Re: [PATCH 03/15] nfsd: make reclaim_str_hashtbl allocated per net

[Stanislav Kinsbursky <skinsbursky@parallels.com>]

14.11.2012 17:02, Jeff Layton пишет:
> On Tue, 13 Nov 2012 18:48:33 +0300
> Stanislav Kinsbursky <skinsbursky@parallels.com> wrote:
>
<snip>
>>
>> +
>> +static int
>> +nfs4_legacy_state_init(struct net *net)
>> +{
>> +     struct nfsd_net *nn = net_generic(net, nfsd_net_id);
>> +     int i;
>> +
>> +     nn->reclaim_str_hashtbl = kmalloc(sizeof(struct list_head) *
>> +                                       CLIENT_HASH_SIZE, GFP_KERNEL);
>> +     if (!nn->reclaim_str_hashtbl)
>> +             return -ENOMEM;
>> +
>> +     for (i = 0; i < CLIENT_HASH_SIZE; i++)
>> +             INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]);
>> +     nn->reclaim_str_hashtbl_size = 0;
>> +
>> +     return 0;
>> +}
>> +
>> +static void
>> +nfs4_legacy_state_shutdown(struct net *net)
>> +{
>> +     struct nfsd_net *nn = net_generic(net, nfsd_net_id);
>> +
>> +     kfree(nn->reclaim_str_hashtbl);
>> +}
>> +
>>   static int
>>   nfsd4_load_reboot_recovery_data(struct net *net)
>>   {
>> @@ -425,13 +454,23 @@ nfsd4_load_reboot_recovery_data(struct net *net)
>>                return -EINVAL;
>>        }
>>
>> +     status = nfs4_legacy_state_init(net);
>> +     if (status)
>> +             return status;
>> +
>
> nit: might be clearer to create a new legacy ->init function and have
> it call the state_init and then load the recovery data.
>

Yep, sure.

>
> Looks like a reasonable and good change, but may need reconciling with
> the patches from me that Bruce merged recently into his for-next tree.
>

Thanks. Already working on it.

-- 
Best regards,
Stanislav Kinsbursky