tjh.dev
-2
tools/arch/alpha/include/uapi/asm/mman.h
-2
tools/arch/alpha/include/uapi/asm/mman.h
-2
tools/arch/mips/include/uapi/asm/mman.h
-2
tools/arch/mips/include/uapi/asm/mman.h
-2
tools/arch/parisc/include/uapi/asm/mman.h
-2
tools/arch/parisc/include/uapi/asm/mman.h
+2
tools/arch/powerpc/include/uapi/asm/kvm.h
+2
tools/arch/powerpc/include/uapi/asm/kvm.h
···
463
463
#define KVM_PPC_CPU_CHAR_BR_HINT_HONOURED (1ULL << 58)
464
464
#define KVM_PPC_CPU_CHAR_MTTRIG_THR_RECONF (1ULL << 57)
465
465
#define KVM_PPC_CPU_CHAR_COUNT_CACHE_DIS (1ULL << 56)
466
+
#define KVM_PPC_CPU_CHAR_BCCTR_FLUSH_ASSIST (1ull << 54)
466
467
467
468
#define KVM_PPC_CPU_BEHAV_FAVOUR_SECURITY (1ULL << 63)
468
469
#define KVM_PPC_CPU_BEHAV_L1D_FLUSH_PR (1ULL << 62)
469
470
#define KVM_PPC_CPU_BEHAV_BNDS_CHK_SPEC_BAR (1ULL << 61)
471
+
#define KVM_PPC_CPU_BEHAV_FLUSH_COUNT_CACHE (1ull << 58)
470
472
471
473
/* Per-vcpu XICS interrupt controller state */
472
474
#define KVM_REG_PPC_ICP_STATE (KVM_REG_PPC | KVM_REG_SIZE_U64 | 0x8c)
+1
tools/arch/x86/include/asm/cpufeatures.h
+1
tools/arch/x86/include/asm/cpufeatures.h
···
344
344
/* Intel-defined CPU features, CPUID level 0x00000007:0 (EDX), word 18 */
345
345
#define X86_FEATURE_AVX512_4VNNIW (18*32+ 2) /* AVX-512 Neural Network Instructions */
346
346
#define X86_FEATURE_AVX512_4FMAPS (18*32+ 3) /* AVX-512 Multiply Accumulation Single precision */
347
+
#define X86_FEATURE_TSX_FORCE_ABORT (18*32+13) /* "" TSX_FORCE_ABORT */
347
348
#define X86_FEATURE_PCONFIG (18*32+18) /* Intel PCONFIG */
348
349
#define X86_FEATURE_SPEC_CTRL (18*32+26) /* "" Speculation Control (IBRS + IBPB) */
349
350
#define X86_FEATURE_INTEL_STIBP (18*32+27) /* "" Single Thread Indirect Branch Predictors */
-2
tools/arch/xtensa/include/uapi/asm/mman.h
-2
tools/arch/xtensa/include/uapi/asm/mman.h
+2
-2
tools/build/feature/test-libopencsd.c
+2
-2
tools/build/feature/test-libopencsd.c
···
4
4
/*
5
5
* Check OpenCSD library version is sufficient to provide required features
6
6
*/
7
-
#define OCSD_MIN_VER ((0 << 16) | (10 << 8) | (0))
7
+
#define OCSD_MIN_VER ((0 << 16) | (11 << 8) | (0))
8
8
#if !defined(OCSD_VER_NUM) || (OCSD_VER_NUM < OCSD_MIN_VER)
9
-
#error "OpenCSD >= 0.10.0 is required"
9
+
#error "OpenCSD >= 0.11.0 is required"
10
10
#endif
11
11
12
12
int main(void)
+23
tools/include/uapi/asm-generic/mman-common-tools.h
+23
tools/include/uapi/asm-generic/mman-common-tools.h
···
1
+
/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2
+
#ifndef __ASM_GENERIC_MMAN_COMMON_TOOLS_ONLY_H
3
+
#define __ASM_GENERIC_MMAN_COMMON_TOOLS_ONLY_H
4
+
5
+
#include <asm-generic/mman-common.h>
6
+
7
+
/* We need this because we need to have tools/include/uapi/ included in the tools
8
+
* header search path to get access to stuff that is not yet in the system's
9
+
* copy of the files in that directory, but since this cset:
10
+
*
11
+
* 746c9398f5ac ("arch: move common mmap flags to linux/mman.h")
12
+
*
13
+
* We end up making sys/mman.h, that is in the system headers, to not find the
14
+
* MAP_SHARED and MAP_PRIVATE defines because they are not anymore in our copy
15
+
* of asm-generic/mman-common.h. So we define them here and include this header
16
+
* from each of the per arch mman.h headers.
17
+
*/
18
+
#ifndef MAP_SHARED
19
+
#define MAP_SHARED 0x01 /* Share changes */
20
+
#define MAP_PRIVATE 0x02 /* Changes are private */
21
+
#define MAP_SHARED_VALIDATE 0x03 /* share + validate extension flags */
22
+
#endif
23
+
#endif // __ASM_GENERIC_MMAN_COMMON_TOOLS_ONLY_H
+1
-3
tools/include/uapi/asm-generic/mman-common.h
+1
-3
tools/include/uapi/asm-generic/mman-common.h
···
15
15
#define PROT_GROWSDOWN 0x01000000 /* mprotect flag: extend change to start of growsdown vma */
16
16
#define PROT_GROWSUP 0x02000000 /* mprotect flag: extend change to end of growsup vma */
17
17
18
-
#define MAP_SHARED 0x01 /* Share changes */
19
-
#define MAP_PRIVATE 0x02 /* Changes are private */
20
-
#define MAP_SHARED_VALIDATE 0x03 /* share + validate extension flags */
18
+
/* 0x01 - 0x03 are defined in linux/mman.h */
21
19
#define MAP_TYPE 0x0f /* Mask for type of mapping */
22
20
#define MAP_FIXED 0x10 /* Interpret addr exactly */
23
21
#define MAP_ANONYMOUS 0x20 /* don't use a file */
+1
-1
tools/include/uapi/asm-generic/mman.h
+1
-1
tools/include/uapi/asm-generic/mman.h
+10
-1
tools/include/uapi/asm-generic/unistd.h
+10
-1
tools/include/uapi/asm-generic/unistd.h
···
824
824
__SYSCALL(__NR_sched_rr_get_interval_time64, sys_sched_rr_get_interval)
825
825
#endif
826
826
827
+
#define __NR_pidfd_send_signal 424
828
+
__SYSCALL(__NR_pidfd_send_signal, sys_pidfd_send_signal)
829
+
#define __NR_io_uring_setup 425
830
+
__SYSCALL(__NR_io_uring_setup, sys_io_uring_setup)
831
+
#define __NR_io_uring_enter 426
832
+
__SYSCALL(__NR_io_uring_enter, sys_io_uring_enter)
833
+
#define __NR_io_uring_register 427
834
+
__SYSCALL(__NR_io_uring_register, sys_io_uring_register)
835
+
827
836
#undef __NR_syscalls
828
-
#define __NR_syscalls 424
837
+
#define __NR_syscalls 428
829
838
830
839
/*
831
840
* 32 bit systems traditionally used different
+64
tools/include/uapi/drm/i915_drm.h
+64
tools/include/uapi/drm/i915_drm.h
···
1486
1486
#define I915_CONTEXT_MAX_USER_PRIORITY 1023 /* inclusive */
1487
1487
#define I915_CONTEXT_DEFAULT_PRIORITY 0
1488
1488
#define I915_CONTEXT_MIN_USER_PRIORITY -1023 /* inclusive */
1489
+
/*
1490
+
* When using the following param, value should be a pointer to
1491
+
* drm_i915_gem_context_param_sseu.
1492
+
*/
1493
+
#define I915_CONTEXT_PARAM_SSEU 0x7
1489
1494
__u64 value;
1495
+
};
1496
+
1497
+
/**
1498
+
* Context SSEU programming
1499
+
*
1500
+
* It may be necessary for either functional or performance reason to configure
1501
+
* a context to run with a reduced number of SSEU (where SSEU stands for Slice/
1502
+
* Sub-slice/EU).
1503
+
*
1504
+
* This is done by configuring SSEU configuration using the below
1505
+
* @struct drm_i915_gem_context_param_sseu for every supported engine which
1506
+
* userspace intends to use.
1507
+
*
1508
+
* Not all GPUs or engines support this functionality in which case an error
1509
+
* code -ENODEV will be returned.
1510
+
*
1511
+
* Also, flexibility of possible SSEU configuration permutations varies between
1512
+
* GPU generations and software imposed limitations. Requesting such a
1513
+
* combination will return an error code of -EINVAL.
1514
+
*
1515
+
* NOTE: When perf/OA is active the context's SSEU configuration is ignored in
1516
+
* favour of a single global setting.
1517
+
*/
1518
+
struct drm_i915_gem_context_param_sseu {
1519
+
/*
1520
+
* Engine class & instance to be configured or queried.
1521
+
*/
1522
+
__u16 engine_class;
1523
+
__u16 engine_instance;
1524
+
1525
+
/*
1526
+
* Unused for now. Must be cleared to zero.
1527
+
*/
1528
+
__u32 flags;
1529
+
1530
+
/*
1531
+
* Mask of slices to enable for the context. Valid values are a subset
1532
+
* of the bitmask value returned for I915_PARAM_SLICE_MASK.
1533
+
*/
1534
+
__u64 slice_mask;
1535
+
1536
+
/*
1537
+
* Mask of subslices to enable for the context. Valid values are a
1538
+
* subset of the bitmask value return by I915_PARAM_SUBSLICE_MASK.
1539
+
*/
1540
+
__u64 subslice_mask;
1541
+
1542
+
/*
1543
+
* Minimum/Maximum number of EUs to enable per subslice for the
1544
+
* context. min_eus_per_subslice must be inferior or equal to
1545
+
* max_eus_per_subslice.
1546
+
*/
1547
+
__u16 min_eus_per_subslice;
1548
+
__u16 max_eus_per_subslice;
1549
+
1550
+
/*
1551
+
* Unused for now. Must be cleared to zero.
1552
+
*/
1553
+
__u32 rsvd;
1490
1554
};
1491
1555
1492
1556
enum drm_i915_oa_format {
+1
tools/include/uapi/linux/fcntl.h
+1
tools/include/uapi/linux/fcntl.h
···
41
41
#define F_SEAL_SHRINK 0x0002 /* prevent file from shrinking */
42
42
#define F_SEAL_GROW 0x0004 /* prevent file from growing */
43
43
#define F_SEAL_WRITE 0x0008 /* prevent writes */
44
+
#define F_SEAL_FUTURE_WRITE 0x0010 /* prevent future writes while mapped */
44
45
/* (1U << 31) is reserved for signed error codes */
45
46
46
47
/*
+4
tools/include/uapi/linux/mman.h
+4
tools/include/uapi/linux/mman.h
···
12
12
#define OVERCOMMIT_ALWAYS 1
13
13
#define OVERCOMMIT_NEVER 2
14
14
15
+
#define MAP_SHARED 0x01 /* Share changes */
16
+
#define MAP_PRIVATE 0x02 /* Changes are private */
17
+
#define MAP_SHARED_VALIDATE 0x03 /* share + validate extension flags */
18
+
15
19
/*
16
20
* Huge page size encoding when MAP_HUGETLB is specified, and a huge page
17
21
* size other than the default is desired. See hugetlb_encode.h.
+2
-2
tools/perf/Makefile.perf
+2
-2
tools/perf/Makefile.perf
···
481
481
mmap_flags_array := $(beauty_outdir)/mmap_flags_array.c
482
482
mmap_flags_tbl := $(srctree)/tools/perf/trace/beauty/mmap_flags.sh
483
483
484
-
$(mmap_flags_array): $(asm_generic_uapi_dir)/mman.h $(asm_generic_uapi_dir)/mman-common.h $(mmap_flags_tbl)
485
-
$(Q)$(SHELL) '$(mmap_flags_tbl)' $(asm_generic_uapi_dir) $(arch_asm_uapi_dir) > $@
484
+
$(mmap_flags_array): $(linux_uapi_dir)/mman.h $(asm_generic_uapi_dir)/mman.h $(asm_generic_uapi_dir)/mman-common.h $(mmap_flags_tbl)
485
+
$(Q)$(SHELL) '$(mmap_flags_tbl)' $(linux_uapi_dir) $(asm_generic_uapi_dir) $(arch_asm_uapi_dir) > $@
486
486
487
487
mount_flags_array := $(beauty_outdir)/mount_flags_array.c
488
488
mount_flags_tbl := $(srctree)/tools/perf/trace/beauty/mount_flags.sh
+4
tools/perf/arch/x86/entry/syscalls/syscall_64.tbl
+4
tools/perf/arch/x86/entry/syscalls/syscall_64.tbl
···
345
345
334 common rseq __x64_sys_rseq
346
346
# don't use numbers 387 through 423, add new calls after the last
347
347
# 'common' entry
348
+
424 common pidfd_send_signal __x64_sys_pidfd_send_signal
349
+
425 common io_uring_setup __x64_sys_io_uring_setup
350
+
426 common io_uring_enter __x64_sys_io_uring_enter
351
+
427 common io_uring_register __x64_sys_io_uring_register
348
352
349
353
#
350
354
# x32-specific system call numbers start at 512 to avoid cache impact
+1
-1
tools/perf/check-headers.sh
+1
-1
tools/perf/check-headers.sh
···
103
103
# diff with extra ignore lines
104
104
check arch/x86/lib/memcpy_64.S '-I "^EXPORT_SYMBOL" -I "^#include <asm/export.h>"'
105
105
check arch/x86/lib/memset_64.S '-I "^EXPORT_SYMBOL" -I "^#include <asm/export.h>"'
106
-
check include/uapi/asm-generic/mman.h '-I "^#include <\(uapi/\)*asm-generic/mman-common.h>"'
106
+
check include/uapi/asm-generic/mman.h '-I "^#include <\(uapi/\)*asm-generic/mman-common\(-tools\)*.h>"'
107
107
check include/uapi/linux/mman.h '-I "^#include <\(uapi/\)*asm/mman.h>"'
108
108
109
109
# diff non-symmetric files
+63
-14
tools/perf/scripts/python/exported-sql-viewer.py
+63
-14
tools/perf/scripts/python/exported-sql-viewer.py
···
107
107
from PySide.QtCore import *
108
108
from PySide.QtGui import *
109
109
from PySide.QtSql import *
110
+
pyside_version_1 = True
110
111
from decimal import *
111
112
from ctypes import *
112
113
from multiprocessing import Process, Array, Value, Event
···
1527
1526
" (" + dsoname(query.value(15)) + ")")
1528
1527
return data
1529
1528
1529
+
def BranchDataPrepWA(query):
1530
+
data = []
1531
+
data.append(query.value(0))
1532
+
# Workaround pyside failing to handle large integers (i.e. time) in python3 by converting to a string
1533
+
data.append("{:>19}".format(query.value(1)))
1534
+
for i in xrange(2, 8):
1535
+
data.append(query.value(i))
1536
+
data.append(tohex(query.value(8)).rjust(16) + " " + query.value(9) + offstr(query.value(10)) +
1537
+
" (" + dsoname(query.value(11)) + ")" + " -> " +
1538
+
tohex(query.value(12)) + " " + query.value(13) + offstr(query.value(14)) +
1539
+
" (" + dsoname(query.value(15)) + ")")
1540
+
return data
1541
+
1530
1542
# Branch data model
1531
1543
1532
1544
class BranchModel(TreeModel):
···
1567
1553
" AND evsel_id = " + str(self.event_id) +
1568
1554
" ORDER BY samples.id"
1569
1555
" LIMIT " + str(glb_chunk_sz))
1570
-
self.fetcher = SQLFetcher(glb, sql, BranchDataPrep, self.AddSample)
1556
+
if pyside_version_1 and sys.version_info[0] == 3:
1557
+
prep = BranchDataPrepWA
1558
+
else:
1559
+
prep = BranchDataPrep
1560
+
self.fetcher = SQLFetcher(glb, sql, prep, self.AddSample)
1571
1561
self.fetcher.done.connect(self.Update)
1572
1562
self.fetcher.Fetch(glb_chunk_sz)
1573
1563
···
2097
2079
return False
2098
2080
return True
2099
2081
2100
-
# SQL data preparation
2101
-
2102
-
def SQLTableDataPrep(query, count):
2103
-
data = []
2104
-
for i in xrange(count):
2105
-
data.append(query.value(i))
2106
-
return data
2107
-
2108
2082
# SQL table data model item
2109
2083
2110
2084
class SQLTableItem():
···
2120
2110
self.more = True
2121
2111
self.populated = 0
2122
2112
self.column_headers = column_headers
2123
-
self.fetcher = SQLFetcher(glb, sql, lambda x, y=len(column_headers): SQLTableDataPrep(x, y), self.AddSample)
2113
+
self.fetcher = SQLFetcher(glb, sql, lambda x, y=len(column_headers): self.SQLTableDataPrep(x, y), self.AddSample)
2124
2114
self.fetcher.done.connect(self.Update)
2125
2115
self.fetcher.Fetch(glb_chunk_sz)
2126
2116
···
2164
2154
def columnHeader(self, column):
2165
2155
return self.column_headers[column]
2166
2156
2157
+
def SQLTableDataPrep(self, query, count):
2158
+
data = []
2159
+
for i in xrange(count):
2160
+
data.append(query.value(i))
2161
+
return data
2162
+
2167
2163
# SQL automatic table data model
2168
2164
2169
2165
class SQLAutoTableModel(SQLTableModel):
···
2198
2182
QueryExec(query, "SELECT column_name FROM information_schema.columns WHERE table_schema = '" + schema + "' and table_name = '" + select_table_name + "'")
2199
2183
while query.next():
2200
2184
column_headers.append(query.value(0))
2185
+
if pyside_version_1 and sys.version_info[0] == 3:
2186
+
if table_name == "samples_view":
2187
+
self.SQLTableDataPrep = self.samples_view_DataPrep
2188
+
if table_name == "samples":
2189
+
self.SQLTableDataPrep = self.samples_DataPrep
2201
2190
super(SQLAutoTableModel, self).__init__(glb, sql, column_headers, parent)
2191
+
2192
+
def samples_view_DataPrep(self, query, count):
2193
+
data = []
2194
+
data.append(query.value(0))
2195
+
# Workaround pyside failing to handle large integers (i.e. time) in python3 by converting to a string
2196
+
data.append("{:>19}".format(query.value(1)))
2197
+
for i in xrange(2, count):
2198
+
data.append(query.value(i))
2199
+
return data
2200
+
2201
+
def samples_DataPrep(self, query, count):
2202
+
data = []
2203
+
for i in xrange(9):
2204
+
data.append(query.value(i))
2205
+
# Workaround pyside failing to handle large integers (i.e. time) in python3 by converting to a string
2206
+
data.append("{:>19}".format(query.value(9)))
2207
+
for i in xrange(10, count):
2208
+
data.append(query.value(i))
2209
+
return data
2202
2210
2203
2211
# Base class for custom ResizeColumnsToContents
2204
2212
···
2908
2868
ok = self.xed_format_context(2, inst.xedp, inst.bufferp, sizeof(inst.buffer), ip, 0, 0)
2909
2869
if not ok:
2910
2870
return 0, ""
2871
+
if sys.version_info[0] == 2:
2872
+
result = inst.buffer.value
2873
+
else:
2874
+
result = inst.buffer.value.decode()
2911
2875
# Return instruction length and the disassembled instruction text
2912
2876
# For now, assume the length is in byte 166
2913
-
return inst.xedd[166], inst.buffer.value
2877
+
return inst.xedd[166], result
2914
2878
2915
2879
def TryOpen(file_name):
2916
2880
try:
···
2930
2886
header = f.read(7)
2931
2887
f.seek(pos)
2932
2888
magic = header[0:4]
2933
-
eclass = ord(header[4])
2934
-
encoding = ord(header[5])
2935
-
version = ord(header[6])
2889
+
if sys.version_info[0] == 2:
2890
+
eclass = ord(header[4])
2891
+
encoding = ord(header[5])
2892
+
version = ord(header[6])
2893
+
else:
2894
+
eclass = header[4]
2895
+
encoding = header[5]
2896
+
version = header[6]
2936
2897
if magic == chr(127) + "ELF" and eclass > 0 and eclass < 3 and encoding > 0 and encoding < 3 and version == 1:
2937
2898
result = True if eclass == 2 else False
2938
2899
return result
+11
-3
tools/perf/trace/beauty/mmap_flags.sh
+11
-3
tools/perf/trace/beauty/mmap_flags.sh
···
1
1
#!/bin/sh
2
2
# SPDX-License-Identifier: LGPL-2.1
3
3
4
-
if [ $# -ne 2 ] ; then
4
+
if [ $# -ne 3 ] ; then
5
5
[ $# -eq 1 ] && hostarch=$1 || hostarch=`uname -m | sed -e s/i.86/x86/ -e s/x86_64/x86/`
6
+
linux_header_dir=tools/include/uapi/linux
6
7
header_dir=tools/include/uapi/asm-generic
7
8
arch_header_dir=tools/arch/${hostarch}/include/uapi/asm
8
9
else
9
-
header_dir=$1
10
-
arch_header_dir=$2
10
+
linux_header_dir=$1
11
+
header_dir=$2
12
+
arch_header_dir=$3
11
13
fi
12
14
15
+
linux_mman=${linux_header_dir}/mman.h
13
16
arch_mman=${arch_header_dir}/mman.h
14
17
15
18
# those in egrep -vw are flags, we want just the bits
···
21
18
regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
22
19
egrep -q $regex ${arch_mman} && \
23
20
(egrep $regex ${arch_mman} | \
21
+
sed -r "s/$regex/\2 \1/g" | \
22
+
xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n")
23
+
egrep -q $regex ${linux_mman} && \
24
+
(egrep $regex ${linux_mman} | \
25
+
egrep -vw 'MAP_(UNINITIALIZED|TYPE|SHARED_VALIDATE)' | \
24
26
sed -r "s/$regex/\2 \1/g" | \
25
27
xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n")
26
28
([ ! -f ${arch_mman} ] || egrep -q '#[[:space:]]*include[[:space:]]+<uapi/asm-generic/mman.*' ${arch_mman}) &&
+1
tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
+1
tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
-29
tools/perf/util/evlist.c
-29
tools/perf/util/evlist.c
···
231
231
}
232
232
}
233
233
234
-
void perf_event_attr__set_max_precise_ip(struct perf_event_attr *pattr)
235
-
{
236
-
struct perf_event_attr attr = {
237
-
.type = PERF_TYPE_HARDWARE,
238
-
.config = PERF_COUNT_HW_CPU_CYCLES,
239
-
.exclude_kernel = 1,
240
-
.precise_ip = 3,
241
-
};
242
-
243
-
event_attr_init(&attr);
244
-
245
-
/*
246
-
* Unnamed union member, not supported as struct member named
247
-
* initializer in older compilers such as gcc 4.4.7
248
-
*/
249
-
attr.sample_period = 1;
250
-
251
-
while (attr.precise_ip != 0) {
252
-
int fd = sys_perf_event_open(&attr, 0, -1, -1, 0);
253
-
if (fd != -1) {
254
-
close(fd);
255
-
break;
256
-
}
257
-
--attr.precise_ip;
258
-
}
259
-
260
-
pattr->precise_ip = attr.precise_ip;
261
-
}
262
-
263
234
int __perf_evlist__add_default(struct perf_evlist *evlist, bool precise)
264
235
{
265
236
struct perf_evsel *evsel = perf_evsel__new_cycles(precise);
-2
tools/perf/util/evlist.h
-2
tools/perf/util/evlist.h
···
315
315
void perf_evlist__set_tracking_event(struct perf_evlist *evlist,
316
316
struct perf_evsel *tracking_evsel);
317
317
318
-
void perf_event_attr__set_max_precise_ip(struct perf_event_attr *attr);
319
-
320
318
struct perf_evsel *
321
319
perf_evlist__find_evsel_by_str(struct perf_evlist *evlist, const char *str);
322
320
+59
-13
tools/perf/util/evsel.c
+59
-13
tools/perf/util/evsel.c
···
295
295
if (!precise)
296
296
goto new_event;
297
297
298
-
perf_event_attr__set_max_precise_ip(&attr);
299
298
/*
300
299
* Now let the usual logic to set up the perf_event_attr defaults
301
300
* to kick in when we return and before perf_evsel__open() is called.
···
303
304
evsel = perf_evsel__new(&attr);
304
305
if (evsel == NULL)
305
306
goto out;
307
+
308
+
evsel->precise_max = true;
306
309
307
310
/* use asprintf() because free(evsel) assumes name is allocated */
308
311
if (asprintf(&evsel->name, "cycles%s%s%.*s",
···
1084
1083
}
1085
1084
1086
1085
if (evsel->precise_max)
1087
-
perf_event_attr__set_max_precise_ip(attr);
1086
+
attr->precise_ip = 3;
1088
1087
1089
1088
if (opts->all_user) {
1090
1089
attr->exclude_kernel = 1;
···
1750
1749
return true;
1751
1750
}
1752
1751
1752
+
static void display_attr(struct perf_event_attr *attr)
1753
+
{
1754
+
if (verbose >= 2) {
1755
+
fprintf(stderr, "%.60s\n", graph_dotted_line);
1756
+
fprintf(stderr, "perf_event_attr:\n");
1757
+
perf_event_attr__fprintf(stderr, attr, __open_attr__fprintf, NULL);
1758
+
fprintf(stderr, "%.60s\n", graph_dotted_line);
1759
+
}
1760
+
}
1761
+
1762
+
static int perf_event_open(struct perf_evsel *evsel,
1763
+
pid_t pid, int cpu, int group_fd,
1764
+
unsigned long flags)
1765
+
{
1766
+
int precise_ip = evsel->attr.precise_ip;
1767
+
int fd;
1768
+
1769
+
while (1) {
1770
+
pr_debug2("sys_perf_event_open: pid %d cpu %d group_fd %d flags %#lx",
1771
+
pid, cpu, group_fd, flags);
1772
+
1773
+
fd = sys_perf_event_open(&evsel->attr, pid, cpu, group_fd, flags);
1774
+
if (fd >= 0)
1775
+
break;
1776
+
1777
+
/*
1778
+
* Do quick precise_ip fallback if:
1779
+
* - there is precise_ip set in perf_event_attr
1780
+
* - maximum precise is requested
1781
+
* - sys_perf_event_open failed with ENOTSUP error,
1782
+
* which is associated with wrong precise_ip
1783
+
*/
1784
+
if (!precise_ip || !evsel->precise_max || (errno != ENOTSUP))
1785
+
break;
1786
+
1787
+
/*
1788
+
* We tried all the precise_ip values, and it's
1789
+
* still failing, so leave it to standard fallback.
1790
+
*/
1791
+
if (!evsel->attr.precise_ip) {
1792
+
evsel->attr.precise_ip = precise_ip;
1793
+
break;
1794
+
}
1795
+
1796
+
pr_debug2("\nsys_perf_event_open failed, error %d\n", -ENOTSUP);
1797
+
evsel->attr.precise_ip--;
1798
+
pr_debug2("decreasing precise_ip by one (%d)\n", evsel->attr.precise_ip);
1799
+
display_attr(&evsel->attr);
1800
+
}
1801
+
1802
+
return fd;
1803
+
}
1804
+
1753
1805
int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
1754
1806
struct thread_map *threads)
1755
1807
{
···
1878
1824
if (perf_missing_features.sample_id_all)
1879
1825
evsel->attr.sample_id_all = 0;
1880
1826
1881
-
if (verbose >= 2) {
1882
-
fprintf(stderr, "%.60s\n", graph_dotted_line);
1883
-
fprintf(stderr, "perf_event_attr:\n");
1884
-
perf_event_attr__fprintf(stderr, &evsel->attr, __open_attr__fprintf, NULL);
1885
-
fprintf(stderr, "%.60s\n", graph_dotted_line);
1886
-
}
1827
+
display_attr(&evsel->attr);
1887
1828
1888
1829
for (cpu = 0; cpu < cpus->nr; cpu++) {
1889
1830
···
1890
1841
1891
1842
group_fd = get_group_fd(evsel, cpu, thread);
1892
1843
retry_open:
1893
-
pr_debug2("sys_perf_event_open: pid %d cpu %d group_fd %d flags %#lx",
1894
-
pid, cpus->map[cpu], group_fd, flags);
1895
-
1896
1844
test_attr__ready();
1897
1845
1898
-
fd = sys_perf_event_open(&evsel->attr, pid, cpus->map[cpu],
1899
-
group_fd, flags);
1846
+
fd = perf_event_open(evsel, pid, cpus->map[cpu],
1847
+
group_fd, flags);
1900
1848
1901
1849
FD(evsel, cpu, thread) = fd;
1902
1850
+8
-12
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+8
-12
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
···
251
251
if (!(decoder->tsc_ctc_ratio_n % decoder->tsc_ctc_ratio_d))
252
252
decoder->tsc_ctc_mult = decoder->tsc_ctc_ratio_n /
253
253
decoder->tsc_ctc_ratio_d;
254
-
255
-
/*
256
-
* Allow for timestamps appearing to backwards because a TSC
257
-
* packet has slipped past a MTC packet, so allow 2 MTC ticks
258
-
* or ...
259
-
*/
260
-
decoder->tsc_slip = multdiv(2 << decoder->mtc_shift,
261
-
decoder->tsc_ctc_ratio_n,
262
-
decoder->tsc_ctc_ratio_d);
263
254
}
264
-
/* ... or 0x100 paranoia */
265
-
if (decoder->tsc_slip < 0x100)
266
-
decoder->tsc_slip = 0x100;
255
+
256
+
/*
257
+
* A TSC packet can slip past MTC packets so that the timestamp appears
258
+
* to go backwards. One estimate is that can be up to about 40 CPU
259
+
* cycles, which is certainly less than 0x1000 TSC ticks, but accept
260
+
* slippage an order of magnitude more to be on the safe side.
261
+
*/
262
+
decoder->tsc_slip = 0x10000;
267
263
268
264
intel_pt_log("timestamp: mtc_shift %u\n", decoder->mtc_shift);
269
265
intel_pt_log("timestamp: tsc_ctc_ratio_n %u\n", decoder->tsc_ctc_ratio_n);
+20
-12
tools/perf/util/machine.c
+20
-12
tools/perf/util/machine.c
···
1421
1421
machine->vmlinux_map->end = ~0ULL;
1422
1422
}
1423
1423
1424
+
static void machine__update_kernel_mmap(struct machine *machine,
1425
+
u64 start, u64 end)
1426
+
{
1427
+
struct map *map = machine__kernel_map(machine);
1428
+
1429
+
map__get(map);
1430
+
map_groups__remove(&machine->kmaps, map);
1431
+
1432
+
machine__set_kernel_mmap(machine, start, end);
1433
+
1434
+
map_groups__insert(&machine->kmaps, map);
1435
+
map__put(map);
1436
+
}
1437
+
1424
1438
int machine__create_kernel_maps(struct machine *machine)
1425
1439
{
1426
1440
struct dso *kernel = machine__get_kernel(machine);
···
1467
1453
goto out_put;
1468
1454
}
1469
1455
1470
-
/* we have a real start address now, so re-order the kmaps */
1471
-
map = machine__kernel_map(machine);
1472
-
1473
-
map__get(map);
1474
-
map_groups__remove(&machine->kmaps, map);
1475
-
1476
-
/* assume it's the last in the kmaps */
1477
-
machine__set_kernel_mmap(machine, addr, ~0ULL);
1478
-
1479
-
map_groups__insert(&machine->kmaps, map);
1480
-
map__put(map);
1456
+
/*
1457
+
* we have a real start address now, so re-order the kmaps
1458
+
* assume it's the last in the kmaps
1459
+
*/
1460
+
machine__update_kernel_mmap(machine, addr, ~0ULL);
1481
1461
}
1482
1462
1483
1463
if (machine__create_extra_kernel_maps(machine, kernel))
···
1607
1599
if (strstr(kernel->long_name, "vmlinux"))
1608
1600
dso__set_short_name(kernel, "[kernel.vmlinux]", false);
1609
1601
1610
-
machine__set_kernel_mmap(machine, event->mmap.start,
1602
+
machine__update_kernel_mmap(machine, event->mmap.start,
1611
1603
event->mmap.start + event->mmap.len);
1612
1604
1613
1605
/*
+10
tools/perf/util/pmu.c
+10
tools/perf/util/pmu.c
···
732
732
733
733
if (!is_arm_pmu_core(name)) {
734
734
pname = pe->pmu ? pe->pmu : "cpu";
735
+
736
+
/*
737
+
* uncore alias may be from different PMU
738
+
* with common prefix
739
+
*/
740
+
if (pmu_is_uncore(name) &&
741
+
!strncmp(pname, name, strlen(pname)))
742
+
goto new_alias;
743
+
735
744
if (strcmp(pname, name))
736
745
continue;
737
746
}
738
747
748
+
new_alias:
739
749
/* need type casts to override 'const' */
740
750
__perf_pmu__new_alias(head, NULL, (char *)pe->name,
741
751
(char *)pe->desc, (char *)pe->event,