The code for gso_max_size was added originally to allow for debugging and
workaround of buggy devices that couldn't support TSO with blocks 64K in
size. The original reason for limiting it to 64K was because that was the
existing limits of IPv4 and non-jumbogram IPv6 length fields.
With the addition of Big TCP we can remove this limit and allow the value
to potentially go up to UINT_MAX and instead be limited by the tso_max_size
value.
So in order to support this we need to go through and clean up the
remaining users of the gso_max_size value so that the values will cap at
64K for non-TCPv6 flows. In addition we can clean up the GSO_MAX_SIZE value
so that 64K becomes GSO_LEGACY_MAX_SIZE and UINT_MAX will now be the upper
limit for GSO_MAX_SIZE.
v6: (edumazet) fixed a compile error if CONFIG_IPV6=n,
in a new sk_trim_gso_size() helper.
netif_set_tso_max_size() caps the requested TSO size
with GSO_MAX_SIZE.
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove napi_weight statics which are set to 64 and never modified,
remnants of the out-of-tree napi_weight module param.
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220512205603.1536771-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If CONFIG_PTP_1588_CLOCK=m and CONFIG_SFC_SIENA=y, the siena driver will fail to link:
drivers/net/ethernet/sfc/siena/ptp.o: In function `efx_ptp_remove_channel':
ptp.c:(.text+0xa28): undefined reference to `ptp_clock_unregister'
drivers/net/ethernet/sfc/siena/ptp.o: In function `efx_ptp_probe_channel':
ptp.c:(.text+0x13a0): undefined reference to `ptp_clock_register'
ptp.c:(.text+0x1470): undefined reference to `ptp_clock_unregister'
drivers/net/ethernet/sfc/siena/ptp.o: In function `efx_ptp_pps_worker':
ptp.c:(.text+0x1d29): undefined reference to `ptp_clock_event'
drivers/net/ethernet/sfc/siena/ptp.o: In function `efx_siena_ptp_get_ts_info':
ptp.c:(.text+0x301b): undefined reference to `ptp_clock_index'
To fix this build error, make SFC_SIENA depends on PTP_1588_CLOCK.
Reported-by: Hulk Robot <hulkci@huawei.com>
Fixes: d48523cb88 ("sfc: Copy shared files needed for Siena (part 2)")
Signed-off-by: Ren Zhijie <renzhijie2@huawei.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220513012721.140871-1-renzhijie2@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
They were removed in the first series since they were not used for EF10.
Put that code back for Siena, with the prototypes in siena_sriov.h
since that file is a more applicable place for it.
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Change the clock name and work queue names to differentiate them from
the names used in sfc.ko.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a Siena Kconfig option and use it in stead of the sfc one.
Rename the internal variable for the 'mcdi_logging_default' module
parameter to avoid a naming conflict with the one in sfc.ko.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a Siena Kconfig option and use it in stead of the sfc one.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a Siena Kconfig option and use it in stead of the sfc one.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a Siena Kconfig option and use it in stead of the sfc one.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In the NIC ->probe() callback, ->mtd_probe() callback is called.
If NIC has 2 ports, ->probe() is called twice and ->mtd_probe() too.
In the ->mtd_probe(), which is efx_ef10_mtd_probe() it allocates and
initializes mtd partiion.
But mtd partition for sfc is shared data.
So that allocated mtd partition data from last called
efx_ef10_mtd_probe() will not be used.
Therefore it must be freed.
But it doesn't free a not used mtd partition data in efx_ef10_mtd_probe().
kmemleak reports:
unreferenced object 0xffff88811ddb0000 (size 63168):
comm "systemd-udevd", pid 265, jiffies 4294681048 (age 348.586s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<ffffffffa3767749>] kmalloc_order_trace+0x19/0x120
[<ffffffffa3873f0e>] __kmalloc+0x20e/0x250
[<ffffffffc041389f>] efx_ef10_mtd_probe+0x11f/0x270 [sfc]
[<ffffffffc0484c8a>] efx_pci_probe.cold.17+0x3df/0x53d [sfc]
[<ffffffffa414192c>] local_pci_probe+0xdc/0x170
[<ffffffffa4145df5>] pci_device_probe+0x235/0x680
[<ffffffffa443dd52>] really_probe+0x1c2/0x8f0
[<ffffffffa443e72b>] __driver_probe_device+0x2ab/0x460
[<ffffffffa443e92a>] driver_probe_device+0x4a/0x120
[<ffffffffa443f2ae>] __driver_attach+0x16e/0x320
[<ffffffffa4437a90>] bus_for_each_dev+0x110/0x190
[<ffffffffa443b75e>] bus_add_driver+0x39e/0x560
[<ffffffffa4440b1e>] driver_register+0x18e/0x310
[<ffffffffc02e2055>] 0xffffffffc02e2055
[<ffffffffa3001af3>] do_one_initcall+0xc3/0x450
[<ffffffffa33ca574>] do_init_module+0x1b4/0x700
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Fixes: 8127d661e7 ("sfc: Add support for Solarflare SFC9100 family")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Link: https://lore.kernel.org/r/20220512054709.12513-1-ap420073@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make the (un)load message more specific to differentiate it from
the sfc.ko messages.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The implementation of each is quite short. This means sriov.c is
not needed any more.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For siena use efx_siena_ as the function prefix.
efx_nic_update_stats_atomic is only used in efx_common.c, so move
it there.
efx_nic_copy_stats is not used in Siena, so it is removed.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For siena use efx_siena_ as the function prefix.
Several functions are not used in Siena, so they are removed.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For siena use efx_siena_ as the function prefix.
This patch covers selftest.h, ptp.h, net_driver.h and ethtool_common.h.
efx_ethtool_fill_self_tests() can become static.
Some functions in ptp.c can also become static.
Rename loopback_mode in net_driver.h.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For siena use efx_siena_ as the function prefix.
Several functions are not used in Siena, so they are removed.
Use a Siena specific variable name for module parameter
efx_separate_tx_channels.
Move efx_fini_tx_queue() to avoid a forward declaration of
efx_dequeue_buffer().
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When building with allyesconfig there are many identical
symbol names.
For siena use efx_siena_ as the function and variable prefix
to avoid build errors.
efx_mtd_remove_partition can become static as it is no longer called
from other files.
efx_ticks_to_usecs and efx_xmit_done_single are not used in Siena, so
they are removed.
Several functions are only used inside efx_channels.c for Siena so
they can become static.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Functionality not supported or needed on Siena includes:
- Anything for EF100
- EF10 specifics such as register access, PIO and TSO offload.
Also only bind to Siena NICs.
Remove EF10 specifics from nic.h.
The functions that start with efx_farch_ will be removed from sfc.ko
with a subsequent patch.
Add the efx_ prefix to siena_prepare_flush() to make it consistent
with the other APIs.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
These are the files starting with m through w.
No changes are done, those will be done with subsequent commits.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
These are the files starting with b through i.
No changes are done, those will be done with subsequent commits.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It fixes memory leak in ring buffer change logic.
When ring buffer size is changed(ethtool -G eth0 rx 4096), sfc driver
works like below.
1. stop all channels and remove ring buffers.
2. allocates new buffer array.
3. allocates rx buffers.
4. start channels.
While the above steps are working, it skips some steps if the channel
doesn't have a ->copy callback function.
Due to ptp channel doesn't have ->copy callback, these above steps are
skipped for ptp channel.
It eventually makes some problems.
a. ptp channel's ring buffer size is not changed, it works only
1024(default).
b. memory leak.
The reason for memory leak is to use the wrong ring buffer values.
There are some values, which is related to ring buffer size.
a. efx->rxq_entries
- This is global value of rx queue size.
b. rx_queue->ptr_mask
- used for access ring buffer as circular ring.
- roundup_pow_of_two(efx->rxq_entries) - 1
c. rx_queue->max_fill
- efx->rxq_entries - EFX_RXD_HEAD_ROOM
These all values should be based on ring buffer size consistently.
But ptp channel's values are not.
a. efx->rxq_entries
- This is global(for sfc) value, always new ring buffer size.
b. rx_queue->ptr_mask
- This is always 1023(default).
c. rx_queue->max_fill
- This is new ring buffer size - EFX_RXD_HEAD_ROOM.
Let's assume we set 4096 for rx ring buffer,
normal channel ptp channel
efx->rxq_entries 4096 4096
rx_queue->ptr_mask 4095 1023
rx_queue->max_fill 4086 4086
sfc driver allocates rx ring buffers based on these values.
When it allocates ptp channel's ring buffer, 4086 ring buffers are
allocated then, these buffers are attached to the allocated array.
But ptp channel's ring buffer array size is still 1024(default)
and ptr_mask is still 1023 too.
So, 3062 ring buffers will be overwritten to the array.
This is the reason for memory leak.
Test commands:
ethtool -G <interface name> rx 4096
while :
do
ip link set <interface name> up
ip link set <interface name> down
done
In order to avoid this problem, it adds ->copy callback to ptp channel
type.
So that rx_queue->ptr_mask value will be updated correctly.
Fixes: 7c236c43b8 ("sfc: Add support for IEEE-1588 PTP")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Switch all Ethernet drivers which use custom napi weights
to the new API.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Drivers should call the TSO setting helper, GSO is controllable
by user space.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
For Siena we do not need new messages that were defined
for the EF100 architecture. Several debug messages have
also been removed.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Disable the build of Siena code until later in this patch series.
Prevent sfc.ko from binding to Siena NICs.
efx_init_sriov/efx_fini_sriov is only used for Siena. Remove calls
to those.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch extends the EF100 PF driver by adding .sriov_configure()
which would allow users to enable and disable virtual functions
using the sriov sysfs.
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/75e74d9e-14ce-0524-9668-5ab735a7cf62@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The strings are only used in efx_common.c so the definitions
can be static in there.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
It is defined both in efx.h and tx_common.h.
Remove the definition in efx.h.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This means we can remove them from efx_channel.h and avoid
naming conflicts later.
efx_channel_dummy_op_void() cannot be static as it is
used in ef100_nic.c.
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The 8000 series and newer NICs all get hardware timestamps from the MAC
and can provide timestamps on a normal TX queue, rather than via a slow
path through the MC. As such we can use this path for any packet where a
hardware timestamp is requested.
This also enables support for PTP over transports other than IPv4+UDP.
Signed-off-by: Bert Kenward <bkenward@solarflare.com>
Signed-off-by: Edward Cree <ecree@xilinx.com>
Link: https://lore.kernel.org/r/510652dc-54b4-0e11-657e-e37ee3ca26a9@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Even if an IOMMU might be present for some PCI segment in the system,
that doesn't necessarily mean it provides translation for the device
we care about. It appears that what we care about here is specifically
whether DMA mapping ops involve any IOMMU overhead or not, so check for
translation actually being active for our device.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/7350f957944ecfce6cce90f422e3992a1f428775.1649166055.git.robin.murphy@arm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In some cases, xdp tx_queue can get used before initialization.
1. interface up/down
2. ring buffer size change
When CPU cores are lower than maximum number of channels of sfc driver,
it creates new channels only for XDP.
When an interface is up or ring buffer size is changed, all channels
are initialized.
But xdp channels are always initialized later.
So, the below scenario is possible.
Packets are received to rx queue of normal channels and it is acted
XDP_TX and tx_queue of xdp channels get used.
But these tx_queues are not initialized yet.
If so, TX DMA or queue error occurs.
In order to avoid this problem.
1. initializes xdp tx_queues earlier than other rx_queue in
efx_start_channels().
2. checks whether tx_queue is initialized or not in efx_xdp_tx_buffers().
Splat looks like:
sfc 0000:08:00.1 enp8s0f1np1: TX queue 10 spurious TX completion id 250
sfc 0000:08:00.1 enp8s0f1np1: resetting (RECOVER_OR_ALL)
sfc 0000:08:00.1 enp8s0f1np1: MC command 0x80 inlen 100 failed rc=-22
(raw=22) arg=789
sfc 0000:08:00.1 enp8s0f1np1: has been disabled
Fixes: f28100cb9c ("sfc: fix lack of XDP TX queues - error XDP TX failed (-22)")
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the page_ring is not used page_ptr_mask is 0.
Do not dereference page_ring[0] in this case.
Fixes: 2768935a46 ("sfc: reuse pages to avoid DMA mapping/unmapping costs")
Reported-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On such systems cpumask_of_node() returns NULL, which bitmap
operations are not happy with.
Fixes: c265b569a4 ("sfc: default config to 1 channel/core in local NUMA node only")
Fixes: 09a99ab16c ("sfc: set affinity hints in local NUMA node only")
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Reviewed-by: Íñigo Huguet <ihuguet@redhat.com>
Link: https://lore.kernel.org/r/164857006953.8140.3265568858101821256.stgit@palantir17.mph.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
seqno could be read as a stale value outside of the lock. The lock is
already acquired to protect the modification of seqno against a possible
race condition. Place the reading of this value also inside this locking
to protect it against a possible race condition.
Signed-off-by: Niels Dossche <dossche.niels@gmail.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Affinity hints were being set to CPUs in local NUMA node first, and then
in other CPUs. This was creating 2 unintended issues:
1. Channels created to be assigned each to a different physical core
were assigned to hyperthreading siblings because of being in same
NUMA node.
Since the patch previous to this one, this did not longer happen
with default rss_cpus modparam because less channels are created.
2. XDP channels could be assigned to CPUs in different NUMA nodes,
decreasing performance too much (to less than half in some of my
tests).
This patch sets the affinity hints spreading the channels only in local
NUMA node's CPUs. A fallback for the case that no CPU in local NUMA node
is online has been added too.
Example of CPUs being assigned in a non optimal way before this and the
previous patch (note: in this system, xdp-8 to xdp-15 are created
because num_possible_cpus == 64, but num_present_cpus == 32 so they're
never used):
$ lscpu | grep -i numa
NUMA node(s): 2
NUMA node0 CPU(s): 0-7,16-23
NUMA node1 CPU(s): 8-15,24-31
$ grep -H . /proc/irq/*/0000:07:00.0*/../smp_affinity_list
/proc/irq/141/0000:07:00.0-0/../smp_affinity_list:0
/proc/irq/142/0000:07:00.0-1/../smp_affinity_list:1
/proc/irq/143/0000:07:00.0-2/../smp_affinity_list:2
/proc/irq/144/0000:07:00.0-3/../smp_affinity_list:3
/proc/irq/145/0000:07:00.0-4/../smp_affinity_list:4
/proc/irq/146/0000:07:00.0-5/../smp_affinity_list:5
/proc/irq/147/0000:07:00.0-6/../smp_affinity_list:6
/proc/irq/148/0000:07:00.0-7/../smp_affinity_list:7
/proc/irq/149/0000:07:00.0-8/../smp_affinity_list:16
/proc/irq/150/0000:07:00.0-9/../smp_affinity_list:17
/proc/irq/151/0000:07:00.0-10/../smp_affinity_list:18
/proc/irq/152/0000:07:00.0-11/../smp_affinity_list:19
/proc/irq/153/0000:07:00.0-12/../smp_affinity_list:20
/proc/irq/154/0000:07:00.0-13/../smp_affinity_list:21
/proc/irq/155/0000:07:00.0-14/../smp_affinity_list:22
/proc/irq/156/0000:07:00.0-15/../smp_affinity_list:23
/proc/irq/157/0000:07:00.0-xdp-0/../smp_affinity_list:8
/proc/irq/158/0000:07:00.0-xdp-1/../smp_affinity_list:9
/proc/irq/159/0000:07:00.0-xdp-2/../smp_affinity_list:10
/proc/irq/160/0000:07:00.0-xdp-3/../smp_affinity_list:11
/proc/irq/161/0000:07:00.0-xdp-4/../smp_affinity_list:12
/proc/irq/162/0000:07:00.0-xdp-5/../smp_affinity_list:13
/proc/irq/163/0000:07:00.0-xdp-6/../smp_affinity_list:14
/proc/irq/164/0000:07:00.0-xdp-7/../smp_affinity_list:15
/proc/irq/165/0000:07:00.0-xdp-8/../smp_affinity_list:24
/proc/irq/166/0000:07:00.0-xdp-9/../smp_affinity_list:25
/proc/irq/167/0000:07:00.0-xdp-10/../smp_affinity_list:26
/proc/irq/168/0000:07:00.0-xdp-11/../smp_affinity_list:27
/proc/irq/169/0000:07:00.0-xdp-12/../smp_affinity_list:28
/proc/irq/170/0000:07:00.0-xdp-13/../smp_affinity_list:29
/proc/irq/171/0000:07:00.0-xdp-14/../smp_affinity_list:30
/proc/irq/172/0000:07:00.0-xdp-15/../smp_affinity_list:31
CPUs assignments after this and previous patch, so normal channels
created only one per core in NUMA node and affinities set only to local
NUMA node:
$ grep -H . /proc/irq/*/0000:07:00.0*/../smp_affinity_list
/proc/irq/116/0000:07:00.0-0/../smp_affinity_list:0
/proc/irq/117/0000:07:00.0-1/../smp_affinity_list:1
/proc/irq/118/0000:07:00.0-2/../smp_affinity_list:2
/proc/irq/119/0000:07:00.0-3/../smp_affinity_list:3
/proc/irq/120/0000:07:00.0-4/../smp_affinity_list:4
/proc/irq/121/0000:07:00.0-5/../smp_affinity_list:5
/proc/irq/122/0000:07:00.0-6/../smp_affinity_list:6
/proc/irq/123/0000:07:00.0-7/../smp_affinity_list:7
/proc/irq/124/0000:07:00.0-xdp-0/../smp_affinity_list:16
/proc/irq/125/0000:07:00.0-xdp-1/../smp_affinity_list:17
/proc/irq/126/0000:07:00.0-xdp-2/../smp_affinity_list:18
/proc/irq/127/0000:07:00.0-xdp-3/../smp_affinity_list:19
/proc/irq/128/0000:07:00.0-xdp-4/../smp_affinity_list:20
/proc/irq/129/0000:07:00.0-xdp-5/../smp_affinity_list:21
/proc/irq/130/0000:07:00.0-xdp-6/../smp_affinity_list:22
/proc/irq/131/0000:07:00.0-xdp-7/../smp_affinity_list:23
/proc/irq/132/0000:07:00.0-xdp-8/../smp_affinity_list:0
/proc/irq/133/0000:07:00.0-xdp-9/../smp_affinity_list:1
/proc/irq/134/0000:07:00.0-xdp-10/../smp_affinity_list:2
/proc/irq/135/0000:07:00.0-xdp-11/../smp_affinity_list:3
/proc/irq/136/0000:07:00.0-xdp-12/../smp_affinity_list:4
/proc/irq/137/0000:07:00.0-xdp-13/../smp_affinity_list:5
/proc/irq/138/0000:07:00.0-xdp-14/../smp_affinity_list:6
/proc/irq/139/0000:07:00.0-xdp-15/../smp_affinity_list:7
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Handling channels from CPUs in different NUMA node can penalize
performance, so better configure only one channel per core in the same
NUMA node than the NIC, and not per each core in the system.
Fallback to all other online cores if there are not online CPUs in local
NUMA node.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Ideally the size would depend on the link speed, but the recycle
ring is created when the interface is brought up before the driver
knows the link speed. So size it for the maximum speed of a given NIC.
PowerPC is only supported on SFN7xxx and SFN8xxx NICs.
With this patch on a 40G NIC the number of calls to alloc_pages and
friends went down from about 18% to under 2%.
On a 10G NIC the number of calls to alloc_pages and friends went down
from about 15% to 0 (perf did not capture any calls during the 60
second test).
On a 100G NIC the number of calls to alloc_pages and friends went down
from about 23% to 4%.
Reported-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220131111054.cp4f6foyinaarwbn@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Clean the following coccicheck warning:
./drivers/net/ethernet/sfc/efx_channels.c:870:36-37: WARNING opportunity
for swap().
./drivers/net/ethernet/sfc/efx_channels.c:824:36-37: WARNING opportunity
for swap().
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RX page_ring is an optional feature that improves
performance. When allocation fails the driver can still
function, but possibly with a lower bandwidth.
Guard against dereferencing a NULL page_ring.
Fixes: 2768935a46 ("sfc: reuse pages to avoid DMA mapping/unmapping costs")
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Reported-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Link: https://lore.kernel.org/r/164111288276.5798.10330502993729113868.stgit@palantir17.mph.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Return value from efx_mcdi_rpc() directly instead
of taking this in another redundant variable.
Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: CGEL ZTE <cgel.zte@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
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 <davem@davemloft.net>
sock.h is pretty heavily used (5k objects rebuilt on x86 after
it's touched). We can drop the include of filter.h from it and
add a forward declaration of struct sk_filter instead.
This decreases the number of rebuilt objects when bpf.h
is touched from ~5k to ~1k.
There's a lot of missing includes this was masking. Primarily
in networking tho, this time.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/bpf/20211229004913.513372-1-kuba@kernel.org
Because of the possible failure of the kcalloc, it should be better to
set rx_queue->page_ptr_mask to 0 when it happens in order to maintain
the consistency.
Fixes: 5a6681e22c ("sfc: separate out SFC4000 ("Falcon") support into new sfc-falcon driver")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20211220140344.978408-1-jiasheng@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Because of the possible failure of the kcalloc, it should be better to
set rx_queue->page_ptr_mask to 0 when it happens in order to maintain
the consistency.
Fixes: 5a6681e22c ("sfc: separate out SFC4000 ("Falcon") support into new sfc-falcon driver")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20211220135603.954944-1-jiasheng@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The return value of kmalloc() needs to be checked.
To avoid use in efx_nic_update_stats() in case of the failure of alloc.
Fixes: b593b6f1b4 ("sfc_ef100: statistics gathering")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 94dd016ae5 ("bond: pass get_ts_info and SIOC[SG]HWTSTAMP
ioctl to active device") the user could get bond active interface's
PHC index directly. But when there is a failover, the bond active
interface will change, thus the PHC index is also changed. This may
break the user's program if they did not update the PHC timely.
This patch adds a new hwtstamp_config flag HWTSTAMP_FLAG_BONDED_PHC_INDEX.
When the user wants to get the bond active interface's PHC, they need to
add this flag and be aware the PHC index may be changed.
With the new flag. All flag checks in current drivers are removed. Only
the checking in net_hwtstamp_validate() is kept.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
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 <pabeni@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/bpf/ddb96bb975cbfddb1546cf5da60e77d5100b533c.1638189075.git.pabeni@redhat.com
dev->gso_max_segs is written under RTNL protection, or when the device is
not yet visible, but is read locklessly.
Add netif_set_gso_max_segs() helper.
Add the READ_ONCE()/WRITE_ONCE() pairs, and use netif_set_gso_max_segs()
where we can to better document what is going on.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
dev->gso_max_size is written under RTNL protection, or when the device is
not yet visible, but is read locklessly.
Add the READ_ONCE()/WRITE_ONCE() pairs, and use netif_set_gso_max_size()
where we can to better document what is going on.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add two new parameters kernel_ringparam and extack for
.get_ringparam and .set_ringparam to extend more ring params
through netlink.
Signed-off-by: Hao Chen <chenhao288@hisilicon.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the macro 'swap()' defined in 'include/linux/minmax.h' to avoid
opencoding it.
Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Yang Guang <yang.guang5@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use pci_info instead to avoid unnamed/uninitialized noise:
[197088.688729] sfc 0000:01:00.0: Solarflare NIC detected
[197088.690333] sfc 0000:01:00.0: Part Number : SFN5122F
[197088.729061] sfc 0000:01:00.0 (unnamed net_device) (uninitialized): no SR-IOV VFs probed
[197088.729071] sfc 0000:01:00.0 (unnamed net_device) (uninitialized): no PTP support
Inspired by fa44821a4d ("sfc: don't use netif_info et al before
net_device is registered") from Heiner Kallweit.
Signed-off-by: Erik Ekman <erik@kryo.se>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The 1/10GbaseT modes were set up for cards with SFP+ cages in
3497ed8c85 ("sfc: report supported link speeds on SFP connections").
10GbaseT was likely used since no 10G fibre mode existed.
The missing fibre modes for 1/10G were added to ethtool.h in 5711a98221
("net: ethtool: add support for 1000BaseX and missing 10G link modes")
shortly thereafter.
The user guide available at https://support-nic.xilinx.com/wp/drivers
lists support for the following cable and transceiver types in section 2.9:
- QSFP28 100G Direct Attach Cables
- QSFP28 100G SR Optical Transceivers (with SR4 modules listed)
- SFP28 25G Direct Attach Cables
- SFP28 25G SR Optical Transceivers
- QSFP+ 40G Direct Attach Cables
- QSFP+ 40G Active Optical Cables
- QSFP+ 40G SR4 Optical Transceivers
- QSFP+ to SFP+ Breakout Direct Attach Cables
- QSFP+ to SFP+ Breakout Active Optical Cables
- SFP+ 10G Direct Attach Cables
- SFP+ 10G SR Optical Transceivers
- SFP+ 10G LR Optical Transceivers
- SFP 1000BASE‐T Transceivers
- 1G Optical Transceivers
(From user guide issue 28. Issue 16 which also includes older cards like
SFN5xxx/SFN6xxx has matching lists for 1/10/40G transceiver types.)
Regarding SFP+ 10GBASE‐T transceivers the latest guide says:
"Solarflare adapters do not support 10GBASE‐T transceiver modules."
Tested using SFN5122F-R7 (with 2 SFP+ ports). Supported link modes do not change
depending on module used (tested with 1000BASE-T, 1000BASE-BX10, 10GBASE-LR).
Before:
$ ethtool ext
Settings for ext:
Supported ports: [ FIBRE ]
Supported link modes: 1000baseT/Full
10000baseT/Full
Supported pause frame use: Symmetric Receive-only
Supports auto-negotiation: No
Supported FEC modes: Not reported
Advertised link modes: Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Advertised FEC modes: Not reported
Link partner advertised link modes: Not reported
Link partner advertised pause frame use: No
Link partner advertised auto-negotiation: No
Link partner advertised FEC modes: Not reported
Speed: 1000Mb/s
Duplex: Full
Auto-negotiation: off
Port: FIBRE
PHYAD: 255
Transceiver: internal
Current message level: 0x000020f7 (8439)
drv probe link ifdown ifup rx_err tx_err hw
Link detected: yes
After:
$ ethtool ext
Settings for ext:
Supported ports: [ FIBRE ]
Supported link modes: 1000baseT/Full
1000baseX/Full
10000baseCR/Full
10000baseSR/Full
10000baseLR/Full
Supported pause frame use: Symmetric Receive-only
Supports auto-negotiation: No
Supported FEC modes: Not reported
Advertised link modes: Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Advertised FEC modes: Not reported
Link partner advertised link modes: Not reported
Link partner advertised pause frame use: No
Link partner advertised auto-negotiation: No
Link partner advertised FEC modes: Not reported
Speed: 1000Mb/s
Duplex: Full
Auto-negotiation: off
Port: FIBRE
PHYAD: 255
Transceiver: internal
Supports Wake-on: g
Wake-on: d
Current message level: 0x000020f7 (8439)
drv probe link ifdown ifup rx_err tx_err hw
Link detected: yes
Signed-off-by: Erik Ekman <erik@kryo.se>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Everything except the first 32 bits was lost when the pause flags were
added. This makes the 50000baseCR2 mode flag (bit 34) not appear.
I have tested this with a 10G card (SFN5122F-R7) by modifying it to
return a non-legacy link mode (10000baseCR).
Signed-off-by: Erik Ekman <erik@kryo.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
This big patch sprinkles const on local variables and
function arguments which may refer to netdev->dev_addr.
Commit 406f42fa0d ("net-next: When a bond have a massive amount
of VLANs...") introduced a rbtree for faster Ethernet address look
up. To maintain netdev->dev_addr in this tree we need to make all
the writes to it got through appropriate helpers.
Some of the changes here are not strictly required - const
is sometimes cast off but pointer is not used for writing.
It seems like it's still better to add the const in case
the code changes later or relevant -W flags get enabled
for the build.
No functional changes.
Link: https://lore.kernel.org/r/20211014142432.449314-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Convert Ethernet from ether_addr_copy() to eth_hw_addr_set():
@@
expression dev, np;
@@
- ether_addr_copy(dev->dev_addr, np)
+ eth_hw_addr_set(dev, np)
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert all Ethernet drivers from memcpy(... ETH_ADDR)
to eth_hw_addr_set():
@@
expression dev, np;
@@
- memcpy(dev->dev_addr, np, ETH_ALEN)
+ eth_hw_addr_set(dev, np)
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previous patch addressed the situation of having some free resources for
xdp tx but not enough for one tx queue per CPU. This patch address the
worst case of not having resources at all for xdp tx.
Instead of using queues dedicated to xdp, normal queues used by network
stack are shared for both cases, using __netif_tx_lock for
synchronization. Also queue stop/restart must be considered in the xdp
path to avoid freezing the queue.
This is not the ideal situation we might want to be, and a performance
penalty is expected both for normal and xdp traffic, but at least XDP
will work in all possible situations (with a warning in the logs),
improving a bit the pain of not knowing in what situations we can use it
and in what situations we cannot.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If there are not enough resources to allocate one TX queue per core for
XDP TX it was completely disabled.
This patch implements a fallback solution for sharing the available
queues using __netif_tx_lock for synchronization. In the normal case that
there is one TX queue per CPU, no locking is done, as it was before.
With this fallback solution, XDP TX will work in much more cases that
were failing, specially in machines with many CPUs. It's hard for XDP
users to know what features are supported across different NICs and
configurations, so they will benefit on having wider support.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use pci_vpd_find_ro_info_keyword() to search for keywords in VPD to
simplify the code.
Replace netif_err() with pci_err() because the netdevice isn't registered
yet, which results in very ugly messages.
Use kmemdup_nul() instead of open-coding it.
This is the same as 37838aa437 ("sfc: Search VPD with
pci_vpd_find_ro_info_keyword()"), just for the falcon chip version.
Link: https://lore.kernel.org/r/898282a1-13bd-17bc-2e9a-d3dcd336b46c@gmail.com
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Use pci_vpd_alloc() to dynamically allocate a properly sized buffer and
read the full VPD data into it.
This avoids having to allocate a buffer on the stack, and we don't have to
make any assumptions on VPD size and location of information in VPD.
This is the same as 5119e20fac ("sfc: Read VPD with pci_vpd_alloc()"),
just for the falcon chip version.
Link: https://lore.kernel.org/r/2a8d069e-9516-50d8-6520-2614222c8f5f@gmail.com
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
In order to support more coalesce parameters through netlink,
add two new parameter kernel_coal and extack for .set_coalesce
and .get_coalesce, then some extra info can return to user with
the netlink API.
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is the same as 37838aa437 "sfc: Search VPD with
pci_vpd_find_ro_info_keyword()", just for the falcon chip version.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the same as 5119e20fac "sfc: Read VPD with pci_vpd_alloc()",
just for the falcon chip version.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use pci_vpd_find_ro_info_keyword() to search for keywords in VPD to
simplify the code.
Replace netif_err() with pci_err() because the netdevice isn't registered
yet, which results in very ugly messages.
Use kmemdup_nul() instead of open-coding it.
Link: https://lore.kernel.org/r/bf5d4ba9-61a9-2bfe-19ec-75472732d74d@gmail.com
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Use pci_vpd_alloc() to dynamically allocate a properly sized buffer and
read the full VPD data into it.
This avoids having to allocate a buffer on the stack, and we don't have to
make any assumptions on VPD size and location of information in VPD.
Link: https://lore.kernel.org/r/e58f1e40-c043-0266-9a0f-e5a7f3f6883c@gmail.com
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The 'imply' keyword does not do what most people think it does, it only
politely asks Kconfig to turn on another symbol, but does not prevent
it from being disabled manually or built as a loadable module when the
user is built-in. In the ICE driver, the latter now causes a link failure:
aarch64-linux-ld: drivers/net/ethernet/intel/ice/ice_main.o: in function `ice_eth_ioctl':
ice_main.c:(.text+0x13b0): undefined reference to `ice_ptp_get_ts_config'
ice_main.c:(.text+0x13b0): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `ice_ptp_get_ts_config'
aarch64-linux-ld: ice_main.c:(.text+0x13bc): undefined reference to `ice_ptp_set_ts_config'
ice_main.c:(.text+0x13bc): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `ice_ptp_set_ts_config'
aarch64-linux-ld: drivers/net/ethernet/intel/ice/ice_main.o: in function `ice_prepare_for_reset':
ice_main.c:(.text+0x31fc): undefined reference to `ice_ptp_release'
ice_main.c:(.text+0x31fc): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `ice_ptp_release'
aarch64-linux-ld: drivers/net/ethernet/intel/ice/ice_main.o: in function `ice_rebuild':
This is a recurring problem in many drivers, and we have discussed
it several times befores, without reaching a consensus. I'm providing
a link to the previous email thread for reference, which discusses
some related problems.
To solve the dependency issue better than the 'imply' keyword, introduce a
separate Kconfig symbol "CONFIG_PTP_1588_CLOCK_OPTIONAL" that any driver
can depend on if it is able to use PTP support when available, but works
fine without it. Whenever CONFIG_PTP_1588_CLOCK=m, those drivers are
then prevented from being built-in, the same way as with a 'depends on
PTP_1588_CLOCK || !PTP_1588_CLOCK' dependency that does the same trick,
but that can be rather confusing when you first see it.
Since this should cover the dependencies correctly, the IS_REACHABLE()
hack in the header is no longer needed now, and can be turned back
into a normal IS_ENABLED() check. Any driver that gets the dependency
wrong will now cause a link time failure rather than being unable to use
PTP support when that is in a loadable module.
However, the two recently added ptp_get_vclocks_index() and
ptp_convert_timestamp() interfaces are only called from builtin code with
ethtool and socket timestamps, so keep the current behavior by stubbing
those out completely when PTP is in a loadable module. This should be
addressed properly in a follow-up.
As Richard suggested, we may want to actually turn PTP support into a
'bool' option later on, preventing it from being a loadable module
altogether, which would be one way to solve the problem with the ethtool
interface.
Fixes: 06c16d89d2 ("ice: register 1588 PTP clock device object for E810 devices")
Link: https://lore.kernel.org/netdev/20210804121318.337276-1-arnd@kernel.org/
Link: https://lore.kernel.org/netdev/CAK8P3a06enZOf=XyZ+zcAwBczv41UuCTz+=0FMf2gBz1_cOnZQ@mail.gmail.com/
Link: https://lore.kernel.org/netdev/CAK8P3a3=eOxE-K25754+fB_-i_0BZzf9a9RfPTX3ppSwu9WZXw@mail.gmail.com/
Link: https://lore.kernel.org/netdev/20210726084540.3282344-1-arnd@kernel.org/
Acked-by: Shannon Nelson <snelson@pensando.io>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20210812183509.1362782-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Most users of ndo_do_ioctl are ethernet drivers that implement
the MII commands SIOCGMIIPHY/SIOCGMIIREG/SIOCSMIIREG, or hardware
timestamping with SIOCSHWTSTAMP/SIOCGHWTSTAMP.
Separate these from the few drivers that use ndo_do_ioctl to
implement SIOCBOND, SIOCBR and SIOCWANDEV commands.
This is a purely cosmetic change intended to help readers find
their way through the implementation.
Cc: Doug Ledford <dledford@redhat.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Vivien Didelot <vivien.didelot@gmail.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Vladimir Oltean <olteanv@gmail.com>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: linux-rdma@vger.kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If it's not possible to allocate enough channels for XDP, XDP_TX and
XDP_REDIRECT don't work. However, only a message saying that not enough
channels were available was shown, but not saying what are the
consequences in that case. The user didn't know if he/she can use XDP
or not, if the performance is reduced, or what.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 99ba0ea616 ("sfc: adjust efx->xdp_tx_queue_count with the real
number of initialized queues") intended to fix a problem caused by a
round up when calculating the number of XDP channels and queues.
However, this was not the real problem. The real problem was that the
number of XDP TX queues had been reduced to half in
commit e26ca4b535 ("sfc: reduce the number of requested xdp ev queues"),
but the variable xdp_tx_queue_count had remained the same.
Once the correct number of XDP TX queues is created again in the
previous patch of this series, this also can be reverted since the error
doesn't actually exist.
Only in the case that there is a bug in the code we can have different
values in xdp_queue_number and efx->xdp_tx_queue_count. Because of this,
and per Edward Cree's suggestion, I add instead a WARN_ON to catch if it
happens again in the future.
Note that the number of allocated queues can be higher than the number
of used ones due to the round up, as explained in the existing comment
in the code. That's why we also have to stop increasing xdp_queue_number
beyond efx->xdp_tx_queue_count.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes: e26ca4b535 sfc: reduce the number of requested xdp ev queues
The buggy commit intended to allocate less channels for XDP in order to
be more unlikely to reach the limit of 32 channels of the driver.
The idea was to use each IRQ/eventqeue for more XDP TX queues than
before, calculating which is the maximum number of TX queues that one
event queue can handle. For example, in EF10 each event queue could
handle up to 8 queues, better than the 4 they were handling before the
change. This way, it would have to allocate half of channels than before
for XDP TX.
The problem is that the TX queues are also contained inside the channel
structs, and there are only 4 queues per channel. Reducing the number of
channels means also reducing the number of queues, resulting in not
having the desired number of 1 queue per CPU.
This leads to getting errors on XDP_TX and XDP_REDIRECT if they're
executed from a high numbered CPU, because there only exist queues for
the low half of CPUs, actually. If XDP_TX/REDIRECT is executed in a low
numbered CPU, the error doesn't happen. This is the error in the logs
(repeated many times, even rate limited):
sfc 0000:5e:00.0 ens3f0np0: XDP TX failed (-22)
This errors happens in function efx_xdp_tx_buffers, where it expects to
have a dedicated XDP TX queue per CPU.
Reverting the change makes again more likely to reach the limit of 32
channels in machines with many CPUs. If this happen, no XDP_TX/REDIRECT
will be possible at all, and we will have this log error messages:
At interface probe:
sfc 0000:5e:00.0: Insufficient resources for 12 XDP event queues (24 other channels, max 32)
At every subsequent XDP_TX/REDIRECT failure, rate limited:
sfc 0000:5e:00.0 ens3f0np0: XDP TX failed (-22)
However, without reverting the change, it makes the user to think that
everything is OK at probe time, but later it fails in an unpredictable
way, depending on the CPU that handles the packet.
It is better to restore the predictable behaviour. If the user sees the
error message at probe time, he/she can try to configure the best way it
fits his/her needs. At least, he/she will have 2 options:
- Accept that XDP_TX/REDIRECT is not available (he/she may not need it)
- Load sfc module with modparam 'rss_cpus' with a lower number, thus
creating less normal RX queues/channels, letting more free resources
for XDP, with some performance penalty.
Anyway, let the calculation of maximum TX queues that can be handled by
a single event queue, and use it only if it's less than the number of TX
queues per channel. This doesn't happen in practice, but could happen if
some constant values are tweaked in the future, such us
EFX_MAX_TXQ_PER_CHANNEL, EFX_MAX_EVQ_SIZE or EFX_MAX_DMAQ_SIZE.
Related mailing list thread:
https://lore.kernel.org/bpf/20201215104327.2be76156@carbon/
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Trivial conflict in net/netfilter/nf_tables_api.c.
Duplicate fix in tools/testing/selftests/net/devlink_port_split.py
- take the net-next version.
skmsg, and L4 bpf - keep the bpf code but remove the flags
and err params.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The sfc driver has 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 <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Cc: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/bpf/20210624160609.292325-17-toke@redhat.com
The fail path of efx_ef10_sriov_alloc_vf_vswitching is identical to the
full content of efx_ef10_sriov_free_vf_vswitching, so replace it for a
single call to efx_ef10_sriov_free_vf_vswitching.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During SRIOV disabling it is checked wether any VF is currently attached
to a guest, using pci_vfs_assigned function. However, this check only
works with VFs attached with Xen, not with vfio/KVM. Added comments
clarifying this point.
Also, replaced manual check of PCI_DEV_FLAGS_ASSIGNED flag and used the
helper function pci_is_dev_assigned instead.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If SRIOV cannot be disabled during device removal or module unloading,
return error code so it can be logged properly in the calling function.
Note that this can only happen if any VF is currently attached to a
guest using Xen, but not with vfio/KVM. Despite that in that case the
VFs won't work properly with PF removed and/or the module unloaded, I
have let it as is because I don't know what side effects may have
changing it, and also it seems to be the same that other drivers are
doing in this situation.
In the case of being called during SRIOV reconfiguration, the behavior
hasn't changed because the function is called with force=false.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If pci_remove was called for a PF with VFs, the removal of the VFs was
called twice from efx_ef10_sriov_fini: one directly with pci_driver->remove
and another implicit by calling pci_disable_sriov, which also perform
the VFs remove. This was leading to crashing the kernel on the second
attempt.
Given that pci_disable_sriov already calls to pci remove function, get
rid of the direct call to pci_driver->remove from the driver.
2 different ways to trigger the bug:
- Create one or more VFs, then attach the PF to a virtual machine (at
least with qemu/KVM)
- Create one or more VFs, then remove the PF with:
echo 1 > /sys/bus/pci/devices/PF_PCI_ID/remove
Removing sfc module does not trigger the error, at least for me, because
it removes the VF first, and then the PF.
Example of a log with the error:
list_del corruption, ffff967fd20a8ad0->next is LIST_POISON1 (dead000000000100)
------------[ cut here ]------------
kernel BUG at lib/list_debug.c:47!
[...trimmed...]
RIP: 0010:__list_del_entry_valid.cold.1+0x12/0x4c
[...trimmed...]
Call Trace:
efx_dissociate+0x1f/0x140 [sfc]
efx_pci_remove+0x27/0x150 [sfc]
pci_device_remove+0x3b/0xc0
device_release_driver_internal+0x103/0x1f0
pci_stop_bus_device+0x69/0x90
pci_stop_and_remove_bus_device+0xe/0x20
pci_iov_remove_virtfn+0xba/0x120
sriov_disable+0x2f/0xe0
efx_ef10_pci_sriov_disable+0x52/0x80 [sfc]
? pcie_aer_is_native+0x12/0x40
efx_ef10_sriov_fini+0x72/0x110 [sfc]
efx_pci_remove+0x62/0x150 [sfc]
pci_device_remove+0x3b/0xc0
device_release_driver_internal+0x103/0x1f0
unbind_store+0xf6/0x130
kernfs_fop_write+0x116/0x190
vfs_write+0xa5/0x1a0
ksys_write+0x4f/0xb0
do_syscall_64+0x5b/0x1a0
entry_SYSCALL_64_after_hwframe+0x65/0xca
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use DEVICE_ATTR_*() helper instead of plain DEVICE_ATTR,
which makes the code a bit shorter and easier to read.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use DEVICE_ATTR_*() helper instead of plain DEVICE_ATTR,
which makes the code a bit shorter and easier to read.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SFC driver can be configured via modparam to work using MSI-X, MSI or
legacy IRQ interrupts. In the last one, the interrupt was not properly
released on module remove.
It was not freed because the flag irqs_hooked was not set during
initialization in the case of using legacy IRQ.
Example of (trimmed) trace during module remove without this fix:
remove_proc_entry: removing non-empty directory 'irq/125', leaking at least '0000:3b:00.1'
WARNING: CPU: 39 PID: 3658 at fs/proc/generic.c:715 remove_proc_entry+0x15c/0x170
...trimmed...
Call Trace:
unregister_irq_proc+0xe3/0x100
free_desc+0x29/0x70
irq_free_descs+0x47/0x70
mp_unmap_irq+0x58/0x60
acpi_unregister_gsi_ioapic+0x2a/0x40
acpi_pci_irq_disable+0x78/0xb0
pci_disable_device+0xd1/0x100
efx_pci_remove+0xa1/0x1e0 [sfc]
pci_device_remove+0x38/0xa0
__device_release_driver+0x177/0x230
driver_detach+0xcb/0x110
bus_remove_driver+0x58/0xd0
pci_unregister_driver+0x2a/0xb0
efx_exit_module+0x24/0xf40 [sfc]
__do_sys_delete_module.constprop.0+0x171/0x280
? exit_to_user_mode_prepare+0x83/0x1d0
do_syscall_64+0x3d/0x80
entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7f9f9385800b
...trimmed...
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix the following kernel build warning when CONFIG_SFC_SRIOV is disabled:
drivers/net/ethernet/sfc/farch.c: In function ‘efx_farch_dimension_resources’:
drivers/net/ethernet/sfc/farch.c:1671:21: warning: variable ‘buftbl_min’ set but not used [-Wunused-but-set-variable]
unsigned vi_count, buftbl_min, total_tx_channels;
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using netif_info() before the net_device is registered results in ugly
messages like the following:
sfc 0000:01:00.1 (unnamed net_device) (uninitialized): Solarflare NIC detected
Therefore use pci_info() et al until net_device is registered.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
