When a FSF request is returned with an error it should be reported
through blktrace for the ziomon tools, but the latency information
should not be read. Fix this by also calling zfcp_fsf_req_trace for
the error case, but skip reading the latencies inside the function.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
zfcp and qeth are setting flags for the qdio-layer, but these flags
are not used in qdio. Patch removes the flag definitions from qdio
and their settings in zfcp and qeth.
Cc: Jan Glauber <jang@linux.vnet.ibm.com>
Cc: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Move the qdio related structs and some helper functions to a new
zfcp_qdio.h header file. While doing this, rename the struct
zfcp_queue_req to zfcp_qdio_req to adhere to the naming scheme used in
zfcp. This allows a better seperation of the qdio code and inlining
the helper functions will save some function calls.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The FC4 types are already available from exchange port. Use this for
reporting the FC4 types, instead of having the value hardcoded in
zfcp.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Move the decision which trace tag and trace level to use for the scsi
result trace to zfcp_dbf.h. zfcp_dbf_scsi_result is already an inline
function, so move the trace code there, simplifying the response
handling in zfcp_fsf.c.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Kernel code uses dev as short name for the struct device. Rename the
sysfs_device in zfcp_unit and zfcp_port to match this convention.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
zfcp_fsf_req_create assigns the same value twice to req_seq_no.
Remove one assignment and move the req_id and seq_no assignments to
one place.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
0 is a valid value for a LUN. It is slightly confusing to also see 0
in the trace entries relating to adapter and port. Change this to use
0xFFFFFFFFFFFFFFFF in the LUN field when the trace entry does not
relate to a LUN or unit.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The smatch tool from http://repo.or.cz/w/smatch.git warns about this:
drivers/s390/scsi/zfcp_scsi.c +64 zfcp_scsi_command_fail(5) warn: variable dereferenced before check 'scpnt->device'
drivers/s390/scsi/zfcp_scsi.c +64 zfcp_scsi_command_fail(5) warn: variable dereferenced before check 'scpnt->device->host'
drivers/s390/scsi/zfcp_scsi.c +93 zfcp_scsi_queuecommand(23) warn: variable dereferenced before check 'unit'
Fix the first two warnings by removing the checks for scpnt->device
and -> host: As long as the SCSI command exists, there is also a
scsi_device and a Scsi_Host.
Fix the last warning by removing the BUG_ON checks in
zfcp_scsi_queuecommand, they are leftovers from previous paranoia
about wrong pointers between data structures.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Remove the unused payload field from the struct zfcp_dbf_san_record,
saving some space in the SAN trace.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
On a link down, the adapter reopen is not strictly necessary, but it
helps flushing pending requests as quickly as possible. Add a comment
mentioning this.
qdio returning a problem on the response queue is an unlikely event.
The recovery mentioned in the comment might resolve it, so implement
it. This also has the advantage that it creates an entry in the
recovery trace to see if and when this is occurring.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Move the code for tracking FSF requests to new file to have this code
in one place. The functions for adding and removing requests on the
I/O path are already inline. The alloc and free functions are only
called once, so it does not hurt to inline them and add them to the
same file.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Always use the FSF request id as a reference to the FSF request. With
this change the function zfcp_reqlist_find_safe is no longer needed
and can be removed.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The status FC_CTELS_STATUS_REJECT for all FC BSG errors is not
appropriate. Instead, report -EIO in the result field if there was a
problem in zfcp with the FC BSG request. If the request is good from
our point of view, report result 0, status FC_CTELS_STATUS_OK and let
userspace read the Accept or Reject from the payload (as documented in
scsi_bsg_fc.h).
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The hardware used with zfcp provides a timer for CT and ELS requests
instead of an abort capability for these commands. To correctly handle
the FC BSG timeouts, pass the timeout from the BSG requests to the
hardware.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Introduce a zfcp callback for timeouts triggered from FC BSG. With
zfcp, the underlying hardware cannot abort CT or ELS requests, so
there is nothing to do when the block layer timeout expires. To avoid
interference with the block layer timeout, simply indicate that the
block layer timer should be reset. The timer running in the hardware
for the pending CT or ELS request will return the request when it
expires.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Advance the correct pointer when inserting the linebreak for the HBA
trace. It was missing in the output since the pointer to the output
buffer was never advanced, and the linebreak character was overwritten
later.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The patch "zfcp: Simplify handling of ct and els requests"
accidentally removed the call to zfcp_fc_wka_port_put for FC CT BSG
requests, thus not issuing a "close" request for the WKA ports.
Introduce a CT specific handler to first call zfcp_fc_wka_port_put and
then continue with the generic handler when returning from FC CT BSG
requests.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The flag ZFCP_STATUS_FSFREQ_TMFUNCNOTSUPP is never set and hence can
be removed. This is a leftover from the time when zfcp had to decide
whether the target supports a "logical unit reset" or not. Nowadays,
the SCSI midlayer calls the eh_device_reset_handler or the
eh_target_reset_handler and zfcp simply maps this to a "logical unit
reset" or a "target reset".
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Enable the display of supported and active fc4s for zfcp in the FC
transport class. zfcp only supports FCP, so simply hard-code this
information. The zfcp hbaapi already has this information hardcoded,
but this would allow to switch from the coding in the zfcp hbaapi to
the common FC transport attributes in the future.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
In case the SCSI error recovery starts because of a SCSI command
timeout, but then something else triggers the rport to be deleted, the
SCSI error recovery will run to the end and set the SCSI device
offline. To prevent this, call the FC transport function
fc_block_scsi_eh which waits until the rport leaves the BLOCKED state.
This guarantees that communication is possible if the rport is ONLINE,
or the SCSI devices will be removed if the rport state switches to
NOT_PRESENT.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The SCSI midlayer retries commands based on the remote port state and
the command status reported by the driver. Returning
DID_TRANSPORT_DISRUPTED is a better approach, use this for reporting
FSF errors back to the SCSI midlayer. See
http://marc.info/?l=linux-scsi&m=125668044215051&w=2 as reference.
There is also no need in special treatment of ABORTED commands, so
remove the ZFCP_STATUS_FSFREQ_ABORTED, the commands are then returned
with DID_TRANSPORT_DISRUPTED.
Also remove the ZFCP_STATUS_FSFREQ_RETRY: It is useless, no retry is
happening in the FSF layer and nobody checks the state of this flag.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Introduce kmem_cache for ELS ADISC data to guarantee the required
hardware alignment and free the allocated memory in case the send
failes.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Remove some redundancies in FC related code and trace:
- drop redundant data from SAN trace (local s_id that only changes
during link down, ls_code that is already part of payload, d_id in
ct response trace that is always the same as in ct request trace)
- use one common fsf struct to hold zfcp data for ct and els requests
- leverage common fsf struct for FC passthrough job data, allocate it
with dd_bsg_data for passthrough requests and unify common code for
ct and els passthrough request
- simplify callback handling in zfcp_fc
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Instead of assigning 4 bytes with the highest byte masked out, use a 3
byte array with the ntoh24 and h24ton helper functions, thus
eliminating the need for the ZFCP_DID_MASK.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The well-known-address (WKA) port handling code is part of the FC code
in zfcp. Move everything WKA related to the zfcp_fc files and use the
common zfcp_fc prefix for structs and functions. Drop the unused key
management service while renaming the struct, no request could ever
reach this service in zfcp and it is obsolete anyway.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Use common code definitions for FC GPN_FT and GID_PN
instead of inventing private ones. Move the private structs still
required inside zfcp to zfcp_fc header file.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Use common code definitions for FC plogi, logo, rscn and adisc structs
instead of inventing private ones. Move the private struct for issuing
ELS ADISC inside zfcp to zfcp_fc header file.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Use common data structures for FCP CMND, FCP RSP and related
definitions and remove zfcp private definitions. Split the FCP CMND
setup and FCP RSP evaluation code in seperate functions. Use inline
functions to not negatively impact the I/O path.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
If an error occurs that triggers the call to fc_remote_port_delete,
ideally this call would happen before any I/O is passed back to the
SCSI midlayer through scsi_done. The SCSI midlayer will retry the
commands and fc_remote_port_chkready will return the correct status
code. But with the delay between calling scsi_done in softirq context
and the call to fc_remote_port_delete from the workqueue, there is a
window where zfcp returns DID_ERROR. This leads to SCSI error recovery
which then leads to offline SCSI devices since all recovery actions
will fail with the rport now being blocked.
In this window, zfcp has to return DID_IMM_RETRY just as the FC
transport class would do in fc_remote_port_chkready for the blocked
fc_rport. As soon as the fc_rport is BLOCKED, fc_remote_port_chkready
will do the right thing.
Additionally, there are two more cases to catch in zfcp_scsi_queuecommand:
- After the port has been opened, the unit has to be opened. During
this period I/O has to be retried. This can also be handled with
DID_IMM_RETRY.
- If the access to the unit fails, but the port is good, then
this single unit cannot be accessed and I/O to this unit has to fail
without involving the FC transport class.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The port_scan work was scheduled to the work_queue provided by the
kernel. This resulted on SMP systems to a likely situation that more
than one scan_work were processed in parallel. This is not required
and openes the possibility of race conditions between the removal of
invalid ports and the enqueue of just scanned ports. This patch
synchronizes the scan_work tasks by scheduling them to adapter local
work_queue.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The flag ZFCP_STATUS_COMMON_REMOVE was used to indicate that a
resource is not ready to be used or about to be removed from the
system. This is now better done by an improved list handling
and therefore the additional indicator is not required anymore.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
With the reference counting for zfcp data structures, it is now
possible to implement module unloading again. Module unloading
requires to free all data structures in the module exit function. This
is done by unregistering zfcp from s390 cio and the SCSI midlayer
first in the module exit function.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The latencies traced per fsf request are traced for sysfs output and
for blktrace, each in one function. Simplify the tracing code by
merging both tracing functions into one.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
When accessing port and unit attributes, use container_of instead of
dev_get_drvdata. This eliminates some code checker warnings about
aliased access of data structures.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The callback for suspend is not required because it contains exactly
the same functionality as the _set_offline routine does.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The global config_mutex was required for the serialization of a
configuration change within the zfcp driver. This global locking is
now obsolete and can be removed. The requirement of serializing the
access to a zfcp_adapter reference via a ccw_device is realized wth a
static spinlock.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Replace the local reference counting by already available mechanisms
offered by kref. Where possible existing device structures were used,
including the same functionality.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The global config_lock was used to protect the configuration organized
in independent lists. It is not necessary to have a lock on driver
level for this purpose. This patch replaces the global config_lock
with a set of local list locks.
Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Adapt the change_queue_depth callback in zfcp for the new reason
parameter. Simply pass each call back to the SCSI midlayer, there are
no resource adjustments necessary for zfcp.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Removes check for (depth <= default_depth) in case of
SCSI_QDEPTH_RAMP_UP call back, not needed after added
max_queue_depth per sdev.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
This patch modifies scsi_host_template->change_queue_depth so that
it takes an argument indicating why it is being called. This will be
used so that if a LLD needs to do some extra processing when
handling queue fulls or later ramp ups, it can do so.
This is a simple port of the drivers setting a change_queue_depth
callback. In the patch I just have these LLDs adjust the queue depth
if the user was requesting it.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
[Vasu.Dev: v2
Also converted pmcraid_change_queue_depth and then verified
all modules compile using "make allmodconfig" for any new build
warnings on X86_64.
Updated original description after combing two original
patches from Mike to make this patch git bisectable.]
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
[jejb: fixed up 53c700]
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
When configuring a LUN for use in zfcp, flush the SCSI work to ensure
the SCSI device has been created before returning. This means that a
configuration procedure can run these commands in a script and the
SCSI device is available immediately after the unit_add:
echo 1 > /sys/bus/ccw/drivers/zfcp/0.0.181d/online
echo 0x401040C300000000 > \
/sys/bus/ccw/drivers/zfcp/0.0.181d/0x500507630313c562/unit_add
lsscsi
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Add HZ since the start_timer function expects jiffies, not seconds.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
After opening a remote port zfcp checks if the WWPN returned in the
PLOGI maches the WWPN of the port that should have been opened. On a
mismatch zfcp assumes that the DID just changed, queries the FC
nameserver and tries again. If the situation persists the erp will
give up.
With this strategy, if the remote port always returns the wrong PLOGI
data, the remote port will not be opened. Introduce a warning, so that
the system administrator knows why the remote port is not being opened
and to have a pointer to investigate the problem on the storage
system.
Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
For ports, zfcp gets the DID from the FC nameserver and tries to open
the port. If the open succeeds, zfcp compares the WWPN from the
nameserver with the WWPN in the PLOGI payload. In case of a mismatch,
zfcp assumes that the DID of the port just changed and we opened the
wrong port. This means that zfcp has to forget the DID, lookup the DID
again and retry.
This error case had a problem that zfcp forgets the DID, but never
looks up a new one, stalling the ERP in this case. Fix this by
triggering the DID lookup and properly exit from the ERP. The DID
lookup will trigger a new ERP action.
Also ensure when trying to open the port again with the new DID, first
close the open port, even in the NOESC case.
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The pointer that is allocated with kmalloc() is passed to strsep()
which modifies it. Later on the modified pointer value will be passed
to kfree. Save the original pointer and pass that one to kfree
instead.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Running chchp --vary 0 and chccwdev -d on a FCP device with scsi
devices attached can lead to this thread hanging:
================================================================
STACK TRACE FOR TASK: 0x2fbfcc00 (kslowcrw)
STACK:
0 schedule+1136 [0x45f99c]
1 schedule_timeout+534 [0x46054e]
2 wait_for_common+374 [0x45f442]
3 blk_execute_rq+160 [0x217a2c]
4 scsi_execute+278 [0x26daf2]
5 scsi_execute_req+150 [0x26dc86]
6 sd_sync_cache+138 [0x28460a]
7 sd_shutdown+130 [0x28486a]
8 sd_remove+104 [0x284c84]
9 __device_release_driver+152 [0x257430]
10 device_release_driver+56 [0x2575c8]
11 bus_remove_device+214 [0x25672a]
12 device_del+352 [0x25456c]
13 __scsi_remove_device+108 [0x272630]
14 scsi_remove_device+66 [0x2726ba]
15 zfcp_ccw_remove+824 [0x335558]
16 ccw_device_remove+62 [0x2b3f2a]
17 __device_release_driver+152 [0x257430]
18 device_release_driver+56 [0x2575c8]
19 bus_remove_device+214 [0x25672a]
20 device_del+352 [0x25456c]
21 ccw_device_unregister+92 [0x2b48c4]
22 io_subchannel_remove+108 [0x2b4950]
23 css_remove+62 [0x2af7ee]
24 __device_release_driver+152 [0x257430]
25 device_release_driver+56 [0x2575c8]
26 bus_remove_device+214 [0x25672a]
27 device_del+352 [0x25456c]
28 device_unregister+38 [0x25464a]
29 css_sch_device_unregister+68 [0x2af97c]
30 ccw_device_call_sch_unregister+78 [0x2b581e]
31 worker_thread+604 [0x69eb0]
32 kthread+154 [0x6ff42]
33 kernel_thread_starter+6 [0x1c952]
================================================================
The problem is that the chchp --vary 0 leads to zfcp first calling
fc_remote_port_delete which blocks all scsi devices on the remote
port. Calling scsi_remove_device later lets the sd driver issue a
SYNCHRONIZE_CACHE command. This command stays on the "stopped" request
requeue because the SCSI device is blocked. Fix this by first removing
the scsi and fc hosts which removes all scsi devices and do not use
scsi_remove_device.
Reviewed-by: Felix Beck <felix.beck@de.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.31-39.x.20090917-s390xdefault #1
-------------------------------------------------------
kslowcrw/83 is trying to acquire lock:
(&adapter->scan_work){+.+.+.}, at: [<0000000000169c5c>] __cancel_work_timer+0x64/0x3d4
but task is already holding lock:
(&zfcp_data.config_mutex){+.+.+.}, at: [<00000000004671ea>] zfcp_ccw_remove+0x66/0x384
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (&zfcp_data.config_mutex){+.+.+.}:
[<0000000000189962>] __lock_acquire+0xe26/0x1834
[<000000000018a4b6>] lock_acquire+0x146/0x178
[<000000000058cb5a>] mutex_lock_nested+0x82/0x3ec
[<0000000000477170>] zfcp_fc_scan_ports+0x3ec/0x728
[<0000000000168e34>] worker_thread+0x278/0x3a8
[<000000000016ff08>] kthread+0x9c/0xa4
[<0000000000109ebe>] kernel_thread_starter+0x6/0xc
[<0000000000109eb8>] kernel_thread_starter+0x0/0xc
-> #0 (&adapter->scan_work){+.+.+.}:
[<0000000000189e60>] __lock_acquire+0x1324/0x1834
[<000000000018a4b6>] lock_acquire+0x146/0x178
[<0000000000169c9a>] __cancel_work_timer+0xa2/0x3d4
[<0000000000465cb2>] zfcp_adapter_dequeue+0x32/0x14c
[<00000000004673e4>] zfcp_ccw_remove+0x260/0x384
[<00000000004250f6>] ccw_device_remove+0x42/0x1ac
[<00000000003cb6be>] __device_release_driver+0x9a/0x10c
[<00000000003cb856>] device_release_driver+0x3a/0x4c
[<00000000003ca94c>] bus_remove_device+0xcc/0x114
[<00000000003c8506>] device_del+0x162/0x21c
[<0000000000425ff2>] ccw_device_unregister+0x5e/0x7c
[<000000000042607e>] io_subchannel_remove+0x6e/0x9c
[<000000000041ff9a>] css_remove+0x3e/0x7c
[<00000000003cb6be>] __device_release_driver+0x9a/0x10c
[<00000000003cb856>] device_release_driver+0x3a/0x4c
[<00000000003ca94c>] bus_remove_device+0xcc/0x114
[<00000000003c8506>] device_del+0x162/0x21c
[<00000000003c85e8>] device_unregister+0x28/0x38
[<0000000000420152>] css_sch_device_unregister+0x46/0x58
[<00000000004276a6>] io_subchannel_sch_event+0x28e/0x794
[<0000000000420442>] css_evaluate_known_subchannel+0x46/0xd0
[<0000000000420ebc>] slow_eval_known_fn+0x88/0xa0
[<00000000003caffa>] bus_for_each_dev+0x7e/0xd0
[<000000000042188c>] for_each_subchannel_staged+0x6c/0xd4
[<0000000000421a00>] css_slow_path_func+0x54/0xd8
[<0000000000168e34>] worker_thread+0x278/0x3a8
[<000000000016ff08>] kthread+0x9c/0xa4
[<0000000000109ebe>] kernel_thread_starter+0x6/0xc
[<0000000000109eb8>] kernel_thread_starter+0x0/0xc
cancel_work_sync is called while holding the config_mutex. But the
work that is being cancelled or flushed also uses the config_mutex.
Fix the resulting deadlock possibility by calling cancel_work_sync
earlier without holding the mutex. The best place to do is is after
offlining the device. No new port scan work will be scheduled for the
offline device, so this is a safe place to call cancel_work_sync.
Reviewed-by: Felix Beck <felix.beck@de.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Christof Schmitt