Skip to content

L
linux
Switch branch/tag
  1. 07 Dec, 2020 12 commits
  3. 06 Dec, 2020 3 commits
  5. 05 Dec, 2020 9 commits
  7. 04 Dec, 2020 16 commits
    • Bongsu Jeon's avatar
      nfc: s3fwrn5: skip the NFC bootloader mode · 4fb7b98c
      Bongsu Jeon authored 
      If there isn't a proper NFC firmware image, Bootloader mode will be
skipped.
Signed-off-by: default avatarBongsu Jeon <bongsu.jeon@samsung.com>
Reviewed-by: default avatarKrzysztof Kozlowski <krzk@kernel.org>
Link: https://lore.kernel.org/r/20201203225257.2446-1-bongsu.jeon@samsung.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      4fb7b98c
    • Jakub Kicinski's avatar
      Merge branch 'perf-optimizations-for-tcp-recv-zerocopy' · 43be3a3c
      Jakub Kicinski authored 
      Arjun Roy says:

====================
Perf. optimizations for TCP Recv. Zerocopy

This patchset contains several optimizations for TCP Recv. Zerocopy.

Summarized:
1. It is possible that a read payload is not exactly page aligned -
that there may exist "straggler" bytes that we cannot map into the
caller's address space cleanly. For this, we allow the caller to
provide as argument a "hybrid copy buffer", turning
getsockopt(TCP_ZEROCOPY_RECEIVE) into a "hybrid" operation that allows
the caller to avoid a subsequent recvmsg() call to read the
stragglers.

2. Similarly, for "small" read payloads that are either below the size
of a page, or small enough that remapping pages is not a performance
win - we allow the user to short-circuit the remapping operations
entirely and simply copy into the buffer provided.

Some of the patches in the middle of this set are refactors to support
this "short-circuiting" optimization.

3. We allow the user to provide a hint that performing a page zap
operation (and the accompanying TLB shootdown) may not be necessary,
for the provided region that the kernel will attempt to map pages
into. This allows us to avoid this expensive operation while holding
the socket lock, which provides a significant performance advantage.

With all of these changes combined, "medium" sized receive traffic
(multiple tens to few hundreds of KB) see significant efficiency gains
when using TCP receive zerocopy instead of regular recvmsg(). For
example, with RPC-style traffic with 32KB messages, there is a roughly
15% efficiency improvement when using zerocopy. Without these changes,
there is a roughly 60-70% efficiency reduction with such messages when
employing zerocopy.
====================

Link: https://lore.kernel.org/r/20201202225349.935284-1-arjunroy.kdev@gmail.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      43be3a3c
    • Arjun Roy's avatar
      net-zerocopy: Defer vm zap unless actually needed. · 94ab9eb9
      Arjun Roy authored 
      Zapping pages is required only if we are calling vm_insert_page into a
region where pages had previously been mapped. Receive zerocopy allows
reusing such regions, and hitherto called zap_page_range() before
calling vm_insert_page() in that range.

zap_page_range() can also be triggered from userspace with
madvise(MADV_DONTNEED). If userspace is configured to call this before
reusing a segment, or if there was nothing mapped at this virtual
address to begin with, we can avoid calling zap_page_range() under the
socket lock. That said, if userspace does not do that, then we are
still responsible for calling zap_page_range().

This patch adds a flag that the user can use to hint to the kernel
that a zap is not required. If the flag is not set, or if an older
user application does not have a flags field at all, then the kernel
calls zap_page_range as before. Also, if the flag is set but a zap is
still required, the kernel performs that zap as necessary. Thus
incorrectly indicating that a zap can be avoided does not change the
correctness of operation. It also increases the batchsize for
vm_insert_pages and prefetches the page struct for the batch since
we're about to bump the refcount.

An alternative mechanism could be to not have a flag, assume by
default a zap is not needed, and fall back to zapping if needed.
However, this would harm performance for older applications for which
a zap is necessary, and thus we implement it with an explicit flag
so newer applications can opt in.

When using RPC-style traffic with medium sized (tens of KB) RPCs, this
change yields an efficency improvement of about 30% for QPS/CPU usage.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      94ab9eb9
    • Arjun Roy's avatar
      net-zerocopy: Set zerocopy hint when data is copied · 0c3936d3
      Arjun Roy authored 
      Set zerocopy hint, event when falling back to copy, so that the
pending data can be efficiently received using zerocopy when
possible.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      0c3936d3
    • Arjun Roy's avatar
      net-zerocopy: Introduce short-circuit small reads. · f21a3c48
      Arjun Roy authored 
      Sometimes, we may call tcp receive zerocopy when inq is 0,
or inq < PAGE_SIZE, or inq is generally small enough that
it is cheaper to copy rather than remap pages.

In these cases, we may want to either return early (inq=0) or
attempt to use the provided copy buffer to simply copy
the received data.

This allows us to save both system call overhead and
the latency of acquiring mmap_sem in read mode for cases where
it would be useless to do so.

This patchset enables this behaviour by:
1. Returning quickly if inq is 0.
2. Attempting to perform a regular copy if a hybrid copybuffer is
   provided and it is large enough to absorb all available bytes.
3. Return quickly if no such buffer was provided and there are less
   than PAGE_SIZE bytes available.

For small RPC ping-pong workloads, normally we would have
1 getsockopt(), 1 recvmsg() and 1 sendmsg() call per RPC. With this
change, we remove the recvmsg() call entirely, reducing the syscall
overhead by about 33%. In testing with small (hundreds of bytes)
RPC traffic, this yields a syscall reduction of about 33% and
an efficiency gain of about 3-5% when defined as QPS/CPU Util.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      f21a3c48
    • Arjun Roy's avatar
      net-zerocopy: Fast return if inq < PAGE_SIZE · 936ced41
      Arjun Roy authored 
      Sometimes, we may call tcp receive zerocopy when inq is 0,
or inq < PAGE_SIZE, in which case we cannot remap pages. In this case,
simply return the appropriate hint for regular copying without taking
mmap_sem.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      936ced41
    • Arjun Roy's avatar
      net-zerocopy: Refactor frag-is-remappable test. · 98917cf0
      Arjun Roy authored 
      Refactor frag-is-remappable test for tcp receive zerocopy. This is
part of a patch set that introduces short-circuited hybrid copies
for small receive operations, which results in roughly 33% fewer
syscalls for small RPC scenarios.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      98917cf0
    • Arjun Roy's avatar
      net-zerocopy: Refactor skb frag fast-forward op. · 7fba5309
      Arjun Roy authored 
      Refactor skb frag fast-forwarding for tcp receive zerocopy. This is
part of a patch set that introduces short-circuited hybrid copies
for small receive operations, which results in roughly 33% fewer
syscalls for small RPC scenarios.

skb_advance_to_frag(), given a skb and an offset into the skb,
iterates from the first frag for the skb until we're at the frag
specified by the offset. Assuming the offset provided refers to how
many bytes in the skb are already read, the returned frag points to
the next frag we may read from, while offset_frag is set to the number
of bytes from this frag that we have already read.

If frag is not null and offset_frag is equal to 0, then we may be able
to map this frag's page into the process address space with
vm_insert_page(). However, if offset_frag is not equal to 0, then we
cannot do so.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      7fba5309
    • Arjun Roy's avatar
      net-tcp: Introduce tcp_recvmsg_locked(). · 2cd81161
      Arjun Roy authored 
      Refactor tcp_recvmsg() by splitting it into locked and unlocked
portions. Callers already holding the socket lock and not using
ERRQUEUE/cmsg/busy polling can simply call tcp_recvmsg_locked().
This is in preparation for a short-circuit copy performed by
TCP receive zerocopy for small (< PAGE_SIZE, or otherwise requested
by the user) reads.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      2cd81161
    • Arjun Roy's avatar
      net-zerocopy: Copy straggler unaligned data for TCP Rx. zerocopy. · 18fb76ed
      Arjun Roy authored 
      When TCP receive zerocopy does not successfully map the entire
requested space, it outputs a 'hint' that the caller should recvmsg().

Augment zerocopy to accept a user buffer that it tries to copy this
hint into - if it is possible to copy the entire hint, it will do so.
This elides a recvmsg() call for received traffic that isn't exactly
page-aligned in size.

This was tested with RPC-style traffic of arbitrary sizes. Normally,
each received message required at least one getsockopt() call, and one
recvmsg() call for the remaining unaligned data.

With this change, almost all of the recvmsg() calls are eliminated,
leading to a savings of about 25%-50% in number of system calls
for RPC-style workloads.
Signed-off-by: default avatarArjun Roy <arjunroy@google.com>
Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      18fb76ed
    • Jakub Kicinski's avatar
      Merge branch 'seg6-add-support-for-srv6-end-dt4-dt6-behavior' · 4be986c8
      Jakub Kicinski authored 
      Andrea Mayer says:

====================
seg6: add support for SRv6 End.DT4/DT6 behavior

This patchset provides support for the SRv6 End.DT4 and End.DT6 (VRF mode)
behaviors.

The SRv6 End.DT4 behavior is used to implement multi-tenant IPv4 L3 VPNs. It
decapsulates the received packets and performs IPv4 routing lookup in the
routing table of the tenant. The SRv6 End.DT4 Linux implementation leverages a
VRF device in order to force the routing lookup into the associated routing
table.
The SRv6 End.DT4 behavior is defined in the SRv6 Network Programming [1].

The Linux kernel already offers an implementation of the SRv6 End.DT6 behavior
which allows us to set up IPv6 L3 VPNs over SRv6 networks. This new
implementation of DT6 is based on the same VRF infrastructure already exploited
for implementing the SRv6 End.DT4 behavior. The aim of the new SRv6 End.DT6 in
VRF mode consists in simplifying the construction of IPv6 L3 VPN services in
the multi-tenant environment.
Currently, the two SRv6 End.DT6 implementations (legacy and VRF mode)
coexist seamlessly and can be chosen according to the context and the user
preferences.

- Patch 1 is needed to solve a pre-existing issue with tunneled packets
  when a sniffer is attached;

- Patch 2 improves the management of the seg6local attributes used by the
  SRv6 behaviors;

- Patch 3 adds support for optional attributes in SRv6 behaviors;

- Patch 4 introduces two callbacks used for customizing the
  creation/destruction of a SRv6 behavior;

- Patch 5 is the core patch that adds support for the SRv6 End.DT4
  behavior;

- Patch 6 introduces the VRF support for SRv6 End.DT6 behavior;

- Patch 7 adds the selftest for SRv6 End.DT4 behavior;

- Patch 8 adds the selftest for SRv6 End.DT6 (VRF mode) behavior.

Regarding iproute2, the support for the new "vrftable" attribute, required by
both SRv6 End.DT4 and End.DT6 (VRF mode) behaviors, is provided in a different
patchset that will follow shortly.

I would like to thank David Ahern for his support during the development of
this patchset.

[1] https://tools.ietf.org/html/draft-ietf-spring-srv6-network-programming
====================

Link: https://lore.kernel.org/r/20201202130517.4967-1-andrea.mayer@uniroma2.itSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      4be986c8
    • Andrea Mayer's avatar
      selftests: add selftest for the SRv6 End.DT6 (VRF) behavior · 2bc03553
      Andrea Mayer authored 
      this selftest is designed for evaluating the new SRv6 End.DT6 (VRF) behavior
used, in this example, for implementing IPv6 L3 VPN use cases.
Signed-off-by: default avatarAndrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: default avatarPaolo Lungaroni <paolo.lungaroni@cnit.it>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      2bc03553
    • Andrea Mayer's avatar
      selftests: add selftest for the SRv6 End.DT4 behavior · 2195444e
      Andrea Mayer authored 
      this selftest is designed for evaluating the new SRv6 End.DT4 behavior
used, in this example, for implementing IPv4 L3 VPN use cases.
Signed-off-by: default avatarAndrea Mayer <andrea.mayer@uniroma2.it>
Reviewed-by: default avatarDavid Ahern <dsahern@kernel.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      2195444e
    • Andrea Mayer's avatar
      seg6: add VRF support for SRv6 End.DT6 behavior · 20a081b7
      Andrea Mayer authored 
      SRv6 End.DT6 is defined in the SRv6 Network Programming [1].

The Linux kernel already offers an implementation of the SRv6
End.DT6 behavior which permits IPv6 L3 VPNs over SRv6 networks. This
implementation is not particularly suitable in contexts where we need to
deploy IPv6 L3 VPNs among different tenants which share the same network
address schemes. The underlying problem lies in the fact that the
current version of DT6 (called legacy DT6 from now on) needs a complex
configuration to be applied on routers which requires ad-hoc routes and
routing policy rules to ensure the correct isolation of tenants.

Consequently, a new implementation of DT6 has been introduced with the
aim of simplifying the construction of IPv6 L3 VPN services in the
multi-tenant environment using SRv6 networks. To accomplish this task,
we reused the same VRF infrastructure and SRv6 core components already
exploited for implementing the SRv6 End.DT4 behavior.

Currently the two End.DT6 implementations coexist seamlessly and can be
used depending on the context and the user preferences. So, in order to
support both versions of DT6 a new attribute (vrftable) has been
introduced which allows us to differentiate the implementation of the
behavior to be used.

A SRv6 End.DT6 legacy behavior is still instantiated using a command
like the following one:

 $ ip -6 route add 2001:db8::1 encap seg6local action End.DT6 table 100 dev eth0

While to instantiate the SRv6 End.DT6 in VRF mode, the command is still
pretty straight forward:

 $ ip -6 route add 2001:db8::1 encap seg6local action End.DT6 vrftable 100 dev eth0.

Obviously as in the case of SRv6 End.DT4, the VRF strict_mode parameter
must be set (net.vrf.strict_mode=1) and the VRF associated with table
100 must exist.

Please note that the instances of SRv6 End.DT6 legacy and End.DT6 VRF
mode can coexist in the same system/configuration without problems.

[1] https://tools.ietf.org/html/draft-ietf-spring-srv6-network-programmingSigned-off-by: default avatarAndrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      20a081b7
    • Andrea Mayer's avatar
      seg6: add support for the SRv6 End.DT4 behavior · 664d6f86
      Andrea Mayer authored 
      SRv6 End.DT4 is defined in the SRv6 Network Programming [1].

The SRv6 End.DT4 is used to implement IPv4 L3VPN use-cases in
multi-tenants environments. It decapsulates the received packets and it
performs IPv4 routing lookup in the routing table of the tenant.

The SRv6 End.DT4 Linux implementation leverages a VRF device in order to
force the routing lookup into the associated routing table.

To make the End.DT4 work properly, it must be guaranteed that the routing
table used for routing lookup operations is bound to one and only one
VRF during the tunnel creation. Such constraint has to be enforced by
enabling the VRF strict_mode sysctl parameter, i.e:
 $ sysctl -wq net.vrf.strict_mode=1.

At JANOG44, LINE corporation presented their multi-tenant DC architecture
using SRv6 [2]. In the slides, they reported that the Linux kernel is
missing the support of SRv6 End.DT4 behavior.

The SRv6 End.DT4 behavior can be instantiated using a command similar to
the following:

 $ ip route add 2001:db8::1 encap seg6local action End.DT4 vrftable 100 dev eth0

We introduce the "vrftable" extension in iproute2 in a following patch.

[1] https://tools.ietf.org/html/draft-ietf-spring-srv6-network-programming
[2] https://speakerdeck.com/line_developers/line-data-center-networking-with-srv6Signed-off-by: default avatarAndrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      664d6f86
    • Andrea Mayer's avatar
      seg6: add callbacks for customizing the creation/destruction of a behavior · cfdf64a0
      Andrea Mayer authored 
      We introduce two callbacks used for customizing the creation/destruction of
a SRv6 behavior. Such callbacks are defined in the new struct
seg6_local_lwtunnel_ops and hereafter we provide a brief description of
them:

 - build_state(...): used for calling the custom constructor of the
   behavior during its initialization phase and after all the attributes
   have been parsed successfully;

 - destroy_state(...): used for calling the custom destructor of the
   behavior before it is completely destroyed.
Signed-off-by: default avatarAndrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      cfdf64a0
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7