Created a thread using alloc_ordered_workqueue() API in order to process
the works from firmware Work-queue sequentially instead of
create_singlethread_workqueue() API.
Changes in v1:
No need to check for backport compatibility in the upstream kernel.
so removing the else section where driver use
create_singlethread_workqueue() API if alloc_ordered_workqueue() API is
not defined, This else section is not required since in the latest upstream
kernel this alloc_ordered_workqueue() API is always defined.
Signed-off-by: Sreekanth Reddy <[email protected]>
---
drivers/scsi/mpt3sas/mpt3sas_scsih.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
index b848458..7e5926c 100644
--- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c
+++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
@@ -8085,8 +8085,8 @@ _scsih_probe(struct pci_dev *pdev, const struct pci_device_id *id)
/* event thread */
snprintf(ioc->firmware_event_name, sizeof(ioc->firmware_event_name),
"fw_event%d", ioc->id);
- ioc->firmware_event_thread = create_singlethread_workqueue(
- ioc->firmware_event_name);
+ ioc->firmware_event_thread = alloc_ordered_workqueue(
+ ioc->firmware_event_name, WQ_MEM_RECLAIM);
if (!ioc->firmware_event_thread) {
pr_err(MPT3SAS_FMT "failure at %s:%d/%s()!\n",
ioc->name, __FILE__, __LINE__, __func__);
--
2.0.2
On 06/16/2015 01:37 AM, Sreekanth Reddy wrote:
> Created a thread using alloc_ordered_workqueue() API in order to process
> the works from firmware Work-queue sequentially instead of
> create_singlethread_workqueue() API.
>
> Changes in v1:
> No need to check for backport compatibility in the upstream kernel.
> so removing the else section where driver use
> create_singlethread_workqueue() API if alloc_ordered_workqueue() API is
> not defined, This else section is not required since in the latest upstream
> kernel this alloc_ordered_workqueue() API is always defined.
>
> Signed-off-by: Sreekanth Reddy <[email protected]>
> ---
> drivers/scsi/mpt3sas/mpt3sas_scsih.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
> index b848458..7e5926c 100644
> --- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c
> +++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
> @@ -8085,8 +8085,8 @@ _scsih_probe(struct pci_dev *pdev, const struct pci_device_id *id)
> /* event thread */
> snprintf(ioc->firmware_event_name, sizeof(ioc->firmware_event_name),
> "fw_event%d", ioc->id);
> - ioc->firmware_event_thread = create_singlethread_workqueue(
> - ioc->firmware_event_name);
> + ioc->firmware_event_thread = alloc_ordered_workqueue(
> + ioc->firmware_event_name, WQ_MEM_RECLAIM);
> if (!ioc->firmware_event_thread) {
> pr_err(MPT3SAS_FMT "failure at %s:%d/%s()!\n",
> ioc->name, __FILE__, __LINE__, __func__);
>
Hi Sreekanth,
Is this change still needed after e09c2c2954684 workqueue: apply
__WQ_ORDERED to create_singlethread_workqueue() ? (3.17+)
In upstream, this change looks cosmetic (unless Tejun has a preference
for one over the other), but maybe converting to alloc_ordered_workqueue
keeps your in house version in closer sync?
Thanks,
-- Joe
On Thu, Jun 18, 2015 at 5:40 PM, Joe Lawrence <[email protected]> wrote:
> On 06/16/2015 01:37 AM, Sreekanth Reddy wrote:
>> Created a thread using alloc_ordered_workqueue() API in order to process
>> the works from firmware Work-queue sequentially instead of
>> create_singlethread_workqueue() API.
>>
>> Changes in v1:
>> No need to check for backport compatibility in the upstream kernel.
>> so removing the else section where driver use
>> create_singlethread_workqueue() API if alloc_ordered_workqueue() API is
>> not defined, This else section is not required since in the latest upstream
>> kernel this alloc_ordered_workqueue() API is always defined.
>>
>> Signed-off-by: Sreekanth Reddy <[email protected]>
>> ---
>> drivers/scsi/mpt3sas/mpt3sas_scsih.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
>> index b848458..7e5926c 100644
>> --- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c
>> +++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
>> @@ -8085,8 +8085,8 @@ _scsih_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>> /* event thread */
>> snprintf(ioc->firmware_event_name, sizeof(ioc->firmware_event_name),
>> "fw_event%d", ioc->id);
>> - ioc->firmware_event_thread = create_singlethread_workqueue(
>> - ioc->firmware_event_name);
>> + ioc->firmware_event_thread = alloc_ordered_workqueue(
>> + ioc->firmware_event_name, WQ_MEM_RECLAIM);
>> if (!ioc->firmware_event_thread) {
>> pr_err(MPT3SAS_FMT "failure at %s:%d/%s()!\n",
>> ioc->name, __FILE__, __LINE__, __func__);
>>
>
> Hi Sreekanth,
>
> Is this change still needed after e09c2c2954684 workqueue: apply
> __WQ_ORDERED to create_singlethread_workqueue() ? (3.17+)
I won't say that it is compulsory required, but I feel it is better if
these changes are included. since initially we thought that thread
created by using create_singlethread_workqueue() will process the
works sequentially but in-between it has broken and then it is fixed
by Tejun. So I thought it is better to directly use the
alloc_ordered_workqueue() as create_singlethead_workqueue() API also
invoked the same API.
>
> In upstream, this change looks cosmetic (unless Tejun has a preference
> for one over the other), but maybe converting to alloc_ordered_workqueue
> keeps your in house version in closer sync?
>
> Thanks,
>
> -- Joe
--
Regards,
Sreekanth
On 06/18/2015 09:06 AM, Sreekanth Reddy wrote:
> On Thu, Jun 18, 2015 at 5:40 PM, Joe Lawrence <[email protected]> wrote:
>> On 06/16/2015 01:37 AM, Sreekanth Reddy wrote:
>>> Created a thread using alloc_ordered_workqueue() API in order to process
>>> the works from firmware Work-queue sequentially instead of
>>> create_singlethread_workqueue() API.
>>>
>>> Changes in v1:
>>> No need to check for backport compatibility in the upstream kernel.
>>> so removing the else section where driver use
>>> create_singlethread_workqueue() API if alloc_ordered_workqueue() API is
>>> not defined, This else section is not required since in the latest upstream
>>> kernel this alloc_ordered_workqueue() API is always defined.
>>>
>>> Signed-off-by: Sreekanth Reddy <[email protected]>
>>> ---
>>> drivers/scsi/mpt3sas/mpt3sas_scsih.c | 4 ++--
>>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
>>> index b848458..7e5926c 100644
>>> --- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c
>>> +++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c
>>> @@ -8085,8 +8085,8 @@ _scsih_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>>> /* event thread */
>>> snprintf(ioc->firmware_event_name, sizeof(ioc->firmware_event_name),
>>> "fw_event%d", ioc->id);
>>> - ioc->firmware_event_thread = create_singlethread_workqueue(
>>> - ioc->firmware_event_name);
>>> + ioc->firmware_event_thread = alloc_ordered_workqueue(
>>> + ioc->firmware_event_name, WQ_MEM_RECLAIM);
>>> if (!ioc->firmware_event_thread) {
>>> pr_err(MPT3SAS_FMT "failure at %s:%d/%s()!\n",
>>> ioc->name, __FILE__, __LINE__, __func__);
>>>
>>
>> Hi Sreekanth,
>>
>> Is this change still needed after e09c2c2954684 workqueue: apply
>> __WQ_ORDERED to create_singlethread_workqueue() ? (3.17+)
>
> I won't say that it is compulsory required, but I feel it is better if
> these changes are included. since initially we thought that thread
> created by using create_singlethread_workqueue() will process the
> works sequentially but in-between it has broken and then it is fixed
> by Tejun. So I thought it is better to directly use the
> alloc_ordered_workqueue() as create_singlethead_workqueue() API also
> invoked the same API.
Ok, I was just wondering if maybe create_singlethread_workqueue was
fixed after this patch was initially written. Since it's effectively
the same...
Reviewed-by: Joe Lawrence <[email protected]>
Regards,
-- Joe
>>>>> Sreekanth Reddy <[email protected]> writes:
> Created a thread using alloc_ordered_workqueue() API in order to
> process the works from firmware Work-queue sequentially instead of
> create_singlethread_workqueue() API.
Reviewed-by: Martin K. Petersen <[email protected]>
--
Martin K. Petersen Oracle Linux Engineering