Subject: Re: [PATCH net-next 3/9] bpf: pre-allocate hash map elements
To: Daniel Borkmann <daniel@iogearbox.net>,
        "David S . Miller" <davem@davemloft.net>
References: <1457315917-1970307-1-git-send-email-ast@fb.com>
 <1457315917-1970307-4-git-send-email-ast@fb.com>
 <56DD6118.7090608@iogearbox.net>
CC: Daniel Wagner <daniel.wagner@bmw-carit.de>,
        Tom Zanussi <tom.zanussi@linux.intel.com>,
        Wang Nan <wangnan0@huawei.com>, He Kuang <hekuang@huawei.com>,
        Martin KaFai Lau <kafai@fb.com>,
        Brendan Gregg <brendan.d.gregg@gmail.com>, <netdev@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>, <kernel-team@fb.com>
From: Alexei Starovoitov <ast@fb.com>
Message-ID: <56DDC872.5010001@fb.com>
Date: Mon, 7 Mar 2016 10:29:06 -0800
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:38.0)
 Gecko/20100101 Thunderbird/38.6.0
MIME-Version: 1.0
In-Reply-To: <56DD6118.7090608@iogearbox.net>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1185
Lines: 42

On 3/7/16 3:08 AM, Daniel Borkmann wrote:
> On 03/07/2016 02:58 AM, Alexei Starovoitov wrote:
> [...]
>> ---
>>   include/linux/bpf.h      |   1 +
>>   include/uapi/linux/bpf.h |   3 +
>>   kernel/bpf/hashtab.c     | 264
>> ++++++++++++++++++++++++++++++++++-------------
>>   kernel/bpf/syscall.c     |   2 +-
>>   4 files changed, 196 insertions(+), 74 deletions(-)
>
> Shouldn't all other map types (like array) need something like this as
> well to
> reserve this for their future flags?
>
>    if (attr->map_flags)
>      return ERR_PTR(-EINVAL);

yeah. good point. will add another patch for that.

>> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
>> index 4b070827200d..c81efb10bbb5 100644
>> --- a/include/linux/bpf.h
>> +++ b/include/linux/bpf.h
>> @@ -37,6 +37,7 @@ struct bpf_map {
>>       u32 key_size;
>>       u32 value_size;
>>       u32 max_entries;
>> +    u32 map_flags;
>
> Just naming this 'flags' doesn't work due to the anonymous struct inside
> that
> union, right? :/

yep. exactly. there is already 'flags' member there.

>>
>> +#define BPF_F_NO_PREALLOC    (1ULL << 0)
>
> Nit: Should better be (1U << 0) as map_flags are of __u32.

right. will do.