This is a cleanup patch, making code a bit more concise.
1) Use skb_network_offset(skb) in place of
(skb_network_header(skb) - skb->data)
2) Use -skb_network_offset(skb) in place of
(skb->data - skb_network_header(skb))
3) Use skb_transport_offset(skb) in place of
(skb_transport_header(skb) - skb->data)
4) Use skb_inner_transport_offset(skb) in place of
(skb_inner_transport_header(skb) - skb->data)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Edward Cree <ecree.xilinx@gmail.com> # for sfc
Signed-off-by: David S. Miller <davem@davemloft.net>
When running workloads heavy unbalanced towards TX (high TX, low RX
traffic), sfc driver can retain the CPU during too long times. Although
in many cases this is not enough to be visible, it can affect
performance and system responsiveness.
A way to reproduce it is to use a debug kernel and run some parallel
netperf TX tests. In some systems, this will lead to this message being
logged:
kernel:watchdog: BUG: soft lockup - CPU#12 stuck for 22s!
The reason is that sfc driver doesn't account any NAPI budget for the TX
completion events work. With high-TX/low-RX traffic, this makes that the
CPU is held for long time for NAPI poll.
Documentations says "drivers can process completions for any number of Tx
packets but should only process up to budget number of Rx packets".
However, many drivers do limit the amount of TX completions that they
process in a single NAPI poll.
In the same way, this patch adds a limit for the TX work in sfc. With
the patch applied, the watchdog warning never appears.
Tested with netperf in different combinations: single process / parallel
processes, TCP / UDP and different sizes of UDP messages. Repeated the
tests before and after the patch, without any noticeable difference in
network or CPU performance.
Test hardware:
Intel(R) Xeon(R) CPU E5-1620 v4 @ 3.50GHz (4 cores, 2 threads/core)
Solarflare Communications XtremeScale X2522-25G Network Adapter
Fixes: 5227ecccea ("sfc: remove tx and MCDI handling from NAPI budget consideration")
Fixes: d19a537218 ("sfc_ef100: TX path for EF100 NICs")
Reported-by: Fei Liu <feliu@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20230615084929.10506-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move declarations into include/net/gso.h and code into net/core/gso.c
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stanislav Fomichev <sdf@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20230608191738.3947077-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A non-null efv in __ef100_enqueue_skb() indicates that the packet is
from that representor, should be transmitted with a suitable option
descriptor (to instruct the switch to deliver it to the representee),
and should not be accounted to the parent PF's stats or BQL.
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code for gso_max_size was added originally to allow for debugging and
workaround of buggy devices that couldn't support TSO with blocks 64K in
size. The original reason for limiting it to 64K was because that was the
existing limits of IPv4 and non-jumbogram IPv6 length fields.
With the addition of Big TCP we can remove this limit and allow the value
to potentially go up to UINT_MAX and instead be limited by the tso_max_size
value.
So in order to support this we need to go through and clean up the
remaining users of the gso_max_size value so that the values will cap at
64K for non-TCPv6 flows. In addition we can clean up the GSO_MAX_SIZE value
so that 64K becomes GSO_LEGACY_MAX_SIZE and UINT_MAX will now be the upper
limit for GSO_MAX_SIZE.
v6: (edumazet) fixed a compile error if CONFIG_IPV6=n,
in a new sk_trim_gso_size() helper.
netif_set_tso_max_size() caps the requested TSO size
with GSO_MAX_SIZE.
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In some cases, xdp tx_queue can get used before initialization.
1. interface up/down
2. ring buffer size change
When CPU cores are lower than maximum number of channels of sfc driver,
it creates new channels only for XDP.
When an interface is up or ring buffer size is changed, all channels
are initialized.
But xdp channels are always initialized later.
So, the below scenario is possible.
Packets are received to rx queue of normal channels and it is acted
XDP_TX and tx_queue of xdp channels get used.
But these tx_queues are not initialized yet.
If so, TX DMA or queue error occurs.
In order to avoid this problem.
1. initializes xdp tx_queues earlier than other rx_queue in
efx_start_channels().
2. checks whether tx_queue is initialized or not in efx_xdp_tx_buffers().
Splat looks like:
sfc 0000:08:00.1 enp8s0f1np1: TX queue 10 spurious TX completion id 250
sfc 0000:08:00.1 enp8s0f1np1: resetting (RECOVER_OR_ALL)
sfc 0000:08:00.1 enp8s0f1np1: MC command 0x80 inlen 100 failed rc=-22
(raw=22) arg=789
sfc 0000:08:00.1 enp8s0f1np1: has been disabled
Fixes: f28100cb9c ("sfc: fix lack of XDP TX queues - error XDP TX failed (-22)")
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the tx_queue->handle_tso function pointer, and just use
tx_queue->tso_version to decide which function to call, thus removing
an indirect call from the fast path.
Instead of passing a tso_v2 flag to efx_mcdi_tx_init(), set the desired
tx_queue->tso_version before calling it.
In efx_mcdi_tx_init(), report back failure to obtain a TSOv2 context by
setting tx_queue->tso_version to 0, which will cause the TX path to
use the GSO-based fallback.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make it possible to have an arbitrary mapping from types to labels,
because when we add inner-csum-offload TXQs there will no longer be a
convenient nesting hierarchy of NIC types (EF10 will have inner-csum
TXQs, while Siena will have HIGHPRI).
Correct a misleading comment on efx_hard_start_xmit().
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of open-coding the calculation with efx_tx_queue_partner(), use
the functions that iterate over numbers of queues other than 2 with
efx_for_each_channel_tx_queue().
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Instead of using efx_tx_queue_partner(), which relies on the assumption
that tx_queues_per_channel is 2, efx_tx_send_pending() iterates over
txqs with efx_for_each_channel_tx_queue().
We unconditionally set tx_queue->xmit_pending (renamed from
xmit_more_available), then condition on xmit_more for the call to
efx_tx_send_pending(), which will clear xmit_pending. Thus, after an
xmit_more TX, the doorbell is un-rung and xmit_pending is true.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Includes checksum offload and TSO, so declare those in our netdev features.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ef100 will need to check this against NIC limits.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define it in nic_common.h, even though the ef100 driver will have a
different implementation backing it (actually a WARN_ON_ONCE as it
should never get called by ef100. But it needs to still exist because
common TX path code references it).
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ef100 will need this if it gets GSO skbs it can't handle (e.g. too long
header length).
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We now ignore the "completion" event when using tx queue timestamping,
and only pay attention to the two (high and low) timestamp events. The
NIC will send a pair of timestamp events for every packet transmitted.
The current firmware may merge the completion events, and it is possible
that future versions may reorder the completion and timestamp events.
As such the completion event is not useful.
Without this patch in place a merged completion event on a queue with
timestamping will cause a "spurious TX completion" error. This affects
SFN8000-series adapters.
Signed-off-by: Tom Zhao <tzhao@solarflare.com>
Acked-by: Martin Habets <mhabets@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code that handles transmission finalization will also be common.
Signed-off-by: Alexandru-Mihai Maftei <amaftei@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Once again, a tiny bit of refactoring was required to stitch the code
together (i.e. adding headers). The moved code deals with managing tx
queues and mappings.
Signed-off-by: Alexandru-Mihai Maftei <amaftei@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
