CoreSight August 2024

coresight@lists.linaro.org

9 participants
48 discussions

Re: [PATCH v1 2/9] perf auxtrace: Remove unused 'pmu' pointer from struct auxtrace_record

by Arnaldo Carvalho de Melo

On Fri, Aug 09, 2024 at 11:02:16AM +0300, Adrian Hunter wrote: > On 6/08/24 23:41, Leo Yan wrote: > > The 'pmu' pointer in the auxtrace_record structure is not used after > > support multiple AUX events, remove it. > > > > Signed-off-by: Leo Yan <leo.yan(a)arm.com> > > Reviewed-by: Adrian Hunter <adrian.hunter(a)intel.com> Applied the first reviewed two patches. - Arnaldo > > --- > > tools/perf/arch/arm/util/cs-etm.c | 1 - > > tools/perf/arch/arm64/util/arm-spe.c | 1 - > > tools/perf/arch/arm64/util/hisi-ptt.c | 1 - > > tools/perf/arch/x86/util/intel-bts.c | 1 - > > tools/perf/arch/x86/util/intel-pt.c | 1 - > > tools/perf/util/auxtrace.h | 1 - > > 6 files changed, 6 deletions(-) > > > > diff --git a/tools/perf/arch/arm/util/cs-etm.c b/tools/perf/arch/arm/util/cs-etm.c > > index da6231367993..96aeb7cdbee1 100644 > > --- a/tools/perf/arch/arm/util/cs-etm.c > > +++ b/tools/perf/arch/arm/util/cs-etm.c > > @@ -888,7 +888,6 @@ struct auxtrace_record *cs_etm_record_init(int *err) > > } > > > > ptr->cs_etm_pmu = cs_etm_pmu; > > - ptr->itr.pmu = cs_etm_pmu; > > ptr->itr.parse_snapshot_options = cs_etm_parse_snapshot_options; > > ptr->itr.recording_options = cs_etm_recording_options; > > ptr->itr.info_priv_size = cs_etm_info_priv_size; > > diff --git a/tools/perf/arch/arm64/util/arm-spe.c b/tools/perf/arch/arm64/util/arm-spe.c > > index d59f6ca499f2..2be99fdf997d 100644 > > --- a/tools/perf/arch/arm64/util/arm-spe.c > > +++ b/tools/perf/arch/arm64/util/arm-spe.c > > @@ -514,7 +514,6 @@ struct auxtrace_record *arm_spe_recording_init(int *err, > > } > > > > sper->arm_spe_pmu = arm_spe_pmu; > > - sper->itr.pmu = arm_spe_pmu; > > sper->itr.snapshot_start = arm_spe_snapshot_start; > > sper->itr.snapshot_finish = arm_spe_snapshot_finish; > > sper->itr.find_snapshot = arm_spe_find_snapshot; > > diff --git a/tools/perf/arch/arm64/util/hisi-ptt.c b/tools/perf/arch/arm64/util/hisi-ptt.c > > index ba97c8a562a0..eac9739c87e6 100644 > > --- a/tools/perf/arch/arm64/util/hisi-ptt.c > > +++ b/tools/perf/arch/arm64/util/hisi-ptt.c > > @@ -174,7 +174,6 @@ struct auxtrace_record *hisi_ptt_recording_init(int *err, > > } > > > > pttr->hisi_ptt_pmu = hisi_ptt_pmu; > > - pttr->itr.pmu = hisi_ptt_pmu; > > pttr->itr.recording_options = hisi_ptt_recording_options; > > pttr->itr.info_priv_size = hisi_ptt_info_priv_size; > > pttr->itr.info_fill = hisi_ptt_info_fill; > > diff --git a/tools/perf/arch/x86/util/intel-bts.c b/tools/perf/arch/x86/util/intel-bts.c > > index 34696f3d3d5d..85c8186300c8 100644 > > --- a/tools/perf/arch/x86/util/intel-bts.c > > +++ b/tools/perf/arch/x86/util/intel-bts.c > > @@ -434,7 +434,6 @@ struct auxtrace_record *intel_bts_recording_init(int *err) > > } > > > > btsr->intel_bts_pmu = intel_bts_pmu; > > - btsr->itr.pmu = intel_bts_pmu; > > btsr->itr.recording_options = intel_bts_recording_options; > > btsr->itr.info_priv_size = intel_bts_info_priv_size; > > btsr->itr.info_fill = intel_bts_info_fill; > > diff --git a/tools/perf/arch/x86/util/intel-pt.c b/tools/perf/arch/x86/util/intel-pt.c > > index 4b710e875953..ea510a7486b1 100644 > > --- a/tools/perf/arch/x86/util/intel-pt.c > > +++ b/tools/perf/arch/x86/util/intel-pt.c > > @@ -1197,7 +1197,6 @@ struct auxtrace_record *intel_pt_recording_init(int *err) > > } > > > > ptr->intel_pt_pmu = intel_pt_pmu; > > - ptr->itr.pmu = intel_pt_pmu; > > ptr->itr.recording_options = intel_pt_recording_options; > > ptr->itr.info_priv_size = intel_pt_info_priv_size; > > ptr->itr.info_fill = intel_pt_info_fill; > > diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h > > index 8a6ec9565835..95304368103b 100644 > > --- a/tools/perf/util/auxtrace.h > > +++ b/tools/perf/util/auxtrace.h > > @@ -411,7 +411,6 @@ struct auxtrace_record { > > int (*read_finish)(struct auxtrace_record *itr, int idx); > > unsigned int alignment; > > unsigned int default_aux_sample_size; > > - struct perf_pmu *pmu; > > struct evlist *evlist; > > }; > >

3 weeks

Re: [PATCH v1 9/9] perf arm-spe: Dump metadata with version 2

by James Clark

On 27/08/2024 5:44 pm, Leo Yan wrote: > This commit dumps metadata with version 2. It uses two string arrays > metadata_hdr_fmts and metadata_per_cpu_fmts as string formats for the > header and per CPU data respectively, and the arm_spe_print_info() > function is enhanced to support dumping metadata with the version 2 > format. > > After: > > 0 0 0x4a8 [0x170]: PERF_RECORD_AUXTRACE_INFO type: 4 > PMU Type :13 > Version :2 > Num of CPUs :8 > CPU # :0 > MIDR :0x410fd801 > Bound PMU Type :-1 > Min Interval :0 > Load Data Source :0 > CPU # :1 > MIDR :0x410fd801 > Bound PMU Type :-1 > Min Interval :0 > Load Data Source :0 > CPU # :2 > MIDR :0x410fd870 > Bound PMU Type :13 > Min Interval :1024 > Load Data Source :1 > CPU # :3 > MIDR :0x410fd870 > Bound PMU Type :13 > Min Interval :1024 > Load Data Source :1 > CPU # :4 > MIDR :0x410fd870 > Bound PMU Type :13 > Min Interval :1024 > Load Data Source :1 > CPU # :5 > MIDR :0x410fd870 > Bound PMU Type :13 > Min Interval :1024 > Load Data Source :1 > CPU # :6 > MIDR :0x410fd850 > Bound PMU Type :14 > Min Interval :1024 > Load Data Source :1 > CPU # :7 > MIDR :0x410fd850 > Bound PMU Type :14 > Min Interval :1024 > Load Data Source :1 > > Signed-off-by: Leo Yan <leo.yan(a)arm.com> > --- > tools/perf/util/arm-spe.c | 43 ++++++++++++++++++++++++++++++++++----- > 1 file changed, 38 insertions(+), 5 deletions(-) > > diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c > index 87cf06db765b..be34d4c4306a 100644 > --- a/tools/perf/util/arm-spe.c > +++ b/tools/perf/util/arm-spe.c > @@ -1067,16 +1067,49 @@ static bool arm_spe_evsel_is_auxtrace(struct perf_session *session __maybe_unuse > return strstarts(evsel->name, ARM_SPE_PMU_NAME); > } > > -static const char * const arm_spe_info_fmts[] = { > - [ARM_SPE_PMU_TYPE] = " PMU Type %"PRId64"\n", > +static const char * const metadata_hdr_fmts[] = { > + [ARM_SPE_PMU_TYPE] = " PMU Type :%"PRId64"\n", > + [ARM_SPE_HEADER_VERSION] = " Version :%"PRId64"\n", > + [ARM_SPE_CPU_NUM] = " Num of CPUs :%"PRId64"\n", > }; > > -static void arm_spe_print_info(__u64 *arr) > +static const char * const metadata_per_cpu_fmts[] = { > + [ARM_SPE_CPU] = " CPU # :%"PRId64"\n", > + [ARM_SPE_CPU_MIDR] = " MIDR :0x%"PRIx64"\n", > + [ARM_SPE_CPU_PMU_TYPE] = " Bound PMU Type :%"PRId64"\n", > + [ARM_SPE_CAP_MIN_IVAL] = " Min Interval :%"PRId64"\n", > + [ARM_SPE_CAP_LDS] = " Load Data Source :%"PRId64"\n", > +}; > + > +static void arm_spe_print_info(struct arm_spe *spe, __u64 *arr) > { > + unsigned int i, cpu, header_size, cpu_num, per_cpu_size; > + > if (!dump_trace) > return; > > - fprintf(stdout, arm_spe_info_fmts[ARM_SPE_PMU_TYPE], arr[ARM_SPE_PMU_TYPE]); > + if (spe->metadata_ver == 1) { > + cpu_num = 0; > + header_size = ARM_SPE_AUXTRACE_V1_PRIV_MAX; > + per_cpu_size = 0; > + } else if (spe->metadata_ver == 2) { Assuming future version updates are backwards compatible and only add new info this should be spe->metadata_ver >= 2, otherwise version bumps end up causing errors when files get passed around. I know there are arguments about what should and shouldn't be supported when opening new files on old perfs, but in this case it's easy to only add new info to the aux header and leave the old stuff intact. > + cpu_num = arr[ARM_SPE_CPU_NUM]; > + header_size = ARM_SPE_AUXTRACE_V2_PRIV_MAX; > + per_cpu_size = ARM_SPE_AUXTRACE_V2_PRIV_PER_CPU_MAX; I think for coresight we also save the size of each per-cpu block rather than use a constant, that way new items can be appended without breaking readers. That kind of leads to another point that this mechanism is mostly duplicated from coresight. It saves a main header version, then per-cpu groups of variable size with named elements. I'm not saying we should definitely try to share the code, but it's worth keeping in mind. > + } else { > + pr_err("Cannot support metadata ver: %ld\n", spe->metadata_ver); > + return; > + } > + > + for (i = 0; i < header_size; i++) > + fprintf(stdout, metadata_hdr_fmts[i], arr[i]); > + > + arr += header_size; > + for (cpu = 0; cpu < cpu_num; cpu++) { > + for (i = 0; i < per_cpu_size; i++) > + fprintf(stdout, metadata_per_cpu_fmts[i], arr[i]); > + arr += per_cpu_size; > + } > } > > static void arm_spe_set_event_name(struct evlist *evlist, u64 id, > @@ -1383,7 +1416,7 @@ int arm_spe_process_auxtrace_info(union perf_event *event, > spe->auxtrace.evsel_is_auxtrace = arm_spe_evsel_is_auxtrace; > session->auxtrace = &spe->auxtrace; > > - arm_spe_print_info(&auxtrace_info->priv[0]); > + arm_spe_print_info(spe, &auxtrace_info->priv[0]); > > if (dump_trace) > return 0;

3 weeks

Re: [PATCH V4 1/4] Add dso__is_pie call to identify ELF PIE

by Steve Clevenger

On 8/28/2024 1:23 AM, Leo Yan wrote: > On 8/28/2024 6:09 AM, Steve Clevenger wrote: >> >> Add dso__is_pie global to read the .dynamic section DT_FLAGS_1 entry for >> the DF_1_PIE flag. This identifies position executable code. >> >> Signed-off-by: Steve Clevenger <scclevenger(a)os.amperecomputing.com> >> --- >> tools/perf/util/symbol-elf.c | 60 ++++++++++++++++++++++++++++++++++++ >> tools/perf/util/symbol.h | 1 + >> 2 files changed, 61 insertions(+) >> >> diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c >> index e398abfd13a0..0e49c1345a67 100644 >> --- a/tools/perf/util/symbol-elf.c >> +++ b/tools/perf/util/symbol-elf.c >> @@ -662,6 +662,66 @@ static int dso__synthesize_plt_got_symbols(struct dso *dso, Elf *elf, >> return err; >> } >> >> +/* >> + * Check dynamic section DT_FLAGS_1 for a Position Independent >> + * Executable (PIE). >> + */ >> +bool dso__is_pie(struct dso *dso) >> +{ >> + Elf *elf = NULL; >> + Elf_Scn *scn = NULL; >> + GElf_Ehdr ehdr; >> + GElf_Shdr shdr; >> + bool is_pie = false; >> + char dso_path[PATH_MAX]; >> + int fd = -1; >> + >> + if (!dso || (elf_version(EV_CURRENT) == EV_NONE)) >> + goto exit; // false >> + >> + dso__build_id_filename(dso, dso_path, sizeof(dso_path), false); >> + >> + fd = open(dso_path, O_RDONLY); >> + >> + if (fd < 0) { >> + pr_debug("%s: cannot read cached %s.\n", __func__, dso_path); >> + goto exit; // false >> + } >> + >> + elf = elf_begin(fd, ELF_C_READ, NULL); >> + gelf_getehdr(elf, &ehdr); >> + >> + if (ehdr.e_type == ET_DYN) { >> + Elf_Data *data; >> + GElf_Dyn *entry; >> + int n_entries = shdr.sh_size / sizeof(GElf_Dyn); >> + >> + scn = elf_section_by_name(elf, &ehdr, &shdr, ".dynamic", NULL); >> + if (!scn) >> + goto exit; // false > > It should goto a new tag (e.g. exit_failed) so can invoke elf_end(elf) and > close(fd). > >> + >> + data = (Elf_Data *) elf_getdata(scn, NULL); >> + if (!data || !data->d_buf) >> + goto exit; // false > > Ditto. > >> + >> + // check DT_FLAGS_1 >> + for (int i = 0; i < n_entries; i++) { >> + entry = ((GElf_Dyn *) data->d_buf) + i; >> + if (entry->d_tag == DT_FLAGS_1) { >> + if ((entry->d_un.d_val & DF_1_PIE) != 0) { >> + is_pie = true; >> + break; >> + } >> + } >> + } // end for >> + } >> + > > Put 'exit_failed' tag at here. > > With the changes: > > Reviewed-by: Leo Yan <leo.yan(a)arm.com> > >> + elf_end(elf); >> + close(fd); >> +exit: >> + return is_pie; >> +} >> + Hi Leo, It's been a long time since I've seen goto's used as a rule rather than an exception. I see it introduced the problem you've mentioned. Is it ok to simply update the cover letter (patch 0), and this one (patch 1) as V5? If I don't hear from you, I'll resubmit all. Thanks, Steve >> /* >> * We need to check if we have a .dynsym, so that we can handle the >> * .plt, synthesizing its symbols, that aren't on the symtabs (be it >> diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h >> index 3fb5d146d9b1..33ea2596ce31 100644 >> --- a/tools/perf/util/symbol.h >> +++ b/tools/perf/util/symbol.h >> @@ -127,6 +127,7 @@ void dso__insert_symbol(struct dso *dso, >> struct symbol *sym); >> void dso__delete_symbol(struct dso *dso, >> struct symbol *sym); >> +bool dso__is_pie(struct dso *dso); >> >> struct symbol *dso__find_symbol(struct dso *dso, u64 addr); >> struct symbol *dso__find_symbol_nocache(struct dso *dso, u64 addr); >> -- >> 2.25.1 >>

3 weeks

[PATCH V4 0/4] arm-cs-trace-disasm.py/perf must accommodate non-zero DSO text offset

by Steve Clevenger

Changes in V4: - In trace-event-python.c, fixed perf-tools-next merge problem. Changes in V3: - Rebased to linux-perf-tools branch. - Squash symbol-elf.c and symbol.h into same commit. - In map.c, merge dso__is_pie() call into existing if statement. - In arm-cs-trace-disasm.py, remove debug artifacts. Changes in V2: - In dso__is_pie() (symbol-elf.c), Decrease indentation, add null pointer checks per Leo Yan review. - Updated mailing list distribution Steve Clevenger (4): Add dso__is_pie call to identify ELF PIE Force MAPPING_TYPE__IDENTIY for PIE Add map pgoff to python dictionary based on MAPPING_TYPE Adjust objdump start/end range per map pgoff parameter .../scripts/python/arm-cs-trace-disasm.py | 9 ++- tools/perf/util/map.c | 4 +- .../scripting-engines/trace-event-python.c | 13 +++- tools/perf/util/symbol-elf.c | 60 +++++++++++++++++++ tools/perf/util/symbol.h | 1 + 5 files changed, 79 insertions(+), 8 deletions(-) -- 2.25.1

3 weeks

[PATCH V3 0/4] arm-cs-trace-disasm.py/perf must accommodate non-zero DSO text offset

by Steve Clevenger

Changes in V3: - Rebased to linux-perf-tools branch. - Squash symbol-elf.c and symbol.h into same commit. - In map.c, merge dso__is_pie() call into existing if statement. - In arm-cs-trace-disasm.py, remove debug artifacts. Changes in V2: - In dso__is_pie() (symbol-elf.c), Decrease indentation, add null pointer checks per Leo Yan review. - Updated mailing list distribution Fedora 37 distributed shared binary and executable mapped files show a zero text section offset. Starting with the Fedora 38 distribution, the shared binary and executable mapped files show a non-zero text section offset for some binaries. The text offset parameter is never passed into the arm-cs-trace-disasm.py script to allow the script to adjust the start/end address range passed to objdump. This adjustment is required to correctly offset into the dso text section. Not doing so results in an incorrect user instruction trace display for Fedora 38 (and later) user trace output. Steve Clevenger (4): Add dso__is_pie call to identify ELF PIE Force MAPPING_TYPE__IDENTIY for PIE Add map pgoff to python dictionary based on MAPPING_TYPE Adjust objdump start/end range per map pgoff parameter .../scripts/python/arm-cs-trace-disasm.py | 9 ++- tools/perf/util/map.c | 4 +- .../scripting-engines/trace-event-python.c | 13 +++- tools/perf/util/symbol-elf.c | 60 +++++++++++++++++++ tools/perf/util/symbol.h | 1 + 5 files changed, 79 insertions(+), 8 deletions(-) -- 2.25.1

3 weeks

Re: [PATCH V2 5/5] Adjust objdump start/end range per map pgoff parameter

by Steve Clevenger

On 8/27/2024 1:23 PM, Leo Yan wrote: > On 8/26/2024 10:35 PM, Steve Clevenger wrote: >> >> Changes in V2: >> - Updated mailing list distribution >> >> Extract map_pgoff parameter from the dictionary, and adjust start/end >> range passed to objdump based on the value. >> >> The start_addr/stop_addr address checks are changed to print a warning >> only if verbose == True. This script repeatedly sees a zero value passed >> in for >> start_addr = cpu_data[str(cpu) + 'addr'] >> >> These zero values are not a new problem. The start_addr/stop_addr warning >> clutters the instruction trace output, hence this change. >> >> Signed-off-by: Steve Clevenger <scclevenger(a)os.amperecomputing.com> >> --- >> .../perf/scripts/python/arm-cs-trace-disasm.py | 17 +++++++++++++++-- >> 1 file changed, 15 insertions(+), 2 deletions(-) >> >> diff --git a/tools/perf/scripts/python/arm-cs-trace-disasm.py b/tools/perf/scripts/python/arm-cs-trace-disasm.py >> index d973c2baed1c..6bf806078f9a 100755 >> --- a/tools/perf/scripts/python/arm-cs-trace-disasm.py >> +++ b/tools/perf/scripts/python/arm-cs-trace-disasm.py >> @@ -187,6 +187,7 @@ def process_event(param_dict): >> dso_start = get_optional(param_dict, "dso_map_start") >> dso_end = get_optional(param_dict, "dso_map_end") >> symbol = get_optional(param_dict, "symbol") >> + map_pgoff = get_optional(param_dict, "map_pgoff") >> >> cpu = sample["cpu"] >> ip = sample["ip"] >> @@ -250,13 +251,25 @@ def process_event(param_dict): >> return >> >> if (start_addr < int(dso_start) or start_addr > int(dso_end)): >> - print("Start address 0x%x is out of range [ 0x%x .. 0x%x ] for dso %s" % (start_addr, int(dso_start), int(dso_end), dso)) >> + if (options.verbose == True): >> + print("Start address 0x%x is out of range [ 0x%x .. 0x%x ] for dso %s" % (start_addr, int(dso_start), int(dso_end), dso)) >> return >> >> if (stop_addr < int(dso_start) or stop_addr > int(dso_end)): >> - print("Stop address 0x%x is out of range [ 0x%x .. 0x%x ] for dso %s" % (stop_addr, int(dso_start), int(dso_end), dso)) >> + if (options.verbose == True): >> + print("Stop address 0x%x is out of range [ 0x%x .. 0x%x ] for dso %s" % (stop_addr, int(dso_start), int(dso_end), dso)) >> return >> >> + if map_pgoff != None and map_pgoff != '[unknown]': >> + if (dso == "[kernel.kallsyms]"): >> + dso_vm_start = 0 >> + map_pgoff = '[unknown]' >> + else: >> + dso_vm_start = int(dso_start) >> + start_addr += map_pgoff >> + stop_addr += map_pgoff >> + map_pgoff = '[unknown]' > > Every sample has its own field `map_pgoff`. So I am confused why we need the > sentence: map_pgoff = '[unknown]'. And you can see with above change, we will > set dso_vm_start delicately with below code section. > > How about a simple change like below? > > @@ -267,7 +268,7 @@ def process_event(param_dict): > > dso_fname = get_dso_file_path(dso, dso_bid) > if path.exists(dso_fname): > - print_disam(dso_fname, dso_vm_start, start_addr, stop_addr) > + print_disam(dso_fname, dso_vm_start, start_addr + map_pgoff, stop_addr + > map_pgoff) > > If this change does not work, I am curious if anything we missed to handle in > the perf's C code (frontend). > > Thanks, > Leo Leo, Sorry, the map_pgoff = '[unknown]' instances are actually leftover debug artifacts used for print type debug. I resorted to print debug after I didn't see success attempting mixed-mode debug with Visual Studio Code on AARCH64. If you have a known working launch configuration, please share it. See V3 of the patch series. I've rebased to the perf-tools-next branch. Steve > >> if (options.objdump_name != None): >> # It doesn't need to decrease virtual memory offset for disassembly >> # for kernel dso and executable file dso, so in this case we set >> -- >> 2.25.1 >>

3 weeks

Re: [PATCH V2 3/5] Force MAPPING_TYPE__IDENTIY for PIE

by Steve Clevenger

On 8/27/2024 12:55 PM, Leo Yan wrote: > On 8/26/2024 10:35 PM, Steve Clevenger wrote: >> >> Changes in V2: >> - Updated mailing list distribution >> >> Use dso__is_pie() to check whether the DSO file is a Position >> Independent Executable (PIE). If PIE, change the MAPPING_TYPE to >> MAPPING_TYPE__IDENTITY so a zero map pgoff (text offset) is passed >> into the script. >> >> >> Signed-off-by: Steve Clevenger <scclevenger(a)os.amperecomputing.com> >> --- >> tools/perf/util/map.c | 5 ++++- >> 1 file changed, 4 insertions(+), 1 deletion(-) >> >> diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c >> index e1d14936a60d..df7c06fc373e 100644 >> --- a/tools/perf/util/map.c >> +++ b/tools/perf/util/map.c >> @@ -171,8 +171,11 @@ struct map *map__new(struct machine *machine, u64 start, u64 len, >> assert(!dso__kernel(dso)); >> map__init(result, start, start + len, pgoff, dso); >> >> + if (map->pgoff && !no_dso) >> + no_dso = dso__is_pie(dso); // PIE check > > Here the logic is whatever the `pgoff` is (zero or non-zero), we need to > always set the mapping type as MAPPING_TYPE__IDENTITY. It is no need to check > pgoff and is no matter with 'no_dso'. So we can simply check dso__is_pie() and > set IDENTITY flag for the true case. > > if (dso__is_pie(dso)) { > map__set_mapping_type(map, MAPPING_TYPE__IDENTITY); > } else if (anon || no_dso) { > ... > } > > BTW, please rebase the series on the top of the perf-tools-next branch [1], I > saw this patch has merging conflict on it. > > Thanks, > Leo Hi Leo, I combined the dso__is_pie() call onto the end of the existing 'if' statement you now show as an 'else if'. Please see V3 of the patch series. Steve > > [1] https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git > branch: perf-tools-next >> + >> if (anon || no_dso) { >> - map->mapping_type = MAPPING_TYPE__IDENTITY; >> + map__set_mapping_type(map, MAPPING_TYPE__IDENTITY); >> >> /* >> * Set memory without DSO as loaded. All map__find_* >> -- >> 2.25.1 >>

3 weeks

[PATCH V2 0/5] arm-cs-trace-disasm.py/perf must accommodate non-zero DSO text offset

by Steve Clevenger

Changes in V2: - Updated mailing list distribution Fedora 37 distributed shared binary and executable mapped files show a zero text section offset. Starting with the Fedora 38 distribution, the shared binary and executable mapped files show a non-zero text section offset for some binaries. The text offset parameter is never passed into the arm-cs-trace-disasm.py script to allow the script to adjust the start/end address range passed to objdump. This adjustment is required to correctly offset into the dso text section. Not doing so results in an incorrect user instruction trace display for Fedora 38 (and later) user trace output. Steve Clevenger (5): Add dso__is_pie call to identify ELF PIE Add dso__is_pie prototype Force MAPPING_TYPE__IDENTIY for PIE Add map pgoff to python dictionary based on MAPPING_TYPE Adjust objdump start/end range per map pgoff parameter .../scripts/python/arm-cs-trace-disasm.py | 17 +++++- tools/perf/util/map.c | 5 +- .../scripting-engines/trace-event-python.c | 13 +++- tools/perf/util/symbol-elf.c | 60 +++++++++++++++++++ tools/perf/util/symbol.h | 1 + 5 files changed, 90 insertions(+), 6 deletions(-) -- 2.25.1

3 weeks, 2 days

[PATCH V2 1/5] Add dso__is_pie call to identify ELF PIE

by Steve Clevenger

Changes in V2: - Updated mailing list distribution - Decrease indentation, add null pointer checks per Leo Yan review Add dso__is_pie global to read the .dynamic section DT_FLAGS_1 entry for the DF_1_PIE flag. This identifies position executable code. Signed-off-by: Steve Clevenger <scclevenger(a)os.amperecomputing.com> --- tools/perf/util/symbol-elf.c | 60 ++++++++++++++++++++++++++++++++++++ 1 file changed, 60 insertions(+) diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c index e398abfd13a0..0e49c1345a67 100644 --- a/tools/perf/util/symbol-elf.c +++ b/tools/perf/util/symbol-elf.c @@ -662,6 +662,66 @@ static int dso__synthesize_plt_got_symbols(struct dso *dso, Elf *elf, return err; } +/* + * Check dynamic section DT_FLAGS_1 for a Position Independent + * Executable (PIE). + */ +bool dso__is_pie(struct dso *dso) +{ + Elf *elf = NULL; + Elf_Scn *scn = NULL; + GElf_Ehdr ehdr; + GElf_Shdr shdr; + bool is_pie = false; + char dso_path[PATH_MAX]; + int fd = -1; + + if (!dso || (elf_version(EV_CURRENT) == EV_NONE)) + goto exit; // false + + dso__build_id_filename(dso, dso_path, sizeof(dso_path), false); + + fd = open(dso_path, O_RDONLY); + + if (fd < 0) { + pr_debug("%s: cannot read cached %s.\n", __func__, dso_path); + goto exit; // false + } + + elf = elf_begin(fd, ELF_C_READ, NULL); + gelf_getehdr(elf, &ehdr); + + if (ehdr.e_type == ET_DYN) { + Elf_Data *data; + GElf_Dyn *entry; + int n_entries = shdr.sh_size / sizeof(GElf_Dyn); + + scn = elf_section_by_name(elf, &ehdr, &shdr, ".dynamic", NULL); + if (!scn) + goto exit; // false + + data = (Elf_Data *) elf_getdata(scn, NULL); + if (!data || !data->d_buf) + goto exit; // false + + // check DT_FLAGS_1 + for (int i = 0; i < n_entries; i++) { + entry = ((GElf_Dyn *) data->d_buf) + i; + if (entry->d_tag == DT_FLAGS_1) { + if ((entry->d_un.d_val & DF_1_PIE) != 0) { + is_pie = true; + break; + } + } + } // end for + } + + elf_end(elf); + close(fd); +exit: + return is_pie; +} + /* * We need to check if we have a .dynsym, so that we can handle the * .plt, synthesizing its symbols, that aren't on the symtabs (be it -- 2.25.1

3 weeks, 2 days

Re: [PATCH 1/5] Add dso__is_pie call to identify ELF PIE

by Steve Clevenger

On 8/26/2024 6:13 AM, Leo Yan wrote: > Hi Steve, > > On Tue, Aug 20, 2024 at 04:11:35PM -0600, Steve Clevenger wrote: >> From: "steve.c.clevenger.ampere" <scclevenger(a)os.amperecomputing.com> >> >> Add dso__is_pie global to read the .dynamic section DT_FLAGS_1 entry for >> the DF_1_PIE flag. This identifies position executable code. >> >> Signed-off-by: steve.c.clevenger.ampere <scclevenger(a)os.amperecomputing.com> >> --- >> tools/perf/util/symbol-elf.c | 55 ++++++++++++++++++++++++++++++++++++ >> 1 file changed, 55 insertions(+) >> >> diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c >> index e398abfd13a0..1d4bd222b314 100644 >> --- a/tools/perf/util/symbol-elf.c >> +++ b/tools/perf/util/symbol-elf.c >> @@ -662,6 +662,61 @@ static int dso__synthesize_plt_got_symbols(struct dso *dso, Elf *elf, >> return err; >> } >> >> +/* >> + * Check dynamic section DT_FLAGS_1 for a Position Independent >> + * Executable (PIE). >> + */ >> +bool dso__is_pie(struct dso *dso) >> +{ >> + Elf *elf = NULL; >> + Elf_Scn *scn = NULL; >> + GElf_Ehdr ehdr; >> + GElf_Shdr shdr; >> + bool is_pie = false; >> + char dso_path[PATH_MAX]; >> + int fd = -1; >> + >> + if (!dso || (elf_version(EV_CURRENT) == EV_NONE)) >> + return is_pie; // false >> + >> + dso__build_id_filename(dso, dso_path, sizeof(dso_path), false); >> + >> + fd = open(dso_path, O_RDONLY); >> + >> + if (fd < 0) { >> + pr_debug("%s: cannot read cached %s.\n", __func__, dso_path); >> + return is_pie; // false >> + } >> + >> + elf = elf_begin(fd, ELF_C_READ, NULL); >> + gelf_getehdr(elf, &ehdr); >> + >> + if (ehdr.e_type == ET_DYN) { > > The code looks good to me, just several nitpicks. > > To avoid indentation, we can firstly check the failure case and directly > exit for it. > > if (ehdr.e_type != ET_DYN) > goto exit_elf_end; > >> + scn = elf_section_by_name(elf, &ehdr, &shdr, ".dynamic", NULL); > > Ditto. > > if (!scn) > goto exit_elf_end; > >> + if (scn) { // check DT_FLAGS_1 >> + Elf_Data *data; >> + GElf_Dyn *entry; >> + int n_entries = shdr.sh_size / sizeof(GElf_Dyn); >> + >> + data = (Elf_Data *) elf_getdata(scn, NULL); > > For a safe code, it is good to check if pointers (data and > data->d_buf) are valid before dereference them. > > if (!data || !data->d_buf) > goto exit_elf_end; > > With above changes: Thanks Leo. I understand your comment about excess indentation, but I don't believe there's an excess here. Valid points about NULL pointer checks. I've made changes based on your review. Please look for V2 of this patch series. Besides addressing your comments, V2 is mostly to update the mailing lists. Steve C. > > Reviewed-by: Leo Yan <leo.yan(a)arm.com> > >> + for (int i = 0; i < n_entries; i++) { >> + entry = ((GElf_Dyn *) data->d_buf) + i; >> + if (entry->d_tag == DT_FLAGS_1) { >> + if ((entry->d_un.d_val & DF_1_PIE) != 0) { >> + is_pie = true; >> + break; >> + } >> + } >> + } // end for >> + } >> + } >> + >> + elf_end(elf); >> + close(fd); >> + >> + return is_pie; >> +} >> + >> /* >> * We need to check if we have a .dynsym, so that we can handle the >> * .plt, synthesizing its symbols, that aren't on the symtabs (be it >> -- >> 2.25.1 >> >>

3 weeks, 2 days

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

CoreSight August 2024