From 22098763a10d9c1340827fcf6edab66f153b27f0 Mon Sep 17 00:00:00 2001 From: Linus Torvalds Date: Sun, 13 Sep 2026 12:27:00 -0700 Subject: Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace Pull tracing fixes from Steven Rostedt: - Don't destroy user event fields when removal fails User event fields are destroyed before the event is removed from visibility. But that can fail leaving the still visible event with no fields. Move the destroying of the fields to after the event is successfully removed from visibility. - Initialize function graph state is fork before calling copy_exec_state() For non-CLONE_VM forks, copy_exec_state() allocates a new task_exec_state. If that allocation fails, ftrace_graph_exit_task() will free the tasks ret_stack pointer. Since that pointer is still using the parent's ret_stack, it mistakenly frees the parent's pointer too. Call ftrace_graph_init() on the task first which will NULL out the new tasks's ret_stack and if the copy fails, it will not free anything. - Remove FGRAPH_MAX_INDEX The macro FGRAPH_MAX_INDEX was added but never used. Remove it. - Save ent_size in function graph printing of nested functions The function graph tracer needs to look at the next event to see if the next event is the return of the current function entry. If it is, it prints a single line: ktime_get(); Otherwise it prints it like a nested function: tick_nohz_irq_exit() { ktime_get(); kcpustat_irq_exit(); } In order to look at the next event, it must save the current event so that it has the information to print from it. It saves the event in the iterator descriptor called "ent". What it doesn't save is the ent_size of the event which is now used to know if the function graph arguments are to be printed. The peek doesn't save the size so the size used happens to be that of the size of the last event that was seen. Save the entry event size in the iterator descriptor so that the correct size is used. - Fix several errors with freeing data in the histogram code The histogram code had a lot of leaked or or incorrect accounting when failures happen. Correct them. - Fix histogram regression of .percent and .graph modifiers Up until 6.3 histogram values could have "percent" or "graph" modifiers that changed how they were printed. But a change that added restricting histograms values from being strings, stack traces and other modifiers inadvertently prevented them from using the percent and graph modifiers, which were legal use cases for values. Put back the percent and graph modifiers. - Fix various typos in the comments - Set the trace_clock before initializing a histogram with clock argument The histogram API allows the user to specific which trace clock to use via a "clock=" string. The histogram is set up first before the clock is checked. If the passed in clock is not valid, it exits without fully fixing up the histogram leaving it on the list and a use-after-free can trigger. Update the clock argument first and if it fails then exit gracefully before the histogram trigger is placed on any lists. - Restore :mod: trailer after parsing in ftrace_set_clr_event The function ftrace_set_clr_event() modifies the parse string and needs to put it back to what was passed in. It searches for ":mod:" via a strsep() but fails to put back the first ':' in the string. Add back the ':' in the passed in string. - Take trace_array reference when opening a tracer options file The options files are dynamically created and some tracers add their own options. When a tracer adds their own list of options, the trace_array holding them has an array to hold the list of options for each tracer. This array increases in size via a krealloc(), and the new entry gets a newly allocated array to hold the options of the new tracer being added. The element in each entry of the tracer's option array holds a pointer back to the trace_array, a pointer to the tracer it is associated to, a pointer to the flags of the option. The issue is that these arrays are freed when the trace_array is freed when its instance it represents is removed from the instances directory. There's a race that an open of one of these options files can happen when the instance is being removed. Add a new helper function to be called by the open function of the options file to iterate all existing trace_arrays under a lock and find the one that has the given option element in one of it's tracer arrays. If found, then update the associated trace_array's reference counter to keep it from being freed. If not found, have the open call return -ENODEV. - Disable interrupts when acquiring the lock in rb_wake_up_waiters() The function rb_wake_up_waiters() assumes it will be called in interrupt context and does not disable irqs when taking cpu_buffer->reader_lock, which can be called in hard interrupt context. The issue is in PREEMPT_RT, this function is called in thread context leaving this lock open to a deadlock. Take the lock with interrupts disabled. - Use rcu_assign_pointer() for tmp_ops filter hash The tmp_ops used in update_ftrace_direct_mod() assigns its filter_hash field directly, but that field is annotated as __rcu and sparse complains. Assign it with rcu_assign_pointer() - Fix use-after-free in enable_trigger_private_data_free() The trace_event_call is accessed through the event_trigger_data's trace_event_file pointer to put the trace_event_call on freeing. The issue is that the trace_event_file data may have been freed already causing a use-after-free. Add a field to the event_trigger_data that points directly to the trace_event_call so that it can decrement its reference directly without needing to go through the trace_event_file. - Fix accounting of buffer data remote headers trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount the number of pages is needed for the asked for size as it doesn't take into account the meta data on each page. Add a helper function to do the calculation properly and use that in these functions. - Catch nr_page_va overflow in ring_buffer_desc sizing The number of pages per remote ring buffer is capped by ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to overflow that field would silently allocate a descriptor smaller than what was asked for. - Do not resize the subbuf order if any per_cpu buffer is disabled The mmapping of ring buffers disables resizing the subbuffers, but it is done per-cpu whereas the subbuf size change is done for all the per_cpu buffers under the buffer->mutex. It could change the size of some while the mapping is happening on others. Have the resize of the subbuf order check all the per_cpu buffers under the lock to see if any of them is disabled before starting and causing an inconsistency between buffers that are being mapped. * tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits) ring-buffer: Check resize_disabled before publishing the new subbuf order tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing tracing/remotes: Account for ring buffer page header in size calculation tracing: Don't dereference trace_event_file in deferred trigger free ftrace: Use rcu_assign_pointer() for tmp_ops filter hash ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters() tracing: Take trace_array reference when opening a tracer options file tracing: Fix ring_buffer_read_page_size() kernel-doc tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event() tracing: Fix memory corruption from a "STACKTRACE" histogram key tracing: Fix memory corruption from the histogram stacktrace modifier tracing: Undo the registration when enabling the histogram trigger fails tracing: Take the reference before publishing the named histogram trigger tracing: Set the trace clock before registering the histogram trigger tracing: Fix typo "preceeded" in comment tracing: Fix typo "availabe" in comment tracing: Let histogram values keep the percent and graph modifiers tracing: Keep the entry count when the histogram stats allocation fails tracing: Free histogram the field rejected for a bad modifier tracing: Free histogram the var ref when its initialization fails ... --- tools/lib/python/abi/__init__.py | 0 tools/lib/python/abi/abi_parser.py | 631 +++++++++++++++++++++++++++++++++ tools/lib/python/abi/abi_regex.py | 247 +++++++++++++ tools/lib/python/abi/helpers.py | 40 +++ tools/lib/python/abi/system_symbols.py | 378 ++++++++++++++++++++ 5 files changed, 1296 insertions(+) create mode 100644 tools/lib/python/abi/__init__.py create mode 100644 tools/lib/python/abi/abi_parser.py create mode 100644 tools/lib/python/abi/abi_regex.py create mode 100644 tools/lib/python/abi/helpers.py create mode 100644 tools/lib/python/abi/system_symbols.py (limited to 'tools/lib/python/abi') diff --git a/tools/lib/python/abi/__init__.py b/tools/lib/python/abi/__init__.py new file mode 100644 index 000000000..e69de29bb diff --git a/tools/lib/python/abi/abi_parser.py b/tools/lib/python/abi/abi_parser.py new file mode 100644 index 000000000..d7bb20ef3 --- /dev/null +++ b/tools/lib/python/abi/abi_parser.py @@ -0,0 +1,631 @@ +#!/usr/bin/env python3 +# pylint: disable=R0902,R0903,R0911,R0912,R0913,R0914,R0915,R0917,C0302 +# Copyright(c) 2025: Mauro Carvalho Chehab . +# SPDX-License-Identifier: GPL-2.0 + +""" +Parse ABI documentation and produce results from it. +""" + +from argparse import Namespace +import logging +import os +import re + +from pprint import pformat +from random import randrange, seed + +# Import Python modules + +from abi.helpers import AbiDebug, ABI_DIR + + +class AbiParser: + """Main class to parse ABI files.""" + + #: Valid tags at Documentation/ABI. + TAGS = r"(what|where|date|kernelversion|contact|description|users)" + + #: ABI elements that will auto-generate cross-references. + XREF = r"(?:^|\s|\()(\/(?:sys|config|proc|dev|kvd)\/[^,.:;\)\s]+)(?:[,.:;\)\s]|\Z)" + + def __init__(self, directory, logger=None, + enable_lineno=False, show_warnings=True, debug=0): + """Stores arguments for the class and initialize class vars.""" + + self.directory = directory + self.enable_lineno = enable_lineno + self.show_warnings = show_warnings + self.debug = debug + + if not logger: + self.log = logging.getLogger("get_abi") + else: + self.log = logger + + self.data = {} + self.what_symbols = {} + self.file_refs = {} + self.what_refs = {} + + # Ignore files that contain such suffixes + self.ignore_suffixes = (".rej", ".org", ".orig", ".bak", "~") + + # Regular expressions used on parser + self.re_abi_dir = re.compile(r"(.*)" + ABI_DIR) + self.re_tag = re.compile(r"(\S+)(:\s*)(.*)", re.I) + self.re_valid = re.compile(self.TAGS) + self.re_start_spc = re.compile(r"(\s*)(\S.*)") + self.re_whitespace = re.compile(r"^\s+") + + # Regular used on print + self.re_what = re.compile(r"(\/?(?:[\w\-]+\/?){1,2})") + self.re_escape = re.compile(r"([\.\x01-\x08\x0e-\x1f\x21-\x2f\x3a-\x40\x7b-\xff])") + self.re_unprintable = re.compile(r"([\x00-\x2f\x3a-\x40\x5b-\x60\x7b-\xff]+)") + self.re_title_mark = re.compile(r"\n[\-\*\=\^\~]+\n") + self.re_doc = re.compile(r"Documentation/(?!devicetree)(\S+)\.rst") + self.re_abi = re.compile(r"(Documentation/ABI/)([\w\/\-]+)") + self.re_xref_node = re.compile(self.XREF) + + def warn(self, fdata, msg, extra=None): + """Displays a parse error if warning is enabled.""" + + if not self.show_warnings: + return + + msg = f"{fdata.fname}:{fdata.ln}: {msg}" + if extra: + msg += "\n\t\t" + extra + + self.log.warning(msg) + + def add_symbol(self, what, fname, ln=None, xref=None): + """Create a reference table describing where each 'what' is located.""" + + if what not in self.what_symbols: + self.what_symbols[what] = {"file": {}} + + if fname not in self.what_symbols[what]["file"]: + self.what_symbols[what]["file"][fname] = [] + + if ln and ln not in self.what_symbols[what]["file"][fname]: + self.what_symbols[what]["file"][fname].append(ln) + + if xref: + self.what_symbols[what]["xref"] = xref + + def _parse_line(self, fdata, line): + """Parse a single line of an ABI file.""" + + new_what = False + new_tag = False + content = None + + match = self.re_tag.match(line) + if match: + new = match.group(1).lower() + sep = match.group(2) + content = match.group(3) + + match = self.re_valid.search(new) + if match: + new_tag = match.group(1) + else: + if fdata.tag == "description": + # New "tag" is actually part of description. + # Don't consider it a tag + new_tag = False + elif fdata.tag != "": + self.warn(fdata, f"tag '{fdata.tag}' is invalid", line) + + if new_tag: + # "where" is Invalid, but was a common mistake. Warn if found + if new_tag == "where": + self.warn(fdata, "tag 'Where' is invalid. Should be 'What:' instead") + new_tag = "what" + + if new_tag == "what": + fdata.space = None + + if content not in self.what_symbols: + self.add_symbol(what=content, fname=fdata.fname, ln=fdata.ln) + + if fdata.tag == "what": + fdata.what.append(content.strip("\n")) + else: + if fdata.key: + if "description" not in self.data.get(fdata.key, {}): + self.warn(fdata, f"{fdata.key} doesn't have a description") + + for w in fdata.what: + self.add_symbol(what=w, fname=fdata.fname, + ln=fdata.what_ln, xref=fdata.key) + + fdata.label = content + new_what = True + + key = "abi_" + content.lower() + fdata.key = self.re_unprintable.sub("_", key).strip("_") + + # Avoid duplicated keys but using a defined seed, to make + # the namespace identical if there aren't changes at the + # ABI symbols + seed(42) + + while fdata.key in self.data: + char = randrange(0, 51) + ord("A") + if char > ord("Z"): + char += ord("a") - ord("Z") - 1 + + fdata.key += chr(char) + + if fdata.key and fdata.key not in self.data: + self.data[fdata.key] = { + "what": [content], + "file": [fdata.file_ref], + "path": fdata.ftype, + "line_no": fdata.ln, + } + + fdata.what = self.data[fdata.key]["what"] + + self.what_refs[content] = fdata.key + fdata.tag = new_tag + fdata.what_ln = fdata.ln + + if fdata.nametag["what"]: + t = (content, fdata.key) + if t not in fdata.nametag["symbols"]: + fdata.nametag["symbols"].append(t) + + return + + if fdata.tag and new_tag: + fdata.tag = new_tag + + if new_what: + fdata.label = "" + + if "description" in self.data[fdata.key]: + self.data[fdata.key]["description"] += "\n\n" + + if fdata.file_ref not in self.data[fdata.key]["file"]: + self.data[fdata.key]["file"].append(fdata.file_ref) + + if self.debug == AbiDebug.WHAT_PARSING: + self.log.debug("what: %s", fdata.what) + + if not fdata.what: + self.warn(fdata, "'What:' should come first:", line) + return + + if new_tag == "description": + fdata.space = None + + if content: + sep = sep.replace(":", " ") + + c = " " * len(new_tag) + sep + content + c = c.expandtabs() + + match = self.re_start_spc.match(c) + if match: + # Preserve initial spaces for the first line + fdata.space = match.group(1) + content = match.group(2) + "\n" + + self.data[fdata.key][fdata.tag] = content + + return + + # Store any contents before tags at the database + if not fdata.tag and "what" in fdata.nametag: + fdata.nametag["description"] += line + return + + if fdata.tag == "description": + content = line.expandtabs() + + if self.re_whitespace.sub("", content) == "": + self.data[fdata.key][fdata.tag] += "\n" + return + + if fdata.space is None: + match = self.re_start_spc.match(content) + if match: + # Preserve initial spaces for the first line + fdata.space = match.group(1) + + content = match.group(2) + "\n" + else: + if content.startswith(fdata.space): + content = content[len(fdata.space):] + + else: + fdata.space = "" + + if fdata.tag == "what": + w = content.strip("\n") + if w: + self.data[fdata.key][fdata.tag].append(w) + else: + self.data[fdata.key][fdata.tag] += content + return + + content = line.strip() + if fdata.tag: + if fdata.tag == "what": + w = content.strip("\n") + if w: + self.data[fdata.key][fdata.tag].append(w) + else: + self.data[fdata.key][fdata.tag] += "\n" + content.rstrip("\n") + return + + # Everything else is error + if content: + self.warn(fdata, "Unexpected content", line) + + def parse_readme(self, nametag, fname): + """Parse ABI README file.""" + + nametag["what"] = ["Introduction"] + nametag["path"] = "README" + with open(fname, "r", encoding="utf8", errors="backslashreplace") as fp: + for line in fp: + match = self.re_tag.match(line) + if match: + new = match.group(1).lower() + + match = self.re_valid.search(new) + if match: + nametag["description"] += "\n:" + line + continue + + nametag["description"] += line + + def parse_file(self, fname, path, basename): + """Parse a single file.""" + + ref = f"abi_file_{path}_{basename}" + ref = self.re_unprintable.sub("_", ref).strip("_") + + # Store per-file state into a namespace variable. This will be used + # by the per-line parser state machine and by the warning function. + fdata = Namespace + + fdata.fname = fname + fdata.name = basename + + pos = fname.find(ABI_DIR) + if pos > 0: + f = fname[pos:] + else: + f = fname + + fdata.file_ref = (f, ref) + self.file_refs[f] = ref + + fdata.ln = 0 + fdata.what_ln = 0 + fdata.tag = "" + fdata.label = "" + fdata.what = [] + fdata.key = None + fdata.xrefs = None + fdata.space = None + fdata.ftype = path.split("/")[0] + + fdata.nametag = {} + fdata.nametag["what"] = [f"ABI file {path}/{basename}"] + fdata.nametag["type"] = "File" + fdata.nametag["path"] = fdata.ftype + fdata.nametag["file"] = [fdata.file_ref] + fdata.nametag["line_no"] = 1 + fdata.nametag["description"] = "" + fdata.nametag["symbols"] = [] + + self.data[ref] = fdata.nametag + + if self.debug & AbiDebug.WHAT_OPEN: + self.log.debug("Opening file %s", fname) + + if basename == "README": + self.parse_readme(fdata.nametag, fname) + return + + with open(fname, "r", encoding="utf8", errors="backslashreplace") as fp: + for line in fp: + fdata.ln += 1 + + self._parse_line(fdata, line) + + if "description" in fdata.nametag: + fdata.nametag["description"] = fdata.nametag["description"].lstrip("\n") + + if fdata.key: + if "description" not in self.data.get(fdata.key, {}): + self.warn(fdata, f"{fdata.key} doesn't have a description") + + for w in fdata.what: + self.add_symbol(what=w, fname=fname, xref=fdata.key) + + def _parse_abi(self, root=None): + """Internal function to parse documentation ABI recursively.""" + + if not root: + root = self.directory + + with os.scandir(root) as obj: + for entry in obj: + name = os.path.join(root, entry.name) + + if entry.is_dir(): + self._parse_abi(name) + continue + + if not entry.is_file(): + continue + + basename = os.path.basename(name) + + if basename.startswith("."): + continue + + if basename.endswith(self.ignore_suffixes): + continue + + path = self.re_abi_dir.sub("", os.path.dirname(name)) + + self.parse_file(name, path, basename) + + def parse_abi(self, root=None): + """Parse documentation ABI.""" + + self._parse_abi(root) + + if self.debug & AbiDebug.DUMP_ABI_STRUCTS: + self.log.debug(pformat(self.data)) + + def desc_txt(self, desc): + """Print description as found inside ABI files.""" + + desc = desc.strip(" \t\n") + + return desc + "\n\n" + + def xref(self, fname): + """ + Converts a Documentation/ABI + basename into a ReST cross-reference. + """ + + xref = self.file_refs.get(fname) + if not xref: + return None + else: + return xref + + def desc_rst(self, desc): + """Enrich ReST output by creating cross-references.""" + + # Remove title markups from the description + # Having titles inside ABI files will only work if extra + # care would be taken in order to strictly follow the same + # level order for each markup. + desc = self.re_title_mark.sub("\n\n", "\n" + desc) + desc = desc.rstrip(" \t\n").lstrip("\n") + + # Python's regex performance for non-compiled expressions is a lot + # than Perl, as Perl automatically caches them at their + # first usage. Here, we'll need to do the same, as otherwise the + # performance penalty is be high + + new_desc = "" + for d in desc.split("\n"): + if d == "": + new_desc += "\n" + continue + + # Use cross-references for doc files where needed + d = self.re_doc.sub(r":doc:`/\1`", d) + + # Use cross-references for ABI generated docs where needed + matches = self.re_abi.findall(d) + for m in matches: + abi = m[0] + m[1] + + xref = self.file_refs.get(abi) + if not xref: + # This may happen if ABI is on a separate directory, + # like parsing ABI testing and symbol is at stable. + # The proper solution is to move this part of the code + # for it to be inside sphinx/kernel_abi.py + self.log.info("Didn't find ABI reference for '%s'", abi) + else: + new = self.re_escape.sub(r"\\\1", m[1]) + d = re.sub(fr"\b{abi}\b", f":ref:`{new} <{xref}>`", d) + + # Seek for cross reference symbols like /sys/... + # Need to be careful to avoid doing it on a code block + if d[0] not in [" ", "\t"]: + matches = self.re_xref_node.findall(d) + for m in matches: + # Finding ABI here is more complex due to wildcards + xref = self.what_refs.get(m) + if xref: + new = self.re_escape.sub(r"\\\1", m) + d = re.sub(fr"\b{m}\b", f":ref:`{new} <{xref}>`", d) + + new_desc += d + "\n" + + return new_desc + "\n\n" + + def doc(self, output_in_txt=False, show_symbols=True, show_file=True, + filter_path=None): + """Print ABI at stdout.""" + + part = None + for key, v in sorted(self.data.items(), + key=lambda x: (x[1].get("type", ""), + x[1].get("what"))): + + wtype = v.get("type", "Symbol") + file_ref = v.get("file") + names = v.get("what", [""]) + + if wtype == "File": + if not show_file: + continue + else: + if not show_symbols: + continue + + if filter_path: + if v.get("path") != filter_path: + continue + + msg = "" + + if wtype != "File": + cur_part = names[0] + if cur_part.find("/") >= 0: + match = self.re_what.match(cur_part) + if match: + symbol = match.group(1).rstrip("/") + cur_part = "Symbols under " + symbol + + if cur_part and cur_part != part: + part = cur_part + msg += part + "\n"+ "-" * len(part) +"\n\n" + + msg += f".. _{key}:\n\n" + + max_len = 0 + for i in range(0, len(names)): # pylint: disable=C0200 + names[i] = "**" + self.re_escape.sub(r"\\\1", names[i]) + "**" + + max_len = max(max_len, len(names[i])) + + msg += "+-" + "-" * max_len + "-+\n" + for name in names: + msg += f"| {name}" + " " * (max_len - len(name)) + " |\n" + msg += "+-" + "-" * max_len + "-+\n" + msg += "\n" + + for ref in file_ref: + if wtype == "File": + msg += f".. _{ref[1]}:\n\n" + else: + base = os.path.basename(ref[0]) + msg += f"Defined on file :ref:`{base} <{ref[1]}>`\n\n" + + if wtype == "File": + msg += names[0] +"\n" + "-" * len(names[0]) +"\n\n" + + desc = v.get("description") + if not desc and wtype != "File": + msg += f"DESCRIPTION MISSING for {names[0]}\n\n" + + if desc: + if output_in_txt: + msg += self.desc_txt(desc) + else: + msg += self.desc_rst(desc) + + symbols = v.get("symbols") + if symbols: + msg += "Has the following ABI:\n\n" + + for w, label in symbols: + # Escape special chars from content + content = self.re_escape.sub(r"\\\1", w) + + msg += f"- :ref:`{content} <{label}>`\n\n" + + users = v.get("users") + if users and users.strip(" \t\n"): + users = users.strip("\n").replace('\n', '\n\t') + msg += f"Users:\n\t{users}\n\n" + + ln = v.get("line_no", 1) + + yield (msg, file_ref[0][0], ln) + + def check_issues(self): + """Warn about duplicated ABI entries.""" + + for what, v in self.what_symbols.items(): + files = v.get("file") + if not files: + # Should never happen if the parser works properly + self.log.warning("%s doesn't have a file associated", what) + continue + + if len(files) == 1: + continue + + f = [] + for fname, lines in sorted(files.items()): + if not lines: + f.append(f"{fname}") + elif len(lines) == 1: + f.append(f"{fname}:{lines[0]}") + else: + m = fname + "lines " + m += ", ".join(str(x) for x in lines) + f.append(m) + + self.log.warning("%s is defined %d times: %s", what, len(f), "; ".join(f)) + + def search_symbols(self, expr): + """ Searches for ABI symbols.""" + + regex = re.compile(expr, re.I) + + found_keys = 0 + for t in sorted(self.data.items(), key=lambda x: [0]): + v = t[1] + + wtype = v.get("type", "") + if wtype == "File": + continue + + for what in v.get("what", [""]): + if regex.search(what): + found_keys += 1 + + kernelversion = v.get("kernelversion", "").strip(" \t\n") + date = v.get("date", "").strip(" \t\n") + contact = v.get("contact", "").strip(" \t\n") + users = v.get("users", "").strip(" \t\n") + desc = v.get("description", "").strip(" \t\n") + + files = [] + for f in v.get("file", ()): + files.append(f[0]) + + what = str(found_keys) + ". " + what + title_tag = "-" * len(what) + + print(f"\n{what}\n{title_tag}\n") + + if kernelversion: + print(f"Kernel version:\t\t{kernelversion}") + + if date: + print(f"Date:\t\t\t{date}") + + if contact: + print(f"Contact:\t\t{contact}") + + if users: + print(f"Users:\t\t\t{users}") + + print("Defined on file(s):\t" + ", ".join(files)) + + if desc: + desc = desc.strip("\n") + print(f"\n{desc}\n") + + if not found_keys: + print(f"Regular expression /{expr}/ not found.") diff --git a/tools/lib/python/abi/abi_regex.py b/tools/lib/python/abi/abi_regex.py new file mode 100644 index 000000000..198ecbf49 --- /dev/null +++ b/tools/lib/python/abi/abi_regex.py @@ -0,0 +1,247 @@ +#!/usr/bin/env python3 +# xxpylint: disable=R0903 +# Copyright(c) 2025: Mauro Carvalho Chehab . +# SPDX-License-Identifier: GPL-2.0 + +""" +Convert ABI what into regular expressions +""" + +import re +import sys + +from pprint import pformat + +from abi.abi_parser import AbiParser +from abi.helpers import AbiDebug + +class AbiRegex(AbiParser): + """ + Extends AbiParser to search ABI nodes with regular expressions. + + There some optimizations here to allow a quick symbol search: + instead of trying to place all symbols altogether an doing linear + search which is very time consuming, create a tree with one depth, + grouping similar symbols altogether. + + Yet, sometimes a full search will be needed, so we have a special branch + on such group tree where other symbols are placed. + """ + + #: Escape only ASCII visible characters. + escape_symbols = r"([\x21-\x29\x2b-\x2d\x3a-\x40\x5c\x60\x7b-\x7e])" + + #: Special group for other nodes. + leave_others = "others" + + # Tuples with regular expressions to be compiled and replacement data + re_whats = [ + # Drop escape characters that might exist + (re.compile("\\\\"), ""), + + # Temporarily escape dot characters + (re.compile(r"\."), "\xf6"), + + # Temporarily change [0-9]+ type of patterns + (re.compile(r"\[0\-9\]\+"), "\xff"), + + # Temporarily change [\d+-\d+] type of patterns + (re.compile(r"\[0\-\d+\]"), "\xff"), + (re.compile(r"\[0:\d+\]"), "\xff"), + (re.compile(r"\[(\d+)\]"), "\xf4\\\\d+\xf5"), + + # Temporarily change [0-9] type of patterns + (re.compile(r"\[(\d)\-(\d)\]"), "\xf4\1-\2\xf5"), + + # Handle multiple option patterns + (re.compile(r"[\{\<\[]([\w_]+)(?:[,|]+([\w_]+)){1,}[\}\>\]]"), r"(\1|\2)"), + + # Handle wildcards + (re.compile(r"([^\/])\*"), "\\1\\\\w\xf7"), + (re.compile(r"/\*/"), "/.*/"), + (re.compile(r"/\xf6\xf6\xf6"), "/.*"), + (re.compile(r"\<[^\>]+\>"), "\\\\w\xf7"), + (re.compile(r"\{[^\}]+\}"), "\\\\w\xf7"), + (re.compile(r"\[[^\]]+\]"), "\\\\w\xf7"), + + (re.compile(r"XX+"), "\\\\w\xf7"), + (re.compile(r"(? + # (this happens on a few IIO definitions) + (re.compile(r"\s*\=.*$"), ""), + + # Escape all other symbols + (re.compile(escape_symbols), r"\\\1"), + (re.compile(r"\\\\"), r"\\"), + (re.compile(r"\\([\[\]\(\)\|])"), r"\1"), + (re.compile(r"(\d+)\\(-\d+)"), r"\1\2"), + + (re.compile(r"\xff"), r"\\d+"), + + # Special case: IIO ABI which a parenthesis. + (re.compile(r"sqrt(.*)"), r"sqrt(.*)"), + + # Simplify regexes with multiple .* + (re.compile(r"(?:\.\*){2,}"), ""), + + # Recover dot characters + (re.compile(r"\xf6"), "\\."), + # Recover plus characters + (re.compile(r"\xf7"), "+"), + ] + + #: Regex to check if the symbol name has a number on it. + re_has_num = re.compile(r"\\d") + + #: Symbol name after escape_chars that are considered a devnode basename. + re_symbol_name = re.compile(r"(\w|\\[\.\-\:])+$") + + #: List of popular group names to be skipped to minimize regex group size + #: Use AbiDebug.SUBGROUP_SIZE to detect those. + skip_names = set(["devices", "hwmon"]) + + def regex_append(self, what, new): + """ + Get a search group for a subset of regular expressions. + + As ABI may have thousands of symbols, using a for to search all + regular expressions is at least O(n^2). When there are wildcards, + the complexity increases substantially, eventually becoming exponential. + + To avoid spending too much time on them, use a logic to split + them into groups. The smaller the group, the better, as it would + mean that searches will be confined to a small number of regular + expressions. + + The conversion to a regex subset is tricky, as we need something + that can be easily obtained from the sysfs symbol and from the + regular expression. So, we need to discard nodes that have + wildcards. + + If it can't obtain a subgroup, place the regular expression inside + a special group (self.leave_others). + """ + + search_group = None + + for search_group in reversed(new.split("/")): + if not search_group or search_group in self.skip_names: + continue + if self.re_symbol_name.match(search_group): + break + + if not search_group: + search_group = self.leave_others + + if self.debug & AbiDebug.SUBGROUP_MAP: + self.log.debug("%s: mapped as %s", what, search_group) + + try: + if search_group not in self.regex_group: + self.regex_group[search_group] = [] + + self.regex_group[search_group].append(re.compile(new)) + if self.search_string: + if what.find(self.search_string) >= 0: + print(f"What: {what}") + except re.error: + self.log.warning("Ignoring '%s' as it produced an invalid regex:\n" + " '%s'", what, new) + + def get_regexes(self, what): + """ + Given an ABI devnode, return a list of all regular expressions that + may match it, based on the sub-groups created by regex_append(). + """ + + re_list = [] + + patches = what.split("/") + patches.reverse() + patches.append(self.leave_others) + + for search_group in patches: + if search_group in self.regex_group: + re_list += self.regex_group[search_group] + + return re_list + + def __init__(self, *args, **kwargs): + """ + Override init method to get verbose argument + """ + + self.regex_group = None + self.search_string = None + self.re_string = None + + if "search_string" in kwargs: + self.search_string = kwargs.get("search_string") + del kwargs["search_string"] + + if self.search_string: + + try: + self.re_string = re.compile(self.search_string) + except re.error as e: + msg = f"{self.search_string} is not a valid regular expression" + raise ValueError(msg) from e + + super().__init__(*args, **kwargs) + + def parse_abi(self, *args, **kwargs): + + super().parse_abi(*args, **kwargs) + + self.regex_group = {} + + print("Converting ABI What fields into regexes...", file=sys.stderr) + + for t in sorted(self.data.items(), key=lambda x: x[0]): + v = t[1] + if v.get("type") == "File": + continue + + v["regex"] = [] + + for what in v.get("what", []): + if not what.startswith("/sys"): + continue + + new = what + for r, s in self.re_whats: + try: + new = r.sub(s, new) + except re.error as e: + # Help debugging troubles with new regexes + raise re.error(f"{e}\nwhile re.sub('{r.pattern}', {s}, str)") from e + + v["regex"].append(new) + + if self.debug & AbiDebug.REGEX: + self.log.debug("%-90s <== %s", new, what) + + # Store regex into a subgroup to speedup searches + self.regex_append(what, new) + + if self.debug & AbiDebug.SUBGROUP_DICT: + self.log.debug("%s", pformat(self.regex_group)) + + if self.debug & AbiDebug.SUBGROUP_SIZE: + biggestd_keys = sorted(self.regex_group.keys(), + key= lambda k: len(self.regex_group[k]), + reverse=True) + + print("Top regex subgroups:", file=sys.stderr) + for k in biggestd_keys[:10]: + print(f"{k} has {len(self.regex_group[k])} elements", file=sys.stderr) diff --git a/tools/lib/python/abi/helpers.py b/tools/lib/python/abi/helpers.py new file mode 100644 index 000000000..2a378d780 --- /dev/null +++ b/tools/lib/python/abi/helpers.py @@ -0,0 +1,40 @@ +#!/usr/bin/env python3 +# Copyright(c) 2025: Mauro Carvalho Chehab . +# pylint: disable=R0903 +# SPDX-License-Identifier: GPL-2.0 + +""" +Helper classes for ABI parser +""" + +ABI_DIR = "Documentation/ABI/" + + +class AbiDebug: + """Debug levels""" + + WHAT_PARSING = 1 #: Enable debug parsing logic. + WHAT_OPEN = 2 #: Enable debug messages on file open. + DUMP_ABI_STRUCTS = 4 #: Enable debug for ABI parse data. + UNDEFINED = 8 #: Enable extra undefined symbol data. + REGEX = 16 #: Enable debug for what to regex conversion. + SUBGROUP_MAP = 32 #: Enable debug for symbol regex subgroups + SUBGROUP_DICT = 64 #: Enable debug for sysfs graph tree variable. + SUBGROUP_SIZE = 128 #: Enable debug of search groups. + GRAPH = 256 #: Display ref tree graph for undefined symbols. + +#: Helper messages for each debug variable +DEBUG_HELP = """ +1 - enable debug parsing logic +2 - enable debug messages on file open +4 - enable debug for ABI parse data +8 - enable extra debug information to identify troubles + with ABI symbols found at the local machine that + weren't found on ABI documentation (used only for + undefined subcommand) +16 - enable debug for what to regex conversion +32 - enable debug for symbol regex subgroups +64 - enable debug for sysfs graph tree variable +128 - enable debug of search groups +256 - enable displaying refrence tree graphs for undefined symbols. +""" diff --git a/tools/lib/python/abi/system_symbols.py b/tools/lib/python/abi/system_symbols.py new file mode 100644 index 000000000..7bbefd274 --- /dev/null +++ b/tools/lib/python/abi/system_symbols.py @@ -0,0 +1,378 @@ +#!/usr/bin/env python3 +# pylint: disable=R0902,R0912,R0914,R0915,R1702 +# Copyright(c) 2025: Mauro Carvalho Chehab . +# SPDX-License-Identifier: GPL-2.0 + +""" +Parse ABI documentation and produce results from it. +""" + +import os +import re +import sys + +from concurrent import futures +from datetime import datetime +from random import shuffle + +from abi.helpers import AbiDebug + +class SystemSymbols: + """Stores arguments for the class and initialize class vars.""" + + def graph_add_file(self, path, link=None): + """ + add a file path to the sysfs graph stored at self.root. + """ + + if path in self.files: + return + + name = "" + ref = self.root + for edge in path.split("/"): + name += edge + "/" + if edge not in ref: + ref[edge] = {"__name": [name.rstrip("/")]} + + ref = ref[edge] + + if link and link not in ref["__name"]: + ref["__name"].append(link.rstrip("/")) + + self.files.add(path) + + def print_graph(self, root_prefix="", root=None, level=0): + """Prints a reference tree graph using UTF-8 characters.""" + + if not root: + root = self.root + level = 0 + + # Prevent endless traverse + if level > 5: + return + + if level > 0: + prefix = "├──" + last_prefix = "└──" + else: + prefix = "" + last_prefix = "" + + items = list(root.items()) + + names = root.get("__name", []) + for k, edge in items: + if k == "__name": + continue + + if not k: + k = "/" + + if len(names) > 1: + k += " links: " + ",".join(names[1:]) + + if edge == items[-1][1]: + print(root_prefix + last_prefix + k) + p = root_prefix + if level > 0: + p += " " + self.print_graph(p, edge, level + 1) + else: + print(root_prefix + prefix + k) + p = root_prefix + "│ " + self.print_graph(p, edge, level + 1) + + def _walk(self, root): + """ + Walk through sysfs to get all devnodes that aren't ignored. + + By default, uses /sys as sysfs mounting point. If another + directory is used, it replaces them to /sys at the patches. + """ + + with os.scandir(root) as obj: + for entry in obj: + path = os.path.join(root, entry.name) + if self.sysfs: + p = path.replace(self.sysfs, "/sys", count=1) + else: + p = path + + if self.re_ignore.search(p): + return + + # Handle link first to avoid directory recursion + if entry.is_symlink(): + real = os.path.realpath(path) + if not self.sysfs: + self.aliases[path] = real + else: + real = real.replace(self.sysfs, "/sys", count=1) + + # Add absfile location to graph if it doesn't exist + if not self.re_ignore.search(real): + # Add link to the graph + self.graph_add_file(real, p) + + elif entry.is_file(): + self.graph_add_file(p) + + elif entry.is_dir(): + self._walk(path) + + def __init__(self, abi, sysfs="/sys", hints=False): + """ + Initialize internal variables and get a list of all files inside + sysfs that can currently be parsed. + + Please notice that there are several entries on sysfs that aren't + documented as ABI. Ignore those. + + The real paths will be stored under self.files. Aliases will be + stored in separate, as self.aliases. + """ + + self.abi = abi + self.log = abi.log + + if sysfs != "/sys": + self.sysfs = sysfs.rstrip("/") + else: + self.sysfs = None + + self.hints = hints + + self.root = {} + self.aliases = {} + self.files = set() + + dont_walk = [ + # Those require root access and aren't documented at ABI + f"^{sysfs}/kernel/debug", + f"^{sysfs}/kernel/tracing", + f"^{sysfs}/fs/pstore", + f"^{sysfs}/fs/bpf", + f"^{sysfs}/fs/fuse", + + # This is not documented at ABI + f"^{sysfs}/module", + + f"^{sysfs}/fs/cgroup", # this is big and has zero docs under ABI + f"^{sysfs}/firmware", # documented elsewhere: ACPI, DT bindings + "sections|notes", # aren't actually part of ABI + + # kernel-parameters.txt - not easy to parse + "parameters", + ] + + self.re_ignore = re.compile("|".join(dont_walk)) + + print(f"Reading {sysfs} directory contents...", file=sys.stderr) + self._walk(sysfs) + + def check_file(self, refs, found): + """Check missing ABI symbols for a given sysfs file.""" + + res_list = [] + + try: + for names in refs: + fname = names[0] + + res = { + "found": False, + "fname": fname, + "msg": "", + } + res_list.append(res) + + re_what = self.abi.get_regexes(fname) + if not re_what: + self.abi.log.warning(f"missing rules for {fname}") + continue + + for name in names: + for r in re_what: + if self.abi.debug & AbiDebug.UNDEFINED: + self.log.debug("check if %s matches '%s'", name, r.pattern) + if r.match(name): + res["found"] = True + if found: + res["msg"] += f" {fname}: regex:\n\t" + continue + + if self.hints and not res["found"]: + res["msg"] += f" {fname} not found. Tested regexes:\n" + for r in re_what: + res["msg"] += " " + r.pattern + "\n" + + except KeyboardInterrupt: + pass + + return res_list + + def _ref_interactor(self, root): + """Recursive function to interact over the sysfs tree.""" + + for k, v in root.items(): + if isinstance(v, dict): + yield from self._ref_interactor(v) + + if root == self.root or k == "__name": + continue + + if self.abi.re_string: + fname = v["__name"][0] + if self.abi.re_string.search(fname): + yield v + else: + yield v + + + def get_fileref(self, all_refs, chunk_size): + """Interactor to group refs into chunks.""" + + n = 0 + refs = [] + + for ref in all_refs: + refs.append(ref) + + n += 1 + if n >= chunk_size: + yield refs + n = 0 + refs = [] + + yield refs + + def check_undefined_symbols(self, max_workers=None, chunk_size=50, + found=None, dry_run=None): + """Seach ABI for sysfs symbols missing documentation.""" + + self.abi.parse_abi() + + if self.abi.debug & AbiDebug.GRAPH: + self.print_graph() + + all_refs = [] + for ref in self._ref_interactor(self.root): + all_refs.append(ref["__name"]) + + if dry_run: + print("Would check", file=sys.stderr) + for ref in all_refs: + print(", ".join(ref)) + + return + + print("Starting to search symbols (it may take several minutes):", + file=sys.stderr) + start = datetime.now() + old_elapsed = None + + # Python doesn't support multithreading due to limitations on its + # global lock (GIL). While Python 3.13 finally made GIL optional, + # there are still issues related to it. Also, we want to have + # backward compatibility with older versions of Python. + # + # So, use instead multiprocess. However, Python is very slow passing + # data from/to multiple processes. Also, it may consume lots of memory + # if the data to be shared is not small. So, we need to group workload + # in chunks that are big enough to generate performance gains while + # not being so big that would cause out-of-memory. + + num_refs = len(all_refs) + print(f"Number of references to parse: {num_refs}", file=sys.stderr) + + if not max_workers: + max_workers = os.cpu_count() + elif max_workers > os.cpu_count(): + max_workers = os.cpu_count() + + max_workers = max(max_workers, 1) + + max_chunk_size = int((num_refs + max_workers - 1) / max_workers) + chunk_size = min(chunk_size, max_chunk_size) + chunk_size = max(1, chunk_size) + + if max_workers > 1: + executor = futures.ProcessPoolExecutor + + # Place references in a random order. This may help improving + # performance, by mixing complex/simple expressions when creating + # chunks + shuffle(all_refs) + else: + # Python has a high overhead with processes. When there's just + # one worker, it is faster to not create a new process. + # Yet, User still deserves to have a progress print. So, use + # python's "thread", which is actually a single process, using + # an internal schedule to switch between tasks. No performance + # gains for non-IO tasks, but still it can be quickly interrupted + # from time to time to display progress. + executor = futures.ThreadPoolExecutor + + not_found = [] + f_list = [] + with executor(max_workers=max_workers) as exe: + for refs in self.get_fileref(all_refs, chunk_size): + if refs: + try: + f_list.append(exe.submit(self.check_file, refs, found)) + + except KeyboardInterrupt: + return + + total = len(f_list) + + if not total: + if self.abi.re_string: + print(f"No ABI symbol matches {self.abi.search_string}") + else: + self.abi.log.warning("No ABI symbols found") + return + + print(f"{len(f_list):6d} jobs queued on {max_workers} workers", + file=sys.stderr) + + while f_list: + try: + t = futures.wait(f_list, timeout=1, + return_when=futures.FIRST_COMPLETED) + + done = t[0] + + for fut in done: + res_list = fut.result() + + for res in res_list: + if not res["found"]: + not_found.append(res["fname"]) + if res["msg"]: + print(res["msg"]) + + f_list.remove(fut) + except KeyboardInterrupt: + return + + except RuntimeError as e: + self.abi.log.warning(f"Future: {e}") + break + + if sys.stderr.isatty(): + elapsed = str(datetime.now() - start).split(".", maxsplit=1)[0] + if len(f_list) < total: + elapsed += f" ({total - len(f_list)}/{total} jobs completed). " + if elapsed != old_elapsed: + print(elapsed + "\r", end="", flush=True, + file=sys.stderr) + old_elapsed = elapsed + + elapsed = str(datetime.now() - start).split(".", maxsplit=1)[0] + print(elapsed, file=sys.stderr) + + for f in sorted(not_found): + print(f"{f} not found.") -- cgit v1.3.1