WangYuli [Mon, 23 Jun 2025 07:19:33 +0000 (15:19 +0800)]
gitignore: allow .pylintrc to be tracked
The .pylintrc file was introduced by commit 02df8e3b333c ("docs: add a
.pylintrc file with sys path for docs scripts") to provide Python path
configuration for documentation scripts. However, the generic ".*" rule
in .gitignore causes this tracked file to be ignored, leading to warnings
during kernel builds.
Add !.pylintrc to the exception list to explicitly allow this
configuration file to be tracked by git, consistent with other
development tool configuration files like .clang-format and .rustfmt.toml.
This resolves the build warning:
.pylintrc: warning: ignored by one of the .gitignore files
Fixes: 02df8e3b333c ("docs: add a .pylintrc file with sys path for docs scripts") Signed-off-by: WangYuli <wangyuli@uniontech.com> Reviewed-by: Miguel Ojeda <ojeda@kernel.org> Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/1A357750FF71847E+20250623071933.311947-1-wangyuli@uniontech.com
Bagas Sanjaya [Fri, 20 Jun 2025 10:56:44 +0000 (17:56 +0700)]
Documentation: ext4: Move inode table short docs into its own file
The short description of inode table is in bitmaps.rst alongside the
proper bitmpas documentation. The docs file is short enough that it fits
whole browser screen on desktop, which implies that when readers click
"Inode Table", they will essentially see bitmaps docs.
Bagas Sanjaya [Fri, 20 Jun 2025 10:56:43 +0000 (17:56 +0700)]
Documentation: ext4: blockgroup: Add explicit title heading
Block groups documentation has three, first-level section headings.
These headings' text become toctree entries and the first one "Layout"
becomes docs title in the output, which isn't conveying the docs
contents.
Bagas Sanjaya [Fri, 20 Jun 2025 10:56:42 +0000 (17:56 +0700)]
Documentation: ext4: atomic_writes: Demote last three sections
Last three sections of atomic block writes documentation are adorned as
first-level title headings, which erroneously increase toctree entries
in overview.rst. Demote them.
Fixes: 0bf1f51e34c4 ("ext4: Add atomic block write documentation") Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com> Acked-by: Theodore Ts'o <tytso@mit.edu> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250620105643.25141-5-bagasdotme@gmail.com
Bagas Sanjaya [Fri, 20 Jun 2025 10:56:40 +0000 (17:56 +0700)]
Documentation: ext4: Convert includes into toctrees
ext4 docs are organized in three master docs (overview.rst, globals.rst,
and dynamic.rst), in which these include other docs via include::
directive. These docs sturcture is better served by toctrees instead.
Convert the master docs to use toctrees.
Fixes: 0bf1f51e34c4 ("ext4: Add atomic block write documentation") Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com> Acked-by: Theodore Ts'o <tytso@mit.edu> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250620105643.25141-3-bagasdotme@gmail.com
Add an introductory paragraph to Part Id - Streaming DMA mappings and move
the explanation of address constraints there, because it applies to all map
functions.
Clarify that streaming DMA can be used with memory which does not meet the
addressing constraints of a device, but it may fail in that case.
Make a note about SWIOTLB and link to the detailed description of it.
Do not mention platform-dependent allocation flags. The note may mislead
device driver authors into thinking that they should poke into and try to
second-guess the DMA API implementation. They definitely shouldn't.
Remove the claim that platforms with an IOMMU may not require physically
contiguous buffers. The current implementation explicitly rejects vmalloc
addresses, regardless of IOMMU.
Signed-off-by: Petr Tesarik <ptesarik@suse.com> Reviewed-by: Bagas Sanjaya <bagasdotme@gmail.com> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250627101015.1600042-8-ptesarik@suse.com
Petr Tesarik [Fri, 27 Jun 2025 10:10:12 +0000 (12:10 +0200)]
docs: dma-api: remove duplicate description of the DMA pool API
Move the DMA pool API documentation from Memory Management APIs to
dma-api.rst, replacing the outdated duplicate description there.
Signed-off-by: Petr Tesarik <ptesarik@suse.com> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250627101015.1600042-6-ptesarik@suse.com
Petr Tesarik [Fri, 27 Jun 2025 10:10:11 +0000 (12:10 +0200)]
docs: dma-api: add a kernel-doc comment for dma_pool_zalloc()
Document the dma_pool_zalloc() wrapper.
Signed-off-by: Petr Tesarik <ptesarik@suse.com> Reviewed-by: Randy Dunlap <rdunlap@infradead.org> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
[jc: fixed up dma_pool_alloc() reference in dmapool.h] Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250627101015.1600042-5-ptesarik@suse.com
Petr Tesarik [Fri, 27 Jun 2025 10:10:10 +0000 (12:10 +0200)]
docs: dma-api: remove remnants of PCI DMA API
The wording sometimes suggests there are multiple functions for an
operation. This was in fact the case before PCI DMA API was removed, but
since there is only one API now, the documentation has become confusing.
To improve readability:
* Remove implicit references to the PCI DMA API (plurals, use of "both",
etc.)
* Where possible, refer to an actual function rather than a more generic
description of the operation.
Signed-off-by: Petr Tesarik <ptesarik@suse.com> Reviewed-by: Bagas Sanjaya <bagasdotme@gmail.com> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250627101015.1600042-4-ptesarik@suse.com
Petr Tesarik [Fri, 27 Jun 2025 10:10:09 +0000 (12:10 +0200)]
docs: dma-api: replace consistent with coherent
For consistency, always use the term "coherent" when talking about memory
that is not subject to CPU caching effects. The term "consistent" is a
relic of a long-removed PCI DMA API (pci_alloc_consistent() and
pci_free_consistent() functions).
Signed-off-by: Petr Tesarik <ptesarik@suse.com> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250627101015.1600042-3-ptesarik@suse.com
Petr Tesarik [Fri, 27 Jun 2025 10:10:08 +0000 (12:10 +0200)]
docs: dma-api: use "DMA API" consistently throughout the document
Make sure that all occurrences are spelled "DMA API" (all uppercase, no
hyphen, no underscore).
Signed-off-by: Petr Tesarik <ptesarik@suse.com> Reviewed-by: Randy Dunlap <rdunlap@infradead.org> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250627101015.1600042-2-ptesarik@suse.com
Alison Schofield [Thu, 26 Jun 2025 02:40:58 +0000 (19:40 -0700)]
docs: ABI: make the KernelVersion field optional
The KernelVersion field has limited practical value. Git history
provides more accurate tracking of when features were introduced
and target kernel versions often change during development and
merge.
Jonathan Corbet [Fri, 27 Jun 2025 18:40:00 +0000 (12:40 -0600)]
docs: kdoc: split the processing of the two remaining inline states
Now that "inline_*" are just ordinary parser states, split them into two
separate functions, getting rid of some nested conditional logic.
The original process_inline() would simply ignore lines that didn't match
any of the regexes (those lacking the initial " * " marker). I have
preserved that behavior, but we should perhaps emit a warning instead.
Jonathan Corbet [Fri, 27 Jun 2025 18:39:59 +0000 (12:39 -0600)]
docs: kdoc: remove the inline states-within-a-state
The processing of inline kerneldoc comments is a state like the rest, but
it was implemented as a set of separate substates. Just remove the
substate logic and make the inline states normal ones like the rest.
INLINE_ERROR was never actually used for anything, so just take it out.
Jonathan Corbet [Fri, 27 Jun 2025 18:39:57 +0000 (12:39 -0600)]
docs: kdoc: rework process_export() slightly
Reorganize process_export() to eliminate duplicated code, don't look for
exports in states where we don't expect them, and don't bother with normal
state-machine processing if an export declaration has been found.
Jonathan Corbet [Fri, 27 Jun 2025 18:39:54 +0000 (12:39 -0600)]
docs: kdoc: Move content handling into KernelEntry
Rather than having other code mucking around with this bit of internal
state, encapsulate it internally. Accumulate the description as a list of
strings, joining them at the end, which is a more efficient way of building
the text.
conf.py is missing a SPDX header and doesn't really have
a proper python coding style. It also has an obsolete
commented LaTeX syntax that doesn't work anymore.
Clean it up a little bit with some help from autolints
and manual adjustments.
scripts: test_doc_build.py: make the script smarter
Most of the time, testing the full range of supported Sphinx
version is a waste of time and resources. Instead, the best is
to focus at the versions that are actually shipped by major
distros.
For it to work properly, we need to adjust the requirements for
them to start from first patch for each distro after the
minimal supported one. The requirements were re-adjusted to
avoid build breakages related to version incompatibilities.
Such builds were tested with:
./scripts/test_doc_build.py -m -a "SPHINXOPTS=-j8" "SPHINXDIRS=networking netlink/specs" --full
Change the logic to pick by default only such versions, adding
another parameter to do a comprehensive test.
While here, improve the script documentation to make it easier
to be used.
scripts: test_doc_build.py: improve dependency list
Change the dependency list to ensure that:
- all docutils versions are covered;
- provide an explanation about the dependencies;
- set a better minimal requirement for 3.4.3.
docs: conf.py: properly handle include and exclude patterns
When one does:
make SPHINXDIRS="foo" htmldocs
All patterns would be relative to Documentation/foo, which
causes the include/exclude patterns like:
include_patterns = [
...
f'foo/*.{ext}',
]
to break. This is not what it is expected. Address it by
adding a logic to dynamically adjust the pattern when
SPHINXDIRS is used.
That allows adding parsers for other file types.
It should be noticed that include_patterns was added on
Sphinx 5.1:
https://www.sphinx-doc.org/en/master/usage/configuration.html#confval-include_patterns
So, a backward-compatible code is needed when we start
using it for real.
Jonathan Corbet [Sat, 21 Jun 2025 20:35:12 +0000 (14:35 -0600)]
docs: kdoc: finish disentangling the BODY and SPECIAL_SECTION states
Move the last SPECIAL_SECTION special case into the proper handler
function, getting rid of more if/then/else logic. The leading-space
tracking was tightened up a bit in the move. Add some comments describing
what is going on.
Jonathan Corbet [Sat, 21 Jun 2025 20:35:11 +0000 (14:35 -0600)]
docs: kdoc: Add some comments to process_decl()
Now that the function can actually fit into a human brain, add a few
comments. While I was at it, I switched to the trim_whitespace() helper
rather than open-coding it.
Jonathan Corbet [Sat, 21 Jun 2025 20:35:09 +0000 (14:35 -0600)]
docs: kdoc: rework the handling of SPECIAL_SECTION
Move the recognition of this state to when we enter it, rather than when we
exit, eliminating some twisty logic along the way.
Some changes in output do result from this shift, generally for kerneldoc
comments that do not quite fit the format. See, for example,
struct irqdomain. As far as I can tell, the new behavior is more correct
in each case.
Jonathan Corbet [Sat, 21 Jun 2025 20:35:07 +0000 (14:35 -0600)]
docs: kdoc: split out the special-section state
The state known as BODY_WITH_BLANK_LINE really, in a convoluted way,
indicates a "special section" that is terminated by a blank line or the
beginning of a new section. That is either "@param: desc" sections, or the
weird "context" section that plays by the same rules.
Rename the state to SPECIAL_SECTION and split its processing into a
separate function; no real changes to the logic yet.
Jonathan Corbet [Sat, 21 Jun 2025 20:35:06 +0000 (14:35 -0600)]
docs: kdoc: separate out the handling of the declaration phase
The BODY_MAYBE state really describes the "we are in a declaration" state.
Rename it accordingly, and split the handling of this state out from that
of the other BODY* states. This change introduces a fair amount of
duplicated code that will be coalesced in a later patch.
Jonathan Corbet [Sat, 21 Jun 2025 20:35:04 +0000 (14:35 -0600)]
docs: kdoc: Make body_with_blank_line parsing more flexible
The regex in the BODY_WITH_BLANK_LINE case was looking for lines starting
with " * ", where exactly one space was allowed before the following text.
There are many kerneldoc comments where the authors have put multiple
spaces instead, leading to mis-formatting of the documentation.
Specifically, in this case, the description portion is associated with the
last of the parameters.
Allow multiple spaces in this context.
See, for example, synchronize_hardirq() and how its documentation is
formatted before and after the change.
Bagas Sanjaya [Wed, 11 Jun 2025 06:52:55 +0000 (13:52 +0700)]
Documentation: treewide: Replace remaining spinics links with lore
Long before introduction of lore.kernel.org, people would link
to LKML threads on third-party archives (here spinics.net), which
in some cases can be unreliable (as these were outside of
kernel.org control). Replace links to them with lore counterparts
(if any).
Salvatore Bonaccorso [Thu, 12 Jun 2025 06:02:04 +0000 (08:02 +0200)]
Documentation/sysctl: coredump: add %F for pidfd number
In commit b5325b2a270f ("coredump: hand a pidfd to the usermode coredump
helper") a new core_pattern specifier, %F, was added to provide a pidfs
to the usermode helper process referring to the crashed process.
Update the documentation to include the new core_pattern specifier.
It appears that folks "less versed in kernel coding" think that its
good style to document every function, even if they have no useful
information to pass to the future readers of the code. This used
to be just a waste of space, but with increased kdoc format linting
it's also a burden when refactoring the code.
Shouye Liu [Fri, 20 Jun 2025 02:16:58 +0000 (10:16 +0800)]
Documentation: amd-pstate:fix minimum performance state label error
In the AMD P-States Performance Scale diagram, the labels for "Max Perf"
and "Lowest Perf" were incorrectly used to define the range for
"Desired Perf".The "Desired performance target" should be bounded by the
"Maximum requested performance" and the "Minimum requested performance",
which corresponds to "Max Perf" and "Min Perf", respectively.
Jonathan Corbet [Thu, 19 Jun 2025 21:17:39 +0000 (15:17 -0600)]
docs: sphinx: avoid using the deprecated node.set_class()
Docutils emits a deprecation warning when the set_class() element method is
used; that warning disappears into the ether, but it also causes a crash
with docutils 0.19.
Avoid the deprecated function and just append directly to the "classes"
attribute like the documentation says instead.
Reported-by: Akira Yokosawa <akiyks@gmail.com> Tested-by: Akira Yokosawa <akiyks@gmail.com> Closes: https://lore.kernel.org/de7bae91-3200-481f-9db2-c0dc382c91dd@gmail.com/ Fixes: d6d1df92c25f ("docs: automarkup: Mark up undocumented entities too") Signed-off-by: Jonathan Corbet <corbet@lwn.net>
docs/sched: Make the sched-stats documentation consistent
pull_task(), the original function to move the task from src_rq to the
dst_rq during load balancing was renamed to move_tasks() in commit ddcdf6e7d991 ("sched: Rename load-balancing fields")
As a part of commit 163122b7fcfa ("sched/fair: Remove
double_lock_balance() from load_balance()"), move_task() was broken down
into detach_tasks() and attach_tasks() pair to avoid holding locks of
both src_rq and dst_rq at the same time during load balancing.
Despite the evolution of pull_task() over the years, the sched-stats
documentation remained unchanged. Update the documentation to refer to
detach_task() instead of pull_task() which is responsible for removing
the task from the src_rq during load balancing.
commit 1c055a0f5d3b ("sched: Move sched domain name out of
CONFIG_SCHED_DEBUG") moves sched domain name out of CONFIG_SCHED_DEBUG.
Update the documentation related to that.
Shashank Balaji [Tue, 27 May 2025 14:55:58 +0000 (23:55 +0900)]
sched_deadline, docs: add affinity setting with cgroup2 cpuset controller
Setting the cpu affinity mask of a SCHED_DEADLINE process using the cgroup v1
cpuset controller is already detailed. Add similar information for cgroup v2's
cpuset controller.
Shashank Balaji [Tue, 27 May 2025 14:55:57 +0000 (23:55 +0900)]
sched_deadline, docs: replace rt-app examples with chrt or use config.json
rt-app no longer accepts command-line arguments. So, replace rt-app example
with chrt and use the JSON format in the other example instead of command-
line arguments.
Brigham Campbell [Wed, 28 May 2025 05:41:47 +0000 (23:41 -0600)]
docs: powerpc: Add htm.rst to table of contents
Fix the following documentation build error, which was introduced when
Documentation/arch/powerpc/htm.rst was added to the repository without
any reference to the document.
Documentation/arch/powerpc/htm.rst: WARNING: document isn't included in any toctree [toc.not_included]
Fixes: ab1456c5aa7a ("powerpc/pseries/htmdump: Add documentation for H_HTM debugfs interface") Acked-by: Randy Dunlap <rdunlap@infradead.org> Tested-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Brigham Campbell <me@brighamcampbell.com> Tested-by: Athira Rajeev <atrajeev@linux.ibm.com> Fixes: ab1456c5aa7a63d5 ("powerpc/pseries/htmdump: Add documentation for H_HTM debugfs interface") Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250528054146.2658537-2-me@brighamcampbell.com
Jonathan Corbet [Wed, 4 Jun 2025 14:18:40 +0000 (08:18 -0600)]
docs: CSS: make cross-reference links more evident
The Sphinx Alabaster theme uses border-bottom to mark reference links; the
result does not render correctly (the underline is missing) in some browser
configurations. Switch to using the standard text-underline property, and
use text-underline-offset to place that underline below any underscores in
the underlined text.
Suggested-by: Nícolas F. R. A. Prado <nfraprado@collabora.com> Tested-by: Nícolas F. R. A. Prado <nfraprado@collabora.com> Reviewed-by: Nícolas F. R. A. Prado <nfraprado@collabora.com> Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Jonathan Corbet [Wed, 4 Jun 2025 14:18:04 +0000 (08:18 -0600)]
docs: automarkup: Mark up undocumented entities too
The automarkup code generates markup and a cross-reference link for
functions, structs, etc. for which it finds kerneldoc documentation.
Undocumented entities are left untouched; that creates an inconsistent
reading experience and has caused some writers to go to extra measures to
cause the markup to happen.
Mark up detected C entities regardless of whether they are documented.
Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Jonathan Corbet [Fri, 6 Jun 2025 16:34:38 +0000 (10:34 -0600)]
docs: kdoc: some final touches for process_name()
Add some comments to process_name() to cover its broad phases of operation,
and slightly restructure the if/then/else structure to remove some early
returns.
Jonathan Corbet [Fri, 6 Jun 2025 16:34:37 +0000 (10:34 -0600)]
docs: kdoc: move the declaration regexes out of process_name()
Move two complex regexes up with the other patterns, decluttering this
function and allowing the compilation to be done once rather than for every
kerneldoc comment.
Jonathan Corbet [Fri, 6 Jun 2025 16:34:34 +0000 (10:34 -0600)]
docs: kdoc: remove the KernelEntry::is_kernel_comment member
entry::is_kernel_comment never had anything to do with the entry itself; it
is a bit of local state in one branch of process_name(). It can, in fact,
be removed entirely; rework the code slightly so that it is no longer
needed.
Jonathan Corbet [Fri, 6 Jun 2025 16:34:33 +0000 (10:34 -0600)]
docs: kdoc: simplify the kerneldoc recognition code
process_name() looks for the first line of a kerneldoc comment. It
contains two nearly identical regular expressions, the second of which only
catches six cases in the kernel, all of the form:
define SOME_MACRO_NAME - description
Simply put the "define" into the regex and discard it, eliminating the loop
and the code to remove it specially.
Note that this still treats these defines as if they were functions, but
that's a separate issue.
Jonathan Corbet [Fri, 6 Jun 2025 16:34:31 +0000 (10:34 -0600)]
docs: kdoc: move the core dispatch into a state table
Since all of the handlers already nicely have the same prototype, put them
into a table and call them from there and take out the extended
if-then-else series.
Matthew Wilcox (Oracle) [Wed, 4 Jun 2025 19:57:51 +0000 (20:57 +0100)]
doc: Include scatterlist APIs in htmldocs
We have all this fine kernel-doc written, and it's not published
anywhere. Expose it to public view.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Tested-by: Randy Dunlap <rdunlap@infradead.org> Acked-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250604195753.823742-1-willy@infradead.org
Matthew Wilcox (Oracle) [Fri, 6 Jun 2025 14:15:42 +0000 (15:15 +0100)]
kernel-doc: Fix symbol matching for dropped suffixes
The support for dropping "_noprof" missed dropping the suffix from
exported symbols. That meant that using the :export: feature would
look for kernel-doc for (eg) krealloc_noprof() and not find the
kernel-doc for krealloc().
Fixes: 51a7bf0238c2 (scripts/kernel-doc: drop "_noprof" on function prototypes) Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Link: https://lore.kernel.org/r/20250606141543.1285671-1-willy@infradead.org
Linus Torvalds [Sun, 8 Jun 2025 18:33:00 +0000 (11:33 -0700)]
Merge tag 'timers-cleanups-2025-06-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull timer cleanup from Thomas Gleixner:
"The delayed from_timer() API cleanup:
The renaming to the timer_*() namespace was delayed due massive
conflicts against Linux-next. Now that everything is upstream finish
the conversion"
* tag 'timers-cleanups-2025-06-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
treewide, timers: Rename from_timer() to timer_container_of()
Linus Torvalds [Sun, 8 Jun 2025 18:27:20 +0000 (11:27 -0700)]
Merge tag 'x86-urgent-2025-06-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 fixes from Thomas Gleixner:
"A small set of x86 fixes:
- Cure IO bitmap inconsistencies
A failed fork cleans up all resources of the newly created thread
via exit_thread(). exit_thread() invokes io_bitmap_exit() which
does the IO bitmap cleanups, which unfortunately assume that the
cleanup is related to the current task, which is obviously bogus.
Make it work correctly
- A lockdep fix in the resctrl code removed the clearing of the
command buffer in two places, which keeps stale error messages
around. Bring them back.
- Remove unused trace events"
* tag 'x86-urgent-2025-06-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
fs/resctrl: Restore the rdt_last_cmd_clear() calls after acquiring rdtgroup_mutex
x86/iopl: Cure TIF_IO_BITMAP inconsistencies
x86/fpu: Remove unused trace events
Zhang Rui [Fri, 30 May 2025 00:09:28 +0000 (08:09 +0800)]
tools/power turbostat: Avoid probing the same perf counters
For the RAPL package energy status counter, Intel and AMD share the same
perf_subsys and perf_name, but with different MSR addresses.
Both rapl_counter_arch_infos[0] and rapl_counter_arch_infos[1] are
introduced to describe this counter for different Vendors.
As a result, the perf counter is probed twice, and causes a failure in
in get_rapl_counters() because expected_read_size and actual_read_size
don't match.
Fix the problem by skipping the already probed counter.
Note, this is not a perfect fix. For example, if different
vendors/platforms use the same MSR value for different purpose, the code
can be fooled when it probes a rapl_counter_arch_infos[] entry that does
not belong to the running Vendor/Platform.
In a long run, better to put rapl_counter_arch_infos[] into the
platform_features so that this becomes Vendor/Platform specific.
Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
Zhang Rui [Sat, 17 May 2025 09:44:50 +0000 (17:44 +0800)]
tools/power turbostat: Allow probing RAPL with platform_features->rapl_msrs cleared
platform_features->rapl_msrs describes the RAPL MSRs supported. While
RAPL Perf counters can be exposed from different kernel backend drivers,
e.g. RAPL MSR I/F driver, or RAPL TPMI I/F driver.
Thus, turbostat should first blindly probe all the available RAPL Perf
counters, and falls back to the RAPL MSR counters if they are listed in
platform_features->rapl_msrs.
With this, platforms that don't have RAPL MSRs can clear the
platform_features->rapl_msrs bits and use RAPL Perf counters only.
Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
probe_rapl_msr() is reused for probing RAPL MSR counters, cstate MSR
counters and MPERF/APERF/SMI MSR counters, thus its name is misleading.
Similar to add_perf_counter(), introduce add_msr_counter() to probe a
counter via MSR. Introduce wrapper function add_rapl_msr_counter() at
the same time to add extra check for Zero return value for specified
RAPL counters.
No functional change intended.
Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
As the only caller of add_cstate_perf_counter_(),
add_cstate_perf_counter() just gives extra debug output on top. There is
no need to keep both functions.
Remove add_cstate_perf_counter_() and move all the logic to
add_cstate_perf_counter().
No functional change.
Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
commit 05a2f07db888 ("tools/power turbostat: read RAPL counters via
perf") that adds support to read RAPL counters via perf defines the
notion of a RAPL domain_id which is set to physical_core_id on
platforms which support per_core_rapl counters (Eg: AMD processors
Family 17h onwards) and is set to the physical_package_id on all the
other platforms.
However, the physical_core_id is only unique within a package and on
platforms with multiple packages more than one core can have the same
physical_core_id and thus the same domain_id. (For eg, the first cores
of each package have the physical_core_id = 0). This results in all
these cores with the same physical_core_id using the same entry in the
rapl_counter_info_perdomain[]. Since rapl_perf_init() skips the
perf-initialization for cores whose domain_ids have already been
visited, cores that have the same physical_core_id always read the
perf file corresponding to the physical_core_id of the first package
and thus the package-energy is incorrectly reported to be the same
value for different packages.
Note: This issue only arises when RAPL counters are read via perf and
not when they are read via MSRs since in the latter case the MSRs are
read separately on each core.
Fix this issue by associating each CPU with rapl_core_id which is
unique across all the packages in the system.
Fixes: 05a2f07db888 ("tools/power turbostat: read RAPL counters via perf") Signed-off-by: Gautham R. Shenoy <gautham.shenoy@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
projects / users / willy / linux.git / log