26 Jul, 2014
1 commit
-
Seems like these counters are missing any sort of synchronization for
updates, as a over 10 year old comment from me noted. Fix this by
using atomic counters, and while we're at it also make sure they are
in the same cacheline as the _busy counters and not needlessly stored
to in every I/O completion.With the new model the _busy counters can temporarily go negative,
so all the readers are updated to check for > 0 values. Longer
term every successful I/O completion will reset the counters to zero,
so the temporarily negative values will not cause any harm.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott
25 Jul, 2014
8 commits
-
Avoid taking the queue_lock to check the per-device queue limit. Instead
we do an atomic_inc_return early on to grab our slot in the queue,
and if necessary decrement it after finishing all checks.Unlike the host and target busy counters this doesn't allow us to avoid the
queue_lock in the request_fn due to the way the interface works, but it'll
allow us to prepare for using the blk-mq code, which doesn't use the
queue_lock at all, and it at least avoids a queue_lock round trip in
scsi_device_unbusy, which is still important given how busy the queue_lock
is.Signed-off-by: Christoph Hellwig
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
Avoid taking the host-wide host_lock to check the per-host queue limit.
Instead we do an atomic_inc_return early on to grab our slot in the queue,
and if necessary decrement it after finishing all checks.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
Avoid taking the host-wide host_lock to check the per-target queue limit.
Instead we do an atomic_inc_return early on to grab our slot in the queue,
and if necessary decrement it after finishing all checks.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
Prepare for not taking a host-wide lock in the dispatch path by pushing
the lock down into the places that actually need it. Note that this
patch is just a preparation step, as it will actually increase lock
roundtrips and thus decrease performance on its own.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
The blk-mq code path will set this to a different function, so make the
code simpler by setting it up in a legacy-request specific place.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
Make sure we only have the logic for requeing commands in one place.
Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
Factor out a helper to set the _blocked values, which we'll reuse for the
blk-mq code path.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke
Reviewed-by: Webb Scales
Acked-by: Jens Axboe
Tested-by: Bart Van Assche
Tested-by: Robert Elliott -
Factor out command setup code that will be shared with the blk-mq code path.
Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Webb Scales
18 Jul, 2014
31 commits
-
Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
Factor out a function to initialize regular read/write commands and leave
sd_init_command as a simple dispatcher to the different prepare routines.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
Currently cmd->allowed is initialized from rq->retries for discard
commands, but retries is always 0 for non-BLOCK_PC requests. Set it
to the standard number of retries instead.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
Currently cmd->allowed is initialized from rq->retries for write same
commands, but retries is always 0 for non-BLOCK_PC requests. Set it
to the standard number of retries instead.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
Simplify handling of discard requests by setting up the command directly
instead of initializing request fields and then calling
scsi_setup_blk_pc_cmnd to propagate the information into the command.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
Simplify handling of write same requests by setting up the command directly
instead of initializing request fields and then calling
scsi_setup_blk_pc_cmnd to propagate the information into the command.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Webb Scales -
Simplify handling of flush requests by setting up the command directly
instead of initializing request fields and then calling
scsi_setup_blk_pc_cmnd to propagate the information into the command.Also rename scsi_setup_flush_cmnd to sd_setup_flush_cmnd for consistency.
Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
The data direction fiel in the SCSI command is derived only from the block
request structure. Move setting it up into common code instead of
duplicating it in the ULDs.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
We should call the device handler prep_fn for all TYPE_FS requests,
not just simple read/write calls that are handled by the disk driver.Restructure the common I/O code to call the prep_fn handler and zero
out the CDB, and just leave the call to scsi_init_io to the ULDs.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
scsi_init_io should only be called for requests that transfer data,
so move the assert that a request has segments from the callers into
scsi_init_io.Signed-off-by: Christoph Hellwig
Reviewed-by: Martin K. Petersen
Reviewed-by: Hannes Reinecke -
During IO with fabric faults, one generally sees several "Unhandled error
code" messages in the syslog as shown below:sd 4:0:6:2: [sdbw] Unhandled error code
sd 4:0:6:2: [sdbw] Result: hostbyte=DID_NO_CONNECT driverbyte=DRIVER_OK
sd 4:0:6:2: [sdbw] CDB: Read(10): 28 00 00 00 00 00 00 00 08 00
end_request: I/O error, dev sdbw, sector 0This comes from scsi_io_completion (in scsi_lib.c) while handling error
codes other than DID_RESET or not deferred sense keys i.e. this is
actually handled by the SCSI mid layer. But what gets displayed here is
"Unhandled error code" which is quite misleading as it indicates
something that is not addressed by the mid layer.The description string is based on the sense key and sometimes on the
additional sense code;
since the ACTION_FAIL case always prints the sense key and the
additional sense code, this patch removes the description string
completely because it does not add useful information.Signed-off-by: Maurizio Lombardi
Reviewed-by: Hannes Reinecke
Signed-off-by: Christoph Hellwig -
While checking what scsi_adjust_queue_depth() did I thought its switch
statement could be clearer:- remove redundant assignment (to sdev->queue_depth)
- re-order cases (thus removing the fall-through)Signed-off-by: Douglas Gilbert
Reviewed-by: Martin K. Petersen
Reviewed-by: Robert Elliott
Tested-by: Robert Elliott
Signed-off-by: Christoph Hellwig -
Signed-off-by: Christoph Hellwig
Reviewed-by: Paolo Bonzini
Reviewed-by: Hannes Reinecke -
Because of the removal of the scsi_tgt kernel module, the kbuild variables
CONFIG_SCSI_TGT, CONFIG_SCSI_SRP_TGT_ATTRS and CONFIG_SCSI_FC_TGT_ATTRS
are obsolete. This patch removes these variables. This patch is the result
of the following command:find -name '*defconfig' | while read f; do grep -vwE 'CONFIG_SCSI_TGT|CONFIG_SCSI_SRP_TGT_ATTRS|CONFIG_SCSI_FC_TGT_ATTRS|CONFIG_SRP' $f >/tmp/t && mv /tmp/t $f; done
Signed-off-by: Bart Van Assche
Signed-off-by: Christoph Hellwig
Reviewed-by: Paolo Bonzini
Reviewed-by: Hannes Reinecke -
Now that the ibmvstgt driver as the only user of scsi_tgt is gone, the
scsi_tgt kernel module, the CONFIG_SCSI_TGT, CONFIG_SCSI_SRP_TGT_ATTRS and
CONFIG_SCSI_FC_TGT_ATTRS kbuild variable, the scsi_host_template
transfer_response method are no longer needed.[hch: minor updates to the current tree, changelog update]
Signed-off-by: Bart Van Assche
Signed-off-by: Christoph Hellwig
Reviewed-by: Paolo Bonzini
Reviewed-by: Hannes Reinecke -
Remove the libsrp module which was only used by the now removed ibmvstgt
driver.Signed-off-by: Christoph Hellwig
Reviewed-by: Paolo Bonzini
Reviewed-by: Hannes Reinecke -
The IBM virtual SCSI protocol has been obsoleted by ibmvfc, and there
are no reported of the driver left.Signed-off-by: Christoph Hellwig
Reviewed-by: Paolo Bonzini
Reviewed-by: Hannes Reinecke -
Using dev_printk variants prefixes the logging message with
the originating device, which makes debugging easier.Signed-off-by: Hannes Reinecke
Reviewed-by: Martin K. Petersen
Reviewed-by: Christoph Hellwig
Signed-off-by: Christoph Hellwig -
Signed-off-by: Hannes Reinecke
Reviewed-by: Martin K. Petersen
Reviewed-by: Christoph Hellwig
Signed-off-by: Christoph Hellwig -
Update the st driver to use dev_printk() variants instead of
plain printk(); this will prefix logging messages with the
appropriate device.Signed-off-by: Hannes Reinecke
Reviewed-by: Martin K. Petersen
Reviewed-by: Christoph Hellwig
Signed-off-by: Christoph Hellwig -
Update the ch driver to use dev_printk() variants instead of
plain printk(); this will prefix logging messages with the
appropriate device.Signed-off-by: Hannes Reinecke
Reviewed-by: Martin K. Petersen
Reviewed-by: Christoph Hellwig
Signed-off-by: Christoph Hellwig -
Update the sg driver to use dev_printk() variants instead of
plain printk(); this will prefix logging messages with the
appropriate device.Signed-off-by: Hannes Reinecke
Acked-by: Doug Gilbert
Reviewed-by: Martin K. Petersen
Reviewed-by: Christoph Hellwig
Signed-off-by: Christoph Hellwig -
Update the sr driver to use dev_printk() variants instead of
plain printk(); this will prefix logging messages with the
appropriate device.Signed-off-by: Hannes Reinecke
Reviewed-by: Martin K. Petersen
Reviewed-by: Christoph Hellwig
Signed-off-by: Christoph Hellwig -
scsilun_to_int() has an error which prevents it from generating
correct LUN numbers for 64bit values.
Also we should remove the misleading comment about portions of
the LUN being ignored; the initiator should treat the LUN as
an opaque value.
And, finally, the example given should use the correct
prefix (here: extended flat space addressing scheme).This patch includes the modifications suggested by
Bart van Assche.Cc: Bart van Assche
Cc: Christoph Hellwig
Signed-off-by: Hannes Reinecke
Reviewed-by: James Bottomley
Signed-off-by: Christoph Hellwig -
Now that we're using 64-bit LUNs internally we need to increase
the size of max_luns to 64 bits, too.Signed-off-by: Hannes Reinecke
Reviewed-by: Christoph Hellwig
Reviewed-by: Ewan Milne
Signed-off-by: Christoph Hellwig -
Some driver might want to pass in an 64-bit value, so introduce
a module param type 'ullong'.Signed-off-by: Hannes Reinecke
Reviewed-by: Christoph Hellwig
Reviewed-by: Ewan Milne
Acked-by: Rusty Russell
Signed-off-by: Christoph Hellwig -
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
Reviewed-by: Christoph Hellwig
Reviewed-by: Ewan Milne
Signed-off-by: Christoph Hellwig -
Older HBAs are only capable of supporting 16-bit LUNs,
so we need to make sure to adjust max_lun accordingly.Signed-off-by: Hannes Reinecke
Acked-by: Chad Dupuis
Reviewed-by: Ewan Milne
Signed-off-by: Christoph Hellwig -
Sequential scan for more than 256 LUNs is very fragile as
LUNs might not be numbered sequentially after that point.SAM revisions later than SCSI-3 impose a structure on
LUNs larger than 256, making LUN numbers between 256
and 16384 illegal.
SCSI-3, however allows for plain 64-bit numbers with
no internal structure.So restrict sequential LUN scan to 256 LUNs and add a
new blacklist flag 'BLIST_SCSI3LUN' to scan up to
max_lun devices.Signed-off-by: Hannes Reinecke
Reviewed-by: Ewan Milne
Signed-off-by: Christoph Hellwig -
Obsolete; either use 'max_lun' if the host supports only a
limited number of LUNs or BLIST_NOLUN if the target has
problems addressing more than one LUN.Signed-off-by: Hannes Reinecke
Reviewed-by: Ewan Milne
Signed-off-by: Christoph Hellwig -
This addresses a problem reported by Vaughan Cao concerning
the correctness of the O_EXCL logic in the sg driver. POSIX
doesn't defined O_EXCL semantics on devices but "allow only
one open file descriptor at a time per sg device" is a rough
definition. The sg driver's semantics have been to wait
on an open() when O_NONBLOCK is not given and there are
O_EXCL headwinds. Nasty things can happen during that wait
such as the device being detached (removed). So multiple
locks are reworked in this patch making it large and hard
to break down into digestible bits.This patch is against Linus's current git repository which
doesn't include any sg patches sent in the last few weeks.
Hence this patch touches as little as possible that it
doesn't need to and strips out most SCSI_LOG_TIMEOUT()
changes in v3 because Hannes said he was going to rework all
that stuff.The sg3_utils package has several test programs written to
test this patch. See examples/sg_tst_excl*.cpp .Not all the locks and flags in sg have been re-worked in
this patch, notably sg_request::done . That can wait for
a follow-up patch if this one meets with approval.Signed-off-by: Douglas Gilbert
Reviewed-by: Hannes Reinecke
