Add support for the multi-planar variants of NV12. The RGA is now able
to exchange DMABUFs with other devices that only support multi-planar
NV12, for example the Hantro JPEG encoder.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Multi-planar formats may have multiple planes that must be handled and
correctly mapped into a continuous buffer for the RGA by using the DMA
descriptors.
The plane offsets in the continuous mapping may now start at page
boundaries and the previous calculation based on the frame sizes is only
valid for planar buffers in a single memory. Therefore, the offsets must
be detected and set while creating the mapping.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Switch to the multi-planar API, which allows to handle buffers with
separate planes.
The RGA driver doesn't expose multi-planar formats, yet. The existing
contiguous planar formats can be used with the multi-planar API as well,
but the multi-planar API is required for multi-planar formats.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Use the provided V4L2_TYPE_IS_{OUTPUT,CAPTURE} macros to check if the
buffer or queue is OUTPUT or CAPTURE. The macros work also work for the
_MPLANE buffer and queue types and make it easier to switch to the
multi-planar API.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The local variable allows to simplify the accessed to the format and
makes it easier to change the type of the format.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Use the pixelformat instead of the v4l2_format to find the rga_fmt. This
avoids knowing the structure and type of v4l2_format in rga_fmt_find and
simplifies the function.
Also cleanup the users of the function. In try_fmt always return the
found pixel format to make sure that the pixel format is always set.
Thus, we can be sure that we will find the rga_fmt in s_fmt and can drop
the check if a given format has been found.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The try_fmt should limit the width and height to the know limits of the
RGA. Use the clamp() helper instead of open coding the clamping.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The RGA DMA descriptor list contains only 32-bit addresses. Set the
dma_mask to only allocate memory that is addressable by the descriptors.
This prevents errors when preparing vb2 buffers that were allocated by
the RGA.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Calculate the plane offsets and store them with the video buffer while
creating the buffer mapping.
This allows the driver to more freely handle the memory of the DMA
mapping as the offsets and the mapping can be kept in sync.
The driver still has to update the offsets to respect the configured
cropping and rotation, but this calculation is now separated from the
calculation of the plane offsets.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Split the register setup for the source and destination video buffers
into separate functions.
This is a cleanup to make the code more readable by separating the
offset calculation for the different buffers and prepares the driver for
using pre-calculated offsets of planes.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The RGA driver allocates two buffers for the DMA descriptors of the
input and output buffers. Whenever a new job is processed, the
descriptor list is updated for the current buffers.
By updating the descriptor list during buf_prepare, it is possible to
correctly fail DMABUF imports if the buffers that shall be imported are
not within the 32 bit address range that can be addressed by the RGA.
Managing the DMA descriptor list with the buffer also makes it easier to
track the buffer mapping and the plane offsets into this mapping.
The cost is that the driver now requires DMA coherent memory per buffer
for the descriptor list. However, the size scales with the size of the
video buffers and is not allocated if the RGA is not used.
While at it, use dma_alloc_coherent to allocate the descriptors and get
rid of the virt_to_phys calls to get the physical addresses.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The IOMMU of the RGA is programmed with a list of DMA descriptors that
contain an 32 bit address per 4k page in the video buffers. The address
in the descriptor points to the start address of the page.
Introduce 'struct rga_dma_desc' to make the handling of the DMA
descriptors explicit instead of hiding them behind standard types.
While at it, use provided helpers for iterating the sg_table instead of
manually calculating the DMA addresses.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
When using 32 bit RGB formats, the RGA on the rk3568 produces wrong
colors as the wrong color channels are read or written. The reason is
that the format description for the channel swizzeling is wrong and the
wrong bits are configured. For example, when converting ARGB32 to NV12,
the alpha channel is used as blue channel.. This doesn't happen if the
color format is the same on both sides.
Fix the color_swap settings of the formats to correctly handle 32 bit
RGB formats.
For RGA_COLOR_FMT_XBGR8888, the RGA_COLOR_ALPHA_SWAP bit doesn't have an
effect. Thus, it isn't possible to handle the V4L2_PIX_FMT_XRGB32. Thus,
it is removed from the list of supported formats.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Stop timer in the 'trigger' and 'sync_stop' callbacks since we want
the timer to be stopped before the DMA buffer is released. Otherwise,
it could trigger a kernel panic in some circumstances, for instance
when the DMA buffer is already released but the timer callback is
still running.
Signed-off-by: Ivan Orlov <ivan.orlov0322@gmail.com>
Link: https://lore.kernel.org/r/20231206223211.12761-1-ivan.orlov0322@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Declare power-domain names "top", "ife0" and "ife1" eponymously for the
power-domains TITAN_TOP_GDSC, IFE_0_GDSC and IFE_1_GDSC respectively.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Some platforms such as SC7280 have 3 CSIDs and 2 CSID-lites but current
code has hardcoded 2 as the maximum number of CSIDs. Remove the hardcoded
maximum number of VFEs to handle all possible combinations of CSIDs and
CSID-lites.
Signed-off-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Some platforms such as SC7280 have three VFEs and two VFE-lites. Current
code has hard-coded two as the maximum number of VFEs. Remove the
hard-coded maximum number of VFEs to handle all possible combinations of
VFEs and VFE-lites.
Signed-off-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202311200405.h6G4L9oe-lkp@intel.com
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Right now we use fixed indexes to assign power-domains, with a
requirement for the TOP GDSC to come last in the list.
Adding support for named power-domains means the declaration in the dtsi
can come in any order.
After this change we continue to support the old indexing - if a SoC
resource declaration or the in-use dtb doesn't declare power-domain names
we fall back to the default legacy indexing.
From this point on though new SoC additions should contain named
power-domains, eventually we will drop support for legacy indexing.
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Moving the location of the hooks to VFE power domains has several
advantages.
1. Separation of concerns and functional decomposition.
vfe.c should be responsible for and know best how manage
power-domains for a VFE, excising from camss.c follows this
principle.
2. Embedding a pointer to genpd in struct camss_vfe{} meas that we can
dispense with a bunch of kmalloc array inside of camss.c.
3. Splitting up titan top gdsc from vfe/ife gdsc provides a base for
breaking up magic indexes in dtsi.
Suggested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
For the various versions of VFE we have a boiler-plate
pm_domain_on/pm_domain_off callback pair of the general form.
- Error check.
Not always done but applicable to all.
- device_link_add (DL_FLAG_STATELESS | DL_FLAG_PM_RUNTIME |
DL_FLAG_RPM_ACTIVE);
- Error check returning -EINVAL on error.
- Return 0
Reduce the pattern down to a common callback. VFE 4.1 is a special case
which to me also indicates that it is worthwhile maintaining an indirection
for the vfe_pm_domain_{on|off} for now.
Otherwise lets chuck out a bunch of needlessly replicated code.
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Suggested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Right now we use the top-level camss structure to provide pointers via
VFE id index back to genpd linkages.
In effect this hard-codes VFE indexes to power-domain indexes in the
dtsi and mandates a very particular ordering of power domains in the
dtsi, which bears no relationship to a real hardware dependency.
As a first step to rationalising the VFE power-domain code and breaking
the magic indexing in dtsi use per-VFE pointers to genpd linkages.
The top-level index in msm_vfe_subdev_init is still used to attain the
initial so no functional or logical change arises from this change.
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
At the moment we have some complex code for determining if a VFE requires a
power-domain attachment. Particularly discordant in this scheme is the
subtle reliance on VFE and VFE Lite declaration ordering in our resources.
VFE id is used to determine if a VFE is lite or not and consequently if a
VFE requires power-domain attachment. VFE Lite though is not a correct
delineation between power-domain and non power-domain state since early
SoCs have neither VFE Lite nor power-domains attached to VFEs.
Introduce has_pd to the VFE resource structure to allow the CAMSS code to
understand if it needs to try to attach a power-domain for a given VFE.
As a side-effect from this we no longer need to care about VFE Lite or
non-Lite or the id number associated with either and which order the
VFE/VFE Lite was declared in.
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The function stk1160_dbg gets called too many times, which causes
the output to get flooded with messages. Since stk1160_dbg uses
printk, it is now replaced with printk_ratelimited.
Suggested-by: Phillip Potter <phil@philpotter.co.uk>
Signed-off-by: Ghanshyam Agrawal <ghanshyam1898@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The dma_q->p_left_data alloced by kzalloc should be freed in all the
following error handling paths. However, it hasn't been freed in the
allocation error paths of dev->video_mode.isoc_ctl.urb and
dev->video_mode.isoc_ctl.transfer_buffer.
On the other hand, the dma_q->p_left_data did be freed in the
error-handling paths after that of dev->video_mode.isoc_ctl.urb and
dev->video_mode.isoc_ctl.transfer_buffer, by calling
cx231xx_uninit_isoc(dev). So the same free operation should be done in
error-handling paths of those two allocation.
Fixes: 64fbf44455 ("[media] cx231xx: Added support for Carraera, Shelby, RDx_253S and VIDEO_GRABBER")
Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The StarFive Camera Subsystem is only present on the StarFive JH7110
SoC. Hence add a dependency on ARCH_STARFIVE, to prevent asking the
user about this driver when configuring a kernel without StarFive SoC
support.
Fixes: bba185d141 ("media: staging: media: starfive: camss: Add core driver")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Changhuang Liang <changhuang.liang@starfivetech.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The Chips&Media Wave 5 Series multi-standard codec IP is currently only
supported on Texas Instruments K3 architecture. Hence add a dependency on
ARCH_K3, to prevent asking the user about this driver when configuring a
kernel without Texas Instruments K3 Multicore SoC support.
Fixes: 9707a6254a ("media: chips-media: wave5: Add the v4l2 layer")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
[hverkuil: small update in commit log, and reformatted the Fixes tag]
These > comparisons should be >=. The formats->fmts[] array is either
a pointer to isp_formats_sink[] or isp_formats_source[] respectively.
Fixes: e57854628f ("media: staging: media: starfive: camss: Add ISP driver")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Changhuang Liang <changhuang.liang@starfivetech.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
For dmabuf import users to be able to use the vaddr from another
videobuf2-dma-sg source, the exporter needs to set a proper vaddr on
vb2_dma_sg_dmabuf_ops_vmap callback. This patch adds vmap on map if
buf->vaddr was not set.
Cc: stable@kernel.org
Fixes: 7938f42181 ("dma-buf-map: Rename to iosys-map")
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Acked-by: Tomasz Figa <tfiga@chromium.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Add AV1 information in visl documentation.
Reviewed-by: Daniel Almeida <daniel.almeida@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Let the visl test driver accept the AV1 pixel format.
Reviewed-by: Daniel Almeida <daniel.almeida@collabora.com>
Tested-by: Daniel Almeida <daniel.almeida@collabora.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Defines prefixed with "CONFIG" should be limited to proper Kconfig options,
that are introduced in a Kconfig file.
In the driver code, there is a define for CONFIG_VB2_GSC_DMA_CONTIG, but
this is not used anywhere in the code.
Just remove this unused definition.
No functional change.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Instead of walking the entire parent node for something that has the
right compatible, use the scp_get() function provided by the MediaTek
SCP remoteproc driver to retrieve a handle to mtk_scp through the
devicetree "mediatek,scp" (phandle) property.
In case of multi-core SCP, this also allows to select a specific core.
Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Alexandre Mergnat <amergnat@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The MDP3 RDMA needs to communicate with the SCP remote processor: allow
specifying a phandle to a SCP core.
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Alexandre Mergnat <amergnat@baylibre.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
- Fix asm operand number out of range build error in bug table
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQS86RI+GtKfB8BJu973ErUQojoPXwUCZXCQDQAKCRD3ErUQojoP
X0agAPoDMDmnwLC6S596+J+6VB4UTRJxdn8NrAb5Mq8GXUlKYQD/blU1KHZ4TUEY
yTpqHzcl6iLjkaMjEvzZsAWGyyc61gs=
=yGBG
-----END PGP SIGNATURE-----
Merge tag 'parisc-for-6.7-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux
Pull parisc fix from Helge Deller:
"A single line patch for parisc which fixes the build in tinyconfig
configurations:
- Fix asm operand number out of range build error in bug table"
* tag 'parisc-for-6.7-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
parisc: Fix asm operand number out of range build error in bug table
It's been reported that DSI host driver's detach can be called without
the attach ever happening:
https://lore.kernel.org/all/20230412073954.20601-1-tony@atomide.com/
After reading the code, I think this is what happens:
We have a DSI host defined in the device tree and a DSI peripheral under
that host (i.e. an i2c device using the DSI as data bus doesn't exhibit
this behavior).
The host driver calls mipi_dsi_host_register(), which causes (via a few
functions) mipi_dsi_device_add() to be called for the DSI peripheral. So
now we have a DSI device under the host, but attach hasn't been called.
Normally the probing of the devices continues, and eventually the DSI
peripheral's driver will call mipi_dsi_attach(), attaching the
peripheral.
However, if the host driver's probe encounters an error after calling
mipi_dsi_host_register(), and before the peripheral has called
mipi_dsi_attach(), the host driver will do cleanups and return an error
from its probe function. The cleanups include calling
mipi_dsi_host_unregister().
mipi_dsi_host_unregister() will call two functions for all its DSI
peripheral devices: mipi_dsi_detach() and mipi_dsi_device_unregister().
The latter makes sense, as the device exists, but the former may be
wrong as attach has not necessarily been done.
To fix this, track the attached state of the peripheral, and only detach
from mipi_dsi_host_unregister() if the peripheral was attached.
Note that I have only tested this with a board with an i2c DSI
peripheral, not with a "pure" DSI peripheral.
However, slightly related, the unregister machinery still seems broken.
E.g. if the DSI host driver is unbound, it'll detach and unregister the
DSI peripherals. After that, when the DSI peripheral driver unbound
it'll call detach either directly or using the devm variant, leading to
a crash. And probably the driver will crash if it happens, for some
reason, to try to send a message via the DSI bus.
But that's another topic.
Tested-by: H. Nikolaus Schaller <hns@goldelico.com>
Acked-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Tested-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230921-dsi-detach-fix-v1-1-d0de2d1621d9@ideasonboard.com
smatch reports:
drivers/gpu/drm/drm_framebuffer.c:654 drm_mode_getfb2_ioctl() error: uninitialized symbol 'ret'.
'ret' is possibly not set when there are no errors, causing the error
above. I can't say if that ever happens in real-life, but in any case I
think it is good to initialize 'ret' to 0.
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231103-uninit-fixes-v2-2-c22b2444f5f5@ideasonboard.com
smatch reports:
drivers/gpu/drm/drm_file.c:967 drm_show_memory_stats() error: uninitialized symbol 'supported_status'.
'supported_status' is only set in one code path. I'm not familiar with
the code to say if that path will always be ran in real life, but
whether that is the case or not, I think it is good to initialize
'supported_status' to 0 to silence the warning (and possibly fix a bug).
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231103-uninit-fixes-v2-1-c22b2444f5f5@ideasonboard.com
The Framework 16" laptop has the same controller as other Framework
models. Apply the presence detection quirk.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20231206193927.2996-1-mario.limonciello@amd.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Support the platform with PCC type3 and interrupt ack. And a version
specific structure is introduced to handle the difference between the
device in the code.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
The email domain name in Contact is wrong. So this patch has to modify it.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
Driver gets the PCC channel id by using the PCC GAS in _CRS.
But, currently, if the firmware has no _CRS method on platform, there
is not any failure log. So this patch adds the log for this.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
Fix some incorrect format strings.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2023-12-05 (ice, i40e, iavf)
This series contains updates to ice, i40e and iavf drivers.
Michal fixes incorrect usage of VF MSIX value and index calculation for
ice.
Marcin restores disabling of Rx VLAN filtering which was inadvertently
removed for ice.
Ivan Vecera corrects improper messaging of MFS port for i40e.
Jake fixes incorrect checking of coalesce values on iavf.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
iavf: validate tx_coalesce_usecs even if rx_coalesce_usecs is zero
i40e: Fix unexpected MFS warning message
ice: Restore fix disabling RX VLAN filtering
ice: change vfs.num_msix_per to vf->num_msix
====================
Link: https://lore.kernel.org/r/20231205211918.2123019-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Due to linkwatch_forget_dev() (and perhaps others?) checking for
list_empty(&dev->link_watch_list), we must have all manipulations
of even the local on-stack list 'wrk' here under spinlock, since
even that list can be reached otherwise via dev->link_watch_list.
This is already the case, but makes this a bit counter-intuitive,
often local lists are used to _not_ have to use locking for their
local use.
Remove the local list as it doesn't seem to serve any purpose.
While at it, move a variable declaration into the loop using it.
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20231205170011.56576dcc1727.I698b72219d9f6ce789bd209b8f6dffd0ca32a8f2@changeid
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In 4a56212774, USXGMII support was added for 6393X, but this was
lost in the PCS conversion (the blamed commit), most likely because
these efforts where more or less done in parallel.
Restore this feature by porting Michal's patch to fit the new
implementation.
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Michal Smulski <michal.smulski@ooma.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Fixes: e5b732a275 ("net: dsa: mv88e6xxx: convert 88e639x to phylink_pcs")
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Link: https://lore.kernel.org/r/20231205221359.3926018-1-tobias@waldekranz.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch is based on a detailed report and ideas from Yepeng Pan
and Christian Rossow.
ACK seq validation is currently following RFC 5961 5.2 guidelines:
The ACK value is considered acceptable only if
it is in the range of ((SND.UNA - MAX.SND.WND) <= SEG.ACK <=
SND.NXT). All incoming segments whose ACK value doesn't satisfy the
above condition MUST be discarded and an ACK sent back. It needs to
be noted that RFC 793 on page 72 (fifth check) says: "If the ACK is a
duplicate (SEG.ACK < SND.UNA), it can be ignored. If the ACK
acknowledges something not yet sent (SEG.ACK > SND.NXT) then send an
ACK, drop the segment, and return". The "ignored" above implies that
the processing of the incoming data segment continues, which means
the ACK value is treated as acceptable. This mitigation makes the
ACK check more stringent since any ACK < SND.UNA wouldn't be
accepted, instead only ACKs that are in the range ((SND.UNA -
MAX.SND.WND) <= SEG.ACK <= SND.NXT) get through.
This can be refined for new (and possibly spoofed) flows,
by not accepting ACK for bytes that were never sent.
This greatly improves TCP security at a little cost.
I added a Fixes: tag to make sure this patch will reach stable trees,
even if the 'blamed' patch was adhering to the RFC.
tp->bytes_acked was added in linux-4.2
Following packetdrill test (courtesy of Yepeng Pan) shows
the issue at hand:
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 1024) = 0
// ---------------- Handshake ------------------- //
// when window scale is set to 14 the window size can be extended to
// 65535 * (2^14) = 1073725440. Linux would accept an ACK packet
// with ack number in (Server_ISN+1-1073725440. Server_ISN+1)
// ,though this ack number acknowledges some data never
// sent by the server.
+0 < S 0:0(0) win 65535 <mss 1400,nop,wscale 14>
+0 > S. 0:0(0) ack 1 <...>
+0 < . 1:1(0) ack 1 win 65535
+0 accept(3, ..., ...) = 4
// For the established connection, we send an ACK packet,
// the ack packet uses ack number 1 - 1073725300 + 2^32,
// where 2^32 is used to wrap around.
// Note: we used 1073725300 instead of 1073725440 to avoid possible
// edge cases.
// 1 - 1073725300 + 2^32 = 3221241997
// Oops, old kernels happily accept this packet.
+0 < . 1:1001(1000) ack 3221241997 win 65535
// After the kernel fix the following will be replaced by a challenge ACK,
// and prior malicious frame would be dropped.
+0 > . 1:1(0) ack 1001
Fixes: 354e4aa391 ("tcp: RFC 5961 5.2 Blind Data Injection Attack Mitigation")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Yepeng Pan <yepeng.pan@cispa.de>
Reported-by: Christian Rossow <rossow@cispa.de>
Acked-by: Neal Cardwell <ncardwell@google.com>
Link: https://lore.kernel.org/r/20231205161841.2702925-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>