The user is allowed to call ->rx_pkt_burst() even without free
mbufs in the pool. In this scenario we'll fail allocating a rep mbuf
on the first iteration (where pkt is still NULL). This would cause us
to deref a NULL pkt (reset refcount and free).
Fix this by checking the pkt before freeing it.
Fixes: a1bdb71a32 ("net/mlx5: fix crash in Rx")
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
RHEL 7.1's GCC for POWER8 reports the following error in one rte_memcpy()
macro call by mlx5:
error: array subscript is above array bounds [-Werror=array-bounds]
It appears to be a GCC bug which can be worked around by making parentheses
more explicit.
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Fixed issue could occur when Mbuf starvation happens in a middle of
reception of a segmented packet. In such a situation, the PMD has to
release all segments of that packet. The end condition was wrong
causing it to free an Mbuf still handled by the NIC.
Fixes: 9964b965ad ("net/mlx5: re-add Rx scatter support")
Reported-by: Yongseok Koh <yskoh@mellanox.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
In mlx5 rx function, the packet_type and ol_flags mbuf fields are not
properly initialized when no rx offload feature is enabled (checksum, l2
tun checksum, vlan_strip, crc). Thus, these fields can have a value
different of 0 depending on their value when the mbuf was freed.
This can result in an incorrect application behavior if invalid
ol_flags/ptype are set, or memory corruptions if IND_ATTACHED_MBUF is
set in ol_flags.
Fixes: 081f7eae24 ("mlx5: process offload flags only when requested")
Signed-off-by: Maxime Leroy <maxime.leroy@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
This commit brings back Rx scatter and related support by the MTU update
function. The maximum number of segments per packet is not a fixed value
anymore (previously MLX5_PMD_SGE_WR_N, set to 4 by default) as it caused
performance issues when fewer segments were actually needed as well as
limitations on the maximum packet size that could be received with the
default mbuf size (supporting at most 8576 bytes).
These limitations are now lifted as the number of SGEs is derived from the
MTU (which implies MRU) at queue initialization and during MTU update.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Compared to its previous incarnation, the software limit on the number of
mbuf segments is no more (previously MLX5_PMD_SGE_WR_N, set to 4 by
default) hence no need for linearization code and related buffers that
permanently consumed a non negligible amount of memory to handle oversized
mbufs.
The resulting code is both lighter and faster.
With the addition of this code, older GCC versions (such
as 4.8.5) may complain about 'wqe' variable being uninitialized, so
initialize it preemptively, even though it is not necessary to do so.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
The space necessary to store segmented packets cannot be known in advance
and must be verified for each of them.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
This feature enables the TX burst function to emit up to 5 packets using
only two work queue entries (WQEs) on devices that support it. Saves PCI
bandwidth and improves performance.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Olga Shern <olgas@mellanox.com>
Implement send inline feature which copies packet data directly into
work queue entries (WQEs) for improved latency. The maximum packet
size and the minimum number of Tx queues to qualify for inline send
are user-configurable.
This feature is effective when HW causes a performance bottleneck.
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Replacing the variable countdown (which depends on the number of
descriptors) with a fixed relative threshold known at compile time improves
performance by reducing the TX queue structure footprint and the amount of
code to manage completions during a burst.
Completions are now requested at most once per burst after threshold is
reached.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>
Mini (compressed) completion queue entries (CQEs) are returned by the
NIC when PCI back pressure is detected, in which case the first CQE64
contains common packet information followed by a number of CQE8
providing the rest, followed by a matching number of empty CQE64
entries to be used by software for decompression.
Before decompression:
0 1 2 6 7 8
+-------+ +---------+ +-------+ +-------+ +-------+ +-------+
| CQE64 | | CQE64 | | CQE64 | | CQE64 | | CQE64 | | CQE64 |
|-------| |---------| |-------| |-------| |-------| |-------|
| ..... | | cqe8[0] | | | . | | | | | ..... |
| ..... | | cqe8[1] | | | . | | | | | ..... |
| ..... | | ....... | | | . | | | | | ..... |
| ..... | | cqe8[7] | | | | | | | | ..... |
+-------+ +---------+ +-------+ +-------+ +-------+ +-------+
After decompression:
0 1 ... 8
+-------+ +-------+ +-------+
| CQE64 | | CQE64 | | CQE64 |
|-------| |-------| |-------|
| ..... | | ..... | . | ..... |
| ..... | | ..... | . | ..... |
| ..... | | ..... | . | ..... |
| ..... | | ..... | | ..... |
+-------+ +-------+ +-------+
This patch does not perform the entire decompression step as it would be
really expensive, instead the first CQE64 is consumed and an internal
context is maintained to interpret the following CQE8 entries directly.
Intermediate empty CQE64 entries are handed back to HW without further
processing.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>
The latest version of Mellanox OFED exposes hardware definitions necessary
to implement data path operation bypassing Verbs. Update the minimum
version requirement to MLNX_OFED >= 3.3 and clean up compatibility checks
for previous releases.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
To keep the data path as efficient as possible, move fields only useful to
the control path into new structure rxq_ctrl.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Inline TX will be fully managed by the PMD after Verbs is bypassed in the
data path. Remove the current code until then.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. RX scatter cannot be maintained during the
transition and will be reimplemented later.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. TX gather cannot be maintained during the
transition and will be reimplemented later.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Except for the first time when memory registration occurs, the lkey is
always cached. Since memory registration is slow and performs system calls,
performance can be improved by moving that code to its own function outside
of the data path so only the lookup code is left in the original inlined
function.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
SR-IOV mode is currently set when dealing with VF devices. PF devices must
be taken into account as well if they have active VFs.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Memory regions are always local with raw Ethernet queues, drop the remote
property as it adds extra processing on the hardware side.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
The behavior of PKT_RX_VLAN_PKT was not very well defined, resulting in
PMDs not advertising the same flags in similar conditions.
Following discussion in [1], introduce 2 new flags PKT_RX_VLAN_STRIPPED
and PKT_RX_QINQ_STRIPPED that are better defined:
PKT_RX_VLAN_STRIPPED: a vlan has been stripped by the hardware and its
tci is saved in mbuf->vlan_tci. This can only happen if vlan stripping
is enabled in the RX configuration of the PMD.
For now, the old flag PKT_RX_VLAN_PKT is kept but marked as deprecated.
It should be removed from applications and PMDs in a future revision.
This patch also updates the drivers. For PKT_RX_VLAN_PKT:
- e1000, enic, i40e, mlx5, nfp, vmxnet3: done, PKT_RX_VLAN_PKT already
had the same meaning than PKT_RX_VLAN_STRIPPED, minor update is
required.
- fm10k: done, PKT_RX_VLAN_PKT already had the same meaning than
PKT_RX_VLAN_STRIPPED, and vlan stripping is always enabled on fm10k.
- ixgbe: modification done (vector and normal), the old flag was set
when a vlan was recognized, even if vlan stripping was disabled.
- the other drivers do not support vlan stripping.
For PKT_RX_QINQ_PKT, it was only supported on i40e, and the behavior was
already correct, so we can reuse the same bit value for
PKT_RX_QINQ_STRIPPED.
[1] http://dpdk.org/ml/archives/dev/2016-April/037837.html,
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Some architectures (ex: Power8) have a cache line size of 128 bytes,
so the drivers should not expect that prefetching the second part of
the mbuf with rte_prefetch0(&m->cacheline1) is valid.
This commit add helpers that can be used by drivers to prefetch the
rx or tx part of the mbuf, whatever the cache line size.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Reviewed-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Introduce rte_mempool_populate_default() which allocates
mempool objects in several memzones.
The mempool header is now always allocated in a specific memzone
(not with its objects). Thanks to this modification, we can remove
many specific behavior that was required when hugepages are not
enabled in case we are using rte_mempool_xmem_create().
This change requires to update how kni and mellanox drivers lookup for
mbuf memory. For now, this will only work if there is only one memory
chunk (like today), but we could make use of rte_mempool_mem_iter() to
support more memory chunks.
We can also remove RTE_MEMPOOL_OBJ_NAME that is not required anymore for
the lookup, as memory chunks are referenced by the mempool.
Note that rte_mempool_create() is still broken (it was the case before)
when there is no hugepages support (rte_mempool_create_xmem() has to be
used). This is fixed in next commit.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Do not use paddr table to store the mempool memory chunks.
This will allow to have several chunks with different virtual addresses.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Now that the mempool objects are chained into a list, we can use it to
browse them. This implies a rework of rte_mempool_obj_iter() API, that
does not need to take as many arguments as before. The previous function
is kept as a private function, and renamed in this commit. It will be
removed in a next commit of the patch series.
The only internal users of this function are the mellanox drivers. The
code is updated accordingly.
Introducing an API compatibility for this function has been considered,
but it is not easy to do without keeping the old code, as the previous
function could also be used to browse elements that were not added in a
mempool. Moreover, the API is already be broken by other patches in this
version.
The library version was already updated in
commit 213af31e09 ("mempool: reduce structure size if no cache needed")
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
This commit removes the const qualifier for the mempool in
rte_mempool_walk() callback prototype.
Indeed, most functions that can be done on a mempool require a non-const
mempool pointer, except the dump and the audit. Therefore, the
mempool_walk() is more useful if the mempool pointer is not const.
This is required by next commit where the mellanox drivers use
rte_mempool_walk() to iterate the mempools, then rte_mempool_obj_iter()
to iterate the objects in each mempool.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Many drivers provide their own implementation of rte_mbuf_raw_alloc(),
duplicating the code. Introduce a new public function in rte_mbuf to
allocate a raw mbuf (uninitialized).
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Once freed, completed mbufs pointers are not set to NULL in the TX queue.
Clean up function must take this into account.
Fixes: 2e22920b85 ("mlx5: support non-scattered Tx and Rx")
Fixes: 7fae69eeff ("mlx4: new poll mode driver")
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
VLAN insertion can be done in hardware when supported in Verbs. A software
fallback is provided otherwise. The software implementation is also used
when multi-packet send is enabled on a queue, as both features are mutually
exclusive.
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Add a new API rte_eth_dev_get_supported_ptypes to query what packet types
can be filled by a given device. The device should be already started or
its PMD RX burst function already decided, since the packet types supported
may vary depending on RX function.
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Change rxq_cq_to_ol_flags() to set checksum flags according to packet type,
so for non L3/L4 packets the mbuf chksum_bad flags will not be set.
Fixes: 67fa62bc67 ("mlx5: support checksum offload")
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
The first and last memory pool elements are usually cache-aligned but not
page-aligned, particularly when using huge pages.
Hardware performance can be improved significantly by registering memory
regions starting and ending on page boundaries.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Avoid dereferencing pointers twice to get to fast Verbs functions by
storing them directly in RX/TX queue structures.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Allows HW to strip the 802.1Q header from incoming frames and report it
through the mbuf structure.
This feature requires MLNX_OFED >= 3.2.
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
The number of available entries in TX rings is taken before performing
completion, effectively making rings smaller than they are and causing
TX performance issues under load.
Fixes: 2e22920b85 ("mlx5: support non-scattered Tx and Rx")
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
When MP to MR cache is full, the last (newest) MR is freed instead of the
first (oldest) one, causing local protection errors during TX.
Fixes: 2e22920b85 ("mlx5: support non-scattered Tx and Rx")
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Pre-registering mbuf memory pools when creating TX queues avoids costly
registrations later in the data path.
Fixes: 2e22920b85 ("mlx5: support non-scattered Tx and Rx")
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Buffers with too many segments are linearized to overcome
MLX5_PMD_SGE_WR_N, unfortunately the last segment is never sent.
Fixes: 3ee8444608 ("mlx5: support scattered Rx and Tx")
Signed-off-by: Jesper Wramberg <jesper.wramberg@gmail.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Indirect mbuf data may come from a different mempool which must be
registered separately as another memory region, otherwise such mbufs cannot
be sent.
Fixes: 2e22920b85 ("mlx5: support non-scattered Tx and Rx")
Signed-off-by: Jesper Wramberg <jesper.wramberg@gmail.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
The new Verbs RSS API is lower-level than the previous one and much more
flexible but requires RX queues to use Work Queues (WQs) internally instead
of Queue Pairs (QPs), which are grouped in an indirection table used by a
new kind of hash RX QPs.
Hash RX QPs and the indirection table together replace the parent RSS QP
while WQs are mostly similar to child QPs.
RSS hash key is not configurable yet.
Summary of changes:
- Individual DPDK RX queues do not store flow properties anymore, this info
is now part of the hash RX queues.
- All functions affecting the parent queue when RSS is enabled or the basic
queues otherwise are modified to affect hash RX queues instead.
- Hash RX queues are also used when a single DPDK RX queue is configured (no
RSS) to remove that special case.
- Hash RX queues and indirection table are created/destroyed when device
is started/stopped in addition to create/destroy flows.
- Contrary to QPs, WQs are moved to the "ready" state before posting RX
buffers, otherwise they are ignored.
- Resource domain information is added to WQs for better performance.
- CQs are not resized anymore when switching between non-SG and SG modes as
it does not work correctly with WQs. Use the largest possible size
instead, since CQ size does not have to be the same as the number of
elements in the RX queue. This also applies to the maximum number of
outstanding WRs in a WQ (max_recv_wr).
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Or Ami <ora@mellanox.com>
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Removing this structure reduces the size of SG and non-SG RX queue elements
significantly to improve performance.
An nice side effect is that the mbuf pointer is now fully stored in
struct rxq_elt instead of relying on the WR ID data offset hack.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Or Ami <ora@mellanox.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
This commit updates mlx5_rx_burst_sp() to use the fast verbs interface for
posting RX buffers just like mlx5_rx_burst(). Doing so avoids a loop in
libmlx5 and an indirect function call through libibverbs.
Note: recv_sg_list() is not implemented in the QP burst API, this commit is
only to prepare transition to the WQ-based API.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
This is the same implementation as mlx4.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>