I'm puzzled.
$ git bisect start fc_bsg master
Bisecting: 8 revisions left to test after this (roughly 3 steps)
[005d51510eee6102636d5dbb06310531c5d46151] scsi: fc: implement kref
backed reference counting
$ git bisect bad
Bisecting: 3 revisions left to test after this (roughly 2 steps)
This patch adds code to disconnect from the client, which will make sure
any outstanding commands have been completed, before continuing on with
the remove operation.
Signed-off-by: Michael Cyr
Signed-off-by: Bryant G. Ly
---
Signed-off-by: Michael Cyr
Signed-off-by: Bryant G. Ly
---
drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c | 13 +++--
1 file changed, 3 insertions(+), 10 deletions(-)
diff --git a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
Signed-off-by: Michael Cyr
Signed-off-by: Bryant G. Ly
---
drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c | 7 +++
1 file changed, 7 insertions(+)
diff --git a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
b/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
Don't set FC_RQST_STATE_DONE before calling fc_bsg_jobdone() as
fc_bsg_jobdone() calls blk_complete_requeust() which raises a soft-IRQ that
ends up in fc_bsg_sofirq_done() and fc_bsg_softirq_done() sets the
FC_RQST_STATE_DONE flag.
Signed-off-by: Johannes Thumshirn
Now that all conversions are done, move the FibreChannel bsg code over to the
bsg library.
This patch is derived from work done by Mike Christie in 2011 [1] but only the
iscsi parts got merged back then.
[1] http://marc.info/?l=linux-scsi=131149780921009=2
Signed-off-by: Johannes Thumshirn
Unexport bsg_softirq_done() again, we don't need it outside of bsg-lib.c
anymore now that scsi_transport_fc is a pure bsg-lib client.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
block/bsg-lib.c | 3 +--
include/linux/bsg-lib.h |
Unify the interfaces of fc_bsg_jobdone and bsg_job_done. This will reduce the
diff when moving from 'struct fc_bsg_job' to a plain 'struct bsg_job' later
on.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
drivers/s390/scsi/zfcp_fc.c |
Don't use fc_bsg_job::request and fc_bsg_job::reply directly, but use
helper variables bsg_request and bsg_reply. This will be helpfull when
transitioning to bsg-lib.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
Export fc_bsg_jobdone so drivers can use it directly instead of doing
the round-trip via struct fc_bsg_job::job_done() and use it in the LLDDs.
As we've converted all LLDDs over to use fc_bsg_jobdone() directly,
we can remove the function pointer from struct fc_bsg_job as well.
Signed-off-by:
This series converts the current bsg usage in the FibreChannel drivers over
to use bsg-lib. SAS will follow once FC is in a good enough shape.
I did take some inspiration from a similar patchset from Mike Christie
dating back to 2011 but it's not a 1:1 copy. Patch 15/16 is heavily based
on his
struct fc_bsg_buffer is just a clone of struct bsg_buffer from bsg-lib,
so use this one instead.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
drivers/scsi/lpfc/lpfc_bsg.c | 3 ++-
drivers/scsi/scsi_transport_fc.c | 2 +-
Signed-off-by: Michael Cyr
Signed-off-by: Bryant G. Ly
---
drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
On Thu, Oct 13, 2016 at 03:10:48PM +0200, Hannes Reinecke wrote:
> When the queue depth is reduced we should print out the reason
> for this; it might be due to a queue full condition.
>
> Signed-off-by: Hannes Reinecke
> ---
Acked-by: Johannes Thumshirn
--
Provide fc_bsg_to_rport() helper that will become handy when we're moving
from struct fc_bsg_job to a plain struct bsg_job. Also move all LLDDs to use
the new helper.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
Add reference counting to 'struct bsg_job' so we can implement a reuqest
timeout handler for bsg_jobs, which is needed for Fibre Channel.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
block/bsg-lib.c | 7 +--
fc_destroy_bsgjob() and bsg_destroy_job() are now 1:1 copies, so use the
later. As bsg_destroy_job() comes from bsg-lib we need to select it in Kconfig
once CONFOG_SCSI_FC_ATTRS is active.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
Add bsg_job_put() and bsg_job_get() so don't need to export
bsg_destroy_job() any more.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
block/bsg-lib.c | 17 ++---
drivers/scsi/scsi_transport_fc.c | 4 ++--
It's not necessary to cast the result of kmalloc, since void pointers are
promoted to any other type. This also fixes following coccinelle warning:
casting value returned by memory allocation function to
(BIG_IOCTL_Command_struct *) is useless.
Signed-off-by: Javier Martinez Canillas
On Thu, Oct 13, 2016 at 01:42:06PM +0200, Hannes Reinecke wrote:
> On 10/12/2016 03:06 PM, Johannes Thumshirn wrote:
> > Implement kref backed reference counting instead of rolling our own. This
> > elimnates the need of the following fields in 'struct fc_bsg_job':
> > * ref_cnt
> > * state_flags
Hm, still behaves for me like I reported for v2:
http://marc.info/?l=linux-scsi=147637177902937=2
On 10/13/2016 05:00 PM, Johannes Thumshirn wrote:
Don't use fc_bsg_job::request and fc_bsg_job::reply directly, but use
helper variables bsg_request and bsg_reply. This will be helpfull when
Provide fc_bsg_to_shost() helper that will become handy when we're moving from
struct fc_bsg_job to a plain struct bsg_job. Also use this little helper in
the LLDDs.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
fc_bsg_jobdone() and bsg_job_done() are 1:1 copies now so use the bsg-lib one
instead of the FC private implementation.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
drivers/s390/scsi/zfcp_fc.c | 2 +-
drivers/scsi/bfa/bfad_bsg.c
Implement kref backed reference counting instead of rolling our own. This
elimnates the need of the following fields in 'struct fc_bsg_job':
* ref_cnt
* state_flags
* job_lock
bringing us close to unification of 'struct fc_bsg_job' and 'struct bsg_job'.
Signed-off-by: Johannes Thumshirn
Change FC drivers to use 'struct bsg_job' from bsg-lib.h instead of 'struct
fc_bsg_job' from scsi_transport_fc.h and remove 'struct fc_bsg_job'.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
drivers/s390/scsi/zfcp_ext.h | 4 +--
bsg_softirq_done() and fc_bsg_softirq_done() are copies of each other, so
ditch the fc specific one.
Signed-off-by: Johannes Thumshirn
Reviewed-by: Hannes Reinecke
---
block/bsg-lib.c | 3 ++-
drivers/scsi/scsi_transport_fc.c | 15
On Thu, Oct 13, 2016 at 05:15:25PM +0200, Steffen Maier wrote:
> I'm puzzled.
>
> $ git bisect start fc_bsg master
> Bisecting: 8 revisions left to test after this (roughly 3 steps)
> [005d51510eee6102636d5dbb06310531c5d46151] scsi: fc: implement kref backed
> reference counting
> $ git bisect
It appears that the mailing list email address doesn't exist anymore:
: host smtp.hp.com[15.73.96.116] said: 550 5.1.1
: Recipient address rejected: User unknown in
virtual alias table (in reply to RCPT TO command)
Signed-off-by: Javier
On 2016-10-13 10:23, Joao Pinto wrote:
Signed-off-by: Joao Pinto
---
MAINTAINERS | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 5ac91d8a..d9855b5 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -12084,7 +12084,8 @@ F:
This patch changes the way the IBM vSCSI server driver manages its
Command/Response Queue (CRQ). We used to register the CRQ with phyp at
probe time. Now we wait until tpg_enable_store. Similarly, when
tpg_enable_store is called to "disable" (i.e. the stored value is 0),
we unregister the queue
It's not necessary to cast the result of kmalloc, since void pointers are
promoted to any other type. This also fixes following coccinelle warning:
casting value returned by memory allocation function to (struct
cciss_scsi_adapter_data_t *) is useless.
Signed-off-by: Javier Martinez Canillas
On Thu, Oct 13, 2016 at 05:55:11PM +0200, Steffen Maier wrote:
> Hm, still behaves for me like I reported for v2:
> http://marc.info/?l=linux-scsi=147637177902937=2
>
Well given what you've wrote for v2 it's kinda expected.
Byte,
Johannes
--
Johannes Thumshirn
Signed-off-by: Joao Pinto
---
MAINTAINERS | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 5ac91d8a..d9855b5 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -12084,7 +12084,8 @@ F: Documentation/scsi/ufs.txt
F:
Various fixes and cleanups for the ibmvscsis driver.
The first is a sort of prequel to the second, which is the primary
change (and the biggest). The rest are fairly small fixes.
Michael Cyr (6):
ibmvscsis: Rearrange functions for future patches
ibmvscsis: Synchronize cmds at
These custom handling are no longer necessary, since we always retry UA
in scsi_execute now.
Signed-off-by: Gabriel Krisman Bertazi
---
drivers/scsi/scsi_lib.c | 21 ++---
drivers/scsi/sr_ioctl.c | 6 ++
2 files changed, 8 insertions(+), 19
Hi James,
Thanks for your review. Please see the v2 below.
> OK, so really this isn't what you want, because blk_execute_req may
> have used several of your retries, so you now get a maximum possible
> set of retries at UNIT_ATTENTION_RETRIES*retries. You need to start
> from the returned
On Thu, Oct 13, 2016 at 12:53 PM, Adam Manzanares
wrote:
> Patch adds an association between iocontext ioprio and the ioprio of a
> request. This value is set in blk_rq_set_prio which takes the request and
> the ioc as arguments. If the ioc is valid in blk_rq_set_prio
Commit afc3f83cb4a5 ("scsi: ipr: Add asynchronous error notification")
introduced the warn on shown below. To fix this, rather than attempting
to send the KOBJ_CHANGE uevent from interrupt context, which is what is
causing the WARN_ON, just wake the ipr worker thread which will send a
KOBJ_CHANGE
This patch checks to see if an ATA device supports NCQ command priorities.
If so and the user has specified an iocontext that indicates IO_PRIO_CLASS_RT
then we build a tf with a high priority command.
Signed-off-by: Adam Manzanares
---
drivers/ata/libata-core.c | 35
On Thu, Oct 13, 2016 at 1:09 PM, Jens Axboe wrote:
> On 10/13/2016 02:06 PM, Dan Williams wrote:
>>
>> On Thu, Oct 13, 2016 at 12:53 PM, Adam Manzanares
>> wrote:
>>>
>>> Patch adds an association between iocontext ioprio and the ioprio of a
>>>
The request priority is now by default coming from the ioc. It was not
clear what this code was trying to do based upon the iopriority class or
data. The driver should check that a device supports priorities and use
them according to the specificiations of ioprio.
Signed-off-by: Adam Manzanares
Patch adds an association between iocontext ioprio and the ioprio of a
request. This value is set in blk_rq_set_prio which takes the request and
the ioc as arguments. If the ioc is valid in blk_rq_set_prio then the
iopriority of the request is set as the iopriority of the ioc. In
This patch builds ATA commands with high priority if the iocontext
of a process is set to real time. The goal of the patch is to
improve tail latencies of workloads that use higher queue depths.
This required setting the iocontext ioprio on the request when
the request is initialized.
This patch
Add a sysfs entry to turn on priority information being passed
to a ATA device. By default this feature is turned off.
This patch depends on ata: Enabling ATA Command Priorities
Signed-off-by: Adam Manzanares
---
drivers/ata/libahci.c | 1 +
On 10/13/2016 02:06 PM, Dan Williams wrote:
On Thu, Oct 13, 2016 at 12:53 PM, Adam Manzanares
wrote:
Patch adds an association between iocontext ioprio and the ioprio of a
request. This value is set in blk_rq_set_prio which takes the request and
the ioc as arguments.
On 10/13/2016 11:10 AM, Javier Martinez Canillas wrote:
It's not necessary to cast the result of kmalloc, since void pointers are
promoted to any other type. This also fixes following coccinelle warning:
casting value returned by memory allocation function to
(BIG_IOCTL_Command_struct *) is
Set driver environment data on adapter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_hw4.h | 12
drivers/scsi/lpfc/lpfc_sli.c | 29
Code cleanup for lpfc_topology parameter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 7 +++
1 file changed, 3 insertions(+), 4
Make lpfc_prot_mask and lpfc_prot_guard per hba parameters
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc.h | 2 ++
drivers/scsi/lpfc/lpfc_attr.c
Fix sg_reset on SCSI device causing kernel crash
Driver could reference stale node pointers in task mgmt call.
Changed to use resetting cmd and look up node pointer in task mgmt
function.
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Correct embedded io wq element size. Embedded element sizes are
128 byte elements
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
v2 mods:
change style of selecting wqesize
This patch set updates the lpfc driver to revision 11.2.0.2
The patches were cut against scsi-misc
James Smart (17):
Fix few small typos in lpfc_scsi.c
originally posted by Milan Gandhi
lpfc: Correct embedded io wq element size
lpfc: Fix sg_reset on SCSI device causing kernel crash
Code cleanup for lpfc_enable_rrq parameter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 8 ++--
1 file changed, 2 insertions(+), 6
Code cleanup for lpfc_aer_support parameter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 43 +++
1
Code clean up for lpfc_iocb_cnt parameter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 8 +---
1 file changed, 1 insertion(+), 7
From: "Milan P. Gandhi"
This patch does a cleanup and fixes few small typos in lpfc_scsi.c
Signed-off-by: Milan P. Gandhi
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
On 10/13/2016 02:59 PM, Dan Williams wrote:
On Thu, Oct 13, 2016 at 1:24 PM, Jens Axboe wrote:
On 10/13/2016 02:19 PM, Dan Williams wrote:
On Thu, Oct 13, 2016 at 1:09 PM, Jens Axboe wrote:
On 10/13/2016 02:06 PM, Dan Williams wrote:
On Thu, Oct 13,
On Thu, Oct 13, 2016 at 1:24 PM, Jens Axboe wrote:
> On 10/13/2016 02:19 PM, Dan Williams wrote:
>>
>> On Thu, Oct 13, 2016 at 1:09 PM, Jens Axboe wrote:
>>>
>>> On 10/13/2016 02:06 PM, Dan Williams wrote:
On Thu, Oct 13, 2016 at 12:53 PM, Adam
The 10/13/2016 14:09, Jens Axboe wrote:
> On 10/13/2016 02:06 PM, Dan Williams wrote:
> >On Thu, Oct 13, 2016 at 12:53 PM, Adam Manzanares
> > wrote:
> >>Patch adds an association between iocontext ioprio and the ioprio of a
> >>request. This value is set in
Synchronize link speed with boot driver
Link speed settings set by the boot driver are reported by the hw.
Driver will attempt to read them, and if set, will respect their
values.
The driver can override the settings with its own if instructed by
user space (via bsg), with the new values being
Code cleanup for lpfc_max_scsicmpl_time parameter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 6 ++
1 file changed, 2 insertions(+), 4
Fix lost target in pt-to-pt connect
Change reject code to something that allows a retry
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_els.c | 2 +-
1
Fix fw download on SLI-4 FC adapters
Driver performs a quick validation of magic numbers in the fw
download image. Driver needed to be updated for more recent
magic numbers.
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
---
v3
Code cleanup for lpfc_sriov_nr_virtfn parameter
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 35 +++
1 file
Revise strings with full lpfc parameter name
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_attr.c | 23 ---
1 file changed, 12
Correct panics with eh_timeout and eh_deadline
We were having double completions on our SLI-3 version of adapters.
Solved by clearing our command pointer before calling scsi_done.
The eh paths potentially ran simulatenously and would see the non-null
value and invoke scsi_done again.
lpfc version changed to 11.2.0.2
Signed-off-by: Dick Kennedy
Signed-off-by: James Smart
Reviewed-by: Johannes Thumshirn
---
drivers/scsi/lpfc/lpfc_version.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff
On 10/13/2016 02:19 PM, Dan Williams wrote:
On Thu, Oct 13, 2016 at 1:09 PM, Jens Axboe wrote:
On 10/13/2016 02:06 PM, Dan Williams wrote:
On Thu, Oct 13, 2016 at 12:53 PM, Adam Manzanares
wrote:
Patch adds an association between iocontext ioprio
By removing the code below, we put all the commands for all the types of
devices (SAS/SATA) as simple-Q (requeue as the device require) and I am
not sure whether it is the intention of this change.
-Original Message-
From: Adam Manzanares [mailto:adam.manzana...@hgst.com]
Sent: Thursday,
This patch checks to see if an ATA device supports NCQ command priorities.
If so and the user has specified an iocontext that indicates
IO_PRIO_CLASS_RT then we build a tf with a high priority command.
This is done to improve the tail latency of commands that are high
priority by passing priority
Hello Tejun,
The 10/13/2016 19:22, Tejun Heo wrote:
> Hello, Adam.
>
> Sorry about late reply. Was on vacation.
NP I was on vacation at the end of the week last week.
>
> On Thu, Oct 13, 2016 at 04:00:31PM -0700, Adam Manzanares wrote:
> > Add a sysfs entry to turn on priority information
Hello,
On Thu, Oct 13, 2016 at 04:00:30PM -0700, Adam Manzanares wrote:
> This patch checks to see if an ATA device supports NCQ command priorities.
> If so and the user has specified an iocontext that indicates
> IO_PRIO_CLASS_RT then we build a tf with a high priority command.
>
> This is done
This patch builds ATA commands with high priority if the iocontext of a process
is set to real time. The goal of the patch is to improve tail latencies of
workloads that use higher queue depths. This requires setting the iocontext
ioprio on the request when it is initialized.
This patch has
On Thu, Oct 13, 2016 at 6:00 PM, Adam Manzanares
wrote:
> The request priority is now by default coming from the ioc. It was not
> clear what this code was trying to do based upon the iopriority class or
> data. The driver should check that a device supports priorities
This patch reorders functions in a manner necessary for a follow-on
patch. It also makes some minor styling changes (mostly removing extra
spaces) and fixes some typos.
There are no code changes in this patch, with one exception: due to the
reordering of the functions, I needed to explicitly
Add a sysfs entry to turn on priority information being passed
to a ATA device. By default this feature is turned off.
This patch depends on ata: Enabling ATA Command Priorities
Signed-off-by: Adam Manzanares
---
drivers/ata/libahci.c | 1 +
The request priority is now by default coming from the ioc. It was not
clear what this code was trying to do based upon the iopriority class or
data. The driver should check that a device supports priorities and use
them according to the specificiations of ioprio.
Signed-off-by: Adam Manzanares
Patch adds an association between iocontext ioprio and the ioprio of a
request. This is done to enable request based drivers the ability to
act on priority information stored in the request. An example being
ATA devices that support command priorities. If the ATA driver discovers
that the device
While stressing memory and IO at the same time we changed SMT settings,
we were able to consistently trigger deadlocks in the mm system, which
froze the entire machine.
I think that under memory stress conditions, the large allocations
performed by blk_mq_init_rq_map may trigger a reclaim, which
On Thu, 2016-10-13 at 15:47 -0300, Gabriel Krisman Bertazi wrote:
> Hi James,
>
> Thanks for your review. Please see the v2 below.
>
> > OK, so really this isn't what you want, because blk_execute_req may
> > have used several of your retries, so you now get a maximum
> > possible set of
> Style wise you can further remove the extra parens around
> SCpnt->device->tagged_supported
> As well as the now redundant braces.
I did send a patch looking just like that earlier :)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to
The 10/13/2016 15:05, Sathya Prakash Veerichetty wrote:
> By removing the code below, we put all the commands for all the types of
> devices (SAS/SATA) as simple-Q (requeue as the device require) and I am
> not sure whether it is the intention of this change.
>
This is the intention of the
On 10/13/2016 09:53 PM, Adam Manzanares wrote:
> The request priority is now by default coming from the ioc. It was not
> clear what this code was trying to do based upon the iopriority class or
> data. The driver should check that a device supports priorities and use
> them according to the
On 10/13/2016 09:53 PM, Adam Manzanares wrote:
> Patch adds an association between iocontext ioprio and the ioprio of a
> request. This value is set in blk_rq_set_prio which takes the request and
> the ioc as arguments. If the ioc is valid in blk_rq_set_prio then the
> iopriority of the request is
These patches are being resent with required changes as per comments
for [PATCH 02/28] be2iscsi: Replace _bh with _irqsave/irqrestore
In IO handling path, there is no spin_lock taken with IRQs disabled
when calling into libiscsi, so no issues were found with use of
spin_lock_bh/spin_unlock_bh
On Wed, Oct 12, 2016 at 01:14:18PM -0700, James Smart wrote:
>
> Synchronize link speed with boot driver
>
> Link speed settings set by the boot driver are reported by the hw.
> Driver will attempt to read them, and if set, will respect their
> values.
> The driver can override the settings with
On Wed, Oct 12, 2016 at 01:14:17PM -0700, James Smart wrote:
>
> Correct panics with eh_timeout and eh_deadline
>
> We were having double completions on our SLI-3 version of adapters.
> Solved by clearing our command pointer before calling scsi_done.
>
> The eh paths potentially ran
The code at free_task label in __iscsi_conn_send_pdu can get executed
from blk_timeout_work which takes queue_lock using spin_lock_irq.
back_lock taken with spin_unlock_bh will cause WARN_ON_ONCE.
The code gets executed either with bottom half or IRQ disabled hence
using spin_lock/spin_unlock for
[ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150
__local_bh_enable_ip+0x6b/0x90
[ 3843.142815] Modules linked in:
...
[ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: GE
4.8.0-rc1+ #3
[ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS
On Wed, Oct 12, 2016 at 01:14:06PM -0700, James Smart wrote:
>
> Fix sg_reset on SCSI device causing kernel crash
>
> Driver could reference stale node pointers in task mgmt call.
> Changed to use resetting cmd and look up node pointer in task mgmt
> function.
>
> Modified per review:
> Revised
Having an I/O priority does not mean we should send all requests as
HEAD OF QUEUE tags.
Reported-by: Adam Manzanares
Signed-off-by: Christoph Hellwig
---
drivers/message/fusion/mptscsih.c | 11 +++
1 file changed, 3 insertions(+), 8 deletions(-)
On 10/12/2016 03:06 PM, Johannes Thumshirn wrote:
> struct fc_bsg_buffer is just a clone of struct bsg_buffer from bsg-lib,
> so use this one instead.
>
> Signed-off-by: Johannes Thumshirn
> ---
> drivers/scsi/lpfc/lpfc_bsg.c | 3 ++-
> drivers/scsi/scsi_transport_fc.c
On Wed, Oct 12, 2016 at 01:14:04PM -0700, James Smart wrote:
>
> From: "Milan P. Gandhi"
>
> This patch does a cleanup and fixes few small typos in lpfc_scsi.c
>
> Signed-off-by: Milan P. Gandhi
> Signed-off-by: James Smart
>
On Wed, Oct 12, 2016 at 01:14:19PM -0700, James Smart wrote:
>
> Fix fw download on SLI-4 FC adapters
>
> Driver performs a quick validation of magic numbers in the fw
> download image. Driver needed to be updated for more recent
> magic numbers.
>
> Signed-off-by: Dick Kennedy
On 13/10/2016 02:20, Lorenzo Stoakes wrote:
> This patch removes the write and force parameters from
> __get_user_pages_unlocked() to make the use of FOLL_FORCE explicit in callers
> as
> use of this flag can result in surprising behaviour (and hence bugs) within
> the
> mm subsystem.
>
>
On Wed, Oct 12, 2016 at 01:14:05PM -0700, James Smart wrote:
>
> Correct embedded io wq element size. Embedded element sizes are
> 128 byte elements
>
> Modified per review:
> change style of selecting wqesize using local variable
This should go below the "---" line after the Signed-off-bys.
>
On Wed, Oct 12, 2016 at 05:54:45PM +0200, Steffen Maier wrote:
> Hi Johannes,
>
> On 10/12/2016 03:06 PM, Johannes Thumshirn wrote:
> > This series converts the current bsg usage in the FibreChannel drivers over
> > to use bsg-lib. SAS will follow once FC is in a good enough shape.
> >
> > I did
On 10/13/2016 10:46 AM, Christoph Hellwig wrote:
> Having an I/O priority does not mean we should send all requests as
> HEAD OF QUEUE tags.
>
> Reported-by: Adam Manzanares
> Signed-off-by: Christoph Hellwig
> ---
> drivers/message/fusion/mptscsih.c | 11
Dan, many thanks for catching this! Sparse did not notice, is there
other tooling that would find such things?
James, Martin, could you please queue this as fix for one of my patches
that went into the 4.9 merge window, so for 4.9-rc I guess?
On 10/12/2016 03:06 PM, Johannes Thumshirn wrote:
> Don't set FC_RQST_STATE_DONE before calling fc_bsg_jobdone() as
> fc_bsg_jobdone() calls blk_complete_requeust() which raises a soft-IRQ that
> ends up in fc_bsg_sofirq_done() and fc_bsg_softirq_done() sets the
> FC_RQST_STATE_DONE flag.
>
>
1 - 100 of 148 matches
Mail list logo