The approach for adding a device to the devices_idr data structure and for
removing it is as follows:
* &dev->dev_group.cg_item is initialized before a device is added to
devices_idr.
* If the reference count of a device drops to zero then
target_free_device() removes the device from devices_idr.
* All devices_idr manipulations are protected by device_mutex.
This means that increasing the reference count of a device is sufficient to
prevent removal from devices_idr and also that it is safe access
dev_group.cg_item for any device that is referenced by devices_idr. Use
this to modify target_find_device() and target_for_each_device() such that
these functions no longer introduce a dependency between device_mutex and
the configfs root inode mutex.
Note: it is safe to pass a NULL pointer to config_item_put() and also to
config_item_get_unless_zero().
This patch prevents that lockdep reports the following complaint:
======================================================
WARNING: possible circular locking dependency detected
4.12.0-rc1-dbg+ #1 Not tainted
------------------------------------------------------
rmdir/12053 is trying to acquire lock:
(device_mutex#2){+.+.+.}, at: [<ffffffffa010afce>]
target_free_device+0xae/0xf0 [target_core_mod]
but task is already holding lock:
(&sb->s_type->i_mutex_key#14){++++++}, at: [<ffffffff811c5c30>]
vfs_rmdir+0x50/0x140
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (&sb->s_type->i_mutex_key#14){++++++}:
lock_acquire+0x59/0x80
down_write+0x36/0x70
configfs_depend_item+0x3a/0xb0 [configfs]
target_depend_item+0x13/0x20 [target_core_mod]
target_xcopy_locate_se_dev_e4_iter+0x87/0x100 [target_core_mod]
target_devices_idr_iter+0x16/0x20 [target_core_mod]
idr_for_each+0x39/0xc0
target_for_each_device+0x36/0x50 [target_core_mod]
target_xcopy_locate_se_dev_e4+0x28/0x80 [target_core_mod]
target_xcopy_do_work+0x2e9/0xdd0 [target_core_mod]
process_one_work+0x1ca/0x3f0
worker_thread+0x49/0x3b0
kthread+0x109/0x140
ret_from_fork+0x31/0x40
-> #0 (device_mutex#2){+.+.+.}:
__lock_acquire+0x101f/0x11d0
lock_acquire+0x59/0x80
__mutex_lock+0x7e/0x950
mutex_lock_nested+0x16/0x20
target_free_device+0xae/0xf0 [target_core_mod]
target_core_dev_release+0x10/0x20 [target_core_mod]
config_item_put+0x6e/0xb0 [configfs]
configfs_rmdir+0x1a6/0x300 [configfs]
vfs_rmdir+0xb7/0x140
do_rmdir+0x1f4/0x200
SyS_rmdir+0x11/0x20
entry_SYSCALL_64_fastpath+0x23/0xc2
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&sb->s_type->i_mutex_key#14);
lock(device_mutex#2);
lock(&sb->s_type->i_mutex_key#14);
lock(device_mutex#2);
*** DEADLOCK ***
3 locks held by rmdir/12053:
#0: (sb_writers#10){.+.+.+}, at: [<ffffffff811e223f>]
mnt_want_write+0x1f/0x50
#1: (&sb->s_type->i_mutex_key#14/1){+.+.+.}, at: [<ffffffff811cb97e>]
do_rmdir+0x15e/0x200
#2: (&sb->s_type->i_mutex_key#14){++++++}, at: [<ffffffff811c5c30>]
vfs_rmdir+0x50/0x140
stack backtrace:
CPU: 3 PID: 12053 Comm: rmdir Not tainted 4.12.0-rc1-dbg+ #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.0.0-prebuilt.qemu-project.org 04/01/2014
Call Trace:
dump_stack+0x86/0xcf
print_circular_bug+0x1c7/0x220
__lock_acquire+0x101f/0x11d0
lock_acquire+0x59/0x80
__mutex_lock+0x7e/0x950
mutex_lock_nested+0x16/0x20
target_free_device+0xae/0xf0 [target_core_mod]
target_core_dev_release+0x10/0x20 [target_core_mod]
config_item_put+0x6e/0xb0 [configfs]
configfs_rmdir+0x1a6/0x300 [configfs]
vfs_rmdir+0xb7/0x140
do_rmdir+0x1f4/0x200
SyS_rmdir+0x11/0x20
entry_SYSCALL_64_fastpath+0x23/0xc2
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
[Rebased to handle conflict withe target_find_device removal]
Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some target code uses config_item_name() while other code accesses .ci_name
directly. Make the target code consistent by switching to
config_item_name().
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: Mike Christie <mchristi@redhat.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fix warning:
smatch warnings:
drivers/target/target_core_user.c:301 tcmu_genl_cmd_done() warn: KERN_*
level not at start of string
Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch based on Xiubo's patches adds 2 tcmu attr to block and reset the
netlink interface. It's used during userspace daemon reinitialization after
the daemon has crashed while there is outstanding nl requests. The daemon
can block the nl interface, kill outstanding requests in the kernel and
then reopen the netlink socket and unblock it to allow new requests.
[mkp: typo]
Signed-off-by: Mike Christie <mchristi@redhat.com>
Tested-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some misc cleanup of the nl rework patches.
1. Fix space instead of tabs use and extra newline
2. Drop initializing variables to 0 when not needed
3. Just pass the skb_buff and msg_header pointers to
tcmu_netlink_event_send.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Tested-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Just return EBUSY if a nl request comes in while processing one. The upper
layers do not support sending multiple create/remove requests at the same
time (you cannot have a create and remove at the same time or do multiple
creates or removes at the same time) and doing a reconfig while a
create/remove is still executing does not make sense.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Tested-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The next patch is going to fix the hung nl command issue so this adds a
list of outstanding nl commands that we can later abort when the daemon is
restarted.
Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The sbitmap and the percpu_ida perform essentially the same task,
allocating tags for commands. The sbitmap outperforms the percpu_ida as
documented here: https://lkml.org/lkml/2014/4/22/553
The sbitmap interface is a little harder to use, but being able to remove
the percpu_ida code and getting better performance justifies the additional
complexity.
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com> # f_tcm
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
SPC5r17 states that the contents of the ADDITIONAL LENGTH field are not
altered based on the allocation length, so always calculate and pack the
full key list length even if the list itself is truncated.
According to Maged:
Yes it fixes the "Storage Spaces Persistent Reservation" test in the
Windows 2016 Server Failover Cluster validation suites when having
many connections that result in more than 8 registrations. I tested
your patch on 4.17 with iblock.
This behaviour can be tested using the libiscsi PrinReadKeys.Truncate test.
Cc: stable@vger.kernel.org
Signed-off-by: David Disseldorp <ddiss@suse.de>
Reviewed-by: Mike Christie <mchristi@redhat.com>
Tested-by: Maged Mokhtar <mmokhtar@petasan.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Since the TCMU_RING_SIZE macro is not using here will discard it and at the
same time clean up the code style.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Acked-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Generally target core and TCMUser seem to work fine for tape devices and
media changers. But there is at least one situation where TCMUser is not
able to support sequential access device emulation correctly.
The situation is when an initiator sends a SCSI READ CDB with a length that
is greater than the length of the tape block to read. We can distinguish
two subcases:
A) The initiator sent the READ CDB with the SILI bit being set.
In this case the sequential access device has to transfer the data from
the tape block (only the length of the tape block) and transmit a good
status. The current interface between TCMUser and the userspace does
not support reduction of the read data size by the userspace program.
The patch below fixes this subcase by allowing the userspace program to
specify a reduced data size in read direction.
B) The initiator sent the READ CDB with the SILI bit not being set.
In this case the sequential access device has to transfer the data from
the tape block as in A), but additionally has to transmit CHECK
CONDITION with the ILI bit set and NO SENSE in the sensebytes. The
information field in the sensebytes must contain the residual count.
With the below patch a user space program can specify the real read data
length and appropriate sensebytes. TCMUser then uses the se_cmd flag
SCF_TREAT_READ_AS_NORMAL, to force target core to transmit the real data
size and the sensebytes. Note: the flag SCF_TREAT_READ_AS_NORMAL is
introduced by Lee Duncan's patch "[PATCH v4] target: transport should
handle st FM/EOM/ILI reads" from Tue, 15 May 2018 18:25:24 -0700.
Signed-off-by: Bodo Stroesser <bstroesser@ts.fujitsu.com>
Acked-by: Mike Christie <mchristi@redhat.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull SCSI updates from James Bottomley:
"This is mostly updates to the usual drivers: ufs, qedf, mpt3sas, lpfc,
xfcp, hisi_sas, cxlflash, qla2xxx.
In the absence of Nic, we're also taking target updates which are
mostly minor except for the tcmu refactor.
The only real core change to worry about is the removal of high page
bouncing (in sas, storvsc and iscsi). This has been well tested and no
problems have shown up so far"
* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (268 commits)
scsi: lpfc: update driver version to 12.0.0.4
scsi: lpfc: Fix port initialization failure.
scsi: lpfc: Fix 16gb hbas failing cq create.
scsi: lpfc: Fix crash in blk_mq layer when executing modprobe -r lpfc
scsi: lpfc: correct oversubscription of nvme io requests for an adapter
scsi: lpfc: Fix MDS diagnostics failure (Rx < Tx)
scsi: hisi_sas: Mark PHY as in reset for nexus reset
scsi: hisi_sas: Fix return value when get_free_slot() failed
scsi: hisi_sas: Terminate STP reject quickly for v2 hw
scsi: hisi_sas: Add v2 hw force PHY function for internal ATA command
scsi: hisi_sas: Include TMF elements in struct hisi_sas_slot
scsi: hisi_sas: Try wait commands before before controller reset
scsi: hisi_sas: Init disks after controller reset
scsi: hisi_sas: Create a scsi_host_template per HW module
scsi: hisi_sas: Reset disks when discovered
scsi: hisi_sas: Add LED feature for v3 hw
scsi: hisi_sas: Change common allocation mode of device id
scsi: hisi_sas: change slot index allocation mode
scsi: hisi_sas: Introduce hisi_sas_phy_set_linkrate()
scsi: hisi_sas: fix a typo in hisi_sas_task_prep()
...
Pull block updates from Jens Axboe:
- clean up how we pass around gfp_t and
blk_mq_req_flags_t (Christoph)
- prepare us to defer scheduler attach (Christoph)
- clean up drivers handling of bounce buffers (Christoph)
- fix timeout handling corner cases (Christoph/Bart/Keith)
- bcache fixes (Coly)
- prep work for bcachefs and some block layer optimizations (Kent).
- convert users of bio_sets to using embedded structs (Kent).
- fixes for the BFQ io scheduler (Paolo/Davide/Filippo)
- lightnvm fixes and improvements (Matias, with contributions from Hans
and Javier)
- adding discard throttling to blk-wbt (me)
- sbitmap blk-mq-tag handling (me/Omar/Ming).
- remove the sparc jsflash block driver, acked by DaveM.
- Kyber scheduler improvement from Jianchao, making it more friendly
wrt merging.
- conversion of symbolic proc permissions to octal, from Joe Perches.
Previously the block parts were a mix of both.
- nbd fixes (Josef and Kevin Vigor)
- unify how we handle the various kinds of timestamps that the block
core and utility code uses (Omar)
- three NVMe pull requests from Keith and Christoph, bringing AEN to
feature completeness, file backed namespaces, cq/sq lock split, and
various fixes
- various little fixes and improvements all over the map
* tag 'for-4.18/block-20180603' of git://git.kernel.dk/linux-block: (196 commits)
blk-mq: update nr_requests when switching to 'none' scheduler
block: don't use blocking queue entered for recursive bio submits
dm-crypt: fix warning in shutdown path
lightnvm: pblk: take bitmap alloc. out of critical section
lightnvm: pblk: kick writer on new flush points
lightnvm: pblk: only try to recover lines with written smeta
lightnvm: pblk: remove unnecessary bio_get/put
lightnvm: pblk: add possibility to set write buffer size manually
lightnvm: fix partial read error path
lightnvm: proper error handling for pblk_bio_add_pages
lightnvm: pblk: fix smeta write error path
lightnvm: pblk: garbage collect lines with failed writes
lightnvm: pblk: rework write error recovery path
lightnvm: pblk: remove dead function
lightnvm: pass flag on graceful teardown to targets
lightnvm: pblk: check for chunk size before allocating it
lightnvm: pblk: remove unnecessary argument
lightnvm: pblk: remove unnecessary indirection
lightnvm: pblk: return NVM_ error on failed submission
lightnvm: pblk: warn in case of corrupted write buffer
...
When a tape drive is exported via LIO using the pscsi module, a read
that requests more bytes per block than the tape can supply returns an
empty buffer. This is because the pscsi pass-through target module sees
the "ILI" illegal length bit set and thinks there is no reason to return
the data.
This is a long-standing transport issue, since it assumes that no data
need be returned under a check condition, which isn't always the case
for tape.
Add in a check for tape reads with the ILI, EOM, or FM bits set, with a
sense code of NO_SENSE, treating such cases as if the read
succeeded. The layered tape driver then "does the right thing" when it
gets such a response.
Signed-off-by: Bodo Stroesser <bstroesser@ts.fujitsu.com>
Signed-off-by: Lee Duncan <lduncan@suse.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Problem:
$ cat /sys/kernel/config/target/core/user_0/block/attrib/qfull_time_out
-1
$ echo "-1" > /sys/kernel/config/target/core/user_0/block/attrib/qfull_time_out
-bash: echo: write error: Invalid argument
Fix:
This patch will help reset qfull_time_out to its default
i.e. qfull_time_out=-1.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Acked-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
There are two advantages:
* Direct I/O allows to avoid the write-back cache, so it reduces affects
to other processes in the system.
* Async I/O allows to handle a few commands concurrently.
DIO + AIO shows a better perfomance for random write operations:
Mode: O_DSYNC Async: 1
$ ./fio --bs=4K --direct=1 --rw=randwrite --ioengine=libaio --iodepth=64 --name=/dev/sda --runtime=20 --numjobs=2
WRITE: bw=45.9MiB/s (48.1MB/s), 21.9MiB/s-23.0MiB/s (22.0MB/s-25.2MB/s), io=919MiB (963MB), run=20002-20020msec
Mode: O_DSYNC Async: 0
$ ./fio --bs=4K --direct=1 --rw=randwrite --ioengine=libaio --iodepth=64 --name=/dev/sdb --runtime=20 --numjobs=2
WRITE: bw=1607KiB/s (1645kB/s), 802KiB/s-805KiB/s (821kB/s-824kB/s), io=31.8MiB (33.4MB), run=20280-20295msec
Known issue:
DIF (PI) emulation doesn't work when a target uses async I/O, because
DIF metadata is saved in a separate file, and it is another non-trivial
task how to synchronize writing in two files, so that a following read
operation always returns a consisten metadata for a specified block.
Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Signed-off-by: Andrei Vagin <avagin@openvz.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Reviewed-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Minor optimization - remove a pointer indirection when using fs_bio_set.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Switch everyone to blk_get_request_flags, and then rename
blk_get_request_flags to blk_get_request.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche
Mike Christie
Matthew Wilcox
David Disseldorp
Xiubo Li
bstroesser@ts.fujitsu.com
Kees Cook
Linus Torvalds
Kent Overstreet
Colin Ian King
Lee Duncan
Prasanna Kumar Kalever
Andrei Vagin
Christoph Hellwig