Message ID | 1493705445-5774-4-git-send-email-qdy220091330@gmail.com |
---|---|
State | Changes Requested |
Headers | show |
>From: Michael Qiu <qiudayu@chinac.com> > >When a packet is from DPDK source, and it contains >multiple segments, data_len is not equal to the >packet size. This patch fix this issue. > >Signed-off-by: Michael Qiu <qiudayu@chinac.com> >Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> >Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> >Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> >Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> >--- > lib/dp-packet.h | 23 +++++++++++++---------- > 1 file changed, 13 insertions(+), 10 deletions(-) > >diff --git a/lib/dp-packet.h b/lib/dp-packet.h >index c73ca19..337a600 100644 >--- a/lib/dp-packet.h >+++ b/lib/dp-packet.h >@@ -23,6 +23,7 @@ > #ifdef DPDK_NETDEV > #include <rte_config.h> > #include <rte_mbuf.h> >+#include "rte_ether.h" > #endif > > #include "netdev-dpdk.h" >@@ -415,17 +416,19 @@ dp_packet_size(const struct dp_packet *b) > static inline void > dp_packet_set_size(struct dp_packet *b, uint32_t v) > { >- /* netdev-dpdk does not currently support segmentation; consequently, for >- * all intents and purposes, 'data_len' (16 bit) and 'pkt_len' (32 bit) may >- * be used interchangably. >- * >- * On the datapath, it is expected that the size of packets >- * (and thus 'v') will always be <= UINT16_MAX; this means that there is no >- * loss of accuracy in assigning 'v' to 'data_len'. >+ /* >+ * Assign current segment length. If total length is greater than >+ * max data length in a segment, additional calculation is needed > */ >- b->mbuf.data_len = (uint16_t)v; /* Current seg length. */ >- b->mbuf.pkt_len = v; /* Total length of all segments linked to >- * this segment. */ >+ if (v > (b->mbuf.buf_len - b->mbuf.data_off)) { You could insert '(OVS_UNLIKELY' after the 'if' >+ b->mbuf.data_len = >+ (uint16_t) (b->mbuf.buf_len - b->mbuf.data_off); >+ } else { >+ b->mbuf.data_len = (uint16_t) v; >+ } >+ >+ /* Total length of all segments linked to this segment. */ >+ b->mbuf.pkt_len = v; > } > > static inline uint16_t >-- >1.8.3.1
在 2017年5月3日,下午11:33,Kavanagh, Mark B <mark.b.kavanagh@intel.com> 写道: >> From: Michael Qiu <qiudayu@chinac.com> >> >> When a packet is from DPDK source, and it contains >> multiple segments, data_len is not equal to the >> packet size. This patch fix this issue. >> >> Signed-off-by: Michael Qiu <qiudayu@chinac.com> >> Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> >> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> >> Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> >> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> >> --- >> lib/dp-packet.h | 23 +++++++++++++---------- >> 1 file changed, 13 insertions(+), 10 deletions(-) >> >> diff --git a/lib/dp-packet.h b/lib/dp-packet.h >> index c73ca19..337a600 100644 >> --- a/lib/dp-packet.h >> +++ b/lib/dp-packet.h >> @@ -23,6 +23,7 @@ >> #ifdef DPDK_NETDEV >> #include <rte_config.h> >> #include <rte_mbuf.h> >> +#include "rte_ether.h" >> #endif >> >> #include "netdev-dpdk.h" >> @@ -415,17 +416,19 @@ dp_packet_size(const struct dp_packet *b) >> static inline void >> dp_packet_set_size(struct dp_packet *b, uint32_t v) >> { >> - /* netdev-dpdk does not currently support segmentation; consequently, for >> - * all intents and purposes, 'data_len' (16 bit) and 'pkt_len' (32 bit) may >> - * be used interchangably. >> - * >> - * On the datapath, it is expected that the size of packets >> - * (and thus 'v') will always be <= UINT16_MAX; this means that there is no >> - * loss of accuracy in assigning 'v' to 'data_len'. >> + /* >> + * Assign current segment length. If total length is greater than >> + * max data length in a segment, additional calculation is needed >> */ >> - b->mbuf.data_len = (uint16_t)v; /* Current seg length. */ >> - b->mbuf.pkt_len = v; /* Total length of all segments linked to >> - * this segment. */ >> + if (v > (b->mbuf.buf_len - b->mbuf.data_off)) { > > You could insert '(OVS_UNLIKELY' after the 'if' What I consider is that after enable hw offloading, multiple segments should be unlikely? But anyway, before enable it, unlikely could be right thing. > >> + b->mbuf.data_len = >> + (uint16_t) (b->mbuf.buf_len - b->mbuf.data_off); >> + } else { >> + b->mbuf.data_len = (uint16_t) v; >> + } >> + >> + /* Total length of all segments linked to this segment. */ >> + b->mbuf.pkt_len = v; >> } >> >> static inline uint16_t >> -- >> 1.8.3.1 > > _______________________________________________ > dev mailing list > dev@openvswitch.org > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
On Tue, May 02, 2017 at 02:10:43PM +0800, Michael Qiu wrote: > From: Michael Qiu <qiudayu@chinac.com> > > When a packet is from DPDK source, and it contains > multiple segments, data_len is not equal to the > packet size. This patch fix this issue. > > Signed-off-by: Michael Qiu <qiudayu@chinac.com> > Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> > Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> > Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> > Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Thank you for working to improve OVS support for DPDK. This is a very simple patch. Can you explain the chain of authorship? > @@ -415,17 +416,19 @@ dp_packet_size(const struct dp_packet *b) > static inline void > dp_packet_set_size(struct dp_packet *b, uint32_t v) > { > - /* netdev-dpdk does not currently support segmentation; consequently, for > - * all intents and purposes, 'data_len' (16 bit) and 'pkt_len' (32 bit) may > - * be used interchangably. > - * > - * On the datapath, it is expected that the size of packets > - * (and thus 'v') will always be <= UINT16_MAX; this means that there is no > - * loss of accuracy in assigning 'v' to 'data_len'. > + /* > + * Assign current segment length. If total length is greater than > + * max data length in a segment, additional calculation is needed > */ > - b->mbuf.data_len = (uint16_t)v; /* Current seg length. */ > - b->mbuf.pkt_len = v; /* Total length of all segments linked to > - * this segment. */ > + if (v > (b->mbuf.buf_len - b->mbuf.data_off)) { > + b->mbuf.data_len = > + (uint16_t) (b->mbuf.buf_len - b->mbuf.data_off); > + } else { > + b->mbuf.data_len = (uint16_t) v; > + } I don't see any value in the two casts to uint16_t above. They convert their operands to uint16_t, but this will also happen just as well without the casts since the destination has type uint16_t. Open vSwitch style omits needless casts. Also, I think that the above can be written more simply as: b->mbuf.data_len = MIN(v, b->mbuf.buf_len - b->mbuf.data_off);
> >On Tue, May 02, 2017 at 02:10:43PM +0800, Michael Qiu wrote: >> From: Michael Qiu <qiudayu@chinac.com> >> >> When a packet is from DPDK source, and it contains >> multiple segments, data_len is not equal to the >> packet size. This patch fix this issue. >> >> Signed-off-by: Michael Qiu <qiudayu@chinac.com> >> Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> >> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> >> Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> >> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> > >Thank you for working to improve OVS support for DPDK. > >This is a very simple patch. Can you explain the chain of authorship? Hi Ben, This code originates from an RFC patch, which was authored by the individuals listed: https://mail.openvswitch.org/pipermail/ovs-dev/2016-June/316414.html Hope this clears things up. Thanks, Mark > >> @@ -415,17 +416,19 @@ dp_packet_size(const struct dp_packet *b) >> static inline void >> dp_packet_set_size(struct dp_packet *b, uint32_t v) >> { >> - /* netdev-dpdk does not currently support segmentation; consequently, for >> - * all intents and purposes, 'data_len' (16 bit) and 'pkt_len' (32 bit) may >> - * be used interchangably. >> - * >> - * On the datapath, it is expected that the size of packets >> - * (and thus 'v') will always be <= UINT16_MAX; this means that there is no >> - * loss of accuracy in assigning 'v' to 'data_len'. >> + /* >> + * Assign current segment length. If total length is greater than >> + * max data length in a segment, additional calculation is needed >> */ >> - b->mbuf.data_len = (uint16_t)v; /* Current seg length. */ >> - b->mbuf.pkt_len = v; /* Total length of all segments linked to >> - * this segment. */ >> + if (v > (b->mbuf.buf_len - b->mbuf.data_off)) { >> + b->mbuf.data_len = >> + (uint16_t) (b->mbuf.buf_len - b->mbuf.data_off); >> + } else { >> + b->mbuf.data_len = (uint16_t) v; >> + } > >I don't see any value in the two casts to uint16_t above. They convert >their operands to uint16_t, but this will also happen just as well >without the casts since the destination has type uint16_t. Open vSwitch >style omits needless casts. > >Also, I think that the above can be written more simply as: > b->mbuf.data_len = MIN(v, b->mbuf.buf_len - b->mbuf.data_off); > > >_______________________________________________ >dev mailing list >dev@openvswitch.org >https://mail.openvswitch.org/mailman/listinfo/ovs-dev
On Fri, May 05, 2017 at 08:57:27AM +0000, Kavanagh, Mark B wrote: > > > >On Tue, May 02, 2017 at 02:10:43PM +0800, Michael Qiu wrote: > >> From: Michael Qiu <qiudayu@chinac.com> > >> > >> When a packet is from DPDK source, and it contains > >> multiple segments, data_len is not equal to the > >> packet size. This patch fix this issue. > >> > >> Signed-off-by: Michael Qiu <qiudayu@chinac.com> > >> Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> > >> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> > >> Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> > >> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> > > > >Thank you for working to improve OVS support for DPDK. > > > >This is a very simple patch. Can you explain the chain of authorship? > > Hi Ben, > > This code originates from an RFC patch, which was authored by the individuals listed: https://mail.openvswitch.org/pipermail/ovs-dev/2016-June/316414.html > > Hope this clears things up. OK, if the listed signoffs are co-authors, then Co-authored-by: tags are needed. The RFC had them but Michael's version drops them. Michael, please add back the Co-authored-by: tags. Thanks, Ben.
在 2017/5/5 23:44, Ben Pfaff 写道: > On Fri, May 05, 2017 at 08:57:27AM +0000, Kavanagh, Mark B wrote: >>> On Tue, May 02, 2017 at 02:10:43PM +0800, Michael Qiu wrote: >>>> From: Michael Qiu <qiudayu@chinac.com> >>>> >>>> When a packet is from DPDK source, and it contains >>>> multiple segments, data_len is not equal to the >>>> packet size. This patch fix this issue. >>>> >>>> Signed-off-by: Michael Qiu <qiudayu@chinac.com> >>>> Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> >>>> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> >>>> Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> >>>> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> >>> Thank you for working to improve OVS support for DPDK. >>> >>> This is a very simple patch. Can you explain the chain of authorship? >> Hi Ben, >> >> This code originates from an RFC patch, which was authored by the individuals listed: https://mail.openvswitch.org/pipermail/ovs-dev/2016-June/316414.html >> >> Hope this clears things up. > OK, if the listed signoffs are co-authors, then Co-authored-by: tags are > needed. The RFC had them but Michael's version drops them. Michael, > please add back the Co-authored-by: tags. OK, I will re-add them and send another version. > Thanks, > > Ben.
>From: Michael Qiu [mailto:qdy220091330@gmail.com] >Sent: Wednesday, May 31, 2017 9:58 AM >To: Ben Pfaff <blp@ovn.org>; Kavanagh, Mark B <mark.b.kavanagh@intel.com> >Cc: dev@openvswitch.org; Lal, PrzemyslawX <przemyslawx.lal@intel.com>; Michael Qiu ><qiudayu@chinac.com>; Ksiadz, MarcinX <marcinx.ksiadz@intel.com> >Subject: Re: [ovs-dev] [PATCH 3/5] lib/dp-packet: Fix data_len issue with multi-segments > > > >在 2017/5/5 23:44, Ben Pfaff 写道: >> On Fri, May 05, 2017 at 08:57:27AM +0000, Kavanagh, Mark B wrote: >>>> On Tue, May 02, 2017 at 02:10:43PM +0800, Michael Qiu wrote: >>>>> From: Michael Qiu <qiudayu@chinac.com> >>>>> >>>>> When a packet is from DPDK source, and it contains >>>>> multiple segments, data_len is not equal to the >>>>> packet size. This patch fix this issue. >>>>> >>>>> Signed-off-by: Michael Qiu <qiudayu@chinac.com> >>>>> Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> >>>>> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> >>>>> Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> >>>>> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> >>>> Thank you for working to improve OVS support for DPDK. >>>> >>>> This is a very simple patch. Can you explain the chain of authorship? >>> Hi Ben, >>> >>> This code originates from an RFC patch, which was authored by the individuals listed: >https://mail.openvswitch.org/pipermail/ovs-dev/2016-June/316414.html >>> >>> Hope this clears things up. >> OK, if the listed signoffs are co-authors, then Co-authored-by: tags are >> needed. The RFC had them but Michael's version drops them. Michael, >> please add back the Co-authored-by: tags. > >OK, I will re-add them and send another version. Hi Michael, A number of people have mentioned to me that your patchset doesn't apply to HEAD of master; if you could also rebase in the next version that'd be great. Thanks, Mark > >> Thanks, >> >> Ben.
在 2017/5/31 18:59, Kavanagh, Mark B 写道: >> From: Michael Qiu [mailto:qdy220091330@gmail.com] >> Sent: Wednesday, May 31, 2017 9:58 AM >> To: Ben Pfaff <blp@ovn.org>; Kavanagh, Mark B <mark.b.kavanagh@intel.com> >> Cc: dev@openvswitch.org; Lal, PrzemyslawX <przemyslawx.lal@intel.com>; Michael Qiu >> <qiudayu@chinac.com>; Ksiadz, MarcinX <marcinx.ksiadz@intel.com> >> Subject: Re: [ovs-dev] [PATCH 3/5] lib/dp-packet: Fix data_len issue with multi-segments >> >> >> >> 在 2017/5/5 23:44, Ben Pfaff 写道: >>> On Fri, May 05, 2017 at 08:57:27AM +0000, Kavanagh, Mark B wrote: >>>>> On Tue, May 02, 2017 at 02:10:43PM +0800, Michael Qiu wrote: >>>>>> From: Michael Qiu <qiudayu@chinac.com> >>>>>> >>>>>> When a packet is from DPDK source, and it contains >>>>>> multiple segments, data_len is not equal to the >>>>>> packet size. This patch fix this issue. >>>>>> >>>>>> Signed-off-by: Michael Qiu <qiudayu@chinac.com> >>>>>> Signed-off-by: Marcin Ksiadz <marcinx.ksiadz@intel.com> >>>>>> Signed-off-by: Mark Kavanagh <mark.b.kavanagh@intel.com> >>>>>> Signed-off-by: Przemyslaw Lal <przemyslawx.lal@intel.com> >>>>>> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> >>>>> Thank you for working to improve OVS support for DPDK. >>>>> >>>>> This is a very simple patch. Can you explain the chain of authorship? >>>> Hi Ben, >>>> >>>> This code originates from an RFC patch, which was authored by the individuals listed: >> https://mail.openvswitch.org/pipermail/ovs-dev/2016-June/316414.html >>>> Hope this clears things up. >>> OK, if the listed signoffs are co-authors, then Co-authored-by: tags are >>> needed. The RFC had them but Michael's version drops them. Michael, >>> please add back the Co-authored-by: tags. >> OK, I will re-add them and send another version. > Hi Michael, > > A number of people have mentioned to me that your patchset doesn't apply to HEAD of master; if you could also rebase in the next version that'd be great. Sure, I will rebase it. Thanks, Michael > Thanks, > Mark > > >>> Thanks, >>> >>> Ben. > _______________________________________________ > dev mailing list > dev@openvswitch.org > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
diff --git a/lib/dp-packet.h b/lib/dp-packet.h index c73ca19..337a600 100644 --- a/lib/dp-packet.h +++ b/lib/dp-packet.h @@ -23,6 +23,7 @@ #ifdef DPDK_NETDEV #include <rte_config.h> #include <rte_mbuf.h> +#include "rte_ether.h" #endif #include "netdev-dpdk.h" @@ -415,17 +416,19 @@ dp_packet_size(const struct dp_packet *b) static inline void dp_packet_set_size(struct dp_packet *b, uint32_t v) { - /* netdev-dpdk does not currently support segmentation; consequently, for - * all intents and purposes, 'data_len' (16 bit) and 'pkt_len' (32 bit) may - * be used interchangably. - * - * On the datapath, it is expected that the size of packets - * (and thus 'v') will always be <= UINT16_MAX; this means that there is no - * loss of accuracy in assigning 'v' to 'data_len'. + /* + * Assign current segment length. If total length is greater than + * max data length in a segment, additional calculation is needed */ - b->mbuf.data_len = (uint16_t)v; /* Current seg length. */ - b->mbuf.pkt_len = v; /* Total length of all segments linked to - * this segment. */ + if (v > (b->mbuf.buf_len - b->mbuf.data_off)) { + b->mbuf.data_len = + (uint16_t) (b->mbuf.buf_len - b->mbuf.data_off); + } else { + b->mbuf.data_len = (uint16_t) v; + } + + /* Total length of all segments linked to this segment. */ + b->mbuf.pkt_len = v; } static inline uint16_t