[Devel] [PATCH rh7] ve: fs -- Export fs.aio-nr and fs.aio-max-nr via ve cgroup

[Kirill Tkhai]

On 09.04.2016 19:38, Konstantin Khorenko wrote:
> Kirill, your input is required here.
> 
> Do we need to migrate both fs.aio-(max)-nr or it's enough to migrate fs.aio-max-nr?

We don't need to migrate fs.aio-nr. It's set when aio requests are queued.

If we migrate requests in the future, the true aio-nr will be set automatically.
Now we don't migrate requests (we are almost wait them), so it's 0 on src and dst.

> 
> -- 
> Best regards,
> 
> Konstantin Khorenko,
> Virtuozzo Linux Kernel Team
> 
> On 04/09/2016 12:49 PM, Cyrill Gorcunov wrote:
>> This members represents fs.aio-(max)-nr sysctl entries.
>> While fs.aio-max-nr can be adjusted from inside of ve with
>> traditional sysctl utility the @aio-nr is read only so we
>> need an interface for its c/r.
>>
>> Moreover I think being able to c/r from libvzctl script is
>> better in terms of keeping c/r work in one place, so instead
>> of handling fs.aio-max-nr with sysctl utility from inside of ve
>> context, I do it via scripts and for this sake fs.aio-max-nr
>> is exported via cgroup api.
>>
>> https://jira.sw.ru/browse/PSBM-40690
>>
>> Signed-off-by: Cyrill Gorcunov <gorcunov at virtuozzo.com>
>> CC: Vladimir Davydov <vdavydov at virtuozzo.com>
>> CC: Konstantin Khorenko <khorenko at virtuozzo.com>
>> ---
>>   kernel/ve/ve.c |   29 +++++++++++++++++++++++++++--
>>   1 file changed, 27 insertions(+), 2 deletions(-)
>>
>> Index: linux-pcs7.git/kernel/ve/ve.c
>> ===================================================================
>> --- linux-pcs7.git.orig/kernel/ve/ve.c
>> +++ linux-pcs7.git/kernel/ve/ve.c
>> @@ -1180,7 +1180,9 @@ enum {
>>       VE_CF_IPTABLES_MASK,
>>       VE_CF_PSEUDOSUPER,
>>       VE_CF_CLOCK_MONOTONIC,
>> -    VE_CF_CLOCK_BOOTBASED
>> +    VE_CF_CLOCK_BOOTBASED,
>> +    VE_CF_AIO_NR,
>> +    VE_CF_AIO_MAX_NR,
>>   };
>>
>>   static int ve_ts_read(struct cgroup *cg, struct cftype *cft, struct seq_file *m)
>> @@ -1240,6 +1242,10 @@ static u64 ve_read_u64(struct cgroup *cg
>>   #endif
>>       else if (cft->private == VE_CF_PSEUDOSUPER)
>>           return cgroup_ve(cg)->is_pseudosuper;
>> +    else if (cft->private == VE_CF_AIO_NR)
>> +        return cgroup_ve(cg)->aio_nr;
>> +    else if (cft->private == VE_CF_AIO_MAX_NR)
>> +        return cgroup_ve(cg)->aio_max_nr;
>>       return 0;
>>   }
>>
>> @@ -1276,7 +1282,8 @@ static int ve_write_u64(struct cgroup *c
>>   {
>>       struct ve_struct *ve = cgroup_ve(cg);
>>
>> -    if (!ve_is_super(get_exec_env()))
>> +    if (!ve_is_super(get_exec_env()) &&
>> +        !ve->is_pseudosuper)
>>           return -EPERM;
>>
>>       down_write(&ve->op_sem);
>> @@ -1291,6 +1298,10 @@ static int ve_write_u64(struct cgroup *c
>>       else if (cft->private == VE_CF_IPTABLES_MASK)
>>           ve->ipt_mask = ve_setup_iptables_mask(value);
>>   #endif
>> +    else if (cft->private == VE_CF_AIO_NR)
>> +        ve->aio_nr = value;
>> +    else if (cft->private == VE_CF_AIO_MAX_NR)
>> +        ve->aio_max_nr = value;
>>       up_write(&ve->op_sem);
>>       return 0;
>>   }
>> @@ -1359,6 +1370,20 @@ static struct cftype ve_cftypes[] = {
>>           .write_string        = ve_ts_write,
>>           .private        = VE_CF_CLOCK_BOOTBASED,
>>       },
>> +    {
>> +        .name            = "aio_nr",
>> +        .flags            = CFTYPE_NOT_ON_ROOT,
>> +        .read_u64        = ve_read_u64,
>> +        .write_u64        = ve_write_u64,
>> +        .private        = VE_CF_AIO_NR,
>> +    },
>> +    {
>> +        .name            = "aio_max_nr",
>> +        .flags            = CFTYPE_NOT_ON_ROOT,
>> +        .read_u64        = ve_read_u64,
>> +        .write_u64        = ve_write_u64,
>> +        .private        = VE_CF_AIO_MAX_NR,
>> +    },
>>       { }
>>   };
>>
>> .
>>