Messages in this thread | | | Date | Mon, 20 Apr 2015 11:21:27 +0800 | From | Lai Jiangshan <> | Subject | Re: [PATCH 1/3 V7] workqueue: split apply_workqueue_attrs() into 3 stages |
| |
On 04/17/2015 10:57 PM, Tejun Heo wrote: > On Tue, Apr 07, 2015 at 07:26:35PM +0800, Lai Jiangshan wrote: >> Current apply_workqueue_attrs() includes pwqs-allocation and pwqs-installation, >> so when we batch multiple apply_workqueue_attrs()s as a transaction, we can't >> ensure the transaction must succeed or fail as a complete unit. > > Lai, can you please > > * Break out threads when posting new version. > > * List the changes since the last version?
My bad.
I thought you had only two comments which ware handled in v7. It was considered (with my laziness) too less to be renarrated.
---
In [2/4 V6]: ```quote from TJ: And we're dropping online_cpus locking before applying the new pwq's. Is that safe? ```
It was my fault, I didn't remember when I wrongly moved the code. The patch had been tried my best keep functionality unchanged. It is fixed in [1/3 V7].
----
The changes from [4/4 V6] to [3/3 V7]: ctx->attrs (the original configured value) is not calculated until it is saved into ctx->attrs, and a corresponding local-var is killed.
Could you please consider it as [0/3 V7] this time?
Thx Lai.
> > Thanks. >
| |