- 18 Apr, 2018 40 commits
-
-
James Smart authored
During target-side port faults, the driver would not recover all target port logins. This resulted in a loss of nvme device discovery. The driver is coded to wait for all GID_FT requests to complete before restarting discovery. A fault is seen where the outstanding GIT_FT counts are not properly decremented, thus discovery would never start. Another fault was found in the clearing of the gidft_inp counter that would be skipped in this condition. And a third fault found with lpfc_nvme_register_port that would remove a reverence on the ndlp which then allows a node swap on a port address change to prematurely remove the reference and release the ndlp. The following changes are made: - Correct the decrementing of the outstanding GID_FT counters. - In RSCN handling, no longer zero the counter before calling to issue another GID_FT. - No longer remove the reference on the dlp when the ndlp->nrport value is not yet null. Signed-off-by:
Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by:
James Smart <james.smart@broadcom.com> Reviewed-by:
Hannes Reinecke <hare@suse.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
James Smart authored
The patch to enlarge WQ/CQ creation keys off of an adapter response that indicates support for the larger values. Older adapters return an incorrect response and are limited in size. Thus the adapters fail the WQ creation steps. Augment the WQ sizing checks with a check on the older adapter types and limit them to the restricted sizes. Fixes: c176ffa0 ("scsi: lpfc: Increase CQ and WQ sizes for SCSI") Signed-off-by:
-
James Smart authored
After making remoteport unregister requests, the ndlp nrport pointer was stale. Track when waiting for waiting for unregister completion callback and adjust nldp pointer assignment. Add a few safety checks for NULL pointer values. Signed-off-by:
-
James Smart authored
After driver unloads, lpfc_wq remains active. The destroy_workqueue calls were not being made in driver unload. Additionally, SLI3 is allocating lpfc_wq resources, but never uses it. Make the destroy_workqueue calls on driver unload. Modify the SLI3 code path no longer allocate lpfc_wq resources. Signed-off-by:
-
James Smart authored
When running loads that generated aborts, io errors where seen. Turns out the abort requests where not placed on the proper WQ resulting in the errors. Closer inspection inspection of this error also showed improper spinlock api use. Correct the WQ selection policy for the abort requests. Correct spin_lock/spin_lock_irq/spin_lock_irqsave usage. Signed-off-by:
-
James Smart authored
Under large io load, the current sizing of asynchronous buffer counts could be exceeded, indicated by a 2885 log message: 2885 Port Status Event: port status reg 0x81800000, port smphr reg 0xc000, error 1=0x52004a01, error 2=0x0 Enlarge the async receive queue size. Allow for a configurable number of buffers to be posted to each RQ, using the new attribute lpfc_nvmet_mrq_post. Signed-off-by: -
James Smart authored
When debugging various issues, per IO channel IO statistics were useful to understand what was happening. However, many of the stats were on a port basis rather than an io channel basis. Move statistics to an io channel basis. Signed-off-by:
-
James Smart authored
The max_scsicmpl_time parameter can be used to perform scsi cmd queue depth mgmt based on io completion time: the queue depth is reduced to make completion time shorter. However, as soon as an io completes and the completion time is within limits, the code immediately bumps the queue depth limit back up to the target queue depth. Thus the procedure restarts, effectively limiting the usefulness of adjusting queue depth to help completion time. This patch makes the following changes: - Removes the code at io completion that resets the queue depth as soon as within limits. - As the code removed was where the target queue depth was first applied, change target queue depth application so that it occurs when the parameter is changed. - Makes target queue depth a standard parameter: both a module parameter and a sysfs parameter. - Optimizes the command pending count by using atomics rather than locks. - Updates the debugfs nodelist stats to allow better debugging of pending command counts. Signed-off-by:
-
James Smart authored
Nodelist entry for SCSI array ends up in UNMAPPED state. This is due to illegal discovery State machine transition because of two PRLIs and the first one failing with LS_RJT. Also, the error path was designed assuming the PRLIs complete in the order they were sent, FCP first, then NVME. In a failing case, the array thinks about the first PRLI (FCP), but issues LS_RJT for the 2nd PRLI immediately. Fix PRLI completion error path for the ordering expectation. Ensure the discovery state machine update is not set until all outstanding PRLIs are complete. Signed-off-by:
-
Shivasharan S authored
Signed-off-by:
Shivasharan S <shivasharan.srikanteshwara@broadcom.com> Signed-off-by:
-
Shivasharan S authored
Hardware could time out Fastpath IOs one second earlier than the timeout provided by the host. For non-RAID devices, driver provides timeout value based on OS provided timeout value. Under certain scenarios, if the OS provides a timeout value of 1 second, due to above behavior hardware will timeout immediately. Increase timeout value for non-RAID fastpath IOs by 1 second. Signed-off-by:
-
Himanshu Jha authored
Use pci_zalloc_consistent for allocating zeroed memory and remove unnecessary memset function. Done using Coccinelle. Generated by: scripts/coccinelle/api/alloc/kzalloc-simple.cocci Suggested-by:
Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by:
Himanshu Jha <himanshujha199640@gmail.com> Signed-off-by:
-
Long Li authored
In Vmbus, we have defined a function to calculate available ring buffer percentage to write. Use that function and remove netvsc's private version. [mkp: typo] Signed-off-by:
Long Li <longli@microsoft.com> Acked-by:
David S. Miller <davem@davemloft.net> Signed-off-by:
-
Long Li authored
Netvsc has a function to calculate how much ring buffer in percentage is available to write. This function is also useful for storvsc and other vmbus devices. Define a similar function in vmbus to be used by other vmbus devices. Signed-off-by:
-
Jason Yan authored
Now ata devices attached with sas controller do not have transport class, so that we can not see any information of these ata devices in /sys/class/ata_port(or ata_link or ata_device). Add transport class for the ata devices attached with sas controller. The /sys/class directory will show the infomation of the ata devices as follows: localhost:/sys/class # ls ata* ata_device: dev1.0 dev2.0 ata_link: link1 link2 ata_port: ata1 ata2 No functional change of the device scanning and io path. The ata transport class was deleted when destroying the sas devices. Signed-off-by:
Jason Yan <yanaijie@huawei.com> CC: Dan Williams <dan.j.williams@intel.com> CC: Tejun Heo <tj@kernel.org> Acked-by:
Tejun Heo <tj@kernel.org> Signed-off-by:
-
John Garry authored
This patch removes unneeded structure elements: - hisi_sas_phy.dev_sas_addr: only ever written - Also remove associated function which writes it, hisi_sas_init_add(). - hisi_sas_device.attached_phy: only ever written - Also remove code to set it in hisi_sas_dev_found() Signed-off-by:
John Garry <john.garry@huawei.com> Reviewed-by:
Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by:
-
John Garry authored
When we find an erroneous slot completion, to help aid debugging add the device index to the current debug log. Signed-off-by:
-
Xiaofei Tan authored
There is a bug of v3 hw development version. When AXI error happen, hw may return an abnormal CQ that IPTT value is 0xffff. This will cause IPTT out-of-bounds reference. This patch adds a check of IPTT in cq_tasklet_v3_hw() and discards invalid slot. This workaround scheme is just to enhance fault-tolerance of the driver. So, we will apply this scheme for all version of v3 hw, although release version has fixed this SoC bug. Signed-off-by:
Xiaofei Tan <tanxiaofei@huawei.com> Signed-off-by:
-
Xiaofei Tan authored
Currently we check the fis->command value in 2 locations in hisi_sas_get_ata_protocol() switch statement. Fix this by consolidating the check for fis->command value to 1 location only. Signed-off-by:
-
Xiang Chen authored
This is a warning coming from Coccinelle, and need to use new interface dma_zalloc_coherent() instead of dma_alloc_coherent()/memset(). Signed-off-by:
-
Xiang Chen authored
Delete timer for v1 and v3 hw when removing hisi_sas driver. Signed-off-by:
-
Xiaofei Tan authored
There is an modification for later revision of v3 hw. More HW errors are reported through RAS interrupt. These errors were originally reported only through MSI. When report to RAS, some combinations are done to port AXI errors and FIFO OMIT errors. For example, each port has 4 AXI errors, and they are combined to one when report to RAS. This patch does two things: 1. Enable RAS interrupt of these errors and handle them in PCI error handlers. 2. Disable MSI interrupts of these errors for this later revision hw. Signed-off-by:
-
Xiang Chen authored
When directly connected with SATA disks in different SAS cores, fill SAS address with scsi_host's id to make it's fake SAS address unique. Signed-off-by:
-
Uma Krishnan authored
The following Oops can occur when there is heavy I/O traffic and the host is reset by a tool such as sg_reset. [c000200fff3fbc90] c00800001690117c process_cmd_doneq+0x104/0x500 [cxlflash] (unreliable) [c000200fff3fbd80] c008000016901648 cxlflash_rrq_irq+0xd0/0x150 [cxlflash] [c000200fff3fbde0] c000000000193130 __handle_irq_event_percpu+0xa0/0x310 [c000200fff3fbea0] c0000000001933d8 handle_irq_event_percpu+0x38/0x90 [c000200fff3fbee0] c000000000193494 handle_irq_event+0x64/0xb0 [c000200fff3fbf10] c000000000198ea0 handle_fasteoi_irq+0xc0/0x230 [c000200fff3fbf40] c00000000019182c generic_handle_irq+0x4c/0x70 [c000200fff3fbf60] c00000000001794c __do_irq+0x7c/0x1c0 [c000200fff3fbf90] c00000000002a390 call_do_irq+0x14/0x24 [c000200e5828fab0] c000000000017b2c do_IRQ+0x9c/0x130 [c000200e5828fb00] c000000000009b04 h_virt_irq_common+0x114/0x120 When a context is reset, the pending commands are flushed and the AFU is notified. Before the AFU handles this request there could be command completion interrupts queued to PHB which are yet to be delivered to the context. In this scenario, a context could receive an interrupt for a command that has been flushed, leading to a possible crash when the memory for the flushed command is accessed. To resolve this problem, a boolean will indicate if the hardware queue is ready to process interrupts or not. This can be evaluated in the interrupt handler before proessing an interrupt. Signed-off-by:Uma Krishnan <ukrishn@linux.vnet.ibm.com> Acked-by:
Matthew R. Ochs <mrochs@linux.vnet.ibm.com> Signed-off-by:
-
Uma Krishnan authored
The following Oops can occur if an internal command sent to the AFU does not complete within the timeout: [c000000ff101b810] c008000016020d94 term_mc+0xfc/0x1b0 [cxlflash] [c000000ff101b8a0] c008000016020fb0 term_afu+0x168/0x280 [cxlflash] [c000000ff101b930] c0080000160232ec cxlflash_pci_error_detected+0x184/0x230 [cxlflash] [c000000ff101b9e0] c00800000d95d468 cxl_vphb_error_detected+0x90/0x150[cxl] [c000000ff101ba20] c00800000d95f27c cxl_pci_error_detected+0xa4/0x240 [cxl] [c000000ff101bac0] c00000000003eaf8 eeh_report_error+0xd8/0x1b0 [c000000ff101bb20] c00000000003d0b8 eeh_pe_dev_traverse+0x98/0x170 [c000000ff101bbb0] c00000000003f438 eeh_handle_normal_event+0x198/0x580 [c000000ff101bc60] c00000000003fba4 eeh_handle_event+0x2a4/0x338 [c000000ff101bd10] c0000000000400b8 eeh_event_handler+0x1f8/0x200 [c000000ff101bdc0] c00000000013da48 kthread+0x1a8/0x1b0 [c000000ff101be30] c00000000000b528 ret_from_kernel_thread+0x5c/0xb4 When an internal command times out, the command buffer is freed while it is still in the pending commands list of the context. This corrupts the list and when the context is cleaned up, a crash is encountered. To resolve this issue, when an AFU command or TMF command times out, the command should be deleted from the hardware queue pending command list before freeing the buffer. Signed-off-by: -
Uma Krishnan authored
The following Oops can be encountered if a device removal or system shutdown is initiated while an EEH recovery is in process: [c000000ff2f479c0] c008000015256f18 cxlflash_pci_slot_reset+0xa0/0x100 [cxlflash] [c000000ff2f47a30] c00800000dae22e0 cxl_pci_slot_reset+0x168/0x290 [cxl] [c000000ff2f47ae0] c00000000003ef1c eeh_report_reset+0xec/0x170 [c000000ff2f47b20] c00000000003d0b8 eeh_pe_dev_traverse+0x98/0x170 [c000000ff2f47bb0] c00000000003f80c eeh_handle_normal_event+0x56c/0x580 [c000000ff2f47c60] c00000000003fba4 eeh_handle_event+0x2a4/0x338 [c000000ff2f47d10] c0000000000400b8 eeh_event_handler+0x1f8/0x200 [c000000ff2f47dc0] c00000000013da48 kthread+0x1a8/0x1b0 [c000000ff2f47e30] c00000000000b528 ret_from_kernel_thread+0x5c/0xb4 The remove handler frees AFU memory while the EEH recovery is in progress, leading to a race condition. This can result in a crash if the recovery thread tries to access this memory. To resolve this issue, the cxlflash remove handler will evaluate the device state and yield to any active reset or probing threads. Signed-off-by: -
Uma Krishnan authored
This commit enables the OCXL operations for the OCXL devices. Signed-off-by:
-
Uma Krishnan authored
The cxlflash core driver resets the AFU when the master contexts are created in the initialization or recovery paths. Today, the OCXL provider service to perform this operation is pending implementation. To avoid a crash due to a missing fop, log an error once and return success to continue with execution. Signed-off-by:
-
Uma Krishnan authored
While enabling a context on the link, a predefined callback can be registered with the OCXL provider services to be notified on translation errors. These errors can in turn be passed back to the user on a read operation. Signed-off-by:
-
Uma Krishnan authored
In order to protect the OCXL hardware contexts from getting clobbered, a simple state machine is added to indicate when a context is in open, close or start state. The expected states are validated throughout the code to prevent illegal operations on a context. A mutex is added to protect writes to the context state field. Signed-off-by:
-
Uma Krishnan authored
The SISLite specification has been updated to define new synchronous interrupt status bits. These bits are set by the AFU when a given PASID or EA is bad and a synchronous interrupt is triggered. The SISLite header file is updated to support these new bits. Note that there are also some formatting updates to some of the existing bits to allow all of the definitions to line up uniformly. Signed-off-by:
-
Uma Krishnan authored
Similar to user contexts, master contexts also require that the per-context LISN registers be programmed for certain AFUs. The mapped trigger page is obtained from underlying transport and registered with AFU for each master context. Signed-off-by:
-
Uma Krishnan authored
The SISLite specification has been updated for OCXL to support communicating data to generate AFU interrupts to the AFU. This includes a new capability bit that is advertised for OCXL AFUs and new registers to hold the object handle and translation PASID of each interrupt. For Power, the object handle is the mapped trigger page. Note that because these mappings are kernel only, the PASID of a kernel context must be used to satisfy the translation. Signed-off-by:
-
Uma Krishnan authored
OCXL requires that AFUs use an opaque object handle to represent an AFU interrupt. The specification does not provide a common means to communicate the object handle to the AFU - each AFU must define this within the AFU specification. To support this model, the object handle must be passed back to the core driver as it manages the AFU specification (SISLite) for cxlflash. Note that for Power systems, the object handle is the effective address of the trigger page. Signed-off-by:
-
Uma Krishnan authored
The cxlflash core fop API requires a way to invoke the fault and release handlers of underlying transports using their native file-based APIs. This provides the core with the ability to insert selectively itself into the processing stream of these operations for cleanup. Implement these two fops to map and release when requested. Signed-off-by:
-
Uma Krishnan authored
The cxlflash userspace API requires that users be able to mmap and release the adapter context. Support mapping by implementing the AFU mmap fop to map the context MMIO space and install the corresponding page table entry upon page fault. Similarly, implement the AFU release fop to terminate and clean up the context when invoked. Signed-off-by:
-
Uma Krishnan authored
The cxlflash userspace API requires that users be able to read the adapter context for any pending events or interrupts from the AFU. Support reading various events by implementing the AFU read fop to copy out event data. Signed-off-by:
-
Uma Krishnan authored
The cxlflash userspace API requires that users be able to poll the adapter context for any pending events or interrupts from the AFU. Support polling on various events by implementing the AFU poll fop using a waitqueue. Signed-off-by:
-
Uma Krishnan authored
User contexts request interrupts and are started using the "start work" interface. Populate the start_work() fop to allocate and map interrupts before starting the user context. As part of starting the context, update the user process identification logic to properly derive the data required by the SPA. Also, introduce a skeleton interrupt handler using a bitmap, flag, and spinlock to track interrupts. This handler will be expanded in future commits. Signed-off-by:
-
Uma Krishnan authored
Add support to map and unmap the irq space and manage irq registrations with the kernel for each allocated AFU interrupt. Also support mapping the physical trigger page to obtain an effective address that will be provided to the cxlflash core in a future commit. Signed-off-by:
-
