summaryrefslogtreecommitdiffstats
path: root/libnetdata/procfile
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--libnetdata/procfile/Makefile.am8
-rw-r--r--libnetdata/procfile/README.md67
-rw-r--r--libnetdata/procfile/procfile.c479
-rw-r--r--libnetdata/procfile/procfile.h108
4 files changed, 662 insertions, 0 deletions
diff --git a/libnetdata/procfile/Makefile.am b/libnetdata/procfile/Makefile.am
new file mode 100644
index 0000000..161784b
--- /dev/null
+++ b/libnetdata/procfile/Makefile.am
@@ -0,0 +1,8 @@
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+AUTOMAKE_OPTIONS = subdir-objects
+MAINTAINERCLEANFILES = $(srcdir)/Makefile.in
+
+dist_noinst_DATA = \
+ README.md \
+ $(NULL)
diff --git a/libnetdata/procfile/README.md b/libnetdata/procfile/README.md
new file mode 100644
index 0000000..6563803
--- /dev/null
+++ b/libnetdata/procfile/README.md
@@ -0,0 +1,67 @@
+<!--
+title: "PROCFILE"
+custom_edit_url: https://github.com/netdata/netdata/edit/master/libnetdata/procfile/README.md
+-->
+
+# PROCFILE
+
+procfile is a library for reading text data files (i.e `/proc` files) in the fastest possible way.
+
+## How it works
+
+The library automatically adapts (through the iterations) its memory so that each file
+is read with single `read()` call.
+
+Then the library splits the file into words, using the supplied separators.
+The library also supported quoted words (i.e. strings within of which the separators are ignored).
+
+### Initialization
+
+Initially the caller:
+
+- calls `procfile_open()` to open the file and allocate the structures needed.
+
+### Iterations
+
+For each iteration, the caller:
+
+- calls `procfile_readall()` to read updated contents.
+ This call also rewinds (`lseek()` to 0) before reading it.
+
+ For every file, a [BUFFER](/libnetdata/buffer/README.md) is used that is automatically adjusted to fit the entire
+ file contents of the file. So the file is read with a single `read()` call (providing atomicity / consistency when
+ the data are read from the kernel).
+
+ Once the data are read, 2 arrays of pointers are updated:
+
+ - a `words` array, pointing to each word in the data read
+ - a `lines` array, pointing to the first word for each line
+
+ This is highly optimized. Both arrays are automatically adjusted to
+ fit all contents and are updated in a single pass on the data.
+
+ The library provides a number of macros:
+
+ - `procfile_lines()` returns the # of lines read
+ - `procfile_linewords()` returns the # of words in the given line
+ - `procfile_word()` returns a pointer the given word #
+ - `procfile_line()` returns a pointer to the first word of the given line #
+ - `procfile_lineword()` returns a pointer to the given word # of the given line #
+
+### Cleanup
+
+When the caller exits:
+
+- calls `procfile_free()` to close the file and free all memory used.
+
+### Performance
+
+- a **raspberry Pi 1** (the oldest single core one) can process 5.000+ `/proc` files per second.
+- a **J1900 Celeron** processor can process 23.000+ `/proc` files per second per core.
+
+To achieve this kind of performance, the library tries to work in batches so that the code
+and the data are inside the processor's caches.
+
+This library is extensively used in Netdata and its plugins.
+
+
diff --git a/libnetdata/procfile/procfile.c b/libnetdata/procfile/procfile.c
new file mode 100644
index 0000000..eb04316
--- /dev/null
+++ b/libnetdata/procfile/procfile.c
@@ -0,0 +1,479 @@
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "../libnetdata.h"
+
+#define PF_PREFIX "PROCFILE"
+
+#define PFWORDS_INCREASE_STEP 2000
+#define PFLINES_INCREASE_STEP 200
+#define PROCFILE_INCREMENT_BUFFER 4096
+
+int procfile_open_flags = O_RDONLY;
+
+int procfile_adaptive_initial_allocation = 0;
+
+// if adaptive allocation is set, these store the
+// max values we have seen so far
+size_t procfile_max_lines = PFLINES_INCREASE_STEP;
+size_t procfile_max_words = PFWORDS_INCREASE_STEP;
+size_t procfile_max_allocation = PROCFILE_INCREMENT_BUFFER;
+
+
+// ----------------------------------------------------------------------------
+
+char *procfile_filename(procfile *ff) {
+ if(ff->filename)
+ return ff->filename;
+
+ char filename[FILENAME_MAX + 1];
+ char buffer[FILENAME_MAX + 1];
+ snprintfz(buffer, FILENAME_MAX, "/proc/self/fd/%d", ff->fd);
+
+ ssize_t l = readlink(buffer, filename, FILENAME_MAX);
+ if(unlikely(l == -1))
+ snprintfz(filename, FILENAME_MAX, "unknown filename for fd %d", ff->fd);
+ else
+ filename[l] = '\0';
+
+
+ ff->filename = strdupz(filename);
+
+ // on non-linux systems, something like this will be needed
+ // fcntl(ff->fd, F_GETPATH, ff->filename)
+
+ return ff->filename;
+}
+
+// ----------------------------------------------------------------------------
+// An array of words
+
+static inline void procfile_words_add(procfile *ff, char *str) {
+ // debug(D_PROCFILE, PF_PREFIX ": adding word No %d: '%s'", fw->len, str);
+
+ pfwords *fw = ff->words;
+ if(unlikely(fw->len == fw->size)) {
+ // debug(D_PROCFILE, PF_PREFIX ": expanding words");
+ size_t minimum = PFWORDS_INCREASE_STEP;
+ size_t optimal = fw->size / 2;
+ size_t wanted = (optimal > minimum)?optimal:minimum;
+
+ ff->words = fw = reallocz(fw, sizeof(pfwords) + (fw->size + wanted) * sizeof(char *));
+ fw->size += wanted;
+ }
+
+ fw->words[fw->len++] = str;
+}
+
+NEVERNULL
+static inline pfwords *procfile_words_create(void) {
+ // debug(D_PROCFILE, PF_PREFIX ": initializing words");
+
+ size_t size = (procfile_adaptive_initial_allocation) ? procfile_max_words : PFWORDS_INCREASE_STEP;
+
+ pfwords *new = mallocz(sizeof(pfwords) + size * sizeof(char *));
+ new->len = 0;
+ new->size = size;
+ return new;
+}
+
+static inline void procfile_words_reset(pfwords *fw) {
+ // debug(D_PROCFILE, PF_PREFIX ": resetting words");
+ fw->len = 0;
+}
+
+static inline void procfile_words_free(pfwords *fw) {
+ // debug(D_PROCFILE, PF_PREFIX ": freeing words");
+
+ freez(fw);
+}
+
+
+// ----------------------------------------------------------------------------
+// An array of lines
+
+NEVERNULL
+static inline size_t *procfile_lines_add(procfile *ff) {
+ // debug(D_PROCFILE, PF_PREFIX ": adding line %d at word %d", fl->len, first_word);
+
+ pflines *fl = ff->lines;
+ if(unlikely(fl->len == fl->size)) {
+ // debug(D_PROCFILE, PF_PREFIX ": expanding lines");
+ size_t minimum = PFLINES_INCREASE_STEP;
+ size_t optimal = fl->size / 2;
+ size_t wanted = (optimal > minimum)?optimal:minimum;
+
+ ff->lines = fl = reallocz(fl, sizeof(pflines) + (fl->size + wanted) * sizeof(ffline));
+ fl->size += wanted;
+ }
+
+ ffline *ffl = &fl->lines[fl->len++];
+ ffl->words = 0;
+ ffl->first = ff->words->len;
+
+ return &ffl->words;
+}
+
+NEVERNULL
+static inline pflines *procfile_lines_create(void) {
+ // debug(D_PROCFILE, PF_PREFIX ": initializing lines");
+
+ size_t size = (unlikely(procfile_adaptive_initial_allocation)) ? procfile_max_words : PFLINES_INCREASE_STEP;
+
+ pflines *new = mallocz(sizeof(pflines) + size * sizeof(ffline));
+ new->len = 0;
+ new->size = size;
+ return new;
+}
+
+static inline void procfile_lines_reset(pflines *fl) {
+ // debug(D_PROCFILE, PF_PREFIX ": resetting lines");
+
+ fl->len = 0;
+}
+
+static inline void procfile_lines_free(pflines *fl) {
+ // debug(D_PROCFILE, PF_PREFIX ": freeing lines");
+
+ freez(fl);
+}
+
+
+// ----------------------------------------------------------------------------
+// The procfile
+
+void procfile_close(procfile *ff) {
+ if(unlikely(!ff)) return;
+
+ debug(D_PROCFILE, PF_PREFIX ": Closing file '%s'", procfile_filename(ff));
+
+ freez(ff->filename);
+ procfile_lines_free(ff->lines);
+ procfile_words_free(ff->words);
+
+ if(likely(ff->fd != -1)) close(ff->fd);
+ freez(ff);
+}
+
+NOINLINE
+static void procfile_parser(procfile *ff) {
+ // debug(D_PROCFILE, PF_PREFIX ": Parsing file '%s'", ff->filename);
+
+ char *s = ff->data // our current position
+ , *e = &ff->data[ff->len] // the terminating null
+ , *t = ff->data; // the first character of a word (or quoted / parenthesized string)
+
+ // the look up array to find our type of character
+ PF_CHAR_TYPE *separators = ff->separators;
+
+ char quote = 0; // the quote character - only when in quoted string
+ size_t opened = 0; // counts the number of open parenthesis
+
+ size_t *line_words = procfile_lines_add(ff);
+
+ while(s < e) {
+ PF_CHAR_TYPE ct = separators[(unsigned char)(*s)];
+
+ // this is faster than a switch()
+ // read more here: http://lazarenko.me/switch/
+ if(likely(ct == PF_CHAR_IS_WORD)) {
+ s++;
+ }
+ else if(likely(ct == PF_CHAR_IS_SEPARATOR)) {
+ if(!quote && !opened) {
+ if (s != t) {
+ // separator, but we have word before it
+ *s = '\0';
+ procfile_words_add(ff, t);
+ (*line_words)++;
+ t = ++s;
+ }
+ else {
+ // separator at the beginning
+ // skip it
+ t = ++s;
+ }
+ }
+ else {
+ // we are inside a quote or parenthesized string
+ s++;
+ }
+ }
+ else if(likely(ct == PF_CHAR_IS_NEWLINE)) {
+ // end of line
+
+ *s = '\0';
+ procfile_words_add(ff, t);
+ (*line_words)++;
+ t = ++s;
+
+ // debug(D_PROCFILE, PF_PREFIX ": ended line %d with %d words", l, ff->lines->lines[l].words);
+
+ line_words = procfile_lines_add(ff);
+ }
+ else if(likely(ct == PF_CHAR_IS_QUOTE)) {
+ if(unlikely(!quote && s == t)) {
+ // quote opened at the beginning
+ quote = *s;
+ t = ++s;
+ }
+ else if(unlikely(quote && quote == *s)) {
+ // quote closed
+ quote = 0;
+
+ *s = '\0';
+ procfile_words_add(ff, t);
+ (*line_words)++;
+ t = ++s;
+ }
+ else
+ s++;
+ }
+ else if(likely(ct == PF_CHAR_IS_OPEN)) {
+ if(s == t) {
+ opened++;
+ t = ++s;
+ }
+ else if(opened) {
+ opened++;
+ s++;
+ }
+ else
+ s++;
+ }
+ else if(likely(ct == PF_CHAR_IS_CLOSE)) {
+ if(opened) {
+ opened--;
+
+ if(!opened) {
+ *s = '\0';
+ procfile_words_add(ff, t);
+ (*line_words)++;
+ t = ++s;
+ }
+ else
+ s++;
+ }
+ else
+ s++;
+ }
+ else
+ fatal("Internal Error: procfile_readall() does not handle all the cases.");
+ }
+
+ if(likely(s > t && t < e)) {
+ // the last word
+ if(unlikely(ff->len >= ff->size)) {
+ // we are going to loose the last byte
+ s = &ff->data[ff->size - 1];
+ }
+
+ *s = '\0';
+ procfile_words_add(ff, t);
+ (*line_words)++;
+ // t = ++s;
+ }
+}
+
+procfile *procfile_readall(procfile *ff) {
+ // debug(D_PROCFILE, PF_PREFIX ": Reading file '%s'.", ff->filename);
+
+ ff->len = 0; // zero the used size
+ ssize_t r = 1; // read at least once
+ while(r > 0) {
+ ssize_t s = ff->len;
+ ssize_t x = ff->size - s;
+
+ if(unlikely(!x)) {
+ size_t minimum = PROCFILE_INCREMENT_BUFFER;
+ size_t optimal = ff->size / 2;
+ size_t wanted = (optimal > minimum)?optimal:minimum;
+
+ debug(D_PROCFILE, PF_PREFIX ": Expanding data buffer for file '%s' by %zu bytes.", procfile_filename(ff), wanted);
+ ff = reallocz(ff, sizeof(procfile) + ff->size + wanted);
+ ff->size += wanted;
+ }
+
+ debug(D_PROCFILE, "Reading file '%s', from position %zd with length %zd", procfile_filename(ff), s, (ssize_t)(ff->size - s));
+ r = read(ff->fd, &ff->data[s], ff->size - s);
+ if(unlikely(r == -1)) {
+ if(unlikely(!(ff->flags & PROCFILE_FLAG_NO_ERROR_ON_FILE_IO))) error(PF_PREFIX ": Cannot read from file '%s' on fd %d", procfile_filename(ff), ff->fd);
+ procfile_close(ff);
+ return NULL;
+ }
+
+ ff->len += r;
+ }
+
+ // debug(D_PROCFILE, "Rewinding file '%s'", ff->filename);
+ if(unlikely(lseek(ff->fd, 0, SEEK_SET) == -1)) {
+ if(unlikely(!(ff->flags & PROCFILE_FLAG_NO_ERROR_ON_FILE_IO))) error(PF_PREFIX ": Cannot rewind on file '%s'.", procfile_filename(ff));
+ procfile_close(ff);
+ return NULL;
+ }
+
+ procfile_lines_reset(ff->lines);
+ procfile_words_reset(ff->words);
+ procfile_parser(ff);
+
+ if(unlikely(procfile_adaptive_initial_allocation)) {
+ if(unlikely(ff->len > procfile_max_allocation)) procfile_max_allocation = ff->len;
+ if(unlikely(ff->lines->len > procfile_max_lines)) procfile_max_lines = ff->lines->len;
+ if(unlikely(ff->words->len > procfile_max_words)) procfile_max_words = ff->words->len;
+ }
+
+ // debug(D_PROCFILE, "File '%s' updated.", ff->filename);
+ return ff;
+}
+
+static PF_CHAR_TYPE procfile_default_separators[256];
+__attribute__((constructor)) void procfile_initialize_default_separators(void) {
+ int i = 256;
+ while(i--) {
+ if(unlikely(i == '\n' || i == '\r'))
+ procfile_default_separators[i] = PF_CHAR_IS_NEWLINE;
+
+ else if(unlikely(isspace(i) || !isprint(i)))
+ procfile_default_separators[i] = PF_CHAR_IS_SEPARATOR;
+
+ else
+ procfile_default_separators[i] = PF_CHAR_IS_WORD;
+ }
+}
+
+NOINLINE
+static void procfile_set_separators(procfile *ff, const char *separators) {
+ // set the separators
+ if(unlikely(!separators))
+ separators = " \t=|";
+
+ // copy the default
+ memcpy(ff->separators, procfile_default_separators, 256 * sizeof(PF_CHAR_TYPE));
+
+ PF_CHAR_TYPE *ffs = ff->separators;
+ const char *s = separators;
+ while(*s)
+ ffs[(int)*s++] = PF_CHAR_IS_SEPARATOR;
+}
+
+void procfile_set_quotes(procfile *ff, const char *quotes) {
+ PF_CHAR_TYPE *ffs = ff->separators;
+
+ // remove all quotes
+ int i = 256;
+ while(i--)
+ if(unlikely(ffs[i] == PF_CHAR_IS_QUOTE))
+ ffs[i] = PF_CHAR_IS_WORD;
+
+ // if nothing given, return
+ if(unlikely(!quotes || !*quotes))
+ return;
+
+ // set the quotes
+ const char *s = quotes;
+ while(*s)
+ ffs[(int)*s++] = PF_CHAR_IS_QUOTE;
+}
+
+void procfile_set_open_close(procfile *ff, const char *open, const char *close) {
+ PF_CHAR_TYPE *ffs = ff->separators;
+
+ // remove all open/close
+ int i = 256;
+ while(i--)
+ if(unlikely(ffs[i] == PF_CHAR_IS_OPEN || ffs[i] == PF_CHAR_IS_CLOSE))
+ ffs[i] = PF_CHAR_IS_WORD;
+
+ // if nothing given, return
+ if(unlikely(!open || !*open || !close || !*close))
+ return;
+
+ // set the openings
+ const char *s = open;
+ while(*s)
+ ffs[(int)*s++] = PF_CHAR_IS_OPEN;
+
+ // set the closings
+ s = close;
+ while(*s)
+ ffs[(int)*s++] = PF_CHAR_IS_CLOSE;
+}
+
+procfile *procfile_open(const char *filename, const char *separators, uint32_t flags) {
+ debug(D_PROCFILE, PF_PREFIX ": Opening file '%s'", filename);
+
+ int fd = open(filename, procfile_open_flags, 0666);
+ if(unlikely(fd == -1)) {
+ if(unlikely(!(flags & PROCFILE_FLAG_NO_ERROR_ON_FILE_IO))) error(PF_PREFIX ": Cannot open file '%s'", filename);
+ return NULL;
+ }
+
+ // info("PROCFILE: opened '%s' on fd %d", filename, fd);
+
+ size_t size = (unlikely(procfile_adaptive_initial_allocation)) ? procfile_max_allocation : PROCFILE_INCREMENT_BUFFER;
+ procfile *ff = mallocz(sizeof(procfile) + size);
+
+ //strncpyz(ff->filename, filename, FILENAME_MAX);
+ ff->filename = NULL;
+ ff->fd = fd;
+ ff->size = size;
+ ff->len = 0;
+ ff->flags = flags;
+
+ ff->lines = procfile_lines_create();
+ ff->words = procfile_words_create();
+
+ procfile_set_separators(ff, separators);
+
+ debug(D_PROCFILE, "File '%s' opened.", filename);
+ return ff;
+}
+
+procfile *procfile_reopen(procfile *ff, const char *filename, const char *separators, uint32_t flags) {
+ if(unlikely(!ff)) return procfile_open(filename, separators, flags);
+
+ if(likely(ff->fd != -1)) {
+ // info("PROCFILE: closing fd %d", ff->fd);
+ close(ff->fd);
+ }
+
+ ff->fd = open(filename, procfile_open_flags, 0666);
+ if(unlikely(ff->fd == -1)) {
+ procfile_close(ff);
+ return NULL;
+ }
+
+ // info("PROCFILE: opened '%s' on fd %d", filename, ff->fd);
+
+ //strncpyz(ff->filename, filename, FILENAME_MAX);
+ freez(ff->filename);
+ ff->filename = NULL;
+ ff->flags = flags;
+
+ // do not do the separators again if NULL is given
+ if(likely(separators)) procfile_set_separators(ff, separators);
+
+ return ff;
+}
+
+// ----------------------------------------------------------------------------
+// example parsing of procfile data
+
+void procfile_print(procfile *ff) {
+ size_t lines = procfile_lines(ff), l;
+ char *s;
+ (void)s;
+
+ debug(D_PROCFILE, "File '%s' with %zu lines and %zu words", procfile_filename(ff), ff->lines->len, ff->words->len);
+
+ for(l = 0; likely(l < lines) ;l++) {
+ size_t words = procfile_linewords(ff, l);
+
+ debug(D_PROCFILE, " line %zu starts at word %zu and has %zu words", l, ff->lines->lines[l].first, ff->lines->lines[l].words);
+
+ size_t w;
+ for(w = 0; likely(w < words) ;w++) {
+ s = procfile_lineword(ff, l, w);
+ debug(D_PROCFILE, " [%zu.%zu] '%s'", l, w, s);
+ }
+ }
+}
diff --git a/libnetdata/procfile/procfile.h b/libnetdata/procfile/procfile.h
new file mode 100644
index 0000000..cae4ad4
--- /dev/null
+++ b/libnetdata/procfile/procfile.h
@@ -0,0 +1,108 @@
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#ifndef NETDATA_PROCFILE_H
+#define NETDATA_PROCFILE_H 1
+
+#include "../libnetdata.h"
+
+// ----------------------------------------------------------------------------
+// An array of words
+
+typedef struct {
+ size_t len; // used entries
+ size_t size; // capacity
+ char *words[]; // array of pointers
+} pfwords;
+
+
+// ----------------------------------------------------------------------------
+// An array of lines
+
+typedef struct {
+ size_t words; // how many words this line has
+ size_t first; // the id of the first word of this line
+ // in the words array
+} ffline;
+
+typedef struct {
+ size_t len; // used entries
+ size_t size; // capacity
+ ffline lines[]; // array of lines
+} pflines;
+
+
+// ----------------------------------------------------------------------------
+// The procfile
+
+#define PROCFILE_FLAG_DEFAULT 0x00000000
+#define PROCFILE_FLAG_NO_ERROR_ON_FILE_IO 0x00000001
+
+typedef enum __attribute__ ((__packed__)) procfile_separator {
+ PF_CHAR_IS_SEPARATOR,
+ PF_CHAR_IS_NEWLINE,
+ PF_CHAR_IS_WORD,
+ PF_CHAR_IS_QUOTE,
+ PF_CHAR_IS_OPEN,
+ PF_CHAR_IS_CLOSE
+} PF_CHAR_TYPE;
+
+typedef struct procfile {
+ char *filename; // not populated until procfile_filename() is called
+ uint32_t flags;
+ int fd; // the file descriptor
+ size_t len; // the bytes we have placed into data
+ size_t size; // the bytes we have allocated for data
+ pflines *lines;
+ pfwords *words;
+ PF_CHAR_TYPE separators[256];
+ char data[]; // allocated buffer to keep file contents
+} procfile;
+
+// close the proc file and free all related memory
+void procfile_close(procfile *ff);
+
+// (re)read and parse the proc file
+procfile *procfile_readall(procfile *ff);
+
+// open a /proc or /sys file
+procfile *procfile_open(const char *filename, const char *separators, uint32_t flags);
+
+// re-open a file
+// if separators == NULL, the last separators are used
+procfile *procfile_reopen(procfile *ff, const char *filename, const char *separators, uint32_t flags);
+
+// example walk-through a procfile parsed file
+void procfile_print(procfile *ff);
+
+void procfile_set_quotes(procfile *ff, const char *quotes);
+void procfile_set_open_close(procfile *ff, const char *open, const char *close);
+
+char *procfile_filename(procfile *ff);
+
+// ----------------------------------------------------------------------------
+
+// set to the O_XXXX flags, to have procfile_open and procfile_reopen use them when opening proc files
+extern int procfile_open_flags;
+
+// set this to 1, to have procfile adapt its initial buffer allocation to the max allocation used so far
+extern int procfile_adaptive_initial_allocation;
+
+// return the number of lines present
+#define procfile_lines(ff) ((ff)->lines->len)
+
+// return the number of words of the Nth line
+#define procfile_linewords(ff, line) (((line) < procfile_lines(ff)) ? (ff)->lines->lines[(line)].words : 0)
+
+// return the Nth word of the file, or empty string
+#define procfile_word(ff, word) (((word) < (ff)->words->len) ? (ff)->words->words[(word)] : "")
+
+// return the first word of the Nth line, or empty string
+#define procfile_line(ff, line) (((line) < procfile_lines(ff)) ? procfile_word((ff), (ff)->lines->lines[(line)].first) : "")
+
+// return the Nth word of the current line
+#define procfile_lineword(ff, line, word) (((line) < procfile_lines(ff) && (word) < procfile_linewords((ff), (line))) ? procfile_word((ff), (ff)->lines->lines[(line)].first + (word)) : "")
+
+// Open file without logging file IO error if any
+#define procfile_open_no_log(filename, separators, flags) procfile_open(filename, separators, flags | PROCFILE_FLAG_NO_ERROR_ON_FILE_IO)
+
+#endif /* NETDATA_PROCFILE_H */
generated by cgit v1.2.3 (git 2.39.1) at 2024-11-18 12:10:23 +0000