| Age | Commit message (Collapse) | Author |
|---|
|
Commit 07d44190a389 ("i40e/i40evf: Detect and recover hung queue
scenario") changes i40e_detect_recover_hung() argument type from
i40e_pf* to i40e_vsi* to be shareable by both i40e and i40evf.
Because the i40evf does not exist anymore and the function is
exclusively used by i40e we can revert this change.
Reviewed-by: Michal Schmidt <mschmidt@redhat.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com>
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
As for ice bug fixed by commit b7306b42beaf ("ice: manage interrupts
during poll exit") followed by commit 23be7075b318 ("ice: fix software
generating extra interrupts") I'm seeing the similar issue also with
i40e driver.
In certain situation when busy-loop is enabled together with adaptive
coalescing, the driver occasionally misses that there are outstanding
descriptors to clean when exiting busy poll.
Try to catch the remaining work by triggering a software interrupt
when exiting busy poll. No extra interrupts will be generated when
busy polling is not used.
The issue was found when running sockperf ping-pong tcp test with
adaptive coalescing and busy poll enabled (50 as value busy_pool
and busy_read sysctl knobs) and results in huge latency spikes
with more than 100000us.
The fix is inspired from the ice driver and do the following:
1) During napi poll exit in case of busy-poll (napo_complete_done()
returns false) this is recorded to q_vector that we were in busy
loop.
2) Extends i40e_buildreg_itr() to be able to add an enforced software
interrupt into built value
2) In i40e_update_enable_itr() enforces a software interrupt trigger
if we are exiting busy poll to catch any pending clean-ups
3) Reuses unused 3rd ITR (interrupt throttle) index and set it to
20K interrupts per second to limit the number of these sw interrupts.
Test results
============
Prior:
[root@dell-per640-07 net]# sockperf ping-pong -i 10.9.9.1 --tcp -m 1000 --mps=max -t 120
sockperf: == version #3.10-no.git ==
sockperf[CLIENT] send on:sockperf: using recvfrom() to block on socket(s)
[ 0] IP = 10.9.9.1 PORT = 11111 # TCP
sockperf: Warmup stage (sending a few dummy messages)...
sockperf: Starting test...
sockperf: Test end (interrupted by timer)
sockperf: Test ended
sockperf: [Total Run] RunTime=119.999 sec; Warm up time=400 msec; SentMessages=2438563; ReceivedMessages=2438562
sockperf: ========= Printing statistics for Server No: 0
sockperf: [Valid Duration] RunTime=119.549 sec; SentMessages=2429473; ReceivedMessages=2429473
sockperf: ====> avg-latency=24.571 (std-dev=93.297, mean-ad=4.904, median-ad=1.510, siqr=1.063, cv=3.797, std-error=0.060, 99.0% ci=[24.417, 24.725])
sockperf: # dropped messages = 0; # duplicated messages = 0; # out-of-order messages = 0
sockperf: Summary: Latency is 24.571 usec
sockperf: Total 2429473 observations; each percentile contains 24294.73 observations
sockperf: ---> <MAX> observation = 103294.331
sockperf: ---> percentile 99.999 = 45.633
sockperf: ---> percentile 99.990 = 37.013
sockperf: ---> percentile 99.900 = 35.910
sockperf: ---> percentile 99.000 = 33.390
sockperf: ---> percentile 90.000 = 28.626
sockperf: ---> percentile 75.000 = 27.741
sockperf: ---> percentile 50.000 = 26.743
sockperf: ---> percentile 25.000 = 25.614
sockperf: ---> <MIN> observation = 12.220
After:
[root@dell-per640-07 net]# sockperf ping-pong -i 10.9.9.1 --tcp -m 1000 --mps=max -t 120
sockperf: == version #3.10-no.git ==
sockperf[CLIENT] send on:sockperf: using recvfrom() to block on socket(s)
[ 0] IP = 10.9.9.1 PORT = 11111 # TCP
sockperf: Warmup stage (sending a few dummy messages)...
sockperf: Starting test...
sockperf: Test end (interrupted by timer)
sockperf: Test ended
sockperf: [Total Run] RunTime=119.999 sec; Warm up time=400 msec; SentMessages=2400055; ReceivedMessages=2400054
sockperf: ========= Printing statistics for Server No: 0
sockperf: [Valid Duration] RunTime=119.549 sec; SentMessages=2391186; ReceivedMessages=2391186
sockperf: ====> avg-latency=24.965 (std-dev=5.934, mean-ad=4.642, median-ad=1.485, siqr=1.067, cv=0.238, std-error=0.004, 99.0% ci=[24.955, 24.975])
sockperf: # dropped messages = 0; # duplicated messages = 0; # out-of-order messages = 0
sockperf: Summary: Latency is 24.965 usec
sockperf: Total 2391186 observations; each percentile contains 23911.86 observations
sockperf: ---> <MAX> observation = 195.841
sockperf: ---> percentile 99.999 = 45.026
sockperf: ---> percentile 99.990 = 39.009
sockperf: ---> percentile 99.900 = 35.922
sockperf: ---> percentile 99.000 = 33.482
sockperf: ---> percentile 90.000 = 28.902
sockperf: ---> percentile 75.000 = 27.821
sockperf: ---> percentile 50.000 = 26.860
sockperf: ---> percentile 25.000 = 25.685
sockperf: ---> <MIN> observation = 12.277
Fixes: 0bcd952feec7 ("ethernet/intel: consolidate NAPI and NAPI exit")
Reported-by: Hugo Ferreira <hferreir@redhat.com>
Reviewed-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Fields .caps in i40e_hw and .hw_features in i40e_pf both indicate
capabilities provided by hardware. Move and merge i40e_pf.hw_features
into i40e_hw.caps as this is more appropriate place for them and
adjust their names to I40E_HW_CAP_... convention.
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20231113231047.548659-9-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Convert flags and hw_features fields from i40e_pf from u32 to
bitmaps and their usage to use bit access functions.
Changes:
- Convert "pf_ptr->(flags|hw_features) & FL" to "test_bit(FL, ...)"
- Convert "pf_ptr->(flags|hw_features) |= FL" to "set_bit(FL, ...)"
- Convert "pf_ptr->(flags|hw_features) &= ~FL" to "clear_bit(FL, ...)"
- Rename flag field to bitno in i40e_priv_flags and adjust ethtool
callbacks to work with flags bitmap
- Rename flag names where '_ENABLED'->'_ENA' and '_DISABLED'->'_DIS'
like in ice driver
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20231113231047.548659-7-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Enum type names should not be suffixed by '_t'. Either to use
'typedef enum name name_t' to so plain 'name_t var' instead of
'enum name_t var'.
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20231113231047.548659-6-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Similarly as for ice driver [1] there are also circular header
dependencies in i40e driver:
i40e.h -> i40e_virtchnl_pf.h -> i40e.h
Another issue is that i40e header files does not contain their own
dependencies on other header files (both private and standard) so their
inclusion in .c file require to add these deps in certain order to
that .c file to make it compilable.
Fix both issues by removal the mentioned circular dependency, by filling
i40e headers with their dependencies so they can be placed anywhere in
a source code. Additionally remove bunch of includes from i40e.h super
header file that are not necessary and include i40e.h only in .c files
that really require it.
[1] 649c87c6ff52 ("ice: remove circular header dependencies on ice.h")
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
This patch is inspired from the multi-buffer support in non-zc path for
i40e as well as from the patch to support zc on ice. Each subsequent
frag is added to skb_shared_info of the first frag for possible xdp_prog
use as well to xsk buffer list for accessing the buffers in af_xdp.
For XDP_PASS, new pages are allocated for frags and contents are copied
from memory backed by xsk_buff_pool.
Replace next_to_clean with next_to_process as done in non-zc path and
advance it for every buffer and change the semantics of next_to_clean to
point to the first buffer of a packet. Driver will use next_to_process
in the same way next_to_clean was used previously.
For the non multi-buffer case, next_to_process and next_to_clean will
always be the same since each packet consists of a single buffer.
Signed-off-by: Tirthendu Sarkar <tirthendu.sarkar@intel.com>
Link: https://lore.kernel.org/r/20230719132421.584801-14-maciej.fijalkowski@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
This patch adds multi-buffer support for the i40e_driver.
i40e_clean_rx_irq() is modified to collate all the buffers of a packet
before calling the XDP program. xdp_buff is built for the first frag of
the packet and subsequent frags are added to it. 'next_to_process' is
incremented for all non-EOP frags while 'next_to_clean' stays at the
first descriptor of the packet. XDP program is called only on receiving
EOP frag.
New functions are added for adding frags to xdp_buff and for post
processing of the buffers once the xdp prog has run. For XDP_PASS this
results in a skb with multiple fragments.
i40e_build_skb() builds the skb around xdp buffer that already contains
frags data. So i40e_add_rx_frag() helper function is now removed. Since
fields before 'dataref' in skb_shared_info are cleared during
napi_skb_build(), xdp_update_skb_shared_info() is called to set those.
For i40e_construct_skb(), all the frags data needs to be copied from
xdp_buffer's shared_skb_info to newly constructed skb's shared_skb_info.
This also means 'skb' does not need to be preserved across i40e_napi_poll()
calls and hence is removed from i40e_ring structure.
Previously i40e_alloc_rx_buffers() was called for every 32 cleaned
buffers. For multi-buffers this may not be optimal as there may be more
cleaned buffers in each i40e_clean_rx_irq() call. So this is now called
when at least half of the ring size has been cleaned.
Signed-off-by: Tirthendu Sarkar <tirthendu.sarkar@intel.com>
Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Store xdp_buff on Rx ring struct in preparation for XDP multi-buffer
support. This will allow us to combine fragmented frames across
separate NAPI cycles in the same way as currently skb fragments are
handled. This means that skb pointer on Rx ring will become redundant
and will be removed in a later patch. As a consequence i40e_trace() now
uses xdp instead of skb pointer.
Truesize only needs to be calculated for page sizes bigger than 4k as it
is always half-page for 4k pages. With xdp_buff on ring, frame size can
now be set during xdp_init_buff() and need not be repopulated in each
NAPI call for 4k pages. As a consequence i40e_rx_frame_truesize() is now
used only for bigger pages.
Signed-off-by: Tirthendu Sarkar <tirthendu.sarkar@intel.com>
Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Add a new field called next_to_process in the i40e_ring that is
advanced for every buffer and change the semantics of next_to_clean to
point to the first buffer of a packet. Driver will use next_to_process
in the same way next_to_clean was used previously.
For the non multi-buffer case, next_to_process and next_to_clean will
always be the same since each packet consists of a single buffer.
Signed-off-by: Tirthendu Sarkar <tirthendu.sarkar@intel.com>
Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Page count of rx_buffer needs to be stored prior to XDP call to prevent
page recycling in case that buffer would be freed within xdp redirect
path. Instead of storing it on the stack, now it is stored in the
rx_buffer struct. This will help in processing multi-buffers as the page
counts of all rx_buffers (of the same packet) don't need to be stored on
stack.
Signed-off-by: Tirthendu Sarkar <tirthendu.sarkar@intel.com>
Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
During reallocation of RX buffers, new DMA mappings are created for
those buffers.
steps for reproduction:
while :
do
for ((i=0; i<=8160; i=i+32))
do
ethtool -G enp130s0f0 rx $i tx $i
sleep 0.5
ethtool -g enp130s0f0
done
done
This resulted in crash:
i40e 0000:01:00.1: Unable to allocate memory for the Rx descriptor ring, size=65536
Driver BUG
WARNING: CPU: 0 PID: 4300 at net/core/xdp.c:141 xdp_rxq_info_unreg+0x43/0x50
Call Trace:
i40e_free_rx_resources+0x70/0x80 [i40e]
i40e_set_ringparam+0x27c/0x800 [i40e]
ethnl_set_rings+0x1b2/0x290
genl_family_rcv_msg_doit.isra.15+0x10f/0x150
genl_family_rcv_msg+0xb3/0x160
? rings_fill_reply+0x1a0/0x1a0
genl_rcv_msg+0x47/0x90
? genl_family_rcv_msg+0x160/0x160
netlink_rcv_skb+0x4c/0x120
genl_rcv+0x24/0x40
netlink_unicast+0x196/0x230
netlink_sendmsg+0x204/0x3d0
sock_sendmsg+0x4c/0x50
__sys_sendto+0xee/0x160
? handle_mm_fault+0xbe/0x1e0
? syscall_trace_enter+0x1d3/0x2c0
__x64_sys_sendto+0x24/0x30
do_syscall_64+0x5b/0x1a0
entry_SYSCALL_64_after_hwframe+0x65/0xca
RIP: 0033:0x7f5eac8b035b
Missing register, driver bug
WARNING: CPU: 0 PID: 4300 at net/core/xdp.c:119 xdp_rxq_info_unreg_mem_model+0x69/0x140
Call Trace:
xdp_rxq_info_unreg+0x1e/0x50
i40e_free_rx_resources+0x70/0x80 [i40e]
i40e_set_ringparam+0x27c/0x800 [i40e]
ethnl_set_rings+0x1b2/0x290
genl_family_rcv_msg_doit.isra.15+0x10f/0x150
genl_family_rcv_msg+0xb3/0x160
? rings_fill_reply+0x1a0/0x1a0
genl_rcv_msg+0x47/0x90
? genl_family_rcv_msg+0x160/0x160
netlink_rcv_skb+0x4c/0x120
genl_rcv+0x24/0x40
netlink_unicast+0x196/0x230
netlink_sendmsg+0x204/0x3d0
sock_sendmsg+0x4c/0x50
__sys_sendto+0xee/0x160
? handle_mm_fault+0xbe/0x1e0
? syscall_trace_enter+0x1d3/0x2c0
__x64_sys_sendto+0x24/0x30
do_syscall_64+0x5b/0x1a0
entry_SYSCALL_64_after_hwframe+0x65/0xca
RIP: 0033:0x7f5eac8b035b
This was caused because of new buffers with different RX ring count should
substitute older ones, but those buffers were freed in
i40e_configure_rx_ring and reallocated again with i40e_alloc_rx_bi,
thus kfree on rx_bi caused leak of already mapped DMA.
Fix this by reallocating ZC with rx_bi_zc struct when BPF program loads. Additionally
reallocate back to rx_bi when BPF program unloads.
If BPF program is loaded/unloaded and XSK pools are created, reallocate
RX queues accordingly in XSP_SETUP_XSK_POOL handler.
Fixes: be1222b585fd ("i40e: Separate kernel allocated rx_bi rings from AF_XDP rings")
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Chandan <chandanx.rout@intel.com> (A Contingent Worker at Intel)
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Track TX queue stop events and export the new stat with ethtool.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Daniel Borkmann says:
====================
pull-request: bpf-next 2022-02-09
We've added 126 non-merge commits during the last 16 day(s) which contain
a total of 201 files changed, 4049 insertions(+), 2215 deletions(-).
The main changes are:
1) Add custom BPF allocator for JITs that pack multiple programs into a huge
page to reduce iTLB pressure, from Song Liu.
2) Add __user tagging support in vmlinux BTF and utilize it from BPF
verifier when generating loads, from Yonghong Song.
3) Add per-socket fast path check guarding from cgroup/BPF overhead when
used by only some sockets, from Pavel Begunkov.
4) Continued libbpf deprecation work of APIs/features and removal of their
usage from samples, selftests, libbpf & bpftool, from Andrii Nakryiko
and various others.
5) Improve BPF instruction set documentation by adding byte swap
instructions and cleaning up load/store section, from Christoph Hellwig.
6) Switch BPF preload infra to light skeleton and remove libbpf dependency
from it, from Alexei Starovoitov.
7) Fix architecture-agnostic macros in libbpf for accessing syscall
arguments from BPF progs for non-x86 architectures,
from Ilya Leoshkevich.
8) Rework port members in struct bpf_sk_lookup and struct bpf_sock to be
of 16-bit field with anonymous zero padding, from Jakub Sitnicki.
9) Add new bpf_copy_from_user_task() helper to read memory from a different
task than current. Add ability to create sleepable BPF iterator progs,
from Kenny Yu.
10) Implement XSK batching for ice's zero-copy driver used by AF_XDP and
utilize TX batching API from XSK buffer pool, from Maciej Fijalkowski.
11) Generate temporary netns names for BPF selftests to avoid naming
collisions, from Hangbin Liu.
12) Implement bpf_core_types_are_compat() with limited recursion for
in-kernel usage, from Matteo Croce.
13) Simplify pahole version detection and finally enable CONFIG_DEBUG_INFO_DWARF5
to be selected with CONFIG_DEBUG_INFO_BTF, from Nathan Chancellor.
14) Misc minor fixes to libbpf and selftests from various folks.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (126 commits)
selftests/bpf: Cover 4-byte load from remote_port in bpf_sk_lookup
bpf: Make remote_port field in struct bpf_sk_lookup 16-bit wide
libbpf: Fix compilation warning due to mismatched printf format
selftests/bpf: Test BPF_KPROBE_SYSCALL macro
libbpf: Add BPF_KPROBE_SYSCALL macro
libbpf: Fix accessing the first syscall argument on s390
libbpf: Fix accessing the first syscall argument on arm64
libbpf: Allow overriding PT_REGS_PARM1{_CORE}_SYSCALL
selftests/bpf: Skip test_bpf_syscall_macro's syscall_arg1 on arm64 and s390
libbpf: Fix accessing syscall arguments on riscv
libbpf: Fix riscv register names
libbpf: Fix accessing syscall arguments on powerpc
selftests/bpf: Use PT_REGS_SYSCALL_REGS in bpf_syscall_macro
libbpf: Add PT_REGS_SYSCALL_REGS macro
selftests/bpf: Fix an endianness issue in bpf_syscall_macro test
bpf: Fix bpf_prog_pack build HPAGE_PMD_SIZE
bpf: Fix leftover header->pages in sparc and powerpc code.
libbpf: Fix signedness bug in btf_dump_array_data()
selftests/bpf: Do not export subtest as standalone test
bpf, x86_64: Fail gracefully on bpf_jit_binary_pack_finalize failures
...
====================
Link: https://lore.kernel.org/r/20220209210050.8425-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In some cases, pages cannot be reused by i40e because the page is busy. Add
a counter for this event.
Busy page count is accessible via ethtool.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Dave Switzer <david.switzer@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
In some cases, pages can not be reused because they are not associated with
the correct NUMA zone. Knowing how often pages are waived helps users to
understand the interaction between the driver's memory usage and their
system.
Pass rx_stats through to i40e_can_reuse_rx_page to allow tracking when
pages are waived.
The page waive count is accessible via ethtool.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Dave Switzer <david.switzer@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Add a counter for new page allocations in the i40e RX path. This stat is
accessible with ethtool.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Dave Switzer <david.switzer@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
After commit 1a557afc4dd5 ("i40e: Refactor receive routine"),
rx_stats.realloc_count is no longer being incremented, so remove it.
The debugfs string was left, but hardcoded to 0. This is intended to
prevent breaking any existing code / scripts that are parsing debugfs
for i40e.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Move desc_array from the driver to the pool. The reason behind this is
that we can then reuse this array as a temporary storage for descriptors
in all zero-copy drivers that use the batched interface. This will make
it easier to add batching to more drivers.
i40e is the only driver that has a batched Tx zero-copy
implementation, so no need to touch any other driver.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Link: https://lore.kernel.org/bpf/20220125160446.78976-6-maciej.fijalkowski@intel.com
|
|
In SW DCB mode the packets sent receive incorrect UP tags. They are
constructed correctly and put into tx_ring, but UP is later remapped by
HW on the basis of TCTUPR register contents according to Tx queue
selected, and BW used is consistent with the new UP values. This is
caused by Tx queue selection in kernel not taking into account DCB
configuration. This patch fixes the issue by implementing the
ndo_select_queue NDO callback.
Fixes: fd0a05ce74ef ("i40e: transmit, receive, and NAPI")
Signed-off-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com>
Tested-by: Imam Hassan Reza Biswas <imam.hassan.reza.biswas@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Output of i40e_rx_offset() is based on ethtool's priv flag setting,
which when changed, causes PF reset (disables napi, frees irqs, loads
different Rx mem model, etc.). This means that within napi its result is
constant and there is no reason to call it per each processed frame.
Add new 'rx_offset' field to i40e_ring that is meant to hold the
i40e_rx_offset() result and use it within i40e_clean_rx_irq().
Furthermore, use it within i40e_alloc_mapped_page().
Last but not least, un-inline the function of interest so that compiler
makes the decision about inlining as it lives in .c file.
Reviewed-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Tony Brelinski <tonyx.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Use the new batched xsk interfaces for the Tx path in the i40e driver
to improve performance. On my machine, this yields a throughput
increase of 4% for the l2fwd sample app in xdpsock. If we instead just
look at the Tx part, this patch set increases throughput with above
20% for Tx.
Note that I had to explicitly loop unroll the inner loop to get to
this performance level, by using a pragma. It is honored by both clang
and gcc and should be ignored by versions that do not support
it. Using the -funroll-loops compiler command line switch on the
source file resulted in a loop unrolling on a higher level that
lead to a performance decrease instead of an increase.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/1605525167-14450-6-git-send-email-magnus.karlsson@gmail.com
|
|
This takes care of all of the trivial W=1 fixes in the Intel
Ethernet drivers, which allows developers and maintainers to
build more of the networking tree with more complete warning
checks.
There are three classes of kdoc warnings fixed:
- cannot understand function prototype: 'x'
- Excess function parameter 'x' description in 'y'
- Function parameter or member 'x' not described in 'y'
All of the changes were trivial comment updates on
function headers.
Inspired by Lee Jones' series of wireless work to do the same.
Compile tested only, and passes simple test of
$ git ls-files *.[ch] | egrep drivers/net/ethernet/intel | \
xargs scripts/kernel-doc -none
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The i40e NIC supports two flavors of HW descriptors, 16 and 32
byte. The latter has, obviously, room for more offloading
information. However, the only fields of the 32B HW descriptor that is
being used by the driver, is also available in the 16B descriptor.
In other words; Reading and writing 32 bytes instead of 16 byte is a
waste of bus bandwidth.
This commit starts using 16 byte descriptors instead of 32 byte
descriptors.
For AF_XDP the rx_drop benchmark was improved by 2%.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Replace the explicit umem reference passed to the driver in AF_XDP
zero-copy mode with the buffer pool instead. This in preparation for
extending the functionality of the zero-copy mode so that umems can be
shared between queues on the same netdev and also between netdevs. In
this commit, only an umem reference has been added to the buffer pool
struct. But later commits will add other entities to it. These are
going to be entities that are different between different queue ids
and netdevs even though the umem is shared between them.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-2-git-send-email-magnus.karlsson@intel.com
|
|
Improve the performance of the AF_XDP zero-copy Tx completion
path. When there are no XDP buffers being sent using XDP_TX or
XDP_REDIRECT, we do not have go through the SW ring to clean up any
entries since the AF_XDP path does not use these. In these cases, just
fast forward the next-to-use counter and skip going through the SW
ring. The limit on the maximum number of entries to complete is also
removed since the algorithm is now O(1). To simplify the code path, the
maximum number of entries to complete for the XDP path is therefore
also increased from 256 to 512 (the default number of Tx HW
descriptors). This should be fine since the completion in the XDP path
is faster than in the SKB path that has 256 as the maximum number.
This patch provides around 4% throughput improvement for the l2fwd
application in xdpsock on my machine.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Remove all the unused defines as they are just dead weight.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Remove MEM_TYPE_ZERO_COPY in favor of the new MEM_TYPE_XSK_BUFF_POOL
APIs. The AF_XDP zero-copy rx_bi ring is now simply a struct xdp_buff
pointer.
v4->v5: Fixed "warning: Excess function parameter 'bi' description in
'i40e_construct_skb_zc'". (Jakub)
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Cc: intel-wired-lan@lists.osuosl.org
Link: https://lore.kernel.org/bpf/20200520192103.355233-9-bjorn.topel@gmail.com
|
|
Continuing the path to support MEM_TYPE_XSK_BUFF_POOL, the AF_XDP
zero-copy/sk_buff rx_bi rings are now separate. Functions to properly
allocate the different rings are added as well.
v3->v4: Made i40e_fd_handle_status() static. (kbuild test robot)
v4->v5: Fix kdoc for i40e_clean_programming_status(). (Jakub)
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Cc: intel-wired-lan@lists.osuosl.org
Link: https://lore.kernel.org/bpf/20200520192103.355233-8-bjorn.topel@gmail.com
|
|
In preparation for unifying the skb_frag and bio_vec, use the fine
accessors which already exist and use skb_frag_t instead of
struct skb_frag_struct.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds zero-copy Rx support for AF_XDP sockets. Instead of
allocating buffers of type MEM_TYPE_PAGE_SHARED, the Rx frames are
allocated as MEM_TYPE_ZERO_COPY when AF_XDP is enabled for a certain
queue.
All AF_XDP specific functions are added to a new file, i40e_xsk.c.
Note that when AF_XDP zero-copy is enabled, the XDP action XDP_PASS
will allocate a new buffer and copy the zero-copy frame prior passing
it to the kernel stack.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
Remove the ndo_xdp_flush call implementation i40e_xdp_flush
as no callers of ndo_xdp_flush are left.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
|
|
This patch only change the API and reject any use of flags. This is an
intermediate step that allows us to implement the flush flag operation
later, for each individual driver in a separate patch.
The plan is to implement flush operation via XDP_XMIT_FLUSH flag
and then remove XDP_XMIT_FLAGS_NONE when done.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
This patch change the API for ndo_xdp_xmit to support bulking
xdp_frames.
When kernel is compiled with CONFIG_RETPOLINE, XDP sees a huge slowdown.
Most of the slowdown is caused by DMA API indirect function calls, but
also the net_device->ndo_xdp_xmit() call.
Benchmarked patch with CONFIG_RETPOLINE, using xdp_redirect_map with
single flow/core test (CPU E5-1650 v4 @ 3.60GHz), showed
performance improved:
for driver ixgbe: 6,042,682 pps -> 6,853,768 pps = +811,086 pps
for driver i40e : 6,187,169 pps -> 6,724,519 pps = +537,350 pps
With frames avail as a bulk inside the driver ndo_xdp_xmit call,
further optimizations are possible, like bulk DMA-mapping for TX.
Testing without CONFIG_RETPOLINE show the same performance for
physical NIC drivers.
The virtual NIC driver tun sees a huge performance boost, as it can
avoid doing per frame producer locking, but instead amortize the
locking cost over the bulk.
V2: Fix compile errors reported by kbuild test robot <lkp@intel.com>
V4: Isolated ndo, driver changes and callers.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
After many years of having a ~30 line copyright and license header to our
source files, we are finally able to reduce that to one line with the
advent of the SPDX identifier.
Also caught a few files missing the SPDX license identifier, so fixed
them up.
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@oracle.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Changing API ndo_xdp_xmit to take a struct xdp_frame instead of struct
xdp_buff. This brings xdp_return_frame and ndp_xdp_xmit in sync.
This builds towards changing the API further to become a bulk API,
because xdp_buff is not a queue-able object while xdp_frame is.
V4: Adjust for commit 59655a5b6c83 ("tuntap: XDP_TX can use native XDP")
V7: Adjust for commit d9314c474d4f ("i40e: add support for XDP_REDIRECT")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Also convert driver i40e, which very recently got XDP_REDIRECT support
in commit d9314c474d4f ("i40e: add support for XDP_REDIRECT").
V7: This patch got added in V7 of this patchset.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The driver now acts upon the XDP_REDIRECT return action. Two new ndos
are implemented, ndo_xdp_xmit and ndo_xdp_flush.
XDP_REDIRECT action enables XDP program to redirect frames to other
netdevs.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Add the SPDX identifiers to all the Intel wired LAN driver files, as
outlined in Documentation/process/license-rules.rst.
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The i40e_detect_recover_hung function uses the i40e_get_tx_pending
function to determine if there are packets stalled on the ring.
i40e_get_tx_pending calculates the pending packets using the head
writeback value and HW tail. If the queue is stopped and we lose the
interrupt to update our next_to_clean then we a) won't get another
interrupt to clean because queue is stopped b) we won't catch the
problem with i40e_detect_recover_hung because the HW values look like
there's no packets waiting to be transmitted. Using the SW values we
can catch the issue because next_to_clean will be out of sync with head
writeback.
This has the added benefit being less CPU intensive because we don't
need to reach into the hardware to get the values.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
This patch replaces the existing mechanism for determining the correct
value to program for adaptive ITR with yet another new and more
complicated approach.
The basic idea from a 30K foot view is that this new approach will push the
Rx interrupt moderation up so that by default it starts in low latency and
is gradually pushed up into a higher latency setup as long as doing so
increases the number of packets processed, if the number of packets drops
to 4 to 1 per packet we will reset and just base our ITR on the size of the
packets being received. For Tx we leave it floating at a high interrupt
delay and do not pull it down unless we start processing more than 112
packets per interrupt. If we start exceeding that we will cut our interrupt
rates in half until we are back below 112.
The side effect of these patches are that we will be processing more
packets per interrupt. This is both a good and a bad thing as it means we
will not be blocking processing in the case of things like pktgen and XDP,
but we will also be consuming a bit more CPU in the cases of things such as
network throughput tests using netperf.
One delta from this versus the ixgbe version of the changes is that I have
made the interrupt moderation a bit more aggressive when we are in bulk
mode by moving our "goldilocks zone" up from 48 to 96 to 56 to 112. The
main motivation behind moving this is to address the fact that we need to
update less frequently, and have more fine grained control due to the
separate Tx and Rx ITR times.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
This patch is mostly prep-work for replacing the current approach to
programming the dynamic aka adaptive ITR. Specifically here what we are
doing is splitting the Tx and Rx ITR each into two separate values.
The first value current_itr represents the current value of the register.
The second value target_itr represents the desired value of the register.
The general plan by doing this is to allow for deferring the update of the
ITR value under certain circumstances. For now we will work with what we
have, but in the future I hope to change the behavior so that we always
only update one ITR at a time using some simple logic to determine which
ITR requires an update.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Instead of using the register value for the defines when setting up the
ring ITR we can just use the actual values and avoid the use of shifts and
macros to translate between the values we have and the values we want.
This helps to make the code more readable as we can quickly translate from
one value to the other.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
The rings are already split out into Tx and Rx rings so it doesn't make
sense to have any single ring store both a Tx and Rx itr_setting value.
Since that is the case drop the pair in favor of storing just a single ITR
value.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
When compared to ixgbe and other previous Intel drivers the i40e and i40evf
drivers actually reserve 2 additional descriptors in maybe_stop_tx for
cache line alignment. We need to update DESC_NEEDED to reflect this as
otherwise we are more likely to return TX_BUSY which will cause issues with
things like xmit_more.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In VFs, there is a known issue which can cause writebacks
to not occur when interrupts are disabled and there are
less than 4 descriptors resulting in TX timeout. Timeout
can also occur due to lost interrupt.
The current implementation for detecting and recovering
from hung queues in the PF is problematic because it actually
actively encourages lost interrupts. By triggering a SW
interrupt, interrupts are forced on. If we are already in
napi_poll and an interrupt fires, napi_poll will not be
rescheduled and the interrupt is effectively lost; thereby
potentially *causing* hung queues.
This patch checks whether packets are being processed between
every watchdog cycle and determine potential hung queue and
fires triggers SW interrupt only for that particular queue.
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
The i40e driver has a special "FDIR" RX-ring (I40E_VSI_FDIR) which is
a sideband channel for configuring/updating the flow director tables.
This (i40e_vsi_)type does not invoke XDP-ebpf code.
As suggested by Björn (V2): Instead of marking this I40E_VSI_FDIR RX-ring
a special case, reverse the logic and only select RX-rings of type
I40E_VSI_MAIN to register xdp_rxq_info's for.
Driver hook points for xdp_rxq_info:
* reg : i40e_setup_rx_descriptors (via i40e_vsi_setup_rx_resources)
* unreg: i40e_free_rx_resources (via i40e_vsi_free_rx_resources)
Tested on actual hardware with samples/bpf program.
V2: Fixed bug in i40e_set_ringparam (memset zero) + match on I40E_VSI_MAIN.
V4: Update patch desc that got out-of-sync with code.
Cc: intel-wired-lan@lists.osuosl.org
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
This patch sets up the infrastructure for offloading TCs and
queue configurations to the hardware by creating HW channels(VSI).
A new channel is created for each of the traffic class
configuration offloaded via mqprio framework except for the first TC
(TC0). TC0 for the main VSI is also reconfigured as per user provided
queue parameters. Queue counts that are not power-of-2 are handled by
reconfiguring RSS by reprogramming LUTs using the queue count value.
This patch also handles configuring the TX rings for the channels,
setting up the RX queue map for channel.
Also, the channels so created are removed and all the queue
configuration is set to default when the qdisc is detached from the
root of the device.
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Double the number of descriptors we'll bundle into one tail bump when
receiving. Empirical testing has shown that we reduce CPU utilization
and don't appear to reduce throughput or packet rate. 32 seems to be the
sweet spot, as it's half the default polling budget, so we'd essentially
reduce from 4 tail writes when polling down to 2. Increasing this up to
64 appears to have negative impacts as it may become possible that we
don't bump the tail each time we get polled, which could cause a long
delay between returning descriptors to the hardware.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
The ITR register expects to be programmed in units of 2 microseconds.
Because of this, all of the drivers I40E_ITR_* constants are in terms of
this 2 microsecond register.
Unfortunately, the rx_itr_default value is expected to be programmed in
microseconds.
Effectively the driver defaults to an ITR value of half the expected
value (in terms of minimum microseconds between interrupts).
Fix this by changing the default values to be calculated using
ITR_REG_TO_USEC macro which indicates that we're converting from the
register units into microseconds.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
