Add helper which retrieves the RQ WQE's head. Use this helper in RX
reporter diagnose callback.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Use txrx.h to contain helper function regarding TX/RX. In the coming
patches, I will add more RQ helpers.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Change the hierarchy of the RX reporter 'Common config' in the diagnose
output to match the 'Common config' of the TX reporter which reflects
that CQ is a helper to the traffic queues.
Before:
$ devlink health diagnose pci/0000:00:0b.0 reporter rx
Common config:
RQ:
type: 2 stride size: 2048 size: 8
CQ:
stride size: 64 size: 1024
RQs:
...
After:
$ devlink health diagnose pci/0000:00:0b.0 reporter rx
Common config:
RQ:
type: 2 stride size: 2048 size: 8
CQ:
stride size: 64 size: 1024
RQs:
...
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When received a CQE error, the driver inspect the syndrome given by the
firmware. RQ recovery is initiated only as a result of a fatal syndrome;
syndrome which set the RQ into an error state. Hence no need to query
the RQ state at the beginning of the recovery process. Add additional
debug prints before recovering.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
During queue's recovery, driver waits for flush. The flush timeout is
set to 2 seconds. Add a define for this value for the benefit of RX and
TX reporters.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Creation of devlink health reporters is not fatal for mlx5e instance load.
In case of error in reporter's creation, the return value is ignored.
Change all reporters creation functions to return void.
In addition, with this change, a failure in creating a reporter, will not
prevent the driver from trying to create the next reporter in the list.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Heavy testing indicates the irqsave() spinlock around the __set_bit() is
insufficient to stop following clear_bit() calls being rarely applied
out-of-order. Also the nearby failed kzalloc() path leading to
SCSI_MLQUEUE_HOST_BUSY does not properly undo the in_use bitmap and
num_in_q, fix.
Link: https://lore.kernel.org/r/20200702145355.522283-1-dgilbert@interlog.com
Signed-off-by: Douglas Gilbert <dgilbert@interlog.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If we are doing async removal of the session, we could be doing a
scsi_remove_target from the removal workqueue, and for the offload case we
could be doing a new session addition and scan to the same host. The
add/scan might then end up trying to use the target_id of the target we are
removing.
This patch just has a delay the freeing of the target_id until after the
scsi_remove_target has completed, so we know it's no longer in use.
Link: https://lore.kernel.org/r/1593632868-6808-2-git-send-email-michael.christie@oracle.com
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull nfsd fixes from Bruce Fields:
"Fixes for a umask bug on exported filesystems lacking ACL support, a
leak and a module unloading bug in the /proc/fs/nfsd/clients/ code,
and a compile warning"
* tag 'nfsd-5.8-1' of git://linux-nfs.org/~bfields/linux:
SUNRPC: Add missing definition of ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
nfsd: fix nfsdfs inode reference count leak
nfsd4: fix nfsdfs reference count loop
nfsd: apply umask on fs without ACL support
The current logging methods typically end up requesting a reproduction with
a different logging level set to figure out what happened. This was mainly
by design to not clutter the kernel log messages with things that were
typically not interesting and the messages themselves could cause other
issues.
When looking to make a better system, it was seen that in many cases when
more data was wanted was when another message, usually at KERN_ERR level,
was logged. And in most cases, what the additional logging that was then
enabled was typically. Most of these areas fell into the discovery machine.
Based on this summary, the following design has been put in place: The
driver will maintain an internal log (256 elements of 256 bytes). The
"additional logging" messages that are usually enabled in a reproduction
will be changed to now log all the time to the internal log. A new logging
level is defined - LOG_TRACE_EVENT. When this level is set (it is not by
default) and a message marked as KERN_ERR is logged, all the messages in
the internal log will be dumped to the kernel log before the KERN_ERR
message is logged.
There is a timestamp on each message added to the internal log. However,
this timestamp is not converted to wall time when logged. The value of the
timestamp is solely to give a crude time reference for the messages.
Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Although the existing implementation is very good at high I/O load, on
tests involving light load, especially on only a few hardware queues,
latency was a little higher than it can be due to using workqueue
scheduling. Other tasks in the system can delay handling.
Change the lower level to use irq_poll by default which uses a softirq for
I/O completion. This gives better latency as variance in when the cq is
processed is reduced over the workqueue interface. However, as high load is
better served by not being in softirq when the CPU is loaded, work queues
are still used under high I/O load.
Link: https://lore.kernel.org/r/20200630215001.70793-13-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Change vocabulary of 0373 log msg from "error" to "cmpl" The current
language of the 0373 message contains the word "error" which caused a
number of customers to inquire about the "error" and if it should be a
concern. It isn't an error, it's simply an io completion status.
Revise the message to replace the word "error" with "cmpl" for completion.
Link: https://lore.kernel.org/r/20200630215001.70793-10-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When the kdump kernel shuts down lpfc calls flush_work_queue on an
interrupt to schedule the cq handler. When there is only one CPU active on
the kdump kernel, it is possible for the work_on to get scheduled on a
non-active CPU causing it to never be scheduled.
When in the kdump environment, per-CPU affinity of cq's to cpus is not
necessary. In those cases, use a general queue_work rather than a
queue_work_on().
Link: https://lore.kernel.org/r/20200630215001.70793-9-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When vports are deleted, it is observed that there is memory/kthread
leakage as the vport isn't fully being released.
There is a shost reference taken in scsi_add_host_dma that is not released
during scsi_remove_host. It was noticed that other drivers resolve this by
doing a scsi_host_put after calling scsi_remove_host.
The vport_delete routine is taking two references one that corresponds to
an access to the scsi_host in the vport_delete routine and another that is
released after the adapter mailbox command completes that destroys the VPI
that corresponds to the vport.
Remove one of the references taken such that the second reference that is
put will complete the missing scsi_add_host_dma reference and the shost
will be terminated.
Link: https://lore.kernel.org/r/20200630215001.70793-8-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Call traces have been observed running different tests that involve aborts
and setting the rrq active flag. The lpfc_set_rrq_active routine is doing
a mempool_alloc under the soft_irq processing level. When the mempool needs
to get a new buffer from the free pool and has to wait for memory to become
free it will check the flags passed in on the alloc and dump the stack if
the thread is running in interrupt context.
Replace the GFP_KERNEL flag with GFP_ATOMIC so that the memory allocation
will not attempt to sleep if there is no mem available.
Link: https://lore.kernel.org/r/20200630215001.70793-7-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
During driver unload/reload testing, the NVMe initiator would not
re-establish connectivity to NVMe controllers on reload.
The failing NVMe array supports concurrent FCP and NVMe operation via
different nport_id's. The array was repeatedly sending an ADISC every 2
seconds after PLOGI completed and while NVMe subsystems were executing
discovery. The target would continue this state for roughly 45 seconds.
The driver's current behavior on ADISC receipt is to validate a the ADISC
vs the device and issue a RESUME_RPI to restore transmission. The receipt
of the ADISC effectively caused a driver to take actions similar to a
logout and login for the remote port, causing the deregistration of the
nvme rport and a subsequent re-registration. This caused a constant reset
and re-connect of the NVMe controller while this 45s window occurred. There
was no need for the state changes as ADISC does not change login state.
This patch corrects this behavior by validating if the remoteport is
already logged in (MAPPED) and when true, avoids the call to set the ndlp
state to MAPPED, which triggers the unreg/re-reg. Thus ADISC does not
change the login state of the node.
Link: https://lore.kernel.org/r/20200630215001.70793-5-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Coverity reported the following error:
Assigned value that is never used may represent unnecessary computation.
The rc variable was initially assigned a value but in several cases, when
an error case is detected, it is reassigned a new value. The initial value
had little use.
In code-reviewing this routine, it could use some cleanup:
- Setting the initialization value to -ENODEV is a much better choice and
lessens code in the routine.
- The wasn't tracking logic errors vs no error and mailbox failure.
Better to resolve by adding a status to track the mailbox failure
and merge it with the logic error when the routine returns.
Link: https://lore.kernel.org/r/20200630215001.70793-2-jsmart2021@gmail.com
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull block fixes from Jens Axboe:
- Use kvfree_sensitive() for the block keyslot free (Eric)
- Sync blk-mq debugfs flags (Hou)
- Memory leak fix in virtio-blk error path (Hou)
* tag 'block-5.8-2020-07-01' of git://git.kernel.dk/linux-block:
virtio-blk: free vblk-vqs in error path of virtblk_probe()
block/keyslot-manager: use kvfree_sensitive()
blk-mq-debugfs: update blk_queue_flag_name[] accordingly for new flags
Pull io_uring fixes from Jens Axboe:
"One fix in here, for a regression in 5.7 where a task is waiting in
the kernel for a condition, but that condition won't become true until
task_work is run. And the task_work can't be run exactly because the
task is waiting in the kernel, so we'll never make any progress.
One example of that is registering an eventfd and queueing io_uring
work, and then the task goes and waits in eventfd read with the
expectation that it'll get woken (and read an event) when the io_uring
request completes. The io_uring request is finished through task_work,
which won't get run while the task is looping in eventfd read"
* tag 'io_uring-5.8-2020-07-01' of git://git.kernel.dk/linux-block:
io_uring: use signal based task_work running
task_work: teach task_work_add() to do signal_wake_up()
Edward Cree says:
====================
sfc: prerequisites for EF100 driver, part 3
Continuing on from [1] and [2], this series assembles the last pieces
of the common codebase that will be used by the forthcoming EF100
driver.
Patch #1 also adds a minor feature to EF10 (setting MTU on VFs) since
EF10 supports the same MCDI extension which that feature will use on
EF100.
Patches #5 & #7, while they should have no externally-visible effect
on driver functionality, change how that functionality is implemented
and how the driver represents TXQ configuration internally, so are
not mere cleanup/refactoring like most of these prerequisites have
(from the perspective of the existing sfc driver) been.
Changes in v2:
* Patch #1: use efx_mcdi_set_mtu() directly, instead of as a fallback,
in the mtu_only case (Jakub)
* Patch #3: fix symbol collision in non-modular builds by renaming
interrupt_mode to efx_interrupt_mode (kernel test robot)
* Patch #6: check for failure of netif_set_real_num_[tr]x_queues (Jakub)
* Patch #12: cleaner solution for ethtool drvinfo (Jakub, David)
[1]: https://lore.kernel.org/netdev/20200629.173812.1532344417590172093.davem@davemloft.net/T/
[2]: https://lore.kernel.org/netdev/20200630.130923.402514193016248355.davem@davemloft.net/T/
====================
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We have yet another new scheme for NVRAM, and a corresponding new MCDI.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since ethtool_common.o will be built into both sfc and sfc_ef100 drivers,
it can't use KBUILD_MODNAME directly. Instead, make it reference a
string provided by the individual driver code.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously this was only happening in ef10-specific code.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
_down() merely removes all our filters and VLANs, it doesn't free
efx->filter_state itself.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since we only allocate VIs for the number of TXQs we actually need, we
cannot naively use "channel * TXQ_TYPES + txq" for the TXQ number, as
this has gaps (when efx->tx_queues_per_channel < EFX_TXQ_TYPES) and
thus overruns the driver's VI allocations, causing the firmware to
reject the MC_CMD_INIT_TXQ based on INSTANCE.
Thus, we distinguish INSTANCE (stored in tx_queue->queue) from LABEL
(tx_queue->label); the former is allocated starting from 0 in
efx_set_channels(), while the latter is simply the txq type (index in
channel->tx_queue array).
To simplify things, rather than changing tx_queues_per_channel after
setting up TXQs, make Siena always probe its HIGHPRI queues at start
of day, rather than deferring it until tc mqprio enables them.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
