Every single i/o or event completion incurs a test and branch to see if
the cycle bit changed. For power-of-2 queue sizes the cycle bit can be
read directly from the rollover of the queue pointer.
Likely premature optimization, but the hidden if() and hidden
assignments / side-effects in the macros were already asking to be
cleaned up.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
A tag is a 16 bit number where the upper four bits is a sequence number
and the remainder is the task context index (tci). Sanitize the macro
names and shave 256-bytes out of scic_sds_controller by reducing the size of
io_request_sequence.
scic_sds_io_tag_construct --> ISCI_TAG
scic_sds_io_tag_get_sequence --> ISCI_TAG_SEQ
scic_sds_io_tag_get_index() --> ISCI_TAG_TCI
scic_sds_io_sequence_increment() [delete / open code]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The circ_buf macros are ~6% faster, as measured by perf, because they take
advantage of power-of-two math assumptions i.e. no test and branch for
rollover. Their semantics are clearer than the hidden side effects in pool.h
(like sci_pool_get() which hides an assignment).
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The old 'core' had aspirations of running in severely memory constrained
environments like bios option-rom, it's not needed for Linux and gets in
the way of other cleanups (like unifying/reducing the number of structure
members in scic_sds_controller/isci_host).
This also fixes a theoretical bug in that the driver would blindly override
the silicon advertised limits for number of ports, task contexts, and remote
node contexts.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
C0 silicon updates the pci revision id and requires new AFE parameters
for phy signal integrity. Support for previous silicon revisions is
deprecated (it's also broken for the theoretical case of multiple
controllers at different silicon revisions, all the more reason to get
it removed as soon as possible)
Signed-off-by: Adam Gruchala <adam.gruchala@intel.com>
[fixed up deprecated silicon support]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Additional state machine cleanups:
o Remove static functions sci_state_machine_exit_state() and
sci_state_machine_enter_state()
o Combines sci_base_state_machine_construct() and
sci_base_state_machine_start() into a single function,
sci_init_sm()
o Remove sci_base_state_machine_stop() which is unused.
o Kill state_machine.[ch]
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
[fixed too large to inline functions]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
This cleans up several areas of the state machine mechanism:
o Rename sci_base_state_machine_change_state to sci_change_state
o Remove sci_base_state_machine_get_state function
o Rename 'state_machine' struct member to 'sm' in client structs
o Shorten the name of request states
o Shorten state machine state names as follows:
SCI_BASE_CONTROLLER_STATE_xxx to SCIC_xxx
SCI_BASE_PHY_STATE_xxx to SCI_PHY_xxx
SCIC_SDS_PHY_STARTING_SUBSTATE_xxx to SCI_PHY_SUB_xxx
SCI_BASE_PORT_STATE_xxx to SCI_PORT_xxx and
SCIC_SDS_PORT_READY_SUBSTATE_xxx to SCI_PORT_SUB_xxx
SCI_BASE_REMOTE_DEVICE_STATE_xxx to SCI_DEV_xxx
SCIC_SDS_STP_REMOTE_DEVICE_READY_SUBSTATE_xxx to SCI_STP_DEV_xxx
SCIC_SDS_SMP_REMOTE_DEVICE_READY_SUBSTATE_xxx to SCI_SMP_DEV_xxx
SCIC_SDS_REMOTE_NODE_CONTEXT_xxx_STATE to SCI_RNC_xxx
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Convert the sata_timeout_timer in the scic_sds_phy struct to
use a struct sci_timer
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Rather than preallocating a list of timers and doling them out at runtime,
embed a struct timerlist in each object that needs one. A struct sci_timer
interface is introduced to manage the timer cancellation semantics which
currently need to guarantee the timer is cancelled while holding
spin_lock(ihost->scic_lock). Since the timeout functions also need to acquire
the lock it currently prevents the driver from using del_timer_sync() for
runtime cancellations.
del_timer_sync() is used however before the objects go out of scope.
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Now that any given object type only has one state_machine we can use
container_of() to get back to the given state machine owner.
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Implement the stop handlers directly in scic_sds_port_stop()
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Piotr Sawicki <piotr.sawicki@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
remove the handler from the port state handler table and implement the
logic directly in scic_sds_port_start().
Signed-off-by: Piotr Sawicki <piotr.sawicki@intel.com>
[remove a level of indirection]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
While cleaning up the driver it is very tempting to convert scic_sds_get_*
macros to their open coded equivalent. They are all just pointer dereferences
*except* scic_sds_phy_get_port() which returns NULL if the phy is assigned to
the dummy port. Clarify this by renaming it to phy_get_non_dummy_port().
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Merged states and substates into one state machine, as we always
unconditionally transitioned to the substate machine it was straightforward to
enter that substate from the starting state.
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Adam Gruchala <adam.gruchala@intel.com>
[fixed construction, starting_state_enter, and starting check]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
* Move port configuration agent implementation
* Merge core/scic_sds_port.[ch] into port.[ch]
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
* Consolidate tiny header files
* Move files out of core/ (drop core/scic_sds_ prefix)
* Merge core/scic_sds_request.[ch] into request.[ch]
* Cleanup request.c namespace (clean forward declarations and global
namespace pollution)
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Now that the data structures are unified unify the implementation in
host.[ch] and cleanup namespace pollution.
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
cross driver constants are spread out over multiple header files, consolidate
them into isci.h, and push some includes out to the source files that need
them.
TODO: remove SCI_MODE_SIZE infrastructure.
TODO: task.h is full of inlines that are too large
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Make scic_sds_request a proper member of isci_request. Also let's us
get rid of the dma pool object size tracking since we now know that all
requests are sizeof(isci_request). While cleaning up the construct
routine incidentally replaced SCI_FIELD_OFFSET with offsetof.
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Make scic_sds_port a member of isci_port and merge their lifetimes which
means removing the port table from scic_sds_controller in favor of the
one at the isci_host level. Merge ihost->sas_ports into ihost->ports.
_
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Make it explicit that isci_host and scic_sds_controller are one in the same
object.
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>
[removed ->ihost back pointer]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The 'struct sci_base_object' was removed from the struct
scic_sds_controller and was replaced by a pointer to
struct isci_host.
Signed-off-by: Maciej Patelczyk <maciej.patelczyk@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Now that the core/lldd remote_device data structures are nominally unified
merge the corresponding sources into the top-level directory. Also move the
remote_node_context infrastructure which has no analog at the lldd level.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Remove the extra logic to poll each controller for interrupts, that's
the core's job for shared interrupts.
While testing noticed that a number of interrupts fire while waiting for
the completion tasklet to run, so added an irq-ack.
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Remove the insane infrastructure for preallocating coheren DMA regions,
and just allocate the memory where needed. This also gets rid of the
aligment adjustments given that Documentation/DMA-API-HOWTO.txt sais:
"The cpu return address and the DMA bus master address are both
guaranteed to be aligned to the smallest PAGE_SIZE order which
is greater than or equal to the requested size. This invariant
exists (for example) to guarantee that if you allocate a chunk
which is smaller than or equal to 64 kilobytes, the extent of the
buffer you receive will not cross a 64K boundary."
Signed-off-by: Christoph Hellwig <hch@lst.de>
[djbw: moved allocation from start to init, re-add memset]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Merge struct sci_base_controller into scic_sds_controller, and also factor
the two types of state machine handlers into one function. While we're at
it also remove lots of duplicate incorrect kerneldoc comments for the state
machine handlers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
A domain_device can always reference back to ->lldd_ha unlike local lldd
structures. Fix up cases where the driver uses local objects to look up the
isci_host. This also changes the calling conventions of some routines to
expect a valid isci_host parameter rather than re-lookup the pointer on entry.
Incidentally cleans up some macros that are longer to type than the open-coded
equivalent:
isci_host_from_sas_ha
isci_dev_from_domain_dev
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Exposing the user config parameters through the kernel module parameters.
The kernel module params will have the default values set and we will no
longer pulling the default values for user params from the core.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
1/ Since commit 858d4aa7 "isci: Move firmware loading to per PCI device" we have
been silently falling back to built-in defaults for the parameter settings by
skipping the call to scic_oem_parameters_set().
2/ The afe parameters from the firmware were not being honored
3/ The latest oem parameter definition flips the mode_type values which are
now 0: for APC 1: for MPC. For APC we need to make sure all the phys
default to the same address otherwise strict_wide_ports will cause duplicate
domains.
4/ Fix up the driver announcement to indicate the source of the
parameters.
5/ Fix up the sas addresses to be unique per controller (in the fallback case)
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
We need to scan the OROM for signature and grab the OEM parameters. We
also need to do the same for EFI. If all fails then we resort to user
binary blob, and if that fails then we go to the defaults.
Share the format with the create_fw utility so that all possible sources
of the parameters are in-sync.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
If there is a pending device reset, the I/O is used to accomplish the reset by setting the
RESET bit in the task status, and then putting the task into the error handler
path using sas abort task.
Signed-off-by: Jeff Skirvin <jeffrey.d.skirvin@intel.com>
Signed-off-by: Jacek Danecki <Jacek.Danecki@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Changes to move management of the reqs_in_process entry for the request here.
Made changes to note when the task is already in the abort path and
cannot be completed through callbacks.
Signed-off-by: Jeff Skirvin <jeffrey.d.skirvin@intel.com>
Signed-off-by: Jacek Danecki <Jacek.Danecki@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Until we synchronize against device removal this limits the damage of
use after free bugs to the driver's own objects. Unless we implement
reference counting we need to ensure at least a subset of a remote
device is valid at all times. We follow the lead of other libsas
drivers that also preallocate devices.
This also enforces maximum remote device accounting at the lldd layer,
but the core may still run out of RNC's before we hit this limit.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Replace the device completion infrastructure with the controller wide
event queue. There was a potential for the stop and ready notifications
to corrupt each other, now that cannot happen.
The stop pending flag cannot be used until devices are statically
allocated. We temporarily need to maintain a completion to handle
waiting for an object that has disappeared, but we can at least stop
scribbling on freed memory.
A future change will also get rid of the "stopping" state as it should
not be exposed to the rest of the driver.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
isci_host_by_id() should have been a clue that an array would have been
a simpler approach.
Reported-by: James Bottomley <James.Bottomley@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
There is a condition whereby TCs (task contexts) can jump to the head of
the round robin queue causing indefinite starvation of pending tasks.
Posting a TC to a suspended RNC (remote node context) causes the
hardware to select that task first, but since the RNC is suspended the
scheduler proceeds to the next task in the expected round robin fashion,
restoring TC arbitration fairness.
Signed-off-by: Tomasz Chudy <tomasz.chudy@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Prepare the timer api for the arrival of dynamic creation and
destruction events from the core. It pretended to do this previously
but the core to date only used it in a static init-time only fashion.
This is an interim fix until a cleaner event queue can be developed.
1/ make all locking external to the api (add WARN_ONCE to verify)
2/ add a timer_destroy interface (to be used by the core)
3/ use del_timer_sync() prior to deallocating timer data
4/ delete the "timer_list" indirection, we only have timers allocated
for the isci_host
5/ fix detection of timer list allocation errors
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Undo the open coded and incorrect translation of the oem parameter sas
address to its libsas expected format.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Moved the firmware loading from per adapter to per PCI device. This should
prevent firmware from being loaded twice becuase of 2 SCU controller per
PCI device. We do have to do it per PCI device because request_firmware()
requires a struct device passed in.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Polling the event queue during scan is an unneeded holdover from the
original driver.
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
[djbw: ensure we flush all port events and domain discovery]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The lldd actively disallows requests in the "starting" state. Retrying
or holding off commands in this state is sub-optimal:
1/ it adds another state check to the fast path
2/ retrying can cause libsas to give up
However, isci's ->lldd_dev_found() routine already waits for controller
start to complete before allowing further progress. Checking the
"starting" state in isci_task_execute_task and the isr is redundant and
misleading. Clean this up and introduce a controller-wide event queue
to start reeling in "completion" proliferation in the driver.
The "stopping" state cleanups are in a similar vein, rely on the the isr
and other paths being precluded from occurring rather than implementing
state checking logic.
Reported-by: Christoph Hellwig <hch@infradead.org>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The indirection is unecessary and broken in the current case that assigns the
handlers based on a not up-to-date pdev->msix_enabled value.
Route the handlers directly to the requisite core routines.
Todo: hook up error interrupt handling
Reported-by: Jeff Garzik <jeff@garzik.org>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Support for the up to 2x4-port 6Gb/s SAS controllers embedded in the
chipset.
This is a snapshot of the first publicly available version of the driver,
commit 4c1db2d0 in the 'historical' branch.
git://git.kernel.org/pub/scm/linux/kernel/git/djbw/isci.git historical
Signed-off-by: Maciej Trela <maciej.trela@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Edmund Nadolski <edmund.nadolski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Dan Williams