Pull networking fixes from Jakub Kicinski:
"Including fixes from bpf, mac80211.
Current release - regressions:
- devlink: don't throw an error if flash notification sent before
devlink visible
- page_pool: Revert "page_pool: disable dma mapping support...",
turns out there are active arches who need it
Current release - new code bugs:
- amt: cancel delayed_work synchronously in amt_fini()
Previous releases - regressions:
- xsk: fix crash on double free in buffer pool
- bpf: fix inner map state pruning regression causing program
rejections
- mac80211: drop check for DONT_REORDER in __ieee80211_select_queue,
preventing mis-selecting the best effort queue
- mac80211: do not access the IV when it was stripped
- mac80211: fix radiotap header generation, off-by-one
- nl80211: fix getting radio statistics in survey dump
- e100: fix device suspend/resume
Previous releases - always broken:
- tcp: fix uninitialized access in skb frags array for Rx 0cp
- bpf: fix toctou on read-only map's constant scalar tracking
- bpf: forbid bpf_ktime_get_coarse_ns and bpf_timer_* in tracing
progs
- tipc: only accept encrypted MSG_CRYPTO msgs
- smc: transfer remaining wait queue entries during fallback, fix
missing wake ups
- udp: validate checksum in udp_read_sock() (when sockmap is used)
- sched: act_mirred: drop dst for the direction from egress to
ingress
- virtio_net_hdr_to_skb: count transport header in UFO, prevent
allowing bad skbs into the stack
- nfc: reorder the logic in nfc_{un,}register_device, fix unregister
- ipsec: check return value of ipv6_skip_exthdr
- usb: r8152: add MAC passthrough support for more Lenovo Docks"
* tag 'net-5.16-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (96 commits)
ptp: ocp: Fix a couple NULL vs IS_ERR() checks
net: ethernet: dec: tulip: de4x5: fix possible array overflows in type3_infoblock()
net: tulip: de4x5: fix the problem that the array 'lp->phy[8]' may be out of bound
ipv6: check return value of ipv6_skip_exthdr
e100: fix device suspend/resume
devlink: Don't throw an error if flash notification sent before devlink visible
page_pool: Revert "page_pool: disable dma mapping support..."
ethernet: hisilicon: hns: hns_dsaf_misc: fix a possible array overflow in hns_dsaf_ge_srst_by_port()
octeontx2-af: debugfs: don't corrupt user memory
NFC: add NCI_UNREG flag to eliminate the race
NFC: reorder the logic in nfc_{un,}register_device
NFC: reorganize the functions in nci_request
tipc: check for null after calling kmemdup
i40e: Fix display error code in dmesg
i40e: Fix creation of first queue by omitting it if is not power of two
i40e: Fix warning message and call stack during rmmod i40e driver
i40e: Fix ping is lost after configuring ADq on VF
i40e: Fix changing previously set num_queue_pairs for PFs
i40e: Fix NULL ptr dereference on VSI filter sync
i40e: Fix correct max_pkt_size on VF RX queue
...
Pull KVM fixes from Paolo Bonzini:
"Selftest changes:
- Cleanups for the perf test infrastructure and mapping hugepages
- Avoid contention on mmap_sem when the guests start to run
- Add event channel upcall support to xen_shinfo_test
x86 changes:
- Fixes for Xen emulation
- Kill kvm_map_gfn() / kvm_unmap_gfn() and broken gfn_to_pfn_cache
- Fixes for migration of 32-bit nested guests on 64-bit hypervisor
- Compilation fixes
- More SEV cleanups
Generic:
- Cap the return value of KVM_CAP_NR_VCPUS to both KVM_CAP_MAX_VCPUS
and num_online_cpus(). Most architectures were only using one of
the two"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (42 commits)
KVM: x86: Cap KVM_CAP_NR_VCPUS by KVM_CAP_MAX_VCPUS
KVM: s390: Cap KVM_CAP_NR_VCPUS by num_online_cpus()
KVM: RISC-V: Cap KVM_CAP_NR_VCPUS by KVM_CAP_MAX_VCPUS
KVM: PPC: Cap KVM_CAP_NR_VCPUS by KVM_CAP_MAX_VCPUS
KVM: MIPS: Cap KVM_CAP_NR_VCPUS by KVM_CAP_MAX_VCPUS
KVM: arm64: Cap KVM_CAP_NR_VCPUS by kvm_arm_default_max_vcpus()
KVM: x86: Assume a 64-bit hypercall for guests with protected state
selftests: KVM: Add /x86_64/sev_migrate_tests to .gitignore
riscv: kvm: fix non-kernel-doc comment block
KVM: SEV: Fix typo in and tweak name of cmd_allowed_from_miror()
KVM: SEV: Drop a redundant setting of sev->asid during initialization
KVM: SEV: WARN if SEV-ES is marked active but SEV is not
KVM: SEV: Set sev_info.active after initial checks in sev_guest_init()
KVM: SEV: Disallow COPY_ENC_CONTEXT_FROM if target has created vCPUs
KVM: Kill kvm_map_gfn() / kvm_unmap_gfn() and gfn_to_pfn_cache
KVM: nVMX: Use a gfn_to_hva_cache for vmptrld
KVM: nVMX: Use kvm_read_guest_offset_cached() for nested VMCS check
KVM: x86/xen: Use sizeof_field() instead of open-coding it
KVM: nVMX: Use kvm_{read,write}_guest_cached() for shadow_vmcs12
KVM: x86/xen: Fix get_attr of KVM_XEN_ATTR_TYPE_SHARED_INFO
...
Pull printk fixes from Petr Mladek:
- Try to flush backtraces from other CPUs also on the local one. This
was a regression caused by printk_safe buffers removal.
- Remove header dependency warning.
* tag 'printk-for-5.16-fixup' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux:
printk: Remove printk.h inclusion in percpu.h
printk: restore flushing of NMI buffers on remote CPUs after NMI backtraces
PSFP rules take effect on the streams from any port of VSC9959 switch.
This patch use ingress port to limit the rule only active on this port.
Each stream can only match two ingress source ports in VSC9959. Streams
from lowest port gets the configuration of SFID pointed by MAC Table
lookup and streams from highest port gets the configuration of (SFID+1)
pointed by MAC Table lookup. This patch defines the PSFP rule on highest
port as dummy rule, which means that it does not modify the MAC table.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Policer was previously automatically assigned from the highest index to
the lowest index from policer pool. But police action of tc flower now
uses index to set an police entry. This patch uses the police index to
set vcap policers, so that one policer can be shared by multiple rules.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
VSC9959 supports Per-Stream Filtering and Policing(PSFP) that complies
with the IEEE 802.1Qci standard. The stream is identified by Null stream
identification(DMAC and VLAN ID) defined in IEEE802.1CB.
For PSFP, four tables need to be set up: stream table, stream filter
table, stream gate table, and flow meter table. Identify the stream by
parsing the tc flower keys and add it to the stream table. The stream
filter table is automatically maintained, and its index is determined by
SGID(flow gate index) and FMID(flow meter index).
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
PSFP support gate and police action. This patch add the gate and police
action to flower parse action, check chain ID to determine which block
to offload. Adding psfp callback functions to add, delete and update gate
and police in PSFP table if hardware supports it.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ocelot_mact_learn_streamdata() can be used in VSC9959 to overwrite an
FDB entry with stream data. The stream data includes SFID and SSID which
can be used for PSFP and FRER set.
ocelot_mact_lookup() can be used to check if the given {DMAC, VID} FDB
entry is exist, and also can retrieve the DEST_IDX and entry type for
the FDB entry.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add missing entries to fix these "make htmldocs" warnings.
./include/linux/skbuff.h:953: warning: Function parameter or member 'll_node' not described in 'sk_buff'
./include/net/sock.h:540: warning: Function parameter or member 'defer_list' not described in 'sock'
Fixes: f35f821935 ("tcp: defer skb freeing after socket lock is released")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit d00e60ee54.
As reported by Guillaume in [1]:
Enabling LPAE always enables CONFIG_ARCH_DMA_ADDR_T_64BIT
in 32-bit systems, which breaks the bootup proceess when a
ethernet driver is using page pool with PP_FLAG_DMA_MAP flag.
As we were hoping we had no active consumers for such system
when we removed the dma mapping support, and LPAE seems like
a common feature for 32 bits system, so revert it.
1. https://www.spinics.net/lists/netdev/msg779890.html
Fixes: d00e60ee54 ("page_pool: disable dma mapping support for 32-bit arch with 64-bit DMA")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Reported-by: "kernelci.org bot" <bot@kernelci.org>
Tested-by: "kernelci.org bot" <bot@kernelci.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
* Fixes for Xen emulation
* Kill kvm_map_gfn() / kvm_unmap_gfn() and broken gfn_to_pfn_cache
* Fixes for migration of 32-bit nested guests on 64-bit hypervisor
* Compilation fixes
* More SEV cleanups
In commit 7e2175ebd6 ("KVM: x86: Fix recording of guest steal time /
preempted status") I removed the only user of these functions because
it was basically impossible to use them safely.
There are two stages to the GFN->PFN mapping; first through the KVM
memslots to a userspace HVA and then through the page tables to
translate that HVA to an underlying PFN. Invalidations of the former
were being handled correctly, but no attempt was made to use the MMU
notifiers to invalidate the cache when the HVA->GFN mapping changed.
As a prelude to reinventing the gfn_to_pfn_cache with more usable
semantics, rip it out entirely and untangle the implementation of
the unsafe kvm_vcpu_map()/kvm_vcpu_unmap() functions from it.
All current users of kvm_vcpu_map() also look broken right now, and
will be dealt with separately. They broadly fall into two classes:
* Those which map, access the data and immediately unmap. This is
mostly gratuitous and could just as well use the existing user
HVA, and could probably benefit from a gfn_to_hva_cache as they
do so.
* Those which keep the mapping around for a longer time, perhaps
even using the PFN directly from the guest. These will need to
be converted to the new gfn_to_pfn_cache and then kvm_vcpu_map()
can be removed too.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Message-Id: <20211115165030.7422-8-dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add support to inet v4 raw sockets for binding to nonlocal addresses
through the IP_FREEBIND and IP_TRANSPARENT socket options, as well as
the ipv4.ip_nonlocal_bind kernel parameter.
Add helper function to inet_sock.h to check for bind address validity on
the base of the address type and whether nonlocal address are enabled
for the socket via any of the sockopts/sysctl, deduplicating checks in
ipv4/ping.c, ipv4/af_inet.c, ipv6/af_inet6.c (for mapped v4->v6
addresses), and ipv4/raw.c.
Add test cases with IP[V6]_FREEBIND verifying that both v4 and v6 raw
sockets support binding to nonlocal addresses after the change. Add
necessary support for the test cases to nettest.
Signed-off-by: Riccardo Paolo Bestetti <pbl@bestov.io>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20211117090010.125393-1-pbl@bestov.io
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are two sites that calls queue_work() after the
destroy_workqueue() and lead to possible UAF.
The first site is nci_send_cmd(), which can happen after the
nci_close_device as below
nfcmrvl_nci_unregister_dev | nfc_genl_dev_up
nci_close_device |
flush_workqueue |
del_timer_sync |
nci_unregister_device | nfc_get_device
destroy_workqueue | nfc_dev_up
nfc_unregister_device | nci_dev_up
device_del | nci_open_device
| __nci_request
| nci_send_cmd
| queue_work !!!
Another site is nci_cmd_timer, awaked by the nci_cmd_work from the
nci_send_cmd.
... | ...
nci_unregister_device | queue_work
destroy_workqueue |
nfc_unregister_device | ...
device_del | nci_cmd_work
| mod_timer
| ...
| nci_cmd_timer
| queue_work !!!
For the above two UAF, the root cause is that the nfc_dev_up can race
between the nci_unregister_device routine. Therefore, this patch
introduce NCI_UNREG flag to easily eliminate the possible race. In
addition, the mutex_lock in nci_close_device can act as a barrier.
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Fixes: 6a2968aaf5 ("NFC: basic NCI protocol implementation")
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Link: https://lore.kernel.org/r/20211116152732.19238-1-linma@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
These are not fast path, there is no point in inlining them.
Also provide netif_freeze_queues()/netif_unfreeze_queues()
so that we can use them from dev_watchdog() in the following patch.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In following patches, dev_watchdog() will no longer stop all queues.
It will read queue->trans_start locklessly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tx_timeout_show() assumed dev_watchdog() would stop all
the queues, to fetch queue->trans_timeout under protection
of the queue->_xmit_lock.
As we want to no longer disrupt transmits, we use an
atomic_long_t instead.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: david decotigny <david.decotigny@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Luiz Augusto von Dentz says:
====================
bluetooth-next pull request for net-next:
- Add support for AOSP Bluetooth Quality Report
- Enables AOSP extension for Mediatek Chip (MT7921 & MT7922)
- Rework of HCI command execution serialization
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
virtio_net_hdr_to_skb does not set the skb's gso_size and gso_type
correctly for UFO packets received via virtio-net that are a little over
the GSO size. This can lead to problems elsewhere in the networking
stack, e.g. ovs_vport_send dropping over-sized packets if gso_size is
not set.
This is due to the comparison
if (skb->len - p_off > gso_size)
not properly accounting for the transport layer header.
p_off includes the size of the transport layer header (thlen), so
skb->len - p_off is the size of the TCP/UDP payload.
gso_size is read from the virtio-net header. For UFO, fragmentation
happens at the IP level so does not need to include the UDP header.
Hence the calculation could be comparing a TCP/UDP payload length with
an IP payload length, causing legitimate virtio-net packets to have
lack gso_type/gso_size information.
Example: a UDP packet with payload size 1473 has IP payload size 1481.
If the guest used UFO, it is not fragmented and the virtio-net header's
flags indicate that it is a GSO frame (VIRTIO_NET_HDR_GSO_UDP), with
gso_size = 1480 for an MTU of 1500. skb->len will be 1515 and p_off
will be 42, so skb->len - p_off = 1473. Hence the comparison fails, and
shinfo->gso_size and gso_type are not set as they should be.
Instead, add the UDP header length before comparing to gso_size when
using UFO. In this way, it is the size of the IP payload that is
compared to gso_size.
Fixes: 6dd912f826 ("net: check untrusted gso_size at kernel entry")
Signed-off-by: Jonathan Davies <jonathan.davies@nutanix.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
mlx5-fixes-2021-11-16
Please pull this mlx5 fixes series, or let me know in case of any problem.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
siphash keys use 16 bytes.
Define siphash_aligned_key_t macro so that we can make sure they
are not crossing a cache line boundary.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Same rationale than prior patch : using the dedicated
section avoid holes and pack all these bool values.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
.data.once contains nicely packed bool variables.
It is used already by DO_ONCE_LITE().
Using it also in DO_ONCE() removes holes in .data section.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Borkmann says:
====================
pull-request: bpf 2021-11-16
We've added 12 non-merge commits during the last 5 day(s) which contain
a total of 23 files changed, 573 insertions(+), 73 deletions(-).
The main changes are:
1) Fix pruning regression where verifier went overly conservative rejecting
previsouly accepted programs, from Alexei Starovoitov and Lorenz Bauer.
2) Fix verifier TOCTOU bug when using read-only map's values as constant
scalars during verification, from Daniel Borkmann.
3) Fix a crash due to a double free in XSK's buffer pool, from Magnus Karlsson.
4) Fix libbpf regression when cross-building runqslower, from Jean-Philippe Brucker.
5) Forbid use of bpf_ktime_get_coarse_ns() and bpf_timer_*() helpers in tracing
programs due to deadlock possibilities, from Dmitrii Banshchikov.
6) Fix checksum validation in sockmap's udp_read_sock() callback, from Cong Wang.
7) Various BPF sample fixes such as XDP stats in xdp_sample_user, from Alexander Lobakin.
8) Fix libbpf gen_loader error handling wrt fd cleanup, from Kumar Kartikeya Dwivedi.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
udp: Validate checksum in udp_read_sock()
bpf: Fix toctou on read-only map's constant scalar tracking
samples/bpf: Fix build error due to -isystem removal
selftests/bpf: Add tests for restricted helpers
bpf: Forbid bpf_ktime_get_coarse_ns and bpf_timer_* in tracing progs
libbpf: Perform map fd cleanup for gen_loader in case of error
samples/bpf: Fix incorrect use of strlen in xdp_redirect_cpu
tools/runqslower: Fix cross-build
samples/bpf: Fix summary per-sec stats in xdp_sample_user
selftests/bpf: Check map in map pruning
bpf: Fix inner map state pruning regression.
xsk: Fix crash on double free in buffer pool
====================
Link: https://lore.kernel.org/r/20211116141134.6490-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
E-Switch encap mode is relevant only when in switchdev mode.
The RDMA driver can query the encap configuration via
mlx5_eswitch_get_encap_mode(). Make sure it returns the currently
used mode and not the set one.
This reverts the cited commit which reset the encap mode
on entering switchdev and fixes the original issue properly.
Fixes: 9a64144d68 ("net/mlx5: E-Switch, Fix default encap mode")
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
This event is received when the controller stops advertising,
specifically for these three reasons:
(a) Connection is successfully created (success).
(b) Timeout is reached (error).
(c) Number of advertising events is reached (error).
(*) This event is NOT generated when the host stops the advertisement.
Refer to the BT spec ver 5.3 vol 4 part E sec 7.7.65.18. Note that the
section was revised from BT spec ver 5.0 vol 2 part E sec 7.7.65.18
which was ambiguous about (*).
Some chips (e.g. RTL8822CE) send this event when the host stops the
advertisement with status = HCI_ERROR_CANCELLED_BY_HOST (due to (*)
above). This is treated as an error and the advertisement will be
removed and userspace will be informed via MGMT event.
On suspend, we are supposed to temporarily disable advertisements,
and continue advertising on resume. However, due to the behavior
above, the advertisements are removed instead.
This patch returns early if HCI_ERROR_CANCELLED_BY_HOST is received.
Btmon snippet of the unexpected behavior:
@ MGMT Command: Remove Advertising (0x003f) plen 1
Instance: 1
< HCI Command: LE Set Extended Advertising Enable (0x08|0x0039) plen 6
Extended advertising: Disabled (0x00)
Number of sets: 1 (0x01)
Entry 0
Handle: 0x01
Duration: 0 ms (0x00)
Max ext adv events: 0
> HCI Event: LE Meta Event (0x3e) plen 6
LE Advertising Set Terminated (0x12)
Status: Operation Cancelled by Host (0x44)
Handle: 1
Connection handle: 0
Number of completed extended advertising events: 5
> HCI Event: Command Complete (0x0e) plen 4
LE Set Extended Advertising Enable (0x08|0x0039) ncmd 2
Status: Success (0x00)
Signed-off-by: Archie Pusaka <apusaka@chromium.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
This work is no longer necessary since all the code using it has been
converted to use hci_passive_scan/hci_passive_scan_sync.
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
This makes MGMT_OP_SET_CONNEABLE use hci_cmd_sync_queue instead of
use a dedicated connetable_update work.
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
This makes MGMT_OP_SET_DISCOVERABLE use hci_cmd_sync_queue instead of
use a dedicated discoverable_update work.
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
This is distracting really, let's make this simpler,
because many callers had to take care of this
by themselves, even if on x86 this adds more
code than really needed.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net->core.sock_inuse is a per cpu variable (int),
while net->core.prot_inuse is another per cpu variable
of 64 integers.
per cpu allocator tend to place them in very different places.
Grouping them together makes sense, since it makes
updates potentially faster, if hitting the same
cache line.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
MPTCP hard codes it, let us instead provide this helper.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sock_prot_inuse_add() is very small, we can inline it.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move gro code and data from net/core/dev.c to net/core/gro.c
to ease maintenance.
gro_normal_list() and gro_normal_one() are inlined
because they are called from both files.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/core/gro.c will contain all core gro functions,
to shrink net/core/skbuff.c and net/core/dev.c
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This helper is used once, no need to keep it in fat net/core/skbuff.c
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/netdevice.h became too big, move gro stuff
into include/net/gro.h
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sk_rx_dst/sk_rx_dst_ifindex/sk_rx_dst_cookie are read in early demux,
and currently spans two cache lines.
Moving them close to sk_refcnt makes more sense, as only one cache
line is needed.
New layout for this hot cache line is :
struct sock {
struct sock_common __sk_common; /* 0 0x88 */
/* --- cacheline 2 boundary (128 bytes) was 8 bytes ago --- */
struct dst_entry * sk_rx_dst; /* 0x88 0x8 */
int sk_rx_dst_ifindex; /* 0x90 0x4 */
u32 sk_rx_dst_cookie; /* 0x94 0x4 */
socket_lock_t sk_lock; /* 0x98 0x20 */
atomic_t sk_drops; /* 0xb8 0x4 */
int sk_rcvlowat; /* 0xbc 0x4 */
/* --- cacheline 3 boundary (192 bytes) --- */
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp recvmsg() (or rx zerocopy) spends a fair amount of time
freeing skbs after their payload has been consumed.
A typical ~64KB GRO packet has to release ~45 page
references, eventually going to page allocator
for each of them.
Currently, this freeing is performed while socket lock
is held, meaning that there is a high chance that
BH handler has to queue incoming packets to tcp socket backlog.
This can cause additional latencies, because the user
thread has to process the backlog at release_sock() time,
and while doing so, additional frames can be added
by BH handler.
This patch adds logic to defer these frees after socket
lock is released, or directly from BH handler if possible.
Being able to free these skbs from BH handler helps a lot,
because this avoids the usual alloc/free assymetry,
when BH handler and user thread do not run on same cpu or
NUMA node.
One cpu can now be fully utilized for the kernel->user copy,
and another cpu is handling BH processing and skb/page
allocs/frees (assuming RFS is not forcing use of a single CPU)
Tested:
100Gbit NIC
Max throughput for one TCP_STREAM flow, over 10 runs
MTU : 1500
Before: 55 Gbit
After: 66 Gbit
MTU : 4096+(headers)
Before: 82 Gbit
After: 95 Gbit
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_segs_in() can be called from BH, while socket spinlock
is held but socket owned by user, eventually reading these
fields from tcp_get_info()
Found by code inspection, no need to backport this patch
to older kernels.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use INDIRECT_CALL_INET() to avoid an indirect call
when/if CONFIG_RETPOLINE=y
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(struct proto)->sk_forward_alloc is currently only used by MPTCP.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move sk_bind_phc next to sk_peer_lock to fill a hole.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
gso_size can be moved after tclass, to use an existing hole.
(8 bytes saved on 64bit arches)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of using a full netdev_features_t, we can use a single bit,
as sk_route_nocaps is only used to remove NETIF_F_GSO_MASK from
sk->sk_route_cap.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
