|
Hello,
On Wed, 5 Dec 2001, Julian Anastasov wrote:
>
> Hello,
>
> On Wed, 5 Dec 2001, Wensong Zhang wrote:
>
> > Thanks a lot for reporting and finding the problem.
> >
> > I made 0.9.8pre1.diff (http://linux-vs.org/~wensong/ipvs-0.9.8pre1.diff),
> > which includes adding nat_icmp_send sysctl variable, changing to use
> > spin_lock_bh to guard data access between network softirqs and bottom half
> > in the slow timer handling, etc.
>
> But is there any _bh change that is needed? There are so
> many _bh changes but I don't see one that is needed. If I understand
> correctly Lionel Bringuier, he never tried the plain IPVS, he patched
> it and makes some calls from user space? This is my understanding
> but I don't know what is the case. IMO, the caller should use the
> local_bh_{enable,disable} calls, not IPVS. I don't remember for such
> problems on SMP and I don't see them in the sources. Am I missing
> something?
>
No, it is my mistake. I don't know why I suddenly thought that network
softirq might be interrupted by a bottom half at a CPU last night. :)
Just had a look at the do_softirq(), it disables bottom half before
calling softirqs. So, there is no need to disable bh again in mod_sltimer.
will remove the _bh change soon.
Thanks,
Wensong
|
