the guest when it has been set already on the host
- Make sure cacheinfo structures memory is allocated to address a boot NULL
ptr dereference on Intel Meteor Lake which has different numbers of subleafs
in its CPUID(4) leaf
- Take care of the GDT restoring on the kexec path too, as expected by
the kernel
- Make sure SMP is not disabled when IO-APIC is disabled on the kernel
cmdline
- Add a PGD flag _PAGE_NOPTISHADOW to instruct machinery not to propagate
changes to the kernelmode page tables, to the user portion, in PTI
- Mark Intel Lunar Lake as affected by an issue where MONITOR wakeups can get
lost and thus user-visible delays happen
- Make sure PKRU is properly restored with XRSTOR on AMD after a PRKU write of
0 (WRPKRU) which will mark PKRU in its init state and thus lose the
actual buffer
-----BEGIN PGP SIGNATURE-----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=02CA
-----END PGP SIGNATURE-----
Merge tag 'x86_urgent_for_v6.13_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 fixes from Borislav Petkov:
- Have the Automatic IBRS setting check on AMD does not falsely fire in
the guest when it has been set already on the host
- Make sure cacheinfo structures memory is allocated to address a boot
NULL ptr dereference on Intel Meteor Lake which has different numbers
of subleafs in its CPUID(4) leaf
- Take care of the GDT restoring on the kexec path too, as expected by
the kernel
- Make sure SMP is not disabled when IO-APIC is disabled on the kernel
cmdline
- Add a PGD flag _PAGE_NOPTISHADOW to instruct machinery not to
propagate changes to the kernelmode page tables, to the user portion,
in PTI
- Mark Intel Lunar Lake as affected by an issue where MONITOR wakeups
can get lost and thus user-visible delays happen
- Make sure PKRU is properly restored with XRSTOR on AMD after a PRKU
write of 0 (WRPKRU) which will mark PKRU in its init state and thus
lose the actual buffer
* tag 'x86_urgent_for_v6.13_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/CPU/AMD: WARN when setting EFER.AUTOIBRS if and only if the WRMSR fails
x86/cacheinfo: Delete global num_cache_leaves
cacheinfo: Allocate memory during CPU hotplug if not done from the primary CPU
x86/kexec: Restore GDT on return from ::preserve_context kexec
x86/cpu/topology: Remove limit of CPUs due to disabled IO/APIC
x86/mm: Add _PAGE_NOPTISHADOW bit to avoid updating userspace page tables
x86/cpu: Add Lunar Lake to list of CPUs with a broken MONITOR implementation
x86/pkeys: Ensure updated PKRU value is XRSTOR'd
x86/pkeys: Change caller of update_pkru_in_sigframe()
The usual bunch of singletons - please see the relevant changelogs for
details.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQTTMBEPP41GrTpTJgfdBJ7gKXxAjgUCZ1U/QwAKCRDdBJ7gKXxA
jnE7AQC0eyNNvaL5pLCIxN/Vmr8YeuWP1dldgI29TjrH/JKjSQEAihZNqVZYjoIT
Gf7Y+IKnc4LbfAXcTe+MfJFeDexM5AU=
=U5LQ
-----END PGP SIGNATURE-----
Merge tag 'mm-hotfixes-stable-2024-12-07-22-39' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull misc fixes from Andrew Morton:
"24 hotfixes. 17 are cc:stable. 15 are MM and 9 are non-MM.
The usual bunch of singletons - please see the relevant changelogs for
details"
* tag 'mm-hotfixes-stable-2024-12-07-22-39' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (24 commits)
iio: magnetometer: yas530: use signed integer type for clamp limits
sched/numa: fix memory leak due to the overwritten vma->numab_state
mm/damon: fix order of arguments in damos_before_apply tracepoint
lib: stackinit: hide never-taken branch from compiler
mm/filemap: don't call folio_test_locked() without a reference in next_uptodate_folio()
scatterlist: fix incorrect func name in kernel-doc
mm: correct typo in MMAP_STATE() macro
mm: respect mmap hint address when aligning for THP
mm: memcg: declare do_memsw_account inline
mm/codetag: swap tags when migrate pages
ocfs2: update seq_file index in ocfs2_dlm_seq_next
stackdepot: fix stack_depot_save_flags() in NMI context
mm: open-code page_folio() in dump_page()
mm: open-code PageTail in folio_flags() and const_folio_flags()
mm: fix vrealloc()'s KASAN poisoning logic
Revert "readahead: properly shorten readahead when falling back to do_page_cache_ra()"
selftests/damon: add _damon_sysfs.py to TEST_FILES
selftest: hugetlb_dio: fix test naming
ocfs2: free inode when ocfs2_get_init_inode() fails
nilfs2: fix potential out-of-bounds memory access in nilfs_find_entry()
...
The kx022a driver supports a few different HW variants. A chip-info
structure is used to describe sensor specific details. Support for
sensors with different measurement g-ranges was added recently,
introducing sensor specific scale arrays.
The members of the chip-info structure have been documented using
kerneldoc. The newly added members omitted the documentation. It is nice
to have all the entries documented for the sake of the consistency.
Furthermore, the scale table format may not be self explatonary, nor how
the amount of scales is informed.
Add documentation to scale table entries to maintain consistency and to
make it more obvious how the scales should be represented.
Suggested-by: Mehdi Djait <mehdi.djait@linux.intel.com>
Signed-off-by: Matti Vaittinen <mazziesaccount@gmail.com>
Reviewed-by: Mehdi Djait <mehdi.djait@linux.intel.com>
Link: https://patch.msgid.link/Z1LDUj-naUdGSM6n@mva-rohm
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
This device returns signed, 16-bit samples as stated in its datasheet
(see 8.5.2 Data Format). That is in line with the scan_type definition
for the IIO_VOLTAGE channel, but 'unsigned int' is being used to read
and push the data to userspace.
Given that the size of that type depends on the architecture (at least
2 bytes to store values up to 65535, but its actual size is often 4
bytes), use the 's16' type to provide the same structure in all cases.
Fixes: a9306887eb ("iio: adc: ti-ads1119: Add driver")
Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Reviewed-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Link: https://patch.msgid.link/20241202-ti-ads1119_s16_chan-v1-1-fafe3136dc90@gmail.com
Cc: <Stable@vger.kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
The 'scan' local struct is used to push data to user space from a
triggered buffer, but it has a hole between the two 16-bit data channels
and the timestamp. This hole is never initialized.
Initialize the struct to zero before using it to avoid pushing
uninitialized information to userspace.
Fixes: 91f75ccf9f ("iio: temperature: tmp006: add triggered buffer support")
Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Link: https://patch.msgid.link/20241204-iio_memset_scan_holes-v2-1-3f941592a76d@gmail.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
In the error path of iio_channel_get_all(), iio_device_put() is called
on all IIO devices, which can cause a refcount imbalance. Fix this error
by calling iio_device_put() only on IIO devices whose refcounts were
previously incremented by iio_device_get().
Fixes: 314be14bb8 ("iio: Rename _st_ functions to loose the bit that meant the staging version.")
Signed-off-by: Joe Hattori <joe@pf.is.s.u-tokyo.ac.jp>
Link: https://patch.msgid.link/20241204111342.1246706-1-joe@pf.is.s.u-tokyo.ac.jp
Cc: <Stable@vger.kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
In some cases, rk_hdptx_phy_runtime_resume() may be invoked before
platform_set_drvdata() is executed in ->probe(), leading to a NULL
pointer dereference when using the return of dev_get_drvdata().
Ensure platform_set_drvdata() is called before devm_pm_runtime_enable().
Reported-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Fixes: 553be2830c ("phy: rockchip: Add Samsung HDMI/eDP Combo PHY driver")
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Link: https://lore.kernel.org/r/20241023-phy-sam-hdptx-rpm-fix-v1-1-87f4c994e346@collabora.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
of_gpio.h is deprecated and subject to remove. The drivers in question
don't use it, simply remove the unused header.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Andre Przywara <andre.przywara@arm.com>
Link: https://lore.kernel.org/r/20241031104631.2454581-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Now exynos-pmu can register its custom regmap for gs101 via
of_syscon_register_regmap() we can switch back to the standard
syscon_regmap_lookup_by_phandle() api for obtaining the regmap.
Additionally add a Kconfig dependency for MFD_SYSCON.
Signed-off-by: Peter Griffin <peter.griffin@linaro.org>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Link: https://lore.kernel.org/r/20241029192107.2344279-1-peter.griffin@linaro.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Define several registers to be used by PCIe QMP PHYs on v6 platforms.
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20241021-sar2130p-phys-v2-5-d883acf170f7@linaro.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The PCS_LANE1 region isn't a part of the PCS_PCIE region. It was handled
this way as it simplified handled of devices with the old bindings.
Nowadays it can be handled as is, without hacks.
Split the PCS_LANE1 region from the PCS_PCIE / PCS_MISC region space.
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20241021-sar2130p-phys-v2-4-d883acf170f7@linaro.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Extend the USB+DP combo QMP PHY driver to support the SAR2130P platform.
It mosly follows the SM8550 QMP PHY, but the QSERDES programming
differs, most likely because of the parent clock rate differences.
NOTE: The DP part wasn't yet tested, but it is not possible to support
just the USB part of the PHY. DP part might require additional fixes
later.
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20241021-sar2130p-phys-v2-3-d883acf170f7@linaro.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add clock IDs for the slow clock controller. Previously, raw numbers
were used (0 or 1) for clocks generated by the slow clock controller. This
leads to confusion and wrong IDs were used on few device trees. To avoid
this add macros.
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20240826173116.3628337-2-claudiu.beznea@tuxon.dev
Signed-off-by: Claudiu Beznea <claudiu.beznea@tuxon.dev>
Currently, the calcuation for fld_tg_code is based on a lookup table,
but there are gaps in the lookup table, and frequencies in these
gaps may not properly use the correct divider. Based on the description
of FLD_CK_DIV, the internal PLL frequency should be less than 50 MHz,
so directly calcuate the value of FLD_CK_DIV from pixclk.
This allow for proper calcuation of any pixel clock and eliminates a
few gaps in the LUT.
Since the value of the int_pllclk is in Hz, do the fixed-point
math in Hz to achieve a more accurate value and reduces the complexity
of the caluation to 24MHz * (256 / int_pllclk).
Fixes: 6ad082bee9 ("phy: freescale: add Samsung HDMI PHY")
Signed-off-by: Adam Ford <aford173@gmail.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Link: https://lore.kernel.org/r/20241026132014.73050-3-aford173@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
There are a series of for-loops which check various values of P and S
for the integer divder PLL. The for loops search all entries and use
the one closest to the nominal, but it continues to searches through
all for loops even after the nominal is achieved. Ending when the
nominal value is found stops wasting time, since it will not find
a better value than a deviation of 0 Hz.
Signed-off-by: Adam Ford <aford173@gmail.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Link: https://lore.kernel.org/r/20241026132014.73050-2-aford173@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The Integer divder uses values of P,M, and S to determine the PLL
rate. Currently, the range of M was set based on a series of
table entries where the range was limited. Since the ref manual
shows it is 8-bit wide, expand the range to be up to 255.
Signed-off-by: Adam Ford <aford173@gmail.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Link: https://lore.kernel.org/r/20241026132014.73050-1-aford173@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Rockchip RK3576 integrates two naneng-combo PHY, PHY0 is used for
PCIE and SATA, PHY1 is used for PCIE, SATA and USB3.
This adds device specific data support.
Signed-off-by: Kever Yang <kever.yang@rock-chips.com>
Signed-off-by: William Wu <william.wu@rock-chips.com>
Signed-off-by: Frank Wang <frank.wang@rock-chips.com>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Test-by: Kever Yang <kever.yang@rock-chips.com>
Link: https://lore.kernel.org/r/20241106021357.19782-2-frawang.cn@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Adds the compatible line to support RK3576 SoC.
Signed-off-by: Frank Wang <frank.wang@rock-chips.com>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://lore.kernel.org/r/20241106021357.19782-1-frawang.cn@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Change how the Backup Unit Power is configured and force the
automatic/hardware mode.
This change eliminates the need for software management of the power
switch, ensuring it transitions to the backup power source before
entering low power modes.
This is done in the only location where this switch was configured. It's
usually done in the bootloader.
Previously, the loss of the VDDANA (or VDDIN33) power source was not
automatically compensated by an alternative power source. This resulted
in the loss of Backup Unit content, including Backup Self-refresh low
power mode information, OTP emulation configuration, and boot
configuration, for instance.
Fixes: ac809e7879 ("ARM: at91: pm: switch backup area to vbat in backup mode")
Signed-off-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Link: https://lore.kernel.org/r/20241125165648.509162-1-nicolas.ferre@microchip.com
Signed-off-by: Claudiu Beznea <claudiu.beznea@tuxon.dev>
A device_node acquired via of_find_node_by_path() requires explicit
calls to of_node_put() when it is no longer needed to avoid leaking the
resource.
Instead of adding the missing calls to of_node_put() in all execution
paths, use the cleanup attribute for 'np' by means of the __free()
macro, which automatically calls of_node_put() when the variable goes
out of scope.
Fixes: 960ddf70cc ("drivers: soc: atmel: Avoid calling at91_soc_init on non AT91 SoCs")
Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Link: https://lore.kernel.org/r/20241031-soc-atmel-soc-cleanup-v2-1-73f2d235fd98@gmail.com
Signed-off-by: Claudiu Beznea <claudiu.beznea@tuxon.dev>
These drivers don't use the driver_data member of struct i2c_device_id,
so don't explicitly initialize this member.
This prepares putting driver_data in an anonymous union which requires
either no initialization or named designators. But it's also a nice
cleanup on its own.
Link: https://lore.kernel.org/lkml/20240918123150.1540161-6-u.kleine-koenig@baylibre.com/
Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
The Extcon (External Connector) subsystem driver lacked proper
documentation. This commit adds comprehensive documentation
explaining the purpose, key components, and usage of the Extcon
framework.
The new documentation includes:
- An overview of the Extcon subsystem
- Descriptions of key structures
- Explanations of core functions
- Information on the sysfs interface
- A usage example for driver developers
Link: https://lore.kernel.org/lkml/20241103025436.69196-1-yesanishhere@gmail.com/
Acked-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: anish kumar <yesanishhere@gmail.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
In extcon_rtk_type_c_probe() devm_kzalloc() may return NULL but this
returned value is not checked.
Fixes: 8a590d7371 ("extcon: add Realtek DHC RTD SoC Type-C driver")
Link: https://lore.kernel.org/lkml/20241025094650.253599-1-hanchunchao@inspur.com/
Signed-off-by: Charles Han <hanchunchao@inspur.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
An of_node_put(node) call was immediately used after a null pointer check
for the local variable “edev” at the end of this function implementation.
Thus call such a function only once instead directly before the check.
This issue was transformed by using the Coccinelle software.
Link: https://lore.kernel.org/lkml/0f103384-376c-41f0-a35c-8ad98327d6cb@web.de/
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Since commit 13b25489b6 ("kbuild: change working directory to external
module directory with M="), the Debian package build fails if a relative
path is specified with the O= option.
$ make O=build bindeb-pkg
[ snip ]
dpkg-deb: building package 'linux-image-6.13.0-rc1' in '../linux-image-6.13.0-rc1_6.13.0-rc1-6_amd64.deb'.
Rebuilding host programs with x86_64-linux-gnu-gcc...
make[6]: Entering directory '/home/masahiro/linux/build'
/home/masahiro/linux/Makefile:190: *** specified kernel directory "build" does not exist. Stop.
This occurs because the sub_make_done flag is cleared, even though the
working directory is already in the output directory.
Passing KBUILD_OUTPUT=. resolves the issue.
Fixes: 13b25489b6 ("kbuild: change working directory to external module directory with M=")
Reported-by: Charlie Jenkins <charlie@rivosinc.com>
Closes: https://lore.kernel.org/all/Z1DnP-GJcfseyrM3@ghost/
Tested-by: Charlie Jenkins <charlie@rivosinc.com>
Reviewed-by: Charlie Jenkins <charlie@rivosinc.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The compiler can fully inline the actual handler function of an interrupt
entry into the .irqentry.text entry point. If such a function contains an
access which has an exception table entry, modpost complains about a
section mismatch:
WARNING: vmlinux.o(__ex_table+0x447c): Section mismatch in reference ...
The relocation at __ex_table+0x447c references section ".irqentry.text"
which is not in the list of authorized sections.
Add .irqentry.text to OTHER_SECTIONS to cure the issue.
Reported-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org # needed for linux-5.4-y
Link: https://lore.kernel.org/all/20241128111844.GE10431@google.com/
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Refine the if statement to improve readability.
Signed-off-by: Justin Lai <justinlai0215@realtek.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Link: https://patch.msgid.link/20241206084851.760475-1-justinlai0215@realtek.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Vladimir Oltean says:
====================
Ocelot PTP fixes
This is another attempt at "net: mscc: ocelot: be resilient to loss of
PTP packets during transmission".
https://lore.kernel.org/netdev/20241203164755.16115-1-vladimir.oltean@nxp.com/
The central change is in patch 4/5. It recovers a port from a very
reproducible condition where previously, it would become unable to take
any hardware TX timestamp.
Then we have patches 1/5 and 5/5 which are extra bug fixes.
Patches 2/5 and 3/5 are logical changes split out of the monolithic
patch previously submitted as v1, so that patch 4/5 is hopefully easier
to follow.
====================
Link: https://patch.msgid.link/20241205145519.1236778-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
An unsupported RX filter will leave the port with TX timestamping still
applied as per the new request, rather than the old setting. When
parsing the tx_type, don't apply it just yet, but delay that until after
we've parsed the rx_filter as well (and potentially returned -ERANGE for
that).
Similarly, copy_to_user() may fail, which is a rare occurrence, but
should still be treated by unwinding what was done.
Fixes: 96ca08c058 ("net: mscc: ocelot: set up traps for PTP packets")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20241205145519.1236778-6-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The Felix DSA driver presents unique challenges that make the simplistic
ocelot PTP TX timestamping procedure unreliable: any transmitted packet
may be lost in hardware before it ever leaves our local system.
This may happen because there is congestion on the DSA conduit, the
switch CPU port or even user port (Qdiscs like taprio may delay packets
indefinitely by design).
The technical problem is that the kernel, i.e. ocelot_port_add_txtstamp_skb(),
runs out of timestamp IDs eventually, because it never detects that
packets are lost, and keeps the IDs of the lost packets on hold
indefinitely. The manifestation of the issue once the entire timestamp
ID range becomes busy looks like this in dmesg:
mscc_felix 0000:00:00.5: port 0 delivering skb without TX timestamp
mscc_felix 0000:00:00.5: port 1 delivering skb without TX timestamp
At the surface level, we need a timeout timer so that the kernel knows a
timestamp ID is available again. But there is a deeper problem with the
implementation, which is the monotonically increasing ocelot_port->ts_id.
In the presence of packet loss, it will be impossible to detect that and
reuse one of the holes created in the range of free timestamp IDs.
What we actually need is a bitmap of 63 timestamp IDs tracking which one
is available. That is able to use up holes caused by packet loss, but
also gives us a unique opportunity to not implement an actual timer_list
for the timeout timer (very complicated in terms of locking).
We could only declare a timestamp ID stale on demand (lazily), aka when
there's no other timestamp ID available. There are pros and cons to this
approach: the implementation is much more simple than per-packet timers
would be, but most of the stale packets would be quasi-leaked - not
really leaked, but blocked in driver memory, since this algorithm sees
no reason to free them.
An improved technique would be to check for stale timestamp IDs every
time we allocate a new one. Assuming a constant flux of PTP packets,
this avoids stale packets being blocked in memory, but of course,
packets lost at the end of the flux are still blocked until the flux
resumes (nobody left to kick them out).
Since implementing per-packet timers is way too complicated, this should
be good enough.
Testing procedure:
Persistently block traffic class 5 and try to run PTP on it:
$ tc qdisc replace dev swp3 parent root taprio num_tc 8 \
map 0 1 2 3 4 5 6 7 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
base-time 0 sched-entry S 0xdf 100000 flags 0x2
[ 126.948141] mscc_felix 0000:00:00.5: port 3 tc 5 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
$ ptp4l -i swp3 -2 -P -m --socket_priority 5 --fault_reset_interval ASAP --logSyncInterval -3
ptp4l[70.351]: port 1 (swp3): INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[70.354]: port 0 (/var/run/ptp4l): INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[70.358]: port 0 (/var/run/ptp4lro): INITIALIZING to LISTENING on INIT_COMPLETE
[ 70.394583] mscc_felix 0000:00:00.5: port 3 timestamp id 0
ptp4l[70.406]: timed out while polling for tx timestamp
ptp4l[70.406]: increasing tx_timestamp_timeout or increasing kworker priority may correct this issue, but a driver bug likely causes it
ptp4l[70.406]: port 1 (swp3): send peer delay response failed
ptp4l[70.407]: port 1 (swp3): clearing fault immediately
ptp4l[70.952]: port 1 (swp3): new foreign master d858d7.fffe.00ca6d-1
[ 71.394858] mscc_felix 0000:00:00.5: port 3 timestamp id 1
ptp4l[71.400]: timed out while polling for tx timestamp
ptp4l[71.400]: increasing tx_timestamp_timeout or increasing kworker priority may correct this issue, but a driver bug likely causes it
ptp4l[71.401]: port 1 (swp3): send peer delay response failed
ptp4l[71.401]: port 1 (swp3): clearing fault immediately
[ 72.393616] mscc_felix 0000:00:00.5: port 3 timestamp id 2
ptp4l[72.401]: timed out while polling for tx timestamp
ptp4l[72.402]: increasing tx_timestamp_timeout or increasing kworker priority may correct this issue, but a driver bug likely causes it
ptp4l[72.402]: port 1 (swp3): send peer delay response failed
ptp4l[72.402]: port 1 (swp3): clearing fault immediately
ptp4l[72.952]: port 1 (swp3): new foreign master d858d7.fffe.00ca6d-1
[ 73.395291] mscc_felix 0000:00:00.5: port 3 timestamp id 3
ptp4l[73.400]: timed out while polling for tx timestamp
ptp4l[73.400]: increasing tx_timestamp_timeout or increasing kworker priority may correct this issue, but a driver bug likely causes it
ptp4l[73.400]: port 1 (swp3): send peer delay response failed
ptp4l[73.400]: port 1 (swp3): clearing fault immediately
[ 74.394282] mscc_felix 0000:00:00.5: port 3 timestamp id 4
ptp4l[74.400]: timed out while polling for tx timestamp
ptp4l[74.401]: increasing tx_timestamp_timeout or increasing kworker priority may correct this issue, but a driver bug likely causes it
ptp4l[74.401]: port 1 (swp3): send peer delay response failed
ptp4l[74.401]: port 1 (swp3): clearing fault immediately
ptp4l[74.953]: port 1 (swp3): new foreign master d858d7.fffe.00ca6d-1
[ 75.396830] mscc_felix 0000:00:00.5: port 3 invalidating stale timestamp ID 0 which seems lost
[ 75.405760] mscc_felix 0000:00:00.5: port 3 timestamp id 0
ptp4l[75.410]: timed out while polling for tx timestamp
ptp4l[75.411]: increasing tx_timestamp_timeout or increasing kworker priority may correct this issue, but a driver bug likely causes it
ptp4l[75.411]: port 1 (swp3): send peer delay response failed
ptp4l[75.411]: port 1 (swp3): clearing fault immediately
(...)
Remove the blocking condition and see that the port recovers:
$ same tc command as above, but use "sched-entry S 0xff" instead
$ same ptp4l command as above
ptp4l[99.489]: port 1 (swp3): INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[99.490]: port 0 (/var/run/ptp4l): INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[99.492]: port 0 (/var/run/ptp4lro): INITIALIZING to LISTENING on INIT_COMPLETE
[ 100.403768] mscc_felix 0000:00:00.5: port 3 invalidating stale timestamp ID 0 which seems lost
[ 100.412545] mscc_felix 0000:00:00.5: port 3 invalidating stale timestamp ID 1 which seems lost
[ 100.421283] mscc_felix 0000:00:00.5: port 3 invalidating stale timestamp ID 2 which seems lost
[ 100.430015] mscc_felix 0000:00:00.5: port 3 invalidating stale timestamp ID 3 which seems lost
[ 100.438744] mscc_felix 0000:00:00.5: port 3 invalidating stale timestamp ID 4 which seems lost
[ 100.447470] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 100.505919] mscc_felix 0000:00:00.5: port 3 timestamp id 0
ptp4l[100.963]: port 1 (swp3): new foreign master d858d7.fffe.00ca6d-1
[ 101.405077] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 101.507953] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 102.405405] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 102.509391] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 103.406003] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 103.510011] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 104.405601] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 104.510624] mscc_felix 0000:00:00.5: port 3 timestamp id 0
ptp4l[104.965]: selected best master clock d858d7.fffe.00ca6d
ptp4l[104.966]: port 1 (swp3): assuming the grand master role
ptp4l[104.967]: port 1 (swp3): LISTENING to GRAND_MASTER on RS_GRAND_MASTER
[ 105.106201] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.232420] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.359001] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.405500] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.485356] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.511220] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.610938] mscc_felix 0000:00:00.5: port 3 timestamp id 0
[ 105.737237] mscc_felix 0000:00:00.5: port 3 timestamp id 0
(...)
Notice that in this new usage pattern, a non-congested port should
basically use timestamp ID 0 all the time, progressing to higher numbers
only if there are unacknowledged timestamps in flight. Compare this to
the old usage, where the timestamp ID used to monotonically increase
modulo OCELOT_MAX_PTP_ID.
In terms of implementation, this simplifies the bookkeeping of the
ocelot_port :: ts_id and ptp_skbs_in_flight. Since we need to traverse
the list of two-step timestampable skbs for each new packet anyway, the
information can already be computed and does not need to be stored.
Also, ocelot_port->tx_skbs is always accessed under the switch-wide
ocelot->ts_id_lock IRQ-unsafe spinlock, so we don't need the skb queue's
lock and can use the unlocked primitives safely.
This problem was actually detected using the tc-taprio offload, and is
causing trouble in TSN scenarios, which Felix (NXP LS1028A / VSC9959)
supports but Ocelot (VSC7514) does not. Thus, I've selected the commit
to blame as the one adding initial timestamping support for the Felix
switch.
Fixes: c0bcf53766 ("net: dsa: ocelot: add hardware timestamping support for Felix")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20241205145519.1236778-5-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ocelot_get_txtstamp() is a threaded IRQ handler, requested explicitly as
such by both ocelot_ptp_rdy_irq_handler() and vsc9959_irq_handler().
As such, it runs with IRQs enabled, and not in hardirq context. Thus,
ocelot_port_add_txtstamp_skb() has no reason to turn off IRQs, it cannot
be preempted by ocelot_get_txtstamp(). For the same reason,
dev_kfree_skb_any_reason() will always evaluate as kfree_skb_reason() in
this calling context, so just simplify the dev_kfree_skb_any() call to
kfree_skb().
Also, ocelot_port_txtstamp_request() runs from NET_TX softirq context,
not with hardirqs enabled. Thus, ocelot_get_txtstamp() which shares the
ocelot_port->tx_skbs.lock lock with it, has no reason to disable hardirqs.
This is part of a larger rework of the TX timestamping procedure.
A logical subportion of the rework has been split into a separate
change.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20241205145519.1236778-4-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This condition, theoretically impossible to trigger, is not really
handled well. By "continuing", we are skipping the write to SYS_PTP_NXT
which advances the timestamp FIFO to the next entry. So we are reading
the same FIFO entry all over again, printing stack traces and eventually
killing the kernel.
No real problem has been observed here. This is part of a larger rework
of the timestamp IRQ procedure, with this logical change split out into
a patch of its own. We will need to "goto next_ts" for other conditions
as well.
Fixes: 9fde506e0c ("net: mscc: ocelot: warn when a PTP IRQ is raised for an unknown skb")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20241205145519.1236778-3-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If ocelot_port_add_txtstamp_skb() fails, for example due to a full PTP
timestamp FIFO, we must undo the skb_clone_sk() call with kfree_skb().
Otherwise, the reference to the skb clone is lost.
Fixes: 52849bcf00 ("net: mscc: ocelot: avoid overflowing the PTP timestamp FIFO")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20241205145519.1236778-2-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Russell King says:
====================
net: net: add negotiation of in-band capabilities (remainder)
Here are the last three patches which were not included in the non-RFC
posting, but were in the RFC posting. These add the .pcs_inband()
method to the Lynx, MTK Lynx and XPCS drivers.
====================
Link: https://patch.msgid.link/Z1F1b8eh8s8T627j@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Report the PCS in-band capabilities to phylink for the Lynx PCS.
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tJ8NM-006L5J-AH@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently tun checks the group permission even if the user have matched.
Besides going against the usual permission semantic, this has a
very interesting implication: if the tun group is not among the
supplementary groups of the tun user, then effectively no one can
access the tun device. CAP_SYS_ADMIN still can, but its the same as
not setting the tun ownership.
This patch relaxes the group checking so that either the user match
or the group match is enough. This avoids the situation when no one
can access the device even though the ownership is properly set.
Also I simplified the logic by removing the redundant inversions:
tun_not_capable() --> !tun_capable()
Signed-off-by: Stas Sergeev <stsp2@yandex.ru>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://patch.msgid.link/20241205073614.294773-1-stsp2@yandex.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 66600fac7a ("net: stmmac: TSO: Fix unbalanced DMA map/unmap
for non-paged SKB data") moved the assignment of tx_skbuff_dma[]'s
members to be later in stmmac_tso_xmit().
The buf (dma cookie) and len stored in this structure are passed to
dma_unmap_single() by stmmac_tx_clean(). The DMA API requires that
the dma cookie passed to dma_unmap_single() is the same as the value
returned from dma_map_single(). However, by moving the assignment
later, this is not the case when priv->dma_cap.addr64 > 32 as "des"
is offset by proto_hdr_len.
This causes problems such as:
dwc-eth-dwmac 2490000.ethernet eth0: Tx DMA map failed
and with DMA_API_DEBUG enabled:
DMA-API: dwc-eth-dwmac 2490000.ethernet: device driver tries to +free DMA memory it has not allocated [device address=0x000000ffffcf65c0] [size=66 bytes]
Fix this by maintaining "des" as the original DMA cookie, and use
tso_des to pass the offset DMA cookie to stmmac_tso_allocator().
Full details of the crashes can be found at:
https://lore.kernel.org/all/d8112193-0386-4e14-b516-37c2d838171a@nvidia.com/https://lore.kernel.org/all/klkzp5yn5kq5efgtrow6wbvnc46bcqfxs65nz3qy77ujr5turc@bwwhelz2l4dw/
Reported-by: Jon Hunter <jonathanh@nvidia.com>
Reported-by: Thierry Reding <thierry.reding@gmail.com>
Fixes: 66600fac7a ("net: stmmac: TSO: Fix unbalanced DMA map/unmap for non-paged SKB data")
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Furong Xu <0x1207@gmail.com>
Link: https://patch.msgid.link/E1tJXcx-006N4Z-PC@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Without -o the tool currently crashes, but it's not marked
as required. The only thing we can't do without it is to
generate the correct #include for user source files, but
we can put a placeholder instead.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20241206113100.89d35bf124d6.I9228fb704e6d5c9d8e046ef15025a47a48439c1e@changeid
Signed-off-by: Jakub Kicinski <kuba@kernel.org>