- 28 Jun, 2017 22 commits
-
-
Johannes Thumshirn authored
libsas uses scsi_queue_work() to queue its internal event notifications. scsi_queue_work() can return -EINVAL if the work queue doesn't exist and it does call queue_work() which can return false if the work is already queued. Make the SAS event code capable of returning errors up to the caller, which is handy when changing to dynamically allocated work in libsas as well, as discussed here: https://lkml.org/lkml/2017/6/14/121. [mkp: fixed typo] Signed-off-by:
Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
Himanshu Madhani authored
Signed-off-by:
Himanshu Madhani <himanshu.madhani@cavium.com> Reviewed-by:
James Smart <james.smart@broadcom.com> Reviewed-by:
Hannes Reinecke <hare@suse.com> Signed-off-by:
-
Duane Grigsby authored
[mkp: fixed typo] Signed-off-by:
Duane Grigsby <duane.grigsby@cavium.com> Signed-off-by:
Darren Trapp <darren.trapp@cavium.com> Signed-off-by:
Anil Gurumurthy <anil.gurumurhty@cavium.com> Signed-off-by:
Giridhar Malavali <giridhar.malavali@cavium.com> Signed-off-by:
-
Duane Grigsby authored
This patch adds switch command support for FC-4 type of FC-NVMe (0x28) for resgistering HBA port to the management server. RFT_ID command is used to register FC-4 type of 0x28 and RFF_ID is used to register FC-4 features bits for FC-NVMe port. Signed-off-by:
-
Duane Grigsby authored
This code provides the interfaces to register remote and local ports of FC4 type 0x28 with the FC-NVMe transport and transports the requests (FC-NVMe FC link services and FC-NVMe commands IUs) to the fabric. It also provides the support for allocating h/w queues and aborting FC-NVMe FC requests. Signed-off-by:
-
Duane Grigsby authored
This patch adds logic to handle the completion of FC-NVMe commands and creates a sub-command in the SRB command structure to manage NVMe commands. Signed-off-by:
-
Duane Grigsby authored
Added logic to change the login process into an optional PRIL step for FC-NVMe ports as a separate operation, such that we can change type to 0x28 (NVMe). Currently, the driver performs the PLOGI/PRLI together as one operation, but if the discovered port is an NVMe port then we first issue the PLOGI and then we issue the PRLI. Also, the fabric discovery logic was changed to mark each discovered FC NVMe port, so that we can register them with the FC-NVMe transport later. Signed-off-by:
-
himanshu.madhani@cavium.com authored
Signed-off-by:
-
Quinn Tran authored
Add missing memory dump of Exchange Offload and Extended login into FW dump. Signed-off-by:
Quinn Tran <quinn.tran@cavium.com> Signed-off-by:
-
Quinn Tran authored
Move counters to qpair to reduce cache miss. Signed-off-by:
-
Quinn Tran authored
These fields only hold one set of value. Replace it with macros to reduce cache thrash. Signed-off-by:
-
Quinn Tran authored
By default this flag is forced to true. Remove this flag and unneccessary check for this flag. Signed-off-by:
-
Quinn Tran authored
For target main path io routines that uses qpair, create new logging & debugging routines to use qpair instead of reaching for scsi_qla_host to reduce cache thrash. Signed-off-by:
-
Quinn Tran authored
Add call back to door bell for qpair. This help reduce access to qla_hw_data structure, in order to reduce cach thrash. Signed-off-by:
-
Quinn Tran authored
For ISP27XX, use shadow register to read FW provided REQQ's consumer index. The shadow register is dma'ed by firmware. Signed-off-by:
-
Quinn Tran authored
- Move chip_reset, enable_class_2 fields from qla_hw_data to qla_qpair to reduce cache thrash for target MQ. - Optimizations to reduce unnecessary memory load for good path io. Signed-off-by:
-
Quinn Tran authored
Add fw_started flag to qpair to reduce cache thrash. This reduce access to qla_hw_data structure by each qpair. Signed-off-by:
-
Quinn Tran authored
For Target mode, user can control the work load by placing qla2xxx's irq vector on certain CPU via the smp_affinity knob. This patch allows user to control the number of QPair's irq to be active. The irqs are allocated at driver load time until unload. The work itself is placed on the QPair based on user setting. Usage: modprobe qla2xxx qlini_mode=disabled ql2xuctrlirq=1 mount -t debugfs none /sys/kernel/debug echo 2 > /sys/kernel/debug/qla2xxx/qla2xxx_[host num]/naqp echo [cpu id] > /proc/irq/[irq id]/smp_affinity_list Signed-off-by:
-
Sawan Chandak authored
In target mode driver, queue pairs are not created during driver load time, instead they are created at the configuration time after chip reset. If a user tries to load/unload driver after queue pairs are created, then there would be mailbox failure, while deleting queue pairs. Flag is added to check if queue pairs are created or not. Queue pairs will be deleted only If they were created during target configuration. Signed-off-by:
Sawan Chandak <sawan.chandak@cavium.com> Signed-off-by:
-
Quinn Tran authored
Enable Multi Queue for Target mode. At Initiator LUN scan time, each LUN is assign to a QPair. Each QPair is affinitize to certain CPU. When new cmd arrives from the wire, the lunid is used to search for qpair. The qpair's affinitized cpuid will be used to queue up the work element. Signed-off-by:
-
Quinn Tran authored
In Current code, Req Q 0, RespQ 0 & hardware_lock are the main resources for sending and process completion of Target IO. These resources are now referenced behind a new qpair/"struct qla_qpair base_qpair". Main path IO handle will access those resources via the qpair pointer in preparation for Target MQ. Signed-off-by:
-
Quinn Tran authored
Merge active/outstanding cmd arrays from target side and initiator side together in prepration for Target Multi Queue support. Signed-off-by:
-
- 26 Jun, 2017 18 commits
-
-
Matthew R. Ochs authored
Currently, the SCSI command presented to the device reset handler is used to send TMFs to the AFU for a device reset. This behavior is incorrect as the command presented is an actual command and not a special notification. As such, it should only be used for reference and not be acted upon. Additionally, the existing TMF transmission routine does not account for actual errors from the hardware, only reflecting failure when a timeout occurs. This can lead to a condition where the device reset handler is presented with a false 'success'. Update send_tmf() to dynamically allocate a private command for sending the TMF command and properly reflect failure when the completed command indicates an error or was aborted. Detect TMF commands during response processing and avoid scsi_done() for these types of commands. Lastly, update comments in the TMF processing paths to describe the new behavior. Signed-off-by:
Matthew R. Ochs <mrochs@linux.vnet.ibm.com> Signed-off-by:
Uma Krishnan <ukrishn@linux.vnet.ibm.com> Signed-off-by:
-
Matthew R. Ochs authored
The SCSI core now zeroes the per-command private data area prior to calling into the LLD. Replace the clearing operation that takes place when the private command data reference is obtained with a routine that performs common initializations. The zeroing that takes place in the device reset path remains intact as the private command data associated with the specified SCSI command is not guaranteed to be cleared. Signed-off-by:
-
Matthew R. Ochs authored
The cxlflash driver supports performing a write-same16 to scrub virtual luns when they are released by a user. To date, AFUs for adapters that are supported by cxlflash do not have the capability to unmap as part of the WS operation. This can lead to fragmented flash devices which results in performance degradation. Future AFUs can optionally support unmap write-same commands and reflects this support via the context control register. This provides userspace applications with direct visibility such that they need not depend on a host API. Detect unmap support during cxlflash initialization by reading the context control register associated with the primary hardware queue. Update the existing write_same16() routine to set the unmap bit in the CDB when unmap is supported by the host. Signed-off-by:
-
Matthew R. Ochs authored
Adopt the SISLite AFU debug capability to allow future CXL Flash adapters the ability to better debug AFU issues. Update the SISLite header with the changes necessary to support AFU debug operations and create a host ioctl interface for user debug software. Also update the cxlflash documentation to describe this new host ioctl. Signed-off-by:
-
Matthew R. Ochs authored
Adopt the SISLite AFU LUN provisioning capability to allow future CXL Flash adapters the ability to better manage storage. Update the SISLite header with the changes necessary to support LUN provision operations and create a host ioctl interface for user LUN management software. Also update the cxlflash documentation to describe this new host ioctl. Signed-off-by:
-
Matthew R. Ochs authored
The existing AFU capability checking infrastructure is closely tied to the command mode capability bits. In order to support new capabilities, refactor the existing infrastructure to be more generic. Signed-off-by:
-
Matthew R. Ochs authored
As staging for supporting various host management functions, add a host ioctl infrastructure to filter ioctl commands and perform operations that are common for all host ioctls. Also update the cxlflash documentation to create a new section for documenting host ioctls. Signed-off-by:
-
Matthew R. Ochs authored
To date the only supported internal AFU command is AFU sync. The logic to send an internal AFU command is embedded in the specific AFU sync handler and would need to be duplicated for new internal AFU commands. In order to support new internal AFU commands, separate code that is common for AFU internal commands into a generic transmission routine and support passing back command status through an IOASA structure. The first user of this new routine is the existing AFU sync command. As a cleanup, use a descriptive name for the AFU sync command instead of a magic number. Signed-off-by:
-
Uma Krishnan authored
The cxlflash driver currently lacks host management interface. Future devices supported by cxlflash will provide a variety of host-wide management functions. Examples include LUN provisioning, hardware debug support, and firmware download. In order to provide a way to manage the device, a character device will be created during probe of each adapter. This device will support a set of ioctls defined in the SISLite specification from which administrators can manage the adapter. Signed-off-by:
-
Uma Krishnan authored
To date, CXL flash devices do not support a single command abort operation. Instead, the SISLite specification provides a context reset operation to cleanup all pending commands for a given context. When a context reset is successful, it is guaranteed that the AFU has aborted all currently pending I/O. This sequence is less invasive than a device or host reset and can be executed to support scsi command abort requests. Add eh_abort_handler callback support to process command timeouts and abort requests. Signed-off-by:
-
Uma Krishnan authored
When the AFU is reset in an error path, pending scsi commands can be silently dropped without completion or a formal abort. This puts the onus on the cxlflash driver to notify mid-layer and indicating that the command can be retried. Once the card has been quiesced, the hardware send queue lock is acquired to prevent any data movement while the pending commands are processed. Signed-off-by:
-
Uma Krishnan authored
Currently, there is no book keeping of the pending scsi commands in the cxlflash driver. This lack of tracking in-flight requests is too restrictive and requires a heavy-hammer reset each time an adapter error is encountered. Additionally, it does not allow for commands to be properly retried. In order to avoid this problem and to better handle error path command cleanup, introduce a linked list for each hardware queue that tracks pending commands. Signed-off-by:
-
Uma Krishnan authored
AFU sync operations are not currently evaluated for failure. This is acceptable for paths where there is not a dependency on the AFU being consistent with the host. Examples include link reset events and LUN cleanup operations. On paths where there is a dependency, such as a LUN open, a sync failure should be acted upon. In the event of AFU sync failures, either log or cleanup as appropriate for operations that are dependent on a successful sync completion. Update documentation to reflect behavior in the event of an AFU sync failure. Signed-off-by:
-
Uma Krishnan authored
A context reset failure indicates the AFU is in a bad state. At present, when such a situation occurs, no further action is taken. This leaves the adapter in an unusable state with no recoverable actions. To avoid this situation, context reset failures will be escalated to a host reset operation. This will be done asynchronously to allow the acting thread to return to the user with a failure. Signed-off-by:
-
Uma Krishnan authored
Per the SISLite specification, context_reset() writes 0x1 to the LSB of the reset register. When the AFU processes this reset request, it is expected to clear the bit after reset is complete. The current implementation simply checks that the entire value read back is not 1, instead of masking off the LSB and evaluating it for a change to 0. Should the AFU manipulate other bits during the reset (reading back a value of 0xF for example), successful completion will be prematurely indicated given the existing logic. Additionally, in the event that the context reset operation fails, there does not currently exist a way to provide feedback to the initiator of the reset. This poses a problem for the rare case that a context reset fails as the caller will proceed on the assumption that all is well. To remedy these issues, refactor the context reset routine to only mask off the LSB when evaluating for success and return status to the caller. Also update the context reset handler parameters to pass a hardware queue reference instead of a single command to better reflect that the entire queue associated with the context is impacted by the reset. Signed-off-by:
-
Uma Krishnan authored
The cxlflash_afu_sync() routine returns a negative one to indicate any kind of failure. This makes it impossible to establish why the error occurred. Update the return codes to clearly indicate the failure cause to the caller. Signed-off-by:
-
Uma Krishnan authored
Currently there are separate spin locks for the two supported I/O queueing models. This makes it difficult to serialize with paths outside the enqueue path. As a design simplification and to support serialization with enqueue operations, move to only a single lock that is used for enqueueing regardless of the queueing model. Signed-off-by:
-
Christoph Hellwig authored
Use dma_alloc_attrs directly instead of the dma_alloc_noncoherent wrapper. [mkp: fixed driver name] Signed-off-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
-
