Christoph writes:
"A couple more updates for 4.12. The biggest pile is fc and lpfc
updates from James, but there are various small fixes and cleanups as
well."
Fixes up a few merge issues, and also a warning in
lpfc_nvmet_rcv_unsol_abort() if CONFIG_NVME_TARGET_FC isn't enabled.
Signed-off-by: Jens Axboe <axboe@fb.com>
lpfc was changing the private pointer that is set/maintained by
the nvme_fc transport. This caused two issues: a) the transport, on
teardown may erroneous attempt to free whatever address was set;
and b) lfpc uses any value set in lpfc_nvme_fcp_abort() and
assumes its a valid io request.
Correct issue by properly defining a context structure for lpfc.
Lpfc also updated to clear the private context structure on io
completion.
Since this bug caused scrutiny of the way lpfc moves local request
structures between lists, also cleaned up list_del()'s to
list_del_inits()'s.
This is a nvme-specific bug. The patch was cut against the
linux-block tree, for-4.12/block tree. It should be pulled in through
that tree.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Our final fix before the 4.12 release (hopefully). It's an error leg
again: the fix to not bug on empty DMA transfers is returning the
wrong code and confusing the block layer.
Signed-off-by: James E.J. Bottomley <jejb@linux.vnet.ibm.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=tuno
-----END PGP SIGNATURE-----
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fix from James Bottomley:
"Our final fix before the 4.12 release (hopefully).
It's an error leg again: the fix to not bug on empty DMA transfers is
returning the wrong code and confusing the block layer"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: return correct blkprep status code in case scsi_init_io() fails.
Update lpfc version to reflect this set of changes.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
The older sli4 adapters only supported the 64 byte WQE entry size.
The new adapter (fw) support both 64 and 128 byte WQE entry sizies.
The Express lane WQ was not being created with the 128 byte WQE sizes
when it was supported.
Not having the right WQE size created for the express lane work queue
caused the the firmware to overwrite the lun indentifier in the FCP header.
This patch correctly creates the express lane work queue with the
supported size.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
The driver with nvme had this routine stubbed.
Right now XRI_ABORTED_CQE is not handled and the FC NVMET
Transport has a new API for the driver.
Missing code path, new NVME abort API
Update ABORT processing for NVMET
There are 3 new FC NVMET Transport API/ template routines for NVMET:
lpfc_nvmet_xmt_fcp_release
This NVMET template callback routine called to release context
associated with an IO This routine is ALWAYS called last, even
if the IO was aborted or completed in error.
lpfc_nvmet_xmt_fcp_abort
This NVMET template callback routine called to abort an exchange that
has an IO in progress
nvmet_fc_rcv_fcp_req
When the lpfc driver receives an ABTS, this NVME FC transport layer
callback routine is called. For this case there are 2 paths thru the
driver: the driver either has an outstanding exchange / context for the
XRI to be aborted or not. If not, a BA_RJT is issued otherwise a BA_ACC
NVMET Driver abort paths:
There are 2 paths for aborting an IO. The first one is we receive an IO and
decide not to process it because of lack of resources. An unsolicated ABTS
is immediately sent back to the initiator as a response.
lpfc_nvmet_unsol_fcp_buffer
lpfc_nvmet_unsol_issue_abort (XMIT_SEQUENCE_WQE)
The second one is we sent the IO up to the NVMET transport layer to
process, and for some reason the NVME Transport layer decided to abort the
IO before it completes all its phases. For this case there are 2 paths
thru the driver:
the driver either has an outstanding TSEND/TRECEIVE/TRSP WQE or no
outstanding WQEs are present for the exchange / context.
lpfc_nvmet_xmt_fcp_abort
if (LPFC_NVMET_IO_INP)
lpfc_nvmet_sol_fcp_issue_abort (ABORT_WQE)
lpfc_nvmet_sol_fcp_abort_cmp
else
lpfc_nvmet_unsol_fcp_issue_abort
lpfc_nvmet_unsol_issue_abort (XMIT_SEQUENCE_WQE)
lpfc_nvmet_unsol_fcp_abort_cmp
Context flags:
LPFC_NVMET_IOP - his flag signifies an IO is in progress on the exchange.
LPFC_NVMET_XBUSY - this flag indicates the IO completed but the firmware
is still busy with the corresponding exchange. The exchange should not be
reused until after a XRI_ABORTED_CQE is received for that exchange.
LPFC_NVMET_ABORT_OP - this flag signifies an ABORT_WQE was issued on the
exchange.
LPFC_NVMET_CTX_RLS - this flag signifies a context free was requested,
but we are deferring it due to an XBUSY or ABORT in progress.
A ctxlock is added to the context structure that is used whenever these
flags are set/read within the context of an IO.
The LPFC_NVMET_CTX_RLS flag is only set in the defer_relase routine when
the transport has resolved all IO associated with the buffer. The flag is
cleared when the CTX is associated with a new IO.
An exchange can has both an LPFC_NVMET_XBUSY and a LPFC_NVMET_ABORT_OP
condition active simultaneously. Both conditions must complete before the
exchange is freed.
When the abort callback (lpfc_nvmet_xmt_fcp_abort) is envoked:
If there is an outstanding IO, the driver will issue an ABORT_WQE. This
should result in 3 completions for the exchange:
1) IO cmpl with XB bit set
2) Abort WQE cmpl
3) XRI_ABORTED_CQE cmpl
For this scenerio, after completion #1, the NVMET Transport IO rsp
callback is called. After completion #2, no action is taken with respect
to the exchange / context. After completion #3, the exchange context is
free for re-use on another IO.
If there is no outstanding activity on the exchange, the driver will send a
ABTS to the Initiator. Upon completion of this WQE, the exchange / context
is freed for re-use on another IO.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
NVMET didn't have any RSCN handling at all and
would not execute implicit LOGO when receiving a PLOGI
from an rport that NVMET had in state UNMAPPED.
Clean up the logic in lpfc_nlp_state_cleanup for
initiators (FCP and NVME). NVMET should not respond to
RSCN including allocating new ndlps so this code was
conditionalized when nvmet_support is true. The check
for NLP_RCV_PLOGI in lpfc_setup_disc_node was moved
below the check for nvmet_support to allow the NVMET
to recover initiator nodes correctly. The implicit
logo was introduced with lpfc_rcv_plogi when NVMET gets
a PLOGI on an ndlp in UNMAPPED state. The RSCN handling
was modified to not respond to an RSCN in NVMET. Instead
NVMET sends a GID_FT and determines if an NVMEP_INITIATOR
it has is UNMAPPED but no longer in the zone membership.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Adding support for Fabric assigned WWPN and WWNN.
Firmware sends first FLOGI to fabric with vendor version changes.
On link up driver gets updated service parameter with FAWWN assigned port
name. Driver sends 2nd FLOGI with updated fawwpn and modifies the
vport->fc_portname in driver.
Note:
Soft wwpn will not be allowed when fawwpn is enabled.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Cannot set NVME segment counts to a large number
The existing module parameter lpfc_sg_seg_cnt is used for both
SCSI and NVME.
Limit the module parameter lpfc_sg_seg_cnt to 128 with the
default being 64 for both NVME and NVMET, assuming NVME is enabled in the
driver for that port. The driver will set max_sgl_segments in the
NVME/NVMET template to lpfc_sg_seg_cnt + 1.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
When RPI is not available, driver sends WQE with invalid RPI value and
rejected by HBA.
lpfc 0000:82:00.3: 1:3154 BLS ABORT RSP failed, data: x3/xa0320008
and
lpfc :2753 PLOGI failure DID:FFFFFA Status:x3/xa0240008
In this case, driver accesses rpi_ids array out of bounds.
Fix:
Check return value of lpfc_sli4_alloc_rpi(). Do not allocate
lpfc_nodelist entry if RPI is not available.
When RPI is not available, we will get discovery timeouts and
command drops for some of the vports as seen below.
lpfc :0273 Unexpected discovery timeout, vport State x0
lpfc :0230 Unexpected timeout, hba link state x5
lpfc :0111 Dropping received ELS cmd Data: x0 xc90c55 x0
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
The symptom is that the driver will fail to login to the fabric.
The reason is because it is out of iocb resources.
There is a one to one relationship between MRQs
(receive buffers for NVMET-FC) and iocbs and the default number of
IOCBs was not accounting for the number of MRQs that were being created.
This fix aligns the number of MRQ resources with the total resources so
that it can handle fabric events when needed.
Also the initialization of ctxlock to be on FCP commands, NOT LS commands.
And modified log messages so that the log output can be correlated with
the analyzer trace.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Unnecessary lock is taken. ring lock should be sufficient to protect the
work queue submission.
This was noticed when doing performance testing. The hbalock is not
needed to issue io to the nvme work queue.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Fix nvme initiator handline when CONFIG_LPFC_NVME_INITIATOR is not enabled.
With update nvme upstream driver sources, loading
the driver with nvme enabled resulting in this Oops.
BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
IP: lpfc_nvme_update_localport+0x23/0xd0 [lpfc]
PGD 0
Oops: 0000 [#1] SMP
CPU: 0 PID: 10256 Comm: lpfc_worker_0 Tainted
Hardware name: ...
task: ffff881028191c40 task.stack: ffff880ffdf00000
RIP: 0010:lpfc_nvme_update_localport+0x23/0xd0 [lpfc]
RSP: 0018:ffff880ffdf03c20 EFLAGS: 00010202
Cause: As the initiator driver completes discovery at different stages,
it call lpfc_nvme_update_localport to hint that the DID and role may have
changed. In the implementation of lpfc_nvme_update_localport, the driver
was not validating the localport or the lport during the execution
of the update_localport routine. With the recent upstream additions to
the driver, the create_localport routine didn't run and so the localport
was NULL causing the page-fault Oops.
Fix: Add the CONFIG_LPFC_NVME_INITIATOR preprocessor inclusions to
lpfc_nvme_update_localport to turn off all routine processing when
the running kernel does not have NVME configured. Add NULL pointer
checks on the localport and lport in lpfc_nvme_update_localport and
dump messages if they are NULL and just exit.
Also one alingment issue fixed.
Repalces the ifdef with the IS_ENABLED macro.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
There are two versions of a structure for queue creation and setup that the
driver shares with FW. The driver was only treating as version 0.
Verify WQ_CREATE with 128B WQEs in V0 and V1.
Code review of another bug showed the driver passing
128B WQEs and 8 pages in WQ CREATE and V0.
Code inspection/instrumentation showed that the driver
uses V0 in WQ_CREATE and if the caller passes queue->entry_size
128B, the driver sets the hdr_version to V1 so all is good.
When I tested the V1 WQ_CREATE, the mailbox failed causing
the driver to unload.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
There are couple of different load/unload issues fixed with this patch.
One of the issues was reported by Junichi Nomura, a patch was submitted
by Johannes Thumsrhirn which did fix one of the problems but the fix in
this patch separates the pring free from the queue free and does not set
the parameter passed in to NULL.
issues:
(1) driver could not be unloaded and reloaded without some Oops or
Panic occurring.
(2) The driver was panicking because of a corruption in the Memory
Manager when the iocb list was getting allocated.
Root cause for the memory corruption was a double free of the Work Queue
ring pointer memory - Freed once in the lpfc_sli4_queue_free when the CQ
was destroyed and again in lpfc_sli4_queue_free when the WQ was destroyed.
The pring free and the queue free were separated, the pring free was moved
to the wq destroy routine because it a better fit logically to delete the
ring with the wq.
The checkpatch flagged several alignmenet issues that were also corrected
with this patch.
The mboxq was never initialed correctly before it was used by the driver
this patch corrects that issue.
Reported-by: Junichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Tested-by: Junichi Nomura <j-nomura@ce.jp.nec.com>
An extra blank line was being added the the rqpair printing.
Remove the extra line feed.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
The check for NULL ptr is not necessary, kfree will check it.
Removing NULL ptr check.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
These defines for the posting of buffers for nvmet target were not used.
Removing the unused defines.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Comment should have said Repost.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
The xri resources are split into pools for NVME and FCP IO when NVME is
enabled. There was not message in the log that identified this allocation.
Added debug message to log XRI split.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
In the lpfc_nvme_io_cmd_wqe_cmpl routine the driver was printing two
pointers and the DID for the rport whenever an IO completed on a now
that had transitioned to a non active state.
There is no need to print the node pointer address for a node that
is not active the DID should be enough to debug.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
In this case, the NVME initiator is sending an LS REQ command on an NDLP
that is not MAPPED. The FW rejects it.
The lpfc_nvme_ls_req routine checks for a NULL ndlp pointer
but does not check the NDLP state. This allows the routine
to send an LS IO when the ndlp is disconnected.
Check the ndlp for NULL, actual node, Target and MAPPED
or Initiator and UNMAPPED. This avoids Fabric nodes getting
the Create Association or Create Connection commands. Initiators
are free to Reject either Create.
Also some of the messages numbers in lpfc_nvme_ls_req were changed because
they were already used in other log messages.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
During some link event testing it was observed that the
wait_for_completion_timeout in the lpfc_nvme_unregister_port
was timing out all the time.
The initiator is claiming the nvme_fc_unregister_remoteport upcall is
not completing the unregister in the time allotted.
[ 2186.151317] lpfc 0000:07:00.0: 0:(0):6169 Unreg nvme wait failed 0
The wait_for_completion_timeout returns 0 when the wait has
been outstanding for the jiffies passed by the caller. In this error
message, the nvme initiator passed value 5 - meaning 5 jiffies -
and this is just wrong.
Calculate 5 seconds in Jiffies and pass that value
from the current jiffies.
Also the log message for the unregister timeout was reduced
because timeout failure is the same as timeout.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Standardize default SGL segment count for nvme target and initiator
The driver needs to make them the same for clarity.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
target transport:
----------------------
There are cases when there is a need to abort in-progress target
operations (writedata) so that controller termination or errors can
clean up. That can't happen currently as the abort is another target
op type, so it can't be used till the running one finishes (and it may
not). Solve by removing the abort op type and creating a separate
downcall from the transport to the lldd to request an io to be aborted.
The transport will abort ios on queue teardown or io errors. In general
the transport tries to call the lldd abort only when the io state is
idle. Meaning: ops that transmit data (readdata or rsp) will always
finish their transmit (or the lldd will see a state on the
link or initiator port that fails the transmit) and the done call for
the operation will occur. The transport will wait for the op done
upcall before calling the abort function, and as the io is idle, the
io can be cleaned up immediately after the abort call; Similarly, ios
that are not waiting for data or transmitting data must be in the nvmet
layer being processed. The transport will wait for the nvmet layer
completion before calling the abort function, and as the io is idle,
the io can be cleaned up immediately after the abort call; As for ops
that are waiting for data (writedata), they may be outstanding
indefinitely if the lldd doesn't see a condition where the initiatior
port or link is bad. In those cases, the transport will call the abort
function and wait for the lldd's op done upcall for the operation, where
it will then clean up the io.
Additionally, if a lldd receives an ABTS and matches it to an outstanding
request in the transport, A new new transport upcall was created to abort
the outstanding request in the transport. The transport expects any
outstanding op call (readdata or writedata) will completed by the lldd and
the operation upcall made. The transport doesn't act on the reported
abort (e.g. clean up the io) until an op done upcall occurs, a new op is
attempted, or the nvmet layer completes the io processing.
fcloop:
----------------------
Updated to support the new target apis.
On fcp io aborts from the initiator, the loopback context is updated to
NULL out the half that has completed. The initiator side is immediately
called after the abort request with an io completion (abort status).
On fcp io aborts from the target, the io is stopped and the initiator side
sees it as an aborted io. Target side ops, perhaps in progress while the
initiator side is done, continue but noop the data movement as there's no
structure on the initiator side to reference.
patch also contains:
----------------------
Revised lpfc to support the new abort api
commonized rsp buffer syncing and nulling of private data based on
calling paths.
errors in op done calls don't take action on the fod. They're bad
operations which implies the fod may be bad.
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
With the advent of the opdone calls changing context, the lldd can no
longer assume that once the op->done call returns for RSP operations
that the request struct is no longer being accessed.
As such, revise the lldd api for a req_release callback that the
transport will call when the job is complete. This will also be used
with abort cases.
Fixed text in api header for change in io complete semantics.
Revised lpfc to support the new req_release api.
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Two new feature flags were added to control whether upcalls to the
transport result in context switches or stay in the calling context.
NVMET_FCTGTFEAT_CMD_IN_ISR:
By default, if the flag is not set, the transport assumes the
lldd is in a non-isr context and in the cpu context it should be
for the io queue. As such, the cmd handler is called directly in the
calling context.
If the flag is set, indicating the upcall is an isr context, the
transport mandates a transition to a workqueue. The workqueue assigned
to the queue is used for the context.
NVMET_FCTGTFEAT_OPDONE_IN_ISR
By default, if the flag is not set, the transport assumes the
lldd is in a non-isr context and in the cpu context it should be
for the io queue. As such, the fcp operation done callback is called
directly in the calling context.
If the flag is set, indicating the upcall is an isr context, the
transport mandates a transition to a workqueue. The workqueue assigned
to the queue is used for the context.
Updated lpfc for flags
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Now that all drivers that call blk_mq_complete_requests have a
->complete callback we can remove the direct call to blk_mq_end_request,
as well as the error argument to blk_mq_complete_request.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This passes on the scsi_cmnd result field to users of passthrough
requests. Currently we abuse req->errors for this purpose, but that
field will go away in its current form.
Note that the old IDE code abuses the errors field in very creative
ways and stores all kinds of different values in it. I didn't dare
to touch this magic, so the abuses are brought forward 1:1.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
The function only returns -EIO if rq->errors is non-zero, which is not
very useful and lets a large number of callers ignore the return value.
Just let the callers figure out their error themselves.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This is seven small fixes which are all for user visible issues that
fortunately only occur in rare circumstances. The most serious is the
sr one in which QEMU can cause us to read beyond the end of a buffer
(I don't think it's exploitable, but just in case). The next is the
sd capacity fix which means all non 512 byte sector drives greater
than 2TB fail to be correctly sized. The rest are either in new
drivers (qedf) or on error legs.
Signed-off-by: James E.J. Bottomley <jejb@linux.vnet.ibm.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=vOpC
-----END PGP SIGNATURE-----
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes from James Bottomley:
"This is seven small fixes which are all for user visible issues that
fortunately only occur in rare circumstances.
The most serious is the sr one in which QEMU can cause us to read
beyond the end of a buffer (I don't think it's exploitable, but just
in case).
The next is the sd capacity fix which means all non 512 byte sector
drives greater than 2TB fail to be correctly sized.
The rest are either in new drivers (qedf) or on error legs"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: ipr: do not set DID_PASSTHROUGH on CHECK CONDITION
scsi: aacraid: fix PCI error recovery path
scsi: sd: Fix capacity calculation with 32-bit sector_t
scsi: qla2xxx: Add fix to read correct register value for ISP82xx.
scsi: qedf: Fix crash due to unsolicited FIP VLAN response.
scsi: sr: Sanity check returned mode data
scsi: sd: Consider max_xfer_blocks if opt_xfer_blocks is unusable
When instrumenting the SCSI layer to run into the
!blk_rq_nr_phys_segments(rq) case the following warning emitted from the
block layer:
blk_peek_request: bad return=-22
This happens because since commit fd3fc0b4d7 ("scsi: don't BUG_ON()
empty DMA transfers") we return the wrong error value from
scsi_prep_fn() back to the block layer.
[mkp: silenced checkpatch]
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Fixes: fd3fc0b4d7 scsi: don't BUG_ON() empty DMA transfers
Cc: <stable@vger.kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
On a dual controller setup with multipath enabled, some MEDIUM ERRORs
caused both paths to be failed, thus I/O got queued/blocked since the
'queue_if_no_path' feature is enabled by default on IPR controllers.
This example disabled 'queue_if_no_path' so the I/O failure is seen at
the sg_dd program. Notice that after the sg_dd test-case, both paths
are in 'failed' state, and both path/priority groups are in 'enabled'
state (not 'active') -- which would block I/O with 'queue_if_no_path'.
# sg_dd if=/dev/dm-2 bs=4096 count=1 dio=1 verbose=4 blk_sgio=0
<...>
read(unix): count=4096, res=-1
sg_dd: reading, skip=0 : Input/output error
<...>
# dmesg
[...] sd 2:2:16:0: [sds] FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
[...] sd 2:2:16:0: [sds] Sense Key : Medium Error [current]
[...] sd 2:2:16:0: [sds] Add. Sense: Unrecovered read error - recommend rewrite the data
[...] sd 2:2:16:0: [sds] CDB: Read(10) 28 00 00 00 00 00 00 00 20 00
[...] blk_update_request: I/O error, dev sds, sector 0
[...] device-mapper: multipath: Failing path 65:32.
<...>
[...] device-mapper: multipath: Failing path 65:224.
# multipath -l
1IBM_IPR-0_59C2AE0000001F80 dm-2 IBM ,IPR-0 59C2AE00
size=5.2T features='0' hwhandler='1 alua' wp=rw
|-+- policy='service-time 0' prio=0 status=enabled
| `- 2:2:16:0 sds 65:32 failed undef running
`-+- policy='service-time 0' prio=0 status=enabled
`- 1:2:7:0 sdae 65:224 failed undef running
This is not the desired behavior. The dm-multipath explicitly checks
for the MEDIUM ERROR case (and a few others) so not to fail the path
(e.g., I/O to other sectors could potentially happen without problems).
See dm-mpath.c :: do_end_io_bio() -> noretry_error() !->! fail_path().
The problem trace is:
1) ipr_scsi_done() // SENSE KEY/CHECK CONDITION detected, go to..
2) ipr_erp_start() // ipr_is_gscsi() and masked_ioasc OK, go to..
3) ipr_gen_sense() // masked_ioasc is IPR_IOASC_MED_DO_NOT_REALLOC,
// so set DID_PASSTHROUGH.
4) scsi_decide_disposition() // check for DID_PASSTHROUGH and return
// early on, faking a DID_OK.. *instead*
// of reaching scsi_check_sense().
// Had it reached the latter, that would
// set host_byte to DID_MEDIUM_ERROR.
5) scsi_finish_command()
6) scsi_io_completion()
7) __scsi_error_from_host_byte() // That would be converted to -ENODATA
<...>
8) dm_softirq_done()
9) multipath_end_io()
10) do_end_io()
11) noretry_error() // And that is checked in dm-mpath :: noretry_error()
// which would cause fail_path() not to be called.
With this patch applied, the I/O is failed but the paths are not. This
multipath device continues accepting more I/O requests without blocking.
(and notice the different host byte/driver byte handling per SCSI layer).
# dmesg
[...] sd 2:2:7:0: [sdaf] Done: SUCCESS Result: hostbyte=0x13 driverbyte=DRIVER_OK
[...] sd 2:2:7:0: [sdaf] CDB: Read(10) 28 00 00 00 00 00 00 00 40 00
[...] sd 2:2:7:0: [sdaf] Sense Key : Medium Error [current]
[...] sd 2:2:7:0: [sdaf] Add. Sense: Unrecovered read error - recommend rewrite the data
[...] blk_update_request: critical medium error, dev sdaf, sector 0
[...] blk_update_request: critical medium error, dev dm-6, sector 0
[...] sd 2:2:7:0: [sdaf] Done: SUCCESS Result: hostbyte=0x13 driverbyte=DRIVER_OK
[...] sd 2:2:7:0: [sdaf] CDB: Read(10) 28 00 00 00 00 00 00 00 10 00
[...] sd 2:2:7:0: [sdaf] Sense Key : Medium Error [current]
[...] sd 2:2:7:0: [sdaf] Add. Sense: Unrecovered read error - recommend rewrite the data
[...] blk_update_request: critical medium error, dev sdaf, sector 0
[...] blk_update_request: critical medium error, dev dm-6, sector 0
[...] Buffer I/O error on dev dm-6, logical block 0, async page read
# multipath -l 1IBM_IPR-0_59C2AE0000001F80
1IBM_IPR-0_59C2AE0000001F80 dm-6 IBM ,IPR-0 59C2AE00
size=5.2T features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
|-+- policy='service-time 0' prio=0 status=active
| `- 2:2:7:0 sdaf 65:240 active undef running
`-+- policy='service-time 0' prio=0 status=enabled
`- 1:2:7:0 sdh 8:112 active undef running
Signed-off-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
During a PCI error recovery, if aac_check_health() is not aware that a
PCI error happened and we have an offline PCI channel, it might trigger
some errors (like NULL pointer dereference) and inhibit the error
recovery process to complete.
This patch makes the health check procedure aware of PCI channel issues,
and in case of error recovery process, the function
aac_adapter_check_health() returns -1 and let the recovery process to
complete successfully. This patch was tested on upstream kernel
v4.11-rc5 in PowerPC ppc64le architecture with adapter 9005:028d
(VID:DID) - the error recovery procedure was able to recover fine.
Fixes: 5c63f7f710 ("aacraid: Added EEH support")
Cc: stable@vger.kernel.org # v4.6+
Signed-off-by: Guilherme G. Piccoli <gpiccoli@linux.vnet.ibm.com>
Reviewed-by: Dave Carroll <david.carroll@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull block fixes from Jens Axboe:
"Here's a pull request for 4.11-rc, fixing a set of issues mostly
centered around the new scheduling framework. These have been brewing
for a while, but split up into what we absolutely need in 4.11, and
what we can defer until 4.12. These are well tested, on both single
queue and multiqueue setups, and with and without shared tags. They
fix several hangs that have happened in testing.
This is obviously larger than I would have preferred at this point in
time, but I don't think we can shave much off this and still get the
desired results.
In detail, this pull request contains:
- a set of five fixes for NVMe, mostly from Christoph and one from
Roland.
- a series from Bart, fixing issues with dm-mq and SCSI shared tags
and scheduling. Note that one of those patches commit messages may
read like an optimization, but it is in fact an important fix for
queue restarts in particular.
- a series from Omar, most importantly fixing a hang with multiple
hardware queues when we fail to get a driver tag. Another important
fix in there is for resizing hardware queues, which nbd does when
handling multiple sockets for one connection.
- fixing an imbalance in putting the ctx for hctx request allocations
from Minchan"
* 'for-linus' of git://git.kernel.dk/linux-block:
blk-mq: Restart a single queue if tag sets are shared
dm rq: Avoid that request processing stalls sporadically
scsi: Avoid that SCSI queues get stuck
blk-mq: Introduce blk_mq_delay_run_hw_queue()
blk-mq: remap queues when adding/removing hardware queues
blk-mq-sched: fix crash in switch error path
blk-mq-sched: set up scheduler tags when bringing up new queues
blk-mq-sched: refactor scheduler initialization
blk-mq: use the right hctx when getting a driver tag fails
nvmet: fix byte swap in nvmet_parse_io_cmd
nvmet: fix byte swap in nvmet_execute_write_zeroes
nvmet: add missing byte swap in nvmet_get_smart_log
nvme: add missing byte swap in nvme_setup_discard
nvme: Correct NVMF enum values to match NVMe-oF rev 1.0
block: do not put mq context in blk_mq_alloc_request_hctx
Separating discards and zeroout operations allows us to remove the LBPRZ
block zeroing constraints from discards and honor the device preferences
for UNMAP commands.
If supported by the device, we'll also choose UNMAP over one of the
WRITE SAME variants for discards.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Now that zeroout and discards are distinct operations we need to
separate the policy of choosing the appropriate command. Create a
zeroing_mode which can be one of:
write: Zeroout assist not present, use regular WRITE
writesame: Allow WRITE SAME(10/16) with a zeroed payload
writesame_16_unmap: Allow WRITE SAME(16) with UNMAP
writesame_10_unmap: Allow WRITE SAME(10) with UNMAP
The last two are conditional on the device being thin provisioned with
LBPRZ=1 and LBPWS=1 or LBPWS10=1 respectively.
Whether to set the UNMAP bit or not depends on the REQ_NOUNMAP flag. And
if none of the _unmap variants are supported, regular WRITE SAME will be
used if the device supports it.
The zeroout_mode is exported in sysfs and the detected mode for a given
device can be overridden using the string constants above.
With this change in place we can now issue WRITE SAME(16) with UNMAP set
for block zeroing applications that require hard guarantees and
logical_block_size granularity. And at the same time use the UNMAP
command with the device's preferred granulary and alignment for discard
operations.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Now that we use the proper REQ_OP_WRITE_ZEROES operation everywhere we can
kill this hack.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Try to use a write same with unmap bit variant if the device supports it
and the caller allows for it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Split sd_setup_discard_cmnd into one function per provisioning type. While
this creates some very slight duplication of boilerplate code it keeps the
code modular for additions of new provisioning types, and for reusing the
write same functions for the upcoming scsi implementation of the Write Zeroes
operation.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
We previously made sure that the reported disk capacity was less than
0xffffffff blocks when the kernel was not compiled with large sector_t
support (CONFIG_LBDAF). However, this check assumed that the capacity
was reported in units of 512 bytes.
Add a sanity check function to ensure that we only enable disks if the
entire reported capacity can be expressed in terms of sector_t.
Cc: <stable@vger.kernel.org>
Reported-by: Steve Magnani <steve.magnani@digidescorp.com>
Cc: Bart Van Assche <Bart.VanAssche@sandisk.com>
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add fix to read correct register value for ISP82xx, during check for
register disconnect.ISP82xx has different base register.
Fixes: a465537ad1 ("qla2xxx: Disable the adapter and skip error recovery in case of register disconnect")
Signed-off-by: Sawan Chandak <sawan.chandak@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Kefeng Wang discovered that old versions of the QEMU CD driver would
return mangled mode data causing us to walk off the end of the buffer in
an attempt to parse it. Sanity check the returned mode sense data.
Cc: <stable@vger.kernel.org>
Reported-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Tested-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If device reports a small max_xfer_blocks and a zero opt_xfer_blocks, we
end up using BLK_DEF_MAX_SECTORS, which is wrong and r/w of that size
may get error.
[mkp: tweaked to avoid setting rw_max twice and added typecast]
Cc: <stable@vger.kernel.org> # v4.4+
Fixes: ca369d51b3 ("block/sd: Fix device-imposed transfer length limits")
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
