MAINTAINERS | 7 + configure | 33 + docs/ebpf_rss.rst | 133 + ebpf/EbpfElf_to_C.py | 36 + ebpf/Makefile.ebpf | 33 + ebpf/ebpf_rss-stub.c | 40 + ebpf/ebpf_rss.c | 186 ++ ebpf/ebpf_rss.h | 44 + ebpf/meson.build | 1 + ebpf/rss.bpf.c | 505 +++ ebpf/tun_rss_steering.h | 5439 ++++++++++++++++++++++++++++++++ hw/net/vhost_net.c | 2 + hw/net/virtio-net.c | 120 +- include/hw/virtio/virtio-net.h | 4 + include/net/net.h | 2 + meson.build | 11 + net/tap-bsd.c | 5 + net/tap-linux.c | 13 + net/tap-linux.h | 1 + net/tap-solaris.c | 5 + net/tap-stub.c | 5 + net/tap.c | 9 + net/tap_int.h | 1 + net/vhost-vdpa.c | 2 + 24 files changed, 6633 insertions(+), 4 deletions(-) create mode 100644 docs/ebpf_rss.rst create mode 100644 ebpf/EbpfElf_to_C.py create mode 100755 ebpf/Makefile.ebpf create mode 100644 ebpf/ebpf_rss-stub.c create mode 100644 ebpf/ebpf_rss.c create mode 100644 ebpf/ebpf_rss.h create mode 100644 ebpf/meson.build create mode 100644 ebpf/rss.bpf.c create mode 100644 ebpf/tun_rss_steering.h
This set of patches introduces the usage of eBPF for packet steering and RSS hash calculation: * RSS(Receive Side Scaling) is used to distribute network packets to guest virtqueues by calculating packet hash * Additionally adding support for the usage of RSS with vhost The eBPF works on kernels 5.8+ On earlier kerneld it fails to load and the RSS feature is reported only without vhost and implemented in 'in-qemu' software. Implementation notes: Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. Added libbpf dependency and eBPF support. The eBPF program is part of the qemu and presented as an array of BPF ELF file data. The compilation of eBPF is not part of QEMU build and can be done using provided Makefile.ebpf(need to adjust 'linuxhdrs'). Added changes to virtio-net and vhost, primary eBPF RSS is used. 'in-qemu' RSS used in the case of hash population and as a fallback option. For vhost, the hash population feature is not reported to the guest. Please also see the documentation in PATCH 5/5. I am sending those patches as RFC to initiate the discussions and get feedback on the following points: * Fallback when eBPF is not supported by the kernel * Live migration to the kernel that doesn't have eBPF support * Integration with current QEMU build * Additional usage for eBPF for packet filtering Known issues: * hash population not supported by eBPF RSS: 'in-qemu' RSS used as a fallback, also, hash population feature is not reported to guests with vhost. * big-endian BPF support: for now, eBPF isn't supported on big-endian systems. Can be added in future if required. * huge .h file with eBPF binary. The size of .h file containing eBPF binary is currently ~5K lines, because the binary is built with debug information. The binary without debug/BTF info can't be loaded by libbpf. We're looking for possibilities to reduce the size of the .h files. Changes since v1: * using libbpf instead of direct 'bpf' system call. * added libbpf dependency to the configure/meson scripts. * changed python script for eBPF .h file generation. * changed eBPF program - reading L3 proto from ethernet frame. * added TUNSETSTEERINGEBPF define for TUN. * changed the maintainer's info. * added license headers. * refactored code. Andrew (5): net: Added SetSteeringEBPF method for NetClientState. ebpf: Added eBPF RSS program. ebpf: Added eBPF RSS loader. virtio-net: Added eBPF RSS to virtio-net. docs: Added eBPF RSS documentation. MAINTAINERS | 7 + configure | 33 + docs/ebpf_rss.rst | 133 + ebpf/EbpfElf_to_C.py | 36 + ebpf/Makefile.ebpf | 33 + ebpf/ebpf_rss-stub.c | 40 + ebpf/ebpf_rss.c | 186 ++ ebpf/ebpf_rss.h | 44 + ebpf/meson.build | 1 + ebpf/rss.bpf.c | 505 +++ ebpf/tun_rss_steering.h | 5439 ++++++++++++++++++++++++++++++++ hw/net/vhost_net.c | 2 + hw/net/virtio-net.c | 120 +- include/hw/virtio/virtio-net.h | 4 + include/net/net.h | 2 + meson.build | 11 + net/tap-bsd.c | 5 + net/tap-linux.c | 13 + net/tap-linux.h | 1 + net/tap-solaris.c | 5 + net/tap-stub.c | 5 + net/tap.c | 9 + net/tap_int.h | 1 + net/vhost-vdpa.c | 2 + 24 files changed, 6633 insertions(+), 4 deletions(-) create mode 100644 docs/ebpf_rss.rst create mode 100644 ebpf/EbpfElf_to_C.py create mode 100755 ebpf/Makefile.ebpf create mode 100644 ebpf/ebpf_rss-stub.c create mode 100644 ebpf/ebpf_rss.c create mode 100644 ebpf/ebpf_rss.h create mode 100644 ebpf/meson.build create mode 100644 ebpf/rss.bpf.c create mode 100644 ebpf/tun_rss_steering.h -- 2.29.2
On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > This set of patches introduces the usage of eBPF for packet steering > and RSS hash calculation: > * RSS(Receive Side Scaling) is used to distribute network packets to > guest virtqueues by calculating packet hash > * Additionally adding support for the usage of RSS with vhost > > The eBPF works on kernels 5.8+ > On earlier kerneld it fails to load and the RSS feature is reported > only without vhost and implemented in 'in-qemu' software. > > Implementation notes: > Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. > Added libbpf dependency and eBPF support. > The eBPF program is part of the qemu and presented as an array > of BPF ELF file data. > The compilation of eBPF is not part of QEMU build and can be done > using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > Added changes to virtio-net and vhost, primary eBPF RSS is used. > 'in-qemu' RSS used in the case of hash population and as a fallback option. > For vhost, the hash population feature is not reported to the guest. > > Please also see the documentation in PATCH 5/5. > > I am sending those patches as RFC to initiate the discussions and get > feedback on the following points: > * Fallback when eBPF is not supported by the kernel > * Live migration to the kernel that doesn't have eBPF support > * Integration with current QEMU build > * Additional usage for eBPF for packet filtering > > Known issues: > * hash population not supported by eBPF RSS: 'in-qemu' RSS used > as a fallback, also, hash population feature is not reported to guests > with vhost. > * big-endian BPF support: for now, eBPF isn't supported on > big-endian systems. Can be added in future if required. > * huge .h file with eBPF binary. The size of .h file containing > eBPF binary is currently ~5K lines, because the binary is built with debug information. > The binary without debug/BTF info can't be loaded by libbpf. > We're looking for possibilities to reduce the size of the .h files. A question here, is this because the binary file contains DWARF data? If yes, is it a building or loading dependency? If it's latter, maybe we can try to strip them out, anyhow it can't be recognized by kernel. Thanks > > Changes since v1: > * using libbpf instead of direct 'bpf' system call. > * added libbpf dependency to the configure/meson scripts. > * changed python script for eBPF .h file generation. > * changed eBPF program - reading L3 proto from ethernet frame. > * added TUNSETSTEERINGEBPF define for TUN. > * changed the maintainer's info. > * added license headers. > * refactored code. > > Andrew (5): > net: Added SetSteeringEBPF method for NetClientState. > ebpf: Added eBPF RSS program. > ebpf: Added eBPF RSS loader. > virtio-net: Added eBPF RSS to virtio-net. > docs: Added eBPF RSS documentation. > > MAINTAINERS | 7 + > configure | 33 + > docs/ebpf_rss.rst | 133 + > ebpf/EbpfElf_to_C.py | 36 + > ebpf/Makefile.ebpf | 33 + > ebpf/ebpf_rss-stub.c | 40 + > ebpf/ebpf_rss.c | 186 ++ > ebpf/ebpf_rss.h | 44 + > ebpf/meson.build | 1 + > ebpf/rss.bpf.c | 505 +++ > ebpf/tun_rss_steering.h | 5439 ++++++++++++++++++++++++++++++++ > hw/net/vhost_net.c | 2 + > hw/net/virtio-net.c | 120 +- > include/hw/virtio/virtio-net.h | 4 + > include/net/net.h | 2 + > meson.build | 11 + > net/tap-bsd.c | 5 + > net/tap-linux.c | 13 + > net/tap-linux.h | 1 + > net/tap-solaris.c | 5 + > net/tap-stub.c | 5 + > net/tap.c | 9 + > net/tap_int.h | 1 + > net/vhost-vdpa.c | 2 + > 24 files changed, 6633 insertions(+), 4 deletions(-) > create mode 100644 docs/ebpf_rss.rst > create mode 100644 ebpf/EbpfElf_to_C.py > create mode 100755 ebpf/Makefile.ebpf > create mode 100644 ebpf/ebpf_rss-stub.c > create mode 100644 ebpf/ebpf_rss.c > create mode 100644 ebpf/ebpf_rss.h > create mode 100644 ebpf/meson.build > create mode 100644 ebpf/rss.bpf.c > create mode 100644 ebpf/tun_rss_steering.h >
On Mon, Nov 23, 2020 at 8:08 AM Jason Wang <jasowang@redhat.com> wrote: > > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > > This set of patches introduces the usage of eBPF for packet steering > > and RSS hash calculation: > > * RSS(Receive Side Scaling) is used to distribute network packets to > > guest virtqueues by calculating packet hash > > * Additionally adding support for the usage of RSS with vhost > > > > The eBPF works on kernels 5.8+ > > On earlier kerneld it fails to load and the RSS feature is reported > > only without vhost and implemented in 'in-qemu' software. > > > > Implementation notes: > > Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. > > Added libbpf dependency and eBPF support. > > The eBPF program is part of the qemu and presented as an array > > of BPF ELF file data. > > The compilation of eBPF is not part of QEMU build and can be done > > using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > > Added changes to virtio-net and vhost, primary eBPF RSS is used. > > 'in-qemu' RSS used in the case of hash population and as a fallback > option. > > For vhost, the hash population feature is not reported to the guest. > > > > Please also see the documentation in PATCH 5/5. > > > > I am sending those patches as RFC to initiate the discussions and get > > feedback on the following points: > > * Fallback when eBPF is not supported by the kernel > > * Live migration to the kernel that doesn't have eBPF support > > * Integration with current QEMU build > > * Additional usage for eBPF for packet filtering > > > > Known issues: > > * hash population not supported by eBPF RSS: 'in-qemu' RSS used > > as a fallback, also, hash population feature is not reported to guests > > with vhost. > > * big-endian BPF support: for now, eBPF isn't supported on > > big-endian systems. Can be added in future if required. > > * huge .h file with eBPF binary. The size of .h file containing > > eBPF binary is currently ~5K lines, because the binary is built with > debug information. > > The binary without debug/BTF info can't be loaded by libbpf. > > We're looking for possibilities to reduce the size of the .h files. > > > A question here, is this because the binary file contains DWARF data? If > yes, is it a building or loading dependency? If it's latter, maybe we > can try to strip them out, anyhow it can't be recognized by kernel. > > Thanks > > After some experiments we can see that stripping of debug sections reduces the size of ELF from ~45K to ~20K (we tried to strip more but the libbpf fails to load it, libbpf needs BTF and symbols) So I suggest to reevaluate the necessity of libbpf. For this specific BPF it does not present advantage and we hardly can create some reusable code related to libbpf, i.e. any further BPF will need its own libbpf wrapper. The BTF is really good feature and in case some later BPF will need an access to kernel structures it will use libbpf loader. What you think about it? > > > > > Changes since v1: > > * using libbpf instead of direct 'bpf' system call. > > * added libbpf dependency to the configure/meson scripts. > > * changed python script for eBPF .h file generation. > > * changed eBPF program - reading L3 proto from ethernet frame. > > * added TUNSETSTEERINGEBPF define for TUN. > > * changed the maintainer's info. > > * added license headers. > > * refactored code. > > > > Andrew (5): > > net: Added SetSteeringEBPF method for NetClientState. > > ebpf: Added eBPF RSS program. > > ebpf: Added eBPF RSS loader. > > virtio-net: Added eBPF RSS to virtio-net. > > docs: Added eBPF RSS documentation. > > > > MAINTAINERS | 7 + > > configure | 33 + > > docs/ebpf_rss.rst | 133 + > > ebpf/EbpfElf_to_C.py | 36 + > > ebpf/Makefile.ebpf | 33 + > > ebpf/ebpf_rss-stub.c | 40 + > > ebpf/ebpf_rss.c | 186 ++ > > ebpf/ebpf_rss.h | 44 + > > ebpf/meson.build | 1 + > > ebpf/rss.bpf.c | 505 +++ > > ebpf/tun_rss_steering.h | 5439 ++++++++++++++++++++++++++++++++ > > hw/net/vhost_net.c | 2 + > > hw/net/virtio-net.c | 120 +- > > include/hw/virtio/virtio-net.h | 4 + > > include/net/net.h | 2 + > > meson.build | 11 + > > net/tap-bsd.c | 5 + > > net/tap-linux.c | 13 + > > net/tap-linux.h | 1 + > > net/tap-solaris.c | 5 + > > net/tap-stub.c | 5 + > > net/tap.c | 9 + > > net/tap_int.h | 1 + > > net/vhost-vdpa.c | 2 + > > 24 files changed, 6633 insertions(+), 4 deletions(-) > > create mode 100644 docs/ebpf_rss.rst > > create mode 100644 ebpf/EbpfElf_to_C.py > > create mode 100755 ebpf/Makefile.ebpf > > create mode 100644 ebpf/ebpf_rss-stub.c > > create mode 100644 ebpf/ebpf_rss.c > > create mode 100644 ebpf/ebpf_rss.h > > create mode 100644 ebpf/meson.build > > create mode 100644 ebpf/rss.bpf.c > > create mode 100644 ebpf/tun_rss_steering.h > > > >
On 2020/11/26 下午8:52, Yuri Benditovich wrote: > > > On Mon, Nov 23, 2020 at 8:08 AM Jason Wang <jasowang@redhat.com > <mailto:jasowang@redhat.com>> wrote: > > > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > > This set of patches introduces the usage of eBPF for packet steering > > and RSS hash calculation: > > * RSS(Receive Side Scaling) is used to distribute network packets to > > guest virtqueues by calculating packet hash > > * Additionally adding support for the usage of RSS with vhost > > > > The eBPF works on kernels 5.8+ > > On earlier kerneld it fails to load and the RSS feature is reported > > only without vhost and implemented in 'in-qemu' software. > > > > Implementation notes: > > Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. > > Added libbpf dependency and eBPF support. > > The eBPF program is part of the qemu and presented as an array > > of BPF ELF file data. > > The compilation of eBPF is not part of QEMU build and can be done > > using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > > Added changes to virtio-net and vhost, primary eBPF RSS is used. > > 'in-qemu' RSS used in the case of hash population and as a > fallback option. > > For vhost, the hash population feature is not reported to the guest. > > > > Please also see the documentation in PATCH 5/5. > > > > I am sending those patches as RFC to initiate the discussions > and get > > feedback on the following points: > > * Fallback when eBPF is not supported by the kernel > > * Live migration to the kernel that doesn't have eBPF support > > * Integration with current QEMU build > > * Additional usage for eBPF for packet filtering > > > > Known issues: > > * hash population not supported by eBPF RSS: 'in-qemu' RSS used > > as a fallback, also, hash population feature is not reported to > guests > > with vhost. > > * big-endian BPF support: for now, eBPF isn't supported on > > big-endian systems. Can be added in future if required. > > * huge .h file with eBPF binary. The size of .h file containing > > eBPF binary is currently ~5K lines, because the binary is built > with debug information. > > The binary without debug/BTF info can't be loaded by libbpf. > > We're looking for possibilities to reduce the size of the .h files. > > > A question here, is this because the binary file contains DWARF > data? If > yes, is it a building or loading dependency? If it's latter, maybe we > can try to strip them out, anyhow it can't be recognized by kernel. > > Thanks > > > After some experiments we can see that stripping of debug sections > reduces the size of > ELF from ~45K to ~20K (we tried to strip more but the libbpf fails to > load it, libbpf needs BTF and symbols) > So I suggest to reevaluate the necessity of libbpf. > For this specific BPF it does not present advantage and we hardly can > create some reusable code > related to libbpf, i.e. any further BPF will need its own libbpf wrapper. > The BTF is really good feature and in case some later BPF will need an > access to kernel > structures it will use libbpf loader. > What you think about it? If we can find a way to use BTF without libbpf, it should be acceptable. Thanks > > > > > Changes since v1: > > * using libbpf instead of direct 'bpf' system call. > > * added libbpf dependency to the configure/meson scripts. > > * changed python script for eBPF .h file generation. > > * changed eBPF program - reading L3 proto from ethernet frame. > > * added TUNSETSTEERINGEBPF define for TUN. > > * changed the maintainer's info. > > * added license headers. > > * refactored code. > > > > Andrew (5): > > net: Added SetSteeringEBPF method for NetClientState. > > ebpf: Added eBPF RSS program. > > ebpf: Added eBPF RSS loader. > > virtio-net: Added eBPF RSS to virtio-net. > > docs: Added eBPF RSS documentation. > > > > MAINTAINERS | 7 + > > configure | 33 + > > docs/ebpf_rss.rst | 133 + > > ebpf/EbpfElf_to_C.py | 36 + > > ebpf/Makefile.ebpf | 33 + > > ebpf/ebpf_rss-stub.c | 40 + > > ebpf/ebpf_rss.c | 186 ++ > > ebpf/ebpf_rss.h | 44 + > > ebpf/meson.build | 1 + > > ebpf/rss.bpf.c | 505 +++ > > ebpf/tun_rss_steering.h | 5439 > ++++++++++++++++++++++++++++++++ > > hw/net/vhost_net.c | 2 + > > hw/net/virtio-net.c | 120 +- > > include/hw/virtio/virtio-net.h | 4 + > > include/net/net.h | 2 + > > meson.build | 11 + > > net/tap-bsd.c | 5 + > > net/tap-linux.c | 13 + > > net/tap-linux.h | 1 + > > net/tap-solaris.c | 5 + > > net/tap-stub.c | 5 + > > net/tap.c | 9 + > > net/tap_int.h | 1 + > > net/vhost-vdpa.c | 2 + > > 24 files changed, 6633 insertions(+), 4 deletions(-) > > create mode 100644 docs/ebpf_rss.rst > > create mode 100644 ebpf/EbpfElf_to_C.py > > create mode 100755 ebpf/Makefile.ebpf > > create mode 100644 ebpf/ebpf_rss-stub.c > > create mode 100644 ebpf/ebpf_rss.c > > create mode 100644 ebpf/ebpf_rss.h > > create mode 100644 ebpf/meson.build > > create mode 100644 ebpf/rss.bpf.c > > create mode 100644 ebpf/tun_rss_steering.h > > >
On Fri, Nov 27, 2020 at 6:36 AM Jason Wang <jasowang@redhat.com> wrote: > > On 2020/11/26 下午8:52, Yuri Benditovich wrote: > > > > > > On Mon, Nov 23, 2020 at 8:08 AM Jason Wang <jasowang@redhat.com > > <mailto:jasowang@redhat.com>> wrote: > > > > > > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > > > This set of patches introduces the usage of eBPF for packet > steering > > > and RSS hash calculation: > > > * RSS(Receive Side Scaling) is used to distribute network packets > to > > > guest virtqueues by calculating packet hash > > > * Additionally adding support for the usage of RSS with vhost > > > > > > The eBPF works on kernels 5.8+ > > > On earlier kerneld it fails to load and the RSS feature is reported > > > only without vhost and implemented in 'in-qemu' software. > > > > > > Implementation notes: > > > Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF > program. > > > Added libbpf dependency and eBPF support. > > > The eBPF program is part of the qemu and presented as an array > > > of BPF ELF file data. > > > The compilation of eBPF is not part of QEMU build and can be done > > > using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > > > Added changes to virtio-net and vhost, primary eBPF RSS is used. > > > 'in-qemu' RSS used in the case of hash population and as a > > fallback option. > > > For vhost, the hash population feature is not reported to the > guest. > > > > > > Please also see the documentation in PATCH 5/5. > > > > > > I am sending those patches as RFC to initiate the discussions > > and get > > > feedback on the following points: > > > * Fallback when eBPF is not supported by the kernel > > > * Live migration to the kernel that doesn't have eBPF support > > > * Integration with current QEMU build > > > * Additional usage for eBPF for packet filtering > > > > > > Known issues: > > > * hash population not supported by eBPF RSS: 'in-qemu' RSS used > > > as a fallback, also, hash population feature is not reported to > > guests > > > with vhost. > > > * big-endian BPF support: for now, eBPF isn't supported on > > > big-endian systems. Can be added in future if required. > > > * huge .h file with eBPF binary. The size of .h file containing > > > eBPF binary is currently ~5K lines, because the binary is built > > with debug information. > > > The binary without debug/BTF info can't be loaded by libbpf. > > > We're looking for possibilities to reduce the size of the .h files. > > > > > > A question here, is this because the binary file contains DWARF > > data? If > > yes, is it a building or loading dependency? If it's latter, maybe we > > can try to strip them out, anyhow it can't be recognized by kernel. > > > > Thanks > > > > > > After some experiments we can see that stripping of debug sections > > reduces the size of > > ELF from ~45K to ~20K (we tried to strip more but the libbpf fails to > > load it, libbpf needs BTF and symbols) > > So I suggest to reevaluate the necessity of libbpf. > > For this specific BPF it does not present advantage and we hardly can > > create some reusable code > > related to libbpf, i.e. any further BPF will need its own libbpf wrapper. > > The BTF is really good feature and in case some later BPF will need an > > access to kernel > > structures it will use libbpf loader. > > What you think about it? > > > If we can find a way to use BTF without libbpf, it should be acceptable. > > But the point is that the RSS BPF does not need the BTF as it does not use any kernel structures. When we have, for example, filter BPF that will need the BTF - we'll use libbpf for it. Anyway we do not have here any infrastructural code related to libbpf, > Thanks > > > > > > > > > > Changes since v1: > > > * using libbpf instead of direct 'bpf' system call. > > > * added libbpf dependency to the configure/meson scripts. > > > * changed python script for eBPF .h file generation. > > > * changed eBPF program - reading L3 proto from ethernet frame. > > > * added TUNSETSTEERINGEBPF define for TUN. > > > * changed the maintainer's info. > > > * added license headers. > > > * refactored code. > > > > > > Andrew (5): > > > net: Added SetSteeringEBPF method for NetClientState. > > > ebpf: Added eBPF RSS program. > > > ebpf: Added eBPF RSS loader. > > > virtio-net: Added eBPF RSS to virtio-net. > > > docs: Added eBPF RSS documentation. > > > > > > MAINTAINERS | 7 + > > > configure | 33 + > > > docs/ebpf_rss.rst | 133 + > > > ebpf/EbpfElf_to_C.py | 36 + > > > ebpf/Makefile.ebpf | 33 + > > > ebpf/ebpf_rss-stub.c | 40 + > > > ebpf/ebpf_rss.c | 186 ++ > > > ebpf/ebpf_rss.h | 44 + > > > ebpf/meson.build | 1 + > > > ebpf/rss.bpf.c | 505 +++ > > > ebpf/tun_rss_steering.h | 5439 > > ++++++++++++++++++++++++++++++++ > > > hw/net/vhost_net.c | 2 + > > > hw/net/virtio-net.c | 120 +- > > > include/hw/virtio/virtio-net.h | 4 + > > > include/net/net.h | 2 + > > > meson.build | 11 + > > > net/tap-bsd.c | 5 + > > > net/tap-linux.c | 13 + > > > net/tap-linux.h | 1 + > > > net/tap-solaris.c | 5 + > > > net/tap-stub.c | 5 + > > > net/tap.c | 9 + > > > net/tap_int.h | 1 + > > > net/vhost-vdpa.c | 2 + > > > 24 files changed, 6633 insertions(+), 4 deletions(-) > > > create mode 100644 docs/ebpf_rss.rst > > > create mode 100644 ebpf/EbpfElf_to_C.py > > > create mode 100755 ebpf/Makefile.ebpf > > > create mode 100644 ebpf/ebpf_rss-stub.c > > > create mode 100644 ebpf/ebpf_rss.c > > > create mode 100644 ebpf/ebpf_rss.h > > > create mode 100644 ebpf/meson.build > > > create mode 100644 ebpf/rss.bpf.c > > > create mode 100644 ebpf/tun_rss_steering.h > > > > > > >
On 2020/11/27 下午2:06, Yuri Benditovich wrote: > > > > After some experiments we can see that stripping of debug sections > > reduces the size of > > ELF from ~45K to ~20K (we tried to strip more but the libbpf > fails to > > load it, libbpf needs BTF and symbols) > > So I suggest to reevaluate the necessity of libbpf. > > For this specific BPF it does not present advantage and we > hardly can > > create some reusable code > > related to libbpf, i.e. any further BPF will need its own libbpf > wrapper. > > The BTF is really good feature and in case some later BPF will > need an > > access to kernel > > structures it will use libbpf loader. > > What you think about it? > > > If we can find a way to use BTF without libbpf, it should be > acceptable. > > But the point is that the RSS BPF does not need the BTF as it does not > use any kernel structures. Kinds of, it tries to access skb. But yes, it doesn't access any metadata of skb. > When we have, for example, filter BPF that will need the BTF - we'll > use libbpf for it. > Anyway we do not have here any infrastructural code related to libbpf, Right, so I think we can probably start from a non BTF version without libbpf. And adding other features on top. Thanks
On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > This set of patches introduces the usage of eBPF for packet steering > and RSS hash calculation: > * RSS(Receive Side Scaling) is used to distribute network packets to > guest virtqueues by calculating packet hash > * Additionally adding support for the usage of RSS with vhost > > The eBPF works on kernels 5.8+ > On earlier kerneld it fails to load and the RSS feature is reported > only without vhost and implemented in 'in-qemu' software. > > Implementation notes: > Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. > Added libbpf dependency and eBPF support. > The eBPF program is part of the qemu and presented as an array > of BPF ELF file data. > The compilation of eBPF is not part of QEMU build and can be done > using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > Added changes to virtio-net and vhost, primary eBPF RSS is used. > 'in-qemu' RSS used in the case of hash population and as a fallback option. > For vhost, the hash population feature is not reported to the guest. > > Please also see the documentation in PATCH 5/5. > > I am sending those patches as RFC to initiate the discussions and get > feedback on the following points: > * Fallback when eBPF is not supported by the kernel > * Live migration to the kernel that doesn't have eBPF support > * Integration with current QEMU build > * Additional usage for eBPF for packet filtering > > Known issues: > * hash population not supported by eBPF RSS: 'in-qemu' RSS used > as a fallback, also, hash population feature is not reported to guests > with vhost. > * big-endian BPF support: for now, eBPF isn't supported on > big-endian systems. Can be added in future if required. > * huge .h file with eBPF binary. The size of .h file containing > eBPF binary is currently ~5K lines, because the binary is built with debug information. > The binary without debug/BTF info can't be loaded by libbpf. > We're looking for possibilities to reduce the size of the .h files. Adding Toke for sharing more idea from eBPF side. We had some discussion on the eBPF issues: 1) Whether or not to use libbpf. Toke strongly suggest to use libbpf 2) Whether or not to use BTF. Toke confirmed that if we don't access any skb metadata, BTF is not strictly required for CO-RE. But it might still useful for e.g debugging. 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed that we can strip debug symbols, but Yuri found some sections can't be stripped, we can keep discussing here. Toke, feel free to correct me if I was wrong. Thanks > > Changes since v1: > * using libbpf instead of direct 'bpf' system call. > * added libbpf dependency to the configure/meson scripts. > * changed python script for eBPF .h file generation. > * changed eBPF program - reading L3 proto from ethernet frame. > * added TUNSETSTEERINGEBPF define for TUN. > * changed the maintainer's info. > * added license headers. > * refactored code. > > Andrew (5): > net: Added SetSteeringEBPF method for NetClientState. > ebpf: Added eBPF RSS program. > ebpf: Added eBPF RSS loader. > virtio-net: Added eBPF RSS to virtio-net. > docs: Added eBPF RSS documentation. > > MAINTAINERS | 7 + > configure | 33 + > docs/ebpf_rss.rst | 133 + > ebpf/EbpfElf_to_C.py | 36 + > ebpf/Makefile.ebpf | 33 + > ebpf/ebpf_rss-stub.c | 40 + > ebpf/ebpf_rss.c | 186 ++ > ebpf/ebpf_rss.h | 44 + > ebpf/meson.build | 1 + > ebpf/rss.bpf.c | 505 +++ > ebpf/tun_rss_steering.h | 5439 ++++++++++++++++++++++++++++++++ > hw/net/vhost_net.c | 2 + > hw/net/virtio-net.c | 120 +- > include/hw/virtio/virtio-net.h | 4 + > include/net/net.h | 2 + > meson.build | 11 + > net/tap-bsd.c | 5 + > net/tap-linux.c | 13 + > net/tap-linux.h | 1 + > net/tap-solaris.c | 5 + > net/tap-stub.c | 5 + > net/tap.c | 9 + > net/tap_int.h | 1 + > net/vhost-vdpa.c | 2 + > 24 files changed, 6633 insertions(+), 4 deletions(-) > create mode 100644 docs/ebpf_rss.rst > create mode 100644 ebpf/EbpfElf_to_C.py > create mode 100755 ebpf/Makefile.ebpf > create mode 100644 ebpf/ebpf_rss-stub.c > create mode 100644 ebpf/ebpf_rss.c > create mode 100644 ebpf/ebpf_rss.h > create mode 100644 ebpf/meson.build > create mode 100644 ebpf/rss.bpf.c > create mode 100644 ebpf/tun_rss_steering.h >
Jason Wang <jasowang@redhat.com> writes: > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: >> This set of patches introduces the usage of eBPF for packet steering >> and RSS hash calculation: >> * RSS(Receive Side Scaling) is used to distribute network packets to >> guest virtqueues by calculating packet hash >> * Additionally adding support for the usage of RSS with vhost >> >> The eBPF works on kernels 5.8+ >> On earlier kerneld it fails to load and the RSS feature is reported >> only without vhost and implemented in 'in-qemu' software. >> >> Implementation notes: >> Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. >> Added libbpf dependency and eBPF support. >> The eBPF program is part of the qemu and presented as an array >> of BPF ELF file data. >> The compilation of eBPF is not part of QEMU build and can be done >> using provided Makefile.ebpf(need to adjust 'linuxhdrs'). >> Added changes to virtio-net and vhost, primary eBPF RSS is used. >> 'in-qemu' RSS used in the case of hash population and as a fallback option. >> For vhost, the hash population feature is not reported to the guest. >> >> Please also see the documentation in PATCH 5/5. >> >> I am sending those patches as RFC to initiate the discussions and get >> feedback on the following points: >> * Fallback when eBPF is not supported by the kernel >> * Live migration to the kernel that doesn't have eBPF support >> * Integration with current QEMU build >> * Additional usage for eBPF for packet filtering >> >> Known issues: >> * hash population not supported by eBPF RSS: 'in-qemu' RSS used >> as a fallback, also, hash population feature is not reported to guests >> with vhost. >> * big-endian BPF support: for now, eBPF isn't supported on >> big-endian systems. Can be added in future if required. >> * huge .h file with eBPF binary. The size of .h file containing >> eBPF binary is currently ~5K lines, because the binary is built with debug information. >> The binary without debug/BTF info can't be loaded by libbpf. >> We're looking for possibilities to reduce the size of the .h files. > > > Adding Toke for sharing more idea from eBPF side. > > We had some discussion on the eBPF issues: > > 1) Whether or not to use libbpf. Toke strongly suggest to use libbpf > 2) Whether or not to use BTF. Toke confirmed that if we don't access any > skb metadata, BTF is not strictly required for CO-RE. But it might still > useful for e.g debugging. > 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed that we > can strip debug symbols, but Yuri found some sections can't be stripped, > we can keep discussing here. I just tried simply running 'strip' on a sample trivial XDP program, which brought its size down from ~5k to ~1k and preserved the BTF information without me having to do anything. As a side note, though, instead of embedding the BPF program into a .h, you could simply ship it as a .o and load it from the file system. We do that with xdp-tools and install the bpf object files into /usr/$LIB/bpf/. -Toke
On Wed, Dec 2, 2020 at 4:18 PM Toke Høiland-Jørgensen <toke@redhat.com> wrote: > Jason Wang <jasowang@redhat.com> writes: > > > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > >> This set of patches introduces the usage of eBPF for packet steering > >> and RSS hash calculation: > >> * RSS(Receive Side Scaling) is used to distribute network packets to > >> guest virtqueues by calculating packet hash > >> * Additionally adding support for the usage of RSS with vhost > >> > >> The eBPF works on kernels 5.8+ > >> On earlier kerneld it fails to load and the RSS feature is reported > >> only without vhost and implemented in 'in-qemu' software. > >> > >> Implementation notes: > >> Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. > >> Added libbpf dependency and eBPF support. > >> The eBPF program is part of the qemu and presented as an array > >> of BPF ELF file data. > >> The compilation of eBPF is not part of QEMU build and can be done > >> using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > >> Added changes to virtio-net and vhost, primary eBPF RSS is used. > >> 'in-qemu' RSS used in the case of hash population and as a fallback > option. > >> For vhost, the hash population feature is not reported to the guest. > >> > >> Please also see the documentation in PATCH 5/5. > >> > >> I am sending those patches as RFC to initiate the discussions and get > >> feedback on the following points: > >> * Fallback when eBPF is not supported by the kernel > >> * Live migration to the kernel that doesn't have eBPF support > >> * Integration with current QEMU build > >> * Additional usage for eBPF for packet filtering > >> > >> Known issues: > >> * hash population not supported by eBPF RSS: 'in-qemu' RSS used > >> as a fallback, also, hash population feature is not reported to guests > >> with vhost. > >> * big-endian BPF support: for now, eBPF isn't supported on > >> big-endian systems. Can be added in future if required. > >> * huge .h file with eBPF binary. The size of .h file containing > >> eBPF binary is currently ~5K lines, because the binary is built with > debug information. > >> The binary without debug/BTF info can't be loaded by libbpf. > >> We're looking for possibilities to reduce the size of the .h files. > > > > > > Adding Toke for sharing more idea from eBPF side. > > > > We had some discussion on the eBPF issues: > > > > 1) Whether or not to use libbpf. Toke strongly suggest to use libbpf > > 2) Whether or not to use BTF. Toke confirmed that if we don't access any > > skb metadata, BTF is not strictly required for CO-RE. But it might still > > useful for e.g debugging. > > 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed that we > > can strip debug symbols, but Yuri found some sections can't be stripped, > > we can keep discussing here. > > I just tried simply running 'strip' on a sample trivial XDP program, > which brought its size down from ~5k to ~1k and preserved the BTF > information without me having to do anything. > With our eBPF code the numbers are slightly different: The code size without BTF: 7.5K (built without '-g') Built with '-g': 45K Stripped: 19K The difference between 7.5 and 19K still seems significant, especially when we do not use any kernel structures and do not need these BTF sections This is only reason to prefer non-libbpf option for this specific eBPF > > As a side note, though, instead of embedding the BPF program into a .h, > you could simply ship it as a .o and load it from the file system. We do > that with xdp-tools and install the bpf object files into /usr/$LIB/bpf/. > Yes, we've discussed this option and decided to go with embedding the BPF https://lists.gnu.org/archive/html/qemu-devel/2020-11/msg02157.html > -Toke > >
Yuri Benditovich <yuri.benditovich@daynix.com> writes: > On Wed, Dec 2, 2020 at 4:18 PM Toke Høiland-Jørgensen <toke@redhat.com> > wrote: > >> Jason Wang <jasowang@redhat.com> writes: >> >> > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: >> >> This set of patches introduces the usage of eBPF for packet steering >> >> and RSS hash calculation: >> >> * RSS(Receive Side Scaling) is used to distribute network packets to >> >> guest virtqueues by calculating packet hash >> >> * Additionally adding support for the usage of RSS with vhost >> >> >> >> The eBPF works on kernels 5.8+ >> >> On earlier kerneld it fails to load and the RSS feature is reported >> >> only without vhost and implemented in 'in-qemu' software. >> >> >> >> Implementation notes: >> >> Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. >> >> Added libbpf dependency and eBPF support. >> >> The eBPF program is part of the qemu and presented as an array >> >> of BPF ELF file data. >> >> The compilation of eBPF is not part of QEMU build and can be done >> >> using provided Makefile.ebpf(need to adjust 'linuxhdrs'). >> >> Added changes to virtio-net and vhost, primary eBPF RSS is used. >> >> 'in-qemu' RSS used in the case of hash population and as a fallback >> option. >> >> For vhost, the hash population feature is not reported to the guest. >> >> >> >> Please also see the documentation in PATCH 5/5. >> >> >> >> I am sending those patches as RFC to initiate the discussions and get >> >> feedback on the following points: >> >> * Fallback when eBPF is not supported by the kernel >> >> * Live migration to the kernel that doesn't have eBPF support >> >> * Integration with current QEMU build >> >> * Additional usage for eBPF for packet filtering >> >> >> >> Known issues: >> >> * hash population not supported by eBPF RSS: 'in-qemu' RSS used >> >> as a fallback, also, hash population feature is not reported to guests >> >> with vhost. >> >> * big-endian BPF support: for now, eBPF isn't supported on >> >> big-endian systems. Can be added in future if required. >> >> * huge .h file with eBPF binary. The size of .h file containing >> >> eBPF binary is currently ~5K lines, because the binary is built with >> debug information. >> >> The binary without debug/BTF info can't be loaded by libbpf. >> >> We're looking for possibilities to reduce the size of the .h files. >> > >> > >> > Adding Toke for sharing more idea from eBPF side. >> > >> > We had some discussion on the eBPF issues: >> > >> > 1) Whether or not to use libbpf. Toke strongly suggest to use libbpf >> > 2) Whether or not to use BTF. Toke confirmed that if we don't access any >> > skb metadata, BTF is not strictly required for CO-RE. But it might still >> > useful for e.g debugging. >> > 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed that we >> > can strip debug symbols, but Yuri found some sections can't be stripped, >> > we can keep discussing here. >> >> I just tried simply running 'strip' on a sample trivial XDP program, >> which brought its size down from ~5k to ~1k and preserved the BTF >> information without me having to do anything. >> > > With our eBPF code the numbers are slightly different: > The code size without BTF: 7.5K (built without '-g') > Built with '-g': 45K > Stripped: 19K > The difference between 7.5 and 19K still seems significant, especially when > we do not use any kernel structures and do not need these BTF sections That does seem like a lot of BTF information. Did you confirm (with objdump) that it's the .BTF* sections that take up these extra 12k? Do you have some really complicated data structures in the file or something? Got a link to the source somewhere that isn't a web mailing list archive? :) In any case, while I do think it smells a little of premature optimisation, you can of course strip the BTF information until you need it. Having it around makes debugging easier (bpftool will expand your map structures for you when dumping maps, and that sort of thing), but it's not really essential if you don't need CO-RE. > This is only reason to prefer non-libbpf option for this specific eBPF You can still use libbpf without BTF. It's using BTF without libbpf that tends to not work so well... >> As a side note, though, instead of embedding the BPF program into a .h, >> you could simply ship it as a .o and load it from the file system. We do >> that with xdp-tools and install the bpf object files into /usr/$LIB/bpf/. >> > > Yes, we've discussed this option and decided to go with embedding the BPF > https://lists.gnu.org/archive/html/qemu-devel/2020-11/msg02157.html Right, okay. I'll note, though, that if your concern is that the BPF code should always match the rest of the code base, omitting the compilation if there's no Clang present seems like it could lead to problems :) Also, if you do go the embedded-bytecode route, you may want to have a look at the upstream 'skeleton' concept. It takes a BPF object file and automatically generates a header file that gives you direct access to maps, programs and global data in C. There are some examples in selftests/bpf on how to use it, but you basically just run 'bpftool gen skeleton mybpf.o'. -Toke
On Fri, Dec 4, 2020 at 12:09 PM Toke Høiland-Jørgensen <toke@redhat.com> wrote: > Yuri Benditovich <yuri.benditovich@daynix.com> writes: > > > On Wed, Dec 2, 2020 at 4:18 PM Toke Høiland-Jørgensen <toke@redhat.com> > > wrote: > > > >> Jason Wang <jasowang@redhat.com> writes: > >> > >> > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > >> >> This set of patches introduces the usage of eBPF for packet steering > >> >> and RSS hash calculation: > >> >> * RSS(Receive Side Scaling) is used to distribute network packets to > >> >> guest virtqueues by calculating packet hash > >> >> * Additionally adding support for the usage of RSS with vhost > >> >> > >> >> The eBPF works on kernels 5.8+ > >> >> On earlier kerneld it fails to load and the RSS feature is reported > >> >> only without vhost and implemented in 'in-qemu' software. > >> >> > >> >> Implementation notes: > >> >> Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. > >> >> Added libbpf dependency and eBPF support. > >> >> The eBPF program is part of the qemu and presented as an array > >> >> of BPF ELF file data. > >> >> The compilation of eBPF is not part of QEMU build and can be done > >> >> using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > >> >> Added changes to virtio-net and vhost, primary eBPF RSS is used. > >> >> 'in-qemu' RSS used in the case of hash population and as a fallback > >> option. > >> >> For vhost, the hash population feature is not reported to the guest. > >> >> > >> >> Please also see the documentation in PATCH 5/5. > >> >> > >> >> I am sending those patches as RFC to initiate the discussions and get > >> >> feedback on the following points: > >> >> * Fallback when eBPF is not supported by the kernel > >> >> * Live migration to the kernel that doesn't have eBPF support > >> >> * Integration with current QEMU build > >> >> * Additional usage for eBPF for packet filtering > >> >> > >> >> Known issues: > >> >> * hash population not supported by eBPF RSS: 'in-qemu' RSS used > >> >> as a fallback, also, hash population feature is not reported to > guests > >> >> with vhost. > >> >> * big-endian BPF support: for now, eBPF isn't supported on > >> >> big-endian systems. Can be added in future if required. > >> >> * huge .h file with eBPF binary. The size of .h file containing > >> >> eBPF binary is currently ~5K lines, because the binary is built with > >> debug information. > >> >> The binary without debug/BTF info can't be loaded by libbpf. > >> >> We're looking for possibilities to reduce the size of the .h files. > >> > > >> > > >> > Adding Toke for sharing more idea from eBPF side. > >> > > >> > We had some discussion on the eBPF issues: > >> > > >> > 1) Whether or not to use libbpf. Toke strongly suggest to use libbpf > >> > 2) Whether or not to use BTF. Toke confirmed that if we don't access > any > >> > skb metadata, BTF is not strictly required for CO-RE. But it might > still > >> > useful for e.g debugging. > >> > 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed that > we > >> > can strip debug symbols, but Yuri found some sections can't be > stripped, > >> > we can keep discussing here. > >> > >> I just tried simply running 'strip' on a sample trivial XDP program, > >> which brought its size down from ~5k to ~1k and preserved the BTF > >> information without me having to do anything. > >> > > > > With our eBPF code the numbers are slightly different: > > The code size without BTF: 7.5K (built without '-g') > > Built with '-g': 45K > > Stripped: 19K > > The difference between 7.5 and 19K still seems significant, especially > when > > we do not use any kernel structures and do not need these BTF sections > > That does seem like a lot of BTF information. Did you confirm (with > objdump) that it's the .BTF* sections that take up these extra 12k? Do > you have some really complicated data structures in the file or > something? Got a link to the source somewhere that isn't a web mailing > list archive? :) > > Looks like the extra size is related to BTF: there are 4 BTF sections that take 12.5K [ 7] .BTF PROGBITS 0000000000000000 00144c 00175d 00 0 0 1 [ 8] .rel.BTF REL 0000000000000000 002bb0 000040 10 14 7 8 [ 9] .BTF.ext PROGBITS 0000000000000000 002bf0 000cd0 00 0 0 1 [10] .rel.BTF.ext REL 0000000000000000 0038c0 000ca0 10 14 9 8 All the sources are at: The branch without libbpf https://github.com/daynix/qemu/tree/eBPF_RFC The branch with libbpf https://github.com/daynix/qemu/tree/eBPF_RFCv2 all the eBPF-related code is under qemu/ebpf directory. > In any case, while I do think it smells a little of premature > optimisation, you can of course strip the BTF information until you need > it. Having it around makes debugging easier (bpftool will expand your > map structures for you when dumping maps, and that sort of thing), but > it's not really essential if you don't need CO-RE. > > > This is only reason to prefer non-libbpf option for this specific eBPF > > You can still use libbpf without BTF. It's using BTF without libbpf that > tends to not work so well... > > If we build the eBPF without '-g' or strip the BTF information out of the object file the libbpf crashes right after issuing printout "libbpf: BTF is required, but is missing or corrupted". We did not investigate this too deeply but on the first glance it looks like the presence of maps automatically makes the libbpf to require BTF. > >> As a side note, though, instead of embedding the BPF program into a .h, > >> you could simply ship it as a .o and load it from the file system. We do > >> that with xdp-tools and install the bpf object files into > /usr/$LIB/bpf/. > >> > > > > Yes, we've discussed this option and decided to go with embedding the BPF > > https://lists.gnu.org/archive/html/qemu-devel/2020-11/msg02157.html > > Right, okay. I'll note, though, that if your concern is that the BPF > code should always match the rest of the code base, omitting the > compilation if there's no Clang present seems like it could lead to > problems :) > > Also, if you do go the embedded-bytecode route, you may want to have a > look at the upstream 'skeleton' concept. It takes a BPF object file and > automatically generates a header file that gives you direct access to > maps, programs and global data in C. There are some examples in > selftests/bpf on how to use it, but you basically just run > 'bpftool gen skeleton mybpf.o'. > > Indeed looks very interesting. We've missed this feature. Thank you very much! > -Toke > >
Yuri Benditovich <yuri.benditovich@daynix.com> writes: > On Fri, Dec 4, 2020 at 12:09 PM Toke Høiland-Jørgensen <toke@redhat.com> > wrote: > >> Yuri Benditovich <yuri.benditovich@daynix.com> writes: >> >> > On Wed, Dec 2, 2020 at 4:18 PM Toke Høiland-Jørgensen <toke@redhat.com> >> > wrote: >> > >> >> Jason Wang <jasowang@redhat.com> writes: >> >> >> >> > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: >> >> >> This set of patches introduces the usage of eBPF for packet steering >> >> >> and RSS hash calculation: >> >> >> * RSS(Receive Side Scaling) is used to distribute network packets to >> >> >> guest virtqueues by calculating packet hash >> >> >> * Additionally adding support for the usage of RSS with vhost >> >> >> >> >> >> The eBPF works on kernels 5.8+ >> >> >> On earlier kerneld it fails to load and the RSS feature is reported >> >> >> only without vhost and implemented in 'in-qemu' software. >> >> >> >> >> >> Implementation notes: >> >> >> Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF program. >> >> >> Added libbpf dependency and eBPF support. >> >> >> The eBPF program is part of the qemu and presented as an array >> >> >> of BPF ELF file data. >> >> >> The compilation of eBPF is not part of QEMU build and can be done >> >> >> using provided Makefile.ebpf(need to adjust 'linuxhdrs'). >> >> >> Added changes to virtio-net and vhost, primary eBPF RSS is used. >> >> >> 'in-qemu' RSS used in the case of hash population and as a fallback >> >> option. >> >> >> For vhost, the hash population feature is not reported to the guest. >> >> >> >> >> >> Please also see the documentation in PATCH 5/5. >> >> >> >> >> >> I am sending those patches as RFC to initiate the discussions and get >> >> >> feedback on the following points: >> >> >> * Fallback when eBPF is not supported by the kernel >> >> >> * Live migration to the kernel that doesn't have eBPF support >> >> >> * Integration with current QEMU build >> >> >> * Additional usage for eBPF for packet filtering >> >> >> >> >> >> Known issues: >> >> >> * hash population not supported by eBPF RSS: 'in-qemu' RSS used >> >> >> as a fallback, also, hash population feature is not reported to >> guests >> >> >> with vhost. >> >> >> * big-endian BPF support: for now, eBPF isn't supported on >> >> >> big-endian systems. Can be added in future if required. >> >> >> * huge .h file with eBPF binary. The size of .h file containing >> >> >> eBPF binary is currently ~5K lines, because the binary is built with >> >> debug information. >> >> >> The binary without debug/BTF info can't be loaded by libbpf. >> >> >> We're looking for possibilities to reduce the size of the .h files. >> >> > >> >> > >> >> > Adding Toke for sharing more idea from eBPF side. >> >> > >> >> > We had some discussion on the eBPF issues: >> >> > >> >> > 1) Whether or not to use libbpf. Toke strongly suggest to use libbpf >> >> > 2) Whether or not to use BTF. Toke confirmed that if we don't access >> any >> >> > skb metadata, BTF is not strictly required for CO-RE. But it might >> still >> >> > useful for e.g debugging. >> >> > 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed that >> we >> >> > can strip debug symbols, but Yuri found some sections can't be >> stripped, >> >> > we can keep discussing here. >> >> >> >> I just tried simply running 'strip' on a sample trivial XDP program, >> >> which brought its size down from ~5k to ~1k and preserved the BTF >> >> information without me having to do anything. >> >> >> > >> > With our eBPF code the numbers are slightly different: >> > The code size without BTF: 7.5K (built without '-g') >> > Built with '-g': 45K >> > Stripped: 19K >> > The difference between 7.5 and 19K still seems significant, especially >> when >> > we do not use any kernel structures and do not need these BTF sections >> >> That does seem like a lot of BTF information. Did you confirm (with >> objdump) that it's the .BTF* sections that take up these extra 12k? Do >> you have some really complicated data structures in the file or >> something? Got a link to the source somewhere that isn't a web mailing >> list archive? :) >> >> > Looks like the extra size is related to BTF: there are 4 BTF sections that > take 12.5K > [ 7] .BTF PROGBITS 0000000000000000 00144c 00175d 00 > 0 0 1 > [ 8] .rel.BTF REL 0000000000000000 002bb0 000040 10 > 14 7 8 > [ 9] .BTF.ext PROGBITS 0000000000000000 002bf0 000cd0 00 > 0 0 1 > [10] .rel.BTF.ext REL 0000000000000000 0038c0 000ca0 10 > 14 9 8 Right, okay, that does not look completely outrageous with the amount of code and type information you have in that file. > All the sources are at: > The branch without libbpf > https://github.com/daynix/qemu/tree/eBPF_RFC > The branch with libbpf > https://github.com/daynix/qemu/tree/eBPF_RFCv2 > > all the eBPF-related code is under qemu/ebpf directory. Ah, cool, thanks! >> In any case, while I do think it smells a little of premature >> optimisation, you can of course strip the BTF information until you need >> it. Having it around makes debugging easier (bpftool will expand your >> map structures for you when dumping maps, and that sort of thing), but >> it's not really essential if you don't need CO-RE. >> >> > This is only reason to prefer non-libbpf option for this specific eBPF >> >> You can still use libbpf without BTF. It's using BTF without libbpf that >> tends to not work so well... >> >> > If we build the eBPF without '-g' or strip the BTF information out of the > object file the libbpf crashes right after issuing printout "libbpf: BTF is > required, but is missing or corrupted". > We did not investigate this too deeply but on the first glance it looks > like the presence of maps automatically makes the libbpf to require BTF. Ah, right. Well, you're using the BTF-based map definition syntax. So yeah, that does require BTF: The __uint() and __type() macros really expand to type definitions that are specifically crafted to be embedded as BTF in the file. You could use the old-style map definitions that don't use BTF[0], but BTF is really where things are going in BPF-land so I think longer term you'll probably end up needing it anyway. So going to this much trouble just to save 10k on binary size seems to me like it's a decision you'll end up regretting :) [0] https://github.com/xdp-project/xdp-tutorial/blob/master/basic03-map-counter/xdp_prog_kern.c#L11 -Toke
On Fri, Dec 4, 2020 at 3:57 PM Toke Høiland-Jørgensen <toke@redhat.com> wrote: > Yuri Benditovich <yuri.benditovich@daynix.com> writes: > > > On Fri, Dec 4, 2020 at 12:09 PM Toke Høiland-Jørgensen <toke@redhat.com> > > wrote: > > > >> Yuri Benditovich <yuri.benditovich@daynix.com> writes: > >> > >> > On Wed, Dec 2, 2020 at 4:18 PM Toke Høiland-Jørgensen < > toke@redhat.com> > >> > wrote: > >> > > >> >> Jason Wang <jasowang@redhat.com> writes: > >> >> > >> >> > On 2020/11/19 下午7:13, Andrew Melnychenko wrote: > >> >> >> This set of patches introduces the usage of eBPF for packet > steering > >> >> >> and RSS hash calculation: > >> >> >> * RSS(Receive Side Scaling) is used to distribute network packets > to > >> >> >> guest virtqueues by calculating packet hash > >> >> >> * Additionally adding support for the usage of RSS with vhost > >> >> >> > >> >> >> The eBPF works on kernels 5.8+ > >> >> >> On earlier kerneld it fails to load and the RSS feature is > reported > >> >> >> only without vhost and implemented in 'in-qemu' software. > >> >> >> > >> >> >> Implementation notes: > >> >> >> Linux TAP TUNSETSTEERINGEBPF ioctl was used to set the eBPF > program. > >> >> >> Added libbpf dependency and eBPF support. > >> >> >> The eBPF program is part of the qemu and presented as an array > >> >> >> of BPF ELF file data. > >> >> >> The compilation of eBPF is not part of QEMU build and can be done > >> >> >> using provided Makefile.ebpf(need to adjust 'linuxhdrs'). > >> >> >> Added changes to virtio-net and vhost, primary eBPF RSS is used. > >> >> >> 'in-qemu' RSS used in the case of hash population and as a > fallback > >> >> option. > >> >> >> For vhost, the hash population feature is not reported to the > guest. > >> >> >> > >> >> >> Please also see the documentation in PATCH 5/5. > >> >> >> > >> >> >> I am sending those patches as RFC to initiate the discussions and > get > >> >> >> feedback on the following points: > >> >> >> * Fallback when eBPF is not supported by the kernel > >> >> >> * Live migration to the kernel that doesn't have eBPF support > >> >> >> * Integration with current QEMU build > >> >> >> * Additional usage for eBPF for packet filtering > >> >> >> > >> >> >> Known issues: > >> >> >> * hash population not supported by eBPF RSS: 'in-qemu' RSS used > >> >> >> as a fallback, also, hash population feature is not reported to > >> guests > >> >> >> with vhost. > >> >> >> * big-endian BPF support: for now, eBPF isn't supported on > >> >> >> big-endian systems. Can be added in future if required. > >> >> >> * huge .h file with eBPF binary. The size of .h file containing > >> >> >> eBPF binary is currently ~5K lines, because the binary is built > with > >> >> debug information. > >> >> >> The binary without debug/BTF info can't be loaded by libbpf. > >> >> >> We're looking for possibilities to reduce the size of the .h > files. > >> >> > > >> >> > > >> >> > Adding Toke for sharing more idea from eBPF side. > >> >> > > >> >> > We had some discussion on the eBPF issues: > >> >> > > >> >> > 1) Whether or not to use libbpf. Toke strongly suggest to use > libbpf > >> >> > 2) Whether or not to use BTF. Toke confirmed that if we don't > access > >> any > >> >> > skb metadata, BTF is not strictly required for CO-RE. But it might > >> still > >> >> > useful for e.g debugging. > >> >> > 3) About the huge (5K lines, see patch #2 Toke). Toke confirmed > that > >> we > >> >> > can strip debug symbols, but Yuri found some sections can't be > >> stripped, > >> >> > we can keep discussing here. > >> >> > >> >> I just tried simply running 'strip' on a sample trivial XDP program, > >> >> which brought its size down from ~5k to ~1k and preserved the BTF > >> >> information without me having to do anything. > >> >> > >> > > >> > With our eBPF code the numbers are slightly different: > >> > The code size without BTF: 7.5K (built without '-g') > >> > Built with '-g': 45K > >> > Stripped: 19K > >> > The difference between 7.5 and 19K still seems significant, especially > >> when > >> > we do not use any kernel structures and do not need these BTF sections > >> > >> That does seem like a lot of BTF information. Did you confirm (with > >> objdump) that it's the .BTF* sections that take up these extra 12k? Do > >> you have some really complicated data structures in the file or > >> something? Got a link to the source somewhere that isn't a web mailing > >> list archive? :) > >> > >> > > Looks like the extra size is related to BTF: there are 4 BTF sections > that > > take 12.5K > > [ 7] .BTF PROGBITS 0000000000000000 00144c 00175d > 00 > > 0 0 1 > > [ 8] .rel.BTF REL 0000000000000000 002bb0 000040 > 10 > > 14 7 8 > > [ 9] .BTF.ext PROGBITS 0000000000000000 002bf0 000cd0 > 00 > > 0 0 1 > > [10] .rel.BTF.ext REL 0000000000000000 0038c0 000ca0 > 10 > > 14 9 8 > > Right, okay, that does not look completely outrageous with the amount of > code and type information you have in that file. > > > All the sources are at: > > The branch without libbpf > > https://github.com/daynix/qemu/tree/eBPF_RFC > > The branch with libbpf > > https://github.com/daynix/qemu/tree/eBPF_RFCv2 > > > > all the eBPF-related code is under qemu/ebpf directory. > > Ah, cool, thanks! > > >> In any case, while I do think it smells a little of premature > >> optimisation, you can of course strip the BTF information until you need > >> it. Having it around makes debugging easier (bpftool will expand your > >> map structures for you when dumping maps, and that sort of thing), but > >> it's not really essential if you don't need CO-RE. > >> > >> > This is only reason to prefer non-libbpf option for this specific eBPF > >> > >> You can still use libbpf without BTF. It's using BTF without libbpf that > >> tends to not work so well... > >> > >> > > If we build the eBPF without '-g' or strip the BTF information out of the > > object file the libbpf crashes right after issuing printout "libbpf: BTF > is > > required, but is missing or corrupted". > > We did not investigate this too deeply but on the first glance it looks > > like the presence of maps automatically makes the libbpf to require BTF. > > Ah, right. Well, you're using the BTF-based map definition syntax. So > yeah, that does require BTF: The __uint() and __type() macros really > expand to type definitions that are specifically crafted to be embedded > as BTF in the file. > Yes, now the EBPF built without '-g' also can be loaded via libbpf and we can enable/disable BTF as we need. Again, thank you very much! > > You could use the old-style map definitions that don't use BTF[0], but > BTF is really where things are going in BPF-land so I think longer term > you'll probably end up needing it anyway. So going to this much trouble > just to save 10k on binary size seems to me like it's a decision you'll > end up regretting :) > > [0] > https://github.com/xdp-project/xdp-tutorial/blob/master/basic03-map-counter/xdp_prog_kern.c#L11 > > -Toke > >
© 2016 - 2024 Red Hat, Inc.