does cross it. This patch sets the
stripe size for the device if the device and vendor ids match one with
this feature and splits IO requests that cross the stripe boundary.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

It is possible a bio request can not be submitted as a single NVMe IO
command if the bio_vec is not mergeable with the NVMe PRP alignement
constraints. This condition was handled by submitting an IO for the
mergeable portion then submitting a follow on IO for the remaining data
after the previous IO completes. The remainder to be sent was tracked
by manipulating the bio->bi_idx and bio->bi_sector. This patch splits
the request as many times as necessary and submits the bios together.

Since submitting the bio may cause it to be requeued on split,
nvme_resubmit_bios had to be modified to remove the wait queue when
the bio list is empty prior to submitting the bio since a split would
have added the wait queue a second time, corrupting the wait queue head
task list.

There are a few other benefits from doing this: it fixes a potential
issue with the previous handling of a non-mergeable bio as the requeuing
method could would use an unlocked nvme_queue if the callback isn't
invoked on the queue's associated cpu; it will be possible to retry a
failed bio if desired at some later time since it does not manipulate
the original bio; the bio integrity extensions require the bio to be in
its original condition for the checks to work correctly if we implement
the end-to-end data protection in the future.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

There is no situation that could occur where we could error out of this
function and require cleaning up allocated namespaces.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

The nvme_queue's depth is not set if we fail to allocate submission queue
entries, which was being used to determine how much coherent memory to
free on error. Use the depth variable instead.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Fixes a potential memory leak if requesting the admin queue irq fails.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Translates a scsi unmap request from SG_IO ioctl to NVMe
data-set-management deallocate.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Acked-by: NVishal Verma <vishal.l.verma@linux.intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Fixes nvme queue usages in scsi-to-nvme translation code to not get
a queue more often than it is being put, and not use the queue in an
unsafe way without it being locked.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Acked-by: NVishal Verma <vishal.l.verma@linux.intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

The kthread has two tasks; handling timeouts (for which it runs once per
second), and submitting queued BIOs.  If a BIO happens to be queued after
the thread has processed the queue but before it calls schedule_timeout(),
the thread will sleep for a second before submitting it, which can cause
performance problems in some rare cases (that will become more common in
a subsequent patch).
Signed-off-by: NArjan van de Ven <arjan@linux.intel.com>
Tested-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Registers a miscellaneous device for each nvme controller probed. This
creates character device files as /dev/nvmeN, where N is the device
instance, and supports nvme admin ioctl commands so devices without
namespaces can be managed.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

When constructing the command, dsmgmt needs to be treated as a 32-bit
value, not a 16-bit value.  reftag, apptag and appmask all need to be
converted from native-endian to little-endian.  Again, sparse's bitwise
warnings caught this problem.  Thanks to Keith for pointing out the
correct way to fix the reftag.
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>
Acked-by: NKeith Busch <keith.busch@intel.com>

The sparse bitwise checks pointed out that I needed to shift the status
before changing its endianness, not after.
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Sparse produced warnings for some instances of
mismatched types and direct userspace dereferences.
This patch fixes those for the scsi emulation layer.
Signed-off-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

The nvme_dev_add() function currently returns the last error code that it
saw, which (if everything else succeeds) happens to be the result of an
optional command, so it can legitimately fail.  Looking at the error path
more closely reveals that we should return success from this function,
even if no device namespaces are added.  So once the queues are created
and the device has responded to Identify, make sure that this function
succeeds.
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>
Acked-by: NKeith Busch <keith.busch@intel.com>

Introduce nvme_block_nr() to help convert sectors to block numbers.
This fixes an integer overflow in the SCSI conversion layer, and it's
slightly less typing than opencoding it.
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>
Acked-by: NKeith Busch <keith.busch@intel.com>

The nvme driver has a "once per second" event where the management kthread
wakes up the system and then reschedules itself for 1 second later.
For power efficiency reasons, I'd like this timer to happen together
with other wakeups in the system.

This patch makes the schedule_timeout() call in the kthread use
round_jiffies_relative(), causing the wakeup to at least align with other
"once per X seconds" events in the kernel.
Signed-off-by: NArjan van de Ven <arjan@linux.intel.com>
Tested-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

Translates SCSI commands in SG_IO ioctl to NVMe commands.
Uses the scsi-nvme translation spec from nvmexpress.org as reference.
Signed-off-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

The SCSI emulation has the ability to send format commands, so we need
to add the definition of the command.  Also add a missing error code.
Signed-off-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

nvme-scsi.c uses several data structures and definitions that were
previously private to nvme-core.c.  Move the definitions to nvme.h,
protected by __KERNEL__.
Signed-off-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

In preparation for adding nvme-scsi.c
It is preferable to retain the module name 'nvme'
Signed-off-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

This adds discard support to block queues if the nvme device is capable of
deallocating blocks as indicated by the controller's optional command support.
A discard flagged bio request will submit an NVMe deallocate Data Set
Management command for the requested blocks.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

The LBA Range Type feature is optional in the NVMe specification,
so we should continue with adding namespaces for controllers that do
not implement this feature.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NMatthew Wilcox <matthew.r.wilcox@intel.com>

I just fixed this in "drivers/block/rbd.c" and I noticed that
"drivers/block/nbd.c" has the same problem.  Fix a warning issued by
sparse by adding some lockdep annotations to indicate the queue lock gets
dropped (because it's held when do_nbd_request() is called) and
re-acquired within the function.
Signed-off-by: NAlex Elder <elder@inktank.com>
Cc: Paul Clements <paul.clements@steeleye.com>
Cc: Paul Clements <paul.clements@us.sios.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Pass the read-only flag to set_device_ro, so that it will be visible to
the block layer and in sysfs.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Cc: Paul Clements <Paul.Clements@steeleye.com>
Cc: Alex Bligh <alex@alex.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

There are two problems with shutdown in the NBD driver.

1: Receiving the NBD_DISCONNECT ioctl does not sync the filesystem.

   This patch adds the sync operation into __nbd_ioctl()'s
   NBD_DISCONNECT handler.  This is useful because BLKFLSBUF is restricted
   to processes that have CAP_SYS_ADMIN, and the NBD client may not
   possess it (fsync of the block device does not sync the filesystem,
   either).

2: Once we clear the socket we have no guarantee that later reads will
   come from the same backing storage.

   The patch adds calls to kill_bdev() in __nbd_ioctl()'s socket
   clearing code so the page cache is cleaned, lest reads that hit on the
   page cache will return stale data from the previously-accessible disk.

Example:

    # qemu-nbd -r -c/dev/nbd0 /dev/sr0
    # file -s /dev/nbd0
    /dev/stdin: # UDF filesystem data (version 1.5) etc.
    # qemu-nbd -d /dev/nbd0
    # qemu-nbd -r -c/dev/nbd0 /dev/sda
    # file -s /dev/nbd0
    /dev/stdin: # UDF filesystem data (version 1.5) etc.

While /dev/sda has:

    # file -s /dev/sda
    /dev/sda: x86 boot sector; etc.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Acked-by: NPaul Clements <Paul.Clements@steeleye.com>
Cc: Alex Bligh <alex@alex.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Currently, the NBD device does not accept flush requests from the Linux
block layer.  If the NBD server opened the target with neither O_SYNC nor
O_DSYNC, however, the device will be effectively backed by a writeback
cache.  Without issuing flushes properly, operation of the NBD device will
not be safe against power losses.

The NBD protocol has support for both a cache flush command and a FUA
command flag; the server will also pass a flag to note its support for
these features.  This patch adds support for the cache flush command and
flag.  In the kernel, we receive the flags via the NBD_SET_FLAGS ioctl,
and map NBD_FLAG_SEND_FLUSH to the argument of blk_queue_flush.  When the
flag is active the block layer will send REQ_FLUSH requests, which we
translate to NBD_CMD_FLUSH commands.

FUA support is not included in this patch because all free software
servers implement it with a full fdatasync; thus it has no advantage over
supporting flush only.  Because I [Paolo] cannot really benchmark it in a
realistic scenario, I cannot tell if it is a good idea or not.  It is also
not clear if it is valid for an NBD server to support FUA but not flush.
The Linux block layer gives a warning for this combination, the NBD
protocol documentation says nothing about it.

The patch also fixes a small problem in the handling of flags: nbd->flags
must be cleared at the end of NBD_DO_IT, but the driver was not doing
that.  The bug manifests itself as follows.  Suppose you two different
client/server pairs to start the NBD device.  Suppose also that the first
client supports NBD_SET_FLAGS, and the first server sends
NBD_FLAG_SEND_FLUSH; the second pair instead does neither of these two
things.  Before this patch, the second invocation of NBD_DO_IT will use a
stale value of nbd->flags, and the second server will issue an error every
time it receives an NBD_CMD_FLUSH command.

This bug is pre-existing, but it becomes much more important after this
patch; flush failures make the device pretty much unusable, unlike
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAlex Bligh <alex@alex.org.uk>
Acked-by: NPaul Clements <Paul.Clements@steeleye.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Convert to the much saner new idr interface.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Convert to the much saner new idr interface.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NJens Axboe <axboe@kernel.dk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

idr_destroy() can destroy idr by itself and idr_remove_all() is being
deprecated.  Drop its usage.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Use the new version of the encoding for osd requests and replies.  In the
process, update the way we are tracking request ops and reply lengths and
results in the struct ceph_osd_request.  Update the rbd and fs/ceph users
appropriately.

The main changes are:
 - we keep pointers into the request memory for fields we need to update
   each time the request is sent out over the wire
 - we keep information about the result in an array in the request struct
   where the users can easily get at it.
Signed-off-by: NSage Weil <sage@inktank.com>
Reviewed-by: NAlex Elder <elder@inktank.com>

The only thing type-specific osd completion functions do with their
osd op parameter is (in some cases) extract the number of bytes
transferred from it.  In the other cases, the xferred bytes field
is not used, and total message data transfer byte count (which may
well be zero) is used.

Just set the object request transfer count in the main osd request
callback function and provide that to the other routines.  There is
then no longer any need to pass the op pointer to the type-specific
completion routines, so drop those parameters.

Stop doing anything with the total message data length.
Signed-off-by: NAlex Elder <elder@inktank.com>
Reviewed-by: NSage Weil <sage@inktank.com>

This function is slightly out of place, probably the result
of an errant automatic merge or something.
Signed-off-by: NAlex Elder <elder@inktank.com>
Reviewed-by: NSage Weil <sage@inktank.com>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Fengguang Wu reminded me that there were outstanding sparse reports
in the ceph and rbd code.  This patch fixes these problems in rbd
that lead to those reports:
    - Convert functions that are never referenced externally to have
      static scope.
    - Add a lockdep annotation to rbd_request_fn(), because it
      releases a lock before acquiring it again.

This partially resolves:
    http://tracker.ceph.com/issues/4184Reported-by: NFengguang Wu <fengguang.wu@intel.com>
Signed-off-by: NAlex Elder <elder@inktank.com>
Reviewed-by: NJosh Durgin <josh.durgin@inktank.com>

Add dout() calls to facilitate tracing of image and object requests.
Change a few existing calls so they use __func__ rather than the
hard-coded function name.  Have calls always add ":" after the name
of the function, and prefix pointer values with a consistent tag
indicating what it represents.  (Note that there remain some older
dout() calls that are left untouched by this patch.)

Issue a warning if rbd_osd_write_callback() ever gets a short write.

This resolves:
    http://tracker.ceph.com/issues/4235Signed-off-by: NAlex Elder <elder@inktank.com>
Reviewed-by: NJosh Durgin <josh.durgin@inktank.com>

Let's go shopping!

I'm afraid this may not have gotten it right:
    07741308  rbd: add barriers near done flag operations

The smp_wmb() should have been done *before* setting the done flag,
to ensure all other data was valid before marking the object request
done.

Switch to use atomic_inc_return() here to set the done flag, which
allows us to verify we don't mark something done more than once.
Doing this also implies general barriers before and after the call.

And although a read memory barrier might have been sufficient before
reading the done flag, convert this to a full memory barrier just
to put this issue to bed.

This resolves:
    http://tracker.ceph.com/issues/4238Signed-off-by: NAlex Elder <elder@inktank.com>
Reviewed-by: NJosh Durgin <josh.durgin@inktank.com>

The old request code simply ignored zero-length requests.  We should
still operate that same way to avoid any changes in behavior.  We
can implement handling for special zero-length requests separately
(see http://tracker.ceph.com/issues/4236).

Add some assertions based on this new constraint.

This resolves:
    http://tracker.ceph.com/issues/4237Signed-off-by: NAlex Elder <elder@inktank.com>
Reviewed-by: NJosh Durgin <josh.durgin@inktank.com>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Negative offset may cause loop device size larger than backing file
size.

 $ fallocate -l 1M a
 $ losetup --offset 0xffffffffffff0000 /dev/loop0 a
 $ blockdev --getsize64 /dev/loop0
 1114112
 $ ls -l a
 -rw-r--r-- 1 root root 1048576 Jan 23 12:46 a
 $ cat /dev/loop0
 cat: /dev/loop0: Input/output error

It makes no sense to do that. Only apply offset when it's positive.

Fix a typo in the comment by the way.
Signed-off-by: NGuo Chao <yan@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Guo Chao <yan@linux.vnet.ibm.com>
Cc: M. Hindess <hindessm@uk.ibm.com>
Cc: Nikanth Karthikesan <knikanth@suse.de>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

When loopdev is built as module and we pass an invalid parameter,
loop_init() will return directly without deregister misc device, which
will cause an oops when insert loop module next time because we left some
garbage in the misc device list.

Test case:
sudo modprobe loop max_part=1024
(failed due to invalid parameter)
sudo modprobe loop
(oops)

Clean up nicely to avoid such oops.
Signed-off-by: NGuo Chao <yan@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Guo Chao <yan@linux.vnet.ibm.com>
Cc: M. Hindess <hindessm@uk.ibm.com>
Cc: Nikanth Karthikesan <knikanth@suse.de>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>