The IO and Abort latency counter counts the time taken to complete the
IO and abort command into broad buckets. This is not intended for
performance measurement, just a debug statistic. current_max_io_time
tries to keep track of the maximum time an IO has taken to complete if
it is > 30sec.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Just a simple counter of number of check conditions encountered on that
host.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If SCSI-ML has already issued abort on a command i.e
FNIC_IOREQ_ABTS_PENDING is set and we get a IO completion, avoid this
being flagged as out-of-order completion by setting the FNIC_IO_DONE
flag in fnic_fcpio_icmnd_cmpl_handler
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fixing the IO stats update (Active IOs and IO completion) to prevent
"Number of Active IOs" from becoming negative in the fnistats output.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Getting rid of else case to make the flow look bit simpler.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This fix is to avoid calling fnic_fw_reset_handler through
fnic_host_reset when a finc reset is alreay in progress.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The ->lport_reset callback only ever had one implementation,
which already is exported. So remove it and use the function
directly.
Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When issuing I/O we check if rport is online through libfc
rport_lookup() function which needs to be protected by mutex lock that
cannot acquired in I/O context. The change is to use midlayer remote
port s dd_data which is preserved until its devloss timeout and no
protection is required. The the scsi_cmnd error code is expected to be
in the left 16 bits of the result field. Changed to correct this. Fnic
driver version changed from 1.6.0.20 to 1.6.0.21
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Reviewed-by: Ewan Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In case of LUN reset, the device reset command is issued with one of the
I/Os that has timed out on that LUN. The change is to also return this
I/O with error status set to DID_RESET. In case when the reset is issued
using the sg_reset tool (from sg3_utils) it is a new command and new_sc
is set to 1. Fnic driver version changed from 1.6.0.19 to 1.6.0.20
[mkp: Fixed checkpatch warning]
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Reviewed-by: Ewan Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If an I/O times out and an abort issued by host, if the abort is
successful we need to set scsi status as DID_ABORT. Or else the
mid-layer error handler which looks for this error code, will offline
the device. Also if the original I/O is not found in fnic firmware, we
will consider the abort as successful. The start_time assignment is
moved because of the new goto. Fnic driver version changed from
1.6.0.17a to 1.6.0.19, version 1.6.0.18 has been skipped
[mkp: Fixed checkpatch warning]
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
the kernel prints some warnings when compiled with CONFIG_DMA_API_DEBUG.
This is because the fnic driver doesn't check the return value of
pci_map_single().
[ 11.942770] scsi host12: fnic
[ 11.950811] ------------[ cut here ]------------
[ 11.950818] WARNING: at lib/dma-debug.c:937 check_unmap+0x47b/0x920()
[ 11.950821] fnic 0000:0c:00.0: DMA-API: device driver failed to check map error[device address=0x0000002020a30040] [size=44 bytes] [mapped as single]
Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed By: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: James Bottomley <JBottomley@Odin.com>
We added changes in fnic driver patch 1.6.0.16 to acquire
io_req_lock in fnic_queuecommand() before issuing I/O so that io completion
is serialized. But when releasing the lock we check for the I/O flag and
this could be modified if IO abort occurs before I/O completion. In this case
we wont release the lock and causes deadlock in some scenerios. Using the
local variable to check the IO lock status will resolve the problem.
Fixes: 41df7b02db
Signed-off-by: Hiral Shah <hishah@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Anil Chintalapati <achintal@cisco.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: James Bottomley <JBottomley@Odin.com>
When I/O is aborted by mid-layer, fnic FW will complete the I/O before
completing the abort task. In some cases abort request is completed before
the I/O, which could lead to inconsistent driver and firmware states.
In this case firmware reset would clear the inconsistent state.
Signed-off-by: Anil Chintalapati <achintal@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Hiral Shah <hishah@cisco.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When issuing I/O request, if the I/O completes before returning from
fnic_queuecommand(), we may be referencing scsi_cmnd structure that may
be freed by interrupt handler. Acquring IO lock would synchronize
fnic_queuecommand and interrupt handler.
- Increment fnic version from 1.6.0.15 to 1.6.0.16
Signed-off-by: Hiral Shah <hishah@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Anil Chintalapati <achintal@cisco.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
IOs belonging to an rport are aborted with Internal terminate option
when rport goes offline. Any new IO issued to the rport during this
time can reuse the terminated exchange which will cause inconsistent
state of the exchange between local port and remote port.
fc_rport_priv is set to RPORT_ST_DELETE before exchanges are aborted by
libfc. Not issuing amy more I/O requests when RPORT_ST_DELETE is set,
will avoid inconsistent state of the exchange between local port and
remote port.
- Increment fnic version from 1.6.0.13 to 1.6.0.14
Signed-off-by: Hiral Shah <hishah@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Anil Chintalapati <achintal@cisco.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Unless we want to build a SPI tag message we should just check SCMD_TAGGED
instead of reverse engineering a tag type through the use of
scsi_populate_tag_msg.
Also rename the function to spi_populate_tag_msg, make it behave like the
other spi message helpers, and move it to the spi transport class.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Current the midlayer fakes up a struct request for the explicit reset
ioctls, and those don't have a tag allocated to them. The fnic driver pokes
into midlayer structures to paper over this design issue, but that won't
work for the blk-mq case.
Either someone who can actually test the hardware will have to come up with
a similar hack for the blk-mq case, or we'll have to bite the bullet and fix
the way the EH ioctls work for real, but until that happens we fail these
explicit requests here.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Webb Scales <webbnh@hp.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Tested-by: Bart Van Assche <bvanassche@acm.org>
Tested-by: Robert Elliott <elliott@hp.com>
Cc: Hiral Patel <hiralpat@cisco.com>
Cc: Suma Ramars <sramars@cisco.com>
Cc: Brian Uchino <buchino@cisco.com>
The SCSI standard defines 64-bit values for LUNs, and large arrays
employing large or hierarchical LUN numbers become more and more
common.
So update the linux SCSI stack to use 64-bit LUN numbers.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Ewan Milne <emilne@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
In fnic abort handler, abort queuing can be failed when hardware queue is full.
The command state is left as abort queued. The command with abort queued state
will never be queued next time for abort or termiantion.
Fix restores the command state in above case.
Signed-off-by: Hiral Shah <hishah@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
This feature gathers active and cumulative per fnic stats for io,
abort, terminate, reset, vlan discovery path and it also includes
various important stats for debugging issues. It also provided
debugfs and ioctl interface for user to retrieve these stats.
It also provides functionality to reset cumulative stats through
user interface.
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Fixed appropriate error codes that returns negative error number on failure,
and 0 on success. fnic_reset() is used directly by the fc transport callback
issue_fc_host_lip which requires a negative error number on failure.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Introduced module params to provide dynamic way of configuring
queue depth.
Added support to get max io throttle count through UCSM to
configure maximum outstanding IOs supported by fnic and push
that value to scsi mid-layer.
Supported IO throttle values:
UCSM IO THROTTLE VALUE FNIC MAX OUTSTANDING IOS
------------------------------------------------------
16 (Default) 2048
<= 256 256
> 256 <ucsm value>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Kernel panics due to NULL lport while executing the log message because
of synchronization issues between libfc and scsi transport fc. Checking
for NULL pointers at the beginning of this routine would resolve the issue
from kernel panic point of view.
Signed-off-by: Sesidhar Baddel <sebaddel@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Hitting BUG_ON(io_req->abts_done) in fnic_rport_exch_reset in case of
timing issue and also to some extent locking issue where abts and terminate
is happening around same timing.
The code changes are intended to update CMD_STATE(sc) and
io_req->abts_done together.
Signed-off-by: Sesidhar Beddel <sebaddel@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Remove fnic driver QUEUE_FULL handling code instead let SCSI mid layer
handle queue full and use its algorithm to ramp down/up queue
Signed-off-by: Suma Ramars <sramars@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
There is an unlock missing if the == FNIC_IOREQ_ABTS_PENDING is
false.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Fnic Trace utility is a tracing functionality built directly into fnic driver
to trace events. The benefit that trace buffer brings to fnic driver is the
ability to see what it happening inside the fnic driver. It also provides the
capability to trace every IO event inside fnic driver to debug panics, hangs
and potentially IO corruption issues. This feature makes it easy to find
problems in fnic driver and it also helps in tracking down strange bugs in a
more manageable way. Trace buffer is shared across all fnic instances for
this implementation.
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Added new fnic debug flags for identifying IO state at every stage of IO while
debugging and also added more log messages for better debugging capability.
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
The issue was observed when LUN Reset is issued through IOCTL or sg_reset
utility.
fnic driver issues LUN RESET to firmware. On successful completion of device
reset, driver cleans up all the pending IOs that were issued prior to device
reset. These pending IOs are expected to be in ABTS_PENDING state. This works
fine, when the device reset operation resulted from midlayer, but not when
device reset was triggered from IOCTL path as the pending IOs were not in
ABTS_PENDING state. execution path hits panic if the pending IO is not in
ABTS_PENDING state.
Changes:
The fix replaces BUG_ON check in fnic_clean_pending_aborts() with marking
pending IOs as ABTS_PENDING if they were not in ABTS_PENDING state and skips
if they were already in ABTS_PENDING state. An extra check is added to validate
the abort status of the commands after a delay of 2 * E_D_TOV using a
helper function. The helper function returns 1 if it finds any pending IO in
ABTS_PENDING state, belong to the LUN on which device reset was issued else 0.
With this, device reset operation returns success only if the helper funciton
returns 0, otherwise it returns failure.
Other changes:
- Removed code in fnic_clean_pending_aborts() that returns failure if it finds
io_req NULL, instead of returning failure added code to continue with next io
- Added device reset flags for debugging in fnic_terminate_rport_io,
fnic_rport_exch_reset, and fnic_clean_pending_aborts
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
1. Handling overlapped firmware resets
This fix serialize multiple firmware resets to avoid situation where fnic
device fails to come up for link up event, when firmware resets are issued
back to back. If there are overlapped firmware resets are issued,
the firmware reset operation checks whether there is any firmware reset in
progress, if so it polls for its completion in a loop with 100ms delay.
2. Handling device reset timeout
fnic_device_reset code has been modified to handle Device reset timeout:
- Issue terminate on device reset timeout.
- Introduced flags field (one of the scratch fields in scsi_cmnd).
With this, device reset request would have DEVICE_RESET flag set for other
routines to determine the type of the request.
Also modified fnic_terminate_rport_io, fnic_rport_exch_rset, completion
routines to handle SCSI commands with DEVICE_RESET flag.
3. LUN/Device Reset hangs when issued through IOCTL using utilities like
sg_reset.
Each SCSI command is associated with a valid tag, fnic uses this tag to
retrieve associated scsi command on completion. the LUN/Device Reset issued
through IOCTL resulting into a SCSI command that is not associated with a
valid tag. So fnic fails to retrieve associated scsi command on completion,
which causes hang. This fix allocates tag, associates it with the
scsi command and frees the tag, when the operation completed.
4. Preventing IOs during firmware reset.
Current fnic implementation allows IO submissions during firmware reset.
This fix synchronizes IO submissions and firmware reset operations.
It ensures that IOs issued to fnic prior to reset will be issued to the
firmware before firmware reset.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Driver was incorrectly using the SLAB_CACHE_DMA flag when creating a cache
for SGLs. fnic device does not have 24-bit DMA restrictions. Remove the flag
and allocations from ZONE_DMA.
Thanks to Roland Dreier and David Rientjes for pointing out the bug.
Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Move the mid-layer's ->queuecommand() invocation from being locked
with the host lock to being unlocked to facilitate speeding up the
critical path for drivers who don't need this lock taken anyway.
The patch below presents a simple SCSI host lock push-down as an
equivalent transformation. No locking or other behavior should change
with this patch. All existing bugs and locking orders are preserved.
Additionally, add one parameter to queuecommand,
struct Scsi_Host *
and remove one parameter from queuecommand,
void (*done)(struct scsi_cmnd *)
Scsi_Host* is a convenient pointer that most host drivers need anyway,
and 'done' is redundant to struct scsi_cmnd->scsi_done.
Minimal code disturbance was attempted with this change. Most drivers
needed only two one-line modifications for their host lock push-down.
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Acked-by: James Bottomley <James.Bottomley@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In fnic_abort_cmd() and fnic_device_reset() assign `rport' earlier to make
FNIC_SCSI_DBG() calls cleaner.
In fnic_clean_pending_aborts() `rport' is not used.
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Robert Love <robert.w.love@intel.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>
Use libfcoe as a common FIP implementation with fcoe.
FIP or non-FIP mode is fully automatic if the firmware
supports and enables it.
Even if FIP is not supported, this uses libfcoe for the non-FIP
handling of FLOGI and its response.
Use the new lport_set_port_id() notification to capture
successful FLOGI responses and port_id resets.
While transitioning between Ethernet and FC mode, all rx and
tx FC frames are queued. In Ethernet mode, all frames are
passed to the exchange manager to capture FLOGI responses.
Change to set data_src_addr to the ctl_src_addr whenever it
would have previously been zero because we're not logged in.
This seems safer so we'll never send a frame with a 0 source MAC.
This also eliminates a special case for sending FLOGI frames.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Move the duplicated code from FC LLDs to SCSI FC transport class.
Acked-by: James Smart <james.smart@emulex.com>
Acked-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Acked-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The shost sg tablesize is set to FNIC_MAX_SG_DESC_CNT and fnic uses
scsi_dma_map, so both BUG_ONs can be removed.
scsi_dma_map may return -ENOMEM, sg_count should be int to catch that.
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
fnic is a driver for the Cisco PCI-Express FCoE HBA
Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
