- 13 Aug, 2019 5 commits
-
-
Bart Van Assche authored
Since the put_unaligned_*() macros are used in this header file, include the header file that defines these macros. Cc: Himanshu Madhani <hmadhani@marvell.com> Fixes: 15b7a68c ("scsi: qla2xxx: Introduce the dsd32 and dsd64 data structures") # v5.2-rc1. Signed-off-by:
Bart Van Assche <bvanassche@acm.org> Tested-by:
Himanshu Madhani <hmadhani@marvell.com> Reviewed-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
Bart Van Assche authored
This patch only modifies whitespace. Cc: Himanshu Madhani <hmadhani@marvell.com> Signed-off-by:
-
Bart Van Assche authored
Insert a space where required, surround complex expressions in macros with parentheses, use the UL suffix instead of the (unsigned long) cast, do not use line continuations when not necessary and do not explicitly initialize static variables to zero. Cc: Himanshu Madhani <hmadhani@marvell.com> Signed-off-by:
-
Bart Van Assche authored
I'm not sure how this happened but the patch that was intended to fix abort handling was incomplete. This patch fixes that patch as follows: - If aborting the SCSI command failed, wait until the SCSI command completes. - Return SUCCESS instead of FAILED if an abort attempt races with SCSI command completion. - Since qla2xxx_eh_abort() increments the sp reference count by calling sp_get(), decrement the sp reference count before returning. Cc: Himanshu Madhani <hmadhani@marvell.com> Fixes: 219d27d7 ("scsi: qla2xxx: Fix race conditions in the code for aborting SCSI commands") Signed-off-by:
-
Bart Van Assche authored
Since qla2x00_abort_srb() starts with increasing the reference count of @sp, decrease that same reference count before returning. Cc: Himanshu Madhani <hmadhani@marvell.com> Fixes: 219d27d7 ("scsi: qla2xxx: Fix race conditions in the code for aborting SCSI commands") # v5.2. Signed-off-by:
-
- 08 Aug, 2019 35 commits
-
-
Suganath Prabu authored
Updated driver version from 29.100.00.00 to 31.100.00.00 which is equivalent to Phase 12 OOB. Signed-off-by:
Suganath Prabu <suganath-prabu.subramani@broadcom.com> Signed-off-by:
-
Suganath Prabu authored
In some cases, like while performing extensive expander reset or phy reset, user may observe that drives are not visible in OS. Driver's firmware-worker thread is blocked for more than 120 seconds resulting in a call trace. 1. Received target add event for Device A and hence driver has registered this device to SML by calling sas_rphy_add(). SML has half added this device and returned the control to the driver by quitting from sas_rphy_add() API, and started some background scanning on this device A. 2. While background scanning is going on device A, driver has received SAS DEVICE STATUS CHANGE EVENT with RC code "Internal device reset" event and hence driver has set tm_busy flag for this Device A from FW worker thread context. When tm_busy flag is set then driver return scsi commands with device busy status asking the kernel to retry the command after some time. So background scanning for device A will be waiting for this tm_busy to be cleared. 3. Meanwhile driver has received a target add event for Device B and hence driver called sas_rphy_add() API to register this device with SML. But since background scanning for Device A is still pending and SML is not quitting from sas_rphy_add(), the driver’s firmware worker thread got blocked. 4. Now driver has received SAS DEVICE STATUS CHANGE EVENT with RC code "Internal device reset complete" event. But as driver’s firmware worker thread got blocked in Step3, it can’t process this event and it was not clearing the tm_busy flag and deadlock occurred (where SML was waiting for tm_busy flag to be cleared and our FW worker thread is waiting for SML to quit from sas_device_rphy_add() API). Same deadlock will be observed even if device B is getting removed in step3. So to limit these types of deadlocks driver will process the SAS DEVICE STATUS CHANGE EVENT events from ISR context instead of processing this event from worker thread context. This improvement avoids above deadlock. Signed-off-by:
-
Suganath Prabu authored
This patch is to reduce the performance drop depth observed on SATA HDD when ATA PT command is outstanding. Driver returns IO commands with status "SAM_STAT_BUSY" whenever ATA PT command is outstanding. With this, IO commands will be retried until this outstanding ATA PT to complete and hence we will observe drop in performance. As the driver is completing the subsequent IOs commands with SAM_STAT_BUSY status, these IOs has to go though the block layer. Hence it adds latency to the IOs and large performance drop is observed. So to reduce this performance dropp, added improvement in driver to return the subsequent IOs with SCSI_MLQUEUE_DEVICE_BUSY status instead of completing the IOs with SAM_STAT_BUSY status when ATA PT command is outstanding. Sending command back with SCSI_MLQUEUE_DEVICE_BUSY does not go through complete block layer stack (as scsi_done won't be called) SML will immediately retry the command and this method will avoid latency of block layer stack and the performance impact will be reduced. On Local setup, ran 512k sequential read IO operation on HGST SATA drive with existing driver & with this improvement drivers and here is the result, 1. With existing driver: IOs are running at bandwidth of ~230 rMB/s and whenever any ATA PT command is outstanding (e.g issued from systemd-udevd daemon) then this bandwidth drops to ~150 rMB/s. 2. With this improvement driver: IOs are running at bandwidth of ~230 rMB/s and whenever any ATA PT command is outstanding then this bandwidth drops to just ~190 rMB/s. Signed-off-by:
-
Suganath Prabu authored
During HBA initialization time, if handshake operation fails due to some firmware fault then currently driver is terminating the HBA initialization. It is possible that HBA may come up properly if diag reset is issued. So improvement is made in driver in such a way that before terminating the HBA initialization, driver checks the IOC state and if IOC state is in fault state then issue diag reset for once. If diag reset is successful then continue with HBA initialization else terminate the HBA initialization. Signed-off-by:
-
Suganath Prabu authored
Currently with sysfs parameter "drv_support_bitmap" driver exposes whether driver supports toolbox memory move command or not. And application should issue the toolbox memory move command only if driver tell that memory move tool box command is supported through this sysfs parameter. In future we can utilize this sysfs parameter if any new feature is added and need to notify the same to applications. Signed-off-by:
-
Suganath Prabu authored
Host uses the Memory Move Tool to copy data from any source/destination combination of system memory and IOC memory. Memory Move Tool box request contains two SGE fields, First SGE field must contains the source buffer details described by an MPI Simple SGE. The second SGE field must contains the destination buffer details described by an MPI Simple SGE. Source -> Destination 1. IOC -> IOC (Both the SGE's will be filled by application) 2. HOST -> HOST (Both the SGE's will be filled by the host, application should give sgl_offset to first SGE offset) 3. IOC -> HOST (Application will fill the first SGE and set the sgl_offset to second SGE and hence driver fills the second SGE) 4. HOST -> IOC (Application will fill IOC buffer information in the first SGE and set the sgl_offset to second SGE. Then driver will fill the second SGE with Host buffer information and just before posting the command to the firmware, driver will swap these two SGEs so that first SGE contains the HOST buffer information and second SGE contains the IOC information. Driver has to take care only of the 4th case, other three cases are by default supported by the current driver design. Signed-off-by: -
Suganath Prabu authored
If driver sees the NVMe drive with "DEVICE_BLOCKED" AccessStatus in its PCIe Device Page0, then driver removes the drive from its internal list and does not allow any IOCTL commands to be sent to the drive and will return the IOCTLs with "-ENODEV" status. The driver will now allow NVMe Encapsulated IOCTL issued to the NVMe device with an access status of DEVICE_BLOCKED. This change allows the user to flash new drive firmware online and revive the drive. Add NVMe device only the driver's internal list even though the device is in the blocked state so that the device will be visible to Apps. This way Apps can send NVMe Encapsulated IOCTLs to this drive and bring the drive online. This NVMe drive with DEVICE_BLOCKED access status won't added to the SML, it will be added only in the driver's internal list. [mkp: clarified desc] Signed-off-by:
-
Suganath Prabu authored
SES device of managed PCIe switch will be enumerated same as NVMe drives. The device info type for this SES device is MPI26_PCIE_DEVINFO_SCSI (0x4), whereas the device info type for NVMe drives is MPI26_PCIE_DEVINFO_NVME (0x3). Based on this device info type driver determines whether the device is NVMe drive or a SES device of a managed PCIe switch. This SES device doesn't have the PCIe device page 2 information like NVMe drives, so driver won't read PCIe device page 2 information for SES device. This SES device uses only IEEE SGL's, So driver build's IEEE SGL's whenever it receives any SCSI commands for this SES device. Signed-off-by: -
Suganath Prabu authored
Updated MPI to 2.6.8 specification and header files to 2.00.54. Signed-off-by:
-
Suganath Prabu authored
Issue: During online Firmware upgrade operations it is possible that MaxDevHandles filled in IOCFacts may change with new FW. With this we may observe kernel panics when driver try to access the pd_handles or blocking_handles buffers at offset greater than the old firmware's MaxDevHandle value. Fix: _base_check_ioc_facts_changes() looks for increase/decrease in IOCFacts attributes during online firmware upgrade and increases the pd_handles, blocking_handles, etc buffer sizes to new firmware's MaxDevHandle value if this new firmware's MaxDevHandle value is greater than the old firmware's MaxDevHandle value. Signed-off-by:
-
Suganath Prabu authored
Driver gets a request frame from the free pool of DMA-able request frames and fill in the required information and pass the address of the frame to IOC/FW to pull the complete request frame. In certain places the driver used the request frame allocated from the free pool without completely clearing the previous data stored in it. The request contents were cleared only for the size of the new request to be issued and that left out some stale data in the unused part of the request. Though the IOC/FW is not expected to access the request beyond the specified size, it is good practice to clear complete request message frame. So reinitialize the complete request message frame with 0s before using it. Signed-off-by:
-
Suganath Prabu authored
PCIe Lane margin tool box request requires IEEE sgl's and hence driver fills the SGL field with IEEE sgl's while issuing the PCIe Lane margin ioctl request to the HBA firmware. Signed-off-by:
-
Colin Ian King authored
There are several occasions where variable rc is being initialized with a value that is never read and error is being re-assigned a little later on. Clean up the code by removing rc entirely and just returning the return value from the call to pm8001_issue_ssp_tmf Addresses-Coverity: ("Unused value") Signed-off-by:Colin Ian King <colin.king@canonical.com> Acked-by:
Jack Wang <jinpu.wang@cloud.ionos.com> Signed-off-by:
-
Luo Jiaxing authored
In hisi_sas_lu_reset(), we call internal abort for SAS and SATA device codepaths -> consolidate into a single call. Signed-off-by:
Luo Jiaxing <luojiaxing@huawei.com> Signed-off-by:
John Garry <john.garry@huawei.com> Signed-off-by:
-
Xiang Chen authored
The format specifier "%p" can leak kernel address, and use "%pK" instead. Signed-off-by:
Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by:
-
Xiang Chen authored
Remove some unnecessary code, including: - Explicit zeroing of memory allocated for dmam_alloc_coherent() - Some duplicated code - Some redundant masking Signed-off-by:
-
Luo Jiaxing authored
For functions which always return 0, which is never checked, make to return void. Signed-off-by:
-
Xiang Chen authored
It will free irq automatically if devm_request_irq() failed, so drop free_irq() if devm_request_irq() failed. Signed-off-by:
-
John Garry authored
The SMP frame response is written to the command table and not the SMP response pointer from libsas, so don't bother DMA mapping (and unmapping) the SMP response from libsas. Suggested-by:
-
John Garry authored
The call to kmap_atomic() in the SMP command completion code is unnecessary, since kmap() is only really concerned with highmem, which is not relevant on arm64. The controller only finds itself in arm64 systems. Signed-off-by:
-
Xiang Chen authored
For a system with PAGE_SIZE of 16K or 64K, the size every time we want to alloc may be small like 4K, but for function dmam_alloc_coherent(), the least size it allocates is PAGE_SIZE, so it will waste much memory for the situation. To solve the issue, limit the minimum allocation size of slot buf to PAGE_SIZE. Signed-off-by:
-
Xiang Chen authored
For struct hisi_sas_status_buffer, it contains struct hisi_sas_err_record and iu[1024]. The struct iu[1024] will be filled fully by the response of disks, so it is not need to initialize them to 0, but for the struct hisi_sas_err_record, SAS controller only fill some fields of hisi_sas_err_record according to hw designer, so it should be initialised to 0. After the change, cpu utilization percentage of memset() is changed from 1.7% to 0.12%. Signed-off-by:
-
Luo Jiaxing authored
Fix a possible out-of-bounds access in hisi_sas_debug_I_T_nexus_reset(). Signed-off-by:
-
Luo Jiaxing authored
The AXI and RAS register values should also should be snapshot at debugfs. Signed-off-by:
-
Luo Jiaxing authored
The value of IOST/ITCT is updated to cache first, and then synchronize to DDR periodically. So the value in IOST/ITCT cache is the latest data and it's important for debugging. So, the HW cache of IOST and ITCT should be snapshot at debugfs. Signed-off-by:
-
Luo Jiaxing authored
Fix how the pointer is set in hisi_sas_debugfs_iost_show() and hisi_sas_debugfs_itct_show(). Signed-off-by:
-
John Garry authored
In commit 1273d65f ("scsi: hisi_sas: change queue depth from 512 to 4096"), the depth of each queue is the same as the max IPTT in the system. As such, as long as we have an IPTT allocated, we will have enough space on any delivery queue. All .get_free_slot functions were checking for space on the queue by reading the DQ read pointer. Drop this, and also raise the code into common code, as there is nothing hw specific remaining. Signed-off-by:
-
John Garry authored
There is a small optimisation to be had by making the max IPTT the same for all hw revisions, that being we can drop the check for read and write pointer being the same in the get free slot function. Change v1 hw to have max IPTT of 4096 - same as v2 and v3 hw - and drop hisi_sas_hw.max_command_entries. Signed-off-by:
-
Helge Deller authored
Signed-off-by:
Helge Deller <deller@gmx.de> Signed-off-by:
-
Li Zhong authored
If the userspace process crashes while we send the nl msg, it is possible that the cmd in curr_nl_cmd of tcmu_dev never gets reset to 0, and and returns busy for other commands after the userspace process is restartd. More details below: /backstores/user:file/file> set attribute dev_size=2048 Cannot set attribute dev_size: [Errno 3] No such process /backstores/user:file/file> set attribute dev_size=2048 Cannot set attribute dev_size: [Errno 16] Device or resource busy with following kernel messages: [173605.747169] Unable to reconfigure device [173616.686674] tcmu daemon: command reply support 1. [173623.866978] netlink cmd 3 already executing on file [173623.866984] Unable to reconfigure device Also, it is not safe to leave the nl_cmd in the list, and not get deleted. This patch removes the nl_cmd from the list, and clear its data if it is not sent successfully. Signed-off-by:
Li Zhong <lizhongfs@gmail.com> Acked-by:
Mike Christie <mchristi@redhat.com> Signed-off-by:
-
Anil Varughese authored
Configure CDNS_UFS_REG_HCLKDIV in .hce_enable_notify() instead of .setup_clock() because if UFSHCD resets the controller ip because of phy or device related errors then CDNS_UFS_REG_HCLKDIV is reset to default value and .setup_clock() is not called later in the sequence whereas .hce_enable_notify will be called everytime controller is reenabled. Signed-off-by:
Anil Varughese <aniljoy@cadence.com> Reviewed-by:
Vignesh Raghavendra <vigneshr@ti.com> Signed-off-by:
-
Bart Van Assche authored
The data structure used for log messages is so large that it can cause a boot failure. Since allocations from that data structure can fail anyway, use kmalloc() / kfree() instead of that data structure. See also https://bugzilla.kernel.org/show_bug.cgi?id=204119. See also commit ded85c19 ("scsi: Implement per-cpu logging buffer") # v4.0. Reported-by:
Jan Palus <jpalus@fastmail.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Hannes Reinecke <hare@suse.com> Cc: Johannes Thumshirn <jthumshirn@suse.de> Cc: Ming Lei <ming.lei@redhat.com> Cc: Jan Palus <jpalus@fastmail.com> Signed-off-by:
-
Bart Van Assche authored
If scsi_target_block() fails that can break the code that calls this function. Hence complain loudly if scsi_target_block() fails. Cc: Christoph Hellwig <hch@lst.de> Cc: Hannes Reinecke <hare@suse.com> Cc: Johannes Thumshirn <jthumshirn@suse.de> Cc: Ming Lei <ming.lei@redhat.com> Signed-off-by:
-
Bart Van Assche authored
The only 'new_state' values passed by upstream kernel code to scsi_internal_device_unblock_nowait() are SDEV_RUNNING and SDEV_TRANSPORT_OFFLINE. These are the only values that should be passed to this function. Hence check the value of the 'new_state' argument to avoid that scsi_internal_device_unblock_nowait() would be used to trigger an illegal SCSI device state transition. In this context 'illegal' means not allowed by scsi_device_set_state(). Cc: Christoph Hellwig <hch@lst.de> Cc: Hannes Reinecke <hare@suse.com> Cc: Johannes Thumshirn <jthumshirn@suse.de> Cc: Ming Lei <ming.lei@redhat.com> Signed-off-by:
-
Tomas Winkler authored
Define new a type: uc_string_id for easier string handling and less casting. Reduce number or string copies in price of a dynamic allocation. Signed-off-by:
Tomas Winkler <tomas.winkler@intel.com> Tested-by:
Avri Altman <avri.altman@wdc.com> Signed-off-by:
-
