| Age | Commit message (Collapse) | Author |
|---|
|
The removal of the supply clock rate check implies a need to remove
some unnecessary left-over data from the driver as well.
Signed-off-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
Add dt-schema for the I2C controller on the RTL9300 Ethernet switch
with integrated SoC.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
*-objs suffix is reserved rather for (user-space) host programs while
usually *-y suffix is used for kernel drivers (although *-objs works
for that purpose for now).
Let's correct the old usages of *-objs in Makefiles.
While at it, fix an obvious typo in help section of the Kconfig.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
Some S32G2/S32G3 SoC I2C particularities exist
such as different <clock divider, register value> pairs.
Those are addressed by adding specific S32G2 and S32G3
compatible strings.
Co-developed-by: Ionut Vicovan <Ionut.Vicovan@nxp.com>
Signed-off-by: Ionut Vicovan <Ionut.Vicovan@nxp.com>
Signed-off-by: Ciprian Marian Costea <ciprianmarian.costea@oss.nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
S32G2 and S32G3 SoCs use the same I2C controller as i.MX.
But there are small differences such as specific
<clock divider, register value> pairs.
So add new compatible strings 'nxp,s32g2-i2c'and 'nxp,s32g3-i2c'
for S32G2/S32G3 Socs.
Signed-off-by: Ciprian Marian Costea <ciprianmarian.costea@oss.nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
Variable cci_clk_rate is not effectively used, so delete it.
drivers/i2c/busses/i2c-qcom-cci.c:526:16: warning: variable ‘cci_clk_rate’ set but not used.
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=11532
Fixes: 8284750a1829 ("i2c: qcom-cci: Stop complaining about DT set clock rate")
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
The i2c-amd756-s4882 and i2c-nforce2-s4985 muxing pseudo-drivers were
written at a time when the i2c core did not support muxing. They are
essentially board-specific hacks. If we had to add support for these
boards today, we would implement it in a completely different way.
These Tyan server boards are 19 years old by now, so I very much doubt
any of these is still running today. So let's just drop this clumsy
code. If anyone really still needs this support and complains, I'll
rewrite it in a proper way on top of i2c-mux.
This also fixes the following warnings:
drivers/i2c/busses/i2c-amd756.c:286:20: warning: symbol 'amd756_smbus' was not declared. Should it be static?
drivers/i2c/busses/i2c-nforce2.c:123:20: warning: symbol 'nforce2_smbus' was not declared. Should it be static?
Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: Andi Shyti <andi.shyti@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
We are experiencing a problem with the i.MX I2C controller when
communicating with SMBus devices. We are seeing devices time-out because
the time between sending/receiving two bytes is too long, and the SMBus
device returns to the idle state. This happens because the i.MX I2C
controller sends and receives byte by byte. When a byte is sent or
received, we get an interrupt and can send or receive the next byte.
The current implementation sends a byte and then waits for an event
generated by the interrupt subroutine. After the event is received, the
next byte is sent and we wait again. This waiting allows the scheduler
to reschedule other tasks, with the disadvantage that we may not send
the next byte for a long time because the send task is not immediately
scheduled. For example, if the rescheduling takes more than 25ms, this
can cause SMBus devices to timeout and communication to fail.
This patch changes the behavior so that we do not reschedule the
send/receive task, but instead send or receive the next byte in the
interrupt subroutine. This prevents rescheduling and drastically reduces
the time between sending/receiving bytes. The cost in the interrupt
subroutine is relatively small, we check what state we are in and then
send/receive the next byte. Before we had to call wake_up, which is even
less expensive. However, we also had to do some scheduling, which
increased the overall cost compared to the new solution. The wake_up
function to wake up the send/receive task is now only called when an
error occurs or when the transfer is complete.
Signed-off-by: Stefan Eichenberger <stefan.eichenberger@toradex.com>
Acked-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
Separate the atomic, dma and non-dma use case as a preparation step for
moving the non-dma use case to the isr to avoid rescheduling while a
transfer is in progress.
Signed-off-by: Stefan Eichenberger <stefan.eichenberger@toradex.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Acked-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
According to the i.MX8M Mini reference manual chapter "16.1.4.2
Generation of Start" it is only necessary to poll for bus busy and
arbitration lost in multi master mode. This helps to avoid rescheduling
while the i2c bus is busy and avoids SMBus devices to timeout. For
backward compatibility, the single-master property needs to be
explicitly set to disable the bus busy polling.
Signed-off-by: Stefan Eichenberger <stefan.eichenberger@toradex.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Acked-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
Define a new ACPI HID for HJMC01
Signed-off-by: Hunter Yu <hunter.yu@hj-micro.com>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
The to-be-fixed commit rightfully reduced a race window, but also
removed a comment which is still helpful after the fix. Bring the
comment back.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Bjorn Andersson <andersson@kernel.org>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
PIC64GX i2c is compatible with the microchip corei2c, just add fallback
Signed-off-by: Pierre-Henry Moussay <pierre-henry.moussay@microchip.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
We never modify abort_sources, mark it as const.
Signed-off-by: Raag Jadav <raag.jadav@intel.com>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
After commit 0edb555a65d1 ("platform: Make platform_driver::remove()
return void") .remove() is (again) the right callback to implement for
platform drivers.
Convert all platform drivers below drivers/i2c to use .remove(), with
the eventual goal to drop struct platform_driver::remove_new(). As
.remove() and .remove_new() have the same prototypes, conversion is done
by just changing the structure member name in the driver initializer.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
In existing socs, I2C serial engine is sourced from XO (19.2MHz).
Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
The existing map table is based on 19.2MHz. This patch incorporates
the clock map table to derive the SCL clock from the 32MHz source
clock frequency.
Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
Reviewed-by: Bjorn Andersson <andersson@kernel.org>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
It is common practice in the downstream and upstream CCI dt to set CCI
clock rates to 19.2 MHz. It appears to be fairly common for initial code to
set the CCI clock rate to 37.5 MHz.
Applying the widely used CCI clock rates from downstream ought not to cause
warning messages in the upstream kernel where our general policy is to
usually copy downstream hardware clock rates across the range of Qualcomm
drivers.
Drop the warning it is pervasive across CAMSS users but doesn't add any
information or warrant any changes to the DT to align the DT clock rate to
the bootloader clock rate.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Link: https://lore.kernel.org/linux-arm-msm/20240824115900.40702-1-bryan.odonoghue@linaro.org
Signed-off-by: Richard Acayan <mailingradian@gmail.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
The CCI on the Snapdragon 670 is the interface for controlling camera
hardware over I2C. Add the compatible so it can be added to the SDM670
device tree.
Signed-off-by: Richard Acayan <mailingradian@gmail.com>
Reviewed-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
If not clearing the BB (bus busy) condition in the BER (bus error)
interrupt, the driver causes a timeout and hence the i2c core
doesn't do the i2c transfer retry but returns the driver's return
value to the upper layer instead.
Clear the BB condition in the BER interrupt and a software flag is
used. The driver does an i2c recovery without causing the timeout
if the flag is set.
Signed-off-by: Tyrone Ting <kfting@nuvoton.com>
Reviewed-by: Tali Perry <tali.perry1@gmail.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
Originally the driver uses the XMIT bit in SMBnST register to decide
the upcoming i2c transaction. If XMIT bit is 1, then it will be an i2c
write operation. If it's 0, then a read operation will be executed.
In slave mode the XMIT bit can simply be used directly to set the state.
XMIT bit can be used as an indication to the current state of the state
machine during slave operation. (meaning XMIT = 1 during writing and
XMIT = 0 during reading).
In master operation XMIT is valid only if there are no bus errors.
For example: in a multi master where the same module is switching from
master to slave at runtime, and there are collisions, the XMIT bit
cannot be trusted.
However the maser already "knows" what the bus state is, so this bit
is not needed and the driver can just track what it is currently doing.
Signed-off-by: Tyrone Ting <kfting@nuvoton.com>
Reviewed-by: Tali Perry <tali.perry1@gmail.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
The len variable is not initialized, which may cause the for loop to
behave unexpectedly.
Fixes: 9b25419ad397 ("i2c: amd-asf: Add routine to handle the ASF slave process")
Signed-off-by: Qianqiang Liu <qianqiang.liu@163.com>
Reviewed-by: Dan Carpenter <dan.carpenter@linaro.org>
Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
|
|
When we load SSDT from efi variable (specified with efivar_ssdt=<var>
boot command line argument) a name for the variable is allocated
dynamically because we traverse all EFI variables. Unlike ACPI table
data, which is later used by ACPI engine, the name is no longer needed
once traverse is complete -- don't forget to free this memory.
Same time we silently ignore any errors happened here let's print a
message if something went wrong (but do not exit since this is not a
critical error and the system should continue to boot).
Also while here -- add a note why we keep SSDT table on success.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
|
|
When CONFIG_CMDLINE_OVERRIDE or CONFIG_CMDLINE_FORCE are configured, the
command line provided by the boot stack should be ignored, and only the
built-in command line should be taken into account.
Add the required handling of this when dealing with initrd= or dtb=
command line options in the EFI stub.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
|
|
CONFIG_CMDLINE, when set, is supposed to serve either as a fallback when
no command line is provided by the bootloader, or to be taken into account
unconditionally, depending on the configured options.
The initrd and dtb loader ignores CONFIG_CMDLINE in either case, and
only takes the EFI firmware provided load options into account. This
means that configuring the kernel with initrd= or dtb= on the built-in
command line does not produce the expected result.
Fix this by doing a separate pass over the built-in command line when
dealing with initrd= or dtb= options.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull hotfixes from Andrew Morton:
"10 hotfixes, 7 of which are cc:stable. All singletons, please see the
changelogs for details"
* tag 'mm-hotfixes-stable-2024-11-16-15-33' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
mm: revert "mm: shmem: fix data-race in shmem_getattr()"
ocfs2: uncache inode which has failed entering the group
mm: fix NULL pointer dereference in alloc_pages_bulk_noprof
mm, doc: update read_ahead_kb for MADV_HUGEPAGE
fs/proc/task_mmu: prevent integer overflow in pagemap_scan_get_args()
sched/task_stack: fix object_is_on_stack() for KASAN tagged pointers
crash, powerpc: default to CRASH_DUMP=n on PPC_BOOK3S_32
mm/mremap: fix address wraparound in move_page_tables()
tools/mm: fix compile error
mm, swap: fix allocation and scanning race with swapoff
|
|
Revert d949d1d14fa2 ("mm: shmem: fix data-race in shmem_getattr()") as
suggested by Chuck [1]. It is causing deadlocks when accessing tmpfs over
NFS.
As Hugh commented, "added just to silence a syzbot sanitizer splat: added
where there has never been any practical problem".
Link: https://lkml.kernel.org/r/ZzdxKF39VEmXSSyN@tissot.1015granger.net [1]
Fixes: d949d1d14fa2 ("mm: shmem: fix data-race in shmem_getattr()")
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Jeongjun Park <aha310510@gmail.com>
Cc: Yu Zhao <yuzhao@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Pull ARM fixes from Russell King:
- Fix kernel mapping for XIP kernels
- Fix SMP support for XIP kernels
- Fix complication corner case with CFI
- Fix a typo in nommu code
- Fix cacheflush syscall when PAN is enabled on LPAE platforms
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rmk/linux:
ARM: fix cacheflush with PAN
ARM: 9435/1: ARM/nommu: Fix typo "absence"
ARM: 9434/1: cfi: Fix compilation corner case
ARM: 9420/1: smp: Fix SMP for xip kernels
ARM: 9419/1: mm: Fix kernel memory mapping for xip kernels
|
|
Pull drm fix from Dave Airlie:
"Alex sent on a last minute revert for a amdgpu/swsmu regression:
- revert patch to fix swsmu regression"
* tag 'drm-fixes-2024-11-17' of https://gitlab.freedesktop.org/drm/kernel:
Revert "drm/amd/pm: correct the workload setting"
|
|
https://gitlab.freedesktop.org/agd5f/linux into drm-fixes
amd-drm-fixes-6.12-2024-11-16:
amdgpu:
- Revert a swsmu patch to fix a regression
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241116145320.2507156-1-alexander.deucher@amd.com
|
|
The hash_combine() could be trapped when compiled with sanitizer like "zig cc"
or clang with signed-integer-overflow option. This patch parameters and return
type to unsigned long to remove the potential overflow.
Signed-off-by: Sidong Yang <sidong.yang@furiosa.ai>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20241116081054.65195-1-sidong.yang@furiosa.ai
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
llvm 19 fails to compile arena self test:
CLNG-BPF [test_progs] verifier_arena_large.bpf.o
progs/verifier_arena_large.c:90:24: error: unsupported signed division, please convert to unsigned div/mod.
90 | pg_idx = (pg - base) / PAGE_SIZE;
Though llvm <= 18 and llvm >= 20 don't have this issue,
fix the test to avoid the build error.
Reported-by: Jiri Olsa <olsajiri@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull ring buffer fixes from Steven Rostedt:
- Revert: "ring-buffer: Do not have boot mapped buffers hook to CPU
hotplug"
A crash that happened on cpu hotplug was actually caused by the
incorrect ref counting that was fixed by commit 2cf9733891a4
("ring-buffer: Fix refcount setting of boot mapped buffers"). The
removal of calling cpu hotplug callbacks on memory mapped buffers was
not an issue even though the tests at the time pointed toward it. But
in fact, there's a check in that code that tests to see if the
buffers are already allocated or not, and will not allocate them
again if they are. Not calling the cpu hotplug callbacks ended up not
initializing the non boot CPU buffers.
Simply remove that change.
- Clear all CPU buffers when starting tracing in a boot mapped buffer
To properly process events from a previous boot, the address space
needs to be accounted for due to KASLR and the events in the buffer
are updated accordingly when read. This also requires that when the
buffer has tracing enabled again in the current boot that the buffers
are reset so that events from the previous boot do not interact with
the events of the current boot and cause confusing due to not having
the proper meta data.
It was found that if a CPU is taken offline, that its per CPU buffer
is not reset when tracing starts. This allows for events to be from
both the previous boot and the current boot to be in the buffer at
the same time. Clear all CPU buffers when tracing is started in a
boot mapped buffer.
* tag 'trace-ringbuffer-v6.12-rc7-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tracing/ring-buffer: Clear all memory mapped CPU ring buffers on first recording
Revert: "ring-buffer: Do not have boot mapped buffers hook to CPU hotplug"
|
|
This operation is used by alienware-wmi driver to avoid brute-forcing
operation 0x03.
Signed-off-by: Kurt Borja <kuurtb@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20241111183639.14726-1-kuurtb@gmail.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
WMAX_METHOD_THERMAL_INFORMATION has a *system description* operation
that outputs a buffer with the following structure:
out[0] -> Number of fans
out[1] -> Number of sensors
out[2] -> 0x00
out[3] -> Number of thermal modes
This is now used by create_thermal_profile() to retrieve available thermal
codes instead of brute-forcing every ID.
Tested on an Alienware x15 R1. Verified by checking ACPI tables of
supported models.
Signed-off-by: Kurt Borja <kuurtb@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20241111183623.14691-1-kuurtb@gmail.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
Adds support to Alienware x17 R2
Tested-by: Samith Castro <SamithNarayam@hotmail.com>
Signed-off-by: Kurt Borja <kuurtb@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20241111183609.14653-1-kuurtb@gmail.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
Adds thermal + gmode quirk to:
- Dell G15 5510
- Dell G15 5511
- Dell G15 5515
- Dell G3 3500
- Dell G3 3590
- Dell G5 5500
Adds thermal quirk to:
- Alienware m18 R2
- Alienware m17 R5 AMD
Support for these models was manually verified by reading their
respective ACPI tables.
Signed-off-by: Kurt Borja <kuurtb@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20241111183546.14617-1-kuurtb@gmail.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
alienware_quirks[] entries are now ordered alphabetically
Signed-off-by: Kurt Borja <kuurtb@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20241111183520.14573-1-kuurtb@gmail.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
This reverts commit 74e1006430a5377228e49310f6d915628609929e.
This causes a regression in the workload selection.
A more extensive fix is being worked on.
For now, revert.
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3618
Fixes: 74e1006430a5 ("drm/amd/pm: correct the workload setting")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Xuan Zhuo says:
====================
virtio-net: support AF_XDP zero copy (tx)
XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero
copy feature of xsk (XDP socket) needs to be supported by the driver. The
performance of zero copy is very good. mlx5 and intel ixgbe already support
this feature, This patch set allows virtio-net to support xsk's zerocopy xmit
feature.
At present, we have completed some preparation:
1. vq-reset (virtio spec and kernel code)
2. virtio-core premapped dma
3. virtio-net xdp refactor
So it is time for Virtio-Net to complete the support for the XDP Socket
Zerocopy.
Virtio-net can not increase the queue num at will, so xsk shares the queue with
kernel.
This patch set includes some refactor to the virtio-net to let that to support
AF_XDP.
The current configuration sets the virtqueue (vq) to premapped mode,
implying that all buffers submitted to this queue must be mapped ahead
of time. This presents a challenge for the virtnet send queue (sq): the
virtnet driver would be required to keep track of dma information for vq
size * 17, which can be substantial. However, if the premapped mode were
applied on a per-buffer basis, the complexity would be greatly reduced.
With AF_XDP enabled, AF_XDP buffers would become premapped, while kernel
skb buffers could remain unmapped.
We can distinguish them by sg_page(sg), When sg_page(sg) is NULL, this
indicates that the driver has performed DMA mapping in advance, allowing
the Virtio core to directly utilize sg_dma_address(sg) without
conducting any internal DMA mapping. Additionally, DMA unmap operations
for this buffer will be bypassed.
ENV: Qemu with vhost-user(polling mode).
Host CPU: Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz
testpmd> show port stats all
######################## NIC statistics for port 0 ########################
RX-packets: 19531092064 RX-missed: 0 RX-bytes: 1093741155584
RX-errors: 0
RX-nombuf: 0
TX-packets: 5959955552 TX-errors: 0 TX-bytes: 371030645664
Throughput (since last show)
Rx-pps: 8861574 Rx-bps: 3969985208
Tx-pps: 8861493 Tx-bps: 3969962736
############################################################################
testpmd> show port stats all
######################## NIC statistics for port 0 ########################
RX-packets: 68152727 RX-missed: 0 RX-bytes: 3816552712
RX-errors: 0
RX-nombuf: 0
TX-packets: 68114967 TX-errors: 33216 TX-bytes: 3814438152
Throughput (since last show)
Rx-pps: 6333196 Rx-bps: 2837272088
Tx-pps: 6333227 Tx-bps: 2837285936
############################################################################
But AF_XDP consumes more CPU for tx and rx napi(100% and 86%).
====================
Link: https://patch.msgid.link/20241112012928.102478-1-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Now, we support AF_XDP(xsk). Add NETDEV_XDP_ACT_XSK_ZEROCOPY to
xdp_features.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-14-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
If send queue sent some packets, we update the tx timeout
record to prevent the tx timeout.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-13-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The driver's tx napi is very important for XSK. It is responsible for
obtaining data from the XSK queue and sending it out.
At the beginning, we need to trigger tx napi.
virtnet_free_old_xmit distinguishes three type ptr(skb, xdp frame, xsk
buffer) by the last bits of the pointer.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-12-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Since xsk's TX queue is consumed by TX NAPI, if sq is bound to xsk, then
we must stop tx napi from being disabled.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-11-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This patch implement the logic of bind/unbind xsk pool to sq and rq.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-10-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Because the af-xdp will introduce a new xmit type, so I refactor the
xmit type mechanism first.
We know both xdp_frame and sk_buff are at least 4 bytes aligned.
For the xdp tx, we do not pass any pointer to virtio core as data,
we just need to pass the len of the packet. So we will push len
to the void pointer. We can make sure the pointer is 4 bytes aligned.
And the data structure of AF_XDP also is at least 4 bytes aligned.
So the last two bits of the pointers are free, we can't use these to
distinguish them.
00 for skb
01 for SKB_ORPHAN
10 for XDP
11 for AF-XDP tx
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-9-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Now, this API is useless. remove it.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-8-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
virtio-net rq submits premapped per-buffer by setting sg page to NULL;
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-7-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Two APIs are introduced to submit premapped per-buffers.
int virtqueue_add_inbuf_premapped(struct virtqueue *vq,
struct scatterlist *sg, unsigned int num,
void *data,
void *ctx,
gfp_t gfp);
int virtqueue_add_outbuf_premapped(struct virtqueue *vq,
struct scatterlist *sg, unsigned int num,
void *data,
gfp_t gfp);
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-6-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The current configuration sets the virtqueue (vq) to premapped mode,
implying that all buffers submitted to this queue must be mapped ahead
of time. This presents a challenge for the virtnet send queue (sq): the
virtnet driver would be required to keep track of dma information for vq
size * 17, which can be substantial. However, if the premapped mode were
applied on a per-buffer basis, the complexity would be greatly reduced.
With AF_XDP enabled, AF_XDP buffers would become premapped, while kernel
skb buffers could remain unmapped.
And consider that some sgs are not generated by the virtio driver,
that may be passed from the block stack. So we can not change the
sgs, new APIs are the better way.
So we pass the new argument 'premapped' to indicate the buffers
submitted to virtio are premapped in advance. Additionally,
DMA unmap operations for these buffers will be bypassed.
Suggested-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-5-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The subsequent commit needs to know whether every indirect buffer is
premapped or not. So we need to introduce an extra struct for every
indirect buffer to record this info.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241112012928.102478-4-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
