linux/tools/perf/util/dso.c

1158 lines
25 KiB
C
Raw Normal View History

#include <asm/bug.h>
#include <sys/time.h>
#include <sys/resource.h>
#include "symbol.h"
#include "dso.h"
#include "machine.h"
#include "util.h"
#include "debug.h"
char dso__symtab_origin(const struct dso *dso)
{
static const char origin[] = {
[DSO_BINARY_TYPE__KALLSYMS] = 'k',
[DSO_BINARY_TYPE__VMLINUX] = 'v',
[DSO_BINARY_TYPE__JAVA_JIT] = 'j',
[DSO_BINARY_TYPE__DEBUGLINK] = 'l',
[DSO_BINARY_TYPE__BUILD_ID_CACHE] = 'B',
[DSO_BINARY_TYPE__FEDORA_DEBUGINFO] = 'f',
[DSO_BINARY_TYPE__UBUNTU_DEBUGINFO] = 'u',
[DSO_BINARY_TYPE__OPENEMBEDDED_DEBUGINFO] = 'o',
[DSO_BINARY_TYPE__BUILDID_DEBUGINFO] = 'b',
[DSO_BINARY_TYPE__SYSTEM_PATH_DSO] = 'd',
[DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE] = 'K',
[DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE_COMP] = 'm',
[DSO_BINARY_TYPE__GUEST_KALLSYMS] = 'g',
[DSO_BINARY_TYPE__GUEST_KMODULE] = 'G',
[DSO_BINARY_TYPE__GUEST_KMODULE_COMP] = 'M',
[DSO_BINARY_TYPE__GUEST_VMLINUX] = 'V',
};
if (dso == NULL || dso->symtab_type == DSO_BINARY_TYPE__NOT_FOUND)
return '!';
return origin[dso->symtab_type];
}
int dso__read_binary_type_filename(const struct dso *dso,
enum dso_binary_type type,
char *root_dir, char *filename, size_t size)
{
char build_id_hex[BUILD_ID_SIZE * 2 + 1];
int ret = 0;
size_t len;
switch (type) {
case DSO_BINARY_TYPE__DEBUGLINK: {
char *debuglink;
len = __symbol__join_symfs(filename, size, dso->long_name);
debuglink = filename + len;
while (debuglink != filename && *debuglink != '/')
debuglink--;
if (*debuglink == '/')
debuglink++;
ret = filename__read_debuglink(filename, debuglink,
size - (debuglink - filename));
}
break;
case DSO_BINARY_TYPE__BUILD_ID_CACHE:
/* skip the locally configured cache if a symfs is given */
if (symbol_conf.symfs[0] ||
(dso__build_id_filename(dso, filename, size) == NULL))
ret = -1;
break;
case DSO_BINARY_TYPE__FEDORA_DEBUGINFO:
len = __symbol__join_symfs(filename, size, "/usr/lib/debug");
snprintf(filename + len, size - len, "%s.debug", dso->long_name);
break;
case DSO_BINARY_TYPE__UBUNTU_DEBUGINFO:
len = __symbol__join_symfs(filename, size, "/usr/lib/debug");
snprintf(filename + len, size - len, "%s", dso->long_name);
break;
case DSO_BINARY_TYPE__OPENEMBEDDED_DEBUGINFO:
{
const char *last_slash;
size_t dir_size;
last_slash = dso->long_name + dso->long_name_len;
while (last_slash != dso->long_name && *last_slash != '/')
last_slash--;
len = __symbol__join_symfs(filename, size, "");
dir_size = last_slash - dso->long_name + 2;
if (dir_size > (size - len)) {
ret = -1;
break;
}
len += scnprintf(filename + len, dir_size, "%s", dso->long_name);
len += scnprintf(filename + len , size - len, ".debug%s",
last_slash);
break;
}
case DSO_BINARY_TYPE__BUILDID_DEBUGINFO:
if (!dso->has_build_id) {
ret = -1;
break;
}
build_id__sprintf(dso->build_id,
sizeof(dso->build_id),
build_id_hex);
len = __symbol__join_symfs(filename, size, "/usr/lib/debug/.build-id/");
snprintf(filename + len, size - len, "%.2s/%s.debug",
build_id_hex, build_id_hex + 2);
break;
case DSO_BINARY_TYPE__VMLINUX:
case DSO_BINARY_TYPE__GUEST_VMLINUX:
case DSO_BINARY_TYPE__SYSTEM_PATH_DSO:
__symbol__join_symfs(filename, size, dso->long_name);
break;
case DSO_BINARY_TYPE__GUEST_KMODULE:
case DSO_BINARY_TYPE__GUEST_KMODULE_COMP:
path__join3(filename, size, symbol_conf.symfs,
root_dir, dso->long_name);
break;
case DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE:
case DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE_COMP:
__symbol__join_symfs(filename, size, dso->long_name);
break;
case DSO_BINARY_TYPE__KCORE:
case DSO_BINARY_TYPE__GUEST_KCORE:
snprintf(filename, size, "%s", dso->long_name);
break;
default:
case DSO_BINARY_TYPE__KALLSYMS:
case DSO_BINARY_TYPE__GUEST_KALLSYMS:
case DSO_BINARY_TYPE__JAVA_JIT:
case DSO_BINARY_TYPE__NOT_FOUND:
ret = -1;
break;
}
return ret;
}
static const struct {
const char *fmt;
int (*decompress)(const char *input, int output);
} compressions[] = {
perf tools: Add gzip decompression support for kernel module Now my Archlinux box shows module symbols correctly. Before: $ perf report --stdio Failed to open /tmp/perf-3477.map, continuing without symbols no symbols found in /usr/bin/date, maybe install a debug package? No kallsyms or vmlinux with build-id 7b4ea0a49ae2111925857099aaf05c3246ff33e0 was found [drm] with build id 7b4ea0a49ae2111925857099aaf05c3246ff33e0 not found, continuing without symbols No kallsyms or vmlinux with build-id edd931629094b660ca9dec09a1b635c8d87aa2ee was found [jbd2] with build id edd931629094b660ca9dec09a1b635c8d87aa2ee not found, continuing without symbols No kallsyms or vmlinux with build-id a7b1eada671c34933e5610bb920b2ca4945a82c3 was found [ext4] with build id a7b1eada671c34933e5610bb920b2ca4945a82c3 not found, continuing without symbols No kallsyms or vmlinux with build-id d69511fa3e5840e770336ef45b06c83fef8d74e3 was found [scsi_mod] with build id d69511fa3e5840e770336ef45b06c83fef8d74e3 not found, continuing without symbols No kallsyms or vmlinux with build-id af0430af13461af058770ee9b87afc07922c2e77 was found [libata] with build id af0430af13461af058770ee9b87afc07922c2e77 not found, continuing without symbols No kallsyms or vmlinux with build-id aaeedff8160ce631a5f0333591c6ff291201d29f was found [libahci] with build id aaeedff8160ce631a5f0333591c6ff291201d29f not found, continuing without symbols No kallsyms or vmlinux with build-id c57907712becaf662dc4981824bb372c0441d605 was found [mac80211] with build id c57907712becaf662dc4981824bb372c0441d605 not found, continuing without symbols No kallsyms or vmlinux with build-id e0589077cc0ec8c3e4c40eb9f2d9e69d236bee8f was found [iwldvm] with build id e0589077cc0ec8c3e4c40eb9f2d9e69d236bee8f not found, continuing without symbols No kallsyms or vmlinux with build-id 2d86086bf136bf374a2f029cf85a48194f9b950b was found [cfg80211] with build id 2d86086bf136bf374a2f029cf85a48194f9b950b not found, continuing without symbols No kallsyms or vmlinux with build-id 4493c48599bdb3d91d0f8db5150e0be33fdd9221 was found [iwlwifi] with build id 4493c48599bdb3d91d0f8db5150e0be33fdd9221 not found, continuing without symbols ... # # Overhead Command Shared Object Symbol # ........ ............... ....................... ........................................................ # 0.03% swapper [ext4] [k] 0x000000000000fe2e 0.03% swapper [kernel.kallsyms] [k] account_entity_enqueue 0.03% swapper [ext4] [k] 0x000000000000fc2b 0.03% irq/50-iwlwifi [iwlwifi] [k] 0x000000000000200b 0.03% swapper [kernel.kallsyms] [k] ktime_add_safe 0.03% swapper [kernel.kallsyms] [k] elv_completed_request 0.03% swapper [libata] [k] 0x0000000000003997 0.03% swapper [libahci] [k] 0x0000000000001f25 0.03% swapper [kernel.kallsyms] [k] rb_next 0.03% swapper [kernel.kallsyms] [k] blk_finish_request 0.03% swapper [ext4] [k] 0x0000000000010248 0.00% perf [kernel.kallsyms] [k] native_write_msr_safe After: $ perf report --stdio Failed to open /tmp/perf-3477.map, continuing without symbols no symbols found in /usr/bin/tr, maybe install a debug package? ... # # Overhead Command Shared Object Symbol # ........ ............... ........................... ...................................................... # 0.04% kworker/u16:3 [ext4] [k] ext4_read_block_bitmap 0.03% kworker/u16:0 [mac80211] [k] ieee80211_sta_reset_beacon_monitor 0.02% irq/50-iwlwifi [mac80211] [k] ieee80211_get_bssid 0.02% firefox [e1000e] [k] __ew32_prepare 0.02% swapper [libahci] [k] ahci_handle_port_interrupt 0.02% emacs libglib-2.0.so.0.4000.0 [.] g_mutex_unlock 0.02% swapper [e1000e] [k] e1000_clean_tx_irq 0.02% dwm [kernel.kallsyms] [k] __schedule 0.02% gnome-terminal- [vdso] [.] __vdso_clock_gettime 0.02% swapper [e1000e] [k] e1000_alloc_rx_buffers 0.02% irq/50-iwlwifi [mac80211] [k] ieee80211_rx 0.01% firefox [vdso] [.] __vdso_gettimeofday 0.01% irq/50-iwlwifi [iwlwifi] [k] iwl_pcie_rxq_restock.part.13 Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/87h9yexshi.fsf@sejong.aot.lge.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-10-31 07:51:38 +00:00
#ifdef HAVE_ZLIB_SUPPORT
{ "gz", gzip_decompress_to_file },
#endif
#ifdef HAVE_LZMA_SUPPORT
{ "xz", lzma_decompress_to_file },
perf tools: Add gzip decompression support for kernel module Now my Archlinux box shows module symbols correctly. Before: $ perf report --stdio Failed to open /tmp/perf-3477.map, continuing without symbols no symbols found in /usr/bin/date, maybe install a debug package? No kallsyms or vmlinux with build-id 7b4ea0a49ae2111925857099aaf05c3246ff33e0 was found [drm] with build id 7b4ea0a49ae2111925857099aaf05c3246ff33e0 not found, continuing without symbols No kallsyms or vmlinux with build-id edd931629094b660ca9dec09a1b635c8d87aa2ee was found [jbd2] with build id edd931629094b660ca9dec09a1b635c8d87aa2ee not found, continuing without symbols No kallsyms or vmlinux with build-id a7b1eada671c34933e5610bb920b2ca4945a82c3 was found [ext4] with build id a7b1eada671c34933e5610bb920b2ca4945a82c3 not found, continuing without symbols No kallsyms or vmlinux with build-id d69511fa3e5840e770336ef45b06c83fef8d74e3 was found [scsi_mod] with build id d69511fa3e5840e770336ef45b06c83fef8d74e3 not found, continuing without symbols No kallsyms or vmlinux with build-id af0430af13461af058770ee9b87afc07922c2e77 was found [libata] with build id af0430af13461af058770ee9b87afc07922c2e77 not found, continuing without symbols No kallsyms or vmlinux with build-id aaeedff8160ce631a5f0333591c6ff291201d29f was found [libahci] with build id aaeedff8160ce631a5f0333591c6ff291201d29f not found, continuing without symbols No kallsyms or vmlinux with build-id c57907712becaf662dc4981824bb372c0441d605 was found [mac80211] with build id c57907712becaf662dc4981824bb372c0441d605 not found, continuing without symbols No kallsyms or vmlinux with build-id e0589077cc0ec8c3e4c40eb9f2d9e69d236bee8f was found [iwldvm] with build id e0589077cc0ec8c3e4c40eb9f2d9e69d236bee8f not found, continuing without symbols No kallsyms or vmlinux with build-id 2d86086bf136bf374a2f029cf85a48194f9b950b was found [cfg80211] with build id 2d86086bf136bf374a2f029cf85a48194f9b950b not found, continuing without symbols No kallsyms or vmlinux with build-id 4493c48599bdb3d91d0f8db5150e0be33fdd9221 was found [iwlwifi] with build id 4493c48599bdb3d91d0f8db5150e0be33fdd9221 not found, continuing without symbols ... # # Overhead Command Shared Object Symbol # ........ ............... ....................... ........................................................ # 0.03% swapper [ext4] [k] 0x000000000000fe2e 0.03% swapper [kernel.kallsyms] [k] account_entity_enqueue 0.03% swapper [ext4] [k] 0x000000000000fc2b 0.03% irq/50-iwlwifi [iwlwifi] [k] 0x000000000000200b 0.03% swapper [kernel.kallsyms] [k] ktime_add_safe 0.03% swapper [kernel.kallsyms] [k] elv_completed_request 0.03% swapper [libata] [k] 0x0000000000003997 0.03% swapper [libahci] [k] 0x0000000000001f25 0.03% swapper [kernel.kallsyms] [k] rb_next 0.03% swapper [kernel.kallsyms] [k] blk_finish_request 0.03% swapper [ext4] [k] 0x0000000000010248 0.00% perf [kernel.kallsyms] [k] native_write_msr_safe After: $ perf report --stdio Failed to open /tmp/perf-3477.map, continuing without symbols no symbols found in /usr/bin/tr, maybe install a debug package? ... # # Overhead Command Shared Object Symbol # ........ ............... ........................... ...................................................... # 0.04% kworker/u16:3 [ext4] [k] ext4_read_block_bitmap 0.03% kworker/u16:0 [mac80211] [k] ieee80211_sta_reset_beacon_monitor 0.02% irq/50-iwlwifi [mac80211] [k] ieee80211_get_bssid 0.02% firefox [e1000e] [k] __ew32_prepare 0.02% swapper [libahci] [k] ahci_handle_port_interrupt 0.02% emacs libglib-2.0.so.0.4000.0 [.] g_mutex_unlock 0.02% swapper [e1000e] [k] e1000_clean_tx_irq 0.02% dwm [kernel.kallsyms] [k] __schedule 0.02% gnome-terminal- [vdso] [.] __vdso_clock_gettime 0.02% swapper [e1000e] [k] e1000_alloc_rx_buffers 0.02% irq/50-iwlwifi [mac80211] [k] ieee80211_rx 0.01% firefox [vdso] [.] __vdso_gettimeofday 0.01% irq/50-iwlwifi [iwlwifi] [k] iwl_pcie_rxq_restock.part.13 Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/87h9yexshi.fsf@sejong.aot.lge.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-10-31 07:51:38 +00:00
#endif
{ NULL, NULL },
};
bool is_supported_compression(const char *ext)
{
unsigned i;
for (i = 0; compressions[i].fmt; i++) {
if (!strcmp(ext, compressions[i].fmt))
return true;
}
return false;
}
bool is_kmodule_extension(const char *ext)
{
if (strncmp(ext, "ko", 2))
return false;
if (ext[2] == '\0' || (ext[2] == '.' && is_supported_compression(ext+3)))
return true;
return false;
}
bool is_kernel_module(const char *pathname, bool *compressed)
{
const char *ext = strrchr(pathname, '.');
if (ext == NULL)
return false;
if (is_supported_compression(ext + 1)) {
if (compressed)
*compressed = true;
ext -= 3;
} else if (compressed)
*compressed = false;
return is_kmodule_extension(ext + 1);
}
bool decompress_to_file(const char *ext, const char *filename, int output_fd)
{
unsigned i;
for (i = 0; compressions[i].fmt; i++) {
if (!strcmp(ext, compressions[i].fmt))
return !compressions[i].decompress(filename,
output_fd);
}
return false;
}
bool dso__needs_decompress(struct dso *dso)
{
return dso->symtab_type == DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE_COMP ||
dso->symtab_type == DSO_BINARY_TYPE__GUEST_KMODULE_COMP;
}
perf tools: Add kmod_path__parse function Provides united way of parsing kernel module path into several components. The new kmod_path__parse function and few defines: int __kmod_path__parse(struct kmod_path *m, const char *path, bool alloc_name, bool alloc_ext); #define kmod_path__parse(__m, __p) __kmod_path__parse(__m, __p, false, false) #define kmod_path__parse_name(__m, __p) __kmod_path__parse(__m, __p, true , false) #define kmod_path__parse_ext(__m, __p) __kmod_path__parse(__m, __p, false, true) parse kernel module @path and updates @m argument like: @comp - true if @path contains supported compression suffix, false otherwise @kmod - true if @path contains '.ko' suffix in right position, false otherwise @name - if (@alloc_name && @kmod) is true, it contains strdup-ed base name of the kernel module without suffixes, otherwise strudup-ed base name of @path @ext - if (@alloc_ext && @comp) is true, it contains strdup-ed string the compression suffix It returns 0 if there's no strdup error, -ENOMEM otherwise. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-9t6eqg8j610r94l743hkntiv@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-02-05 14:40:25 +00:00
/*
* Parses kernel module specified in @path and updates
* @m argument like:
*
* @comp - true if @path contains supported compression suffix,
* false otherwise
* @kmod - true if @path contains '.ko' suffix in right position,
* false otherwise
* @name - if (@alloc_name && @kmod) is true, it contains strdup-ed base name
* of the kernel module without suffixes, otherwise strudup-ed
* base name of @path
* @ext - if (@alloc_ext && @comp) is true, it contains strdup-ed string
* the compression suffix
*
* Returns 0 if there's no strdup error, -ENOMEM otherwise.
*/
int __kmod_path__parse(struct kmod_path *m, const char *path,
bool alloc_name, bool alloc_ext)
{
const char *name = strrchr(path, '/');
const char *ext = strrchr(path, '.');
memset(m, 0x0, sizeof(*m));
name = name ? name + 1 : path;
/* No extension, just return name. */
if (ext == NULL) {
if (alloc_name) {
m->name = strdup(name);
return m->name ? 0 : -ENOMEM;
}
return 0;
}
if (is_supported_compression(ext + 1)) {
m->comp = true;
ext -= 3;
}
/* Check .ko extension only if there's enough name left. */
if (ext > name)
m->kmod = !strncmp(ext, ".ko", 3);
if (alloc_name) {
if (m->kmod) {
if (asprintf(&m->name, "[%.*s]", (int) (ext - name), name) == -1)
return -ENOMEM;
} else {
if (asprintf(&m->name, "%s", name) == -1)
return -ENOMEM;
}
strxfrchar(m->name, '-', '_');
}
if (alloc_ext && m->comp) {
m->ext = strdup(ext + 4);
if (!m->ext) {
free((void *) m->name);
return -ENOMEM;
}
}
return 0;
}
/*
* Global list of open DSOs and the counter.
*/
static LIST_HEAD(dso__data_open);
static long dso__data_open_cnt;
static void dso__list_add(struct dso *dso)
{
list_add_tail(&dso->data.open_entry, &dso__data_open);
dso__data_open_cnt++;
}
static void dso__list_del(struct dso *dso)
{
list_del(&dso->data.open_entry);
WARN_ONCE(dso__data_open_cnt <= 0,
"DSO data fd counter out of bounds.");
dso__data_open_cnt--;
}
static void close_first_dso(void);
static int do_open(char *name)
{
int fd;
char sbuf[STRERR_BUFSIZE];
do {
fd = open(name, O_RDONLY);
if (fd >= 0)
return fd;
pr_debug("dso open failed: %s\n",
strerror_r(errno, sbuf, sizeof(sbuf)));
if (!dso__data_open_cnt || errno != EMFILE)
break;
close_first_dso();
} while (1);
return -1;
}
static int __open_dso(struct dso *dso, struct machine *machine)
{
int fd;
char *root_dir = (char *)"";
char *name = malloc(PATH_MAX);
if (!name)
return -ENOMEM;
if (machine)
root_dir = machine->root_dir;
if (dso__read_binary_type_filename(dso, dso->binary_type,
root_dir, name, PATH_MAX)) {
free(name);
return -EINVAL;
}
fd = do_open(name);
free(name);
return fd;
}
static void check_data_close(void);
/**
* dso_close - Open DSO data file
* @dso: dso object
*
* Open @dso's data file descriptor and updates
* list/count of open DSO objects.
*/
static int open_dso(struct dso *dso, struct machine *machine)
{
int fd = __open_dso(dso, machine);
if (fd >= 0) {
dso__list_add(dso);
/*
* Check if we crossed the allowed number
* of opened DSOs and close one if needed.
*/
check_data_close();
}
return fd;
}
static void close_data_fd(struct dso *dso)
{
if (dso->data.fd >= 0) {
close(dso->data.fd);
dso->data.fd = -1;
dso->data.file_size = 0;
dso__list_del(dso);
}
}
/**
* dso_close - Close DSO data file
* @dso: dso object
*
* Close @dso's data file descriptor and updates
* list/count of open DSO objects.
*/
static void close_dso(struct dso *dso)
{
close_data_fd(dso);
}
static void close_first_dso(void)
{
struct dso *dso;
dso = list_first_entry(&dso__data_open, struct dso, data.open_entry);
close_dso(dso);
}
static rlim_t get_fd_limit(void)
{
struct rlimit l;
rlim_t limit = 0;
/* Allow half of the current open fd limit. */
if (getrlimit(RLIMIT_NOFILE, &l) == 0) {
if (l.rlim_cur == RLIM_INFINITY)
limit = l.rlim_cur;
else
limit = l.rlim_cur / 2;
} else {
pr_err("failed to get fd limit\n");
limit = 1;
}
return limit;
}
static bool may_cache_fd(void)
{
static rlim_t limit;
if (!limit)
limit = get_fd_limit();
if (limit == RLIM_INFINITY)
return true;
return limit > (rlim_t) dso__data_open_cnt;
}
/*
* Check and close LRU dso if we crossed allowed limit
* for opened dso file descriptors. The limit is half
* of the RLIMIT_NOFILE files opened.
*/
static void check_data_close(void)
{
bool cache_fd = may_cache_fd();
if (!cache_fd)
close_first_dso();
}
/**
* dso__data_close - Close DSO data file
* @dso: dso object
*
* External interface to close @dso's data file descriptor.
*/
void dso__data_close(struct dso *dso)
{
close_dso(dso);
}
/**
* dso__data_fd - Get dso's data file descriptor
* @dso: dso object
* @machine: machine object
*
* External interface to find dso's file, open it and
* returns file descriptor.
*/
int dso__data_fd(struct dso *dso, struct machine *machine)
{
enum dso_binary_type binary_type_data[] = {
DSO_BINARY_TYPE__BUILD_ID_CACHE,
DSO_BINARY_TYPE__SYSTEM_PATH_DSO,
DSO_BINARY_TYPE__NOT_FOUND,
};
int i = 0;
if (dso->data.status == DSO_DATA_STATUS_ERROR)
return -1;
if (dso->data.fd >= 0)
goto out;
if (dso->binary_type != DSO_BINARY_TYPE__NOT_FOUND) {
dso->data.fd = open_dso(dso, machine);
goto out;
}
do {
dso->binary_type = binary_type_data[i++];
dso->data.fd = open_dso(dso, machine);
if (dso->data.fd >= 0)
goto out;
} while (dso->binary_type != DSO_BINARY_TYPE__NOT_FOUND);
out:
if (dso->data.fd >= 0)
dso->data.status = DSO_DATA_STATUS_OK;
else
dso->data.status = DSO_DATA_STATUS_ERROR;
return dso->data.fd;
}
bool dso__data_status_seen(struct dso *dso, enum dso_data_status_seen by)
{
u32 flag = 1 << by;
if (dso->data.status_seen & flag)
return true;
dso->data.status_seen |= flag;
return false;
}
static void
dso_cache__free(struct rb_root *root)
{
struct rb_node *next = rb_first(root);
while (next) {
struct dso_cache *cache;
cache = rb_entry(next, struct dso_cache, rb_node);
next = rb_next(&cache->rb_node);
rb_erase(&cache->rb_node, root);
free(cache);
}
}
static struct dso_cache *dso_cache__find(const struct rb_root *root, u64 offset)
{
struct rb_node * const *p = &root->rb_node;
const struct rb_node *parent = NULL;
struct dso_cache *cache;
while (*p != NULL) {
u64 end;
parent = *p;
cache = rb_entry(parent, struct dso_cache, rb_node);
end = cache->offset + DSO__DATA_CACHE_SIZE;
if (offset < cache->offset)
p = &(*p)->rb_left;
else if (offset >= end)
p = &(*p)->rb_right;
else
return cache;
}
return NULL;
}
static void
dso_cache__insert(struct rb_root *root, struct dso_cache *new)
{
struct rb_node **p = &root->rb_node;
struct rb_node *parent = NULL;
struct dso_cache *cache;
u64 offset = new->offset;
while (*p != NULL) {
u64 end;
parent = *p;
cache = rb_entry(parent, struct dso_cache, rb_node);
end = cache->offset + DSO__DATA_CACHE_SIZE;
if (offset < cache->offset)
p = &(*p)->rb_left;
else if (offset >= end)
p = &(*p)->rb_right;
}
rb_link_node(&new->rb_node, parent, p);
rb_insert_color(&new->rb_node, root);
}
static ssize_t
dso_cache__memcpy(struct dso_cache *cache, u64 offset,
u8 *data, u64 size)
{
u64 cache_offset = offset - cache->offset;
u64 cache_size = min(cache->size - cache_offset, size);
memcpy(data, cache->data + cache_offset, cache_size);
return cache_size;
}
static ssize_t
dso_cache__read(struct dso *dso, u64 offset, u8 *data, ssize_t size)
{
struct dso_cache *cache;
ssize_t ret;
do {
u64 cache_offset;
ret = -ENOMEM;
cache = zalloc(sizeof(*cache) + DSO__DATA_CACHE_SIZE);
if (!cache)
break;
cache_offset = offset & DSO__DATA_CACHE_MASK;
ret = pread(dso->data.fd, cache->data, DSO__DATA_CACHE_SIZE, cache_offset);
if (ret <= 0)
break;
cache->offset = cache_offset;
cache->size = ret;
dso_cache__insert(&dso->data.cache, cache);
ret = dso_cache__memcpy(cache, offset, data, size);
} while (0);
if (ret <= 0)
free(cache);
return ret;
}
static ssize_t dso_cache_read(struct dso *dso, u64 offset,
u8 *data, ssize_t size)
{
struct dso_cache *cache;
cache = dso_cache__find(&dso->data.cache, offset);
if (cache)
return dso_cache__memcpy(cache, offset, data, size);
else
return dso_cache__read(dso, offset, data, size);
}
/*
* Reads and caches dso data DSO__DATA_CACHE_SIZE size chunks
* in the rb_tree. Any read to already cached data is served
* by cached data.
*/
static ssize_t cached_read(struct dso *dso, u64 offset, u8 *data, ssize_t size)
{
ssize_t r = 0;
u8 *p = data;
do {
ssize_t ret;
ret = dso_cache_read(dso, offset, p, size);
if (ret < 0)
return ret;
/* Reached EOF, return what we have. */
if (!ret)
break;
BUG_ON(ret > size);
r += ret;
p += ret;
offset += ret;
size -= ret;
} while (size);
return r;
}
static int data_file_size(struct dso *dso)
{
struct stat st;
char sbuf[STRERR_BUFSIZE];
if (!dso->data.file_size) {
if (fstat(dso->data.fd, &st)) {
pr_err("dso mmap failed, fstat: %s\n",
strerror_r(errno, sbuf, sizeof(sbuf)));
return -1;
}
dso->data.file_size = st.st_size;
}
return 0;
}
/**
* dso__data_size - Return dso data size
* @dso: dso object
* @machine: machine object
*
* Return: dso data size
*/
off_t dso__data_size(struct dso *dso, struct machine *machine)
{
int fd;
fd = dso__data_fd(dso, machine);
if (fd < 0)
return fd;
if (data_file_size(dso))
return -1;
/* For now just estimate dso data size is close to file size */
return dso->data.file_size;
}
static ssize_t data_read_offset(struct dso *dso, u64 offset,
u8 *data, ssize_t size)
{
if (data_file_size(dso))
return -1;
/* Check the offset sanity. */
if (offset > dso->data.file_size)
return -1;
if (offset + size < offset)
return -1;
return cached_read(dso, offset, data, size);
}
/**
* dso__data_read_offset - Read data from dso file offset
* @dso: dso object
* @machine: machine object
* @offset: file offset
* @data: buffer to store data
* @size: size of the @data buffer
*
* External interface to read data from dso file offset. Open
* dso data file and use cached_read to get the data.
*/
ssize_t dso__data_read_offset(struct dso *dso, struct machine *machine,
u64 offset, u8 *data, ssize_t size)
{
if (dso__data_fd(dso, machine) < 0)
return -1;
return data_read_offset(dso, offset, data, size);
}
/**
* dso__data_read_addr - Read data from dso address
* @dso: dso object
* @machine: machine object
* @add: virtual memory address
* @data: buffer to store data
* @size: size of the @data buffer
*
* External interface to read data from dso address.
*/
ssize_t dso__data_read_addr(struct dso *dso, struct map *map,
struct machine *machine, u64 addr,
u8 *data, ssize_t size)
{
u64 offset = map->map_ip(map, addr);
return dso__data_read_offset(dso, machine, offset, data, size);
}
struct map *dso__new_map(const char *name)
{
struct map *map = NULL;
struct dso *dso = dso__new(name);
if (dso)
map = map__new2(0, dso, MAP__FUNCTION);
return map;
}
struct dso *dso__kernel_findnew(struct machine *machine, const char *name,
const char *short_name, int dso_type)
{
/*
* The kernel dso could be created by build_id processing.
*/
struct dso *dso = __dsos__findnew(&machine->kernel_dsos, name);
/*
* We need to run this in all cases, since during the build_id
* processing we had no idea this was the kernel dso.
*/
if (dso != NULL) {
dso__set_short_name(dso, short_name, false);
dso->kernel = dso_type;
}
return dso;
}
perf symbols: Improve DSO long names lookup speed with rbtree With workload that spawns and destroys many threads and processes, it was found that perf-mem could took a long time to post-process the perf data after the target workload had completed its operation. The performance bottleneck was found to be the lookup and insertion of the new DSO structures (thousands of them in this case). In a dual-socket Ivy-Bridge E7-4890 v2 machine (30-core, 60-thread), the perf profile below shows what perf was doing after the profiled AIM7 shared workload completed: - 83.94% perf libc-2.11.3.so [.] __strcmp_sse42 - __strcmp_sse42 - 99.82% map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main - 13.17% perf perf [.] __dsos__findnew __dsos__findnew map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main So about 97% of CPU times were spent in the map__new() function trying to insert new DSO entry into the DSO linked list. The whole post-processing step took about 9 minutes. The DSO structures are currently searched linearly. So the total processing time will be proportional to n^2. To overcome this performance problem, the DSO code is modified to also put the DSO structures in a RB tree sorted by its long name in additional to being in a simple linked list. With this change, the processing time will become proportional to n*log(n) which will be much quicker for large n. However, the short name will still be searched using the old linear searching method. With that patch in place, the same perf-mem post-processing step took less than 30 seconds to complete. Signed-off-by: Waiman Long <Waiman.Long@hp.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Douglas Hatch <doug.hatch@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Scott J Norton <scott.norton@hp.com> Link: http://lkml.kernel.org/r/1412098575-27863-3-git-send-email-Waiman.Long@hp.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-09-30 17:36:15 +00:00
/*
* Find a matching entry and/or link current entry to RB tree.
* Either one of the dso or name parameter must be non-NULL or the
* function will not work.
*/
static struct dso *dso__findlink_by_longname(struct rb_root *root,
struct dso *dso, const char *name)
{
struct rb_node **p = &root->rb_node;
struct rb_node *parent = NULL;
if (!name)
name = dso->long_name;
/*
* Find node with the matching name
*/
while (*p) {
struct dso *this = rb_entry(*p, struct dso, rb_node);
int rc = strcmp(name, this->long_name);
parent = *p;
if (rc == 0) {
/*
* In case the new DSO is a duplicate of an existing
* one, print an one-time warning & put the new entry
* at the end of the list of duplicates.
*/
if (!dso || (dso == this))
return this; /* Find matching dso */
/*
* The core kernel DSOs may have duplicated long name.
* In this case, the short name should be different.
* Comparing the short names to differentiate the DSOs.
*/
rc = strcmp(dso->short_name, this->short_name);
if (rc == 0) {
pr_err("Duplicated dso name: %s\n", name);
return NULL;
}
}
if (rc < 0)
p = &parent->rb_left;
else
p = &parent->rb_right;
}
if (dso) {
/* Add new node and rebalance tree */
rb_link_node(&dso->rb_node, parent, p);
rb_insert_color(&dso->rb_node, root);
}
return NULL;
}
static inline struct dso *
dso__find_by_longname(const struct rb_root *root, const char *name)
{
return dso__findlink_by_longname((struct rb_root *)root, NULL, name);
}
void dso__set_long_name(struct dso *dso, const char *name, bool name_allocated)
{
if (name == NULL)
return;
if (dso->long_name_allocated)
free((char *)dso->long_name);
dso->long_name = name;
dso->long_name_len = strlen(name);
dso->long_name_allocated = name_allocated;
}
void dso__set_short_name(struct dso *dso, const char *name, bool name_allocated)
{
if (name == NULL)
return;
if (dso->short_name_allocated)
free((char *)dso->short_name);
dso->short_name = name;
dso->short_name_len = strlen(name);
dso->short_name_allocated = name_allocated;
}
static void dso__set_basename(struct dso *dso)
{
/*
* basename() may modify path buffer, so we must pass
* a copy.
*/
char *base, *lname = strdup(dso->long_name);
if (!lname)
return;
/*
* basename() may return a pointer to internal
* storage which is reused in subsequent calls
* so copy the result.
*/
base = strdup(basename(lname));
free(lname);
if (!base)
return;
dso__set_short_name(dso, base, true);
}
int dso__name_len(const struct dso *dso)
{
if (!dso)
return strlen("[unknown]");
if (verbose)
return dso->long_name_len;
return dso->short_name_len;
}
bool dso__loaded(const struct dso *dso, enum map_type type)
{
return dso->loaded & (1 << type);
}
bool dso__sorted_by_name(const struct dso *dso, enum map_type type)
{
return dso->sorted_by_name & (1 << type);
}
void dso__set_sorted_by_name(struct dso *dso, enum map_type type)
{
dso->sorted_by_name |= (1 << type);
}
struct dso *dso__new(const char *name)
{
struct dso *dso = calloc(1, sizeof(*dso) + strlen(name) + 1);
if (dso != NULL) {
int i;
strcpy(dso->name, name);
dso__set_long_name(dso, dso->name, false);
dso__set_short_name(dso, dso->name, false);
for (i = 0; i < MAP__NR_TYPES; ++i)
dso->symbols[i] = dso->symbol_names[i] = RB_ROOT;
dso->data.cache = RB_ROOT;
dso->data.fd = -1;
dso->data.status = DSO_DATA_STATUS_UNKNOWN;
dso->symtab_type = DSO_BINARY_TYPE__NOT_FOUND;
dso->binary_type = DSO_BINARY_TYPE__NOT_FOUND;
dso->is_64_bit = (sizeof(void *) == 8);
dso->loaded = 0;
dso->rel = 0;
dso->sorted_by_name = 0;
dso->has_build_id = 0;
dso->has_srcline = 1;
dso->a2l_fails = 1;
dso->kernel = DSO_TYPE_USER;
dso->needs_swap = DSO_SWAP__UNSET;
perf symbols: Improve DSO long names lookup speed with rbtree With workload that spawns and destroys many threads and processes, it was found that perf-mem could took a long time to post-process the perf data after the target workload had completed its operation. The performance bottleneck was found to be the lookup and insertion of the new DSO structures (thousands of them in this case). In a dual-socket Ivy-Bridge E7-4890 v2 machine (30-core, 60-thread), the perf profile below shows what perf was doing after the profiled AIM7 shared workload completed: - 83.94% perf libc-2.11.3.so [.] __strcmp_sse42 - __strcmp_sse42 - 99.82% map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main - 13.17% perf perf [.] __dsos__findnew __dsos__findnew map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main So about 97% of CPU times were spent in the map__new() function trying to insert new DSO entry into the DSO linked list. The whole post-processing step took about 9 minutes. The DSO structures are currently searched linearly. So the total processing time will be proportional to n^2. To overcome this performance problem, the DSO code is modified to also put the DSO structures in a RB tree sorted by its long name in additional to being in a simple linked list. With this change, the processing time will become proportional to n*log(n) which will be much quicker for large n. However, the short name will still be searched using the old linear searching method. With that patch in place, the same perf-mem post-processing step took less than 30 seconds to complete. Signed-off-by: Waiman Long <Waiman.Long@hp.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Douglas Hatch <doug.hatch@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Scott J Norton <scott.norton@hp.com> Link: http://lkml.kernel.org/r/1412098575-27863-3-git-send-email-Waiman.Long@hp.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-09-30 17:36:15 +00:00
RB_CLEAR_NODE(&dso->rb_node);
INIT_LIST_HEAD(&dso->node);
INIT_LIST_HEAD(&dso->data.open_entry);
}
return dso;
}
void dso__delete(struct dso *dso)
{
int i;
perf symbols: Improve DSO long names lookup speed with rbtree With workload that spawns and destroys many threads and processes, it was found that perf-mem could took a long time to post-process the perf data after the target workload had completed its operation. The performance bottleneck was found to be the lookup and insertion of the new DSO structures (thousands of them in this case). In a dual-socket Ivy-Bridge E7-4890 v2 machine (30-core, 60-thread), the perf profile below shows what perf was doing after the profiled AIM7 shared workload completed: - 83.94% perf libc-2.11.3.so [.] __strcmp_sse42 - __strcmp_sse42 - 99.82% map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main - 13.17% perf perf [.] __dsos__findnew __dsos__findnew map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main So about 97% of CPU times were spent in the map__new() function trying to insert new DSO entry into the DSO linked list. The whole post-processing step took about 9 minutes. The DSO structures are currently searched linearly. So the total processing time will be proportional to n^2. To overcome this performance problem, the DSO code is modified to also put the DSO structures in a RB tree sorted by its long name in additional to being in a simple linked list. With this change, the processing time will become proportional to n*log(n) which will be much quicker for large n. However, the short name will still be searched using the old linear searching method. With that patch in place, the same perf-mem post-processing step took less than 30 seconds to complete. Signed-off-by: Waiman Long <Waiman.Long@hp.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Douglas Hatch <doug.hatch@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Scott J Norton <scott.norton@hp.com> Link: http://lkml.kernel.org/r/1412098575-27863-3-git-send-email-Waiman.Long@hp.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-09-30 17:36:15 +00:00
if (!RB_EMPTY_NODE(&dso->rb_node))
pr_err("DSO %s is still in rbtree when being deleted!\n",
dso->long_name);
for (i = 0; i < MAP__NR_TYPES; ++i)
symbols__delete(&dso->symbols[i]);
if (dso->short_name_allocated) {
zfree((char **)&dso->short_name);
dso->short_name_allocated = false;
}
if (dso->long_name_allocated) {
zfree((char **)&dso->long_name);
dso->long_name_allocated = false;
}
dso__data_close(dso);
dso_cache__free(&dso->data.cache);
dso__free_a2l(dso);
zfree(&dso->symsrc_filename);
free(dso);
}
void dso__set_build_id(struct dso *dso, void *build_id)
{
memcpy(dso->build_id, build_id, sizeof(dso->build_id));
dso->has_build_id = 1;
}
bool dso__build_id_equal(const struct dso *dso, u8 *build_id)
{
return memcmp(dso->build_id, build_id, sizeof(dso->build_id)) == 0;
}
void dso__read_running_kernel_build_id(struct dso *dso, struct machine *machine)
{
char path[PATH_MAX];
if (machine__is_default_guest(machine))
return;
sprintf(path, "%s/sys/kernel/notes", machine->root_dir);
if (sysfs__read_build_id(path, dso->build_id,
sizeof(dso->build_id)) == 0)
dso->has_build_id = true;
}
int dso__kernel_module_get_build_id(struct dso *dso,
const char *root_dir)
{
char filename[PATH_MAX];
/*
* kernel module short names are of the form "[module]" and
* we need just "module" here.
*/
const char *name = dso->short_name + 1;
snprintf(filename, sizeof(filename),
"%s/sys/module/%.*s/notes/.note.gnu.build-id",
root_dir, (int)strlen(name) - 1, name);
if (sysfs__read_build_id(filename, dso->build_id,
sizeof(dso->build_id)) == 0)
dso->has_build_id = true;
return 0;
}
bool __dsos__read_build_ids(struct list_head *head, bool with_hits)
{
bool have_build_id = false;
struct dso *pos;
list_for_each_entry(pos, head, node) {
if (with_hits && !pos->hit)
continue;
if (pos->has_build_id) {
have_build_id = true;
continue;
}
if (filename__read_build_id(pos->long_name, pos->build_id,
sizeof(pos->build_id)) > 0) {
have_build_id = true;
pos->has_build_id = true;
}
}
return have_build_id;
}
void dsos__add(struct dsos *dsos, struct dso *dso)
{
list_add_tail(&dso->node, &dsos->head);
perf symbols: Improve DSO long names lookup speed with rbtree With workload that spawns and destroys many threads and processes, it was found that perf-mem could took a long time to post-process the perf data after the target workload had completed its operation. The performance bottleneck was found to be the lookup and insertion of the new DSO structures (thousands of them in this case). In a dual-socket Ivy-Bridge E7-4890 v2 machine (30-core, 60-thread), the perf profile below shows what perf was doing after the profiled AIM7 shared workload completed: - 83.94% perf libc-2.11.3.so [.] __strcmp_sse42 - __strcmp_sse42 - 99.82% map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main - 13.17% perf perf [.] __dsos__findnew __dsos__findnew map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main So about 97% of CPU times were spent in the map__new() function trying to insert new DSO entry into the DSO linked list. The whole post-processing step took about 9 minutes. The DSO structures are currently searched linearly. So the total processing time will be proportional to n^2. To overcome this performance problem, the DSO code is modified to also put the DSO structures in a RB tree sorted by its long name in additional to being in a simple linked list. With this change, the processing time will become proportional to n*log(n) which will be much quicker for large n. However, the short name will still be searched using the old linear searching method. With that patch in place, the same perf-mem post-processing step took less than 30 seconds to complete. Signed-off-by: Waiman Long <Waiman.Long@hp.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Douglas Hatch <doug.hatch@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Scott J Norton <scott.norton@hp.com> Link: http://lkml.kernel.org/r/1412098575-27863-3-git-send-email-Waiman.Long@hp.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-09-30 17:36:15 +00:00
dso__findlink_by_longname(&dsos->root, dso, NULL);
}
struct dso *dsos__find(const struct dsos *dsos, const char *name,
bool cmp_short)
{
struct dso *pos;
perf symbols: Fix vdso list searching When "perf record" was used on a large machine with a lot of CPUs, the perf post-processing time (the time after the workload was done until the perf command itself exited) could take a lot of minutes and even hours depending on how large the resulting perf.data file was. While running AIM7 1500-user high_systime workload on a 80-core x86-64 system with a 3.9 kernel (with only the -s -a options used), the workload itself took about 2 minutes to run and the perf.data file had a size of 1108.746 MB. However, the post-processing step took more than 10 minutes. With a gprof-profiled perf binary, the time spent by perf was as follows: % cumulative self self total time seconds seconds calls s/call s/call name 96.90 822.10 822.10 192156 0.00 0.00 dsos__find 0.81 828.96 6.86 172089958 0.00 0.00 rb_next 0.41 832.44 3.48 48539289 0.00 0.00 rb_erase So 97% (822 seconds) of the time was spent in a single dsos_find() function. After analyzing the call-graph data below: ----------------------------------------------- 0.00 822.12 192156/192156 map__new [6] [7] 96.9 0.00 822.12 192156 vdso__dso_findnew [7] 822.10 0.00 192156/192156 dsos__find [8] 0.01 0.00 192156/192156 dsos__add [62] 0.01 0.00 192156/192366 dso__new [61] 0.00 0.00 1/45282525 memdup [31] 0.00 0.00 192156/192230 dso__set_long_name [91] ----------------------------------------------- 822.10 0.00 192156/192156 vdso__dso_findnew [7] [8] 96.9 822.10 0.00 192156 dsos__find [8] ----------------------------------------------- It was found that the vdso__dso_findnew() function failed to locate VDSO__MAP_NAME ("[vdso]") in the dso list and have to insert a new entry at the end for 192156 times. This problem is due to the fact that there are 2 types of name in the dso entry - short name and long name. The initial dso__new() adds "[vdso]" to both the short and long names. After that, vdso__dso_findnew() modifies the long name to something like /tmp/perf-vdso.so-NoXkDj. The dsos__find() function only compares the long name. As a result, the same vdso entry is duplicated many time in the dso list. This bug increases memory consumption as well as slows the symbol processing time to a crawl. To resolve this problem, the dsos__find() function interface was modified to enable searching either the long name or the short name. The vdso__dso_findnew() will now search only the short name while the other call sites search for the long name as before. With this change, the cpu time of perf was reduced from 848.38s to 15.77s and dsos__find() only accounted for 0.06% of the total time. 0.06 15.73 0.01 192151 0.00 0.00 dsos__find Signed-off-by: Waiman Long <Waiman.Long@hp.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: "Chandramouleeswaran, Aswin" <aswin@hp.com> Cc: "Norton, Scott J" <scott.norton@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1368110568-64714-1-git-send-email-Waiman.Long@hp.com [ replaced TRUE/FALSE with stdbool.h equivalents, fixing builds where those macros are not present (NO_LIBPYTHON=1 NO_LIBPERL=1), fix from Jiri Olsa ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-05-09 14:42:48 +00:00
if (cmp_short) {
list_for_each_entry(pos, &dsos->head, node)
perf symbols: Fix vdso list searching When "perf record" was used on a large machine with a lot of CPUs, the perf post-processing time (the time after the workload was done until the perf command itself exited) could take a lot of minutes and even hours depending on how large the resulting perf.data file was. While running AIM7 1500-user high_systime workload on a 80-core x86-64 system with a 3.9 kernel (with only the -s -a options used), the workload itself took about 2 minutes to run and the perf.data file had a size of 1108.746 MB. However, the post-processing step took more than 10 minutes. With a gprof-profiled perf binary, the time spent by perf was as follows: % cumulative self self total time seconds seconds calls s/call s/call name 96.90 822.10 822.10 192156 0.00 0.00 dsos__find 0.81 828.96 6.86 172089958 0.00 0.00 rb_next 0.41 832.44 3.48 48539289 0.00 0.00 rb_erase So 97% (822 seconds) of the time was spent in a single dsos_find() function. After analyzing the call-graph data below: ----------------------------------------------- 0.00 822.12 192156/192156 map__new [6] [7] 96.9 0.00 822.12 192156 vdso__dso_findnew [7] 822.10 0.00 192156/192156 dsos__find [8] 0.01 0.00 192156/192156 dsos__add [62] 0.01 0.00 192156/192366 dso__new [61] 0.00 0.00 1/45282525 memdup [31] 0.00 0.00 192156/192230 dso__set_long_name [91] ----------------------------------------------- 822.10 0.00 192156/192156 vdso__dso_findnew [7] [8] 96.9 822.10 0.00 192156 dsos__find [8] ----------------------------------------------- It was found that the vdso__dso_findnew() function failed to locate VDSO__MAP_NAME ("[vdso]") in the dso list and have to insert a new entry at the end for 192156 times. This problem is due to the fact that there are 2 types of name in the dso entry - short name and long name. The initial dso__new() adds "[vdso]" to both the short and long names. After that, vdso__dso_findnew() modifies the long name to something like /tmp/perf-vdso.so-NoXkDj. The dsos__find() function only compares the long name. As a result, the same vdso entry is duplicated many time in the dso list. This bug increases memory consumption as well as slows the symbol processing time to a crawl. To resolve this problem, the dsos__find() function interface was modified to enable searching either the long name or the short name. The vdso__dso_findnew() will now search only the short name while the other call sites search for the long name as before. With this change, the cpu time of perf was reduced from 848.38s to 15.77s and dsos__find() only accounted for 0.06% of the total time. 0.06 15.73 0.01 192151 0.00 0.00 dsos__find Signed-off-by: Waiman Long <Waiman.Long@hp.com> Acked-by: Ingo Molnar <mingo@kernel.org> Cc: "Chandramouleeswaran, Aswin" <aswin@hp.com> Cc: "Norton, Scott J" <scott.norton@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1368110568-64714-1-git-send-email-Waiman.Long@hp.com [ replaced TRUE/FALSE with stdbool.h equivalents, fixing builds where those macros are not present (NO_LIBPYTHON=1 NO_LIBPERL=1), fix from Jiri Olsa ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-05-09 14:42:48 +00:00
if (strcmp(pos->short_name, name) == 0)
return pos;
return NULL;
}
perf symbols: Improve DSO long names lookup speed with rbtree With workload that spawns and destroys many threads and processes, it was found that perf-mem could took a long time to post-process the perf data after the target workload had completed its operation. The performance bottleneck was found to be the lookup and insertion of the new DSO structures (thousands of them in this case). In a dual-socket Ivy-Bridge E7-4890 v2 machine (30-core, 60-thread), the perf profile below shows what perf was doing after the profiled AIM7 shared workload completed: - 83.94% perf libc-2.11.3.so [.] __strcmp_sse42 - __strcmp_sse42 - 99.82% map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main - 13.17% perf perf [.] __dsos__findnew __dsos__findnew map__new machine__process_mmap_event perf_session_deliver_event perf_session__process_event __perf_session__process_events cmd_record cmd_mem run_builtin main __libc_start_main So about 97% of CPU times were spent in the map__new() function trying to insert new DSO entry into the DSO linked list. The whole post-processing step took about 9 minutes. The DSO structures are currently searched linearly. So the total processing time will be proportional to n^2. To overcome this performance problem, the DSO code is modified to also put the DSO structures in a RB tree sorted by its long name in additional to being in a simple linked list. With this change, the processing time will become proportional to n*log(n) which will be much quicker for large n. However, the short name will still be searched using the old linear searching method. With that patch in place, the same perf-mem post-processing step took less than 30 seconds to complete. Signed-off-by: Waiman Long <Waiman.Long@hp.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Douglas Hatch <doug.hatch@hp.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Scott J Norton <scott.norton@hp.com> Link: http://lkml.kernel.org/r/1412098575-27863-3-git-send-email-Waiman.Long@hp.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-09-30 17:36:15 +00:00
return dso__find_by_longname(&dsos->root, name);
}
struct dso *dsos__addnew(struct dsos *dsos, const char *name)
{
struct dso *dso = dso__new(name);
if (dso != NULL) {
dsos__add(dsos, dso);
dso__set_basename(dso);
}
return dso;
}
struct dso *__dsos__findnew(struct dsos *dsos, const char *name)
{
struct dso *dso = dsos__find(dsos, name, false);
return dso ? dso : dsos__addnew(dsos, name);
}
size_t __dsos__fprintf_buildid(struct list_head *head, FILE *fp,
bool (skip)(struct dso *dso, int parm), int parm)
{
struct dso *pos;
size_t ret = 0;
list_for_each_entry(pos, head, node) {
if (skip && skip(pos, parm))
continue;
ret += dso__fprintf_buildid(pos, fp);
ret += fprintf(fp, " %s\n", pos->long_name);
}
return ret;
}
size_t __dsos__fprintf(struct list_head *head, FILE *fp)
{
struct dso *pos;
size_t ret = 0;
list_for_each_entry(pos, head, node) {
int i;
for (i = 0; i < MAP__NR_TYPES; ++i)
ret += dso__fprintf(pos, i, fp);
}
return ret;
}
size_t dso__fprintf_buildid(struct dso *dso, FILE *fp)
{
char sbuild_id[BUILD_ID_SIZE * 2 + 1];
build_id__sprintf(dso->build_id, sizeof(dso->build_id), sbuild_id);
return fprintf(fp, "%s", sbuild_id);
}
size_t dso__fprintf(struct dso *dso, enum map_type type, FILE *fp)
{
struct rb_node *nd;
size_t ret = fprintf(fp, "dso: %s (", dso->short_name);
if (dso->short_name != dso->long_name)
ret += fprintf(fp, "%s, ", dso->long_name);
ret += fprintf(fp, "%s, %sloaded, ", map_type__name[type],
dso__loaded(dso, type) ? "" : "NOT ");
ret += dso__fprintf_buildid(dso, fp);
ret += fprintf(fp, ")\n");
for (nd = rb_first(&dso->symbols[type]); nd; nd = rb_next(nd)) {
struct symbol *pos = rb_entry(nd, struct symbol, rb_node);
ret += symbol__fprintf(pos, fp);
}
return ret;
}
enum dso_type dso__type(struct dso *dso, struct machine *machine)
{
int fd;
fd = dso__data_fd(dso, machine);
if (fd < 0)
return DSO__TYPE_UNKNOWN;
return dso__type_fd(fd);
}