| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
If the channels fail to reopen after setting an XDP program, return the
error code instead of 0. A proper fix is still needed, as now any error
while reopening the channels brings the interface down. This patch only
adds error reporting.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
params is unused in mlx5e_init_di_list.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Upon high packet rate with multiple CPUs TX workloads, much of the HCA's
resources are spent on prefetching TX descriptors, thus affecting
transmission rates.
This patch comes to mitigate this problem by moving some workload to the
CPU and reducing the HW data prefetch overhead for small packets (<= 256B).
When forwarding packets with XDP, a packet that is smaller
than a certain size (set to ~256 bytes) would be sent inline within
its WQE TX descrptor (mem-copied), when the hardware tx queue is congested
beyond a pre-defined water-mark.
This is added to better utilize the HW resources (which now makes
one less packet data prefetch) and allow better scalability, on the
account of CPU usage (which now 'memcpy's the packet into the WQE).
To load balance between HW and CPU and get max packet rate, we use
watermarks to detect how much the HW is congested and move the work
loads back and forth between HW and CPU.
Performance:
Tested packet rate for UDP 64Byte multi-stream
over two dual port ConnectX-5 100Gbps NICs.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
* Tested with hyper-threading disabled
XDP_TX:
| | before | after | |
| 24 rings | 51Mpps | 116Mpps | +126% |
| 1 ring | 12Mpps | 12Mpps | same |
XDP_REDIRECT:
** Below is the transmit rate, not the redirection rate
which might be larger, and is not affected by this patch.
| | before | after | |
| 32 rings | 64Mpps | 92Mpps | +43% |
| 1 ring | 6.4Mpps | 6.4Mpps | same |
As we can see, feature significantly improves scaling, without
hurting single ring performance.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The buffers mapping of the Multi-Packet WQEs (of Striding RQ)
is done via UMR posts, one UMR WQE per an RX MPWQE.
A single MPWQE is capable of serving many incoming packets,
usually larger than the budget of a single napi cycle.
Hence, posting a single UMR WQE per napi cycle (and handling its
completion in the next cycle) works fine in many common cases,
but not always.
When an XDP program is loaded, every MPWQE is capable of serving less
packets, to satisfy the packet-per-page requirement.
Thus, for the same number of packets more MPWQEs (and UMR posts)
are needed (twice as much for the default MTU), giving less latency
room for the UMR completions.
In this patch, we add support for multiple outstanding UMR posts,
to allow faster gap closure between consuming MPWQEs and reposting
them back into the WQ.
For better SW and HW locality, we combine the UMR posts in bulks of
(at least) two.
This is expected to improve packet rate in high CPU scale.
Performance test:
As expected, huge improvement in large-scale (48 cores).
xdp_redirect_map, 64B UDP multi-stream.
Redirect from ConnectX-5 100Gbps to ConnectX-6 100Gbps.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz.
Before: Unstable, 7 to 30 Mpps
After: Stable, at 70.5 Mpps
No degradation in other tested scenarios.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|\| |
| | |
| | |
| | |
| | |
| | | |
Conflict resolution of af_smc.c from Stephen Rothwell.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Although XOR hash function can perform very well on some special use
cases, to align with all drivers, mlx5 driver should use Toeplitz hash
by default.
Toeplitz is more stable for the general use case and it is more standard
and reliable.
On top of that, since XOR (MLX5_RX_HASH_FN_INVERTED_XOR8) gives only a
repeated 8 bits pattern. When used for udp tunneling RSS source port
manipulation it results in fixed source port, which will cause bad RSS
spread.
Fixes: 2be6967cdbc9 ("net/mlx5e: Support ETH_RSS_HASH_XOR")
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
XDP programs might change packets data contents which will make the
reported skb checksum (checksum complete) invalid.
When XDP programs are loaded/unloaded set/clear rx RQs
MLX5E_RQ_STATE_NO_CSUM_COMPLETE flag.
Fixes: 86994156c736 ("net/mlx5e: XDP fast RX drop bpf programs support")
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When requested to recover from error, the tx reporter might open new
channels and close the existing ones. Use safe channels switch flow in
order to guarantee opened channels at the end of the recover flow.
For this purpose, define mlx5e_safe_reopen_channels function and use it
within those flows.
Fixes: de8650a82071 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Expose a new helper that wraps the logic for setting the
netdevice's MTU boundaries.
Use it for the different components (Eth, rep, IPoIB).
Set the netdevice min MTU to ETH_MIN_MTU, and the max according
to both the FW capability and the kernel definition.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The LRO WQE size is a constant, obsolete the parameter field
that holds it.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently support only default geneve udp port (6081).
For the tx side, the HW is assisted by SW parsing, which sets the
headers offset to offload tunneled LSO and csum. Note that for udp
tunnels, we don't use special rx offloads, as rss on the outer headers
is enough, we support checksum complete and GRO takes care of
aggregation.
Geneve TSO BW and CPU load results (tested using iperf single tcp
stream).
In this patch we add TSO support over Geneve, so the "before" result
doesn't actually get to using the TSO HW offload even when turned on.
Tested on ConnectX-5, Intel(R) Xeon(R) CPU E5-2660 v2 @2.20GHz.
__________________________________
| Before | After |
|________________|_________________|
| 12.6 Gbits/sec | 21.7 Gbits/sec |
| 100% CPU load | 61.5% CPU load |
|________________|_________________|
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Acked-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Once channels were successfully opened, update tx reporter health state to
healthy. This is needed for the following scenario:
- SQ has an un-recovered error reported to the devlink health, resulting tx
reporter state to be error.
- Current channels (including this SQ) are closed
- New channels are opened
After that flow, the original error was "solved", and tx reporter state
should be healthy. However, as it was resolved as a side effect, and not
via tx reporter recover method, driver needs to inform devlink health
about it.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
|
|
| |
Allow enabling VEPA mode on the HCA's port in legacy devlink mode.
Example:
bridge link set dev ens1f0 hwmode vepa
will turn on VEPA mode on the netdev ens1f0.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When TX reporter was introduced, it took ownership over TX timeout error
handling. this introduced a regression in case TX reporter is not valid
(NET_DEVLINK is not set, or devlink_health_reporter_create failure).
Fix mlx5e_tx_reporter_timeout function so it can be called at all times.
In addition, remove a warning print that indicates that a TX timeout won't
be handled in case of no valid TX reporter.
Fixes: 7d91126b1aea ("net/mlx5e: Add tx timeout support for mlx5e tx reporter")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
| |
Take into a function the common code structure of opening
a side set of channels followed by a call to apply them.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The netfilter conflicts were rather simple overlapping
changes.
However, the cls_tcindex.c stuff was a bit more complex.
On the 'net' side, Cong is fixing several races and memory
leaks. Whilst on the 'net-next' side we have Vlad adding
the rtnl-ness support.
What I've decided to do, in order to resolve this, is revert the
conversion over to using a workqueue that Cong did, bringing us back
to pure RCU. I did it this way because I believe that either Cong's
races don't apply with have Vlad did things, or Cong will have to
implement the race fix slightly differently.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently mlx5 driver creates xdp redirect hw queues unconditionally on
netdevice open, This is great until someone starts redirecting XDP traffic
via ndo_xdp_xmit on mlx5 device and changes the device configuration at
the same time, this might cause crashes, since the other device's napi
is not aware of the mlx5 state change (resources un-availability).
To fix this we must synchronize with other devices napi's on the system.
Added a new flag under mlx5e_priv to determine XDP TX resources are
available, set/clear it up when necessary and use synchronize_rcu()
when the flag is turned off, so other napi's are in-sync with it, before
we actually cleanup the hw resources.
The flag is tested prior to committing to transmit on mlx5e_xdp_xmit, and
it is sufficient to determine if it safe to transmit or not. The other
two internal flags (MLX5E_STATE_OPENED and MLX5E_SQ_STATE_ENABLED) become
unnecessary. Thus, they are removed from data path.
Fixes: 58b99ee3e3eb ("net/mlx5e: Add support for XDP_REDIRECT in device-out side")
Reported-by: Toke Høiland-Jørgensen <toke@redhat.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
With this patch, ndo_tx_timeout callback will be redirected to the tx
reporter in order to detect a tx timeout error and report it to the
devlink health. (The watchdog detects tx timeouts, but the driver verify
the issue still exists before launching any recover method).
In addition, recover from tx timeout in case of lost interrupt was added
to the tx reporter recover method. The tx timeout recover from lost
interrupt is not a new feature in the driver, this patch re-organize the
functionality and move it to the tx reporter recovery flow.
tx timeout example:
(with auto_recover set to false, if set to true, the manual recover and
diagnose sections are irrelevant)
$cat /sys/kernel/debug/tracing/trace
...
devlink_health_report: bus_name=pci dev_name=0000:00:09.0
driver_name=mlx5_core reporter_name=tx: TX timeout on queue: 0, SQ: 0x8a,
CQ: 0x35, SQ Cons: 0x2 SQ Prod: 0x2, usecs since last trans: 14912000
$devlink health show
pci/0000:00:09.0:
name tx
state healthy #err 1 #recover 0 last_dump_ts N/A
parameters:
grace_period 500 auto_recover false
$devlink health diagnose pci/0000:00:09.0 reporter tx -j -p
{
"SQs": [ {
"sqn": 138,
"HW state": 1,
"stopped": true
},{
"sqn": 142,
"HW state": 1,
"stopped": false
} ]
}
$devlink health diagnose pci/0000:00:09.0 reporter tx
SQs:
sqn: 138 HW state: 1 stopped: true
sqn: 142 HW state: 1 stopped: false
$devlink health recover pci/0000:00:09 reporter tx
$devlink health show
pci/0000:00:09.0:
name tx
state healthy #err 1 #recover 1 last_dump_ts N/A
parameters:
grace_period 500 auto_recover false
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Add mlx5e tx reporter to devlink health reporters. This reporter will be
responsible for diagnosing, reporting and recovering of tx errors.
This patch declares the TX reporter operations and creates it using the
devlink health API. Currently, this reporter supports reporting and
recovering from send error CQE only. In addition, it adds diagnose
information for the open SQs.
For a local SQ recover (due to driver error report), in case of SQ recover
failure, the recover operation will be considered as a failure.
For a full tx recover, an attempt to close and open the channels will be
done. If this one passed successfully, it will be considered as a
successful recover.
The SQ recover from error CQE flow is not a new feature in the driver,
this patch re-organize the functions and adapt them for the devlink
health API. For this purpose, move code from en_main.c to a new file
named reporter_tx.c.
Diagnose output:
$devlink health diagnose pci/0000:00:09.0 reporter tx -j -p
{
"SQs": [ {
"sqn": 138,
"HW state": 1,
"stopped": false
},{
"sqn": 142,
"HW state": 1,
"stopped": false
} ]
}
$devlink health diagnose pci/0000:00:09.0 reporter tx
SQs:
sqn: 138 HW state: 1 stopped: false
sqn: 142 HW state: 1 stopped: false
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|\| |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
MLX5E_PFLAG_* definitions were changed from bitmask to enumerated
values. However, in mlx5e_open_rq(), the proper API (MLX5E_GET_PFLAG macro)
was not used to read the flag value of MLX5E_PFLAG_RX_NO_CSUM_COMPLETE.
Fixed it.
Fixes: 8ff57c18e9f6 ("net/mlx5e: Improve ethtool private-flags code structure")
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Representors software stats are basic, this patch is reusing the
mlx5e_fold_sw_stats in representors, which sums up the basic stats64 for a
mlx5e netdevice.
Fixes: 8bfaf07f7806 ("net/mlx5e: Present SW stats when state is not opened")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
mlx5e_grp_sw_update_stats can be called from two threads,
1) ndo_get_stats64
2) get_ethtool_stats
For this reason and to minimize concurrency issue impact on 64bit machines
mlx5e_grp_sw_update_stats folds the software stats into a temporary
variable then copies it to the global driver stats, both ethtool and ndo
statistics callbacks will use the global software stats variable to report
whatever stats they need.
Actually ndo_get_stats64 doesn't need to fold the whole software stats
(mlx5e_grp_sw_update_stats), all it needs is five counters to fill the
rtnl_link_stats64 relevant stats parameter.
Hence this patch introduces a simpler helper function to fold software
stats for ndo_get_stats64 which will work directly on rtnl_link_stats64
stats parameter and not on the global or even temporary mlx5e_sw_stats
variable.
Since now mlx5e_grp_sw_update_stats is not called by ndo_get_stats64 we
can make it static and remove the temp var.
Unlike mlx5e_grp_sw_update_stats the new fold stats function doesn't
need to zero out the output statistics parameter since it is already
done by the stack @dev_get_stats().
This patch is fixing stack usage of mlx5e_grp_sw_update_stats on
x86 gcc-4.9 and higher, the concurrency issue between mlx5's
ndo_get_stats64 and get_ethtool_stats is resolved as well.
Fixes: 8bfaf07f7806 ("net/mlx5e: Present SW stats when state is not opened")
Reported-by: Arnd Bergmann <arnd@arndb.de>
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently we have one cpu in XPS cpumask per tx queue, this is good
enough for default configuration where there is a tx queue per cpu.
However, once configuration changes to use less tx queues, part of the
cpus are not XPS-mapped and so the select queue decision falls back to
hash calculation and balancing is not guaranteed.
Expand XPS cpumask to enable using all cpus even when number of tx
queues is smaller than number of cpus.
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
In the non-linear SKB memory scheme of Striding RQ, a packet header
could cross page boundary. This requires special care in fast path
that costs LoC, additional runtime instructions and branches.
It could happen when the header (up to 256B) does not fit in
a single stride. Avoid this by working with a stride size that fits
the maximum possible header. Stride is increased form 64B to 256B.
Performance:
Tested packet rate for UDP streams, single ring, on ConnectX-5.
Configuration:
Set Striding RQ and LRO ON (to enabled the non-linear SKB scheme).
GRO OFF, early drop by TC rule.
64B: 4x worse memory utilization, no page-crossers headers
- No degradation (5,887,305 pps).
- The reduction in memory utilization is compensated by the saving of
branches tests.
192B: 1.33x worse memory utilization, avoid page-crossers headers
- Before: 5,727,252. After: 5,777,037. ~1% gain.
256B: Same memory util, no page-crossers
- Before: 5,691,885. After: 5,748,007. ~1% gain.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This reverts the devlink health changes from 9/17/2019,
Jiri wants things to be designed differently and it was
agreed that the easiest way to do this is start from the
beginning again.
Commits reverted:
cb5ccfbe73b389470e1dc11061bb185ef4bc9aec
880ee82f0313453ec5a6cb122866ac057263066b
c7af343b4e33578b7de91786a3f639c8cfa0d97b
ff253fedab961b22117a73ab808fcfa9e6852b50
6f9d56132eb6d2603d4273cfc65bed914ec47acb
fcd852c69d776c0f46c8f79e8e431e5cc6ddc7b7
8a66704a13d9713593342e29b4f0c19762f5746b
12bd0dcefe88782ac1c9fff632958dd1b71d27e5
aba25279c10094c5c97d09c3491ca86d00b4ad5e
ce019faa70f81555fa17ebc1d5a03651f2e7e15a
b8c45a033acc607201588f7665ba84207e5149e0
And the follow-on build fix:
o33a0efa4baecd689da9474ce0e8b673eb6931c60
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
With this patch, ndo_tx_timeout callback will be redirected to the TX
reporter in order to detect a TX timeout error and report it to the
devlink health. (The watchdog detects TX timeouts, but the driver verify
the issue still exists before launching any recover method).
In addition, recover from TX timeout in case of lost interrupt was added
to the TX reporter recover method. The TX timeout recover from lost
interrupt is not a new feature in the driver, this patch re-organize the
functionality and move it to the TX reporter recovery flow.
TX timeout example:
(with auto_recover set to false, if set to true, the manual recover and
diagnose sections are irrelevant)
$cat /sys/kernel/debug/tracing/trace
...
devlink_health_report: bus_name=pci dev_name=0000:00:09.0
driver_name=mlx5_core reporter_name=TX: TX timeout on queue: 0, SQ: 0xd8a, CQ:
0x406, SQ Cons: 0x2 SQ Prod: 0x2, usecs since last trans: 13972000
$devlink health diagnose pci/0000:00:09 reporter TX
SQ 0xd8a: HW state: 1, stopped: 1
SQ 0xe44: HW state: 1, stopped: 0
SQ 0xeb4: HW state: 1, stopped: 0
SQ 0xf1f: HW state: 1, stopped: 0
SQ 0xf80: HW state: 1, stopped: 0
SQ 0xfe5: HW state: 1, stopped: 0
$devlink health recover pci/0000:00:09 reporter TX
$devlink health show
pci/0000:00:09.0:
name TX state healthy #err 1 #recover 1 last_dump_ts N/A dump_available false
attributes:
grace_period 500 auto_recover false
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add mlx5e tx reporter to devlink health reporters. This reporter will be
responsible for diagnosing, reporting and recovering of TX errors.
This patch declares the TX reporter operations and allocate it using the
devlink health API. Currently, this reporter supports reporting and
recovering from send error CQE only. In addition, it adds diagnose
information for the open SQs.
For a local SQ recover (due to driver error report), in case of SQ recover
failure, the recover operation will be considered as a failure.
For a full TX recover, an attempt to close and open the channels will be
done. If this one passed successfully, it will be considered as a
successful recover.
The SQ recover from error CQE flow is not a new feature in the driver,
this patch re-organize the functions and adapt them for the devlink
health API. For this purpose, move code from en_main.c to a new file
named reporter_tx.c.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
|
|
|
|
|
|
|
| |
Add ethtool private flag 'xdp_tx_mpwqe' to control the feature
from userspace.
Feature is set ON by default, if supported.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add support for the HW feature of multi-packet WQE in XDP
xmit flow.
The conventional TX descriptor (WQE, Work Queue Element) serves
a single packet. Our HW has support for multi-packet WQE (MPWQE)
in which a single descriptor serves multiple TX packets.
This reduces both the PCI overhead and the CPU cycles wasted on
writing them.
In this patch we add support for the HW feature, which is supported
starting from ConnectX-5.
Performance:
Tested packet rate for UDP 64Byte multi-stream over ConnectX-5 NICs.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
XDP_TX:
We see a huge gain on single port ConnectX-5, and reach the 100 Mpps
milestone.
* Single-port HCA:
Before: 70 Mpps
After: 100 Mpps (+42.8%)
* Dual-port HCA:
Before: 51.7 Mpps
After: 57.3 Mpps (+10.8%)
* In both cases we tested traffic on one port and for now On Dual-port HCAs
we see only small gain, we are working to overcome this bottleneck, but
for the moment only with experimental firmware on dual port HCAs we can
reach the wanted numbers as seen on Single-port HCAs.
XDP_REDIRECT:
Redirect from (A) ConnectX-5 to (B) ConnectX-5.
Due to a setup limitation, (A) and (B) are on different NUMA nodes,
so absolute performance numbers are not optimal.
Note:
Below is the transmit rate of (B), not the redirect rate of (A)
which is in some cases higher.
* (B) is single-port:
Before: 77 Mpps
After: 90 Mpps (+16.8%)
* (B) is dual-port:
Before: 61 Mpps
After: 72 Mpps (+18%)
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
|
|
| |
Each xdp_wqe_info instance describes the number of data-segments
and WQEBBs of the WQE.
This is useful for a downstream patch that adds support for
Multi-Packet TX WQE feature.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
| |
This provides infrastructure to have multiple xdp_info instances
for the same consumer index.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Do not maintain an SQ state bit to indicate whether an
XDP SQ serves redirect operations.
Instead, rely on the fact that such an XDP SQ doesn't reside
in an RQ instance, while the others do.
This info is not known to the XDP SQ functions themselves,
and they rely on their callers to distinguish between the cases.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Lots of conflicts, by happily all cases of overlapping
changes, parallel adds, things of that nature.
Thanks to Stephen Rothwell, Saeed Mahameed, and others
for their guidance in these resolutions.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
TXQ SQ closure is followed by closing the corresponding CQ. A pending
DIM work would try to modify the now non-existing CQ.
This would trigger an error:
[85535.835926] mlx5_core 0000:af:00.0: mlx5_cmd_check:769:(pid 124399):
MODIFY_CQ(0x403) op_mod(0x0) failed, status bad resource state(0x9), syndrome (0x1d7771)
Fix by making sure to cancel any pending DIM work before destroying the SQ.
Fixes: cbce4f444798 ("net/mlx5e: Enable adaptive-TX moderation")
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Add check of MPWQE stride size is within range supported by HW. In case
calculated MPWQE stride size exceed range, linear SKB can't be used and
we should use non linear MPWQE instead.
Fixes: 619a8f2a42f1 ("net/mlx5e: Use linear SKB in Striding RQ")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Some of the sriov ndo calls are needed also on the switchdev mode -
e.g setup VF mac and reading vport stats. Add them to the uplink rep
netdev ops. Same for the UDP tunnel ones, need them there to identify
offloaded udp tunnel ports.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Take care of setup/teardown for the port link, dcb, lag as well as
dealing with port mtu and carrier for e-switch uplink representors.
This is achieved by adding a dedicated profile instance for uplink
representors which includes the enable/disable and more profile routines
which are invoked by the general mlx5e code for netdev attach/detach.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Now, when we have a dedicated uplink representor, the netdev instance
set over the esw manager vport (PF) is of no-use. As such, remove it
once we're on switchdev mode and get it back to life when off switchdev.
This is done by reloading the Ethernet interface as well (we already
do that for the IB interface) from the eswitch code while going in/out
of switchdev mode.
The Eth add/remove entries are modified to act differently when called in
switchdev mode. In this case we only deal with registration of the eth
vport representors. The rep netdevices are created from the eswitch call
to load the registered eth representors.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Remove some last leftovers from using the PF netdev as
the e-switch uplink representor.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently, when running in sriov switchdev mode, we are using the PF
netdevice as the uplink representor, this is problematic from few aspects:
- will break when the PF isn't eswitch manager (e.g smart NIC env)
- misalignment with other NIC switchdev drivers
- makes us have and maintain special code, hurts the driver quality/robustness
- which in turn opens the door for future bugs
As of each and all of the above, we move to have a dedicated netdev representor
for the uplink vport in a similar manner done for for the VF vports.
This includes the following:
1. have an uplink rep netdev as we have for VF reps
2. all reps use same load/unload functions
3. HW stats for uplink based on physical port counters and not vport counters
4. link state for the uplink managed through PAOS and not vport state
5. the uplink rep has sysfs link to the PF PCI function && uses the PF MAC address
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
- Adding new notifier block (struct mlx5_nb) monitor_counters_nb
for handeling MONITOR_COUNTER new event type.
- Adding work queue element: monitor_counters_work for re-arm and
update stats.
- We re-queue the update stat work, only when working over firmware
that doesn't support the monitored counters.
Signed-off-by: Eyal Davidovich <eyald@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Writing 64B CQEs to 128B cache lines results in a RMW operation. Padding
the CQEs to 128B if possible improves performance on 128B cache line
systems like PPC.
Testing on PPC showed up to a 24% improvement in small packet throughput
vs the default behavior, depending on the workload and system topology.
Signed-off-by: Daniel Jurgens <danielj@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently the driver controls flower filters that are installed on its
devices. However, with the introduction of the indirect block
notifications platform the driver may receive control events for filters
that are installed on higher level net devices (e.g. tunnel devices).
Therefore, the driver filter control API will not be able to implicitly
assume the filter's net device.
Explicitly specify the filter's net device, no functional change
Signed-off-by: Oz Shlomo <ozsh@mellanox.com>
Reviewed-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
mlx5-next shared branch with rdma subtree to avoid mlx5 rdma v.s. netdev
conflicts.
Highlights:
1) RDMA ODP (On Demand Paging) improvements and moving ODP logic to
mlx5 RDMA driver
2) Improved mlx5 core driver and device events handling and provided API
for upper layers to subscribe to device events.
3) RDMA only code cleanup from mlx5 core
4) Add helper to get CQE opcode
5) Rework handling of port module events
6) shared mlx5_ifc.h updates to avoid conflicts
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This isn't used anywhere across the mlx5 driver stack,
remove it.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Remove the deprecated mlx5_interface->event mlx5e callback and use new
mlx5 notifier API to subscribe for mlx5 events, handle port change event
as received from FW rather than handling the mlx5 core processed port
change software version event.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
In mlx5 we have three types of usages for EQs,
1. Asynchronous EQs, used internally by mlx5 core for
a. FW command completions
b. FW page requests
c. one EQ for all other Asynchronous events
2. Completion EQs, used for CQ completion (we create one per core)
3. *Special type of EQ (page fault) used for RDMA on demand paging
(ODP).
*The 3rd type shouldn't be special at least in mlx5 core, it is yet
another async events EQ with specific use case, it will be removed in
the next two patches, and will completely move its logic to mlx5_ib,
as it is rdma specific.
In this patch we remove use case (eq type) specific fields from
struct mlx5_eq into a new eq type specific structures.
struct mlx5_eq_async;
truct mlx5_eq_comp;
struct mlx5_eq_pagefault;
Separate between their type specific flows.
In the future we will allow users to create there own generic EQs.
for now we will allow only one for ODP in next patches.
We will introduce event listeners registration API for those who
want to receive mlx5 async events.
After that mlx5 eq handling will be clean from feature/user specific
handling.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Move unnecessary EQ table structures and declaration from the
public include/linux/mlx5/driver.h into the private area of mlx5_core
and into eq.c/eq.h.
Introduce new mlx5 EQ APIs:
mlx5_comp_vectors_count(dev);
mlx5_comp_irq_get_affinity_mask(dev, vector);
And use them from mlx5_ib or mlx5e netdevice instead of direct access to
mlx5_core internal structures.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
irq_info and rmap are EQ properties of the driver, and only needed for
EQ objects, move them to the eq_table EQs database structure.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
|