In commit 6f8b12d661 ("net: napi: add hard irqs deferral feature")
napi_defer_irqs was added to net_device and napi_defer_irqs_count was
added to napi_struct, both as type int.
This value never goes below zero, so there is not reason for it to be a
signed int. Change the type for both from int to u32, and add an
overflow check to sysfs to limit the value to S32_MAX.
The limit of S32_MAX was chosen because the practical limit before this
patch was S32_MAX (anything larger was an overflow) and thus there are
no behavioral changes introduced. If the extra bit is needed in the
future, the limit can be raised.
Before this patch:
$ sudo bash -c 'echo 2147483649 > /sys/class/net/eth4/napi_defer_hard_irqs'
$ cat /sys/class/net/eth4/napi_defer_hard_irqs
-2147483647
After this patch:
$ sudo bash -c 'echo 2147483649 > /sys/class/net/eth4/napi_defer_hard_irqs'
bash: line 0: echo: write error: Numerical result out of range
Similarly, /sys/class/net/XXXXX/tx_queue_len is defined as unsigned:
include/linux/netdevice.h: unsigned int tx_queue_len;
And has an overflow check:
dev_change_tx_queue_len(..., unsigned long new_len):
if (new_len != (unsigned int)new_len)
return -ERANGE;
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20240904153431.307932-1-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Frank Sae says:
====================
Add driver for Motorcomm yt8821 2.5G ethernet phy
yt8521 and yt8531s as Gigabit transceiver use bit15:14(bit9 reserved
default 0) as phy speed mask, yt8821 as 2.5G transceiver uses bit9 bit15:14
as phy speed mask.
Be compatible to yt8821, reform phy speed mask and phy speed macro.
Based on update above, add yt8821 2.5G phy driver.
====================
Link: https://patch.msgid.link/20240901083526.163784-1-Frank.Sae@motor-comm.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add a driver for the motorcomm yt8821 2.5G ethernet phy. Verified the
driver on BPI-R3(with MediaTek MT7986(Filogic 830) SoC) development board,
which is developed by Guangdong Bipai Technology Co., Ltd..
yt8821 2.5G ethernet phy works in AUTO_BX2500_SGMII or FORCE_BX2500
interface, supports 2.5G/1000M/100M/10M speeds, and wol(magic package).
Signed-off-by: Frank Sae <Frank.Sae@motor-comm.com>
Reviewed-by: Sai Krishna <saikrishnag@marvell.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
yt8521 and yt8531s as Gigabit transceiver use bit15:14(bit9 reserved
default 0) as phy speed mask, yt8821 as 2.5G transceiver uses bit9 bit15:14
as phy speed mask.
Be compatible to yt8821, reform phy speed mask and phy speed macro.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Frank Sae <Frank.Sae@motor-comm.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Marc Kleine-Budde says:
====================
pull-request: can-next 2024-09-04-2
this is a pull request of 18 patches for net-next/master.
All 18 patches add support for CAN-FD IP core found on Rockchip
RK3568.
The first patch is co-developed by Elaine Zhang and me and adds DT
bindings documentation.
The remaining 17 patches are by me and add the driver in several
stages.
linux-can-next-for-6.12-20240904-2
* tag 'linux-can-next-for-6.12-20240904-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next:
can: rockchip_canfd: add support for CAN_CTRLMODE_BERR_REPORTING
can: rockchip_canfd: add support for CAN_CTRLMODE_LOOPBACK
can: rockchip_canfd: add hardware timestamping support
can: rockchip_canfd: enable full TX-FIFO depth of 2
can: rockchip_canfd: prepare to use full TX-FIFO depth
can: rockchip_canfd: add stats support for errata workarounds
can: rockchip_canfd: rkcanfd_get_berr_counter_corrected(): work around broken {RX,TX}ERRORCNT register
can: rockchip_canfd: implement workaround for erratum 12
can: rockchip_canfd: implement workaround for erratum 6
can: rockchip_canfd: add TX PATH
can: rockchip_canfd: rkcanfd_register_done(): add warning for erratum 5
can: rockchip_canfd: rkcanfd_handle_rx_int_one(): implement workaround for erratum 5: check for empty FIFO
can: rockchip_canfd: add notes about known issues
can: rockchip_canfd: add support for rk3568v3
can: rockchip_canfd: add quirk for broken CAN-FD support
can: rockchip_canfd: add quirks for errata workarounds
can: rockchip_canfd: add driver for Rockchip CAN-FD controller
dt-bindings: can: rockchip_canfd: add rockchip CAN-FD controller
====================
Link: https://patch.msgid.link/20240904130256.1965582-1-mkl@pengutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Hangbin Liu says:
====================
Bonding: support new xfrm state offload functions
Add 2 new xfrm state offload functions xdo_dev_state_advance_esn and
xdo_dev_state_update_stats for bonding. The xdo_dev_state_free will be
added by Jianbo's patchset [1]. I will add the bonding xfrm policy offload
in future.
v7: no update, just rebase the code.
v6: Use "Return: " based on ./scripts/kernel-doc (Simon Horman)
v5: Rebase to latest net-next, update function doc (Jakub Kicinski)
v4: Ratelimit pr_warn (Sabrina Dubroca)
v3: Re-format bond_ipsec_dev, use slave_warn instead of WARN_ON (Nikolay Aleksandrov)
Fix bond_ipsec_dev defination, add *. (Simon Horman, kernel test robot)
Fix "real" typo (kernel test robot)
v2: Add a function to process the common device checking (Nikolay Aleksandrov)
Remove unused variable (Simon Horman)
v1: lore.kernel.org/netdev/20240816035518.203704-1-liuhangbin@gmail.com
====================
Link: https://patch.msgid.link/20240904003457.3847086-1-liuhangbin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Currently, users can see that bonding supports IPSec HW offload via ethtool.
However, this functionality does not work with NICs like Mellanox cards when
ESN (Extended Sequence Numbers) is enabled, as ESN functions are not yet
supported. This patch adds ESN support to the bonding IPSec device offload,
ensuring proper functionality with NICs that support ESN.
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Acked-by: Jay Vosburgh <jv@jvosburgh.net>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This patch adds a common function to check the status of IPSec devices.
This function will be useful for future implementations, such as IPSec ESN
and state offload callbacks.
Suggested-by: Nikolay Aleksandrov <razor@blackwall.org>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Acked-by: Jay Vosburgh <jv@jvosburgh.net>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This NIC doesn't have hardware IRQ coalescing. Under high load,
interrupts can adversely affect performance. To mitigate this, enable
software IRQ coalescing by default. On my system this increases receive
throughput with iperf3 from 853 MBit/sec to 934 MBit/s, decreases
interrupts from 69489/sec to 2016/sec, and decreases CPU utilization
from 27% (4x Cortex-A53) to 14%. Latency is not affected (as far as I
can tell).
Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20240903184912.4151926-1-sean.anderson@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When reading registers from the PHY using the SIOCGMIIREG IOCTL any
errors returned from either mdiobus_read() or mdiobus_c45_read() are
ignored, and parts of the returned error is passed as the register value
back to user-space.
For example, if mdiobus_c45_read() is used with a bus that do not
implement the read_c45() callback -EOPNOTSUPP is returned. This is
however directly stored in mii_data->val_out and returned as the
registers content. As val_out is a u16 the error code is truncated and
returned as a plausible register value.
Fix this by first checking the return value for errors before returning
it as the register content.
Before this patch,
# phytool read eth0/0:1/0
0xffa1
After this change,
$ phytool read eth0/0:1/0
error: phy_read (-95)
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://patch.msgid.link/20240903171536.628930-1-niklas.soderlund+renesas@ragnatech.se
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Ido Schimmel says:
====================
Unmask upper DSCP bits - part 3
tl;dr - This patchset continues to unmask the upper DSCP bits in the
IPv4 flow key in preparation for allowing IPv4 FIB rules to match on
DSCP. No functional changes are expected.
The TOS field in the IPv4 flow key ('flowi4_tos') is used during FIB
lookup to match against the TOS selector in FIB rules and routes.
It is currently impossible for user space to configure FIB rules that
match on the DSCP value as the upper DSCP bits are either masked in the
various call sites that initialize the IPv4 flow key or along the path
to the FIB core.
In preparation for adding a DSCP selector to IPv4 and IPv6 FIB rules, we
need to make sure the entire DSCP value is present in the IPv4 flow key.
This patchset continues to unmask the upper DSCP bits, but this time in
the output route path, specifically in the callers of
ip_route_output_ports().
The next patchset (last) will handle the callers of
ip_route_output_key(). Split from this patchset to avoid going over the
15 patches limit.
No functional changes are expected as commit 1fa3314c14 ("ipv4:
Centralize TOS matching") moved the masking of the upper DSCP bits to
the core where 'flowi4_tos' is matched against the TOS selector.
====================
Link: https://patch.msgid.link/20240903135327.2810535-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The function is passed the full DS field in its 'tos' argument by its
two callers. It then masks the upper DSCP bits using RT_TOS() when
passing it to ip_route_output_ports().
Unmask the upper DSCP bits when passing 'tos' to ip_route_output_ports()
so that in the future it could perform the FIB lookup according to the
full DSCP value.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20240903135327.2810535-2-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Replace comma between expressions with semicolons.
Using a ',' in place of a ';' can have unintended side effects.
Although that is not the case here, it is seems best to use ';'
unless ',' is intended.
Found by inspection.
No functional change intended.
Compile tested only.
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20240904014441.1065753-1-nichen@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
RT_TOS() from include/uapi/linux/in_route.h is defined using
IPTOS_TOS_MASK from include/uapi/linux/ip.h. This is problematic for
files such as include/net/ip_fib.h that want to use RT_TOS() as without
including both header files kernel compilation fails:
In file included from ./include/net/ip_fib.h:25,
from ./include/net/route.h:27,
from ./include/net/lwtunnel.h:9,
from net/core/dst.c:24:
./include/net/ip_fib.h: In function ‘fib_dscp_masked_match’:
./include/uapi/linux/in_route.h:31:32: error: ‘IPTOS_TOS_MASK’ undeclared (first use in this function)
31 | #define RT_TOS(tos) ((tos)&IPTOS_TOS_MASK)
| ^~~~~~~~~~~~~~
./include/net/ip_fib.h:440:45: note: in expansion of macro ‘RT_TOS’
440 | return dscp == inet_dsfield_to_dscp(RT_TOS(fl4->flowi4_tos));
Therefore, cited commit changed linux/in_route.h to include linux/ip.h.
However, as reported by David, this breaks iproute2 compilation due
overlapping definitions between linux/ip.h and
/usr/include/netinet/ip.h:
In file included from ../include/uapi/linux/in_route.h:5,
from iproute.c:19:
../include/uapi/linux/ip.h:25:9: warning: "IPTOS_TOS" redefined
25 | #define IPTOS_TOS(tos) ((tos)&IPTOS_TOS_MASK)
| ^~~~~~~~~
In file included from iproute.c:17:
/usr/include/netinet/ip.h:222:9: note: this is the location of the previous definition
222 | #define IPTOS_TOS(tos) ((tos) & IPTOS_TOS_MASK)
Fix by changing include/net/ip_fib.h to include linux/ip.h. Note that
usage of RT_TOS() should not spread further in the kernel due to recent
work in this area.
Fixes: 1fa3314c14 ("ipv4: Centralize TOS matching")
Reported-by: David Ahern <dsahern@kernel.org>
Closes: https://lore.kernel.org/netdev/2f5146ff-507d-4cab-a195-b28c0c9e654e@kernel.org/
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Link: https://patch.msgid.link/20240903133554.2807343-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Marc Kleine-Budde <mkl@pengutronix.de> says:
This series adds support for the CAN-FD IP core found on the Rockchip
RK3568.
The IP core is a bit complicated and has several documented errata.
The driver is added in several stages, first the base driver including
the RX-path. Then several workarounds for errata and the TX-path, and
finally features like hardware time stamping, loop-back mode and
bus error reporting.
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-0-8ae22bcb27cc@pengutronix.de
[mkl: removed DTS patches]
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
So far the TX-FIFO is only used with a depth of 1, although the
hardware offers a depth of 2.
The workaround for the chips that are affected by erratum 6, i.e. EFF
frames may be send as standard frames, is to re-send the EFF frame.
This means the driver cannot queue the next frame for sending, as long
ad the EFF frame has not been successfully send out.
Introduce rkcanfd_get_effective_tx_free() that returns "0" space in
the TX-FIFO if an EFF frame is pending and the actual free space in
the TX-FIFO otherwise. Then replace rkcanfd_get_tx_free() with
rkcanfd_get_effective_tx_free() everywhere.
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-16-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Tests show that sometimes both CAN bus error counters read 0x0, even
if the controller is in warning mode
(RKCANFD_REG_STATE_ERROR_WARNING_STATE in RKCANFD_REG_STATE
set).
To work around this issue, if both error counters read from hardware
are 0x0, use the structure priv->bec, otherwise save the read value in
priv->bec.
In rkcanfd_handle_rx_int_one() decrement the priv->bec.rxerr for
successfully RX'ed CAN frames.
In rkcanfd_handle_tx_done_one() decrement the priv->bec.txerr for
successfully TX'ed CAN frames.
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-14-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The rk3568 CAN-FD errata sheet as of Tue 07 Nov 2023 11:25:31 +08:00
says:
| A dominant bit at the third bit of the intermission may cause a
| transmission error.
|
| When sampling the third bit of the intermission as a dominant bit, if
| tx_req is configured to transmit extended frames at this time, the
| extended frame may be sent to the bus in the format of a standard
| frame. The extended frame will be sent as a standard frame and will not
| result in error frames
Turn on "Interframe Spaceing RX Mode" only during TX to work around
erratum 12, according to rock-chip:
| Spaceing RX Mode = 1, the third Bit between frames cannot receive
| and send, and the fourth Bit begins to receive and send.
|
| Spaceing RX Mode = 0, allowing the third Bit between frames to
| receive and send.
Message-ID: <be72939f-0a9e-0608-dfff-7b0096a26eba@rock-chips.com>
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-13-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The rk3568 CAN-FD errata sheet as of Tue 07 Nov 2023 11:25:31 +08:00
says:
| The CAN controller's transmission of extended frames may
| intermittently change into standard frames.
|
| When using the CAN controller to send extended frames, if the
| 'tx_req' is configured as 1 and coincides with the internal
| transmission point, the extended frame will be transmitted onto the
| bus in the format of a standard frame.
To work around Erratum 6, the driver is in self-receiving mode (RXSTX)
and all received CAN frames are passed through rkcanfd_rxstx_filter().
Add a check in rkcanfd_rxstx_filter() whether the received frame
corresponds to the current outgoing frame, but the extended CAN ID has
been mangled to a standard ID. In this case re-send the original CAN
frame.
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-12-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The IP core has a TX event FIFO. In other IP cores, this type of FIFO
usually contains the events that a CAN frame has been successfully
sent. However, the IP core on the rk3568v2 the FIFO also holds events
of unsuccessful transmission attempts.
It turned out that the best way to work around this problem is to set
the IP core to self-receive mode (RXSTX), filter out the self-received
frames and insert them into the complete TX path.
Add a pair new functions to check if 2 struct canfd_frame are equal.
The 1st checks if the header of the CAN frames are equal, the 2nd
checks if the data portion are equal:
- rkcanfd_can_frame_header_equal()
- rkcanfd_can_frame_data_equal()
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-11-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The rk3568 CAN-FD errata sheet as of Tue 07 Nov 2023 11:25:31 +08:00
says:
| Erratum 5: Counters related to the TXFIFO and RXFIFO exhibit
| abnormal counting behavior.
|
| Due to a bug in the cross-asynchronous logic of the enable signals
| for rx_fifo_cnt and txe_fifo_frame_cnt counters, the counts of these
| two counters become inaccurate. This issue has resulted in the
| inability to use the TXFIFO and RXFIFO functions.
The errata sheet mentioned above states that only the rk3568v2 is
affected by this erratum, but tests with the rk3568v2 and rk3568v3
show that the RX_FIFO_CNT is sometimes too high. This leads to CAN
frames being read from the FIFO, which is then already empty.
Further tests on the rk3568v2 and rk3568v3 show that in this
situation (i.e. empty FIFO) all elements of the FIFO
header (frameinfo, id, ts) contain the same data.
On the rk3568v2 and rk3568v3, this problem only occurs extremely
rarely with the standard clock of 300 MHz, but almost immediately at
80 MHz.
To workaround this problem, check for empty FIFO with
rkcanfd_fifo_header_empty() in rkcanfd_handle_rx_int_one() and exit
early.
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-9-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Add driver for the Rockchip CAN-FD controller.
The IP core on the rk3568v2 SoC has 12 documented errata. Corrections
for these errata will be added in the upcoming patches.
Since several workarounds are required for the TX path, only add the
base driver that only implements the RX path.
Although the RX path implements CAN-FD support, it's not activated in
ctrlmode_supported, as the IP core in the rk3568v2 has problems with
receiving or sending certain CAN-FD frames.
Tested-by: Alibek Omarov <a1ba.omarov@gmail.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20240904-rockchip-canfd-v5-4-8ae22bcb27cc@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Add support for group stats for mac. The fbnic_set_counter help preserve
the default values for counters which are not touched by the driver.
The 'reset' flag in 'get_eth_mac_stats' allows to choose between
resetting the counter to recent most value or fetching the aggregated
values of the counter.
The 'fbnic_stat_rd64' read 64b stats counters in an atomic fashion using
read-read-read approach. This allows to isolate cases where counter is
moving too fast making accuracy of the counter questionable.
Command: ethtool -S eth0 --groups eth-mac
Example Output:
eth-mac-FramesTransmittedOK: 421644
eth-mac-FramesReceivedOK: 3849708
eth-mac-FrameCheckSequenceErrors: 0
eth-mac-AlignmentErrors: 0
eth-mac-OctetsTransmittedOK: 64799060
eth-mac-FramesLostDueToIntMACXmitError: 0
eth-mac-OctetsReceivedOK: 5134513531
eth-mac-FramesLostDueToIntMACRcvError: 0
eth-mac-MulticastFramesXmittedOK: 568
eth-mac-BroadcastFramesXmittedOK: 454
eth-mac-MulticastFramesReceivedOK: 276106
eth-mac-BroadcastFramesReceivedOK: 26119
eth-mac-FrameTooLongErrors: 0
Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add ethtool ops support and enable 'get_drvinfo' for fbnic. The driver
provides firmware version information while the driver name and bus
information is provided by ethtool_get_drvinfo().
Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the SO_PEEK_OFF selftest for UDP. In this patch, I mainly do
three things:
1. rename tcp_so_peek_off.c
2. adjust for UDP protocol
3. add selftests into it
Suggested-by: Jon Maloy <jmaloy@redhat.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Machon says:
====================
net: microchip: add FDMA library and use it for Sparx5
This patch series is the first of a 2-part series, that adds a new
common FDMA library for Microchip switch chips Sparx5 and lan966x. These
chips share the same FDMA engine, and as such will benefit from a
common library with a common implementation. This also has the benefit
of removing a lot open-coded bookkeeping and duplicate code for the two
drivers.
Additionally, upstreaming efforts for a third chip, lan969x, will begin
in the near future. This chip will use the new library too.
In this first series, the FDMA library is introduced and used by the
Sparx5 switch driver.
###################
# Example of use: #
###################
- Initialize the rx and tx fdma structs with values for: number of
DCB's, number of DB's, channel ID, DB size (data buffer size), and
total size of the requested memory. Also provide two callbacks:
nextptr_cb() and dataptr_cb() for getting the nextptr and dataptr.
- Allocate memory using fdma_alloc_phys() or fdma_alloc_coherent().
- Initialize the DCB's with fdma_dcb_init().
- Add new DCB's with fdma_dcb_add().
- Free memory with fdma_free_phys() or fdma_free_coherent().
#####################
# Patch breakdown: #
#####################
Patch #1: introduces library and selects it for Sparx5.
Patch #2: includes the fdma_api.h header and removes old symbols.
Patch #3: replaces old rx and tx variables with equivalent ones from the
fdma struct. Only the variables that can be changed without
breaking traffic is changed in this patch.
Patch #4: uses the library for allocation of rx buffers. This requires
quite a bit of refactoring in this single patch.
Patch #5: uses the library for adding DCB's in the rx path.
Patch #6: uses the library for freeing rx buffers.
Patch #7: uses the library helpers in the rx path.
Patch #8: uses the library for allocation of tx buffers. This requires
quite a bit of refactoring in this single patch.
Patch #9: uses the library for adding DCB's in the tx path.
Patch #10: uses the library helpers in the tx path.
Patch #11: ditches the existing linked list for storing buffer addresses,
and instead uses offsets into contiguous memory.
Patch #12: modifies existing rx and tx functions to be direction
independent.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
