- 15 Apr, 2016 26 commits
-
-
Sumit Saxena authored
In OCR path, before calling chip reset calls function megasas_wait_for_outstanding_fusion to check reason for OCR. In case of firmware FAULT initiated OCR and DCMD timeout initiated timeout, driver will clear any outstanding reply (yet to be processed by driver) in reply queues before going for chip reset. This code is added to handle a scenario when IO timeout initiated adapter reset and management application initiated adapter reset (by sending command to FAULT firmware) happens simultaneously since adapter reset function is safe-guarded by reset_mutex so only thread will be doing controller reset. Consider IO timeout thread gets mutex and proceeds with adapter reset process after disabling interrupts and by the time management application has fired command to firmware to do adapter reset and the same command is completed by firmware but since interrupts are disabled, driver will not get completion and the same command will be in outstanding/pending commands list of driver and refires same command from IO timeout thread after chip reset which will again FAULT firmware and eventually causes kill adapter. Signed-off-by:
Sumit Saxena <sumit.saxena@broadcom.com> Reviewed-by:
Hannes Reinicke <hare@suse.de> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
Sumit Saxena authored
This patch will reduce memory footprints of megaraid_sas driver when booted in kdump mode. Driver will not allocate memory for optional and perfromance oriented features. Below are key changes done in megaraid_sas driver to do this: 1. Limit Controller's queue depth to 100 in kdump mode. 2. Do not allocate memory for system info buffer and PD info buffer. 3. Disable performance oriented features e.g. Disable RDPQ mode, disable dual queue depth, restrict to single MSI-x vector. Signed-off-by:
-
Bart Van Assche authored
This patch avoids that building with W=1 causes gcc to report the following type of warning: no previous prototype for ... [-Wmissing-prototypes] Signed-off-by:Bart Van Assche <bart.vanassche@sandisk.com> Reviewed-by:
-
Tina Ruchandani authored
'struct timeval' will have its tv_sec value overflow on 32-bit systems in year 2038 and beyond. This patch replaces the use of struct timeval for computing mpi_request.TimeStamp, and instead uses ktime_t which provides 64-bit seconds value. The timestamp computed remains unaffected (milliseconds since Unix epoch). Signed-off-by:
Tina Ruchandani <ruchandani.tina@gmail.com> Reviewed-by:
Arnd Bergmann <arnd@arndb.de> Reviewed-by:
Johannes Thumshirn <jthumshirn@suse.de> Acked-by:
Sathya Prakash <sathya.prakash@broadcom.com> Signed-off-by:
-
David Daney authored
It is unclear what the original intent of the masking was, but it is clearly incorrect to truncate a physical address before calling ioremap(). On systems where there are valid physical address bits above bit-31 (arm64 for example) the result is an eventual OOPs when initializing the driver. Remove the bogus code to fix it. Signed-off-by:
David Daney <david.daney@cavium.com> Acked-by:
Jack Wang <jinpu.wang@profitbricks.com> Signed-off-by:
-
Tina Ruchandani authored
struct register_host_info stores a 64-bit UTC system time timestamp. This patch removes the use of 'struct timeval' to obtain that timestamp as its tv_sec value will overflow on 32-bit systems in year 2038 beyond. The patch uses ktime_get_real_seconds() which returns a 64-bit seconds value. Signed-off-by:
Himanshu Madhani <himanshu.madhani@qlogic.com> Signed-off-by:
-
John Garry authored
Signed-off-by:
John Garry <john.garry@huawei.com> Reviewed-by:
Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by:
-
John Garry authored
This patch fixes the ITCT table setup as it should be configured differently for SAS and SATA devices. For SATA disks there is no need to set qw2 (already zeroed). Also, link parameters for Bus inactive limit, max connection time limit, and reject to open limit timers parameters are changed to match global config register, MAX_CON_TIME_LIMIT_TIME, as recommended by hw team. Signed-off-by:
-
John Garry authored
This patch adds support for directly attaching SATA disks to phy 4-8. The problem was that only registers concerned with phy 0-3 were being considered in sata_int_v2_hw(). The issue was not detected previously as the development board only exposed phy 0-3; the new board provides access to 8 phys. Signed-off-by:
-
John Garry authored
Intermittently it is found that when multiple SATA disks are directly connected to the host that some disks are not detected. The problem is that all set bitfields in ENT_INT_SRC1 are cleared for all phys in sata_int_v2_hw() - it should clear the set bit for the phy being serviced. Also unnecessary double-write to ENT_INT_SRC1 and ENT_INT_SRC_MSK1 is removed (remaining writes are done at end label). Signed-off-by:
-
John Garry authored
Contrary to the field name, the MCR (max connection rate) in the ITCT should hold the device linkrate (linkrate of the connected phy), and not the max linkrate. This fixes an issue seen where some SATA drives connected through an expander which would not attach. Signed-off-by:
-
Ming Lin authored
Now it's ready to move the mempool based SG chained allocator code from SCSI driver to lib/sg_pool.c, which will be compiled only based on a Kconfig symbol CONFIG_SG_POOL. SCSI selects CONFIG_SG_POOL. Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
Ming Lin <ming.l@ssi.samsung.com> Reviewed-by:
Sagi Grimberg <sagi@grimberg.me> Signed-off-by:
-
Ming Lin authored
Rename SCSI_MAX_SG_SEGMENTS to SG_CHUNK_SIZE, which means the amount we fit into a single scatterlist chunk. Rename SCSI_MAX_SG_CHAIN_SEGMENTS to SG_MAX_SEGMENTS. Will move these 2 generic definitions to scatterlist.h later. Reviewed-by:
Tejun Heo <tj@kernel.org> Reviewed-by:
-
Ming Lin authored
Rename SCSI specific struct and functions to more genenic names. Reviewed-by:
Sagi Grimberg <sgi@grimberg.me> Signed-off-by:
-
Ming Lin authored
Parameter "bool mq" is block driver specific. Change it to "first_chunk" to make it more generic. Reviewed-by:
-
Ming Lin authored
Replace parameter "struct scsi_data_buffer" with "struct sg_table" in SG alloc/free functions to make them generic. Reviewed-by:
-
Hannes Reinecke authored
zfcp has its own mechanism for selective scanning, so revert to the original scanning behaviour to not confuse users. Fixes: 4e91e876e9b8b6eb4255aa0d690778a89d3f1d28 Suggested-by:
Benjamin Block <bblock@linux.vnet.ibm.com> Signed-off-by:
Hannes Reinecke <hare@suse.com> Reviewed-by:
-
Leonid Moiseichuk authored
Claim Marvell 9485 controllers regardless of subdevice ID. Tested on ASUS P9A-I/C2550/SAS/4L which uses vendor-specific 1043:8635. [mkp: Tweaked commit message] Signed-off-by:
Leonid Moiseichuk <leonid.moiseichuk@gmail.com> Reviewed-by:
Andy Shevchenko <andy.shevchenko@gmail.com> Signed-off-by:
-
Chad Dupuis authored
Signed-off-by:
Chad Dupuis <chad.dupuis@qlogic.com> Reviewed-by:
-
Chad Dupuis authored
When we are in connection recovery and the internal command timer on a request pops, either the scsi_cmnd->device or scsi_cmnd->device->host back pointers may be NULL as the device that the command that the request was submitted on may have been subsequently reaped due to the connection recovery. This can cause one or both of the pointers above to be NULL and cause a system crash if we try to return the command to the midlayer. Instead, double check the pointers before the return to the midlayer so as to prevent the crash and let the upper layers finish the session recovery and rediscover the device. Signed-off-by:
-
Chad Dupuis authored
Signed-off-by:
-
Chad Dupuis authored
Signed-off-by:
-
Joe Carnuccio authored
Per customer request, add the following driver tunables: o devloss_tmo o max_luns o queue_depth o tm_timeout tm_timeout is set per scsi_host in /sys/class/scsi_host/hostX/tm_timeout. Signed-off-by:
Joe Carnuccio <joe.carnuccio@qlogic.com> Signed-off-by:
-
Hannes Reinecke authored
VPD pages 0x0 and 0x83 are mandatory even for SPC-2, so we should be lowering the restriction to avoid having to whitelist every SPC-2 compliant device. Signed-off-by:
-
Johannes Thumshirn authored
Now that we've done a more comprehensive fix with the intermediate target state we can remove the previous hack introduced with commit 90a88d6e ("scsi: fix soft lockup in scsi_remove_target() on module removal"). Signed-off-by:
Ewan D. Milne <emilne@redhat.com> Reviewed-by:
-
Johannes Thumshirn authored
Add intermediate STARGET_REMOVE state to scsi_target_state to avoid running into the BUG_ON() in scsi_target_reap(). The STARGET_REMOVE state is only valid in the path from scsi_remove_target() to scsi_target_destroy() indicating this target is going to be removed. This re-fixes the problem introduced in commits bc3f02a7 ("[SCSI] scsi_remove_target: fix softlockup regression on hot remove") and 40998193 ("scsi: restart list search after unlock in scsi_remove_target") in a more comprehensive way. [mkp: Included James' fix for scsi_target_destroy()] Signed-off-by:
Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Tested-by:
James Bottomley <jejb@linux.vnet.ibm.com> Signed-off-by:
-
- 11 Apr, 2016 14 commits
-
-
Narsimhulu Musini authored
scsi host is added after negotiating the max number of IOs with Firmware. Signed-off-by:
Narsimhulu Musini <nmusini@cisco.com> Signed-off-by:
Sesidhar Baddela <sebaddel@cisco.com> Signed-off-by:
-
Narsimhulu Musini authored
While HBA reset is in progress, if IO/TM completion is received for the same IO then IO/TM completion path releases the driver private resources associated with IO. This fix prevents releasing the resources in IO and TM completion path if HBA reset is in progress. Signed-off-by:
-
Narsimhulu Musini authored
- On posting an IO to the firmware, adapter generates an interrupt. Due to hardware issues, sometimes the adapter fails to generate the interrupt. This behavior skips updating transmit queue- counters, which in turn causes the queue full condition. The fix addresses the queue full condition. - The fix also reserves a slot in transmit queue for hba reset. when queue full is observed during IO, there will always be room to post hba reset command. Signed-off-by:
-
Narsimhulu Musini authored
Fix deletes the snic targets synchronously prior to deletion of host. Signed-off-by:
-
Narsimhulu Musini authored
Fix handles control path queue issues such as queue full and sudden removal of hardware. Signed-off-by:
-
Narsimhulu Musini authored
- LUN goes offline if there are at least two scsi command timeouts Completing the IO with scsi_done() fixes the issue. Signed-off-by:
-
Narsimhulu Musini authored
Adding additional stats, and fixed logging messages. - Added qdepth change stats - Added separate isr stats for each type of interrupt - Fixed race in updating active IOs - Suppressed Link event message for DAS backend. Signed-off-by:
-
James Smart authored
Signed-off-by:
Dick Kennedy <dick.kennedy@avagotech.com> Signed-off-by:
James Smart <james.smart@avagotech.com> Signed-off-by:
-
James Smart authored
Signed-off-by:
-
James Smart authored
Signed-off-by:
-
James Smart authored
Revise versions to reflect SmartSAN 2.0 support RDP updated to support additional descriptors: Credit descriptor Optical Element Data descriptors for Temperature, Voltage, Bias current, TX power and TX power. Optical Product Data descriptor. Signed-off-by: -
James Smart authored
When the switch is rebooted, the lpfc driver fails to log into the fabric, and Unexpected timeout message is seen. Fix: Do not issue RegVFI if the FLOGI was internally aborted. Signed-off-by:
-
James Smart authored
Pass cmd iu payloads inline to adapter job structure rather than as separate dma buffers. Signed-off-by:
-
James Smart authored
The default rpi completion handler does back to back puts to force the removal of the ndlp. This ends up calling lpfc_unreg_rpi after the reference count is at 0. Fix: Check the reference count of the ndlp before getting the ref to make sure we are not getting a reference on a removed object. Signed-off-by:
-
