[ Upstream commit c41f59884b ]
We cannot wait on a completion object in the lpfc_nvme_targetport structure
in the _destroy_targetport() code path because the NVMe/fc transport will
free that structure immediately after the .targetport_delete() callback.
This results in a use-after-free, and a hang if slub_debug=FZPU is enabled.
Fix this by putting the completion on the stack.
Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Acked-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 7961cba6f7 ]
We cannot wait on a completion object in the lpfc_nvme_lport structure in
the _destroy_localport() code path because the NVMe/fc transport will free
that structure immediately after the .localport_delete() callback. This
results in a use-after-free, and a hang if slub_debug=FZPU is enabled.
Fix this by putting the completion on the stack.
Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Acked-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 30e196cace ]
After a LOGO in response to an ABTS timeout, a PLOGI wasn't issued to
re-establish the login. An nlp_type check in the LOGO completion
handler failed to restart discovery for NVME targets. Revised the
nlp_type check for NVME as well as SCSI.
While reviewing the LOGO handling a few other issues were seen and
were addressed:
- Better lock synchronization around ndlp data types
- When the ABTS times out, unregister the RPI before sending the LOGO
so that all local exchange contexts are cleared and nothing received
while awaiting LOGO/PLOGI handling will be accepted.
- LOGO handling optimized to:
Wait only R_A_TOV for a response.
It doesn't need to be retried on timeout. If there wasn't a
response, a PLOGI will be sent, thus an implicit logout
applies as well when the other port sees it.
If there is a response, any kind of response is considered "good"
and the XRI quarantined for a exchange qualifier window.
- PLOGI is issued as soon a LOGO state is resolved.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b114d9009d ]
When LCB's are rejected, if beaconing was already in progress, the
Reason Code Explanation was not being set. Should have been set to
command in progress.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 4e87eb2f46 upstream.
Certain older adapters such as the OneConnect OCe10100 may not have a valid
wqpcnt value. In this case, do not set queue->page_count to 0 in
lpfc_sli4_queue_alloc() as this will prevent the driver from initializing.
Fixes: 895427bd01 ("scsi: lpfc: NVME Initiator: Base modifications")
Cc: stable@vger.kernel.org # 4.11+
Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Reviewed-by: Laurence Oberman <loberman@redhat.com>
Tested-by: Laurence Oberman <loberman@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f8d2943245 ]
The addition of a spinlock in lpfc_debugfs_nodelist_data() introduced
a bug that lets us not skip NULL pointers correctly, as noticed by
gcc-8:
drivers/scsi/lpfc/lpfc_debugfs.c: In function 'lpfc_debugfs_nodelist_data.constprop':
drivers/scsi/lpfc/lpfc_debugfs.c:728:13: error: 'nrport' may be used uninitialized in this function [-Werror=maybe-uninitialized]
if (nrport->port_role & FC_PORT_ROLE_NVME_INITIATOR)
This changes the logic back to what it was, while keeping the added
spinlock.
Fixes: 9e21017826 ("scsi: lpfc: Synchronize access to remoteport via rport")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit ca7fb76e09 ]
On io completion, the driver is taking an adapter wide lock and nulling the
scsi command back pointer. The nulling of the back pointer is to signify the
io was completed and the scsi_done() routine was called. However, the routine
makes no check to see if the abort routine had done the same thing and
possibly nulled the pointer. Thus it may doubly-complete the io.
Make the following mods:
- Check to make sure forward progress (call scsi_done()) only happens if the
command pointer was non-null.
- As the taking of the lock, which is adapter wide, is very costly on a system
under load, null the pointer using an xchg operation rather than under lock.
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>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 0ef01a2d95 ]
When running an mds diagnostic that passes frames with the switch, soft
lockups are detected. The driver is in a CQE processing loop and has
sufficient amount of traffic that it never exits the ring processing routine,
thus the "lockup".
Cap the number of elements in the work processing routine to 64 elements. This
ensures that the cpu will be given up and the handler reschedule to process
additional items.
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>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The driver currently uses the ndlp to get the local rport which is then used
to get the nvme transport remoteport pointer. There can be cases where a stale
remoteport pointer is obtained as synchronization isn't done through the
different dereferences.
Correct by using locks to synchronize the dereferences.
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>
A recent change added some MDS processing in the lpfc_drain_txq routine
that relies on the fcp_wq being allocated. For nvmet operation the fcp_wq
is not allocated because it can only be an nvme-target. When the original
MDS support was added LS_MDS_LOOPBACK was defined wrong, (0x16) it should
have been 0x10 (decimal value used for hex setting). This incorrect value
allowed MDS_LOOPBACK to be set simultaneously with LS_NPIV_FAB_SUPPORTED,
causing the driver to crash when it accesses the non-existent fcp_wq.
Correct the bad value setting for LS_MDS_LOOPBACK.
Fixes: ae9e28f36a ("lpfc: Add MDS Diagnostic support.")
Cc: <stable@vger.kernel.org> # v4.12+
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Tested-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Enablement of the PBDE optimization brought out some incompatible behaviors
under error scenarios.
Best to disable and remove the PBDE optimization.
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>
After memory allocation for the LCB response frame, the memory wasn't zero
initialized, and not all fields are set. Thus garbage shows up in the
payload.
Fix by zeroing the memory at allocation. Also properly set the Capability
field based on duration support.
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>
Performance is affected when target queue depth is tracked. An atomic
counter is incremented on the submission path which competes with it being
decremented on the completion path. In addition, multiple CPUs can
simultaniously be manipulating this counter for the same ndlp.
Reduce the overhead by only performing the target increment/decrement when
the target queue depth is less than the overall adapter depth, thus is
actually meaningful.
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>
During remote port loss fault testing, the driver crashed with the
following trace:
general protection fault: 0000 [#1] SMP
RIP: ... lpfc_nvme_register_port+0x250/0x480 [lpfc]
Call Trace:
lpfc_nlp_state_cleanup+0x1b3/0x7a0 [lpfc]
lpfc_nlp_set_state+0xa6/0x1d0 [lpfc]
lpfc_cmpl_prli_prli_issue+0x213/0x440
lpfc_disc_state_machine+0x7e/0x1e0 [lpfc]
lpfc_cmpl_els_prli+0x18a/0x200 [lpfc]
lpfc_sli_sp_handle_rspiocb+0x3b5/0x6f0 [lpfc]
lpfc_sli_handle_slow_ring_event_s4+0x161/0x240 [lpfc]
lpfc_work_done+0x948/0x14c0 [lpfc]
lpfc_do_work+0x16f/0x180 [lpfc]
kthread+0xc9/0xe0
ret_from_fork+0x55/0x80
After registering a new remoteport, the driver is pulling an ndlp pointer
from the lpfc rport associated with the private area of a newly registered
remoteport. The private area is uninitialized, so it's garbage.
Correct by pulling the the lpfc rport pointer from the entering ndlp point,
then ndlp value from at rport. Note the entering ndlp may be replacing by
the rport->ndlp due to an address change swap.
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>
Enabling list_debug showed the drivers txcmplq was suffering list
corruption. The systems will eventually crash because the iocb free list
gets crossed linked with the prings txcmplq. Most systems will run for a
while after the corruption, but will eventually crash when a scsi eh reset
occurs and the txcmplq is attempted to be flushed. The flush gets stuck in
an endless loop.
The problem is the abort handler does not hold the sli4 ring lock while
validating the IO so the IO could complete while the driver is still
preping the abort. The erroneously generated abort, when it completes, has
pointers to the original IO that has already completed, and the IO
manipulation (for the second time) corrupts the list.
Correct by taking the ring lock early in the abort handler so the erroneous
abort won't be sent if the io has/is completing.
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>
For ABORT_XRI_CN command, firmware identifies XRI to abort by IOTAG and RPI
combination. For ELS aborts, driver specifies IOTAG correctly but RPI is
not specified.
Fix by setting RPI in WQE.
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>
The null checks on nvmebuf are redundant as nvmebuf is always obtained from
a container_of() and hence can never be null. Remove all the redundant null
checks. This also cleans up a static analysis warning.
Detected by CoverityScan, CID#1471753 ("Dereference before null check")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Change references from "Broadcom Limited" to "Broadcom Inc." in the
copyright message. Update copyright duration if not yet updated for 2018.
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>
A race condition between the context of devloss timeout handler and I/O
completion caused devloss timeout handler de-referencing pointer that had
been released.
Added the check in lpfc_sli_validate_fcp_iocb() on LPFC_IO_ON_TXCMPLQ to
capture the race condition of I/O completion and devloss timeout handler
attemption for aborting the I/O. Also, added check on lpfc_cmd->rdata
pointer before de-referenceing lpfc_cmd->rdata->pnode.
Also, added protection in lpfc_sli_abort_iocb() routine on driver performed
FCP I/O FLUSHING already under way before proceeding to aborting I/Os.
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>
Kernel occasionally crashed with the following
ops on NVME Target:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000058
IP: [<ffffffffa042ee50>] lpfc_nvmet_defer_rcv+0x50/0x70 [lpfc]
Callback routine was called for deferred rcv when it should be treated as a
normal rcv.
Added code in callback routine to detect this condition and log a message,
then bail.
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>
The PBDE optimizations aren't supported in all firmware revs.
Make optimizations configurable in case there's a side effect on old
firmware.
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>
rmmod of driver hangs
As driver instances were being unloaded, the NVME target port was unloaded
first. During the unload, the NVME initiator port sent a heartbeat
IO. Because of the target port state, that IO was scheduled for an Abort;
however, that abort subsequently failed. The failure was not cleaned up
properly and lpfc_sli4_xri_exchange_busy_wait silently hung forever.
Clean failed abort properly and make lpfc_sli4_xri_exchange_busy_wait not
hangs silently while waiting for aborts to complete.
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>
System crashes when the lpfc module is unloaded after making the port
offline
The nvme queue pointers were freed during port offline, but were later
accessed in pci remove path.
Validate the pointers in pci remove path before accessing them.
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>
Driver is incorrectly formatting a register on new hardware, using a format
for an older chip. This can result in non-deterministic behavior.
Ensure driver is not setting "workqueue index" in the WQ doorbell when
making a non-dpp doorbell write. The field must be zero when non-dpp.
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>
Kernel crashes during fill_read_buffer when nvme_info sysfs file read.
With multiple NVME targets, approx 40, nvme_info may grow larger than
PAGE_SIZE bytes. snprintf(buf + len, PAGE_SIZE - len, ...) logic is flawed
as PAGE_SIZE - len can be < 0 and is accepted by snprintf. This results in
buffer overflow, and is detected with check from dev_attr_show and
fill_read_buffer.
Change to use scnprintf to a tmp array, before calling strlcat to ensure no
buffer overflow over PAGE_SIZE bytes.
Message "6314" created as a new message indicating when there is more nvme
info, but is truncated to fit within PAGE_SIZE bytes.
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>
The get_seconds() function suffers from a possible overflow in 2038 or
2106, as well as jitter due to settimeofday or leap second updates, and is
deprecated.
As we are interested in elapsed time only, using ktime_get_seconds() to
read the CLOCK_MONOTONIC timebase is ideal here. This also lets us remove
the hack that tries to deal with get_seconds() going slightly backwards,
which cannot happen with montonic timestamps.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The driver exits port setup after failing the lpfc_sli4_get_parameters
command (messages 0356, 2541, & 1412).
The older CNA adapters do not support the MBX command. In the past
the code was allowed to fail and continue on with initialization.
However a nvme change moved a closing bracket and now makes all
failures terminal.
Revise the logic so that terminal failure only occurs if the command
failed on the newer adapters. Additionally, if parameters are set
that require information from the command and the command failed,
the parameters are erroneous and port set up should fail even on
the older adapters.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The lancer G5 chip family fails the CQ create with 16k page size. The
hardware incorrectly reports it supports large page sizes when it is
actually limited to 4k pages.
A prior patch resolved this for the A0 chip revision only. This patch
excludes all revisions of the G5 asic from using large page sizes. As
knowing the actual chip revision is unnecessary, the now unused definitions
are removed
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
modprobe -r lpfc produces the following:
Call Trace:
__blk_mq_run_hw_queue+0xa2/0xb0
__blk_mq_delay_run_hw_queue+0x9d/0xb0
? blk_mq_hctx_has_pending+0x32/0x80
blk_mq_run_hw_queue+0x50/0xd0
blk_mq_sched_insert_request+0x110/0x1b0
blk_execute_rq_nowait+0x76/0x180
nvme_keep_alive_work+0x8a/0xd0 [nvme_core]
process_one_work+0x17f/0x440
worker_thread+0x126/0x3c0
? manage_workers.isra.24+0x2a0/0x2a0
kthread+0xd1/0xe0
? insert_kthread_work+0x40/0x40
ret_from_fork_nospec_begin+0x21/0x21
? insert_kthread_work+0x40/0x40
However, rmmod lpfc would run correctly.
When an nvme remoteport is unregistered with the host nvme transport, it
needs to set the remoteport->dev_loss_tmo value 0 to indicate an immediate
termination of device loss and prevent any further keep alives to that
rport. The driver was never setting dev_loss_tmo causing the nvme
transport to continue to send the keep alive.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Under large configurations, the driver would start to log message 6065 -
NVME out of buffers (exchanges).
The driver is using the ndlp cmd_qdepth value when determining the max
outstanding ios for an adapter. This value, by default, is set to 65536,
which exceeds the maximum exchange counts supported on an adapter. The ndlp
cmd_qdepth has no relevance and outstanding io count should be capped at
the max exchange count with IO requests beyond that level getting bounced
back with an EBUSY status so that they are retried by the block layer.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
MDS diagnostics fail because of frame count mismatch.
Unavailability of SGL is the trigger for this issue. If ELS SGL is not
available to process MDS frame, IOCB is put in FCP txq but not attempted to
post afterwards. So, driver stops processing incoming frames as it runs out
of IOCB. lpfc_drain_txq attempts to submit IOCBS that are queued in ELS
txq but MDS frames are posted to FCP WQ.
Attempt to submit IOCBs that are present in FCP txq when MDS loopback is
running.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Trivial fix to spelling mistakes in lpfc_printf_log log message
"mabilbox" -> "mailbox"
"maibox" -> "mailbox"
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
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: 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>
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: 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>
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: 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>
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: 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>
