Return-Path: Subject: Re: [RFCv2 bluetooth-next 2/3] ipv6: add ipv6_addr_prefix_cpy To: Alexander Aring References: <1447799594-6050-1-git-send-email-alex.aring@gmail.com> <1447799594-6050-3-git-send-email-alex.aring@gmail.com> <5655E50C.5090906@osg.samsung.com> <20151125181713.GC4652@omega> Cc: linux-wpan@vger.kernel.org, linux-bluetooth@vger.kernel.org, netdev@vger.kernel.org, kernel@pengutronix.de, mcr@sandelman.ca, lukasz.duda@nordicsemi.no, martin.gergeleit@hs-rm.de, "David S . Miller" , Alexey Kuznetsov , James Morris , Hideaki YOSHIFUJI , Patrick McHardy From: Stefan Schmidt Message-ID: <5656E8CB.5050903@osg.samsung.com> Date: Thu, 26 Nov 2015 12:11:07 +0100 MIME-Version: 1.0 In-Reply-To: <20151125181713.GC4652@omega> Content-Type: text/plain; charset=utf-8; format=flowed Sender: linux-wpan-owner@vger.kernel.org List-ID: Hello. On 25/11/15 19:17, Alexander Aring wrote: > On Wed, Nov 25, 2015 at 05:42:52PM +0100, Stefan Schmidt wrote: >> Hello. >> >> On 17/11/15 23:33, Alexander Aring wrote: >>> This patch adds a static inline function ipv6_addr_prefix_cpy which >>> copies a ipv6 address prefix(argument pfx) into the ipv6 address prefix. >>> The prefix len is given by plen as bits. This function mainly based on >>> ipv6_addr_prefix which copies one address prefix from address into a new >>> ipv6 address destination and zero all other address bits. >>> >>> The difference is that ipv6_addr_prefix_cpy don't get a prefix from an >>> ipv6 address, it sets a prefix to an ipv6 address with keeping other >>> address bits. The use case is for context based address compression >>> inside 6LoWPAN IPHC header which keeping ipv6 prefixes inside a context >>> table to lookup address-bits without sending them. >>> >>> Cc: David S. Miller >>> Cc: Alexey Kuznetsov >>> Cc: James Morris >>> Cc: Hideaki YOSHIFUJI >>> Cc: Patrick McHardy >>> Signed-off-by: Alexander Aring >>> --- >>> include/net/ipv6.h | 15 +++++++++++++++ >>> 1 file changed, 15 insertions(+) >>> >>> diff --git a/include/net/ipv6.h b/include/net/ipv6.h >>> index e1a10b0..9d38fc2 100644 >>> --- a/include/net/ipv6.h >>> +++ b/include/net/ipv6.h >>> @@ -382,6 +382,21 @@ static inline void ipv6_addr_prefix(struct in6_addr *pfx, >>> pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b); >>> } >>> +static inline void ipv6_addr_prefix_cpy(struct in6_addr *addr, >>> + const struct in6_addr *pfx, >>> + int plen) >>> +{ >> Naming it ipv6_addr_prefix_cop, as Sergei suggested, would be easier to >> read. > ok. > >>> + /* caller must guarantee 0 <= plen <= 128 */ >>> + int o = plen >> 3, >>> + b = plen & 0x7; >>> + >> Any reason you are not doing the memset here before memcpy like it is done >> in ipv6_addr_prefi()? > ipv6_addr_prefix is more a "getter" for a prefix from an address. The > destination address (which should be the result) will memset the whole > address before copy the prefix from address to it. Ah, right. Missed that part. > THe ipv6_addr_prefix_copy will set a prefix given by second parameter > pfx and set the prefix to the address (given by first parameter). > > Maybe the ipv6_addr_prefix could call: > > 1. memset(pfx, sizeof...) > 2. ipv6_addr_prefix_copy(pfx, addr, plen); > > but ipv6_addr_prefix_copy need to care about other bits which are set at > destination address, if (plen/8 != 0). I don't want to touch > ipv6_addr_prefix function and slow-down some mechanism. No need to change this from my side. I just missed the memset being handled in the destination address. With the name change to _copy you can add my Reviewed-by: Stefan Schmidt regards Stefan Schmidt