iSER currently has a couple places that set max_sectors in either the host
template or SCSI host, and all of them get it wrong.
This patch instead uses a single assignment that (hopefully) gets it right:
the max_sectors value must be derived from the number of segments in the
FR or FMR structure, but actually be one lower than the page size multiplied
by the number of sectors, as it has to handle the case of non-aligned I/O.
Without this I get trivial to reproduce hangs when running xfstests
(on XFS) over iSER to Linux targets.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Max Gurtovoy <[email protected]>
Acked-by: Sagi Grimberg <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
We need to have the WWN fully initialized before addig default groups to it,
so add a new method to add these groups after the WWN has been initialized.
Also remove the default groups in the core while we're at it.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
The iSCSI targets wants to add a default group, for which we need to
have the list of default groups initialized previously.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
Instead we can clean up the list of default ACLs in core code.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
This is a simple extension to the block layout driver to use SCSI
persistent reservations for access control and fencing, as well as
SCSI VPD pages for device identification.
For this we need to pass the nfs4_client to the proc_getdeviceinfo method
to generate the reservation key, and add a new fence_client method
to allow for fence actions in the layout driver.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Trivial reorganization, no change in behavior. Move some code around,
pull some code out of block layoutcommit that will be useful for the
scsi layout.
[[email protected]: split off from "nfsd: add SCSI layout support"]
Signed-off-by: J. Bruce Fields <[email protected]>
Split the config symbols into a generic pNFS one, which is invisible
and gets selected by the layout drivers, and one for the block layout
driver.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
This is a trivial extension to the block layout driver to support the
new SCSI layouts draft. There are three changes:
- device identifcation through the SCSI VPD page. This allows us to
directly use the udev generated persistent device names instead of
requiring an expensive lookup by crawling every block device node
in /dev and reading a signature for it.
- use of SCSI persistent reservations to protect device access and
allow for robust fencing. On the client sides this just means
registering and unregistering a server supplied key.
- an optimized LAYOUTCOMMIT payload that doesn't send unessecary
fields to the server.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Trond Myklebust <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Based on draft-ietf-nfsv4-scsi-layout-05 after the WG last call.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Just call inode_dio_wait directly instead of through a pointless wrapper.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
The only difference to nfs_file_fsync is the call to pnfs_sync_inode. But
pnfs_sync_inode is just an inline that calls a pNFS layout driver method
if CONFIG_PNFS is designed, and thus can be called just fine from the core
NFS module.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
xfs_dir2_node_trim_free can return with setting the rvalp argument
pointer. Initialize it to 0 at the beginning of the function and
only update it to 1 if we succeeded trimming a freespace block.
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Carlos Maiolino <[email protected]>
Signed-off-by: Dave Chinner <[email protected]>
We use bio chaining during most I/Os these days due to the delayed
bio splitting. Additionally XFS will start using it, and there is
a pending direct I/O rewrite also making heavy use for it. Don't
pretend it's always unlikely, and let the branch predictor do it's
job instead.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Replace the while loop that unecessarily checks for a NULL bio in the fast
path with a simple goto loop.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Factor common code between bio_chain_endio and bio_endio into a common
helper.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Only overwrite the parents bi_error if it was zero. That way a successful
bio completion doesn't reset the error pointer.
Reported-by: Brian Foster <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
We only use the pointer when processing regular iSER commands, and it then
always points to the struct iser_cmd that contains the TX descriptor.
Remove it and rely on container_of to save a little space and avoid a
pointer that is updated multiple times per processed command.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Sagi Grimberg <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
There is exactly one instance per struct isert_cmd, so merge the two to
simplify everyones life.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Sagi Grimberg <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
Use the workqueue based CQ type similar to what isert was using previously,
and properly split up the completion handlers.
Note that this also takes special care to handle the magic login WRs
separately, and also renames the submission functions so that it's clear
that they are only to be used for the login buffers.
(Fix up isert_print_wc usage in isert_beacon_done - nab)
Signed-off-by: Christoph Hellwig <[email protected]>
[sagig: added iscsi conn reinstatement in non-flush
error completions and added error completion type print]
Signed-off-by: Sagi Grimberg <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
The login receive buffer is used as a iser_rx_desc, so type it as such
in struct isert_conn and allocate the exactly right space for it. The
TX buffer is moved to a separate variable and properly sized as well.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Sagi Grimberg <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
This is the same as ISCSI_DEF_MAX_RECV_SEG_LEN (and must be the same given
the structure layouts), so just use that constant instead. This also
allows removing ISER_RX_LOGIN_SIZE in favor of ISER_RX_PAYLOAD_SIZE.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Sagi Grimberg <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
This patch converts existing loopback, usb-gadget, and
xen-scsiback demo-mode only fabric drivers to use the
new target_alloc_session API caller.
This includes adding a new alloc_session callback for
fabric driver internal nexus pointer assignments.
(Fixes for early for-next nexus breakage - Dan Carpenter)
Cc: Christoph Hellwig <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Acked-by: Juergen Gross <[email protected]>
Tested-by: Andrzej Pietrasiewicz <[email protected]>
Tested-by: Chris Boot <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
For a long time all architectures implement the pci_dma_* functions using
the generic DMA API, and they all use the same header to do so.
Move this header, pci-dma-compat.h, to include/linux and include it from
the generic pci.h instead of having each arch duplicate this include.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Bjorn Helgaas <[email protected]>
FRV doesn't implement pci_{alloc,free}_consistent(), so remove the
declarations for them.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Bjorn Helgaas <[email protected]>
Acked-by: David S. Miller <[email protected]>
Acked-by: David Howells <[email protected]>
Replace the current NULL-terminated array of default groups with a linked
list. This gets rid of lots of nasty code to size and/or dynamically
allocate the array.
While we're at it also provide a conveniant helper to remove the default
groups.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Felipe Balbi <[email protected]> [drivers/usb/gadget]
Acked-by: Joel Becker <[email protected]>
Acked-by: Nicholas Bellinger <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Now that applications need to explicitly ask for polling we can enable it
by default in blk-mq drivers. Note that this will only have an affect
on driver that supply a poll function, which currently only includes nvme.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Stephen Bates <[email protected]>
Tested-by: Stephen Bates <[email protected]>
Reviewed-by: Jeff Moyer <[email protected]>
Signed-off-by: Al Viro <[email protected]>
This adds a flag that tells the file system that this is a high priority
request for which it's worth to poll the hardware. The flag is purely
advisory and can be ignored if not supported.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Stephen Bates <[email protected]>
Tested-by: Stephen Bates <[email protected]>
Acked-by: Jeff Moyer <[email protected]>
Signed-off-by: Al Viro <[email protected]>
This way we can set kiocb flags also from the sync read/write path for
the read_iter/write_iter operations. For now there is no way to pass
flags to plain read/write operations as there is no real need for that,
and all flags passed are explicitly rejected for these files.
Signed-off-by: Milosz Tanski <[email protected]>
[hch: rebased on top of my kiocb changes]
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Stephen Bates <[email protected]>
Tested-by: Stephen Bates <[email protected]>
Acked-by: Jeff Moyer <[email protected]>
Signed-off-by: Al Viro <[email protected]>
Simplifies the code, and makes it more fair vs other users by using a
softirq for polling.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Haggai Eran <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
This patch adds support for larger requests in blk_rq_map_user_iov by
allowing it to build multiple bios for a request. This functionality
used to exist for the non-vectored blk_rq_map_user in the past, and
this patch reuses the existing functionality for it on the unmap side,
which stuck around. Thanks to the iov_iter API supporting multiple
bios is fairly trivial, as we can just iterate the iov until we've
consumed the whole iov_iter.
Signed-off-by: Christoph Hellwig <[email protected]>
Reported-by: Jeff Lien <[email protected]>
Tested-by: Jeff Lien <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Driver private request types should not get the artifical cap for the
FS requests. This is important to use the full device capabilities
for internal command or NVMe pass through commands.
Signed-off-by: Christoph Hellwig <[email protected]>
Reported-by: Jeff Lien <[email protected]>
Tested-by: Jeff Lien <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Updated by me to use an explicit check for the one command type that
does support extended checking, instead of relying on the ordering
of the enum command values - as suggested by Keith.
Signed-off-by: Jens Axboe <[email protected]>
The block layer uses an unsigned short for max_segments. The way we
calculate the value for NVMe tends to generate very large 32-bit values,
which after integer truncation may lead to a zero value instead of
the desired outcome.
Signed-off-by: Christoph Hellwig <[email protected]>
Reported-by: Jeff Lien <[email protected]>
Tested-by: Jeff Lien <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Factor out a helper to set all the device specific queue limits and apply
them to the admin queue in addition to the I/O queues. Without this the
command size on the admin queue is arbitrarily low, and the missing
other limitations are just minefields waiting for victims.
Signed-off-by: Christoph Hellwig <[email protected]>
Reported-by: Jeff Lien <[email protected]>
Tested-by: Jeff Lien <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Just use the t_blk_res field directly instead of obsfucating the reference
by a macro.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Brian Foster <[email protected]>
Signed-off-by: Dave Chinner <[email protected]>
The only work left in the kthread is the periodic health check for each
controller. There is no need to run this from process context or keep
a thread context around for it, so replace it with a simpler timer.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Reviewed-by: Johannes Thumshirn <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Use a dedicated work item to submit async event requests instead of the
global kthread. This simplifies the code and reduces the latencies to
resubmit a request once an even notification happened.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Reviewed-by: Johannes Thumshirn <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
-EBADF is a rather confusing error if an operations is not supported,
and nfsd gets rather upset about it.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Al Viro <[email protected]>
The newly added NFS v4.2 operations (ALLOCATE, DEALLOCATE, SEEK and CLONE)
use a helper called nfs42_set_rw_stateid to select a stateid that is sent
to the server. But they don't set the inode and state fields in the
nfs4_exception structure, and this don't partake in the stateid recovery
protocol. Because of this they will simply return errors insted of trying
to recover a stateid when the server return a BAD_STATEID error.
Additionally CLONE has the problem that it operates on two files and thus
two stateids, and thus needs to call the exception handler twice to
recover stateids.
While we're at it stop grabbing an addititional reference to the open
context in all these operations - having the file open guarantees that
the open context won't go away.
All this can be produces with the generic/168 and generic/170 tests in
xfstests which stress the CLONE stateid handling.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
Pass the right private data to device_create_with_groups from the
beginning, and remove the superflous call to dev_set_drvdata.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Jon Derrick <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
See http: //www.infradead.org/rpr.html
X-Evolution-Source: [email protected]
Content-Transfer-Encoding: 8bit
Mime-Version: 1.0
We support OFFSET_MAX just fine, so don't round down below it. Also
switch to using min_t to make the helper more readable.
Signed-off-by: Christoph Hellwig <[email protected]>
Fixes: 433c92379d ("NFS: Clean up nfs_size_to_loff_t()")
Cc: [email protected] # 2.6.23+
Signed-off-by: Trond Myklebust <[email protected]>
We only need to communicate two bits of information to the direct I/O
completion handler:
(1) do we need to convert any unwritten extents in the range
(2) do we need to check if we need to update the inode size based
on the range passed to the completion handler
We can use the private data passed to the get_block handler and the
completion handler as a simple bitmask to communicate this information
instead of the current complicated infrastructure reusing the ioends
from the buffer I/O path, and thus avoiding a memory allocation and
a context switch for any non-trivial direct write. As a nice side
effect we also decouple the direct I/O path implementation from that
of the buffered I/O path.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Brian Foster <[email protected]>
This way we can pass back errors to the file system, and allow for
cleanup required for all direct I/O invocations.
Also allow the ->end_io handlers to return errors on their own, so that
I/O completion errors can be passed on to the callers.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Dave Chinner <[email protected]>
Signed-off-by: Dave Chinner <[email protected]>
We'll soon merge <asm-generic/dma-mapping-common.h> into
<linux/dma-mapping.h> and the reference to dma_capable in the tile
dma_set_mask would create a circular dependency.
Fix this by moving the implementation out of line.
Signed-off-by: Christoph Hellwig <[email protected]>
Cc: Chris Metcalf <[email protected]>
Cc: Christian Borntraeger <[email protected]>
Cc: Joerg Roedel <[email protected]>
Cc: Sebastian Ott <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Sparc already uses the same code as the generic code for the PCI
implementation but just fails the call sbus. This moves to the generic
implemenation which eventually return -EIO due to the NULL dma_mask
pointer in the device.
Signed-off-by: Christoph Hellwig <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Christian Borntraeger <[email protected]>
Cc: Joerg Roedel <[email protected]>
Cc: Sebastian Ott <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Translation SCSI commands to NVMe commands is rather pointless in general
as applications must not expext to be able to use SCSI commands on a
generic block device.
Make the huge translation layer optional and hope no one will ever enable
it in the future.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Make sure we synchronize access to the namespaces list and grab a reference
to the namespace before doing I/O. Make sure to reject the ioctl if multiple
namespaces are present as it's entirely unsafe, and warn when using it even
with a single namespace.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Currently traversal and modification of ctrl->namespaces happens completely
unsynchronized, which can be fixed by the addition of a simple mutex.
Note: nvme_dev_ioctl will be handled in the next patch.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
And propagate the error up the stack so we can add the stripe
to no_stripes_list and retry our log operation later. This avoids
blocking raid5d due to reclaim, an it allows to get rid of the
deadlock-prone GFP_NOFAIL allocation.
shli: add missing mempool_destroy()
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: NeilBrown <[email protected]>
We only have a limited number in flight, so use a page based mempool.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: NeilBrown <[email protected]>
Once the I/O completed we don't need the meta page anymore. As the iounits
can live on for a long time this reduces memory pressure a bit.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Shaohua Li <[email protected]>
Signed-off-by: NeilBrown <[email protected]>
It's only used for one kind of move, so make that explicit. Also clean
up the code a bit by using list_for_each_safe.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Shaohua Li <[email protected]>
Signed-off-by: NeilBrown <[email protected]>
Fold simplified versions of build_phys_page_list and
iwch_register_phys_mem into iwch_get_dma_wr now that no other callers
are left.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]> [core]
Reviewed-by: Steve Wise <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
Remove the unused ib_allow_mw and ib_bind_mw functions, remove the
unused IB_WR_BIND_MW and IB_WC_BIND_MW opcodes and move ib_dealloc_mw
into the uverbs module.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]> [core]
Reviewed-by: Steve Wise <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
We have stopped using phys MRs in the kernel a while ago, so let's
remove all the cruft used to implement them.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]> [core]
Reviewed-By: Devesh Sharma<[email protected]> [ocrdma]
Reviewed-by: Steve Wise <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
Just IB_DEVICE_LOCAL_DMA_LKEY and IB_DEVICE_MEM_MGT_EXTENSIONS for now
as I'm most familar with those.
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Sagi Grimberg <[email protected]>
Reviewed-By: Jason Gunthorpe <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>
This was added for the 'magic' AEN requests in the NVMe driver that never
return. We now handle them purely inside the driver and don't need this
core hack any more.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Merge the two per-request structures in the nvme driver into a single
one.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
We need to move freeing of resources to the ->complete handler to ensure
they are also freed when we cancel the command.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Now that all commands are executed as block layer requests we can remove the
internal completion in the NVMe driver. Note that we can simply call
blk_mq_complete_request to abort commands as the block layer will protect
against double copletions internally.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
AEN requests are different from other requests in that they don't time out
or can easily be cancelled. Because of that we should not use the blk-mq
infrastructure but just special case them in the completion path.
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
There is no lock to sychronize access to the abort_limit field of
struct nvme_ctrl, so switch it to an atomic_t.
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Keith Busch <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
Christoph Hellwig