Message ID | 20230523073238.54236-1-jasowang@redhat.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On 5/23/23 00:31, Jason Wang wrote: > The following changes since commit 886c0453cbf10eebd42a9ccf89c3e46eb389c357: > > Merge tag 'pull-qapi-2023-05-17-v2' ofhttps://repo.or.cz/qemu/armbru into staging (2023-05-22 15:54:21 -0700) > > are available in the git repository at: > > https://github.com/jasowang/qemu.git tags/net-pull-request > > for you to fetch changes up to 792676c165159c11412346870fd58fd243ab2166: > > rtl8139: fix large_send_mss divide-by-zero (2023-05-23 15:20:15 +0800) > > ---------------------------------------------------------------- Applied, thanks. Please update https://wiki.qemu.org/ChangeLog/8.1 as appropriate. r~
23.05.2023 10:31, Jason Wang wrote: > https://github.com/jasowang/qemu.git tags/net-pull-request I might be wrong but it looks like besides the rtl8139 fix which went to -stable already, and "e1000e: Fix tx/rx counters" change which has been agreed upon before, there are a few other changes which are worth to have in -stable, in 8.0 and some even in 7.2. What do you think, which ones are important for -stable? If there's stable material in the future, please Cc: qemu-stable@nongnu.org. Thanks, /mjt > Akihiko Odaki (48): > hw/net/net_tx_pkt: Decouple implementation from PCI > hw/net/net_tx_pkt: Decouple interface from PCI > e1000x: Fix BPRC and MPRC > igb: Fix Rx packet type encoding > igb: Do not require CTRL.VME for tx VLAN tagging > igb: Clear IMS bits when committing ICR access > net/net_rx_pkt: Use iovec for net_rx_pkt_set_protocols() > e1000e: Always copy ethernet header > igb: Always copy ethernet header > Fix references to igb Avocado test > tests/avocado: Remove unused imports > tests/avocado: Remove test_igb_nomsi_kvm > hw/net/net_tx_pkt: Remove net_rx_pkt_get_l4_info > net/eth: Rename eth_setup_vlan_headers_ex > e1000x: Share more Rx filtering logic > e1000x: Take CRC into consideration for size check > e1000x: Rename TcpIpv6 into TcpIpv6Ex > e1000e: Always log status after building rx metadata > igb: Always log status after building rx metadata > igb: Remove goto > igb: Read DCMD.VLE of the first Tx descriptor > e1000e: Reset packet state after emptying Tx queue > vmxnet3: Reset packet state after emptying Tx queue > igb: Add more definitions for Tx descriptor > igb: Share common VF constants > igb: Fix igb_mac_reg_init coding style alignment > igb: Clear EICR bits for delayed MSI-X interrupts > e1000e: Rename a variable in e1000e_receive_internal() > igb: Rename a variable in igb_receive_internal() > net/eth: Use void pointers > net/eth: Always add VLAN tag > hw/net/net_rx_pkt: Enforce alignment for eth_header > tests/qtest/libqos/igb: Set GPIE.Multiple_MSIX > igb: Implement MSI-X single vector mode > igb: Use UDP for RSS hash > igb: Implement Rx SCTP CSO > igb: Implement Tx SCTP CSO > igb: Strip the second VLAN tag for extended VLAN > igb: Filter with the second VLAN tag for extended VLAN > igb: Implement igb-specific oversize check > igb: Implement Rx PTP2 timestamp > igb: Implement Tx timestamp > e1000e: Notify only new interrupts > igb: Notify only new interrupts > igb: Clear-on-read ICR when ICR.INTA is set > vmxnet3: Do not depend on PC > MAINTAINERS: Add a reviewer for network packet abstractions > docs/system/devices/igb: Note igb is tested for DPDK
On Wed, May 24, 2023 at 3:54 AM Michael Tokarev <mjt@tls.msk.ru> wrote: > > 23.05.2023 10:31, Jason Wang wrote: > > > https://github.com/jasowang/qemu.git tags/net-pull-request > > I might be wrong but it looks like besides the rtl8139 fix which went > to -stable already, and "e1000e: Fix tx/rx counters" change which has > been agreed upon before, there are a few other changes which are worth > to have in -stable, in 8.0 and some even in 7.2. What do you think, > which ones are important for -stable? I'm not sure, most fixes are for igb, considering igb is complicated and just introduced since 8.0, it should be considered as a technical preview somehow. Then I'm not sure whether or not it's worth it to backport them. For the rest, it looks like it should be fine to go for -stable. Akihiko, what's your thoughts? Thanks > > If there's stable material in the future, please Cc: qemu-stable@nongnu.org. > > Thanks, > > /mjt > > > Akihiko Odaki (48): > > hw/net/net_tx_pkt: Decouple implementation from PCI > > hw/net/net_tx_pkt: Decouple interface from PCI > > e1000x: Fix BPRC and MPRC > > igb: Fix Rx packet type encoding > > igb: Do not require CTRL.VME for tx VLAN tagging > > igb: Clear IMS bits when committing ICR access > > net/net_rx_pkt: Use iovec for net_rx_pkt_set_protocols() > > e1000e: Always copy ethernet header > > igb: Always copy ethernet header > > Fix references to igb Avocado test > > tests/avocado: Remove unused imports > > tests/avocado: Remove test_igb_nomsi_kvm > > hw/net/net_tx_pkt: Remove net_rx_pkt_get_l4_info > > net/eth: Rename eth_setup_vlan_headers_ex > > e1000x: Share more Rx filtering logic > > e1000x: Take CRC into consideration for size check > > e1000x: Rename TcpIpv6 into TcpIpv6Ex > > e1000e: Always log status after building rx metadata > > igb: Always log status after building rx metadata > > igb: Remove goto > > igb: Read DCMD.VLE of the first Tx descriptor > > e1000e: Reset packet state after emptying Tx queue > > vmxnet3: Reset packet state after emptying Tx queue > > igb: Add more definitions for Tx descriptor > > igb: Share common VF constants > > igb: Fix igb_mac_reg_init coding style alignment > > igb: Clear EICR bits for delayed MSI-X interrupts > > e1000e: Rename a variable in e1000e_receive_internal() > > igb: Rename a variable in igb_receive_internal() > > net/eth: Use void pointers > > net/eth: Always add VLAN tag > > hw/net/net_rx_pkt: Enforce alignment for eth_header > > tests/qtest/libqos/igb: Set GPIE.Multiple_MSIX > > igb: Implement MSI-X single vector mode > > igb: Use UDP for RSS hash > > igb: Implement Rx SCTP CSO > > igb: Implement Tx SCTP CSO > > igb: Strip the second VLAN tag for extended VLAN > > igb: Filter with the second VLAN tag for extended VLAN > > igb: Implement igb-specific oversize check > > igb: Implement Rx PTP2 timestamp > > igb: Implement Tx timestamp > > e1000e: Notify only new interrupts > > igb: Notify only new interrupts > > igb: Clear-on-read ICR when ICR.INTA is set > > vmxnet3: Do not depend on PC > > MAINTAINERS: Add a reviewer for network packet abstractions > > docs/system/devices/igb: Note igb is tested for DPDK >
On 2023/05/24 13:06, Jason Wang wrote: > On Wed, May 24, 2023 at 3:54 AM Michael Tokarev <mjt@tls.msk.ru> wrote: >> >> 23.05.2023 10:31, Jason Wang wrote: >> >>> https://github.com/jasowang/qemu.git tags/net-pull-request >> >> I might be wrong but it looks like besides the rtl8139 fix which went >> to -stable already, and "e1000e: Fix tx/rx counters" change which has >> been agreed upon before, there are a few other changes which are worth >> to have in -stable, in 8.0 and some even in 7.2. What do you think, >> which ones are important for -stable? > > I'm not sure, most fixes are for igb, considering igb is complicated > and just introduced since 8.0, it should be considered as a technical > preview somehow. Then I'm not sure whether or not it's worth it to > backport them. For the rest, it looks like it should be fine to go for > -stable. > > Akihiko, what's your thoughts? > > Thanks I think it is fine to backport all of patches with Fixes: tags as there are only five such patches for igb. Namely, they are: - [PULL 05/50] igb: Fix Rx packet type encoding - [PULL 06/50] igb: Do not require CTRL.VME for tx VLAN tagging - [PULL 07/50] igb: Clear IMS bits when committing ICR access - [PULL 08/50] net/net_rx_pkt: Use iovec for net_rx_pkt_set_protocols() - [PULL 10/50] igb: Always copy ethernet header That said, if that looks too many, you may backport only the last two patches, which fixes possible out-of-bounds accesses: - [PULL 08/50] net/net_rx_pkt: Use iovec for net_rx_pkt_set_protocols() - [PULL 10/50] igb: Always copy ethernet header The other patches are just behavioral changes which do not corrupt QEMU state so they are not that important. Regards, Akihiko Odaki