| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Unfortunately, the ice driver doesn't respect the RCU critical section that
XSK wakeup is surrounded with. To fix this, add synchronize_rcu() calls to
paths that destroy resources that might be in use.
This was addressed in other AF_XDP ZC enabled drivers, for reference see
for example commit b3873a5be757 ("net/i40e: Fix concurrency issues
between config flow and XSK")
Fixes: efc2214b6047 ("ice: Add support for XDP")
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Shwetha Nagaraju <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
This can happen with big budget values and some breakage of re-filling
descriptors as we do not clear the entry that ntu is pointing at the end
of ice_alloc_rx_bufs_zc. So if ntc is at ntu then it might be the case
that status_error0 has an old, uncleared value and ntc would go over
with processing which would result in false results.
Break Rx loop when ntc == ntu to avoid broken behavior.
Fixes: 3876ff525de7 ("ice: xsk: Handle SW XDP ring wrap and bump tail more often")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
The NIC Tx ring completion routine cleans entries from the ring in
batches. However, it processes one more batch than it is supposed
to. Note that this does not matter from a functionality point of view
since it will not find a set DD bit for the next batch and just exit
the loop. But from a performance perspective, it is faster to
terminate the loop before and not issue an expensive read over PCIe to
get the DD bit.
Fixes: 126cdfe1007a ("ice: xsk: Improve AF_XDP ZC Tx and use batching API")
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Tony Nguyen says:
====================
100GbE Intel Wired LAN Driver Updates 2022-02-09
This series contains updates to ice driver only.
Brett adds support for QinQ. This begins with code refactoring and
re-organization of VLAN configuration functions to allow for
introduction of VSI VLAN ops to enable setting and calling of
respective operations based on device support of single or double
VLANs. Implementations are added for outer VLAN support.
To support QinQ, the device must be set to double VLAN mode (DVM).
In order for this to occur, the DDP package and NVM must also support
DVM. Functions to determine compatibility and properly configure the
device are added as well as setting the proper bits to advertise and
utilize the proper offloads. Support for VIRTCHNL_VF_OFFLOAD_VLAN_V2
is also included to allow for VF to negotiate and utilize this
functionality.
====================
Signed-off-by: David S. Miller <[email protected]>
|
|
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/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Currently the driver only supports 802.1Q VLAN insertion and stripping.
However, once Double VLAN Mode (DVM) is fully supported, then both 802.1Q
and 802.1ad VLAN insertion and stripping will be supported. Unfortunately
the VSI context parameters only allow for one VLAN ethertype at a time
for VLAN offloads so only one or the other VLAN ethertype offload can be
supported at once.
To support this, multiple changes are needed.
Rx path changes:
[1] In DVM, the Rx queue context l2tagsel field needs to be cleared so
the outermost tag shows up in the l2tag2_2nd field of the Rx flex
descriptor. In Single VLAN Mode (SVM), the l2tagsel field should remain
1 to support SVM configurations.
[2] Modify the ice_test_staterr() function to take a __le16 instead of
the ice_32b_rx_flex_desc union pointer so this function can be used for
both rx_desc->wb.status_error0 and rx_desc->wb.status_error1.
[3] Add the new inline function ice_get_vlan_tag_from_rx_desc() that
checks if there is a VLAN tag in l2tag1 or l2tag2_2nd.
[4] In ice_receive_skb(), add a check to see if NETIF_F_HW_VLAN_STAG_RX
is enabled in netdev->features. If it is, then this is the VLAN
ethertype that needs to be added to the stripping VLAN tag. Since
ice_fix_features() prevents CTAG_RX and STAG_RX from being enabled
simultaneously, the VLAN ethertype will only ever be 802.1Q or 802.1ad.
Tx path changes:
[1] In DVM, the VLAN tag needs to be placed in the l2tag2 field of the Tx
context descriptor. The new define ICE_TX_FLAGS_HW_OUTER_SINGLE_VLAN was
added to the list of tx_flags to handle this case.
[2] When the stack requests the VLAN tag to be offloaded on Tx, the
driver needs to set either ICE_TX_FLAGS_HW_OUTER_SINGLE_VLAN or
ICE_TX_FLAGS_HW_VLAN, so the tag is inserted in l2tag2 or l2tag1
respectively. To determine which location to use, set a bit in the Tx
ring flags field during ring allocation that can be used to determine
which field to use in the Tx descriptor. In DVM, always use l2tag2,
and in SVM, always use l2tag1.
Signed-off-by: Brett Creeley <[email protected]>
Tested-by: Gurucharan G <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
For now, if the XDP prog returns XDP_PASS on XSK, the metadata will
be lost as it doesn't get copied to the skb.
Copy it along with the frame headers. Account its size on skb
allocation, and when copying just treat it as a part of the frame
and do a pull after to "move" it to the "reserved" zone.
net_prefetch() xdp->data_meta and align the copy size to speed-up
memcpy() a little and better match ice_construct_skb().
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Suggested-by: Jesper Dangaard Brouer <[email protected]>
Suggested-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Alexander Lobakin <[email protected]>
Reviewed-by: Michal Swiatkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
{__,}napi_alloc_skb() allocates and reserves additional NET_SKB_PAD
+ NET_IP_ALIGN for any skb.
OTOH, ice_construct_skb_zc() currently allocates and reserves
additional `xdp->data - xdp->data_hard_start`, which is
XDP_PACKET_HEADROOM for XSK frames.
There's no need for that at all as the frame is post-XDP and will
go only to the networking stack core.
Pass the size of the actual data only to __napi_alloc_skb() and
don't reserve anything. This will give enough headroom for stack
processing.
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Alexander Lobakin <[email protected]>
Reviewed-by: Michal Swiatkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
One of the things that commit 5574ff7b7b3d ("i40e: optimize AF_XDP Tx
completion path") introduced was the @xdp_tx_active field. Its usage
from i40e can be adjusted to ice driver and give us positive performance
results.
If the descriptor that @next_dd points to has been sent by HW (its DD
bit is set), then we are sure that at least quarter of the ring is ready
to be cleaned. If @xdp_tx_active is 0 which means that related xdp_ring
is not used for XDP_{TX, REDIRECT} workloads, then we know how many XSK
entries should placed to completion queue, IOW walking through the ring
can be skipped.
Signed-off-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Alexander Lobakin <[email protected]>
Acked-by: Magnus Karlsson <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
Apply the logic that was done for regular XDP from commit 9610bd988df9
("ice: optimize XDP_TX workloads") to the ZC side of the driver. On top
of that, introduce batching to Tx that is inspired by i40e's
implementation with adjustments to the cleaning logic - take into the
account NAPI budget in ice_clean_xdp_irq_zc().
Separating the stats structs onto separate cache lines seemed to improve
the performance.
Signed-off-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Alexander Lobakin <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
Currently, if ice_clean_rx_irq_zc() processed the whole ring and
next_to_use != 0, then ice_alloc_rx_buf_zc() would not refill the whole
ring even if the XSK buffer pool would have enough free entries (either
from fill ring or the internal recycle mechanism) - it is because ring
wrap is not handled.
Improve the logic in ice_alloc_rx_buf_zc() to address the problem above.
Do not clamp the count of buffers that is passed to
xsk_buff_alloc_batch() in case when next_to_use + buffer count >=
rx_ring->count, but rather split it and have two calls to the mentioned
function - one for the part up until the wrap and one for the part after
the wrap.
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Alexander Lobakin <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
With the upcoming introduction of batching to XSK data path,
performance wise it will be the best to have the ring descriptor count
to be aligned to power of 2.
Check if ring sizes that user is going to attach the XSK socket fulfill
the condition above. For Tx side, although check is being done against
the Tx queue and in the end the socket will be attached to the XDP
queue, it is fine since XDP queues get the ring->count setting from Tx
queues.
Suggested-by: Alexander Lobakin <[email protected]>
Signed-off-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Alexander Lobakin <[email protected]>
Acked-by: Magnus Karlsson <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
Alexei Starovoitov says:
====================
pull-request: bpf-next 2021-12-30
The following pull-request contains BPF updates for your *net-next* tree.
We've added 72 non-merge commits during the last 20 day(s) which contain
a total of 223 files changed, 3510 insertions(+), 1591 deletions(-).
The main changes are:
1) Automatic setrlimit in libbpf when bpf is memcg's in the kernel, from Andrii.
2) Beautify and de-verbose verifier logs, from Christy.
3) Composable verifier types, from Hao.
4) bpf_strncmp helper, from Hou.
5) bpf.h header dependency cleanup, from Jakub.
6) get_func_[arg|ret|arg_cnt] helpers, from Jiri.
7) Sleepable local storage, from KP.
8) Extend kfunc with PTR_TO_CTX, PTR_TO_MEM argument support, from Kumar.
====================
Signed-off-by: David S. Miller <[email protected]>
|
|
Currently cleaned_count is initialized to ICE_DESC_UNUSED(rx_ring) and
later on during the Rx processing it is incremented per each frame that
driver consumed. This can result in excessive buffers requested from xsk
pool based on that value.
To address this, just drop cleaned_count and pass
ICE_DESC_UNUSED(rx_ring) directly as a function argument to
ice_alloc_rx_bufs_zc(). Idea is to ask for buffers as many as consumed.
Let us also call ice_alloc_rx_bufs_zc unconditionally at the end of
ice_clean_rx_irq_zc. This has been changed in that way for corresponding
ice_clean_rx_irq, but not here.
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Commit ac6f733a7bd5 ("ice: allow empty Rx descriptors") stated that ice
HW can produce empty descriptors that are valid and they should be
processed.
Add this support to xsk ZC path to avoid potential processing problems.
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
The descriptor that ntu is pointing at when we exit
ice_alloc_rx_bufs_zc() should not have its corresponding DD bit cleared
as descriptor is not allocated in there and it is not valid for HW
usage.
The allocation routine at the entry will fill the descriptor that ntu
points to after it was set to ntu + nb_buffs on previous call.
Even the spec says:
"The tail pointer should be set to one descriptor beyond the last empty
descriptor in host descriptor ring."
Therefore, step away from clearing the status_error0 on ntu + nb_buffs
descriptor.
Fixes: db804cfc21e9 ("ice: Use the xsk batched rx allocation interface")
Reported-by: Elza Mathew <[email protected]>
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
The 'if (ntu == rx_ring->count)' block in ice_alloc_rx_buffers_zc()
was previously residing in the loop, but after introducing the
batched interface it is used only to wrap-around the NTU descriptor,
thus no more need to assign 'xdp'.
Fixes: db804cfc21e9 ("ice: Use the xsk batched rx allocation interface")
Signed-off-by: Alexander Lobakin <[email protected]>
Acked-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Currently, the zero-copy data path is reusing the memory region that was
initially allocated for an array of struct ice_rx_buf for its own
purposes. This is error prone as it is based on the ice_rx_buf struct
always being the same size or bigger than what the zero-copy path needs.
There can also be old values present in that array giving rise to errors
when the zero-copy path uses it.
Fix this by freeing the ice_rx_buf region and allocating a new array for
the zero-copy path that has the right length and is initialized to zero.
Fixes: 57f7f8b6bc0b ("ice: Use xdp_buf instead of rx_buf for xsk zero-copy")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Currently we only NULL the xdp_buff pointer in the internal SW ring but
we never give it back to the xsk buffer pool. This means that buffers
can be leaked out of the buff pool and never be used again.
Add missing xsk_buff_free() call to the routine that is supposed to
clean the entries that are left in the ring so that these buffers in the
umem can be used by other sockets.
Also, only go through the space that is actually left to be cleaned
instead of a whole ring.
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
In non trivial scenarios, the action id alone is not sufficient to
identify the program causing the warning. Before the previous patch,
the generated stack-trace pointed out at least the involved device
driver.
Let's additionally include the program name and id, and the relevant
device name.
If the user needs additional infos, he can fetch them via a kernel
probe, leveraging the arguments added here.
Signed-off-by: Paolo Abeni <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Acked-by: Toke Høiland-Jørgensen <[email protected]>
Link: https://lore.kernel.org/bpf/ddb96bb975cbfddb1546cf5da60e77d5100b533c.1638189075.git.pabeni@redhat.com
|
|
Fix a bug in which the receiving of packets can stop in the zero-copy
driver. Ice HW ignores 3 lower bits from QRX_TAIL register, which means
that tail is bumped only on intervals of 8. Currently with XSK RX
batching in place, ice_alloc_rx_bufs_zc() clears the status_error0 only
of the last descriptor that has been allocated/taken from the XSK buffer
pool. status_error0 includes DD bit that is looked upon by the
ice_clean_rx_irq_zc() to tell if a descriptor can be processed.
The bug can be triggered when driver updates the ntu but not the
QRX_TAIL, so HW wouldn't have a chance to write to the ready
descriptors. Later on driver moves the ntc to the mentioned set of
descriptors and interprets them as a ready to be processed, since
corresponding DD bits were not cleared nor any writeback has happened
that would clear it. This can then lead to ntc == ntu case which means
that ring is empty and no further packet processing.
Fix the XSK traffic hang that can be observed when l2fwd scenario from
xdpsock is used by making sure that status_error0 is cleared for each
descriptor that is fed to HW and therefore we are sure that driver will
not processed non-valid DD bits. This will also prevent the driver from
processing the descriptors that were allocated in favor of the
previously processed ones, but writeback didn't happen yet.
Fixes: db804cfc21e9 ("ice: Use the xsk batched rx allocation interface")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Reviewed-by: Alexander Lobakin <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Fix the following coccicheck warning:
./drivers/net/ethernet/intel/i40e/i40e_xsk.c:229:35-40: WARNING:
conversion to bool not needed here
./drivers/net/ethernet/intel/ice/ice_xsk.c:399:35-40: WARNING:
conversion to bool not needed here
Reported-by: Abaci Robot <[email protected]>
Signed-off-by: Yang Li <[email protected]>
Tested-by: Tony Brelinski <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
With rings being split, it is now convenient to introduce a pointer to
XDP ring within the Rx ring. For XDP_TX workloads this means that
xdp_rings array access will be skipped, which was executed per each
processed frame.
Also, read the XDP prog once per NAPI and if prog is present, set up the
local xdp_ring pointer. Reading prog a single time was discussed in [1]
with some concern raised by Toke around dispatcher handling and having
the need for going through the RCU grace period in the ndo_bpf driver
callback, but ice currently is torning down NAPI instances regardless of
the prog presence on VSI.
Although the pointer to XDP ring introduced to Rx ring makes things a
lot slimmer/simpler, I still feel that single prog read per NAPI
lifetime is beneficial.
Further patch that will introduce the fallback path will also get a
profit from that as xdp_ring pointer will be set during the XDP rings
setup.
[1]: https://lore.kernel.org/bpf/[email protected]/
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: George Kuruvinakunnel <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
While it was convenient to have a generic ring structure that served
both Tx and Rx sides, next commits are going to introduce several
Tx-specific fields, so in order to avoid hurting the Rx side, let's
pull out the Tx ring onto new ice_tx_ring and ice_rx_ring structs.
Rx ring could be handled by the old ice_ring which would reduce the code
churn within this patch, but this would make things asymmetric.
Make the union out of the ring container within ice_q_vector so that it
is possible to iterate over newly introduced ice_tx_ring.
Remove the @size as it's only accessed from control path and it can be
calculated pretty easily.
Change definitions of ice_update_ring_stats and
ice_fetch_u64_stats_per_ring so that they are ring agnostic and can be
used for both Rx and Tx rings.
Sizes of Rx and Tx ring structs are 256 and 192 bytes, respectively. In
Rx ring xdp_rxq_info occupies its own cacheline, so it's the major
difference now.
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Gurucharan G <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Use the new xsk batched rx allocation interface for the zero-copy data
path. As the array of struct xdp_buff pointers kept by the driver is
really a ring that wraps, the allocation routine is modified to detect
a wrap and in that case call the allocation function twice. The
allocation function cannot deal with wrapped rings, only arrays. As we
now know exactly how many buffers we get and that there is no
wrapping, the allocation function can be simplified even more as all
if-statements in the allocation loop can be removed, improving
performance.
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
In order to use the new xsk batched buffer allocation interface, a
pointer to an array of struct xsk_buff pointers need to be provided so
that the function can put the result of the allocation there. In the
ice driver, we already have a ring that stores pointers to
xdp_buffs. This is only used for the xsk zero-copy driver and is a
union with the structure that is used for the regular non zero-copy
path. Unfortunately, that structure is larger than the xdp_buffs
pointers which mean that there will be a stride (of 20 bytes) between
each xdp_buff pointer. And feeding this into the xsk_buff_alloc_batch
interface will not work since it assumes a regular array of xdp_buff
pointers (each 8 bytes with 0 bytes in-between them on a 64-bit
system).
To fix this, remove the xdp_buff pointer from the rx_buf union and
move it one step higher to the union above which only has pointers to
arrays in it. This solves the problem and we can directly feed the SW
ring of xdp_buff pointers straight into the allocation function in the
next patch when that interface is used. This will improve performance.
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
Daniel Borkmann says:
====================
pull-request: bpf-next 2021-06-28
The following pull-request contains BPF updates for your *net-next* tree.
We've added 37 non-merge commits during the last 12 day(s) which contain
a total of 56 files changed, 394 insertions(+), 380 deletions(-).
The main changes are:
1) XDP driver RCU cleanups, from Toke Høiland-Jørgensen and Paul E. McKenney.
2) Fix bpf_skb_change_proto() IPv4/v6 GSO handling, from Maciej Żenczykowski.
3) Fix false positive kmemleak report for BPF ringbuf alloc, from Rustam Kovhaev.
4) Fix x86 JIT's extable offset calculation for PROBE_LDX NULL, from Ravi Bangoria.
5) Enable libbpf fallback probing with tracing under RHEL7, from Jonathan Edwards.
6) Clean up x86 JIT to remove unused cnt tracking from EMIT macro, from Jiri Olsa.
7) Netlink cleanups for libbpf to please Coverity, from Kumar Kartikeya Dwivedi.
8) Allow to retrieve ancestor cgroup id in tracing programs, from Namhyung Kim.
9) Fix lirc BPF program query to use user-provided prog_cnt, from Sean Young.
10) Add initial libbpf doc including generated kdoc for its API, from Grant Seltzer.
11) Make xdp_rxq_info_unreg_mem_model() more robust, from Jakub Kicinski.
12) Fix up bpfilter startup log-level to info level, from Gary Lin.
====================
Signed-off-by: David S. Miller <[email protected]>
|
|
The Intel drivers all have rcu_read_lock()/rcu_read_unlock() pairs around
XDP program invocations. However, the actual lifetime of the objects
referred by the XDP program invocation is longer, all the way through to
the call to xdp_do_flush(), making the scope of the rcu_read_lock() too
small. This turns out to be harmless because it all happens in a single
NAPI poll cycle (and thus under local_bh_disable()), but it makes the
rcu_read_lock() misleading.
Rather than extend the scope of the rcu_read_lock(), just get rid of it
entirely. With the addition of RCU annotations to the XDP_REDIRECT map
types that take bh execution into account, lockdep even understands this to
be safe, so there's really no reason to keep it around.
Signed-off-by: Toke Høiland-Jørgensen <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Tested-by: Jesper Dangaard Brouer <[email protected]> # i40e
Cc: Jesse Brandeburg <[email protected]>
Cc: Tony Nguyen <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/bpf/[email protected]
|
|
The hardware is reporting the type of the hash used for RSS
as a PTYPE field in the receive descriptor. Use this value to set
the skb packet hash type by extending the hash type table to
cover all 10-bits of possible values (requiring some variables
to be changed from u8 to u16), and then use that table to convert
to one of the possible values in enum pkt_hash_types.
While we're here, remove the unused ptype struct value, which
makes table init easier for the zero entries, and use ranged
initializer to remove a bunch of code (works with gcc and clang).
Without this change, the kernel will recalculate the hash in software,
which can consume extra CPU cycles.
Co-developed-by: Kiran Patil <[email protected]>
Signed-off-by: Kiran Patil <[email protected]>
Signed-off-by: Jesse Brandeburg <[email protected]>
Tested-by: Tony Brelinski <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Tony Nguyen says:
====================
100GbE Intel Wired LAN Driver Updates 2021-06-07
This series contains updates to virtchnl header file and ice driver.
Brett adds capability bits to virtchnl to specify whether a primary or
secondary MAC address is being requested and adds the implementation to
ice. He also adds storing of VF MAC address so that it will be preserved
across reboots of VM and refactors VF queue configuration to remove the
expectation that configuration be done all at once.
Krzysztof refactors ice_setup_rx_ctx() to remove configuration not
related to Rx context into a new function, ice_vsi_cfg_rxq().
Liwei Song extends the wait time for the global config timeout.
Salil Mehta refactors code in ice_vsi_set_num_qs() to remove an
unnecessary call when the user has requested specific number of Rx or Tx
queues.
Jesse converts define macros to static inlines for NOP configurations.
Jake adds messaging when devlink fails to read device capabilities and
when pldmfw cannot find the requested firmware. Adds a wait for reset
completion when reporting devlink info and reinitializes NVM during
rebuild to ensure values are current.
Ani adds detection and reporting of modules exceeding supported power
levels and changes an error message to a debug message.
Paul fixes a clang warning for deadcode.DeadStores.
====================
Signed-off-by: David S. Miller <[email protected]>
|
|
Move AF_XDP logic and buffer allocation out of ice_setup_rx_ctx() to a
new function ice_vsi_cfg_rxq(), so the function actually sets up the Rx
context.
Signed-off-by: Krzysztof Kazimierczak <[email protected]>
Co-developed-by: Tony Nguyen <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
|
|
Commit c7a219048e45 ("ice: Remove xsk_buff_pool from VSI structure")
silently introduced a regression and broke the Tx side of AF_XDP in copy
mode. xsk_pool on ice_ring is set only based on the existence of the XDP
prog on the VSI which in turn picks ice_clean_tx_irq_zc to be executed.
That is not something that should happen for copy mode as it should use
the regular data path ice_clean_tx_irq.
This results in a following splat when xdpsock is run in txonly or l2fwd
scenarios in copy mode:
<snip>
[ 106.050195] BUG: kernel NULL pointer dereference, address: 0000000000000030
[ 106.057269] #PF: supervisor read access in kernel mode
[ 106.062493] #PF: error_code(0x0000) - not-present page
[ 106.067709] PGD 0 P4D 0
[ 106.070293] Oops: 0000 [#1] PREEMPT SMP NOPTI
[ 106.074721] CPU: 61 PID: 0 Comm: swapper/61 Not tainted 5.12.0-rc2+ #45
[ 106.081436] Hardware name: Intel Corporation S2600WFT/S2600WFT, BIOS SE5C620.86B.02.01.0008.031920191559 03/19/2019
[ 106.092027] RIP: 0010:xp_raw_get_dma+0x36/0x50
[ 106.096551] Code: 74 14 48 b8 ff ff ff ff ff ff 00 00 48 21 f0 48 c1 ee 30 48 01 c6 48 8b 87 90 00 00 00 48 89 f2 81 e6 ff 0f 00 00 48 c1 ea 0c <48> 8b 04 d0 48 83 e0 fe 48 01 f0 c3 66 66 2e 0f 1f 84 00 00 00 00
[ 106.115588] RSP: 0018:ffffc9000d694e50 EFLAGS: 00010206
[ 106.120893] RAX: 0000000000000000 RBX: ffff88984b8c8a00 RCX: ffff889852581800
[ 106.128137] RDX: 0000000000000006 RSI: 0000000000000000 RDI: ffff88984cd8b800
[ 106.135383] RBP: ffff888123b50001 R08: ffff889896800000 R09: 0000000000000800
[ 106.142628] R10: 0000000000000000 R11: ffffffff826060c0 R12: 00000000000000ff
[ 106.149872] R13: 0000000000000000 R14: 0000000000000040 R15: ffff888123b50018
[ 106.157117] FS: 0000000000000000(0000) GS:ffff8897e0f40000(0000) knlGS:0000000000000000
[ 106.165332] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 106.171163] CR2: 0000000000000030 CR3: 000000000560a004 CR4: 00000000007706e0
[ 106.178408] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 106.185653] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 106.192898] PKRU: 55555554
[ 106.195653] Call Trace:
[ 106.198143] <IRQ>
[ 106.200196] ice_clean_tx_irq_zc+0x183/0x2a0 [ice]
[ 106.205087] ice_napi_poll+0x3e/0x590 [ice]
[ 106.209356] __napi_poll+0x2a/0x160
[ 106.212911] net_rx_action+0xd6/0x200
[ 106.216634] __do_softirq+0xbf/0x29b
[ 106.220274] irq_exit_rcu+0x88/0xc0
[ 106.223819] common_interrupt+0x7b/0xa0
[ 106.227719] </IRQ>
[ 106.229857] asm_common_interrupt+0x1e/0x40
</snip>
Fix this by introducing the bitmap of queues that are zero-copy enabled,
where each bit, corresponding to a queue id that xsk pool is being
configured on, will be set/cleared within ice_xsk_pool_{en,dis}able and
checked within ice_xsk_pool(). The latter is a function used for
deciding which napi poll routine is executed.
Idea is being taken from our other drivers such as i40e and ixgbe.
Fixes: c7a219048e45 ("ice: Remove xsk_buff_pool from VSI structure")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Add missing exception tracing to XDP when a number of different
errors can occur. The support was only partial. Several errors
where not logged which would confuse the user quite a lot not
knowing where and why the packets disappeared.
Fixes: efc2214b6047 ("ice: Add support for XDP")
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Reported-by: Jesper Dangaard Brouer <[email protected]>
Signed-off-by: Magnus Karlsson <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Introduce several new helpers for writing ITR and GLINT_RATE
registers, and refactor the code calling them. This resulted
in removal of several duplicate functions and rolled a bunch
of simple code back into the calling routines.
In particular this removes some code that was doing both
a store and a set in a helper function, which seems better
done as separate tasks in the caller (and generally takes
less lines of code even with a tiny bit of repetition).
Signed-off-by: Jesse Brandeburg <[email protected]>
Tested-by: Tony Brelinski <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Remove the leading underscores in enum ice_pf_state. This is not really
communicating anything and is unnecessary. No functional change.
Signed-off-by: Anirudh Venkataramanan <[email protected]>
Tested-by: Tony Brelinski <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Signed-off-by: David S. Miller <[email protected]>
|
|
Optimize ice_run_xdp_zc() for the XDP program verdict being
XDP_REDIRECT in the xsk zero-copy path. This path is only used when
having AF_XDP zero-copy on and in that case most packets will be
directed to user space. This provides a little over 100k extra packets
in throughput on my server when running l2fwd in xdpsock.
Signed-off-by: Magnus Karlsson <[email protected]>
Tested-by: George Kuruvinakunnel <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Fix the wrong napi work done reporting in the xsk path of the ice
driver. The code in the main Rx processing loop was written to assume
that the buffer allocation code returns true if all allocations where
successful and false if not. In contrast with all other Intel NIC xsk
drivers, the ice_alloc_rx_bufs_zc() has the inverted logic messing up
the work done reporting in the napi loop.
This can be fixed either by inverting the return value from
ice_alloc_rx_bufs_zc() in the function that uses this in an incorrect
way, or by changing the return value of ice_alloc_rx_bufs_zc(). We
chose the latter as it makes all the xsk allocation functions for
Intel NICs behave in the same way. My guess is that it was this
unexpected discrepancy that gave rise to this bug in the first place.
Fixes: 5bb0c4b5eb61 ("ice, xsk: Move Rx allocation out of while-loop")
Reported-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Magnus Karlsson <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Whole zero-copy variant of clean Rx IRQ is executed when xsk_pool is
attached to rx_ring and it can happen only when XDP program is present
on interface. Therefore it is safe to assume that program is always
!NULL and there is no need for checking it in ice_run_xdp_zc.
Reviewed-by: Björn Töpel <[email protected]>
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Current implementation of netdev already contains xsk_buff_pools.
We no longer have to contain these structures in ice_vsi.
Refactor the code to operate on netdev-provided xsk_buff_pools.
Move scheduling napi on each queue to a separate function to
simplify setup function.
Signed-off-by: Michal Swiatkowski <[email protected]>
Reviewed-by: Maciej Fijalkowski <[email protected]>
Tested-by: Kiran Bhandare <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
On the Rx side, the next_to_use index points to the next item in the
HW ring to be refilled/allocated, and next_to_clean points to the next
item to potentially be processed.
When the HW Rx ring is fully refilled, i.e. no packets has been
processed, the next_to_use will be next_to_clean - 1. When the ring is
fully processed next_to_clean will be equal to next_to_use. The latter
case is where a bug is triggered.
If the next_to_use bits are not cleared, and the "fully processed"
state is entered, a stale descriptor can be processed.
The skb-path correctly clear the status bit for the next_to_use
descriptor, but the AF_XDP zero-copy path did not do that.
This change adds the status bits clearing of the next_to_use
descriptor.
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Björn Töpel <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Instead doing the check for allocation in each loop, move it outside
the while loop and do it every NAPI loop.
This change boosts the xdpsock rxdrop scenario with 15% more
packets-per-second.
Reviewed-by: Maciej Fijalkowski <[email protected]>
Signed-off-by: Björn Töpel <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
performance
Test for dma_need_sync earlier to increase
performance. xsk_buff_dma_sync_for_cpu() takes an xdp_buff as
parameter and from that the xsk_buff_pool reference is dug out. Perf
shows that this dereference causes a lot of cache misses. But as the
buffer pool is now sent down to the driver at zero-copy initialization
time, we might as well use this pointer directly, instead of going via
the xsk_buff and we can do so already in xsk_buff_dma_sync_for_cpu()
instead of in xp_dma_sync_for_cpu. This gets rid of these cache
misses.
Throughput increases with 3% for the xdpsock l2fwd sample application
on my machine.
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Acked-by: Björn Töpel <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
Rename the AF_XDP zero-copy driver interface functions to better
reflect what they do after the replacement of umems with buffer
pools in the previous commit. Mostly it is about replacing the
umem name from the function names with xsk_buff and also have
them take the a buffer pool pointer instead of a umem. The
various ring functions have also been renamed in the process so
that they have the same naming convention as the internal
functions in xsk_queue.h. This so that it will be clearer what
they do and also for consistency.
Signed-off-by: Magnus Karlsson <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Acked-by: Björn Töpel <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
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 <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Acked-by: Björn Töpel <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
This is a collection of minor fixes including typos, white space, and
style. No functional changes.
Signed-off-by: Tony Nguyen <[email protected]>
Tested-by: Andrew Bowers <[email protected]>
|
|
This is a port of i40e commit 705639572e8c ("i40e: need_wakeup flag might
not be set for Tx").
Quoting the original commit message:
"The need_wakeup flag for Tx might not be set for AF_XDP sockets that
are only used to send packets. This happens if there is at least one
outstanding packet that has not been completed by the hardware and we
get that corresponding completion (which will not generate an interrupt
since interrupts are disabled in the napi poll loop) between the time we
stopped processing the Tx completions and interrupts are enabled again.
In this case, the need_wakeup flag will have been cleared at the end of
the Tx completion processing as we believe we will get an interrupt from
the outstanding completion at a later point in time. But if this
completion interrupt occurs before interrupts are enable, we lose it and
should at that point really have set the need_wakeup flag since there
are no more outstanding completions that can generate an interrupt to
continue the processing. When this happens, user space will see a Tx
queue need_wakeup of 0 and skip issuing a syscall, which means will
never get into the Tx processing again and we have a deadlock."
As a result, packet processing stops. This patch introduces a fix for
this issue, by always setting the need_wakeup flag at the end of an
interrupt processing. This ensures that the deadlock will not happen.
Signed-off-by: Krzysztof Kazimierczak <[email protected]>
Tested-by: Andrew Bowers <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Convert the pre-C90-extension "C struct hack" method (using a single-
element array at the end of a structure for implementing variable-length
types) to the preferred use of C99 flexible array member.
Additional code cleanups were done near areas affected by this change.
Signed-off-by: Bruce Allan <[email protected]>
Tested-by: Andrew Bowers <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Daniel Borkmann says:
====================
pull-request: bpf-next 2020-05-23
The following pull-request contains BPF updates for your *net-next* tree.
We've added 50 non-merge commits during the last 8 day(s) which contain
a total of 109 files changed, 2776 insertions(+), 2887 deletions(-).
The main changes are:
1) Add a new AF_XDP buffer allocation API to the core in order to help
lowering the bar for drivers adopting AF_XDP support. i40e, ice, ixgbe
as well as mlx5 have been moved over to the new API and also gained a
small improvement in performance, from Björn Töpel and Magnus Karlsson.
2) Add getpeername()/getsockname() attach types for BPF sock_addr programs
in order to allow for e.g. reverse translation of load-balancer backend
to service address/port tuple from a connected peer, from Daniel Borkmann.
3) Improve the BPF verifier is_branch_taken() logic to evaluate pointers
being non-NULL, e.g. if after an initial test another non-NULL test on
that pointer follows in a given path, then it can be pruned right away,
from John Fastabend.
4) Larger rework of BPF sockmap selftests to make output easier to understand
and to reduce overall runtime as well as adding new BPF kTLS selftests
that run in combination with sockmap, also from John Fastabend.
5) Batch of misc updates to BPF selftests including fixing up test_align
to match verifier output again and moving it under test_progs, allowing
bpf_iter selftest to compile on machines with older vmlinux.h, and
updating config options for lirc and v6 segment routing helpers, from
Stanislav Fomichev, Andrii Nakryiko and Alan Maguire.
6) Conversion of BPF tracing samples outdated internal BPF loader to use
libbpf API instead, from Daniel T. Lee.
7) Follow-up to BPF kernel test infrastructure in order to fix a flake in
the XDP selftests, from Jesper Dangaard Brouer.
8) Minor improvements to libbpf's internal hashmap implementation, from
Ian Rogers.
====================
Signed-off-by: David S. Miller <[email protected]>
|
|
To make the function easier to identify as being part of the ice driver,
prepend ice to the function name.
Signed-off-by: Tony Nguyen <[email protected]>
Tested-by: Andrew Bowers <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
|