Monitoring tools periodically scan /proc/interrupts to export metrics as a
timeseries for future analysis and investigation.
In large fleets, /proc/interrupts is polled (often every few seconds) on
every machine. The cumulative overhead adds up quickly across thousands
of nodes, so reducing the cost of generating these stats does have a
measurable operational impact. With the ongoing trend toward higher core
counts per machine, this cost becomes even more noticeable over time,
since interrupt counters are per-CPU. In Meta's fleet, we have observed
this overhead at scale.
Although a binary /proc interface would be a better long-term solution
due to lower formatting (kernel side) and parsing (userspace side)
overhead, the text interface will remain in use for some time, even if
better solutions will be available. Optimizing the /proc/interrupts
printing code is therefore still beneficial.
Function seq_printf() supports rich format string for decimals printing,
but it doesn't required for printing /proc/interrupts per CPU counters,
seq_put_decimal_ull_width() function can be used instead to print per
CPU counters, because very limited formatting is required for this case.
Similar optimization idea is already used in show_interrupts().
As a side effect this aligns the x86 decriptions with the generic
interrupts event descriptions.
Performance counter stats (truncated) for 'sh -c cat /proc/interrupts
Before:
3.42 msec task-clock # 0.802 CPUs utilized ( +- 0.05% )
1 context-switches # 291.991 /sec ( +- 0.74% )
0 cpu-migrations # 0.000 /sec
343 page-faults # 100.153 K/sec ( +- 0.01% )
8,932,242 instructions # 1.66 insn per cycle ( +- 0.34% )
5,374,427 cycles # 1.569 GHz ( +- 0.04% )
1,483,154 branches # 433.068 M/sec ( +- 0.22% )
28,768 branch-misses # 1.94% of all branches ( +- 0.31% )
0.00427182 +- 0.00000215 seconds time elapsed ( +- 0.05% )
After:
2.39 msec task-clock # 0.796 CPUs utilized ( +- 0.06% )
1 context-switches # 418.541 /sec ( +- 0.70% )
0 cpu-migrations # 0.000 /sec
343 page-faults # 143.560 K/sec ( +- 0.01% )
7,020,982 instructions # 1.30 insn per cycle ( +- 0.52% )
5,397,266 cycles # 2.259 GHz ( +- 0.06% )
1,569,648 branches # 656.962 M/sec ( +- 0.08% )
25,419 branch-misses # 1.62% of all branches ( +- 0.72% )
0.00299996 +- 0.00000206 seconds time elapsed ( +- 0.07% )
Relative speed up in time elapsed is around 29%.
[ tglx: Fixed it up so it applies to current mainline ]
Signed-off-by: Dmitry Ilvokhin <d@ilvokhin.com>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Michael Kelley <mhklinux@outlook.com>
Reviewed-by: Thomas Gleixner <tglx@kernel.org>
Reviewed-by: Radu Rendec <radu@rendec.net>
Link: https://patch.msgid.link/aQj5mGZ6_BBlAm3B@shell.ilvokhin.com
Link: https://patch.msgid.link/20260517194930.949709489@kernel.org
When build-testing this driver without CONFIG_IRQ_DOMAIN causes
a compile-time error:
drivers/regulator/mt6363-regulator.c: In function 'mt6363_regulator_probe':
drivers/regulator/mt6363-regulator.c:884:18: error: implicit declaration of function 'irq_find_host' [-Wimplicit-function-declaration]
884 | domain = irq_find_host(interrupt_parent);
| ^~~~~~~~~~~~~
drivers/regulator/mt6363-regulator.c:884:16: error: assignment to 'struct irq_domain *' from 'int' makes pointer from integer without a cast [-Wint-conversion]
884 | domain = irq_find_host(interrupt_parent);
| ^
drivers/regulator/mt6363-regulator.c:896:30: error: implicit declaration of function 'irq_create_fwspec_mapping'; did you mean 'irq_create_of_mapping'? [-Wimplicit-function-declaration]
896 | info->virq = irq_create_fwspec_mapping(&fwspec);
| ^~~~~~~~~~~~~~~~~~~~~~~~~
| irq_create_of_mapping
This is rather hard to trigger because so many other drivers
enable IRQ_DOMAIN already, but I ran into this on an s390
randconfig build.
Ensure this is always enabled using a Kconfig 'select IRQ_DOMAIN'
entry, as we do for all other users of this.
Fixes: 3c36965df8 ("regulator: Add support for MediaTek MT6363 SPMI PMIC Regulators")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://patch.msgid.link/20260526102003.2527570-1-arnd@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Use the correct kernel-doc notation for struct members to eliminate
kernel-doc warnings:
Warning: drivers/gpu/drm/drm_panel.c:119 function parameter 'dev'
not described in 'devm_drm_panel_add'
Warning: drivers/gpu/drm/drm_panel.c:119 function parameter 'dev'
not described in 'devm_drm_panel_add'
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202605210648.RI4ufD66-lkp@intel.com/
Signed-off-by: Dongyang Jin <jindongyang@kylinos.cn>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patch.msgid.link/20260521094545.3039207-1-jindongyang@kylinos.cn
Add dt-bindings for 10.1" TFT LCD module from Ampire Co. Ltd.
as part of panel-simple.
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patch.msgid.link/20260515082232.1766586-1-dario.binacchi@amarulasolutions.com
The .driver_data member of the struct pci_device_id array were
initialized by a list expressions to zero without making use of that
value. In this case it's better to not specify a value at all and let
the compiler fill in the zeros. Same for the list terminator that can
better be completely empty.
This patch doesn't introduce changes to the compiled array.
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Acked-by: Alexandra Winter <wintera@linux.ibm.com>
Link: https://patch.msgid.link/20260522153010.777081-2-u.kleine-koenig@baylibre.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The CRC uncompressed buffer pointer array has the same lifetime as
struct crc_data, but it is currently allocated separately. That adds
another allocation failure path and a matching cleanup branch without
providing any extra flexibility.
Store the pointer array as a flexible array member and allocate it
together with the crc_data using kzalloc_flex(). The array remains
zero-initialized, while the allocation and error handling become
simpler.
Assisted-by: Codex:GPT-5.5
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20260510213948.41750-1-rosenp@gmail.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add regression tests for the zero-length entry and root directory
bounds fixes:
- tb_test_property_parse_zero_length: TEXT entry with length 0
must be rejected by the validator.
- tb_test_property_parse_rootdir_overflow: root directory whose
content_offset + content_len exceeds block_len must be rejected.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
pskb_trim_rcsum_slow() keeps CHECKSUM_COMPLETE valid by subtracting
the checksum of the bytes removed from the skb tail. That assumes the
removed bytes can be read.
io_uring zcrx skbs may contain unreadable net_iov frags. With fbnic
header/data split, small TCP/IPv4 packets can carry Ethernet padding
in such a frag. ip_rcv_core() trims the skb to iph->tot_len before TCP
sees it, and the CHECKSUM_COMPLETE adjustment then calls
skb_checksum() on the padding.
This is exposed by IPv4 because small TCP/IPv4 frames can be shorter
than the Ethernet minimum payload. TCP/IPv6 frames are large enough in
the normal zcrx path, so they do not hit the same padding trim.
Keep the existing checksum adjustment for readable skbs. If the
remaining packet is fully linear, drop CHECKSUM_COMPLETE and let the
stack validate the packet after trimming. If unreadable payload would
remain, fail the trim; the checksum cannot be adjusted without reading
the trimmed tail.
Also clear skb->unreadable when trimming removes all frags.
Fixes: 65249feb6b ("net: add support for skbs with unreadable frags")
Signed-off-by: Björn Töpel <bjorn@kernel.org>
Reviewed-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260522120643.242974-1-bjorn@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The i915 and xe calls to display, in particular for
probe/cleanup/suspend/resume, need to be unified. It does not help to
have the related calls scattered around. As a small step forward,
relocate the intel_hpd_cancel_work() call from intel_irq_uninstall() to
i915_driver_remove().
Note that the other intel_irq_uninstall() call sites don't need the
call, as they're on error paths where hotplug hasn't been enabled yet.
Reviewed-by: Michał Grzelak <michal.grzelak@intel.com>
Link: https://patch.msgid.link/20260516101852.1373108-1-jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
tb_xdomain_copy() copies req->response_size bytes from the received
packet buffer regardless of the actual frame size. When a short
response arrives, this reads past the valid frame data in the DMA
pool buffer into stale contents from previous transactions.
Use the minimum of frame size and expected response size for the
copy length.
Fixes: cdae7c07e3 ("thunderbolt: Add support for XDomain properties")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
tb_xdp_handle_request() casts the received packet buffer to
protocol-specific structs without verifying that the allocation
is large enough for the target type. A peer can send a minimal
XDomain packet that passes the generic header length check but is
shorter than the struct accessed after the cast, causing out-of-
bounds reads from the kmemdup allocation.
Plumb the packet length through xdomain_request_work and validate
it against the expected struct size before each cast.
Fixes: 8e1de70425 ("thunderbolt: Add support for XDomain lane bonding")
Fixes: cdae7c07e3 ("thunderbolt: Add support for XDomain properties")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
tb_xdp_properties_request() derives the per-packet copy length from
the response header without checking that it fits in the previously
allocated data buffer. A malicious peer can set its length field
larger than the declared data_length, causing memcpy to write past
the kcalloc allocation.
Clamp the per-packet copy length so that the cumulative offset
never exceeds data_len.
Fixes: cdae7c07e3 ("thunderbolt: Add support for XDomain properties")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
__tb_property_parse_dir() does not check that content_offset +
content_len fits within block_len for the root directory case.
When rootdir->length equals or exceeds block_len - 2, the entry
loop reads past the allocated property block.
Add a bounds check after computing content_offset and content_len
to reject directories whose content extends past the block.
Fixes: cdae7c07e3 ("thunderbolt: Add support for XDomain properties")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
tb_property_entry_valid() accepts entries with length == 0 for
DIRECTORY, DATA, and TEXT types. A zero-length TEXT entry passes
validation but causes an underflow in the null-termination logic:
property->value.text[property->length * 4 - 1] = '\0';
When property->length is 0 this writes to offset -1 relative to
the allocation.
Reject zero-length entries early in the validator since they have no
valid representation in the XDomain property protocol.
Fixes: cdae7c07e3 ("thunderbolt: Add support for XDomain properties")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
To maintain the split where ib_uverbs.ko should not be depended on by
drivers, add a new module ib_uverbs_support.ko which contains the driver
called functions that are too large or too rare to be placed in
ib_uverbs_core.ko
Start by moving most of rdma_core.c into this module, making some
adjustments to split it from the actual uverbs FD code.
This was not done originally because we lacked EXPORT_SYMBOL_NS and I had
a fear that drivers would abuse this interface surface.
Link: https://patch.msgid.link/r/4-v3-43aba1969751+1988-ib_uverbs_support_ko_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Instead of having an alternative fops release always use the standard
uverbs_uobject_fd_release() and route the special async behavior back up
through uverbs_obj_fd_type ops pointer.
This removes a dependency where the technically lower level rdma_core.c is
referring to a symbol from uverbs_std_types_async_fd.c.
Link: https://patch.msgid.link/r/3-v3-43aba1969751+1988-ib_uverbs_support_ko_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Remove the entire ib_core_uverbs.c from the build if
CONFIG_INFINIBAND_USER_ACCESS is not set. These functions are only used to
support uverbs and are never callable even if they happen to get linked
in.
Provide inlines for the missing ones to return errors to further push code
elimination in drivers.
Link: https://patch.msgid.link/r/1-v3-43aba1969751+1988-ib_uverbs_support_ko_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
When topology_physical_package_id()/topology_logical_die_id() returns
a negative value, rapl_add_package_cpuslocked() returns ERR_PTR(-EINVAL)
directly without freeing the rapl_package structure that was just
allocated by kzalloc_obj(), leaking memory on every failed package
addition.
Use the existing err_free_package label so that the allocation is
released on the error path.
Signed-off-by: Sumeet Pawnikar <sumeet4linux@gmail.com>
Link: https://patch.msgid.link/20260515182616.227707-1-sumeet4linux@gmail.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add functions to be able to do memory read under NIC access,
in order to use them later during firmware dump. These may
drop and re-acquire the spinlock, but will not acquire and
release the NIC access.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260517100550.7bb1ea51c347.I91420a24fb0c481c75a2600d60e1365c15c1c5a9@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
There are two dump paths: new "ini" style, and the old,
more driver-centric, way. Separate out the code for the
old way into a separate dbg-old.c file, to simplify.
Fix a typo ('trasport') along the way.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260517100550.1de0f1110d5a.I4faebe15192a1f27cb4b7270fb52154f06eb2a10@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
The iwl_{read,write}_prph_no_grab() functions will trace each
access, but in debug dump a lot of accesses already use the
transport versions of these functions directly. Since the data
(register addresses and their content) is going into the dump
file, tracing isn't really needed. Use the transport functions
in all places.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260517100550.bbc7623a0348.I827729916da8f264befbcb90ac6509c359ee97a3@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
This is all fairly manual, with the shifts etc., and
the documentation has to call out the specific bits
(which also isn't usable for our tracing data). Add
an enum that directly declares the bit masks for it
and then use the more modern u8_encode_bits().
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260517100550.bc09f4f6fd29.I315f15856eb36f0490b8f5008537d3e0ffc215f2@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
The iwl_txq, iwl_pcie_first_tb_buf and iwl_pcie_txq_entry don't need to
be exported to the op_mode in iwl-trans.h. Declare those in the
transport's internal header file to avoid pollution.
iwl_trans_pcie_send_hcmd can also be moved to the internal header file.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Link: https://patch.msgid.link/20260517100550.208921548b4b.I76b1ac8499275e6d231880861e3843278f278c34@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
When EMLSR exits due to beacon loss on the current link, the driver
should keep the link that is still receiving beacons. The previous
code always called get_primary_link(), keeping the primary link
regardless of which link is actually losing beacons. If the primary
link is the one losing beacons, the driver exits EMLSR onto the
degraded link and the connection is lost eventually.
When both links lose beacons, keep the primary link. When only the
current link loses beacons -- whether due to signal loss or a BSS
parameter change -- keep the other link.
Signed-off-by: Avinash Bhatt <avinash.bhatt@intel.com>
Link: https://patch.msgid.link/20260517100550.b2d700f7775e.I8e9189ce6cf4388878beab14e56341becd5f427c@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
When FW PHY statistics report high channel load (MCLM), the driver
triggers an internal MLO scan to find a better link. However, link grade
computation uses the QBSS Load IE from AP beacons, not MCLM data. If
the AP does not broadcast a QBSS Load IE, the scan produces no benefit
as the grade falls back to static band defaults regardless, and the same
bad link will be selected anyway as the active link.
Skip the MLO scan trigger when the AP does not advertise a QBSS Load IE.
Signed-off-by: Avinash Bhatt <avinash.bhatt@intel.com>
Link: https://patch.msgid.link/20260517100550.621538e20244.I7fdccb759508f32991cc06774cc7621725a58bd3@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
This fixes a kernel panic in reconfig failure:
1. we have a BSS connection
2. we have a NAN connection
3. FW error occurs
4. reconfig restores the BSS connection
5. however, restoring the NAN connection fails due to a FW error.
6. erroneously, ieee80211_handle_reconfig_failure is called and marks all
interfaces as not-in-driver (will be fixed in a different patch).
7. mac80211 frees the links of the BSS connection but doesn't tell the
driver about that, as it thinks that this vif is not in the driver.
8. in ieee80211_stop_device, *ALL* wiphy works are getting flushed
(erroneously?)
9. Therefore, async_handlers_wk is being executed, processing the
statistics notification that was received after we restored the BSS
connection.
10. the notification handler dereferences fw_id_to_bss_conf[id], which is
now a dangling pointer, as mac80211 already freed this link in (7).
11. On the first access to one of the links fields, we panic.
While this can and should be fixed by removing the call to
ieee80211_handle_reconfig_failure in (6), it is also not a good idea to
carry and maybe handle notifications from a dead FW.
We do purge the notifications when we stop the FW, but in reconfig
failure we stop the FW too late, after the notifications are processed.
In addition, async_handlers_wk can always be scheduled before the
reconfig work.
Purge the notifications immediately when transport notifies about a nic
error.
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260517100550.4414228bf1d1.I1926a2b2e7827eaac22882699880ec04a3cb95f0@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
iwl_mld_cancel_async_notifications does 2 things: it purges the list of
notifications and flushes the async_handlers_wk.
We call iwl_mld_cancel_async_notifications after we stopped or suspended
the fw. So in that stage we don't expect any new notification coming,
and if erroneously there are new notifications coming, the work will be
queued again anyway, so the flush is pretty much pointless.
iwl_mld_cancel_async_notifications will need to be called in a context
w/o the wiphy lock held, and the only reason why this function requires
the lock being held is for flushing the work.
Remove the flush.
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260517100550.70dddca96191.I06d3c6433ec22f81f2fb3fb2ee43881e662c5212@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
FM continues to follow the BIOS/MCC policy, while WH sets
DISALLOW_PUNCTURING for US/CA and clears it for other MCC values.
Update the MCC handling accordingly.
Signed-off-by: Pagadala Yesu Anjaneyulu <pagadala.yesu.anjaneyulu@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Link: https://patch.msgid.link/20260515150751.df1f1fdd141f.I900c9e2e3dd722619db12ba10d0879a56a2a55f2@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Store beacon_average_energy from per-link FW statistics and expose it
via station_info as rx_beacon_signal_avg in sta statistics.
This fixes missing beacon average signal reporting to upper layers.
Signed-off-by: Shahar Tzarfati <shahar.tzarfati@intel.com>
Link: https://patch.msgid.link/20260515150751.a74a22d90890.I74d596359c5b69364fb977fdf31396eb57ca0927@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
When a GTK is configured for a station, it can only be used for Rx
and never for Tx. In such a case, set the IWL_SEC_KEY_FLAG_NO_TX
flag when the key is installed, so the FW will not use it for Tx.
Specifically, this is needed for per-station GTK installed on NAN
NDI stations associated with NAN Data interfaces.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260515150751.b004744087cb.I25fb83f9e3dc563d122a160da150d793155513fa@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>