- 13 Jun, 2017 40 commits
-
-
James Smart authored
Administrator intervention is currently required to get good numbers when switching from running latency tests to IOPS tests. The configured interrupt coalescing values will greatly effect the results of these tests. Currently, the driver has a single coalescing value set by values of the module attribute. This patch changes the driver to support auto-configuration of the coalescing value based on the total number of outstanding IOs and average number of CQEs processed per interrupt for an EQ. Values are checked every 5 seconds. The driver defaults to the automatic selection. Automatic selection can be disabled by the new lpfc_auto_imax module_parameter. Older hardware can only change interrupt coalescing by mailbox command. Newer hardware supports change via a register. The patch support both. Signed-off-by:
Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by:
James Smart <james.smart@broadcom.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
James Smart authored
Addressed the following reported defects: ** CID 1411552: Control flow issues (MISSING_BREAK) /drivers/scsi/lpfc/lpfc_sli.c: 13259 in lpfc_sli4_nvmet_handle_rcqe() ** CID 1411553: Memory - illegal accesses (OVERRUN) /drivers/scsi/lpfc/lpfc_sli.c: 16218 in lpfc_fc_frame_check() ** CID 1411553: Memory - illegal accesses (OVERRUN) Overrunning array "lpfc_rctl_names" of 202 8-byte elements at element index 244 (byte offset 1952) using index "fc_hdr->fh_r_ctl" (which evaluates to 244). ** CID 1411554: Null pointer dereferences (REVERSE_INULL) /drivers/scsi/lpfc/lpfc_nvmet.c: 2131 in lpfc_nvmet_unsol_fcp_abort_cmp() ** CID 1411555: Memory - illegal accesses (UNINIT) /drivers/scsi/lpfc/lpfc_nvmet.c: 180 in lpfc_nvmet_ctxbuf_post() Signed-off-by:
-
James Smart authored
vports cannot login to target. For vports, lpfc_nodelist is allocated for targets only on completion of GFF_ID command. Driver checks if lpfc_nodelist exists for target before sending GFF_ID. So, GFF_ID and PLOGI are not sent. As mentioned by the comment in lpfc_prep_node_fc4type() routine, do not send GFF_ID only if this NPortID is previously identified as FCP target. Send GFF_ID if it is a newly identified remote port from GID_FT response. Signed-off-by:
-
James Smart authored
The nvmet driver was rejecting the initiator's PRLI because its reg_rpi for the PLOGI was still outstanding. The initiator would resend the PRLI without delay and get the same answer. The PRLI retries would exhaust causing the nvme initiator to set the nvmet ndlp to UNMAPPED. The driver's lpfc_els_retry handler did not have a policy for an LS_RJT with explanation CMD_IN_PROGRESS for PRLI or NVME_PRLI. This caused the delay to remain at 0 but retry set 1. Fix: When the ELS response is LS_RJT, TPC and the command was PRLI or NVME_PRLI, just set the delay to 1000 mS to get a 1 second delay on the PRLI retry. This was enough to allow the REG_RPI to complete at the target. Signed-off-by:
-
James Smart authored
Kernel panic when log_verbose is set to 0xffffffff phba->pport is dereferenced before it is initialized Fix: Do not dereference phba->pport if it is NULL Signed-off-by:
-
James Smart authored
System panic with general protection fault during driver load The driver uses a static array sli4_hba.handler_name to store the irq handler names. If the io_channel_irqs exceeds the pre-allocated size (32+1), then the driver will overwrite other fields of sli4_hba. Fix: Dynamically allocate handler_name. Signed-off-by:
-
James Smart authored
Null pointer dereference when BFS VM is powered off The driver incorrectly uses sli3_ring on SLI-4 adapters Use the correct ring structure based on sli_rev Signed-off-by:
Raphael Silva <raphasil@linux.vnet.ibm.com> Signed-off-by:
-
James Smart authored
On hbacmd reset failure, observing wrong string "nline" in kernel log. On failure, non negative value (1) is returned from sysfs store routine. It is interpreted as count by kernel and store routine is called again with the remaining characters as input. Fix: Return negative error code (-EIO) in case of failure. Signed-off-by:
-
James Smart authored
NVME FC counters don't reflect actual results Since counters are not atomic, or protected by a lock, the values often get screwed up. Make them atomic, like NVMET. Fix up sysfs and debugfs display accordingly Added Outstanding IOs to stats display Signed-off-by:
-
James Smart authored
Observing lpfc port down after issuing hbacmd reset command Failure in posting SGL buffers. If there is only one SGL buffer and rrq is valid for its XRI, we are rightly returning NULL but not adding the buffer back to the SGL list. So, number of buffers become less than total count and repost fails during reset. Add SGL buffer back to list before returning NULL. Signed-off-by:
-
James Smart authored
When unloading the driver, the NVMET driver would wait the full 30 seconds for its UNMAPPED initiator node to get removed before continuing with the unload process. NVMEI worked correctly. For each rport put into UNMAPPED or MAPPED state by NVMET, the driver puts a reference on the NDLP. The difference is that NVMEI has a unregister call for its rports and the extra reference is removed in the unregister process. For NVMET, the driver has to remove the reference explicitly when dropping out of UNMAPPED or MAPPED because there is no unregister call. Add a call to lpfc_nlp_put on the ndlp when NVMET and the old state was UNMAPPED or MAPPED. Signed-off-by:
-
James Smart authored
Lun Priority level shown as NA Remote port is not getting registered for nameserver and fdmi. Due to which dfc SendCTPassThru cmd is failing. Made changes to register the remote port for both. Signed-off-by:
-
James Smart authored
Inconsistent error messages and context state checks Context state sanity checks were not accurate or inconsistent in the code paths. Separated LS context states from FCP. Added and modified context state sanity checks. Use context state to determine if a sol or unsol ABORT is needed. Signed-off-by:
-
James Smart authored
While debugging Devloss and recovery, debugfs and sysfs were found to not show the NVME port roles consistently. The port role FC_PORT_ROLE_NVME_DISCOVERY was added with the devloss bringup and the other issues were just oversight. Add NVME Target and DISCSRVC to debugfs nodeinfo and sysfs nvme info handlers. The full port role was added to the NVME data only not the generic nodelist. Signed-off-by:
-
James Smart authored
As the devloss API was implemented in the nvmei driver, an evaluation of the nvme transport and the lpfc driver showed dual management of the rports. This creates a bug possibility when the thread count and SAN size increases. The nvmei driver code was based on a very early transport and was not revisited until the devloss API was introduced. Remove the listhead in the driver's rport data structure and the listhead in the driver's lport data structure. Remove all rport_list traversal. Convert the driver to use the nrport (nvme rport) pointer that is now NULL or nonNULL depending on a devloss action. Convert debugfs and nvme_info in sysfs to use the fc_nodes list in the vport. Signed-off-by:
-
James Smart authored
Add nvme initiator devloss support The existing implementation was based on no devloss behavior in the transport (e.g. immediate teardown) so code didn't properly handle delayed nvme rport device unregister calls. In addition, the driver was not correctly cycling the rport port role for each register-unregister-reregister process. This patch does the following: Rework the code to properly handle rport device unregister calls and potential re-allocation of the remoteport structure if the port comes back in under dev_loss_tmo. Correct code that was incorrectly cycling the rport port role for each register-unregister-reregister process. Prep the code to enable calling the nvme_fc transport api to dynamically update dev_loss_tmo when the scsi sysfs interface changes it. Memset the rpinfo structure in the registration call to enforce "accept nvme transport defaults" in the registration call. Driver parameters do influence the dev_loss_tmo transport setting dynamically. Simplifies the register function: the driver was incorrectly searching its local rport list to determine resume or new semantics, which is not valid as the transport already handles this. The rport was resumed if the rport handed back matches the ndlp->nrport pointer. Otherwise, devloss fired and the ndlp's nrport is NULL. Signed-off-by:
-
Christophe JAILLET authored
Merge some quoted strings to improve readability and to save some lines of code. Signed-off-by:
Christophe JAILLET <christophe.jaillet@wanadoo.fr> Acked-by:
Chad Dupuis <chad.dupuis@cavium.com> Signed-off-by:
-
Christophe JAILLET authored
Replace some 'dma_alloc_coherent+memset' by some quivalent 'dma_zalloc_coherent' in order to reduce code verbosity Signed-off-by:
-
Christophe JAILLET authored
We should return -ENOMEM in case of memory allocation error, as done elsewhere in this function. [mkp: fixed typo] Fixes: 61d8658b ("scsi: qedf: Add QLogic FastLinQ offload FCoE driver framework.") Signed-off-by:
-
Binoy Jayan authored
The semaphore 'fs_api_semaphore' is used as a simple mutex, so it should be written as one. Semaphores are going away in the future. Signed-off-by:
Binoy Jayan <binoy.jayan@linaro.org> Reviewed-by:
Arnd Bergmann <arnd@arndb.de> Signed-off-by:
-
Binoy Jayan authored
The semaphore 'fm_api_semaphore' is used as a simple mutex, so it should be written as one. Semaphores are going away in the future. Signed-off-by:
-
Himanshu Madhani authored
Fixes following 0-day kernel build warnings: drivers/scsi/qla2xxx/qla_init.c:6407:50: warning: format '%lx' expects argument of type 'long unsigned int', but argument 5 has type 'unsigned int' [-Wformat=] drivers/scsi/qla2xxx/qla_init.c:6709:50: warning: format '%lx' expects argument of type 'long unsigned int', but argument 5 has type 'unsigned int' [-Wformat=] Fixes: b95b9452 ("scsi: qla2xxx: Fix crash due to mismatch mumber of Q-pair creation for Multi queue") Signed-off-by:
Himanshu Madhani <himanshu.madhani@cavium.com> Signed-off-by:
Giridhar Malavali <giridhar.malavali@cavium.com> Signed-off-by:
-
Colin Ian King authored
An earlier commit ed7fb808477b846bb2 ("scsi: qla2xxx: Remove redundant wait when target is stopped.") removed a null check on ha->tgt.tgt_ops and replaced it with a new check that null checked tgt, thus making the subsequent null check on tgt totally redundant. Remove it. Detected by CoverityScan, CID#1440452 ("Logically Dead Code") Signed-off-by:Colin Ian King <colin.king@canonical.com> Acked-by:
-
Adrian Hunter authored
Add PCI id and variant ops for Intel CNL UFS host controller. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
-
Adrian Hunter authored
Put PM functions under correct config options and use standard PM macros to set callbacks. Signed-off-by:
-
Colin Ian King authored
functions lpfc_nvmet_cleanup_io_context and lpfc_nvmet_setup_io_context can be made static as they do not need to be in global scope. Cleans up sparse warnings: "warning: symbol 'lpfc_nvmet_cleanup_io_context' was not declared. Should it be static?" "warning: symbol 'lpfc_nvmet_setup_io_context' was not declared. Should it be static?" Signed-off-by:
-
Bart Van Assche authored
Since the SCSI core zeroes driver-private command data, remove that code from the xen-scsifront driver. Signed-off-by:
Bart Van Assche <bart.vanassche@sandisk.com> Reviewed-by:
Hannes Reinecke <hare@suse.com> Reviewed-by:
Juergen Gross <jgross@suse.com> Reviewed-by:
-
Bart Van Assche authored
Since the SCSI core zeroes driver-private command data, remove that code from the virtio driver. Signed-off-by:
-
Bart Van Assche authored
Since the SCSI core zeroes driver-private command data, remove that code from the snic driver. Signed-off-by:
Narsimhulu Musini <nmusini@cisco.com> Reviewed-by:
-
Bart Van Assche authored
This patch reduces code duplication. There are two functional changes in this patch: - It causes scsi_mq_prep_fn() to clear driver-private command data, just like the already upstream commit 1bad6c4a ("scsi: zero per-cmd private driver data for each MQ I/O"). - The initialization of .prot_sdb is moved from scsi_mq_prep_fn() into scsi_init_request(). [mkp: applied by hand] Signed-off-by:
-
Bart Van Assche authored
This patch does not change any functionality but makes the next patch easier to read. Signed-off-by:
-
Bart Van Assche authored
Just like for the scsi-mq code path, in the single queue SCSI code path only add commands to the per-device command list if required by the SCSI LLD. This patch will make it easier to merge the single-queue and multiqueue command initialization code. Signed-off-by:
-
Bart Van Assche authored
If a device is blocked, make __scsi_remove_device() cause it to transition to the DEL state. This means that all the commands issued in .shutdown() will error in the mid-layer, thus making the removal proceed without being stopped. This patch is a slightly modified version of a patch from James Bottomley. This patch avoids that the following lockup occurs: Call Trace: schedule+0x35/0x80 schedule_timeout+0x237/0x2d0 io_schedule_timeout+0xa6/0x110 wait_for_completion_io+0xa3/0x110 blk_execute_rq+0xdf/0x120 scsi_execute+0xce/0x150 [scsi_mod] scsi_execute_req_flags+0x8f/0xf0 [scsi_mod] sd_sync_cache+0xa9/0x190 [sd_mod] sd_shutdown+0x6a/0x100 [sd_mod] sd_remove+0x64/0xc0 [sd_mod] __device_release_driver+0x8d/0x120 device_release_driver+0x1e/0x30 bus_remove_device+0xf9/0x170 device_del+0x127/0x240 __scsi_remove_device+0xc1/0xd0 [scsi_mod] scsi_forget_host+0x57/0x60 [scsi_mod] scsi_remove_host+0x72/0x110 [scsi_mod] srp_remove_work+0x8b/0x200 [ib_srp] Reported-by:
Israel Rukshin <israelr@mellanox.com> Signed-off-by:
Hannes Reinecke <hare@suse.de> Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Cc: Israel Rukshin <israelr@mellanox.com> Cc: Max Gurtovoy <maxg@mellanox.com> Cc: Benjamin Block <bblock@linux.vnet.ibm.com> Signed-off-by:
-
Bart Van Assche authored
This patch does not change any functionality. Signed-off-by:
-
Bart Van Assche authored
Serializing SCSI device state changes avoids that two state changes can occur concurrently, e.g. the state changes in scsi_target_block() and __scsi_remove_device(). This serialization is essential to make patch "Make __scsi_remove_device go straight from BLOCKED to DEL" work reliably. Enable this mechanism for all scsi_target_*block() callers but not for the scsi_internal_device_unblock() calls from the mpt3sas driver because that driver can call scsi_internal_device_unblock() from atomic context. Signed-off-by:
-
Bart Van Assche authored
This will make it easier to serialize SCSI device state changes through a mutex. Signed-off-by:
Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by:
-
Bart Van Assche authored
Instead of passing a "wait" argument to scsi_internal_device_block(), split this function into a function that waits and a function that doesn't wait. This will make it easier to serialize SCSI device state changes through a mutex. Signed-off-by:
-
Bart Van Assche authored
Dereferencing shost from scsi_exit_rq() is not safe because the SCSI host may already have been freed when scsi_exit_rq() is called. Increasing the shost reference count in scsi_init_rq() and dropping that reference in scsi_exit_rq() is nontrivial since scsi_host_dev_release() may sleep and since scsi_exit_rq() may be called from interrupt context. Since scsi_exit_rq() only needs a single bit from shost, copy that bit into struct scsi_cmnd. Reported-by:
Scott Bauer <scott.bauer@intel.com> Fixes: e9c787e6 ("scsi: allocate scsi_cmnd structures as part of struct request") Signed-off-by:
-
Jia-Ju Bai authored
The driver may sleep under a spin lock, and the function call path is: mraid_mm_attach_buf (acquire the lock by spin_lock_irqsave) pci_pool_alloc(GFP_KERNEL) --> may sleep To fix it, the "GFP_KERNEL" is replaced with "GFP_ATOMIC". [mkp: fixed whitespace] Signed-off-by:
Jia-Ju Bai <baijiaju1990@163.com> Acked-by:
Sumit Saxena <sumit.saxena@broadcom.com> Signed-off-by:
-
Szymon Mielczarek authored
Move the initialization of clocks list head to ufshcd_alloc_host() so that every driver doesn't have to do it. Remove checks for the list head being NULL because that is not possible. Signed-off-by:
Szymon Mielczarek <szymonx.mielczarek@intel.com> Signed-off-by:
Subhash Jadavani <subhashj@codeaurora.org> Signed-off-by:
-
