- 15 May, 2018 2 commits
-
-
Andrei Vagin authored
There are two advantages: * Direct I/O allows to avoid the write-back cache, so it reduces affects to other processes in the system. * Async I/O allows to handle a few commands concurrently. DIO + AIO shows a better perfomance for random write operations: Mode: O_DSYNC Async: 1 $ ./fio --bs=4K --direct=1 --rw=randwrite --ioengine=libaio --iodepth=64 --name=/dev/sda --runtime=20 --numjobs=2 WRITE: bw=45.9MiB/s (48.1MB/s), 21.9MiB/s-23.0MiB/s (22.0MB/s-25.2MB/s), io=919MiB (963MB), run=20002-20020msec Mode: O_DSYNC Async: 0 $ ./fio --bs=4K --direct=1 --rw=randwrite --ioengine=libaio --iodepth=64 --name=/dev/sdb --runtime=20 --numjobs=2 WRITE: bw=1607KiB/s (1645kB/s), 802KiB/s-805KiB/s (821kB/s-824kB/s), io=31.8MiB (33.4MB), run=20280-20295msec Known issue: DIF (PI) emulation doesn't work when a target uses async I/O, because DIF metadata is saved in a separate file, and it is another non-trivial task how to synchronize writing in two files, so that a following read operation always returns a consisten metadata for a specified block. Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org> Cc: Christoph Hellwig <hch@infradead.org> Cc: Bryant G. Ly <bryantly@linux.vnet.ibm.com> Tested-by:
Bryant G. Ly <bryantly@linux.vnet.ibm.com> Signed-off-by:
Andrei Vagin <avagin@openvz.org> Reviewed-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
Mike Christie <mchristi@redhat.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com>
-
Kees Cook authored
On the quest to remove all VLAs from the kernel[1] this rearranges the code to avoid a VLA warning under -Wvla (gcc doesn't recognize "const" variables as not triggering VLA creation). Additionally cleans up variable naming to avoid 80 character column limit. [1] https://lkml.kernel.org/r/CA+55aFzCG-zNmZwX4A2FQpadafLfEzK6CC=qPXydAacU1RqZWA@mail.gmail.comSigned-off-by:
Kees Cook <keescook@chromium.org> Acked-by:
Boaz Harrosh <ooo@electrozaur.com> Signed-off-by:
-
- 08 May, 2018 38 commits
-
-
Zhu Lingshan authored
use new netlink events helpers tcmu_netlink_init() and tcmu_netlink_send() to refactor netlink event attribute TCMU_ATTR_WRITECACHE(belongs to TCMU_CMD_RECONFIG_DEVICE) which is also emulate_write_cache in configFS. Removed tcmu_netlink_event() since we have new netlink events helpers now. Signed-off-by:
Zhu Lingshan <lszhu@suse.com> Acked-by:
-
Zhu Lingshan authored
use new netlink events helpers tcmu_netlink_init() and tcmu_netlink_send() to refactor netlink event attribute TCMU_ATTR_DEV_SIZE(belongs to TCMU_CMD_RECONFIG_DEVICE) which is also dev_size in configFS. Signed-off-by:
-
Zhu Lingshan authored
use new netlink events helpers tcmu_netlink_init() and tcmu_netlink_send() to refactor netlink event attribute TCMU_ATTR_DEV_CFG(belongs to TCMU_CMD_RECONFIG_DEVICE) which is also dev_config in configFS. Signed-off-by:
-
Zhu Lingshan authored
use new netlink events helpers tcmu_netlink_init() and tcmu_netlink_send() to refactor netlink event TCMU_CMD_REMOVED_DEVICE Signed-off-by:
-
Zhu Lingshan authored
use new netlink events helpers tcmu_netlink_init() and tcmu_netlink_send() to refactor netlink event TCMU_CMD_ADDED_DEVICE Signed-off-by:
-
Zhu Lingshan authored
Add new netlink events helpers tcmu_netlink_event_init() and tcmu_netlink_event_send(). These new functions intend to replace existing netlink events helper function tcmu_netlink_event(). The existing function tcmu_netlink_event() works well for events like TCMU_ADDED_DEVICE and TCMU_REMOVED_DEVICE which only has one netlink attribute. But if there is a command requires more than one attributes to send out, we have to use a struct to adapt the paremeter reconfig_data, it is hard to use one struct or a union in one struct to adapt every command with different attributes, it may get long and ugly. With the new two functions, we can call tcmu_netlink_event_init() to initialize a netlink event, then add all attributes we need by using nla_put_xxx(), at last use tcmu_netlink_event_send() to send it out. So that we don't need to use a long struct or union if we want to send mulitple attributes for different commands. [mkp: typos] Signed-off-by:
-
Wenwen Wang authored
In tw_chrdev_ioctl(), the length of the data buffer is firstly copied from the userspace pointer 'argp' and saved to the kernel object 'data_buffer_length'. Then a security check is performed on it to make sure that the length is not more than 'TW_MAX_IOCTL_SECTORS * 512'. Otherwise, an error code -EINVAL is returned. If the security check is passed, the entire ioctl command is copied again from the 'argp' pointer and saved to the kernel object 'tw_ioctl'. Then, various operations are performed on 'tw_ioctl' according to the 'cmd'. Given that the 'argp' pointer resides in userspace, a malicious userspace process can race to change the buffer length between the two copies. This way, the user can bypass the security check and inject invalid data buffer length. This can cause potential security issues in the following execution. This patch checks for capable(CAP_SYS_ADMIN) in tw_chrdev_open() to avoid the above issues. Signed-off-by:
Wenwen Wang <wang6495@umn.edu> Acked-by:
Adam Radford <aradford@gmail.com> Signed-off-by:
-
Wenwen Wang authored
In twa_chrdev_ioctl(), the ioctl driver command is firstly copied from the userspace pointer 'argp' and saved to the kernel object 'driver_command'. Then a security check is performed on the data buffer size indicated by 'driver_command', which is 'driver_command.buffer_length'. If the security check is passed, the entire ioctl command is copied again from the 'argp' pointer and saved to the kernel object 'tw_ioctl'. Then, various operations are performed on 'tw_ioctl' according to the 'cmd'. Given that the 'argp' pointer resides in userspace, a malicious userspace process can race to change the buffer size between the two copies. This way, the user can bypass the security check and inject invalid data buffer size. This can cause potential security issues in the following execution. This patch checks for capable(CAP_SYS_ADMIN) in twa_chrdev_open()t o avoid the above issues. Signed-off-by:
-
Tomohiro Kusumi authored
MPT2_MAGIC_NUMBER as well as drivers/scsi/mpt2sas/mpt2sas_ctl.h were removed to reuse mpt3sas code since commit 09ec55ed ("mpt2sas: Remove .c and .h files from mpt2sas driver"). Signed-off-by:
Tomohiro Kusumi <kusumi.tomohiro@osnexus.com> Signed-off-by:
-
Tomohiro Kusumi authored
drivers/scsi/mpt2sas/ no longer exists after commit c84b06a4 ("mpt3sas: Single driver module which supports both SAS 2.0 & SAS 3.0 HBAs") merged/removed it. Signed-off-by:
-
Dan Carpenter authored
If we had more than 32 megaraid cards then it would cause memory corruption. That's not likely, of course, but it's handy to enforce it and make the static checker happy. Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in warning message Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistakes in lpfc_printf_log log message "mabilbox" -> "mailbox" "maibox" -> "mailbox" Signed-off-by:
-
Andrei Vagin authored
Signed-off-by:
Laurence Oberman <loberman@redhat.com> Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in text string. Signed-off-by:
-
Xiaofei Tan authored
There is an SoC bug of v3 hw development version. When hot- unplugging a directly attached disk, the PHY down interrupt may not happen. It is very easy to appear on some boards. When this issue occurs, the controller will receive many invalid dword frames, and the "alos" fields of register HILINK_ERR_DFX can indicate that disk was unplugged. As an workaround solution, this patch detects this issue in the channel interrupt, and workaround it by following steps: - Disable the PHY - Clear error code and interrupt - Enable the PHY Then the HW will reissue PHY down interrupt. Signed-off-by:
Xiaofei Tan <tanxiaofei@huawei.com> Signed-off-by:
John Garry <john.garry@huawei.com> Signed-off-by:
-
John Garry authored
It is common to use readl poll timeout helpers in the driver, so create custom wrappers. Signed-off-by:
-
Xiaofei Tan authored
Event95 is used for DFX purpose. The relevant bit for this interrupt in the ENT_INT_SRC_MSK3 register has been disabled, so remove the processing. Signed-off-by:
-
Xiang Chen authored
As a unconstrained command, a command can be sent to SATA disk even if SATA disk status is BUSY, ERR or DRQ. If an ATA reset assert is successful but ATA reset de-assert fails, then it will retry the reset de-assert. If reset de- assert retry is successful, we think it is okay to probe the device but actually it still has Err status. Apparently we need to retry the ATA reset assertion and de- assertion instead for this mentioned scenario. As such, we config ATA reset assert as a constrained command, if ATA reset de-assert fails, then ATA reset de-assert retry will also fail. Then we will retry the proper process of ATA reset assert and de-assert again. Signed-off-by:
Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by:
-
Xiang Chen authored
After the controller is reset, we currently may not honour the PHY max linkrate set via sysfs, in that after a reset we always revert to max linkrate of 12Gbps, ignoring the value set via sysfs. This patch modifies to policy to set the programmed PHY linkrate, honouring the max linkrate programmed via sysfs. Signed-off-by:
-
John Garry authored
We should only have the timer enabled after PHY up after controller reset, so disable prior to reset. Signed-off-by:
-
Xiang Chen authored
It is possible to dereference a NULL-pointer in hisi_sas_abort_task() in special scenario when the device has been removed. If an SMP task times-out, it will call hisi_sas_abort_task() to recover. And currently there is a check in hisi_sas_abort_task() to avoid the situation of processing the abort for the removed device. However we have an ordering problem, in that we may reference a task for the removed device before checking if the device has been removed. Fix this by only referencing the sas_dev after we know it is still present. Signed-off-by:
-
Xiang Chen authored
There are 28 bytes of protection information record of SSP for v3 hw, 16 bytes for v2 hw, and probably 24 for v1 hw (forgotten now). So use a value big enough in hisi_sas_command_table_ssp.prot to cover all cases. Signed-off-by:
-
Xiang Chen authored
When the host is frozen in SCSI EH state, at any point after the LLDD sets SAS_TASK_STATE_DONE for the sas_task task state, libsas may free the task; see sas_scsi_find_task(). This puts the LLDD in a difficult position, in that once it sets SAS_TASK_STATE_DONE for the task state it should not reference the sas_task again. But the LLDD needs will check the sas_task indirectly in calling task->task_done()->sas_scsi_task_done() or sas_ata_task_done() (to check if the host is frozen state actually). And the LLDD cannot set SAS_TASK_STATE_DONE for the task state after task->task_done() is called (as the sas_task is free'd at this point). This situation would seem to be a problem made by libsas. To work around, check in the LLDD whether the host is in frozen state to ensure it is ok to call task->task_done() function. If in the frozen state, we rely on SCSI EH and libsas to free the sas_task directly. We do not do this for the following IO types: - SMP - they are managed in libsas directly, outside SCSI EH - Any internally originated IO, for similar reason Signed-off-by:
-
Xiang Chen authored
If the SCSI host enters EH, any pending IO will be processed by SCSI EH. However it is possible that SCSI EH will try to abort the IO and also at the same time the IO completes in the driver. In this situation there is a small chance of freeing the sas_task twice. Then if another IO re-uses freed sas_task before the second time of free'ing sas_task, it is possible to free incorrect sas_task. To avoid this situation, add some checks to increase reliability. The sas_task task state flag SAS_TASK_STATE_ABORTED is used to mutually protect the LLDD and libsas freeing the task. Signed-off-by:
-
Xiang Chen authored
In the DQ tasklet processing it is not necessary to take the DQ lock, as there is no contention between adding slots to the CQ and removing slots from the matching DQ. In addition, since we run each DQ in a separate tasklet context, there would be no possible contention between DQ processing running for the same queue in parallel. It is still necessary to take hisi_hba lock when free'ing slots. Signed-off-by:
-
James Smart authored
Fix small formatting and wording nits in Broadcom copyright header Signed-off-by:
Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by:
James Smart <james.smart@broadcom.com> Signed-off-by:
-
James Smart authored
Update the driver version to 12.0.0.3 Signed-off-by:
-
James Smart authored
Enhance log messages for CQEs as they were not reporting certain fields. Signed-off-by:
-
James Smart authored
Fix up log messages and add an fcp error stat counter in the IO submit code path to make diagnosing problems easier Signed-off-by:
-
James Smart authored
If the cpu count is larger than the number of WQ resources available, adapter attachment eventually failes due to a WQ_CREATE failure. Calculate the number of WQs desired (which initializes to cpu count) after accounting for the number of queues the adapter supports and the number allocated to SCSI and the control/ELS path, and scale down if necessary. Signed-off-by:
-
James Smart authored
The driver encounters a link event ACQE with a fault code it doesn't recognize, it logs an "Invalid" fault type and futher treats the unknown value as a mailbox command failure. First off, there is no "invalid" value, only values that are unknown. Secondly, the fault code doesn't indicate status - the rest of the ACQE contains that status so there is no reason to "fail the commands". Change the "Invalid" to "Unknown". There is no "invalid" code value. Separate fault code parsing and message genaration from any mbx handling status. Signed-off-by:
-
James Smart authored
In situations when the firmware image in inappropriate for the chip type, initial validation checks were light, allowing the checks to pass, thus allowing the firmware to be downloaded. Eventually, after the download, the chip rejects the firmware but it is logged as a generic firmware download error. Revise the initial checks to validate the image vs asic type so that the correct message is displayed and the download process is avoided. Signed-off-by:
-
James Smart authored
The driver builds the control structures in host memory using definitions that are based on 32-bit words. After building the structure it is then written to the adapter. This patch slightly optimizes LE hosts by copying the structures via 64-bit copies. This is doable as the adapter interface is LE thus there is no byteswapping as the copy is performed. The same optimization would be nice on BE systems, but when byteswapping occurs, it swaps 32-bit words as well, thus trashing the control structure. Given amount of code that is dependent upon the 32-bit word definition, it was decided to not change things for the minor optimization. Thus PPC 64-bit systems sticks with doing 32-bit copies. Signed-off-by:
-
James Smart authored
I/O submission paths in the lpfc nvme path are rejecting the io with an error code that reflects back to the callee as a hard io failure. Many of these conditions are transient and would likely resolve if retried. Correct by returning -EBUSY, which the FC transport triggers off of to return busy status codes to the blk-mq layer. Signed-off-by:
-
Chad Dupuis authored
Signed-off-by:
Chad Dupuis <chad.dupuis@cavium.com> Signed-off-by:
-
Chad Dupuis authored
Signed-off-by:
-
Chad Dupuis authored
During an uplink toggle test all error handling is done via timeout and firmware error conditions which can occur concurrently: - SCSI layer timeouts - Error detect CQEs - Firmware detected underruns - ABTS timeouts All these concurrent events require more defensive checks in the driver including: - Check both internally and externally generated aborts to make sure the xid is not already been aborted in another context or in cleanup. - Check back pointers in qedf_cmd_timeout to verify the context of the io_req, fcport and qedf_ctx - Check rport state in host reset handler to not reset the whole host if the rport is already uploaded or in the process of relogin - Check to state for an fcport before initiating a middle path ELS request Signed-off-by:
-
