Messages in this thread | | | Date | Fri, 2 May 2008 13:21:24 +0800 (CST) | Subject | Re: [Patch]net/xfrm/xfrm_policy.c: replace timer with delayed_work | From | WANG Cong <> |
| |
On Thu, 1 May 2008, Herbert Xu wrote:
> On Thu, May 01, 2008 at 12:06:55AM +0800, WANG Cong wrote: >> >> Hi, Herbert. >> >> I referenced the book Linux Device Drivers, it is said that >> the work function in workqueue can sleep, that is to say, it only >> works in process context, so we don't need the disable the lock >> in softirq, which is interrupt context, right? > > Because it can sleep, you must disable BH for locks that can > be taken on the softirq path as otherwise a softirq can come > in, try to take the lock again and dead-lock.
Yes, I was an idiot.
I will resend the revised patches right now.
| |