This patch converts se_lun->lun_deve_lock acquire/release access
to use a normal, non bottom-half spin_lock_t for protecting
se_lun->lun_deve_list access.
Reported-by: NChristoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

LUN allocation is now fully dynamic, so there is no need to
artificially restrain the number of exported LUNs.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

As we're now using a list to hold the LUNs the target core
can now converted to use 64-bit LUNs internally.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Reported-by: NBart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

With se_port and t10_alua_tg_pt_gp_member being absored into se_lun,
there is no need for an extra lock to protect se_lun->lun_se_dev
assignment.

This patch also converts backend drivers to use call_rcu() release
to allow any se_device readers to complete.  The call_rcu() instead
of kfree_rcu() is required here because se_device is embedded into
the backend driver specific structure.

Also, convert se_lun->lun_stats to use atomic_long_t within the
target_complete_ok_work() completion callback, and add FIXME for
transport_lookup_tmr_lun() with se_lun->lun_ref.

Finally, update sbp_update_unit_directory() special case usage with
proper rcu_dereference_raw() and configfs symlink comment.
Reported-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Chris Boot <bootc@bootc.net>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch eliminates all se_port + t10_alua_tg_pt_gp_member usage,
and converts current users to direct se_lun pointer dereference.

This includes the removal of core_export_port(), core_release_port()
core_dev_export() and core_dev_unexport().  Along with conversion
of special case se_lun pointer dereference within PR ALL_TG_PT=1
and ALUA access state transition UNIT_ATTENTION handling.

Also, update core_enable_device_list_for_node() to reference the
new per se_lun->lun_deve_list when creating a new entry, or
replacing an existing one via RCU.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Instead of starting a thread from transport_clear_lun_ref() that
waits for LUN shutdown, wait in that function for LUN shutdown
to finish. Additionally, change the return type of
transport_clear_lun_ref() from int to void.
Signed-off-by: NBart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Consolidate the implementation of the backend attributes in a single file
and single function per attribute show/store function instead of splitting
it into multiple functions in multiple files.

Also use the proper strto* helpers for exposed data types, add macros to
implement the store methods for the most common data types and share the
show methods between the two different attribute implementations.

(Fix bogus store_pi_prot_format flag=0 return value - nab)
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Rewrite the backend driver registration based on what we did to the fabric
drivers:  introduce a read-only struct target_bakckend_ops that the driver
registers, which is then instanciate as a struct target_backend by the
core.  This allows the ops vector to be smaller and allows us to mark it
const.  At the same time the registration function can set up the
configfs attributes, avoiding the need to add additional boilerplate code
for that to the drivers.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch changes core_tpg_add_node_to_devs() to avoid unnecessarly
resetting every se_dev_entry in se_node_acl->tpg_lun_hlist when the
operation is driven by an explicit configfs se_lun->lun_group creation
via core_dev_add_lun() to only update a single se_lun.

Otherwise for the second core_tpg_check_initiator_node_acl() case, go
ahead and continue to scan the full set of currently active se_lun in
se_portal_group->tpg_lun_hlist.
Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch converts se_tpg->acl_node_lock to struct mutex, so that
->acl_node_acl walkers in core_clear_lun_from_tpg() can block when
calling core_disable_device_list_for_node().

It also updates core_dev_add_lun() to hold ->acl_node_mutex when
calling core_tpg_add_node_to_devs() to build ->lun_entry_hlist
for dynamically generated se_node_acl.
Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch converts the fixed size se_portal_group->tpg_lun_list[]
to use modern RCU with hlist_head in order to support an arbitary
number of se_lun ports per target endpoint.

It includes dropping core_tpg_alloc_lun() from core_dev_add_lun(),
and calling it directly from target_fabric_make_lun() to allocate
a new se_lun.  And add a new target_fabric_port_release() configfs
item callback to invoke kfree_rcu() to release memory during
se_lun->lun_group shutdown.

Also now that se_node_acl->lun_entry_hlist is using RCU, convert
existing tpg_lun_lock to struct mutex so core_tpg_add_node_to_devs()
can perform RCU updater logic without releasing ->tpg_lun_mutex.

Also, drop core_tpg_clear_object_luns() and it's single consumer
in iscsi-target, which is duplicating TPG LUN shutdown logic and
is current code results in a NOP.

Finally, sbp-target and xen-scsiback fabric driver conversions are
included, which are required due to the non-standard way they use
->tpg_lun_hlist.
Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Chris Boot <bootc@bootc.net>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch converts the core_scsi3_pr_seq_non_holder() check for non
reservation holding registrations to use an atomic bitop in ->deve_flags
to determine if a registration is currently active.

It also includes associated a set_bit() in __core_scsi3_add_registration()
and clear_bit() in __core_scsi3_free_registration(), if se_dev_entry still
exists, and has not already been released via se_dev_entry shutdown path
in core_disable_device_list_for_node().

Also, clear_bit in core_disable_device_list_for_node as well to ensure
the read-critical path in core_scsi3_pr_seq_non_holder() picks up the
new state, preceeding the final kfree_rcu() call.
Reported-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch converts se_node_acl->device_list[] table for mappedluns
to modern RCU hlist_head usage in order to support an arbitrary number
of node_acl lun mappings.

It converts transport_lookup_*_lun() fast-path code to use RCU read path
primitives when looking up se_dev_entry.  It adds a new hlist_head at
se_node_acl->lun_entry_hlist for this purpose.

For transport_lookup_cmd_lun() code, it works with existing per-cpu
se_lun->lun_ref when associating se_cmd with se_lun + se_device.
Also, go ahead and update core_create_device_list_for_node() +
core_free_device_list_for_node() to use ->lun_entry_hlist.

It also converts se_dev_entry->pr_ref_count access to use modern
struct kref counting, and updates core_disable_device_list_for_node()
to kref_put() and block on se_deve->pr_comp waiting for outstanding PR
special-case PR references to drop, then invoke kfree_rcu() to wait
for the RCU grace period to complete before releasing memory.

So now that se_node_acl->lun_entry_hlist fast path access uses RCU
protected pointers, go ahead and convert remaining non-fast path
RCU updater code using ->lun_entry_lock to struct mutex to allow
callers to block while walking se_node_acl->lun_entry_hlist.

Finally drop the left-over core_clear_initiator_node_from_tpg() that
originally cleared lun_access during se_node_acl shutdown, as post
RCU conversion it now becomes duplicated logic.
Reviewed-by: NHannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Signed-off-by: NChristophe Vu-Brugier <cvubrugier@fastmail.fm>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Drivers may override the WCE flag, in which case the DPOFUA flag in
MODE SENSE might differ from the check used to reject invalid FUA
bits in sbc_check_dpofua.  Also now that we reject invalid FUA
bits early there is no need to duplicate the same buggy check
down in the fileio code.

As the DPOFUA flag controls th support for FUA bits on read and
write commands as well as DPO key off all the checks off a single
helper, and deprecate the emulate_dpo and emulate_fua_read attributs.

This fixes various failures in the libiscsi testsuite.

Personally I'd prefer to also remove the emulate_fua_write attribute
as there is no good reason to disable it, but I'll leave that for
a separate discussion.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

It seems like we only care if a transport is passthrough or not. Convert
transport_type to a flags field and replace TRANSPORT_PLUGIN_* with a
flag, TRANSPORT_FLAG_PASSTHROUGH.
Signed-off-by: NAndy Grover <agrover@redhat.com>
Reviewed-by: NIlias Tsitsimpis <iliastsi@arrikto.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Aside from whether they handle BIDI ops or not, parsing of the CDB by
kernel and user SCSI passthrough modules should be identical. Move this
into a new passthrough_parse_cdb() and call it from tcm-pscsi and tcm-user.
Reported-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NIlias Tsitsimpis <iliastsi@arrikto.com>
Signed-off-by: NAndy Grover <agrover@redhat.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Before 4.0, reading attrib/emulate_fua_write has returned 1. Saved
configs created on a pre-4.0 kernel will try to write that back when
restoring LIO configuration. This should succeed with no effect,
and issue a warning.

See https://bugzilla.redhat.com/show_bug.cgi?id=1206184Reported-by: NYanko Kaneti <yaneti@declera.com>
Reported-by: NJuan Quintela <quintela@redhat.com>
Signed-off-by: NAndy Grover <agrover@redhat.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

A check that rejects a CDB with FUA bit set if no write cache is
emulated was added by the following commit:

  fde9f50f target: Add sanity checks for DPO/FUA bit usage

The condition is as follows:

  if (!dev->dev_attrib.emulate_fua_write ||
      !dev->dev_attrib.emulate_write_cache)

However, this check is wrong if the backend device supports WCE but
"emulate_write_cache" is disabled.

This patch uses se_dev_check_wce() (previously named
spc_check_dev_wce) to invoke transport->get_write_cache() if the
device has a write cache or check the "emulate_write_cache" attribute
otherwise.
Reported-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NChristophe Vu-Brugier <cvubrugier@fastmail.fm>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch fixes a NULL pointer dereference triggered by a late
target_configure_device() -> alloc_workqueue() failure that results
in target_free_device() being called with DF_CONFIGURED already set,
which subsequently OOPses in destroy_workqueue() code.

Currently this only happens at modprobe target_core_mod time when
core_dev_setup_virtual_lun0() -> target_configure_device() fails,
and the explicit target_free_device() gets called.

To address this bug originally introduced by commit 0fd97ccf, go
ahead and move DF_CONFIGURED to end of target_configure_device()
code to handle this special failure case.
Reported-by: NClaudio Fleiner <cmf@daterainc.com>
Cc: Claudio Fleiner <cmf@daterainc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org> # v3.7+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Now that incoming FUA=1 bit check is enforced for backends with FUA or
WCE disabled, go ahead and disallow the changing of related backend
attributes when active fabric exports exist.

This is required to avoid potential failures with existing initiator
LUN registrations that have been previously created with FUA=1.
Reported-by: NChristoph Hellwig <hch@lst.de>
Cc: Doug Gilbert <dgilbert@interlog.com>
Cc: James Bottomley <JBottomley@Parallels.com>
Cc: Ronnie Sahlberg <ronniesahlberg@gmail.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Now that fabric_max_sectors is no longer used to enforce the maximum
I/O size, go ahead and drop it's left-over usage in target-core and
associated backend drivers.

Cc: Christoph Hellwig <hch@lst.de>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Roland Dreier <roland@purestorage.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch drops the arbitrary maximum I/O size limit in sbc_parse_cdb(),
which currently for fabric_max_sectors is hardcoded to 8192 (4 MB for 512
byte sector devices), and for hw_max_sectors is a backend driver dependent
value.

This limit is problematic because Linux initiators have only recently
started to honor block limits MAXIMUM TRANSFER LENGTH, and other non-Linux
based initiators (eg: MSFT Fibre Channel) can also generate I/Os larger
than 4 MB in size.

Currently when this happens, the following message will appear on the
target resulting in I/Os being returned with non recoverable status:

  SCSI OP 28h with too big sectors 16384 exceeds fabric_max_sectors: 8192

Instead, drop both [fabric,hw]_max_sector checks in sbc_parse_cdb(),
and convert the existing hw_max_sectors into a purely informational
attribute used to represent the granuality that backend driver and/or
subsystem code is splitting I/Os upon.

Also, update FILEIO with an explicit FD_MAX_BYTES check in fd_execute_rw()
to deal with the one special iovec limitiation case.

v2 changes:
  - Drop hw_max_sectors check in sbc_parse_cdb()
Reported-by: NLance Gropper <lance.gropper@qosserver.com>
Reported-by: NStefan Priebe <s.priebe@profihost.ag>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Roland Dreier <roland@purestorage.com>
Cc: stable@vger.kernel.org # 3.4
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Now that PSCSI is only exposing four hw_* read-only device attributes,
go ahead and drop the left-over -> legacy PHBA_PDEV checks in various
se_dev_set_* code, since it's now only used by virtual devices.
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Now that target_core_backend_configfs.h macros will be using these
se_dev_set attribute functions externally to allow backend drivers
to populate different attributes, go ahead and add EXPORT_SYMBOL()
for the existing default set of 30 device attributes.

Also update target_core_backend.h with proper function prototypes.
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch adds a force_pr_aptpl device attribute used to force SPC-3 PR
Activate Persistence across Target Power Loss (APTPL) operation.  This
makes PR metadata write-out occur during state change regardless if new
PERSISTENT_RESERVE_OUT CDBs have their APTPL feature bit set.

This is useful during H/A failover in active/passive setups where all PR
state is being re-created on a different node, driven by configfs backend
device + export layout and pre-loaded $DEV/pr/res_aptpl_metadata.

Cc: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch fixes a bug in handling of SPC-3 PR Activate Persistence
across Target Power Loss (APTPL) logic where re-creation of state for
MappedLUNs from dynamically generated NodeACLs did not occur during
I_T Nexus establishment.

It adds the missing core_scsi3_check_aptpl_registration() call during
core_tpg_check_initiator_node_acl() -> core_tpg_add_node_to_devs() in
order to replay any pre-loaded APTPL metadata state associated with
the newly connected SCSI Initiator Port.

Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: <stable@vger.kernel.org>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

The target code has a rather generous helping of smp_mb__after_atomic()
throughout the code base.  Most atomic operations were followed by one
and none were preceded by smp_mb__before_atomic(), nor accompanied by a
comment explaining the need for a barrier.

Instead of trying to prove for every case whether or not it is needed,
this patch introduces atomic_inc_mb() and atomic_dec_mb(), which
explicitly include the memory barriers before and after the atomic
operation.  For now they are defined in a target header, although they
could be of general use.

Most of the existing atomic/mb combinations were replaced by the new
helpers.  In a few cases the atomic was sandwiched in
spin_lock/spin_unlock and I simply removed the barrier.

I suspect that in most cases the correct conversion would have been to
drop the barrier.  I also suspect that a few cases exist where a) the
barrier was necessary and b) a second barrier before the atomic would
have been necessary and got added by this patch.
Signed-off-by: NJoern Engel <joern@logfs.org>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Remove core_tpg_pre_dellun entirely, since we don't need to get/check
a pointer we already have.

Nothing else can return an error, so core_dev_del_lun can return void.

Rename core_tpg_post_dellun to remove_lun - a clearer name, now that
pre_dellun is gone.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAndy Grover <agrover@redhat.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch fixes a left-over se_lun->lun_sep pointer OOPs when one
of the /sys/kernel/config/target/$FABRIC/$WWPN/$TPGT/lun/$LUN/alua*
attributes is accessed after the $DEVICE symlink has been removed.

To address this bug, go ahead and clear se_lun->lun_sep memory in
core_dev_unexport(), so that the existing checks for show/store
ALUA attributes in target_core_fabric_configfs.c work as expected.
Reported-by: NSebastian Herbszt <herbszt@gmx.de>
Tested-by: NSebastian Herbszt <herbszt@gmx.de>
Cc: stable@vger.kernel.org
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Just like for pSCSI, if the transport sets get_write_cache, then it is
not valid to enable write cache emulation for it. Return an error.

see https://bugzilla.redhat.com/show_bug.cgi?id=1082675Reviewed-by: NChris Leech <cleech@redhat.com>
Signed-off-by: NAndy Grover <agrover@redhat.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Userspace tools assume if a value is read from configfs, it is valid
and will not cause an error if the same value is written back. The only
valid value for pi_prot_type for backends not supporting DIF is 0, so allow
this particular value to be set without returning an error.
Reported-by: NKrzysztof Chojnowski <frirajder@gmail.com>
Signed-off-by: NAndy Grover <agrover@redhat.com>
Reviewed-by: NSagi Grimberg <sagig@mellanox.com>
Cc: stable@vger.kernel.org # 3.14+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Mostly scripted conversion of the smp_mb__* barriers.
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Acked-by: NPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/n/tip-55dhyhocezdw1dg7u19hmh1u@git.kernel.org
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-arch@vger.kernel.org
Signed-off-by: NIngo Molnar <mingo@kernel.org>

This patch adds support for exposing DIF protection device
attributes via configfs.  This includes:

   pi_prot_type: Protection Type (0, 1, 3 currently support)
   pi_prot_format: Protection Format Operation (FILEIO only)

Within se_dev_set_pi_prot_type() it also adds the se_subsystem_api
device callbacks to setup per device protection information.

v2 changes:
  - Drop pi_guard_type + pi_prot_version related code (MKP)
  - Add pi_prot_format logic (Sagi)
  - Add ->free_prot callback in target_free_device
  - Add hw_pi_prot_type read-only attribute

Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.de>
Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Referrals need an LBA map, which needs to be kept
consistent across all target port groups. So
instead of tying the map to the target port groups
I've implemented a single attribute containing the
entire map.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Add infrastructure for referrals.

v2 changes:

 - Fix unsigned long long division in core_alua_state_lba_dependent on
   32-bit  (Fengguang + Chen + Hannes)
 - Fix compile warning in core_alua_state_lba_dependent (nab)
 - Convert segment_* + sectors variables in core_alua_state_lba_dependent
   to u64 (Hannes)
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch allows FILEIO to update hw_max_sectors based on the current
max_bytes_per_io.  This is required because vfs_[writev,readv]() can accept
a maximum of 2048 iovecs per call, so the enforced hw_max_sectors really
needs to be calculated based on block_size.

This addresses a >= v3.5 bug where block_size=512 was rejecting > 1M
sized I/O requests, because FD_MAX_SECTORS was hardcoded to 2048 for
the block_size=4096 case.

(v2: Use max_bytes_per_io instead of ->update_hw_max_sectors)
Reported-by: NHenrik Goldman <hg@x-formation.com>
Cc: <stable@vger.kernel.org> #3.5+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

"pre" is really an allocation function. The only time it isn't called is
for virtual_lun0, which is statically allocated. Renaming that to "alloc"
lets the other function not need to be "post", and just be called
core_tpg_add_lun.

(nab: fix minor applying fuzz in core_tpg_setup_virtual_lun0)
Signed-off-by: NAndy Grover <agrover@redhat.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>