With MTU sized receive buffers it is not expected to have CQE_RX
with multiple receive buffer pointers. But since same physcial link
is shared by PF and it's VFs, the max receive packet configured
at link could be morethan MTU. Hence there is a chance of receiving
plts morethan MTU which then gets DMA'ed into multiple buffers
and notified in a single CQE_RX. This patch treats such pkts as errors
and frees up receive buffers pointers back to hardware.
Also on the transmit side this patch sets SMQ MAXLEN to max value to avoid
HW length errors for the packets whose size > MTU, eg due to path MTU.
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
VF shares physical link with PF. Admin function (AF) sends
notification to PF whenever a link change event happens. PF
has to forward the same notification to each of the enabled VF.
PF traps START/STOP_RX messages sent by VF to AF to keep track of
VF's enabled/disabled state.
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added ethtool support for VF devices for
- Driver stats, Tx/Rx perqueue stats
- Set/show Rx/Tx queue count
- Set/show Rx/Tx ring sizes
- Set/show IRQ coalescing parameters
- RSS configuration etc
It's the PF which owns the interface, hence VF
cannot display underlying CGX interface stats.
Except for this rest ethtool support reuses PF's
APIs.
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On OcteonTx2 silicon there two two types VFs, VFs that share the
physical link with their parent SR-IOV PF and the VFs which work
in pairs using internal HW loopback channels (LBK). Except for the
underlying Rx/Tx channel mapping from netdev functionality perspective
they are almost identical. This patch adds netdev driver support
for these VFs.
Unlike it's parent PF a VF cannot directly communicate with admin
function (AF) and it has to go through PF for the same. The mailbox
communication with AF works like 'VF <=> PF <=> AF'.
Also functionality wise VF and PF are identical, hence to avoid code
duplication PF driver's APIs are resued here for HW initialization,
packet handling etc etc ie almost everything. For VF driver to compile
as module exported few of the existing PF driver APIs.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When FLR is initiated for a VF (PCI function level reset),
the parent PF gets a interrupt. PF then sends a message to
admin function (AF), which then cleanups all resources attached
to that VF.
Also handled IRQs triggered when master enable bit is cleared
or set for VFs. This handler just clears the transaction pending
ie TRPEND bit.
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added 'sriov_configure' to enable/disable virtual functions (VFs).
Also added handling of mailbox messages from these VFs.
Admin function (AF) is the only one with all priviliges to configure
HW, alloc resources etc etc, PFs and it's VFs have to request AF
via mbox for all their needs. But unlike PFs, their VFs cannot
send a mbox request directly. A VF shares a mailbox region with
it's parent PF, so VF sends a mailbox msg to PF and then PF forwards
it to AF. Then AF after processing sends response to PF which it
again forwards to VF.
This patch adds support for this 'VF <=> PF <=> AF' mailbox
communication.
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Heiner Kallweit says:
====================
net: phy: add and use phy_check_downshift
So far PHY drivers have to check whether a downshift occurred to be
able to notify the user. To make life of drivers authors a little bit
easier move the downshift notification to phylib. phy_check_downshift()
compares the highest mutually advertised speed with the actual value
of phydev->speed (typically read by the PHY driver from a
vendor-specific register) to detect a downshift.
v2: Add downshift hint to phy_print_status().
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that phylib notifies the user of a downshift we can remove
this functionality from the driver.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that phylib notifies the user of a downshift we can remove
this functionality from the driver.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So far PHY drivers have to check whether a downshift occurred to be
able to notify the user. To make life of drivers authors a little bit
easier move the downshift notification to phylib. phy_check_downshift()
compares the highest mutually advertised speed with the actual value
of phydev->speed (typically read by the PHY driver from a
vendor-specific register) to detect a downshift.
v2:
- Add downshift hint to phy_print_status
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jose Abreu says:
====================
net: phy: xpcs: Improvements for -next
Misc set of improvements for XPCS. All for net-next.
Patch 1/4, returns link error upon 10GKR faults are detected.
Patch 2/4, resets XPCS upon probe so that we start from well known state.
Patch 3/4, sets Link as down if AutoNeg is enabled but did not finish with
success.
Patch 4/4, restarts AutoNeg process if previous outcome was not valid.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Restart AutoNeg if we didn't get a valid result from previous run.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set XPCS Link as down when AutoNeg is enabled but it didn't finish with
success.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reset the XPCS upon probe stage so that we start it from well known
state.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For 10GKR rate, when link errors are found we need to return fault
status so that XPCS is correctly resumed.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When building arm32 allyesconfig:
ld.lld: error: undefined symbol: __aeabi_uldivmod
>>> referenced by spectrum_cnt.c
>>> net/ethernet/mellanox/mlxsw/spectrum_cnt.o:(mlxsw_sp_counter_resources_register) in archive drivers/built-in.a
>>> did you mean: __aeabi_uidivmod
>>> defined in: arch/arm/lib/lib.a(lib1funcs.o)
pool_size and bank_size are u64; use div64_u64 so that 32-bit platforms
do not error.
Fixes: ab8c4cc604 ("mlxsw: spectrum_cnt: Move config validation along with resource register")
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 53eca1f347 ("net: rename flow_action_hw_stats_types* ->
flow_action_hw_stats*") renamed just the flow action types and
helpers. For consistency rename variables, enums, struct members
and UAPI too (note that this UAPI was not in any official release,
yet).
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
it's still possible for packetdrill to hang in mptcp_sendmsg(), when the
MPTCP socket falls back to regular TCP (e.g. after receiving unsupported
flags/version during the three-way handshake). Adjust MPTCP socket state
earlier, to ensure correct functionality of mptcp_sendmsg() even in case
of TCP fallback.
Fixes: 767d3ded5f ("net: mptcp: don't hang before sending 'MP capable with data'")
Fixes: 1954b86016 ("mptcp: Check connection state before attempting send")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean says:
====================
MSCC PHY: RGMII delays and VSC8502 support
This series makes RGMII delays configurable as they should be on
Vitesse/Microsemi/Microchip RGMII PHYs, and adds support for a new RGMII
PHY.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a dual copper PHY with support for MII/GMII/RGMII on MAC side,
as well as a bunch of other features such as SyncE and Ring Resiliency.
I haven't tested interrupts and WoL, but I am confident that they work
since support is already present in the driver and the register map is
no different for this PHY.
PHY statistics work, PHY tunables appear to work, suspend/resume works.
Signed-off-by: Wes Li <wes.li@nxp.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver appears to be secretly enabling the RX clock skew
irrespective of PHY interface type, which is generally considered a big
no-no.
Make them configurable instead, and add TX internal delays when
necessary too.
While at it, configure a more canonical clock skew of 2.0 nanoseconds
than the current default of 1.1 ns.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The helper for configuring the pinout of the MII side of the PHY should
do so irrespective of whether RGMII delays are used or not. So accept
the ID, TXID and RXID variants as well, not just the no-delay RGMII
variant.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is nothing RX-specific about these clock skew values. So remove
"RX" from the name in preparation for the next patch where TX delays are
also going to be configured.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Kernel TLS forms TLS header in kernel during encryption and removes
while decryption before giving packet back to user application. The
similar logic is introduced in chtls code as well.
v1->v2:
- tls_proccess_cmsg() uses tls_handle_open_record() which is not required
in TOE-TLS. Don't mix TOE with other TLS types.
Signed-off-by: Vinay Kumar Yadav <vinay.yadav@chelsio.com>
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Problem:
TCP checksum in the output path is not being offloaded during GSO
in the following case:
The network driver does not support scatter-gather but supports
checksum offload with NETIF_F_HW_CSUM.
Cause:
skb_segment calls skb_copy_and_csum_bits if the network driver
does not announce NETIF_F_SG. It does not check if the driver
supports NETIF_F_HW_CSUM.
So for devices which might want to offload checksum but do not support SG
there is currently no way to do so if GSO is enabled.
Solution:
In skb_segment check if the network controller does checksum and if so
call skb_copy_bits instead of skb_copy_and_csum_bits.
Testing:
Without the patch, ran iperf TCP traffic with NETIF_F_HW_CSUM enabled
in the network driver. Observed the TCP checksum offload is not happening
since the skbs received by the driver in the output path have
skb->ip_summed set to CHECKSUM_NONE.
With the patch ran iperf TCP traffic and observed that TCP checksum
is being offloaded with skb->ip_summed set to CHECKSUM_PARTIAL.
Also tested with the patch by disabling NETIF_F_HW_CSUM in the driver
to cover the newly introduced if-else code path in skb_segment.
Link: https://lore.kernel.org/netdev/CA+FuTSeYGYr3Umij+Mezk9CUcaxYwqEe5sPSuXF8jPE2yMFJAw@mail.gmail.com
Signed-off-by: Yadu Kishore <kyk.segfault@gmail.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Huazhong Tan says:
====================
net: hns3: add three optimizations for mailbox handling
This patchset includes three code optimizations for mailbox handling.
[patch 1] adds a response code conversion.
[patch 2] refactors some structure definitions about PF and
VF mailbox.
[patch 3] refactors the condition whether PF responds VF's mailbox.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, PF responds to VF depending on what mailbox it is
handling, it is a bit inflexible. The correct way is, PF should
check the mbx_need_resp field to decide whether gives response
to VF.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For making the code more readable, this adds several new
structure to replace the msg field in structure
hclge_mbx_vf_to_pf_cmd and hclge_mbx_pf_to_vf_cmd.
Also uses macro to instead of some magic number.
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, when mailbox handling fails, the PF driver
just responds 1 to the VF driver. It is not sufficient
for the VF driver to find out why its mailbox fails.
So the error should be responded to VF, but the error
is type int and the response field in struct
hclge_mbx_pf_to_vf_cmd is type u16, a conversion is
needed.
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes gcc '-Wunused-but-set-variable' warning:
net/mptcp/options.c: In function 'mptcp_established_options_dss':
net/mptcp/options.c:338:7: warning:
variable 'can_ack' set but not used [-Wunused-but-set-variable]
commit dc093db5cc ("mptcp: drop unneeded checks")
leave behind this unused, remove it.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jian Yang says:
====================
selftests: expand txtimestamp with new features
Current txtimestamp selftest issues requests with no delay, or fixed 50
usec delay. Nsec granularity is useful to measure fine-grained latency.
A configurable delay is useful to simulate the case with cold
cachelines.
This patchset adds new flags and features to the txtimestamp selftest,
including:
- Printing in nsec (-N)
- Polling interval (-b, -S)
- Using epoll (-E, -e)
- Printing statistics
- Running individual tests in txtimestamp.sh
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Statistics on timestamps is useful to quantify average and tail latency.
Print timestamp statistics in count/avg/min/max format.
Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the following new flags:
-e: use level-triggered epoll() instead of poll().
-E: use event-triggered epoll() instead of poll().
Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A longer sleep duration between sendmsg()s makes more cachelines to be
evicted and results in higher latency. Making the duration configurable.
Add the following new flags:
-S: Configurable sleep duration.
-b: Busy loop instead of poll().
Remove the following flag:
-D: No delay between packets: subsumed by -S.
Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Txtimestamp reports latencies in uses resolution, while nsec is needed
in cases such as measuring latencies on localhost.
Add the following new flag:
-N: print timestamps and durations in nsec (instead of usec)
Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The wrapper script txtimestamp.sh executes a pre-defined list of testcases
sequentially without configuration options available.
Add an option (-r/--run) to setup the test namespace and pass remaining
arguments to txtimestamp binary. The script still runs all tests when no
argument is passed.
Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Sitnicki says:
====================
net/tls: Annotate lockless access to sk_prot
We have recently noticed that there is a case of lockless read/write to
sk->sk_prot [0]. sockmap code on psock tear-down writes to sk->sk_prot,
while holding sk_callback_lock. Concurrently, tcp can access it. Usually to
read out the sk_prot pointer and invoke one of the ops,
sk->sk_prot->handler().
The lockless write (lockless in regard to concurrent reads) happens on the
following paths:
tcp_bpf_{recvmsg|sendmsg} / sock_map_unref
sk_psock_put
sk_psock_drop
sk_psock_restore_proto
WRITE_ONCE(sk->sk_prot, proto)
To prevent load/store tearing [1], and to make tooling aware of intentional
shared access [2], we need to annotate sites that access sk_prot with
READ_ONCE/WRITE_ONCE.
This series kicks off the effort to do it. Starting with net/tls.
[0] https://lore.kernel.org/bpf/a6bf279e-a998-84ab-4371-cd6c1ccbca5d@gmail.com/
[1] https://lwn.net/Articles/793253/
[2] https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
sockmap performs lockless writes to sk->sk_prot on the following paths:
tcp_bpf_{recvmsg|sendmsg} / sock_map_unref
sk_psock_put
sk_psock_drop
sk_psock_restore_proto
WRITE_ONCE(sk->sk_prot, proto)
To prevent load/store tearing [1], and to make tooling aware of intentional
shared access [2], we need to annotate other sites that access sk_prot with
READ_ONCE/WRITE_ONCE macros.
Change done with Coccinelle with following semantic patch:
@@
expression E;
identifier I;
struct sock *sk;
identifier sk_prot =~ "^sk_prot$";
@@
(
E =
-sk->sk_prot
+READ_ONCE(sk->sk_prot)
|
-sk->sk_prot = E
+WRITE_ONCE(sk->sk_prot, E)
|
-sk->sk_prot
+READ_ONCE(sk->sk_prot)
->I
)
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Apart from being a "tremendous" win when it comes to generated machine
code (see bloat-o-meter output for x86-64 below) this mainly prepares
ground for annotating access to sk_prot with READ_ONCE, so that we don't
pepper the code with access annotations and needlessly repeat loads.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-46 (-46)
Function old new delta
tls_init 851 805 -46
Total: Before=21063, After=21017, chg -0.22%
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The helper that builds kTLS proto ops doesn't need to and should not modify
the base proto ops. Annotate the parameter as read-only.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Shannon Nelson says:
====================
ionic error recovery fixes
These are a few little patches to make error recovery a little
more safe and successful.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Make sure the queue structs exist before trying to tear
them down to make for safer error recovery.
Fixes: 0f3154e6bc ("ionic: Add Tx and Rx handling")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a little more cleanup when tearing down the queues.
Fixes: 1d062b7b6f ("ionic: Add basic adminq support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Don't worry if the rx filter add firmware request fails on
EEXIST, at least we know the filter is there. Same for
the delete request, at least we know it isn't there.
Fixes: 2a654540be ("ionic: Add Rx filter and rx_mode ndo support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Don't save the lif->dentry until we know we have
a good value.
Fixes: 1a58e19646 ("ionic: Add basic lif support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is possible (but unlikely) that FW was busy and missed a heartbeat
check but is still alive and will process the pending request, so don't
clean the dev_cmd in this case. This occasionally occurs when working
with a card that is supporting many devices and is trying to shut them
all down at once, but still wants to see that last LIF disable request.
Fixes: 97ca486592 ("ionic: add heartbeat check")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Short circuit the cleanup if we get a timeout error from
ionic_qcq_disable() so as to not have to wait too long
on shutdown when we already know the FW is not responding.
Fixes: 0f3154e6bc ("ionic: Add Tx and Rx handling")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Don't assume the receive buffer size is a power-of-2 number of pages.
Instead, define the receive buffer size independently, and then
compute the page order from that size when needed.
This fixes a build problem that arises when the ARM64_PAGE_SHIFT
config option is set to have a page size greater than 4KB. The
problem was identified by Linux Kernel Functional Testing.
The IPA code basically assumed the page size to be 4KB. A larger page
size caused the receive buffer size to become correspondingly larger
(32KB or 128KB for ARM64_16K_PAGES and ARM64_64K_PAGES, respectively).
The receive buffer size is used to compute an "aggregation byte limit"
value that gets programmed into the hardware, and the large page sizes
caused that limit value to be too big to fit in a 5 bit field. This
triggered a BUILD_BUG_ON() call in ipa_endpoint_validate_build().
This fix causes a lot of receive buffer memory to be wasted if
system is configured for page size greater than 4KB. But such a
misguided configuration will now build successfully.
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Alex Elder <elder@linaro.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
